ABSTRACT 
 
 
 
 
Title: SUBSURFACE HETEROGENEITIES, 
INTERFACES AND BIODEGRADATION: 
DEFINING THE LIMITS ON IN SITU 
BIOREMEDIATION   
  
 Xin Song, Ph.D., 2005 
Directed By: Associate Professor, Eric A. Seagren, Civil and 
Environmental Engineering 
 
 
Subsurface heterogeneities and the associated interfacial processes impact in 
situ bioremediation by affecting the availability of substrates to the microorganisms. 
This research hypothesized that using the scales of subsurface heterogeneities as an 
organizational principle, a quantitative framework based on a set of dimensionless 
numbers could be developed to capture the effects of the competing interfacial and 
biokinetic processes and define the limits for successful application of in situ 
bioremediation. The overall goal of this study was to use an integrated experimental 
and numerical modeling approach to evaluate the developed quantitative framework 
under different simulated scenarios relevant to the subsurface.  
Three experimental scenarios were selected to simulate field sites limited by 
either (1) macro-scale vertical transverse dispersion (Scenario #1), (2) micro-scale 
biokinetics (Scenario #2), or (3) meso-scale sorption/desorption (Scenario #3). 
Experiments were performed in a saturated, heterogeneous intermediate-scale flow 
cell (ISFC) with two layers of contrasting hydraulic conductivities and monitored the 
  
transport of a naphthalene plume through two phases:  Phase I, simulating an intrinsic 
biodegradation; and Phase II, simulating an engineered bioremediation, with selected 
system perturbations. In the first Phase II perturbation, nitrogen (N) and phosphorus 
(P) amendments in excess of stoichiometric requirements were made, while the 
second perturbation was selected based on the rate-limiting process identified via the 
quantitative framework. A numerical model was used to simulate the Phase I 
experiments and verify the independently determined mass transport and biokinetic 
parameters, which were then used in the dimensionless parameters of the proposed 
quantitative framework.  
Scenario #3 was not completed due to the time constraints, but Scenarios #1 
and #2 successfully demonstrated application of the quantitative framework. In 
Scenario #1, Phase I, vertical dispersion was identified as the overall rate-limiting 
process. Correspondingly, increased advection and mechanical dispersion in Phase II 
increased naphthalene biodegradation by ~ 2.7 times, whereas the N and P addition 
had no effect. In Scenario #2, Phase I, dispersion and biokinetics were identified as 
rate-limiting processes. Thus, in Phase II, N and P addition moderately improved 
biodegradation, but removal of inhibitory, high salinity conditions to improve the 
biokinetics increased naphthalene mass loss ~2.7 times. These results demonstrate the 
potential for application of the proposed quantitative framework to predict the rate-
limiting process for in situ bioremediation and aide in the appropriate selection of any 
system perturbations for enhancing in situ bioremediation. 
 
 
  
 
 
 
 
 
 
 
 
 
 
 
SUBSURFACE HETEROGENEITIES, INTERFACES AND BIODEGRADATION: 
DEFINING THE LIMITS ON IN SITU BIOREMEDIATION   
 
 
 
By 
 
 
Xin Song 
 
 
 
 
 
Dissertation submitted to the Faculty of the Graduate School of the  
University of Maryland, College Park, in partial fulfillment 
of the requirements for the degree of 
Doctor of Philosophy 
2005 
 
 
 
 
 
 
 
 
 
Advisory Committee: 
Associate Professor Eric A. Seagren, Chair 
Professor Allen P. Davis 
Professor Alba Torrents 
Associate Professor Karen L. Prestegaard 
Assistant Professor Jennifer G. Becker 
 
  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
? Copyright by 
Xin Song 
2005 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 ii 
 
Dedication 
For my parents Aifang Dai and Shurong Song, especially for my mother, who has always 
believed in me and who is always there for me. 
 
 iii 
 
Acknowledgements 
It is my pleasure to take this opportunity to thank a group of people who 
contribute to the completion of this dissertation.  
First of all, I would like to dedicate my tremendous thanks to Dr. Seagren for his 
guidance, encourage and patience. Without him, I would not have accomplished this 
work. Not only has he been invaluable in the development of my Ph.D. thesis, but it has 
always been a great pleasure working with him. 
My appreciation goes to my other committee members as well. To Drs. Davis, 
Hao, Torrents, Prestegaard and Becker for spending time in reviewing this work and 
assisting me to complete the degree. 
I would like to acknowledge my fellow graduate students Eunyoung Hong and 
Mark Johnson for their assistance and cooperation on this project. And Thank Jim Stagge 
and other undergraduates for their contribution in the lab. I would like to thank Tong Li 
for his advice on the FORTRAN program.  
My special thanks go to Joe Penzien, who has always been very supportive in my 
academic development. 
I owe a great deal to my best friend Naji and his family. They have always been 
with me in the past 5 years and support me in pursuing my Ph.D. degree. My special 
thanks to Naji for helping me when I needed to take samples at midnight. 
Lastly, I would like to thank my mother, for her love, support and encouragement. 
Without her, I would not have made it this far.  
 
 
 iv 
 
Table of Contents 
 
 
Dedication........................................................................................................................... ii 
Acknowledgements............................................................................................................iii 
Table of Contents...............................................................................................................iv 
List of Tables ..................................................................................................................... ix 
List of Figures.................................................................................................................... xi 
Chapter 1 Introduction..................................................................................................... 1 
Chapter 2 Objective and Scope of Work......................................................................... 4 
Chapter 3 Literature Review............................................................................................ 7 
3.1 Macro -scale heterogeneities............................................................................... 8 
3.2 Meso/pore-scale heterogeneity ......................................................................... 12 
3.3 Micro-scale heterogeneities .............................................................................. 21 
3.4 Summary and Conclusion................................................................................. 30 
Chapter 4 Theoretical Basis and Framework Development.......................................... 32 
4.1 Mathematical Model ......................................................................................... 32 
4.1.1 Governing Equations ................................................................................ 33 
4.1.2 Rate-Limited Sorption Reactions.............................................................. 35 
4.1.3 Double Monod Biodegradation Kinetic Model ........................................ 36 
4.1.4 Numerical Solution Techniques................................................................ 38 
4.2 Framework Development.................................................................................. 42 
4.3 Research Approach ........................................................................................... 46 
Chapter 5 Experimental Materials and Methods ........................................................... 54 
5.1 Model Contaminant, Artificial Groundwater and Microorganism ................... 54 
 
 v 
 
5.1.1 Model contaminant ................................................................................... 54 
5.1.2 Artificial Groundwater.............................................................................. 56 
5.1.3 Microorganism.......................................................................................... 59 
5.2 Intermediate-Scale Flow Cell (ISFC) ............................................................... 61 
5.2.1 Sizing of ISFC .......................................................................................... 61 
5.2.2 Flow Cell Material .................................................................................... 65 
5.2.3 Inlet and Outlet Assemblies...................................................................... 67 
5.2.4 ISFC Construction .................................................................................... 72 
5.2.5 Pore Water Sampling Ports Installation.................................................... 74 
5.2.6 Sand Selection, Characterization and Packing.......................................... 79 
5.3 Parameter Estimation Experiments................................................................... 81 
5.3.1 Hydraulic Conductivity............................................................................. 82 
5.3.2 Porous Media Density and Porosity.......................................................... 86 
5.3.3 Dispersivity............................................................................................... 88 
5.3.4 Biodegradation Kinetics............................................................................ 94 
5.3.5 Sorption/Desorption Equilibrium and Kinetics ...................................... 101 
5.3.5.1 Unmodified Sand Sorption ................................................................. 102 
5.3.5.2 HMN-Coated sand sorption ................................................................ 103 
5.3.5.3 Surfactant-coated sand sorption.......................................................... 104 
5.3.5.4 Mixing sands with Amberlite resins ................................................... 106 
5.4 Bioremediation Simulation Experiments........................................................ 111 
5.4.1 Control Experiments ............................................................................... 112 
5.4.1.1 Phase I: Intrinsic Control Experiments............................................... 112 
5.4.1.2 Phase IIA: Nutrients Addition Control Experiments.......................... 115 
5.4.1.3 Phase IIB: Engineered Control Experiments ...................................... 116 
 
 vi 
 
5.4.2 General Bioremediation Simulation Experimental Procedures .............. 116 
5.4.2.1 ISFC preparation and Inoculation....................................................... 116 
5.4.2.2 General ISFC Operation ..................................................................... 118 
5.4.3 Scenario #1: Dispersion-Limited Biodegradation .................................. 120 
5.4.3.1 Phase I: Intrinsic Biodegradation........................................................ 120 
5.4.3.2 Phase IIA: Nutrient Enhanced Biodegradation................................... 122 
5.4.3.3 Phase IIB: Engineered Biodegradation............................................... 122 
5.4.4 Scenario #2: Biodegradation-Limited Biodegradation ........................... 123 
5.4.4.1 Phase I: Intrinsic Biodegradation........................................................ 124 
5.4.4.2 Phase IIA: Nutrient Enhanced Biodegradation................................... 124 
5.4.4.3 Phase IIB: Engineered Biodegradation............................................... 125 
5.4.5 Scenario #3: Sorption-Limited Biodegradation...................................... 125 
5.5 Analytical Methods......................................................................................... 127 
5.5.1 Naphthalene Analysis ............................................................................. 127 
5.5.1.1 Gas Chromatography .......................................................................... 127 
5.5.1.2 Spectrofluophotometry ....................................................................... 133 
5.5.2 Non Reactive Tracer Analysis ................................................................ 134 
5.5.2.1 Fluorescein.......................................................................................... 134 
5.5.2.2 Bromide (Br)....................................................................................... 134 
5.5.3 CPC Analysis.......................................................................................... 136 
5.5.4 Chemical Oxygen Demand (COD) Test................................................. 136 
5.5.5 Dissolved Oxygen................................................................................... 137 
5.5.6 Heterotrophic Plate Count (HPC) ........................................................... 138 
Chapter 6 Results and Discussion ............................................................................... 142 
6.1 Parameter Estimation Experiments................................................................. 142 
 
 vii 
 
6.1.1 Hydraulic Conductivity........................................................................... 142 
6.1.2 Porous Media Density and Porosity........................................................ 143 
6.1.3 Dispersivity............................................................................................. 144 
6.1.4 Biodegradation Kinetics.......................................................................... 153 
6.1.5 Sorption/Desorption Equilibrium and Kinetics ...................................... 163 
6.1.5.1 Unmodified Sand Sorption ................................................................. 163 
6.1.5.2 HMN-Coated Sand Sorption............................................................... 163 
6.1.5.3 Surfactant (CPC)-coated sand sorption............................................... 167 
6.1.5.4 Mixing sands with Amberlite Resins.................................................. 170 
6.2 Bioremediation Simulation Control Experiments........................................... 181 
6.2.1 Phase I: Intrinsic Control Experiments................................................... 181 
6.2.2 Phase IIA: Nutrients Enhanced Control Experiments ............................ 187 
6.2.3 Phase IIB: Engineered Control Experiments .......................................... 188 
6.2.4 Summary and Conclusion....................................................................... 188 
6.3 Bioremediation Simulation Experiments, Scenario #1: Dispersion-Limited 
Scenario....................................................................................................................... 190 
6.3.1 Phase I: Intrinsic Biodegradation............................................................ 190 
6.3.2 Phase IIA: Nutrient Enhanced Biodegradation....................................... 208 
6.3.3 Phase IIB: Engineered Biodegradation................................................... 215 
6.3.4 Summary and Conclusion....................................................................... 223 
6.4 Bioremediation Simulation Experiments, Scenario #2: Biodegradation-Limited 
Scenario....................................................................................................................... 224 
6.4.1 Phase I: Intrinsic Biodegradation............................................................ 224 
6.4.2 Phase IIA: Nutrient Enhanced Biodegradation....................................... 236 
6.4.3 Phase IIB: Engineered Biodegradation................................................... 243 
6.4.4 Summary and Conclusion....................................................................... 254 
 
 viii 
 
Chapter 7 Conclusions and Recommendations ........................................................... 256 
7.1 Summary and Conclusions ............................................................................. 256 
7.2 Recommendations for Future Work................................................................ 260 
Appendix......................................................................................................................... 263 
Appendix 1: Materials and Parts for ISFC.................................................................. 263 
Appendix 2 FORTRAN Program ?trafit3d?............................................................... 266 
Appendix 3: FORTRAN Program ?nvolma?.............................................................. 270 
Bibliography ................................................................................................................... 283 
 
 
 ix 
 
List of Tables 
Table 4.1 Definition of dimensionless numbers ............................................................... 44 
Table 4.2 Description of the flow cell experiments.......................................................... 49 
Table 4.3 Application of the dimensionless numbers in the quantitative framework to 
simulated bioremediation scenarios.................................................................................. 49 
Table 5.1 Physicochemical properties of naphthalene...................................................... 55 
Table 5.2 Mineral medium compositions ......................................................................... 58 
Table 5.3 Summary from literature study of flow cells and their applications................. 63 
Table 5.4 Placement points of the sampling needle tips................................................... 77 
Table 5.5 Point-source tracer experimental operational conditions.................................. 92 
Table 5.6 Physical properties of Amberlite resins XAD-2 and XAD-7 ......................... 107 
Table 5.7 Operation parameters for GC analysis............................................................ 130 
Table 6.1 Physical properties for Mystic White II and Filtersil sands............................ 143 
Table 6.2 FORTRAN ?trafit3d? calibration results for the longitudinal dispersion....... 149 
Table 6.3 Summary of non-reactive tracer results.......................................................... 149 
Table 6.4 Vertical transverse dispersivities calculation summarization table ................ 151 
Table 6.5 Biokinetic parameters for respirometry experiments...................................... 154 
Table 6.6 Freundlich constants for sorption of naphthalene from XAD-2 ..................... 170 
Table 6.7 Partition coefficient of naphthalene onto XAD-7........................................... 177 
Table 6.8 Hydro-geological parameters utilized in RT3D simulation............................ 186 
Table 6.9 Biological parameters utilized in RT3D simulation for Scenario #1.............. 199 
Table 6.10 Moment analysis summarization for Scenario #1......................................... 205 
Table 6.11 Parameters for quantitative framework evaluation for Scenario #1 ............. 207 
Table 6.12  Quantitative framework used to identify the rate-limiting process ............. 208 
Table 6.13 Parameters utilized in RT3D simulation for Scenario #2............................. 226 
 
 x 
 
Table 6.14 Moment analysis summarization for Scenario #2......................................... 233 
Table 6.15 Parameters for quantitative framework evaluation for Scenario #2 ............. 235 
Table 6.16  Quantitative framework used to identify the rate-limiting process ............. 236 
 
 
 
 xi 
 
List of Figures 
Figure 3.1 Naphthalene degradation pathway under aerobic condition ........................... 23 
Figure 3.2 Naphthalene degradation pathway under methanogenic conditions ............... 25 
Figure 3.3 Degradative pathway of naphthalene under sulfate-reducing conditions........ 26 
Figure 4.1 Block diagram illustrating the numerical solution scheme. ............................ 41 
Figure 4.2 Quantitative framework for predicting the rate-limiting phenomenon. .......... 43 
Figure 4.3 Quantitative framework for predicting the rate-limiting phenomenon 
Application of the flow chart to Scenario #1 is highlighted by encircling the experimental 
conditions.......................................................................................................................... 50 
Figure 4.4 Quantitative framework for predicting the rate-limiting phenomenon 
Application of the flow chart to Scenario #2 is highlighted by encircling the experimental 
conditions.......................................................................................................................... 51 
Figure 4.5 Quantitative framework for predicting the rate-limiting phenomenon 
Application of the flow chart to Scenario #3 is highlighted by encircling the experimental 
conditions.......................................................................................................................... 53 
Figure 5.1 Schematic diagram of experimental aquifer assembly with influent and 
effluent appurtenances, and sampling port identification................................................. 66 
Figure 5.2 Schematic of in-line flow meter. ..................................................................... 70 
Figure 5.3 Placement of Sampling Needles...................................................................... 76 
Figure 5.4 Glass ISFC with screens, sampling ports, and aluminum frame and cart. ...... 78 
Figure 5.5 Schematic of the permeameter for the constant head method......................... 83 
Figure 5.6 Schematic of the permeameter set up for the constant head method. ............. 84 
Figure 5.7 Sample dilution steps used for microbial plate counts. ................................. 141 
Figure 6.1 Comparison between breakthrough curves at sampling port B3 for the Br and 
fluorescein tracer experiments. ....................................................................................... 145 
Figure 6.2 Breakthrough curves from continuous point source fluorescein experimental 
data along with the model fitting data in the high-K layer. Graphs (a), (b) and (c) are data 
from triplicate experiments............................................................................................. 146 
 
 xii 
 
Figure 6.3 Breakthrough curves from continuous point source fluorescein experimental 
data along with the model fitting data in the low-K layer. Graphs (a), (b), (c) and (d) are 
data from replicate experiments...................................................................................... 147 
Figure 6.4 Typical plot of the rate of oxygen uptake vs. the cumulative oxygen uptake for 
a valid data set (Run 2) (an inoculum collected from the ISFC effluent during Scenario #1 
Phase I)............................................................................................................................ 158 
Figure 6.5 Cumulative oxygen uptake profiles during respirometry batch growth with 
Uper-1 from different cell sources: (a) an inoculum from the Uper-1 primary culture, (b) 
an inoculum aseptically collected from the ISFC effluent during the Scenario #1 Phase I 
experiment, and (c) an inoculum from the Uper-1 primary culture in MSNS with high 
salinity............................................................................................................................. 160 
Figure 6.6 Naphthalene sorption isotherm onto unmodified coarse sand....................... 164 
Figure 6.7 Naphthalene sorption isotherms onto HMN-Coated sands: (a) HMN-Coated 
Mystic White II coarse sand, and (b) HMN-Coated Filtersil fine sand.......................... 165 
Figure 6.8 Naphthalene sorption isotherms onto CPC-coated Mystic White II sand..... 169 
Figure 6.9 Naphthalene sorption isotherm onto XAD-2................................................. 171 
Figure 6.10 Naphthalene sorption isotherm onto XAD-7............................................... 174 
Figure 6.11 Naphthalene sorption isotherms onto the mixture of Mystic White II coarse 
sand and XAD-7 (10% w/w of XAD-7): (a) Sorbed concentration based on the XAD-7 
mass; and (b) Sorbed concentration based on the mass of the mixture of XAD-7 with 
Mystic White II coarse sand. .......................................................................................... 175 
Figure 6.12 Naphthalene sorption isotherms onto the mixture of Filtersil fine sand and 
XAD-7 (10% w/w of XAD-7): (a) Sorbed concentration based on the XAD-7 mass; and 
(b) Sorbed concentration based on the mass of the mixture of XAD-7 with Filtersil fine 
sand. ................................................................................................................................ 176 
Figure 6.13 Naphthalene sorption kinetics onto a mixture of coarse sand XAD-7. ....... 178 
Figure 6.14 Naphthalene desorption kinetics from a mixture of preloaded coarse sand and 
XAD-7............................................................................................................................. 178 
Figure 6.15 Operational conditions for Phase I control experiment: (a) flow rate; and (b) 
naphthalene concentration from the influent sampling port. .......................................... 182 
Figure 6.16 Example breakthrough curves from the Phase I control experiment for 
sampling ports: (a) A4, (b) B9, (c) C4 (d) C9, (e) E4 and (f) D9. Each graph shows the 
experimental naphthalene concentration along with the RT3D model prediction. ........ 184 
 
 xiii 
 
Figure 6.17 Example breakthrough curves from Phase II control experiment for sampling 
ports: (a) B9, (b) C9 and (d) D9. Each graph shows the naphthalene data from Phase IIA, 
Phase IIB and Phase I for comparison. ........................................................................... 189 
Figure 6.18 Breakthrough curves for influent and effluent from the Phase I (2): (a) 
Naphthalene, Br and DO concentration in the influent and (b) Normalized naphthalene 
and Br in the influent and effluent, and normalized DO in the effluent. ........................ 192 
Figure 6.19 Contour plot of naphthalene plume at 48 hours for Scenario #1 Phase I (2).
......................................................................................................................................... 194 
Figure 6.20 Example breakthrough curves from the Phase I (2) experiment for sampling 
ports: (a) A2, (b) B7, (c) C2 and (d) C7, (e) D1 and (f) E2. Each graph shows Br tracer, 
naphthalene, and DO data, as well as the RT3D predictions for Br and naphthalene.... 196 
Figure 6.21 Aqueous heterotrophic plate count data at sampling ports A2, C2, E2 and 
Effluent for Phase I (2). .................................................................................................. 202 
Figure 6.22 Breakthrough curves for influent and effluent from the Phase IIA: (a) 
Naphthalene, Br and DO concentration in the influent and (b) Normalized naphthalene 
and Br in the influent and effluent, and normalized DO in the effluent. ........................ 210 
Figure 6.23 Example breakthrough curves from Phase IIA experiment for sampling ports: 
(a) B7, (b) E2, (c) C2 and (d) C7. Each graph shows Br tracer data, naphthalene data, 
dissolved oxygen (DO) data and naphthalene data from Phase I (2) for comparison. ... 211 
Figure 6.24 Aqueous heterotrophic plate count data at sampling ports A2, C2, E2 and 
Effluent for Phase IIA..................................................................................................... 213 
Figure 6.25 Breakthrough curves for influent and effluent from the Phase IIB (2): (a) 
Naphthalene, Br and DO concentration in the influent and (b) Normalized naphthalene 
and Br in the influent and effluent, and normalized DO in the effluent. ........................ 216 
Figure 6.26 Example breakthrough curves from Phase IIB (2) experiment for sampling 
ports: (a) B7, (b) E2, (c) C2, (d) C7 and (e) D7. Each graph shows Br tracer data, 
naphthalene data and DO data along with naphthalene data from Phase I for comparison.
......................................................................................................................................... 217 
Figure 6.27 Aqueous heterotrophic plate count data at sampling ports A2, C2, E2 and 
Effluent for Phase IIB (2). .............................................................................................. 219 
Figure 6.28 Breakthrough curves for influent and effluent from the Phase I (2): (a) 
Naphthalene, Br and DO concentration in the influent and (b) Normalized naphthalene 
and Br in the influent and effluent, and normalized DO in the effluent. ........................ 225 
Figure 6.29 Example breakthrough curves from the Phase I (2) experiment for sampling 
ports: (a) B7, (b) C2, (c) C7, (d) D1 and (e) E2. Each graph shows Br tracer, naphthalene, 
RT3D fitting data for Br and naphthalene, and DO data. ............................................... 227 
 
 xiv 
 
Figure 6.30 Aqueous heterotrophic plate count data at sampling ports A2, C2, E2 and 
Effluent for Phase I (2). .................................................................................................. 230 
Figure 6.31 Breakthrough curves for influent and effluent from the Phase IIA: (a) 
Naphthalene, Br and DO concentration in the influent and (b) Normalized naphthalene 
and Br in the influent and effluent, and normalized DO in the effluent. ........................ 237 
Figure 6.32 Example breakthrough curves from Phase IIA experiment for sampling ports: 
(a) B7, (b) C2, (c) C7, (d) D1 and (e) E2. Each graph shows Br tracer data, naphthalene 
data, DO data and naphthalene data from Phase I (2) for comparison. .......................... 238 
Figure 6.33 Aqueous HPC at sampling ports A2, C2, E2 and Effluent for Phase IIA. .. 242 
Figure 6.34 Breakthrough curves for influent and effluent from the Phase IIB (2): (a) 
Naphthalene, Br and DO concentration in the influent and (b) Normalized naphthalene 
and Br in the influent and effluent, and normalized DO in the effluent. ........................ 245 
Figure 6.35 Example breakthrough curves from Phase IIB (2) experiment for sampling 
ports: (a) B7, (b) C2, (c) C7, (d) D1, (e) E2 and (f) Effluent. Each graph shows Br tracer 
data, naphthalene data and DO data along with naphthalene data from Phase I (2) for 
comparison...................................................................................................................... 246 
Figure 6.36 Breakthrough curves for influent and effluent from the Phase IIB (3): (a) 
Naphthalene, Br and DO concentration in the influent and (b) Normalized naphthalene 
and Br in the influent and effluent, and normalized DO in the effluent ......................... 249 
Figure 6.37 Example breakthrough curves from Phase IIB (3) experiment for sampling 
ports: (a) B7, (b) C2, (c) C7, (d) D1, (e) E2 and (f) Effluent. Each graph shows Br tracer 
data, naphthalene data and DO data along with naphthalene data from Phase I (2) for 
comparison...................................................................................................................... 250 
Figure 6.38 Aqueous heterotrophic plate count data at sampling ports A2, C2, E2 and 
Effluent for Phase IIB (3). .............................................................................................. 252 
 
 
 1 
 
Chapter 1 Introduction 
Organic contaminants in soils and sediments originate from a variety of 
anthropogenic activities. Many organic contaminants, such as benzene, toluene, 
ethylbenzene and xylenes (BTEX), polycyclic-aromatic hydrocarbons (PAHs), poly-
chlorinated biphenyls (PCBs), methyl-tertiary butyl ether (MTBE) and tertiary-butyl 
alcohol (TBA), chlorinated solvents, and pesticides and herbicides have documented or 
suspected mutagenic and carcinogenic effects. Therefore, the fate of these compounds at 
contaminated sites and restoration of these sites are of high concern with respect to 
environmental and public health. 
Bioremediation offers great potential for cleaning up environmental contaminants 
because: it can treat them in situ with little disturbance to the contaminated matrix; the 
contaminants can often be completely mineralized to inorganic materials (Head 1998); 
and bioremediation is relatively inexpensive compared to other remedial technologies 
such as incineration, soil washing, and pump and treat (Hughes et al. 1997; Singleton 
1994). However, the wide application of this technology has been hindered because of 
the uncertainty of success. The success of in situ bioremediation is made technologically 
challenging by the inherently complex and heterogeneous nature of the subsurface 
environment (National Research Council 1993). These physical and chemical 
heterogeneities of the subsurface occur at several scales and affect in situ bioremediation 
by controlling the availability of nutrients and substrates that drive the microbiological 
processes. This is important because many field and laboratory studies suggest that a 
large fraction of pollutants present in the environmental systems are unavailable for 
 
 2 
 
microbial degradation (Alexander 1995; Beck et al. 1995). Therefore, not understanding 
or accounting for the interactions between these scale-dependent physical/chemical 
heterogeneities and microbiological processes may reduce the effectiveness of field-scale 
in situ bioremediation. However, the interactions between these processes and in situ 
biodegradation, albeit important, are still not well understood. Indeed, reviews of field 
methods available for bioremediation have concluded that consideration of scale-
dependent phenomena, such as mass transport and interfacial transfer mechanisms, is a 
prerequisite to success in the field (Sturman et al. 1995). 
In order to describe subsurface heterogeneities and associated physical, chemical 
and microbiological processes, it is helpful to apply three scales of observation: macro-, 
meso- and micro-scale (Sturman et al. 1995). Advection, dispersion and geologic spatial 
heterogeneity are examples of macro-scale phenomena (Sturman et al. 1995). The 
corresponding physical scale for these phenomena ranges from ~10
-2
 to ~10
2
 m or even 
larger. The meso-scale is defined as the scale at which transport phenomena are apparent, 
with the exclusion of advection and mixing processes. Thus, the meso-scale phenomena 
include diffusion, sorption and interphase-mass transfer. Possible physical scales for 
meso-scale phenomena include the size of pore channels or soil particles, the 
characteristic diffusion length, or the dimension of microbial aggregates (10
-5
 to 10
-2
 m). 
The micro-scale is taken as the scale at which chemical and microbiological species exist 
and can be characterized independently of any transport phenomena. Examples of micro-
scale features are the composition of microbial consortia and the kinetics and 
stoichiometry of transformation reactions. The physical scale of these phenomena is the 
dimension of the microbial cell, on the order of ~10
-6
 to ~10
-5
 m. 
 
 3 
 
These scale-dependent physical and chemical heterogeneities and associated 
processes in the subsurface create, either directly or indirectly, interfaces between two 
phases, where strong contrasts in physical and chemical properties exist over short 
distances (centimeter to meters). The physical/chemical interface zones comprise a wide 
variety of types including: aqueous-solid interfaces; aqueous-nonaqueous phase liquid 
(NAPL) interfaces; and aqueous-aqueous interfaces. Specifically, such strong contrasts in 
physical and chemical properties at these interfaces control moisture nutrient fluxes and 
redox conditions, which, in turn, drive the distribution and activities of microbes in the 
subsurface, e.g., (McMahon and Chapelle 1991). For example, many organic 
contaminants of concern (e.g., PAHs and PCBs) are hydrophobic, resulting in low 
aqueous solubility and a strong tendency to sorb to the matrix of organic material in soil 
or to partition into oily phases at contaminated sites (Harayama 1997). Due to these 
chemical characteristics, the bioavailability of sorbed contaminants in natural sorbents 
has been an important issue in applying bioremediation technologies to remediation of 
contaminated sites (Willumsem and Arvin 1999).  
Importantly, several of these scale-dependent, interfacial processes are occurring 
and impacting biodegradation simultaneously at any site. Hence, Rittmann et al. (1992) 
concluded that the major research needs required to facilitate bioremediation success in 
the field are the means to quantify mass-transfer kinetics and microbial kinetics - all of 
which are scale-dependent. Therefore, an integrated kinetic study of several such 
interfacial physicochemical mass-transfer and biokinetic processes occurring 
simultaneously at different scales is of significant importance in understanding the 
bioavailability of pollutants, which in turn impacts the effectiveness of bioremediation.
 
 4 
 
Chapter 2 Objective and Scope of Work 
Despite the importance of heterogeneities in bioremediation, the impact of 
geochemical and hydraulic interfaces on in situ biodegradation is still not well 
understood. Because the subsurface physical, chemical, and microbiological 
processes and their interactions are extremely complex, it is useful to apply the scales 
of heterogeneity as an organizational tool and use relevant dimensionless parameters 
to succinctly capture the complexity of the interactions between these processes. 
Therefore, it is hypothesized that using the scales of subsurface heterogeneities and 
associated interfacial processes as the organizational principle, a quantitative 
framework based on a set of dimensionless numbers can be developed to capture the 
effects of the competing interfacial and biokinetic processes and to define the limits 
for the successful application of in situ bioremediation.  
The overall goal of this research, and the larger project of which it was a part, 
was to develop such a quantitative framework and to evaluate its utility under 
different scenarios relevant to the subsurface. To accomplish this, a systematic and 
integrated modeling and laboratory investigation was used to evaluate the impact of a 
wide range of scales and magnitudes of heterogeneities on in situ bioremediation and 
the utility of the developed quantitative framework for predicting what is the overall 
limiting process and what engineering actions will positively impact the in situ 
biodegradation rates. This investigation had three specific major objectives.  
The first objective was to develop and refine the proposed integrated 
quantitative framework of dimensionless numbers. A systematic comparison of the 
dimensionless numbers was used to integrate and compare all the phenomena 
 
 5 
 
occurring at different scales of heterogeneities and to predict the overall rate-limiting 
process for a given environmental system. In the framework as used in this work, 
there are three steps. The first two steps are used to identify the slowest mass-transfer 
process, and the third step compares that process rate with the biokinetics to 
determine the overall rate-limiting process. A mathematical model, Reactive Multi-
species Transport in 3-Dimensional (RT3D) Groundwater Aquifers, which 
mathematically describes the relationships between the processes of interest in natural 
subsurface systems, was used to provide a preliminary evaluation of the proposed 
framework.   
The second objective was to design and build an intermediate-scale flow cell 
(ISFC) system that represented a realistic, yet experimentally tractable natural 
subsurface system for evaluating the physical, chemical, and microbiological 
processes occurring at heterogeneous interfaces, especially for aqueous-aqueous and 
aqueous-solid interfaces. As part of this objective, batch experiments were also 
conducted to measure the kinetic parameters describing groundwater flow and 
contaminant fate and transport in the simulated subsurface system in order to apply 
the quantitative framework and model the system using RT3D. 
The final objective was to utilize three specially selected scenarios to evaluate 
the quantitative framework with both modeling and experimental investigations. The 
scenarios tested include cases in which either macro-scale mixing (Scenario #1), 
biokinetics (Scenario #2) or sorption/desorption (Scenario #3) was the rate-limiting 
process controlling the overall contaminant biodegradation rate and preventing 
formation of a biologically active zone (BAZ). These investigations were broken 
 
 6 
 
down into two phases. Phase I represented biodegradation under natural conditions, 
as simulated in the ISFC. The results from these experiments were used to evaluate if 
the factors limiting the biodegradation rate under natural conditions were what was 
expected based on the quantitative framework. Phase II represented perturbations to 
the system as part of a simulated engineered in situ bioremediation. Engineered 
remedial actions involve efforts to overcome the mass-transfer and microbial 
limitations and create a BAZ where it is needed. Importantly, only an appropriate 
remedial strategy that alleviates the overall rate-limiting phenomenon will enhance 
the in situ biodegradation rate. Therefore, the results of the Phase II experiments were 
used to test the quantitative framework utility for predicting what, if any, engineered 
actions would augment the intrinsic in situ biodegradation. 
 In the next chapter (Chapter 3), background literature is reviewed with respect 
to the key processes relevant to this project that occur at the macro-scale (e.g., 
advection and dispersion), meso-scale (e.g., sorption/desorption), and micro-scale 
(e.g., microbial activities). Subsequent chapters review the theoretical concepts 
underlying this research (Chapter 4), as well as the experimental materials and 
methods (Chapter 5), and the results and discussion (Chapter 6). Finally, the 
conclusions are outlined in the last chapter (Chapter 7), along with recommendations 
for future work. 
 
 7 
 
Chapter 3  Literature Review 
As introduced in Chapter 1, the subsurface is a geologically, chemically, and 
microbiologically heterogeneous environment. Furthermore, the scale-dependent 
physical and chemical heterogeneities create various types of interfacial zones, such 
as aqueous-solid interfaces, aqueous-NAPL interfaces, and aqueous-aqueous 
interfaces. These interfaces have physical and chemical properties that are different 
from any one of the individual phases involved and, as a result, their presence can 
support a greater density, activity and/or diversity of microorganisms, which is of 
great concern for bioremediation.  
This situation occurs because the subsurface has geological units with 
different chemical compositions that exist at different scales, such as strata, beds, 
laminae, and pores. For example, at a pore scale, different phases and different types 
of sedimentary organic matter may exist, creating different aqueous-solid interfaces 
and influencing rates of substrate flux via sorption and desorption, the rate of which 
can affect the microbial ecology and biodegradation rate. Therefore, accurate 
prediction of the transport and fate of organic solutes in groundwater requires an 
understanding of the controlling hydrogeologic chemical and microbiological 
processes and their interactions. 
Thus, the subsurface should always be investigated at different scales, the 
scale of interest being a function of the scientific question or applied problem. For 
example, the most appropriate scales for studying bioremediation design and 
engineering include the macro-, meso- and micro-scales as discussed above and 
defined further below. Although these scale definitions are arbitrary, they serve as a 
 
 8 
 
useful conceptual structure for approaching the engineering problem. The goal of this 
chapter is to review the current state of knowledge of these heterogeneities and 
associated interfaces and their impact on microbiological processes. For convenience, 
the discussion is broken down into three scales: macro-, meso-, and micro-scale. 
3.1 Macro -scale heterogeneities 
Macro-scale heterogeneity refers to heterogeneity at the scale of laminae, thin 
lenses or beds, or inclusions (scales of millimeters to 10s of centimeters thick) 
(Brockman and Murray 1997).  At the macro-scale, mass transport via advection and 
dispersion can significantly impact contaminant distribution and the availability of 
substrates, nutrients, and electron acceptors to microorganisms (Sturman et al. 1995), 
which may, in turn, drive the distribution and activity of the microbes in the 
subsurface (Brockman and Murray 1997).  
Advection refers to the movement of dissolved contaminants due to the bulk 
flow of groundwater, which is important in bioremediation in terms of the transport of 
dissolved microbial substrates. According to Darcy?s law, the one dimensional 
volumetric flow rate, Q [L
3
 T
-1
], through a cross sectional area, A [L
2
], of a porous 
medium can be described as (Fetter 2001), 
dx
dh
K
A
Q
?=       (3.1) 
where: K = hydraulic conductivity, [LT
-1
]; and 
dx
dh
 = hydraulic gradient in the 
direction of groundwater flow. The magnitude of K varies widely, and for a given 
geologic formation, is a function of a variety of physical factors, e.g., porosity, 
particle size and distribution, particle shape, particle arrangement, and secondary 
 
 9 
 
features such as fracturing and dissolution. Using Equation 3.1, and accounting for 
the fact that flow only passes through the pore openings of the full cross-sectional 
area, the average pore-water velocity, v  , can be calculated for one-dimension as, 
dx
dhK
A
Q
v
??
?==     (3.2) 
where: A? = the effective area of flow, [L
2
]; and ? = effective porosity, i.e., the 
percentage of interconnected pore space.  
Another key subsurface process at the macro-scale that has an important role 
in the transport/mixing and biodegradation of contaminants is hydrodynamic 
dispersion. It refers to the combined effect of mechanical dispersion and molecular 
diffusion (Fetter, 2001). Importantly, diffusion and heterogeneity induced mechanical 
dispersion are the only mixing processes for solute in the deep subsurface. The 
mathematical description of hydrodynamic dispersion for a two-dimensional domain 
is, 
2
2
2
2
z
C
D
x
C
D
t
C
ZX
?
?
+
?
?
=
?
?
   (3.3) 
where: *DvD
XX
+=? = longitudinal hydrodynamic dispersion coefficient, [L
2
T
-1
]; 
=
X
? longitudinal dispersivity, [L]; *DvD
xZZ
+=? = vertical transverse 
hydrodynamic dispersion coefficient, [L
2
T
-1
]; =
Z
? vertical transverse dispersivity, 
[L]; =
x
v  average pore water velocity in x-direction [LT
-1
]; and D* = effective 
diffusion coefficient [L
2
T
-1
]. In the following discussion, the term "dispersion" is 
meant the hydrodynamic dispersion. 
 
 10 
 
 In particular, vertical dispersion is an important process for developing zones 
of mixing in the subsurface (Sudicky et al. 1985). However, vertical transverse 
dispersion has generally been found to be a weak process in groundwater systems 
(Gelhar et al. 1992). Therefore, both mechanical dispersion and molecular diffusion 
can have a significant effect on the vertical transverse dispersion for a wide range of 
relevant pore water velocities because of low values of vertical transverse 
dispersivities. Two examples of situations in which vertical transverse dispersion is 
critical are: oxygen transfer across the water table, and heterogeneities in hydraulic 
conductivity. Theoretical and field studies, e.g., (Borden and Bedient 1986; Borden et 
al. 1986; Thornton et al. 2001) indicate that the vertical transverse dispersion 
coefficient can have a significant impact on oxygen exchange between the 
unsaturated and saturated zones and, thus, affect the aerobic microbial activities. 
Because of the weakness of the vertical dispersion process and the magnitude of the 
microbial oxygen demand in contaminant plumes, steep vertical oxygen gradients 
result, and the impact of transverse dispersion on oxygen supply is limited to 
relatively shallow plumes (e.g., less than about 2 to 3 m below the water table) 
(MacQuarrie and Sudicky 1990). 
Hydraulic layer interfaces in groundwater systems resulting from 
heterogeneities in hydraulic conductivity, K, may be particularly important to 
microorganisms because interlayer mass transfer of solutes can increase the supply of 
limiting nutrients near the interface. For example, one key macro-scale interfacial 
process that can significantly impact in situ bioremediation is the hydraulic mixing at 
the interface between the contaminant plume and the ?clean? groundwater, which is 
 
 11 
 
induced by hydraulic conductivity heterogeneity. Indeed, Carrera (1993) comments 
that, ??most of the difference between the actual behavior of solutes and that 
predicted by the advection-dispersion equation can be attributed to the spatial 
variability of hydraulic conductivity?. 
Laboratory and modeling studies indicate that varying layers of hydraulic 
conductivity are an important factor affecting in situ biodegradation, with 
contaminants predominantly persisting in low hydraulic conductivity layers, e.g., 
(Murphy et al. 1997; Szecsody et al. 1994; Wood et al. 1994; Yang et al. 1994). For 
example, Szecsody et al. (1994) performed a laboratory study of the transport and 
biodegradation of quinoline in a two-dimensional, horizontally-stratified porous 
media under dual substrate limitation. Szecsody et al. (1994) found that the interlayer 
mass transfer resulted in arrival of substrate and oxygen 10?s to 100?s of hours sooner 
in the low-K layer near the interface compared to other locations within the low-K 
layer, where substrates arrived via only advection. Early arrival of substrates near the 
interface resulted in biodegradation of quinoline for a longer period than within 
layers, yielding increased growth in a 1-3 cm thick zone. Although these studies 
illustrate the importance of understanding processes near interfaces in subsurface 
systems, the data are only suggestive, as the studies were not designed to identify and 
quantify these mechanisms. 
In addition to systems with varying layers of hydraulic conductivity, hydraulic 
conductivity heterogeneity can also occur at a smaller scale (> 10
-3
 to < 1m) than that 
typically associated with measurement of ?field-scale? heterogeneity. The existence 
of regions of small hydraulic conductivity within the flow domain creates a spatially 
 
 12 
 
variable velocity field at these smaller scales. In some cases, there is a relatively 
minimal flow and advection in the low-conductivity domains. Due to the small 
advective flux, these domains act as sink/source components, with rate-limited 
diffusional mass transfer between the advective and nonadvective domains causing 
spreading of the solute front. These sink/source regions can take various forms, 
including the internal porosity of aggregates or porous particles, dead-end pores, the 
bulk matrix of fractured media, and the small conductivity micro-layers or laminae 
typically found in aquifers of sedimentary origin. For example, some researchers have 
shown that these structures can cause asymmetrical and tailed breakthrough curves 
and enhanced dispersion (Brusseau 1994; Brusseau and Rao 1991).  
Importantly for this study, modeling and experimental studies (Murphy et al. 
1997) have demonstrated that small-scale K heterogeneity combined with dual 
substrate limitation can result in the creation of regions of increased electron donor 
and electron acceptor mixing and enhanced microbial activity and growth. However, 
Schafer and Kinzelbach (1992), using stochastic modeling of in situ bioremediation, 
demonstrated that small scale K heterogeneity with dual substrate limitation can also 
decrease contaminant biodegradation, e.g., by preventing electron donor and acceptor 
mixing. Therefore, hydraulic conductivity heterogeneities can have significant 
impacts on microbial activity and biomass production, sometimes positive and 
sometimes negative.  
3.2 Meso/pore-scale heterogeneity 
Bioavailability of organic contaminants has long been identified as a major 
limitation to complete bioremediation of contaminated subsurface environments, 
 
 13 
 
affecting clean-up time, cost, and the end-point of the process (Head 1998). In 
particular, the low mass-transfer rate caused by strong sorptive interactions with 
subsurface organic matter and intraparticle diffusion limitations is one of the main 
factors responsible for the low bioavailability of PAHs in subsurface systems 
(Brusseau et al. 1991; Guerin and Boyd 1992; Volkering 1996). In attempting to 
understand how sorption/desorption and biodegradation processes interactively 
control bioavailability, it must be considered that biodegradation of sorbed materials 
involves at least two rate processes: sorption/desorption and biodegradation. 
Due to the field-scale spatial variability in hydraulic conductivity, it is logical 
to expect meso-scale reaction-related properties in the subsurface to be spatially 
variable as well. For example, it has been shown that spatially variable sorption can 
cause non-ideal transport (Brusseau 1991; Brusseau and Zachara 1993). Such 
physical and chemical heterogeneity at the meso/pore-scale may result in 
microbiological heterogeneity in terms of different microbial physiological types, or 
the presence or absence of microbes. The research into the bioavailability of organic 
pollutants as microbial substrates has largely focused on pore-scale chemical 
heterogeneity resulting from interfacial phenomena such as sorption to, and 
desorption from, sediment particles, and that is the focus here. 
 Contaminant sorption can have a variety of different types of effects on 
microbial activity. Of primary interest in this study is the impact of sorption on 
contaminant biodegradation. For example, if the sorption sink is significant, it can 
reduce the aqueous phase solute concentration. If only the substrate in solution is 
available for biodegradation, sorption will thus reduce the rate and extent of 
 
 14 
 
biodegradation, as has been demonstrated in some studies, e.g., (Mihelcic and Luthy 
1991; Miller and Alexander 1991; Rijnaarts et al. 1990). However, there has been 
some debate regarding the possible occurrence of direct bioavailability, whereby 
organisms can directly utilize the sorbed chemicals, as opposed to indirect 
bioavailability, in which the sorbed chemicals can only be utilized by mcrobes after 
desorption into the aqueous phase.  
Most research addressing the issue of sorbed solute bioavailability has 
concluded that sorbed PAHs are not directly available for biodegradation (Smith et al. 
1992). Organic polymers within the soil attract organic contaminants like PAHs via 
their non-polar gel structure. Once within the polymer matrix, the contaminants may 
bond (via covalent or hydrogen bonding) to specific sites, or may diffuse through 
micro-pores in the gel. It has been suggested that if such tightly bound contaminants 
desorb at all, it is at a rate that certainly limits biodegradation. For example, Angley et 
al. (1992) concluded that biodegradation of alkylbenzenes occurs only in solution and 
that desorption was generally diffusion-rate limited. Ogram et al. (1985) proposed 
three models to describe (2,4-dichlorophenoxy) acetic acid biodegradation in soil. 
The model that most successfully described their experimental data assumed that both 
attached and planktonic bacteria were capable of utilizing dissolved contaminant, but 
neither was able to directly degrade sorbed contaminant. The authors concluded that 
this was either because the contaminant may have been adsorbed sufficiently deep 
within the soil-organic matter matrix that bacteria were unable to reach it, or because 
enzymes responsible for degradation may have been incapable of attacking sorbed 
contaminants. In another attempt to address the question of whether desorption limits 
 
 15 
 
biodegradation, Rijnaarts et al. (1990), investigated hexachlorocyclohexane 
biodegradation in a mixed soil system that was desorption limited. In this case, 
contaminant diffusion through the soil micropores was found to limit the 
biodegradation rate. To further investigate the effects of desorption rate on 
biodegradation, Fry and Istok (1994) utilized an analytical solution to the solute 
transport equation with rate-limited desorption and first-order decay. Their results 
showed that when the desorption coefficient is small relative to the degradation 
coefficient (desorption is the rate-limiting process), increasing the biodegradation rate 
any further will not improve the performance of in situ bioremediation. Similarly, 
Bouwer et al. (1994) found that naphthalene biodegradation proceeded more slowly 
in high organic carbon system, suggesting desorption rate limitation.  
However, some researchers have concluded that enhanced biodegradation can 
occur, that is, the microbial utilization of a compound can occur at a rate faster than 
its abiotic desorption into the bulk-water phase. This requires a mechanism that 
facilitates degradation of sorbed chemicals. Although the mechanisms of microbial 
uptake that lead to such results are not precisely known, several possibilities have 
been proposed. One possible mechanism involves the in situ production of 
biosurfactants, which may enhance desorption of hydrocarbons from soil, e.g., 
(Scheibenbogen et al. 1994). A second possible mechanism is the direct utilization of 
chemicals by microorganisms, possibly involving direct contact between the 
hydrophobic cell surface and the chemicals, e.g., (Calvillo and Alexander 1996; 
Efroymson and Alexander 1991). Interestingly, Guerin and Boyd (1992) performed 
experiments using naphthalene that indicated the potential for direct uptake may be 
 
 16 
 
organism specific. Thus, the uptake mechanism for the sorbed PAHs might also vary 
depending on the consortia present. A third possible mechanism is reduced mass 
transfer resistance brought about by the microorganisms being located very near the 
solid surface, which greatly reduces the diffusion distance from sorption site to the 
microorganisms, e.g., (Chang and Rittmann 1987; Harms and Zehnder 1995). Indeed, 
several researchers have also observed biodegradation of sorbed contaminants via 
attached organisms in sediments and granular activated carbon, e.g., (Speitel and 
DiGiano 1987), suggesting that the attached organisms may enhance contaminant 
concentration gradients, thereby promoting desorption directly into the biofilm.  
Another possible explanation for the observation of desorption limitation in 
some experiments and not in others, maybe be the differences in the 
sorption/desorption properties of the solid phase material. For example, Brusseau 
(1992) was able to predict contaminant breakthrough for four different aquifer 
systems by representing sorption as a dual process. Specifically, sorption was 
modeled to be instantaneous for a portion of the aquifer material while being 
diffusion rate-limited for the remainder. Thus, the fraction of the sorption/desorption 
that is diffusion rate-limited will influence the bioavailability. 
Based on these studies, the bioavailability of sorbed chemicals appears to 
depend on the properties of microorganisms present as well as the characteristics of 
the sorbents and experimental methodology. Unfortunately, variations among the 
studies in terms of the soil organic matter content and composition, contaminant 
hydrophobicity, and the length of time the contaminant had been in place, make 
generalization about sorptive behavior and its impact on biodegradation difficult. 
 
 17 
 
 A variety of conceptual and mathematical models have been developed to 
describe the sorption of PAHs on soils and sediments as summarized by Weber et al. 
(1991), among others. The sorption term can be described via equilibrium and non-
equilibrium rate models. There are several different equations or models available for 
describing the equilibrium partitioning of a sorbate between the aqueous solution and 
the sorbent at a constant temperature, i.e., the sorption isotherm.  This relationship 
between the contaminant concentrations in both phases can be described by either 
linear or nonlinear isotherms. The simple linear isotherm assumes that there are an 
infinite number of equal sites, independent of each other, available for sorption. The 
equation is, 
    CKC
d
=
~
     (3.4) 
where: C = the concentration of contaminant in the mobile phase, [ML
-3
]; C
~
 = the 
concentration of the contaminant in the immobile phase (mass of the contaminant per 
unit mass of porous media), [MM
-1
] ; and
d
K = the linear partitioning coefficient, 
[L
3
M
-1
]. 
 Several nonlinear isotherm models are also available, but the most common 
alternatives to the linear isotherm are the Langmuir isotherm, 
    
bC
abC
C
+
=
1
~
     (3.5) 
 and the Freundlich isotherm: 
    
n
d
CKC =
~
     (3.6) 
 
 18 
 
where a, b, and n (0<n<1 or n>1) are empirical constants (Tchobanoglous and Burton 
1991). In particular, the Freundlich model is frequently applied to describe sorption of 
organic chemicals from groundwater. 
Nonequilibrium sorption conditions have also been observed under field 
conditions, e.g., (Brusseau and Rao 1989). There are many mathematical models that 
have been developed to describe the sorption/desorption process. Based on the rate 
characteristics, three types of models were used widely in the past research: (1) a 
first-order rate model; (2) a second-order rate model; and (3) a dual-resistance 
diffusion model. 
 Several investigations have applied a simple approach to describe the mass-
transfer kinetics between soil and the mobile aqueous phase, with a linear driving 
force and a lumped first-order mass-transfer coefficient (Lapidus and Amundson 
1952): 
?
?
?
?
?
?
?
?
?=
?
?
d
ls
K
C
CK
t
C
~~
?
?
   (3.7) 
where:? = the bulk density of the soil matrix, [ML
-3
]; 
ls
K  = the lumped first-order, 
mass-transfer rate parameter, [T
-1
] =
w
lssa
S
ka
?
 , where == VAa
sasa
/specifc 
soil/aqueous interfacial surface area, [L
-1
], 
ls
k = the mass-transfer-rate coefficient, 
[LT
-1
]; and S
w
 = water saturation. This modeling approach is justified when mass-
transfer is fast, i.e., when the diffusional lengths are extremely short.  
 The second-order model describes the sorption as a second-order process. 
However, this model has received only minor attention for the characterization of 
 
 19 
 
sorption data in natural solid systems (Weber and Miller 1988), and is not discussed 
further here. 
  Diffusion formulations have been used widely in natural systems for nonlinear 
and hysteretic sorption equilibrium (Brusseau 1992; Miller 1984). The dual-resistance 
diffusion model describes sorption as a process of coupled mass transfer through a 
boundary film external to the solid particle followed by diffusion within a 
representative portion of the particle of itself (Weber and Miller 1988). For spherical 
solid particles, the intraparticle-diffusion process can be described by the following 
mass balance (Mulder et al. 2000; Mulder et al. 2001): 
)
2
()1(
2
2
r
C
rr
C
D
t
Q
t
C
effs
?
?
+
?
?
=?
?
?
+
?
?
???   (3.8) 
where ? is the particle volumetric porosity (m
3
/m
3
), Q is the sorbed naphthalene 
concentration in the pores of solid particles (kg/kg), t is the time (s), 
s
? is the skeletal 
density of the solid phase (kg/m
3
), 
eff
D is the effective diffusion coefficient through 
the porous matrix ( sm /
2
), and r is the distance from the center of the particle (m). 
The equation was developed based on the following assumptions: (1) the solid 
particles have a spherical shape, (2) the diffusion coefficient is concentration 
independent, and (3) local equilibrium is reached instantaneously at a certain location 
in the porous particle. The effective diffusion coefficient is defined as the binary 
diffusion coefficient of the naphthalene in dilute water solutions corrected for 
tortuosity and constrictivity effects imposed on this coefficient by the matrix 
geometry by means of a lumped matrix factor (? ) (Mulder et al. 2000; Mulder et al. 
2001) 
 
 20 
 
?
?
AB
eff
D
D =      (3.9) 
where 
AB
D is the binary diffusion coefficient of the naphthalene in water (m
2
/s) and k 
is the dimensionless matrix factor. 
In addition to intraparticle mass transfer resistance, diffusion across a stagnant 
liquid film surrounding the particle influences the overall transfer rate to a well mixed 
aqueous phase. This additional external mass transfer resistance can be modeled by 
assuming a linear concentration gradient across the laminar liquid film and thus 
defining the mass flux through the film as  
)(
bRrl
CCkN ?=
=
?     (3.10) 
where N is the mass flux through the laminar layer (kg/m
2
/s), 
l
k is the film mass 
transfer coefficient of naphthalene (m/s), r is the distance from the center of the 
particle (m), R is the particle radius, 
b
C  is the bulk liquid naphthalene concentration 
(kg/m
3
), and 
Rr
C
=
 is the naphthalene concentration at the interface of the particle with 
the laminar layer (kg/m
3
). 
In turbulent flow reactors, a combination of film diffusion and intraparticle 
diffusion very often controls the sorption rate. Film diffusion may control initially; 
then, intraparticle transport may control after naphthalene accumulates within the 
pore. To determine whether film transfer or intraparticle mass transfer is limiting the 
overall sorption/desorption rate of naphthalene, the dimensionless Biot number is 
used (Mulder et al., 2000) 
eff
l
D
Rk
Bi =      (3.1) 
 
 21 
 
At high Biot numbers ( 1>>Bi ), film diffusion limitations can be neglected and 
intraparticle mass transfer is the rate-limiting step. At low Biot numbers ( 1<<Bi ), 
the film diffusion resistances dominate the over sorption/desorption process. 
In addition to the dual-resistance diffusion model, Brusseau et al. (1991) 
utilized another approach and assumed that the intra-organic matter diffusion was 
responsible for the slow desorption. Accordingly, these authors developed a 
bicontinuum model with two sorption domains, one with instantaneous exchange of 
contaminant and one in which the exchange was considered to be rate-limited. This 
model is known as the intra-organic matter diffusion model. 
3.3 Micro-scale heterogeneities 
Microbial-mediated contaminant transformation reactions occur at the micro-
scale. Importantly, the stoichiometry and kinetics of microbially-mediated 
biodegradation are two of the key micro-scale features of interest (Sturman et al. 
1995). With respect to this project, the primary focus is on the stoichiometry and 
kinetics of naphthalene biodegradation. Many different species of bacteria (both 
gram-negative and gram-positive), fungi, and algae are known to degrade PAHs, such 
as naphthalene. As PAHs are naturally occurring compounds, it is not surprising that 
many different PAH-degrading microorganisms can be found in pristine 
environments (Sims and Overcash 1983). Most of the microorganisms that have been 
isolated, however, originate from PAH-contaminated sites.  
The microbial metabolism of PAHs like naphthalene has been studied 
extensively, e.g., (Sutherland et al. 1995). In fact, examples of bacteria, fungi, yeasts, 
cyanobacteria, and algae have been shown to have the enzymatic capacity to oxidize 
 
 22 
 
PAHs. Because the bacterial metabolism of naphthalene under aerobic and is of 
importance for this research, it is reviewed here in detail. However, naphthalene 
degradation under anaerobic conditions is also reviewed briefly for completeness.  
An overview of the specific steps in the aerobic naphthalene degradation 
pathway by Pseudomonas spp. is presented in Figure 3.1. The initial attack on an 
aromatic ring is usually performed by a dioxygenase, forming naphthalene cis-1, 2-
dihydrodiol. This is then converted by a dehydrogenase into a dihydroxylated 
derivative, 1, 2-dihydroxynaphthalene. In the next step in the degradation pathway, a 
second dioxygenase oxidizes 1, 2-dihydroxynaphthalene to 2-hydroxychromene-2-
carboxylic acid (HCCA). HCCA is then converted to an isomer of trans-o-
hydroxybenzylidene pyruvic acid by an isomerase. Next, an aldolase converts the 
trans-o-hdyroxybenzylidene pyruvate to salicylaldehyde, which is converted to 
salicylate by salicylaldehyde dihydrogenase. The decarboxylation and hydroxylation 
of salicylate to catechol is then catalyzed by salicylate hydroxylase. Catechol is next 
transformed by catechol 2, 3-dioxygenase via meta-cleavage to 2-hydroxymuconic 
semi-aldehyde, which is further degraded to completion.  
 
 23 
 
 
Figure 3.1 Naphthalene degradation pathway under aerobic condition 
(Zeng 2004)
 
 24 
 
Recent studies have also demonstrated PAH degradation under nitrate-
reducing condition (Macrae and Kenneth 1998; Mihelcic and Luthy 1991; Rockne et 
al. 2000; Rockne and Strand 1998; Rockne and Strand 2001). For example, Rockne 
and Strand (2001) studied anaerobic naphthalene degradation under nitrate-reducing 
conditions and reported that the naphthalene degrading pure cultures they obtained 
were related to Pseudomonas stutzeri and Vibrio pelagius. In addition, degradation of 
naphthalene under methanogenic condtions has been clearly documented, as 
illustrated in Figure  
3.2 (Grblic-Galic 1988). The methoanogenic degradation pathway for naphthalene 
follows a similar route to that of monoaromatic hydrocarbons, in which phenol 
appears to be a major intermediate. Furthermore, laboratory studies of enriched 
sulphidogenic consortia have shown that naphthalene is metabolized via pathways 
(Figure 3.3) that lead to the attachment of a carboxylic acid group to one of the 
aromatic ring, resulting in the formation of 2-napthonic acid (2-NA) which is further 
degraded by sequential ring-reduction mechanisms through 5, 6,7,8-tetrahydro-2-
naphthoic acid (TH-2-NA), hexahydro-2-naphthoic acid (HH-A-NA) and decalin-2-
carboxylic acid (D-2-CA) before ring cleavage (Zhang et al. 2000). Further studies 
are required to determine the metabolic pathways of anaerobic naphthalene 
mineralization. This knowledge may help to identify potential cometabolites and 
determine the potential for anaerobic naphthalene utilization in the environment 
(Rockne et al. 2000).
 
 25 
 
 
Figure 3.2 Naphthalene degradation pathway under methanogenic conditions 
(Grblic-Galic 1988) 
 
 
 26 
 
 
Figure 3.3 Degradative pathway of naphthalene under sulfate-reducing conditions 
(Zhang et al. 2000)
 
 27 
 
Biodegradation rates in contaminated aquifers may vary due to variations in 
microbial species (numbers and activities), which are related to availability of 
nutrients and electron donors/acceptors, toxicity of contaminants, predation, and other 
factors. The heterogeneities in a subsurface system at the micro-scale are related to: 
(1) the microscopic distribution of microbes and pollutants/substrates, and (2) 
microbial transport. Because in this work it is assumed that microbes have already 
been attached to the solid phase, the discussion below focuses on the microscopic 
distribution of microbes and pollutants and the kinetics of PAH biodegradation under 
aerobic conditions. 
The microscopic distributions of microbes and pollutants have been studied as 
well as how those distributions affect biodegradation in subsurface systems. 
Interestingly, soils and aquifers are desolate environments from the point of view 
microbes (Bosma et al. 1997), which impacts their micro-scale distribution. The 
average distance between colonies containing up to 100 individual cells can be 
estimated to be at least 100 ?m based upon bacterial densities and the observation that 
bacteria mainly live in pores with sizes of 0.8-3 ?m. On the other hand, pollutants can 
be present in extremely small pores. Upon entering a porous system in the subsurface, 
pollutants first contaminate the macropores and the particle surfaces containing 
relatively very few bacteria. Then, the pollutants diffuse into smaller pores where 
biotransformation may take place when the environmental conditions favor microbial 
activity. However, pollutants can also slowly diffuse into extremely small pores (<< 1 
?m) where microorganisms are absent. Thus, pollutants and bacteria have a different 
microscopic distribution in contaminated subsurface materials. As a result, the 
 
 28 
 
biodegradation rate and extent are affected by a number of physical-chemical 
processes related to substrate bioavailability such as sorption and desorption, 
diffusion, and dissolution, as discussed above, because it is generally assumed that 
the pollutants must diffuse to the bacteria before they can be taken up and degraded. 
Therefore, it is very necessary to study the distribution of the microorganisms and 
pollutants at the micro-scale, which, in turn, is influenced by the heterogeneities in 
natural systems. 
A variety of approaches have been used to describe biodegradation kinetics in 
subsurface systems. Three different conceptual models have been used to describe 
biodegradation in the subsurface (Baveye and Valocchi 1989): (1) the strictly 
macroscopic model, which assumes the bacteria are immobilized and the biokinetics 
are a function of the macroscopic biomass and bulk fluid substrate concentrations; (2) 
the microcolony model, which assumes the biomass is present as small colonies 
growing on solid surface, and accounts for external mass transfer resistance from the 
bulk solution to the colony surface; and (3) the biofilm model, which assumes that the 
bacteria and their extracellular products are distributed as a continuous film on the 
solid surface, and accounts for external mass transport resistance as well as internal 
mass transfer resistance due to diffusion within the biofilm. Odencrantz (1991) 
compared the macroscopic and biofilm models for realistic groundwater conditions, 
and found that the two solutions converged for the organic substrate plume and 
biomass distribution. Therefore, the added complexity of the biofilm model (and the 
micro-colony model) is probably not needed for modeling the solute concentration for 
model groundwater scenarios. Further, the macroscopic model does not require any 
 
 29 
 
assumptions regarding the spatial distribution of the biomass, nor does it require 
determination of biofilm parameters (Baveye and Valocchi 1989). Thus, the strictly 
macroscopic model was selected for this research, as described in Chapter 4. 
In addition to the conceptual model, a mathematical model for the substrate 
utilization rate is required too. Several such models have been used including the: (1) 
first-order model, (2) second-order model, (3) instantaneous reaction model, and (4) 
Monod models. As described above, in systems with heterogeneities and the 
associated interfacial mass transfer processes, the resulting mixing of the limiting 
substrates, may cause both the electron donor and electron acceptor substrates to each 
limit the biokinetics. Therefore, it is necessary to describe the microbial growth and 
substrate utilization with an equation that accounts for dual-substrate limitation. The 
term dual-substrate limitation refers to a type of multiple-substrate limitation in which 
the electron-donor and electron-acceptor substrates together limit the overall 
biodegradation rate (Bae and Rittmann 1996). A commonly applied model for this 
case is the double Monod equation which has been utilized in the advection-
dispersion transport equation to predict the fate of pollutants in the subsurface: 
?
?
?
?
?
?
?
?
+
?
?
?
?
?
?
?
?
+
?=
AK
A
SK
S
Mqq
AS
m
   (3.12) 
where: =
m
q maximum specific substrate utilization rate, [MM
-1
T
-1
]; M = biomass 
concentration (pore volume basis), [ML
-3
]; and K
S
 and K
A
 = half maximum rate 
constants for substrate and electron-acceptor, respectively, [ML
-3
]. The model 
assumes that, if electron donor and electron acceptor are present at subsaturating 
 
 30 
 
concentrations, both directly limit the overall biodegradation, and the limitation 
effects are multiplicative (Bae and Rittmann 1996). 
 It has commonly been assumed that slow biodegradation is the overall rate-
limiting process for in situ bioremediation. In fact, biostimulation (e.g., addition of 
nutrients) and bioaugmentation, which are often practiced in the field, are based on 
the assumption that slow biokinetics are limiting the overall bioremediation (Bosma 
et al. 1997). In fact, many field tests have been conducted to evaluate the 
biostimulation and bioaugmentation effects on the overall in situ bioremediation and 
successfully demonstrated enhanced bioremediation, e.g., (Eguchi et al. 2001; Major 
et al. 2002; Salanitro et al. 2000). Nevertheless, in reviewing bioremediation data, 
Bosma et al. (1997) note that the intrinsic microbial activity (i.e., the biokinetics) 
actually only limited bioremediation in a few cases; in most cases, the full 
exploitation of the microbial biodegradation potential was prevented by mass transfer 
limitations. 
3.4 Summary and Conclusion 
As reviewed in this chapter, the physical and chemical heterogeneities of the 
subsurface occur at several scales and affect in situ bioremediation by controlling the 
availability of nutrients and substrates that drive the microbiological processes. There 
have been a significant number of studies that have focused on specific processes that 
affect bioremediation in the field and some definitive results have been shown. 
However, as summarized above, most studies have investigated in situ biodegradation 
at a single scale. For example, even in a very extensive series of studies performed at 
Stanford University, e.g., (Kissel et al. 1984; Rittmann and McCarty 1980), the work 
 
 31 
 
only focused on mechanisms at the pore scale and was not aimed at simulating large-
scale groundwater transport. Only a few experimental studies, e.g., (Murphy et al. 
1997; Szecsody et al. 1994; Wood et al. 1994) and several modeling-based studies, 
e.g., (MacQuarrie and Sudicky 1990; Odencrantz 1991; Schafer and Kinzelbach 
1992; Wood et al. 1994; Yang et al. 1994) have explicitly examined the influence on 
in situ biodegradation of several, coupled physicochemical heterogeneity factors 
occurring at different scales. Of course, the field studies are even more severely 
limited due to the complexity and financial constraints. 
Clearly, the interactions between the scale-dependent physical and chemical 
heterogeneities, microbiological processes and in situ biodegradation are still not well 
understood. This indicates there is a ?need for the process engineering approach to 
site remediation? that can be used to develop predictions of the effectiveness of 
different remedial approaches based on the information from each scale of 
observation (Sturman et al. 1995). In order to develop such a quantitative approach, it 
is necessary to study the relevant processes at a scale that represents the heterogeneity 
of natural systems, but on a controlled level that can be described completely. The 
multi-scale, multi-factor mathematical and experimental investigation at an 
intermediate-scale performed in this research provided a useful way to quantify the 
degree of heterogeneity and study the impact of coupled microbial and transport 
processes on in situ biodegradation in representative simulated subsurface systems. 
 
 32 
 
Chapter 4  Theoretical Basis and Framework Development 
The research hypothesis for this work was investigated using an integrated 
modeling and experimental approach. This Chapter?s goal is to present the 
mathematical and theoretical framework used to investigate the research problem. 
First, the mathematic model used and the numerical solution techniques are 
presented. Then the model?s governing equations are used to develop a system of 
dimensionless numbers and an integrated quantitative framework for identifying the 
overall rate-limiting process. 
4.1 Mathematical Model 
Reactive Transport in 3-Dimensions (RT3D), a modular computer code for 
simulating reactive multi-species transport in 3-dimensional groundwater aquifers 
(Clement 1998), was chosen as the mathematical model for evaluating the proposed 
quantitative framework and supporting experiments. Use of RT3D requires the 
groundwater flow code MODFLOW, which was developed by U.S. Geological 
Survey (USGS), for computing spatial and temporal variations in groundwater head 
distribution and flow velocities. For this research, the program used to operate the 
RT3D and MODFLOW codes was Visual MODFLOW (VMOD) (Version 2.8.2) 
(Waterloo Hydrogeologic Inc., Waterloo, Ontario, Canada). RT3D was selected for 
this research for several reasons. First, as a research tool, RT3D has been used for 
modeling several laboratory, pilot-scale and field-scale bioremediation experiments, 
e.g., (Clement et al. 1998; Phanikumar and McGuire 2004; Sun et al. 1999). Second, 
the RT3D code is unique in that it includes an implicit reaction solver that makes the 
 
 33 
 
code sufficiently flexible for simulating various types of chemical and microbial 
reaction kinetics. Third, RT3D includes many of the processes and reactions of 
interest in this research. The mathematical details of the RT3D code are discussed in 
the following paragraphs.  
4.1.1 Governing Equations 
 
The general macroscopic advection-dispersion-reaction (ADR) equations 
describing the fate and transport of aqueous- and solid-phase species, respectively, in 
multi-dimensional saturated porous media are written as, 
 ()
dacsk
s
ki
ij
k
ij
i
k
rrrC
q
Cv
xx
C
D
xt
C
+?++
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
=
?
?
?
,  where k=1, 2?m  
          (4.1) 
and, 
  
dac
im
rrr
dt
Cd
?+=
~
~
, where im = 1, 2 ? (n-m)  (4.2) 
where:  n = total number of species; 
 m = total number of aqueous-phase (mobile) species; 
 im = total number of immobile species, = n-m; 
 
i
x and 
j
x = distance in the direction of 
i
x and 
j
x , respectively, [L]; 
 t = time, [T]; 
 
k
C = the aqueous-phase concentration of the k
th
 species, [ML
-3
]; 
 
im
C
~
= the solid-phase concentration of the im
th
 species, [MM
-1
]; 
 
ij
D = hydrodynamic dispersion coefficient, [L
2
T
-1
]; 
 
 34 
 
 
i
v  = the average pore water velocity, [L T
-1
]; 
s
q = the volumetric flux of water per unit volume of aquifer representing 
sources and sinks, [T
-1
], e.g., recharge; 
 
ks
C = the concentration of species k in the source/sink water, [ML
-3
]; 
c
r = the reaction rate that describes the mass of the species removed or 
produced per unit volume per unit time, [ML
-3
 T
-1
]; 
 
c
r
~
= the reaction rate at the solid phase, [M M
-1
T
-1
]; 
a
r and 
d
r = attachment (or sorption) and detachment (or desorption) rate that 
describe the kinetic exchange of the transported species between aqueous and 
solid phases, [ML
-3
 T
-1
]; 
 ? = the soil porosity. 
The processes on the right hand side of Equation 4.1 represent dispersion, advection, 
sources and sinks of groundwater and reactions in ground water.  
 The saturated groundwater flow velocities,
i
v , are calculated using Darcy?s 
law (Equation 3.2) and the hydraulic-head values that are computed by solving the 
three-dimensional groundwater flow model in VMOD. Thus, the flow equations used 
are (Zheng 1990), 
    
s
i
ii
i
s
q
x
h
K
xt
h
S +
?
?
?
?
?
?
?
?
?
?
?
?
=
?
?
   (4.3) 
    
i
ii
i
x
hK
v
?
?
?=
?
     (4.) 
where: h  = the hydraulic head, [L]; 
 
s
S = the specific storage coefficient, [L
-1
]; 
 
 35 
 
 
ii
K = the principal components of the hydraulic conductivity tensor, [L T
-1
]. 
 In the following subsections, the reactions rate terms pertinent to this research 
are discussed, followed by a description of the numerical solution technique used by 
RT3D. 
4.1.2 Rate-Limited Sorption Reactions 
 Sorption mechanisms are usually described in terms of mass-transfer kinetics 
and the type of equilibrium isotherm, as reviewed in Chapter 3. When sorption is 
assumed to be rate-limited, it is necessary to track the contaminant concentrations in 
both mobile (groundwater) and immobile (soil) phases.  
 In this research, the fate and transport of a sorbing solute in the aqueous and 
soil phases can be predicted using the following transport equations modified from 
Haggerty and Gorelick?s (1994) approach. First, Equation 4.1 is written with only the 
sorption reaction rate term, 
t
C
r
a
?
?
=
~
?
?
: 
   ()
t
C
C
q
Cv
xx
C
D
xt
C
s
s
i
ij
ij
i
?
?
?+
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
=
?
?
~
?
?
?
 (4.5) 
where the subscripts were dropped for Equation 4.1 because only one solute was of 
interest to this research, i.e., k = 1 and im = 1. Then, the first-order, linear driving 
force model (Equation 3.7) is substituted for
a
r , except that the possibility of 
nonlinear equilibrium is incorporated using the Freundlich model (Weber and Miller 
1988): 
    
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?=
?
?
n
d
ls
K
C
CK
t
C
1
~~
?
?
   (4.6) 
 
 36 
 
when the empirical constant, n =1, the linear isotherm equation (Equation 3.4) is 
integrated into the equation, and when 0<n<1 or n>1, the Freundlich isotherm 
(Equation 3.6) is incorporated into the equation. 
Although Equation 4.5 and 4.6 are adopted for the proposed study, more 
complicated models, which incorporate internal (Szecsody and Bales 1989) or 
external (Weber and Smith 1987) mass transfer resistances, may be required to 
describe some systems, e.g., Equation 3.8 and 3.10, respectively. This was 
encountered in this work, as discussed in Chapter 6. 
4.1.3 Double Monod Biodegradation Kinetic Model 
 Incorporating the Double Monod kinetic model (Equation 3.12) into Equation 
4.1, along with the sorption reaction rate term (Equation 4.6, with n=1), gives the 
following governing equation for describing the fate and transport of an electron 
donor (a hydrocarbon, for example) in a multi-dimensional saturated porous media: 
() []
?
?
?
?
?
?
?
?
+
?
?
?
?
?
?
?
?
+
?
?
?
?
?
?
?
?
+?
?
?
?
?
?
?
?
?
??+
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
=
?
?
AK
A
SK
SM
Mq
K
S
SKS
q
Sv
xx
S
D
xt
S
AS
m
d
lss
s
i
ij
ij
i
?
?
?
~~
 
          (4.7) 
where:  S = the electron donor concentration in the aqueous phase, replacing the 
generic term C, used earlier [ML
-1
]; 
 S
s
 = the electron donor concentration in the sources/sinks, [ML
-3
]; 
 M = the aqueous phase bacterial cell concentration, [ML
-3
]; 
M
~
= the solid-phase cell concentration (mass of bacterial cells per unit mass 
of porous media, [MM
-1
]; 
 
 37 
 
A = the electron acceptor concentration in the aqueous phase, replacing the 
generic term C, used earlier [ML
-3
]; 
 
S
K = the half-saturation coefficient for the electron donor, [ML
-3
]; 
 
A
K = the half-saturation coefficient for the electron acceptor, [ML
-3
]; 
 
m
q = the specific maximum electron donor utilization rate, [MM
-1
T
-1
]; 
 The assumptions implicit in the double Monod model as implemented in 
Equation 4.7 are as follows: 
1. The degradation reactions occur only in the aqueous phase, which is usually a 
conservative assumption; 
2. No specific microscopic biomass structure is assumed, and diffusion 
limitations across biofilm are neglected, i.e., it is the strictly macroscopic 
model discussed in Chapter 3; 
3. First-order kinetic expressions are assumed to represent the exchange of 
bacteria between aqueous and solid phases (Hornberger et al. 1992; Peyton et 
al. 1995; Taylor and Jaffe 1990), as discussed further below. 
4. Permeability and porosity changes caused by the bacterial growth are ignored 
in the formulation (Rittmann 1993; Taylor and Jaffe 1990). 
 Similarly, the fate and transport of the electron acceptor (oxygen, for 
example) can be modeled using the following equation: 
()
?
?
?
?
?
?
?
?
+
?
?
?
?
?
?
?
?
+
?
?
?
?
?
?
?
?
+?+
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
=
?
?
AK
A
SK
SM
MqYA
q
Av
xx
A
D
xt
A
AS
mSAS
s
i
ij
ij
i
?
?
?
~
/
(4.8) 
 
 38 
 
where 
s
A = the electron acceptor concentration in the sources/sinks, [ML
-3
];; and 
SA
Y
/
 
is the stoichiometric coefficient, relating electron acceptor consumption to utilization 
of the electron donor, [MM
-1
].  
 In addition, the fate and transport of bacteria in the aqueous phase can be 
described using the equation: 
()
MK
AK
A
SK
SM
MqY
MK
MK
M
q
Mv
xx
M
D
xt
M
e
AS
mSXatt
s
s
i
ij
ij
i
?
?
?
?
?
?
?
?
?
+
?
?
?
?
?
?
?
?
+
?
?
?
?
?
?
?
?
+++?
+
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
=
?
?
?
?
?
?
?
~~
/
det
 (4.9) 
where: 
s
M = the bacterial concentration in the sources/sinks [ML
-1
]; 
att
K  = the 
bacterial attachment coefficient, [T
-1
]; 
det
K  = the bacterial detachment coefficient, [T
-
1
]; 
e
K = the endogenous cell death or decay coefficient, [T
-1
]; and 
SX
Y
/
 = the true 
yield coefficient, [MM
-1
]. 
 The growth of attached-phase bacteria can be described using an ordinary 
differential equation: 
 MK
AK
A
SK
S
MqYMK
MK
dt
Md
e
AS
mSX
att
~~~
~
/det
?
?
?
?
?
?
?
?
?
+
?
?
?
?
?
?
?
?
+
+?=
?
?
 (4.10) 
4.1.4 Numerical Solution Techniques 
 The RT3D code utilizes a standard reaction Operator-Splitting (OS) strategy 
to develop a general numerical solution scheme for solving the coupled 
partial/ordinary differential equations described above. The OS strategy is 
demonstrated below using Equation 4.1 as an example. 
 
 39 
 
 Employing the OS strategy, first the mobile species transport equation, 
Equation 4.1, is divided into four distinct differential equations:  
the advection equation, 
( )
i
i
x
Cv
t
C
?
?
?=
?
?
    (4.1) 
the dispersion equation, 
?
?
?
?
?
?
?
?
?
?
?
?
=
?
?
j
ij
i
x
C
D
xt
C
    (4.12) 
the source/sink-mixing equation, 
s
s
C
q
t
C
?
=
?
?
     (4.13) 
and, the reaction equation, 
r
dt
dC
=      (4.14) 
where: the term r represents all possible reaction terms that appear in a typical 
mobile-species transport equation. In a typical immobile species equation (Equation 
4.2), the advection, dispersion, and source/sink-mixing terms are zero and only the 
reaction term exists. 
 Each of the differential equations described above is solved numerically by 
RT3D using various solution solvers. The advection package can use either a method 
of characteristics (MOC), modified method of characteristics (MMOC), hybrid 
method of characteristics (HMOC), or an upstream finite-difference solution schemes 
(Zheng 1990). In this research, the MOC method was applied because it is virtually 
free of numerical dispersion. The dispersion and source-sink mixing packages use 
 
 40 
 
explicit finite-difference approximations, while the reaction package has an improved 
implicit reaction solver. The logical steps involved in the numerical solution 
procedure are illustrated in Figure 4.1. As shown in the figure, the solution algorithm 
initially solves the advection, dispersion, and source-sink mixing steps for all mobile 
components. After solving the transport, the coupled reaction equations are solved 
implicitly. Use of the modular OS approach for solving the reactive transport problem 
facilitates representation of different contaminant transport systems through a set of 
pre-programmed reaction packages. 
 It should be noted that RT3D as included in Visual MODFLOW has the rate-
limited sorption package (Equations 4.5 and 4.6) and Double Monod biokinetics 
(Equations 4.7-4.10) built in. However, it does not allow their simultaneous use. 
Therefore, in a related project, Mark Johnson (M.S. student), with assistance from 
Waterloo Hydrogeologic program, created a new module to add that capability to 
RT3D. As a result, a new reaction transport engine with the title of ?University of 
Maryland-biokinetic model? was added to the Transport Engine Options in RT3D, 
which incorporates the lumped first-order mass-transfer sorption kinetics and Double-
Monod biodegradation kinetics into one package, allowing their simultaneous 
utilization.  
 
 
 41 
 
 
 
Figure 4.1 Block diagram illustrating the numerical solution scheme. 
Solve advection for ?m? mobile 
Solve dispersion for ?m? mobile 
Source/sink-mixing for ?m? mobile 
t
No advection 
No dispersion 
No source/sink 
Mobile Components, m species Immobile Components, im species 
Solve coupled reactions for all ?n? components 
 
 42 
 
4.2 Framework Development 
 In previous research, the various scales of heterogeneities and associated 
interfacial processes have been studied extensively individually. However, in real 
natural systems, several of these processes occur simultaneously at different scales of 
heterogeneities. Nonetheless, there is one process that will limit the overall 
biodegradation rate given a certain combination of environmental conditions. If that 
limitation is significant, then the bioremediation is usually not successful (Sturman et 
al. 1995). In fact, Sturman et al. (1995) concluded that an assessment of the feasibility 
of an in situ bioremediation project is dominated by the need to identify and estimate 
the appropriate rate-controlling phenomenon.  
 Therefore, in order to have successful in situ bioremediation in the field, a 
systematic method is needed to identify the rate-limiting process and then the 
appropriate engineering action can be taken for enhancing the biodegradation rate. An 
integrated quantitative framework that can be used to identify the overall rate-limiting 
process was developed in this research (Figure 4.2), which is based on a systematic 
comparison of a set of dimensionless numbers. The definition of dimensionless 
numbers is presented in Table 4.1. The developed framework is an expansion of the 
one presented in Weiner et al. (1999), which was adapted from Ramasmami and 
Luthy (1997). 
 
 43 
 
 
 
 
Figure 4.2 Quantitative framework for predicting the rate-limiting phenomenon. 
 
Pe
Sh
2
?St
2
Da
2 Da
5
Da
6 
Advection  
control
 
Biokinetics 
Control
 
Sorption  
control
Dispersion  
control
 
>>1 >>1 
>>1 
>>1 >>1 >>1 
<<1 <<1 
<<1 <<1 
<<1 
<<1 
 
 44 
 
Table 4.1 Definition of dimensionless numbers  
Symbol  Meaning and Equations Indication of rate-limiting process 
Pe
l
 (Longitudinal 
Peclet No.) 
=(advection rate/dispersion rate) 
=
x
x
wx
x
D
Lv
SD
Lq
=
?
 
If Pe
l
 >>1: longitudinal dispersion 
limits 
Pe
t
 (Transverse Peclet 
No.) 
=(advection rate/dispersion rate) 
=
z
x
wz
x
D
Lv
SD
Lq
=
?
 
If Pe
t
 >>1: transverse dispersion 
limits 
St
1
 (Stanton No. 1) =(NAPL dissolution/advection rate) 
=
xx
na
v
LK
q
Lak
lnln
=  
If St
1
>>1: advection limits 
St
2
 (Stanton No. 2) =(soil mass transfer rate/advection rate) 
=
x
ls
x
sals
v
LK
q
Lak
=  
If St
2
>>1: advection limits 
Sh
1
? (Modified 
Sherwood No. 1) 
=(NAPL dissolution/dispersion rate) 
=
zz
D
LK
or
D
Lk
2
lnln
 
If Sh
1
?>>1: transverse dispersion 
limits 
Sh
2
? (Modified 
Sherwood No. 2) 
=(soil mass transfer rate/dispersion rate) 
=
z
ls
z
ls
D
LK
or
D
Lk
2
 
If Sh
2
?>>1: transverse dispersion 
limits 
Da
2
 (Damk?hler No. 
2) 
= (biodegradation rate/advection rate) 
=
x
vS
LMq
0
0max
 
If Da
2
 >>1: advection limits 
 
Da
6
 (Damk?hler No. 
6) 
= (biodegradation rate/dispersion rate) 
=
z
DS
LMq
0
2
0max
 
If Da
6
 >>1: transverse dispersion 
limits 
Da
3
 (Damk?hler No. 
3) 
= (biodegradation rate/NAPL 
dissolution rate) 
=
ln0
0max
1
2
KS
Mq
St
Da
=  
If Da
3
 >>1: dissolution limits 
Da
5
 (Damk?hler No. 
5) 
= (biodegradation rate/soil mass transfer 
rate) 
=
ls
KS
Mq
St
Da
0
0max
2
2
=  
If Da
5
 >>1: desorption limits 
 where: L = characteristic length, in all cases, L=aquifer thickness (Mukherji et al. 1997; Oya and 
Valocchi 1998); 
x
q = specific discharge [LT
-1
]; =
x
v average pore water velocity [LT
-1
]; =
0
M initial biomass 
concentration [ML
-3
]; =
0
S initial substrate concentration [ML
-3
]; =
ln
k the mass-transfer-rate coefficient 
between liquid and NAPL [LT
-1
]; 
na
a = =
V
A
na
specific NAPL/aqueous phase interfacial area [L
-1
]; 
==
w
na
nS
ka
K
ln
ln
the lumped mass-transfer-rate coefficient between liquid and NAPL [T
-1
]; 
 
 45 
 
 
 The dimensionless numbers in Figure 4.2 and Table 4.1 can be developed 
from the governing equations describing transport of contaminants presented in 
Section 4.1. For example, the advection and dispersion equation describing transport 
of an electron donor solute influenced by nonlinear kinetic sorption, nonlinear 
double-Monod biodegradation kinetics, and biomass growth and decay, for two-
dimensional steady-state flow can be written as, 
()
?
?
?
?
?
?
?
?
+
?
?
?
?
?
?
?
?
+
?
?
?
?
?
?
?
?
+?
?
?
?
?
?
?
?
?
??
?
?
?
?
?
?
?
?
?
?
+
?
?
?
?
?
?
?
?
?
?
+
?
?
?=
?
?
AK
A
SK
SM
Mq
K
S
SK
z
S
D
zx
S
D
x
vS
xt
S
AS
m
d
lszx
?
?
~~
           
          (4.15) 
Equation 4.15 can be converted into dimensionless form by using the dimensionless 
numbers defined in Table 4.1 and the following dimensionless parameters: 
L
x
x =* ,  
L
z
z =* ,  
x
v
L
t
t =*, 
0
*
S
S
S = ,  
0
~
*
~
S
S
S =  
0
*
A
A
A = , 
0
*
S
K
K
S
S
= ,  
0
*
A
A
K
A
= ,  
0
~
~
*
~
M
M
M
?
?
=  
Using this series of dimensionless parameters, results in the following equation: 
()
?
?
?
?
?
?
?
?
+
?
?
?
?
?
?
?
?
+
???
?
?
+
?
?
+
?
?
?=
?
?
**
*
**
*
*
~
*
*
*1
*
*1
*
*
*
*
22
2
2
2
2
AK
A
SK
S
DaSSSt
z
S
Pex
S
Pex
S
t
S
AStl
 (4.16) 
 Although not shown here, a similar dimensionless equation can be derived for 
the electron acceptor, as well as the biomass in aqueous and solid phases using 
Equations 4.8, 4.9 and 4.10. The advantage of using the dimensionless form of the 
equations is that the number of independent parameters in the system is minimized. 
 
 46 
 
Further, the resulting dimensionless parameters succinctly capture the complexity of 
the system in a way that can be used in the quantitative framework presented in 
Figure 4.2.  
 The quantitative framework was used to integrate and compare the key 
interfacial processes that occur in contaminated natural systems and predict the rate-
limiting process under certain environmental conditions. The first two steps were 
used to identify the slowest mass-transfer process among advection, dispersion, 
sorption/desorption and the third step was used to compare the slowest mass transfer 
process with the biokinetics to determine the overall rate-limiting process. The 
framework is viable for offering insight into the rate-limiting process for the overall 
bioremediation application as long as the dimensionless numbers are significantly 
smaller or larger than unity (e.g., <0.2 or >5) (Ramaswami and Luthy, 1997). If the 
rate-limiting process can be properly identified, then an appropriate remedial 
approach can be chosen to alleviate that limitation. For example, in the study of 
Falatko and Novak (1992), a biosurfactant was added to the influent to enhance 
desorption, which was the rate-limiting step for the overall biodegradation rate.  
4.3 Research Approach 
The integrated modeling and experimental approach for evaluating the 
quantitative framework and the research hypothesis was comprised of four 
components (as discussed in subsequent chapters, some of these components were 
performed in conjunction with other graduate students). First, the quantitative 
framework based on a set of dimensionless numbers described above was developed 
to evaluate the effects of different scales of heterogeneities, associated mass transfer 
 
 47 
 
process, and biokinetics on in situ bioremediation. Second, the RT3D was chosen and 
utilized to perform a preliminary evaluation of the relationship between the 
heterogeneities, interfacial processes, and biokinetics to test the framework. Third, an 
intermediate-scale flow cell (ISFC) was designed and constructed, representing a 
realistic, yet experimentally tractable system to investigate the rate-limiting process 
that affects the successful application of in situ bioremediation. Finally, selected 
bioremediation scenarios were simulated in the constructed ISFC. These scenarios 
were designed to provide a useful evaluation of the quantitative framework. 
Specifically, in each of the simulated scenarios, the experimental conditions were 
varied such that a different single process limited the overall bioremediation rate in 
each case. The results of these experiments were used to evaluate the utility of the 
quantitative framework for defining the rate-limiting process and selecting an 
appropriate engineered bioremediation approach.  
 Specifically, three sets of experiments simulating three different scenarios 
were performed, with each scenario being representative of one common rate-limiting 
phenomenon occurring in natural subsurface systems. Each scenario had a 
combination of processes occurring at the macro-scale, meso-scale, and micro-scale. 
All these experiments were performed in a two-dimensional domain with horizontal 
flow and vertical (hydraulic conductivity) stratification. Different hydraulic 
conductivity (K) strata (i.e., a high-K layer and a low-K layer) were used because 
they are a relatively simple type of macro-scale heterogeneity of practical importance 
and because of the importance of vertical transverse dispersion as discussed in 
Chapter 3. Meso-scale interfacial processes were focused on sorption/desorption, 
 
 48 
 
while micro-scale processes were focused on biodegradation. In all three cases, the 
simulated groundwater contained the model contaminant naphthalene and 
conservative tracer Br, along with oxygen, to be representative of field conditions. 
The experimental conditions were varied in such a way that only one single mass-
transfer process or biokinetics limited the overall bioremediation rate in each case. 
The three scenarios are illustrated schematically in Table 4.2 and the corresponding 
dimensionless numbers are presented in Table 4.3.  
Scenario #1 was performed with relatively fast aerobic biokinetics, in a porous 
media with relatively low sorption, resulting in macro-scale vertical transverse 
dispersion from the fast-K to the slow-K layer being the rate-limiting process (Figure 
4.3). It was expected that the bioremediation would be limited by transverse 
dispersion from the fast-K layer into the slow-K layer and that the biodegradation 
would occur primarily at the interface between the layers where substrate mixing 
occurs. 
Scenario #2 was designed to simulate a situation where slow biokinetics 
limited the overall biodegradation rate (Figure 4.4). It utilized the same porous media 
and sorptive characteristics as Scenario #1. The experiment was run with the same 
microbial culture, but with inhibited slower biokinetics, which was accomplished by 
adding high concentration of salt (5% of NaCl solution by w/v) as an inhibitor to 
reduce the water potential of the system so as to slow down the biodegradation rate 
(Brock 1977). 
 
 
 
 49 
 
Table 4.2 Description of the flow cell experiments 
Scenarios Schematic experimental settings Description 
Scenario #1: macro-
scale: transverse 
dispersion controls 
 
 
 
- Relatively fast biokinetics 
- Limited sorption 
- Relatively slow transverse 
dispersion 
Scenario #2: micro-
scale: biokinetics 
controls 
 
 
 
 
- Slow biokinetics 
- Limited sorption 
- Relatively slow transverse 
dispersion 
Scenario #3: meso-
scale: 
sorption/desorption 
controls 
 
 
 
- Relatively fast biokinetics 
- Increased sorption 
- Relatively slow transverse 
dispersion 
 
 
Table 4.3 Application of the dimensionless numbers in the quantitative framework to 
simulated bioremediation scenarios 
Scenario description First two-step result Final step result 
Scenario #1: transverse dispersion is 
the overall rate-limiting process 
t
Pe >> 1 
'
2
Sh >>1 
6
Da >>1 
Scenario #2: sorption/desorption is the 
overall rate-limiting process 
t
Pe >> 1 
'
2
Sh <<1 
5
Da >>1 
Scenario #3: biokinetics is the overall 
rate-limiting process 
t
Pe >> 1 
'
2
Sh >>1 
6
Da <<1 
 
Fast K 
Slow K 
Fast K 
Slow K 
Fast K 
Slow K 
 
 50 
 
 
Pe
Sh
2
?St
2
Da
2
Da
5
Da
6
Advection 
control
Biokinetics Control
Sorption 
control
Dispersion 
control
>>1>>1
>>1
>>1 >>1>>1
<<1 <<1
<<1 <<1
<<1
<<1
 
Figure 4.3 Quantitative framework for predicting the rate-limiting phenomenon 
Application of the flow chart to Scenario #1 is highlighted by encircling the 
experimental conditions. 
 
 
 51 
 
Pe
Sh
2
?St
2
Da
2
Da
5
Da
6
Advection 
control
Biokinetics Control
Sorption 
control
Dispersion 
control
>>1>>1
>>1
>>1 >>1>>1
<<1 <<1
<<1 <<1
<<1
<<1
 
 
Figure 4.4 Quantitative framework for predicting the rate-limiting phenomenon 
Application of the flow chart to Scenario #2 is highlighted by encircling the 
experimental conditions. 
 
 52 
 
Finally, in Scenario #3, the influent solution, the two-layered porous media, 
and the biokinetics were the same as in Scenario #1, but the sorptive characteristics of 
the porous media were increased so that the meso-scale desorption rate was the 
overall rate-controlling mass-transfer process (Figure 4.5). Thus, it was expected that 
the overall bioremediation rate would be limited by the naphthalene 
sorption/desorption rate. Due to time constraints, this scenario was not implemented 
completely. The sorption/desorption equilibrium and kinetic studies were performed 
in batch experiments and parameter estimates were obtained from model fittings. 
However, the evaluation of this scenario in ISFC is still under investigation and will 
be conducted by Ms. Eunyoung Hong.
 
 53 
 
Pe
Sh
2
?St
2
Da
2
Da
5
Da
6
Advection 
control
Biokinetics Control
Sorption 
control
Dispersion 
control
>>1>>1
>>1
>>1 >>1>>1
<<1 <<1
<<1 <<1
<<1
<<1
 
 
Figure 4.5 Quantitative framework for predicting the rate-limiting phenomenon 
Application of the flow chart to Scenario #3 is highlighted by encircling the 
experimental conditions. 
 
 
 
 54 
 
 
Chapter 5 Experimental Materials and Methods 
 To experimentally investigate the research hypothesis and evaluate the 
quantitative framework using the three experimental scenarios presented in Table 4.2, 
appropriate experimental materials and methods were developed and applied, as 
described in this chapter. First, a model contaminant was selected, an artificial 
groundwater was designed to be representative of natural groundwater chemistry, and 
a microorganism was chosen as inoculum for biodegradation. Second, an 
intermediate-scale flow cell (ISFC) was designed and constructed to simulate a 
contaminated aquifer with various scales of heterogeneities. Third, batch experiments 
were conducted to estimate the kinetic parameters for the physical, chemical and 
biological processes occurring in the simulated subsurface of the ISFC. Fourth, the 
bioremediation simulation experiments were carried out in the ISFC to investigate the 
research hypothesis. Finally, analytical methods were chosen to analyze the model 
contaminant concentration, tracer concentration, biomass concentration and dissolved 
oxygen (DO) concentration, which are summarized below.  
5.1 Model Contaminant, Artificial Groundwater and Microorganism 
5.1.1 Model contaminant 
Naphthalene (Fisher Scientific, 99%) was selected as the test contaminant, 
because PAHs are a  class of contaminants that has critical bioavailability issues at 
numerous contaminated sites (Rittmann et al. 1994), and naphthalene has often been 
used as a model experimental compound for PAHs, e.g., (Park et al. 2002). Although 
 
 55 
 
like other PAHs naphthalene has generally poor mobility in the environment (i.e., low 
solubility, high sorption, low volatility), it has a sufficiently high aqueous solubility 
to facilitate experimentation. The properties of naphthalene are summarized in Table 
5.1. 
 
Table 5.1 Physicochemical properties of naphthalene 
(Sims and Overcash 1983) 
Chemical Naphthalene (C
10
H
8
) 
Molecular weight (g/mol) 128.2 
Vapor pressure (Pa at 25?C) 10.4 
Log (K
ow
) 3.37 
Solubility at 30?C (mg/L) 31.7 
Molecule surface area (A
2
) 155 
 
Once naphthalene was selected, an experimental concentration needed to be 
chosen. There were two factors that affected the selection of the naphthalene 
concentration: (1) the concentration should be representative of the typical 
concentration found at sites contaminated with naphthalene, and (2) the concentration 
should be high enough so the experimental trends could be followed during the 
course of an experiment. With respect to the first constraint, in coal tar/water system 
studies (Peters and Luthy 1993), an equilibrium concentration for naphthalene was 
measured to be 3.3 mg/L, and Kramer and Hayes (1987) measured an equilibrium 
aqueous concentration of 2.540 mg/L naphthalene in a No. 2 heating oil/water 
 
 56 
 
system. Based on these examples, the naphthalene concentration in the research was 
originally chosen to be 5 mg/L as utilized in the control experiments. The 
concentration selected was higher than the above examples, because it was easier to 
study experimental trends compared to if the concentration was lower. Later, in the 
bioremediation simulation experiments, the naphthalene concentration was adjusted 
to 10 mg/L due to the fast biodegradation rate in the simulated aquifer, which 
necessitated that the naphthalene concentration be adjusted to make it more feasible 
to monitor the plume development. 
Aqueous naphthalene solutions were prepared either from methanol stock 
solutions (sorption experiments) or N, N-Dimethylformamide (DMF, Sigma-Aldrich, 
99.8%, A.C.S. reagent) (Whitman et al. 1998) stock solutions (batch respirometry 
experiments and bioremediation simulation experiments). The naphthalene methanol 
stock solutions were initially prepared by dissolving solid naphthalene in methanol. 
These methanolic solutions were then used to prepare the naphthalene aqueous 
solutions. In all cases, it was ensured that the resulting concentration of methanol in 
the aqueous solutions was lower than 2% by volume.  It had previously shown that 
the presence of methanol up to 5% by volume did not modify the sorption properties 
of naphthalene (Appert-Collin et al. 1999). Naphthalene DMF stock solutions were 
prepared by using a similar technique, and following the same constraints. 
5.1.2 Artificial Groundwater 
A dilute mineral salt nutrient solution (MSNS) was used in the parameter 
estimation batch experiments and ISFC bioremediation simulation experiments to 
represent dilute groundwater chemistry. The MSNS was prepared in de-ionized (DI) 
 
 57 
 
water as described by Murphy et al. (1997), with the mineral compositions as 
presented in Table 5.2. 
The MSNS was prepared from three stock solutions. The macro nutrient stock 
solution was made by dissolving 10 mg of FeSO
4
?7H
2
O, 200 mg of MgSO
4
?7H
2
O, 
300 mg NH
4
Cl, and 60 mg NaH
2
PO
4
?H
2
O into 1 L deionized water (= final 
concentration ?100). The trace element stock solution was made by first dissolving 
50 mg each of MnCl
2
, Na
2
SeO
3
, H
3
BO
3
, Na
2
MoO
4
?2H
2
O, CoCl
2
?6H
2
O, NiSO
4
?6H
2
O, 
CaSO
4
?5H
2
O, and ZnSO
4
?7H
2
O into 1 L deionized water, from which 10 ml was 
taken out and diluted into 1 L with distilled water to obtain the trace stock solution (= 
final concentration ?100). The PIPES (Sigma Chemical Co., 99%) stock buffer 
solution was prepared with 151.2 g PIPES dissolved into 2L deionized water, which 
was adjusted to pH = 6.8 with 4 N NaOH and then diluted to 2 L with deionized 
water (= final concentration ? 25). To make MSNS, appropriate volumes of the 
macro nutrient stock solution, trace element stock solution and PIPES buffer solution 
were combined and diluted to the desired volume with deionized water. Different 
concentrations of naphthalene solution were then prepared in the MSNS by adding 
naphthalene in methanol or DMF stock solutions as needed. On at least one occasion, 
the pH of the effluent from the bioremediation experiments described below was 
measured and found to be 6.8, indicating that the buffering capacity of the MSNS was 
adequate. 
 
 
 58 
 
Table 5.2 Mineral medium compositions  
Macro nutrient compounds Concentration, mg/L 
FeSO
4
?7H
2
O 
MgSO
4
?7H
2
O 
NH
4
Cl 
Na
2
HPO
4
 
0.1 
2 
3 
0.6 
Trace element compounds Concentration, ?g/L 
MnCl
2
 
Na
2
SeO
3
 
H
3
BO
3
 
Na
2
MoO
4
?2H
2
O 
CoCl
2
?6H
2
O 
NiSO
4
?6H
2
O 
CaSO
4
?5H
2
O 
ZnSO
4
?7H
2
O 
5 
5 
5 
5 
5 
5 
5 
5 
Buffer Solution Concentration, mM 
PIPES 10 
 
 
 59 
 
In the MSNS, NH
4
Cl and Na
2
HPO
4 
served as the N and P source, 
respectively. The overall stoichiometry for aerobic naphthalene degradation, 
including biomass synthesis, with ammonia as the nitrogen source, is as follows 
(McCarty 1987): 
OHCONOHCNHOHC
2227532810
6.142.12.16 ++?++   (5.1) 
This equation was used as the stoichiometric basis for the nutrient requirement 
calculation. Based on stoichiometry of biological reaction of naphthalene and 
assuming the phosphorus required is approximately one-fifth of that for nitrogen, the 
mass ratio of naphthalene C
10
H
8
: NH
4
Cl: Na
2
HPO
4 
for the biodegradation under 
aerobic conditions is 10:5:1. Therefore, for an initial naphthalene concentration of 10 
mg/L, the stoichiometric amount of NH
4
Cl and Na
2
HPO
4
 required to degrade the 
naphthalene are 5 mg/L and 1 mg/L, respectively. Correspondingly, the amounts of 
NH
4
Cl and Na
2
HPO
4 
in the MSNS, 3 mg/L and 0.6 mg/L, respectively, meet 60% of 
the stoichiometric nutrient requirements for biodegradation of 10 mg/L of 
naphthalene.  
5.1.3 Microorganism 
The microbial culture used in this work was selected as part of the larger 
project, of which this research was a part, by Eunyoung Hong, a Ph.D. student in the 
Department of Civil and Environmental Engineering, University of Maryland, 
College Park. Based on the selection process, Pseudomonas fluorescens Uper-1 was 
chosen as the organism for these experiments, and was obtained from Dr. D.R. 
Lueking (Department of Biological Sciences, Michigan Technological University, 
Houghton, MI, USA). This organism is able to aerobically utilize naphthalene as a 
 
 60 
 
sole source of carbon for growth and displays no other growth factor requirement 
(Whitman et al. 1998). A stock culture of Uper-1 was maintained by periodic 
subculture on nutrient agar in the presence of naphthalene vapor and was stored at 4 
?C in a refrigerator. Batch growth of Uper-1 was routinely conducted in 50-mL side-
arm flasks at 30 ?C with shaking (240 rpm) in a glycerol-based basal salt medium 
(BSM). The BSM contained (per liter at pH 7.0): 4 g KH
2
PO
4
, 4 g NaH
2
PO
4
, 2 g 
(NH
4
)
2
SO
4
, 0.2 g MgSO
4
?7H
2
O, 0.001 g CaCl
2
?2H
2
O, 0.001 g FeSO
4
?7H
2
O, and 1.5% 
w/v glycerol. Batch culture growth in the side-arm flasks was monitored 
turbidimetrically (wavelength = 510 nm) utilizing a Spectronic 21 spectrophotometer 
(Bausch & Lomb).  
The Uper-1 inoculum utilized for the preliminary batch respirometry 
experimental studies was prepared in 50 ml side-armed flasks using the following 
procedures. The inoculum was first grown on glycerol (ACROS, 99+ %) in MSNS as 
a primary culture, followed by growth on naphthalene in MSNS as a secondary 
culture. Five flasks were used for each batch preparation. Fifty ml of MSNS was 
added to each side-arm flask, two of which also contained glycerol (1.5% by volume). 
The flasks were closed with cotton balls, wrapped with foil paper, and autoclaved 
(121?C for 15 minutes).Using aseptic techniques, a Uper-1 plate culture was used to 
inoculate the duplicate flasks with glycerol. A third flask with MSNS only was used 
as a control. The three flasks were shaken in a water bath at the temperature of 30?C 
for about 24 hours. The absorbance was monitored to determine the phase of growth. 
Once the culture arrived in the late exponential phase, 0.5 ml (1%) of the primary 
culture was transferred into the remaining two flasks, into which filter-sterilized 
 
 61 
 
naphthalene in DMF had been introduced to reach the desired concentration of 5 
mg/L. These two flasks were then shaken in the same water bath for about 6 hours. 
Again the absorbance was monitored to assess when the culture had reached the 
middle of the exponential phase, at which point the culture was used as the inoculum 
in the preliminary respirometry studies described in Section 5.3.4. 
The Uper-1 culture used for inoculating the ISFC was initially prepared in five 
250 ml flasks containing autoclaved BSM, by following the same procedures as 
described for the batch respirometry experiments. Once the initial culture reached the 
late exponential phase, the prepared cells were harvested by centrifugation (12 000 x 
g for 10 min), and freed of residual glycerol by washing three times in BSM minus 
glycerol. The washed culture was then resuspended in 500 ml BSM lacking glycerol, 
and used to inoculate 30 L of the autoclaved MSNS artificial groundwater containing 
5 mg/L naphthalene. The second culture was then incubated for 24 h at a room 
temperature, after which the heterotrophic plate count (HPC) was 8.5 x 10
11 
CFU/L. 
The secondary culture was used to ensure that the microorganisms had fully adapted 
to growth on naphthalene medium before inoculating the ISFC. 
5.2 Intermediate-Scale Flow Cell (ISFC) 
5.2.1 Sizing of ISFC 
In this research it was very important to size the flow cell reactor in such 
dimensions that it could satisfactorily represent the heterogeneity in natural systems, 
but in a controlled way that allowed a complete description. In reviewing previous 
studies in the literature, there are a variety of sizes of intermediate-scale flow cell 
reactors that have been used for various purposes, with flow cell lengths ranging from 
 
 62 
 
less than 1m to up to 10m.  These flow cells are summarized in Table 5.3 in terms of 
their size and experimental use. 
For this research, the flow cell dimensions were selected such that it is 
essentially a two dimensional reactor, that is, the contaminant plume is only varying 
in the longitudinal and vertical directions. This was done because the two-
dimensionality reduced the complexity of the monitoring problem in the 
bioremediation simulation experiments as well as the mathematical modeling of these 
experiments. Therefore, the flow cell needed to be sufficiently narrow to ensure 
homogeneity across its width, yet wide enough to cause negligible wall effects for the 
contaminant migration. In the study of Klotz et al. (1980), they found that in order to 
be able to neglect boundary disturbances, the relationship between the average grain 
diameter, d
50
 and the column diameter, d
c
, must meet the following relationship: d
c
 ? 
25 d
50
. The maximum sand size d
50
 (for the coarse sand) in our experiment is 1.2 mm. 
Therefore, the flow cell diameter needs be larger than 3 cm. On the other hand, the 
flow cell should be narrow enough to eliminate significant effects of dispersion in the 
horizontal transverse direction. In order to study the dispersion effect in the horizontal 
transverse direction, model simulations were made using RT3D in a three 
dimensional scenario and it was verified that the transverse dispersion is negligible 
with a width of 10 cm. Based on the d
c
 calculation, the model simulations, and the 
values from the literature, a flow cell width of 10 cm was selected. This selection was 
somewhat conservative, but reasonable considering that, some operations, such as 
sampling, would be more challenging with a narrower width. 
 
 
 63 
 
Table 5.3 Summary from literature study of flow cells and their applications 
Authors 
Flow cell size 
 (length ?height ?width) 
Purpose of study 
Bath et al. (2001) 10 m ?1.2 m ?0.06 m 
To investigate the flow and transport in a 
two-dimensional intermediate-scale, 
heterogeneous medium 
Silliman and 
Dunlap (2001) 
1.6 m ?0.67 m ?0.094 m 
To study a bacterial transport in porous 
media 
Murphy and Ginn 
(1997) 
1 m ?0.2 m ?0.1 m 
To investigate the influence of physical 
heterogeneity on microbial degradation and 
distribution in porous media 
Pearce et al. 
(1994) 
0.75 m?0.37 m?0.21m 
To evaluate the dissolution of 
tricholorethylene (TCE) and 
trichloroethane (TCA ) pools in saturated 
subsurface systems 
Szecsody et al. 
(1994) 
1 m ?0.2 m ? 0.1 m 
To study the transport and biodegradation 
of quinoline in horizontally stratified 
porous media 
Voudrias and Yeh 
(1993) 
1 m ?0.3 m ?0.2 m 
To evaluate dissolution of a toluene pool 
under constant and variable hydraulic 
gradients with implications for aquifer 
remediation 
  
In addition to the appropriate selection of the flow cell width, the flow cell 
longitudinal and vertical dimensions were also very important factors. As for the 
longitudinal dimension, it was important to select a length that would allow 
subsequent bioremediation simulation experiments to be run for a significant length 
of time during which interlayer mass transfer could be observed before the flow cell 
boundaries were encountered by the plume. This allowed for the study of the 
contaminant plume and associated chemical and microbiological gradients through 
the whole length of the flow cell. With respect to the vertical dimension, the same 
constraints held as in the longitudinal direction. In addition, it was very important to 
have enough space in the vertical direction to create the desired K heterogeneity. 
Based on the constraints described above and the literature review, different 
flow cell sizes were tested using RT3D in VMOD. With regard to the length, because 
 
 64 
 
of space limitations in the laboratory, flow cell lengths on the order of 10 m, such as 
used by Bath et al. (2001) were not feasible. Therefore, based on the laboratory space 
constraints, lengths of 1 m, 1.2 m, and 1.5 m were chosen for model simulations of 
the transport of a step input of conservative tracer in the flow cell. The results 
demonstrated that there were no real significance differences among these three 
lengths. The only difference was that the longer the flow cell, the longer was the time 
required for the solute plume to reach the end of the flow cell. Correspondingly, the 
time to reach steady state was also different in each case. Based on these model runs, 
and assuming an observation period of about one-week's length during the transitory 
phase of each experiment would be sufficient for adequate data collection, a flow cell 
length of 1.2 meter was selected. 
The flow cell heights chosen for testing in the modeling simulations were 0.2 
m and 0.3 m, based on the literature review. Based on model simulations, a height of 
0.3 m was selected over 0.2 m, because of the greater flexibility with 0.3 m. For 
example, with a height of 0.3 m, the bottom, low-K layer, could be set as deep as 0.2 
m, which provided more space for the observation of dispersion. In addition, with the 
height of 0.3 m, it is be more convenient to set up multiple layers to simulate more 
complex heterogeneities in future experimental studies using the flow cell.  
In summary, based on the literature study and model simulations, the 
dimensions for the flow cell were selected to be 1.2 meter long (horizontal direction, 
x coordinate), 0.3 meter high (vertical transverse direction, z coordinate), and 0.1 
meter wide (horizontal transverse direction, y coordinate). 
 
 65 
 
5.2.2 Flow Cell Material 
After sizing the flow cell, the next step was the design and construction of the 
flow cell reactor. A schematic of the basic design of the reactor assembly is presented 
in Figure 5.1. The flow cell was designed to simulate a heterogeneous sand aquifer, 
with the groundwater flow represented by the water pumped through the reactor.  
The walls, bottom, and the lid were made from untempered glass (Glass 
Distributors Inc., Bladensburg, MD, USA). Glass has been a very common choice in 
previous similar studies, e.g., (Voudrias and Yeh 1994; Whelan et al. 1994). There 
were at least two reasons for the selection of glass: (1) it allows visual observation of 
the reactor, and (2) glass is a hydrophilic surface, which prevents the preferential 
wetting and spreading of hydrophobic organics. Untempered glass was chosen so that 
it would not shatter when holes for sampling needles, influent fittings and effluent 
fittings were drilled. A glass thickness of "
2
1
was chosen based on literature review 
and consultation with professionals from Glass Distributors Inc.. The dimensions of 
the glass plates used are as follows: two side wall plates "
4
1
52  long ? "
4
3
16  wide; 
two end wall plates "
4
3
16  long? "5 wide; one top plate "
4
1
52 long ? "5 wide; and 
one bottom plate "
4
1
53 long ? "5 wide (Item 1, Appendix A). 
 
 
 66 
 
 
Effluent
Top View
Side View
Constant Head 
Tank
Sampling Port
Sampling 
Needle
Influent 
Clearwell
Effluent 
Clearwell
120 cm
Flow meter
Feed Bottle
Screen
Glass Plate
Valve (A)
Sampling Port
Pump
Break Tube
Valve (B)
30 cm
35
 cm
130 cm
10
 c
m
A
B
C
D
E
1234567891011
Valve
 
Figure 5.1 Schematic diagram of experimental aquifer assembly with influent and effluent appurtenances, and sampling port 
identification. 
 
 67 
 
To support the weight of the filled flow cell reactor and ensure the integrity of 
the glass flow cell, it was necessary to provide a support structure. This structure had 
two components. First, an aluminum frame was selected for supporting the reactor 
walls. An aluminum frame provided a strong support structure, yet allowed flexibility 
in fabrication. Second, because of the large size and heavy weight of the flow cell 
when filled with the porous media, it would be very difficult to move the flow cell in 
the laboratory. In order to solve this problem and provide increased flexibility and 
ease of operation during experiments, the aluminum frame was attached to an 
aluminum cart. The cart was made of 3.8 cm?0.3 cm aluminum strips with overall 
dimensions of 6' (L) ?16" (W) ?33" (H), which supported a top surface made from a 
piece of 6' (L) ?16" (W) ? "
4
1
 (Thickness) aluminum sheet. Four thermoplastic 
Rubber Wheels (Item 2, Appendix A) with brakes (Item 3, Appendix A) were 
installed on the cart so that it can be easily moved around in the laboratory, or held in 
place. 
5.2.3 Inlet and Outlet Assemblies 
The flow cell was equipped with vertical clearwells at the influent and effluent 
ends, which created medium zone of 1.2 m in length and functioned to provide an 
even distribution of the inflow and outflow across the sand. To form the clearwells 
and contain the sand, stainless-steel screens were installed in the flow cell. The 
screens were made out of stainless-steel to provide chemical resistance, with 
dimensions of 10 cm wide and 35 cm high. The screens were positioned the correct 
distance from the end walls to create the desired size of clearwells, which were 
nominally 5 cm long (Figure 5.1). In designing the screens, two requirements had to 
 
 68 
 
be met. One, the screens should distribute the inflow and outflow evenly. Two, the 
screens had to contain the sand. Most designs used by other researchers, e.g., (Pearce 
et al. 1994; Voudrias and Yeh 1994; Whelan et al. 1994) have used an 80-mesh 
stainless-steel wire cloth (mesh opening = 177 ?m) to provide the permeable screen. 
However, in this study, a very fine sand (d
10
 = 0.15 mm) was used to create the 
hydraulic conductivity heterogeneity; therefore, a 150-mesh stainless-steel wire cloth 
(mesh opening = 66 ?m) was chosen to form the screens (Item 4, Appendix A). The 
screens were fabricated in the Clark School of Engineering machine shop by Mr. 
Bernie LaFrance. The wire cloth was sandwiched in between two single one-piece 
frames fabricated from stainless steel (Item 5, Appendix A), and the assembly was 
held together using 24 stainless steel bolts. Eleven of the bolts were chosen to be 
longer than the others to position the screens the correct distance from the end walls. 
The bolts were secured with nuts (Item 6, Appendix A). In addition, two hex nuts 
were placed at each end of the longer bolts to provide a larger area to set against the 
walls. 
Another key aspect of designing the inlet and outlet assemblies was the 
selection of appropriate influent and effluent tubing and fittings for providing the 
flow to and from the influent and effluent clearwells, respectively. Because both the 
influent and effluent contained simulated groundwater with the naphthalene model 
contaminant, it was important that all the parts be made of a material that was 
resistant to the contaminant. Teflon was selected to meet the requirement and all 
tubing and fittings mentioned below are made of Teflon unless noted otherwise. For 
the  influent flow, a Teflon bulkhead Union (Item 7, Appendix A) and white Teflon 
 
 69 
 
washer (Item 8, Appendix A) were used to connect the influent tubing (Item 9, 
Appendix A) to the influent clearwell. The union had a compression fitting into which 
the "
8
3
 O.D. Teflon influent tubing was inserted. Next, a reducing union (Item 10, 
Appendix A) was used to change from the "
8
3
O.D. Teflon tubing into "
4
1
Teflon 
tubing (Item 11, Appendix A) because of connection convenience. An influent 
sampling port was then installed in the influent flow line. The sampling port consisted 
of a branch tee (Item 12, Appendix A) and a Mininert valve (Item 13, Appendix A). 
Next, a Masterflex
?
 L/S
?
 PTFE-tubing pump (Item 14, Appendix A) was installed 
upstream of the sampling port, to control the simulated groundwater flow. Finally, a 
simple flow-meter was installed upstream of the sampling port, between the feed 
bottle and the pump. The flow meter was constructed using a branch tee (Item 12, 
Appendix A) inserted into the influent line, with one valve (valve A) (Item 15, 
Appendix A) installed in the upstream line, and one valve (valve B) in the branch 
line. In addition, there is a 20 ml glass pipette installed in the branch line. A 
schematic of the flow-meter is shown in Figure 5.2. During normal flow cell 
operation, the valve B was closed. However, to measure the flow rate, valve B was 
opened and the feed solution was allowed to fill the flow meter pipette. The valve A 
was then closed and the flow rate was determined by measuring the time required to 
pump a certain volume of solution from the pipette. After measuring the flow rate, 
valve A was opened and valve B was closed.  
 
 70 
 
Feed Bottle
To tank
Valve A
Valve B
Pump
Pipet
 
 
Figure 5.2 Schematic of in-line flow meter. 
 
 
 71 
 
The effluent flow system involved an assembly of three "
4
1
O.D. Teflon tubes 
(Item 11, Appendix A) leaving the effluent clearwell at different elevations to form a 
single mixed effluent stream (Figure 5.1). The three outlets were placed 10 cm, 17.5 
cm and 25 cm above the bottom of the flow cell, respectively. This effluent system 
yielded representative, vertically-averaged effluent concentrations. Three bulkhead 
unions (Item 16, Appendix A) and white Teflon washers (Item 17, Appendix A) were 
used to connect the tubing and clearwell. These three effluent tubing lines were 
brought together in a union cross (Item 18, Appendix A). The effluent line then 
continued as a single line to a valve (Item 15, Appendix A), which allowed the flow 
leaving the flow cell to be stopped, and then to a sampling port, the same as described 
for the influent line, which allowed sampling of the flow cell effluent. The union 
cross and the sampling valve were held by three-prong clamps on the same support 
stand. The effluent continued to a constant-head reservoir, which could be raised or 
lowered to set the hydraulic head at required elevation to create the desired flow 
(Figure 5.1). The effluent tubing was connected to the constant-head reservoir by a 
male pipe adapter (Item 19, Appendix A). The constant-head reservoir was placed on 
a support stand to facilitate its upward and downward movement for adjusting the 
water-level in the flow cell. From the constant-head reservoir, the effluent flowed 
through a long segment of "
4
1
O.D. Teflon tubing that led into an activated carbon 
trap (Barnstead/Thermolyne, Dubuque, IA U.S.A.), which was used to sorb the 
naphthalene remaining in effluent. Subsequently, the effluent line was drained into a 
sink. 
 
 72 
 
The dimensions of the adjustable constant-head reservoir were 8" (L) ?5" (W) 
?4"(H), and it was constructed from 0.23" thick sheet of Extruded acrylic sheet (Item 
20, Appendix A). The constant-head tank was also fabricated in the Clark School of 
Engineering machine shop by Mr. Bernie LaFrance. The constant-head tank was 
designed as a closed system to minimize the release of volatized naphthalene into the 
laboratory. Inside the constant-head tank, there was a 2.5" high barrier wall that 
formed a water basin at the influent end of the tank, the water surface of which 
defined the water level in the flow cell. Threads were drilled in the influent wall, 
effluent wall and top cover into which previously described male pipe adapters (Item 
19, Appendix A) were screwed. The influent and effluent pipe adapters were used to 
connect the constant-head reservoir to the effluent tubing. The pipe adapter in the top 
cover was used to connect to an activated-carbon tube. This arrangement kept the 
pressure in the constant-head at atmospheric pressure, while permitting capture of any 
volatized naphthalene. 
5.2.4 ISFC Construction 
The glass plates for the flow cell wall were purchased and cut to size by Glass 
Distributors Inc. (Bladensburg, MD, USA). In addition, holes of appropriate size for 
the chosen influent and effluent fittings were drilled through the glass by Glass 
Distributors Inc. as well. The influent hole was "
8
5
diameter, and the three effluent 
holes were "
2
1
diameter. The holes for the sampling needles were drilled with a 
diameter of "
8
1
, which was the smallest hole that the glass shop could drill. 
 
 73 
 
After the glass was cut and drilled, the next step was to bond the plates 
together to build up the flow cell. The glass plates were bonded with silicone caulk 
(Item 21, Appendix A). Three persons were required to perform the assembly given 
the size and weight of the glass plates. During the assembly, the stainless-steel end 
screens were used as a guide to correctly place the sidewalls. After the finishing the 
assembly, the flow cell was left to dry for 48 hours. 
After the flow cell was dried, all the excessive caulking was removed using a 
razor blade. Then, several water leakage tests were run. To do this, the needle holes 
were temporarily closed by silicone caulking (Item 21, Appendix A) and the influent 
and effluent fittings were put in place and sealed with silicone caulking. The flow cell 
was then filled with water to test for leakage. In the initial test, there were quite a few 
leaks detected. Therefore, more silicone caulking was placed along the inner edges of 
the flow cell to close the leaks, and provide a better seal. Another leakage test was 
then conducted and was 100% successful. At this point, the screens were put in place, 
and silicone caulking was used to close the small gap between the screen frame edge 
and the glass wall to prevent sand grains from slipping around the screen into the 
clearwells. A third water leakage test was carried out at this point and successfully 
demonstrated no leaks. 
The final step was to construct and install the aluminum frame around the 
flow cell for structural support and ease of handling. The aluminum frame, which was 
also constructed in the Clark School of Engineering machine shop by Mr. Bernie 
LaFrance, was made such that it did not touch the glass flow cell. Instead, a 5 mm gap 
 
 74 
 
was left open between the glass and the frame. The gap was filled with rubber pads 
that held the flow cell tightly in place. 
5.2.5 Pore Water Sampling Ports Installation 
A two-dimensional network of 
8
1
"sampling ports was drilled in the front 
glass wall, as discussed above. The positions and labels of the sampling ports are 
shown schematically in Figure 5.1 as well. The locations of the ports were selected on 
the basis of the designs used in previous studies found in the literature and economic 
considerations. The nominal distance between ports in a horizontal row was 10 cm 
and the distance between each row of ports in the vertical direction was 5 cm.  
Stainless-steel, Luer-lock, 18-gauge hypodermic needles (Item 22, Appendix A) were 
used to create the sampling ports. The needles were originally 4" long, but were cut 
down to a length of 
16
9
2 "in order that the end could be positioned at the center of 
the flow cell. The needles were inserted into the holes drilled in the glass and fixed in 
place. The major difficulty in this process was to locate the sampling needles exactly 
in place through the drilled holes, with the needle tip at the center of the flow cell in 
the desired vertical and horizontal location. This was made more challenging by the 
fact that the drilled holes (
8
1
" diameter) were much larger than the needle diameter 
(0.022" O.D.) 
To assist in precisely locating the needle tips, a balsa wood (Specialized Balsa 
Wood, Loveland, CO, USA) board was fashioned that had pin holes punched through 
it that were located to match the center of the holes on the front glass wall. The board 
was fabricated in such a way that when it was placed in the flow cell, the needles 
 
 75 
 
passed through the center of glass holes, and then through the holes on the board, 
with the needle tips just reaching to the opposite side of the board, at which point the 
needle tips were located at the center of the flow cell. When the needles were 
properly located, a small amount of 5-minute Epoxy caulking was first put in the 
holes in the glass wall to position the needles because Epoxy caulking can dry and 
harden quickly. Then, silicone caulking was used to seal the needles in place by 
filling the holes in the front glass. When the silicone caulking dried sufficiently, the 
balsa wood board was gently removed. After being fixed in place, the Luer hubs of 
the needles were plugged with removable Teflon Mininert sampling valves (Item 23, 
Appendix A).  
A schematic of the flow cell?s front glass wall showing the placement of each 
sampling needle is presented in Figure 5.3. Each needle location is referenced by a 
letter for its row and a number for its column, which were used to label each sampling 
location. The exact position of each needle tip was measured after the installation of 
the sampling needles. Each needle tip location is defined by an x- and a z- value, as 
measured from the inside lower corner of the screen on the influent side. The 
locations are listed in Table 5.4. A photograph of the glass flow cell with the screens, 
sampling ports, and aluminum frame and cart in place is presented in Figure 5.4. 
 
 76 
 
12354678
B
D
E
C
A
91011
Z
X
 
Figure 5.3 Placement of Sampling Needles. 
 
 77 
 
Table 5.4 Placement points of the sampling needle tips 
Needle No. X location (cm) Z location (cm) 
B1 15.5 21.3 
C1 15.5 15.9 
D1 15.5 10.5 
A2 25.6 26.7 
C2 25.6 16 
E2 25.6 5.1 
B3 35.7 21.3 
C3 35.7 16 
D3 35.7 10.5 
A4 45.9 26.7 
C4 45.9 16 
E4 45.9 5.1 
B5 56.0 21.2 
C5 56.0 15.8 
D5 56.0 10.5 
A6 66.2 26.7 
C6 66.2 15.8 
E6 66.2 5.1 
B7 76.3 21.2 
C7 76.3 15.8 
D7 76.3 10.5 
A8 86.5 26.7 
C8 86.5 15.8 
E8 86.5 5.0 
B9 96.6 21.1 
C9 96.6 15.9 
D9 96.6 10.5 
A10 106.7 26.6 
C10 106.7 15.7 
E10 106. 5.0 
B11 116.8 21.1 
C11 116. 15.6 
D11 116.8 10.2 
 
 78 
 
 
 
Figure 5.4 Glass ISFC with screens, sampling ports, and aluminum frame and cart. 
 
 
 
 79 
 
5.2.6 Sand Selection, Characterization and Packing 
 To simulate macro-scale heterogeneities, the sand flow cell reactor used in 
these experiments was set up with two layers of sand with differing hydraulic 
conductivities. Therefore, a variety of sands were evaluated for possible use in this 
study as the simulated aquifer media. The two sands ultimately selected for use in this 
research were: (1) Mystic White II from U.S. Silica Company (Berkeley Springs, 
WV, USA), and (2) Filtersil sand from Unimin Corporation (New Canaan, CT, USA). 
Specifically, Mystic White II was selected for use in the fast-K layer and Filtersil 
sand was chosen to be used in the low-K layer. These sands were selected because of 
the difference between their hydraulic conductivities, as discussed below, which was 
sufficient to create a model aquifer with the stratified hydraulic conductivity 
heterogeneity required for this study. The coarse sand and fine sand have d
10
 values 
of 0.8 mm and 0.15mm, respectively, based on data provided by the manufactures. 
The bulk density, particle density and porosity of each type of sand in the 
experimental aquifer were determined as discussed in the Parameter Estimation 
Experiments Section (5.3.2 Porous Media Density and Porosity). In addition, the 
following properties were determined by Agri Analysis, Inc. (Leola, PA USA) for the 
Mystic White II sand after it was prepared as described below. The cation exchange 
capacity (CEC) = 0.3, percentage of organic matter = 0.4%. 
Prior to their emplacement in the flow cell, the sands were prepared following 
the procedures of Murphy et al. (1997). First, trace contaminants were removed from 
sand by washing with 0.25 M NaOH for 12 hours, followed by 0.25 M HNO
3
 for 12 
 
 80 
 
hours. Then, after thorough rinsing with deionized water until the pH reached 
approximately 7, the sands were autoclaved.  
Based on the literature, both dry, e.g., (Oliviera et al. 1996) and wet packing, 
e.g. (Pearce et al. 1994; Voudrias and Yeh 1994; Whelan et al. 1994) methods have 
been used in previous sand-tank laboratory experiments. Oliviera et al. (1996) studied 
the packing of sands for the production of homogeneous porous media and concluded 
that a dry packing consisting of the deposition of sequential 0.2-cm layers, each 
followed by compaction with a metal pestle, or a wet packing while vibrating 
columns, are the best options. In either operation, the sand was always poured or 
sprinkled in small batches to prevent the preferential deposition of large grained 
material. Furthermore, the sand was always deposited from a height of 0.5 cm to 
prevent heap formation. Neither of these techniques could be applied in this research 
exactly as described: the dry packing technique could not be used because of the 
interference to the pestle by the sampling methods, and no method of vibrating the 
flow cell was available for the wet packing method. However, in some previous 
research studies (Pearce et al. 1994; Voudrias and Yeh 1994; Whelan et al. 1994), 
another wet packing method was applied for packing the flow cells. In preliminary 
experiments by these authors, small slugs of a methylene blue dye injected at 
different places in the aquifer showed that the flow was uniform without field 
variations introduced by the packing procedure.   
Therefore, the same packing technique used by Voudrias and his coworkers 
was adopted to pack the sands into the ISFC. Specifically, the ISFC was filled with 
sand by applying about 0.5 kg at a time with a beaker, under a water head of 5 cm 
 
 81 
 
maximum. Sand was always poured through ? 5cm of water in order to prevent 
entrapment of air bubbles and reduce sand stratification. The same sequence of water 
addition and deposition of 5-cm sand layers was repeated until the entire aquifer was 
formed, with the water level always maintained above the top of the sand. The ISFC 
was filled with fine sand to a level of 15 cm, followed by another 15 cm of coarse 
sand. After the aquifer was formed, it was flushed with DI water for three days. Then 
the water level was lowered 3 cm, thereby creating a capillary fringe zone of ~ 3cm, 
which was controlled by the adjustable constant-head reservoir. The top of the sand 
was covered with a sheet of aluminum foil to minimize volatilization from, and 
spilling of undesired materials or chemicals into, the simulated aquifer. The ISFC was 
at that point ready to be utilized in the bioremediation experimental investigation. 
5.3 Parameter Estimation Experiments 
In past studies of heterogeneous systems, a rarely stated underlying 
assumption of scaling up in engineered bioremediation is that the heterogeneity can 
be explicitly represented, with parameter values that are measured in batch and/or 
column studies producing accurate flow and mass transport results at a larger scale. 
An improved understanding of how measured hydraulic and transport parameter 
values relate to in situ values is needed to examine this assumption, however, the 
same assumption was inherent in this work. Specifically, to describe the controlled 
heterogeneous experiments in the ISFC, and the associated detailed observations of 
flow and transport, it was necessary to have the following parameter estimates: 
hydraulic conductivity, bulk and particle densities, porosity, dispersivity, biokinetics, 
and sorption/desorption equilibrium and kinetics. 
 
 82 
 
5.3.1 Hydraulic Conductivity 
There are three commonly used methods for hydraulic conductivity, K, 
measurement: field tests, laboratory tests, and empirical and semi-empirical methods 
based on grain diameters and grain size distributions. Of most interest here are the 
laboratory testing techniques, which typically include the constant head and falling 
head methods. Of these, the constant head method was employed in these 
experiments to measure the saturated hydraulic conductivity. Prior to the hydraulic 
conductivity experiment, the sands were treated following the procedures from 
(Murphy et al. 1997), as previous described in Section 5.2.6.  
The hydraulic conductivities of the sands were measured by the constant-head 
method using a gradient ratio permeameter (Figure 5.5). The method requires setting 
up a cylindrical clear permeameter with a soil sample (in this case sand) of length L 
and cross-sectional area A, and passing water through the system by applying 
differential heads, H. Measurements of the differential heads with steady a flow rate, 
Q, were taken at different time intervals to determine hydraulic conductivity by 
applying the following equation, derived from Darcy?s law, 
AH
QL
K =     (5.2) 
where K = hydraulic conductivity, [LT
-1
]; L = length of sample, [L]; A = cross-
sectional area of sample, [L
2
]; Q = outflow rate, [L
3
 T
-1
]; H = fluid head difference 
across the sample, [L]. The major pieces of equipment required for performing the 
constant head test are illustrated in Figure 5.6, and include the permeameter, two 
constant water head devices (one mounted on a jack stand adjustable and one 
stationary), and manometer boards. 
 
 83 
 
 
 
Figure 5.5 Schematic of the permeameter for the constant head method  
(ASTM-D2434 2000). 
 
 84 
 
 
 
 
 
Figure 5.6 Schematic of the permeameter set up for the constant head method. 
(ASTM-D2434 2000)
 
 85 
 
The specific steps in performing the constant head test were as follows, 
beginning with sand placement. First, before sand placement, the center section of the 
permeameter was weighed on a scale. In the second step, the sand was placed above 
the support cloth to a total depth of 110 mm. The final depth of sand should be 
approximately 100 mm, which was checked after completion of the experiment. 
During placement, the clean sand was poured carefully into the permeameter with a 
maximum drop of soil no greater than 25 mm. No consolidation procedure was 
utilized. A sand leveler was used to remove any excess sand. Third, the weight of the 
center section was taken again for determination of the bulk density, which provided 
an indicator of how close the packing of the sand in permeameter was close to that of 
ISFC. 
Permeameter assembly followed the sand placement. First, the support screen 
and the bottom section of the permeameter were aligned and inserted into the center 
section. When the bottom was secured, the permeameter was positioned into the 
holding stand. The sand was now compressed from 110mm to 100 mm. 
Subsequently, the inflow and outflow constant head devices were connected to their 
corresponding permeameter ports (Figure 5.6) with plastic tubing.  Specifically, the 
outflow device was attached to the bottom permeameter port and the inflow device 
was attached to the top permeameter port. All of the manometer tubes were then 
connected to their corresponding permeameter manometer port, and all over-flow 
tubes to their corresponding outlet ports. 
Following the permeameter assembly, the next step was to saturate the 
system. To accomplish this, a piece of long tubing with very small I.D. was inserted 
 
 86 
 
into the bottom section of the permeameter through the tubing used to connect the 
water outlet and the outflow constant head device, until one end of the tubing reached 
the ceiling of the bottom section, with the other end extruding out the outflow 
constant head device and into the air. Next, the top vent valve on the permeameter 
was opened and the permeameter water outlet was closed off. The permeameter was 
then backfilled with water through the outflow constant head device. The air trapped 
in the system were sucked out via the small I.D. tubing through the open end for 
about 5 minutes until all air bubbles were removed from the system. The top vent 
valve on the permeameter was then clamped off and the system was allowed to stand 
overnight in a no-flow static condition to ensure complete saturation of water. After 
the saturation of the system, a recorded volume, V, of the outflow water was then 
collected in a container over a time period, to calculate the flow rate, Q (=
t
V
) at the 
applied differential heads. The tests were performed in duplicate for each sand to 
investigate the effect of any variability in the packing sand. 
5.3.2 Porous Media Density and Porosity 
The bulk density,
b
? , and particle density, 
p
? , were determined 
experimentally for the Mystic White II sand and the Filtersil sand. Both sands were 
dried at 104?C in the oven for 24 hours before the density experiments. 
The bulk density was determined in two ways: (1) batch experiment in 100 ml 
graduate cylinders; and (2) experimental estimation while packing the ISFC. First, in 
the batch experiment, each type of sand was added to about the 20 ml mark of a 100 
ml graduate cylinder. The sand was compacted by tapping the cylinder base on the 
 
 87 
 
palm of the hand. Following that, about 20 ml more of the sand was added and 
compacted as above. The steps were repeated until 80 to 100 ml of the sand was in 
the cylinder. The sand volume was recorded (ml) and the weight was measured.  The 
experiments were carried out in triplicate. The bulk density was calculated using the 
equation of bulk density, 
b
?  = (mass of dry sand / total volume of sand and air). 
Second, when the sands were packed into the ISFC, the mass of sand added was 
measured for the Mystic White II and Filtersil sands. The bulk density was then 
calculated by dividing the weight of each type of sand by half of the porous media 
volume of the ISFC. 
The particle density was measured using a water-displacement technique 
(Blake and Hartge 1986). First, the mass of an empty flask was measured. Then, 
twenty five grams of dried sand was measured out, placed in the flask using a funnel, 
and the mass of the flask containing the sand was weighed. The mass of sand was 
doubled checked by deducting the mass of empty flask from the mass of flask with 
sand. Subsequently, about 50 ml of DI water was added in the flask. The sand/water 
mixture was brought to a gentle boil by placing the flask on a hot plate to ensure 
complete air bubble removal. The flask was then removed from the heat and allowed 
to cool. The flask was capped and allowed to sit for 24 hours. After 24 hours, the cap 
was removed and the flask was filled with DI water so that the bottom of the 
meniscus was at the 100 ml line. The 100 ml-sand/water mixture was then weighed in 
the flask, and the temperature was recorded using a thermometer. The mass of water 
was calculated by deducting the mass of flask with sand from the mass of sand/water 
mixture. The volume of water was obtained from the mass of water by using the 
 
 88 
 
appropriate density of water at the experimental temperature, while the volume of 
sand was found by taking 100 ml minus the volume of water. Finally, the particle 
density was obtained using the equation of particle density, 
p
?  = (mass of dry sand / 
volume of sand particles (air removed)). 
Once the bulk density and particle density were obtained for the Mystic White 
II and the Filtersil sands, the porosity, n, was calculated by the equation, 
 
p
b
?
?
? ?=1    (5.3) 
5.3.3 Dispersivity 
 With regard to dispersivity, three different dispersivities were 
evaluated: (1) vertical transverse dispersivity (
z
? ) within the two layers, (2) the 
longitudinal dispersivity (
x
? ) within the two layers, and (3) the interlayer mass 
transfer (
z
? '). A continuous point-source tracer study method that was adapted from 
Robbins (1989) was chosen to estimate the transverse dispersivity and longitudinal 
dispersivity within each layer. The interlayer mass transfer coefficient was estimated 
in the control experiments which are discussed in a later section (5.4.1 Control 
Experiments). 
The basic approach of the continuous point-source tracer technique was as 
follows. First, a continuous injection of the conservative tracer was started via a 
sampling needle inserted to the centerline of each layer. The tracer concentration 
variations were then monitored with time at a down gradient sampling port at the 
same vertical location as the injection port. Prior to, and during, tracer injection, 
 
 89 
 
simulated groundwater was flowing through the flow cell at a constant average pore 
water velocity,
x
v , in the longitudinal direction.  
The first step in performing these tracer studies was the selection of non-
reactive tracer, which should be transported through the water system in a manner 
similar to the item of interest. A good tracer should be a stable substance, show no 
reactivity with the system components, and must have reliably quantitative detection 
at low concentrations (Behrens 1986). Based on these characteristics, two non-
reactive tracers were selected for use in these experiments: bromide (Br) and 
fluorescein. 
The non-reactive tracer initially used in the experiments was chosen to be 
potassium bromide (KBr), with mercury chloride (HgCl
2
) added at a concentration of 
400 mg/L to inhibit microbial activities in the flow cell. Bromide was chosen as the 
conservative tracer because it had been used extensively, e.g., (Bath et al. 2001) and 
the presence of mercury chloride as an inhibitor did not interfere with its 
measurement. The tracer injection solution concentration,
0
C , was selected to be 0.1 
M KBr, to maximize the tracer breakthrough concentrations, while avoiding density-
induced flow effects. The concentrations of Br in samples taken from sampling port 
were quantified by ion chromatography (IC). However, after running three tracer 
studies with potassium bromide, the IC instrument was non-functional and the repair 
process was lengthy. Therefore, it was necessary to select an alternative tracer.  
The second non-reactive tracer used was the organic compound fluorescein 
(sodium salt). Fluorescein is an excellent tracer because it is easy to detect and can be 
measured using either its strong fluorescence or highly absorptive character (Klonis 
 
 90 
 
and Sawyer 1996). Compared to absorbance methods, fluorescence sensitivity is tens 
to thousands times better. Therefore, fluorescence was utilized to measure the 
fluorescein tracer concentration in this study. The other advantages of fluorescein 
include its low sorption tendency (Behrens 1986) and its relatively low temperature 
exponent (Feuerstein and Selleck 1963). The disadvantages are its pH sensitivity and 
its photochemical instability. However, Smith and Pretorius (2002) demonstrated that 
conservative behavior of fluorescein is possible if one correctly evaluates the system 
parameters. For example, Smith and Pretorius (2002) noted that for optimum results, 
pH should be between 6 ~ 10. They further concluded that the test area and samples 
must be protected from bright light when using fluorescein as a tracer so that the 
fluorescein can be used quantitatively. The simplest way to block light is to perform 
the test at night and to store samples in the dark.  
For this study, the fluorescein dye solution (
0
C = 10 mg/L) and bulk feed 
solution were made using sodium bicarbonate-sodium carbonate in deionized water 
buffer solution (0.0138 M NaHCO
3
, 0.012 M Na
2
CO
3
, pH = 10), with AgNO
3
 added 
as a biocide. The experiment was set up as closely as possible to the approach 
described by Robbins (1989). This method involves three important factors: (1) the 
bulk flow of the reactor, (2) continuous point source injection of a known 
concentration of tracer, and (3) continuous sampling at a point downgradient of the 
point source. An average pore water velocity of 9 m/d was used in Robins' study 
(1989), which corresponded to a bulk flow rate of 75 ml/min for the reactor used in 
this study. However, the Teflon pump described Section 5.2.3 has a maximum flow 
rate of 65 ml/min. Therefore, for this study, a flow rate of 60 ml/min was selected, 
 
 91 
 
corresponding to average pore water velocities of approximately 10 m/d and 0.75 m/d 
in the high-K and low-K layers, respectively. 
The tracer point source injection was provided using a Harvard Apparatus 
Syringe Pump Model 22. To perform the injection, a 5-ml plastic syringe filled with 
tracer solution, which was connected via Tygon tubing to a stainless-steel microtube, 
was set in the syringe pump. The stainless-steel microtube was inserted through one 
of the sampling needles (B3 for the coarse sand in the high-K layer, and D3 for the 
fine sand in the low-K layer) in such a way that its tip matched the sampling needle 
tip location. Before the steel microtube was inserted into the sampling needle, it was 
ensured that the tracer solution reached the end of the microtube. The tracer injection 
was then begun, at a flow-rate, 
n
Q , of 0.05 ml/min, which was selected so as to not 
disturb the flow in the flow cell and was utilized successfully by Schicke (1996).  
After the tracer injection began, samples were taken through a sampling 
needle (B3 for the coarse sand in the high-K layer, and D3 for the fine sand in the 
low-K layer), that was 20.2 cm downgradient of the injection needle at the same 
vertical location. The proper selection of the sampling location is an important factor 
to get valid results for dispersivity. Specific requirements have been described in 
detail by Robbins (1989). In addition, the sampling flow rate, Q
sample
, should be a 
small fraction of the bulk flow so that the flow disturbances can be neglected. In the 
high-K layer, the sampling flow rate of less than 1.15 ml/min was achieved by 
opening the Mininert valve at sampling port B3, which corresponded to less than 2% 
of the bulk flow rate. In a low-K layer, the samples were taken using a pump (Gilson 
Miniplus 2, Vilier-le-Bell, France) at a flow rate of 0.7 ml/min to ensure negligible 
 
 92 
 
disturbance to the slow flow in the low-K layer. In both layers, sample volumes of 2 
ml were collected. To correct for the time difference between when the samples were 
taken and when the solution entered the sampling needle, a calculation based on the 
sampling flow rate and microtube diameter was performed to determine the time 
delay between when the tracer reached the needle tip and when it exited the Mininert 
valve. In this way, the ?actual? sampling times were obtained by the subtracting the 
delay time from the original sampling times. The general experimental conditions for 
these tracer studies are summarized in Table 5.5. 
 
Table 5.5 Point-source tracer experimental operational conditions 
Fluorescein 
Experiments 
Bulk flow rate 
(ml/min) 
Sampling flow rate 
(ml/min) 
injection flow rate 
Q (ml/min) 
Concentration 
0
C  (mg/L) 
High-K layer (a) 60 1.15 0.05 10 
 
High-K layer (b) 60 1.15 0.05 10 
 
High-K layer (c) 60 1.15 0.05 10 
 
Low-K layer (a) 60 0.7 0.05 10 
 
Low-K layer (b) 60 0.7 0.05 14 
 
Low-K layer (c) 60 0.7 0.05 22 
 
Low-K layer (d) 60 0.7 0.05 22 
 
 
 93 
 
For each layer, the steady-state tracer concentration, C', at the sampling 
location was calculated by averaging the sample concentrations after they became 
constant.  The vertical transverse hydrodynamic dispersion coefficient,
z
D , was then 
calculated from an analytical expression, which assumes that the horizontal and 
vertical transverse dispersivities are equal (Robbins, 1989): 
( )
'
0
4 xC
QC
D
z
?
=    (5.4) 
where: 
0
C = the injection concentration; Q = the tracer injection rate; and x = the 
distance between the tracer injection port and the sampling port. The vertical 
transverse dispersivity (
z
? ) was calculated using, 
n
xzz
vDD ?+= *    (5.5) 
where *D is the effective coefficient of molecular diffusion in the porous medium, 
x
v  is the average interstitial velocity, and n is 1.0 (Rose, 1977). In the experimental 
design, 
x
v  was selected to be sufficiently large that molecular diffusion was 
negligible compared to mechanical dispersion, so that 
x
z
z
v
D
=? . 
Longitudinal dispersivities were determined by using the tracer breakthrough 
curve at the sampling port, and the following equation (Robbins 1989): 
()
()( )
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
=
2
1
2
2
1
'
tD
tvx
erfc
C
C
x
x
   (5.6) 
where C = the tracer concentration measured down gradient at distance x over time, t, 
until steady state is achieved, and 
x
D is the longitudinal hydrodynamic dispersion 
coefficient. Only 
x
? and 
x
D are unknown in Equation 5.6.  Therefore, the best-fit 
 
 94 
 
values for 
x
v  and 
x
D  were obtained by using non-linear regression to fit Equation 5.6 
to the experimental breakthrough curves.  
The non-linear regression was performed using a FORTRAN program 
?trafit3d? (Schicke 1996), which calculates the sum of the squares of either the 
absolute or relative residuals between the normalized experimental conservative 
tracer data and the normalized flux-averaged concentration calculated using the 
continuous point source model at steady state as described by Robbins (1989) and 
gives as an output of the best fit longitudinal hydrodynamic dispersion coefficient, the 
average pore-water velocity and the porosity (Appendix 2). In this research, the 
absolute non-linear regression was utilized. The best fit 
x
v  and 
x
D parameters are 
obtained using a modified Levenberg-Marquardt method to minimize the sums of the 
squares of the residuals between the observed and calculated concentrations. The 
longitudinal dispersivity (
x
? ) was calculated using, 
n
xxx
vDD ?+= *    (5.7) 
As in the determination of 
z
? , 
x
v  was selected to be sufficiently large that molecular 
diffusion was negligible compared to mechanical dispersion, so that 
x
x
x
v
D
=? . 
5.3.4 Biodegradation Kinetics 
Microbial kinetic parameters for growth of Uper-1 on naphthalene were 
estimated in batch experiments conducted in support of the simulated bioremediation 
experiments in ISFC. Some of these experiments were performed in conjunction with, 
or by, Ms. Eunyoung Hong. Biokinetic parameters are commonly measured in batch 
systems using suspended cells in a liquid medium. One of the advantages of this 
 
 95 
 
approach is that mass-transfer limitations are minimized. Another key advantage of 
batch tests is the ease of performing the test and the quickness with which the test can 
be done (Grady Jr. et al. 1996). However, Monod kinetic parameters estimated using 
non-steady-state batch analysis are subject to large uncertainties. These uncertainties 
are a function of the initial experimental conditions, the values of parameters, the type 
and magnitude of the measurement errors, and the number of samples (Liu and 
Zachara, 2001).  
Due to the large uncertainties related to the kinetic parameters estimation, 
careful manipulation of experimental conditions was required to reduce the 
correlations between the Monod parameters, thereby allowing for the estimation of 
unique parameters with the lowest degree of uncertainty. Liu and Zachara (2001) 
found that the correlation and relative standard deviations of Monod parameters are a 
function of a few dimensionless variables involving the initial substrate 
concentration, 
0
S  , and the initial biomass concentration,
0
X . Analysis of these 
dimensionless variables allowed for identification of the optimal experimental 
conditions for estimation of unique and accurate Monod kinetic parameters.  
For the purpose of this research, the concentrations of naphthalene and 
biomass were expressed as chemical oxygen demand (COD). Thus, all mass-related 
parameters, such as 
s
K  and the yield coefficient,Y , were expressed as mg/L as COD 
and units of biomass COD formed per unit of naphthalene COD removed, 
respectively. Grady and co-workers define kinetic parameters obtained with a high
0
S  
to 
0
X  ratio (e.g., 
0
0
X
S
> 20, both as COD) as the ?intrinsic? kinetics, that is, the 
 
 96 
 
maximum capability of the culture (Grady Jr. et al. 1996). On the other hand, ?extant? 
biokinetics, which represent the capability of the culture in the source reactor from 
which it was obtained are determined with a low 
0
0
X
S
(e.g., 
0
0
X
S
< 0.025, both as 
COD) (Grady Jr. et al. 1996).  
In this work, preliminary batch biokinetics determinations were performed 
under intrinsic and extant conditions using batch culture of Uper-1 as the inoculum. 
In addition, batch experiments were also carried out to estimate the biodegradation 
kinetics for the biomass in the flow cell once the ISFC experiments were started. It 
was originally planned to perform these experiments under extant conditions. 
However, the cell concentration collected from the effluent samples was very low, 
which made it impossible to set up the required low 
0
0
X
S
. Therefore, the batch 
experiments utilizing the cell culture from the effluent samples were all set up with a 
high 
0
0
X
S
to represent the ?intrinsic? biokinetics in the flow cell. This approach 
worked for the Scenario #1 Phase I sample. However, during Scenario #2 Phase I, 
with inhibited biodegradation, no usable data were obtained using the ISFC effluent 
inoculum under intrinsic conditions, as discussed in Chapter 6. Therefore, it was 
necessary to use the results of an ?extant? kinetics study with the primary culture 
Uper-1 inhibited by high salinity to obtain the kinetic parameters for Scenario #2 
Phase I. 
In the kinetic parameters study, respirometry (Challenge AER-200, 
Aerobic/Anaerobic respirometer system, Challenge Environmental Systems, Inc.) was 
used to estimate the kinetic parameters of aerobic naphthalene biodegradation by 
 
 97 
 
Uper-1. The basis for batch respirometry is the balanced equation for microbial 
growth that links substrate consumption, biomass growth, microbial product 
formation and oxygen uptake (Smets et al. 2001): 
PYYXOYS
Pu
+???+ )1(    (5.8) 
where, S is the instantaneous substrate concentration (M/L
3
), Y is the true biomass 
yield coefficient (M/M), O
u
 is the accumulative oxygen uptake (M/L
3
), X is the 
instantaneous biomass concentration, Y
p
 is the product yield coefficient (M/M), and P 
is instantaneous microbial product concentration (M/L
3
). 
 During a batch growth assay using respirometry, time-series profiles for 
oxygen uptake are obtained. In this study, the true biomass yield coefficient, Y, was 
calculated directly from the raw oxygen accumulative uptake profile using the 
formula (and assuming 
p
Y = 0 mg P COD/ mg S COD (Brown et al. 1990)): 
p
uplateau
Y
S
O
Y ??=
0
1     (5.9) 
where 
uplateau
O is the oxygen accumulative uptake at the beginning of the plateau in the 
respirogram, indicative of the substrate exhaustion. The kinetic parameters (
Sm
K,? ) 
were estimated by fitting the experimentally measured oxygen data to the Monod 
kinetic model for a batch culture using nonlinear parameter estimation techniques 
(Smets et al. 2001), as discussed further below.  
The inoculum for the preliminary estimates of the intrinsic biokinetics of 
Uper-1 was prepared by following the protocol described in Section 5.1.3  
Microorganism.  Thus, the cells utilized were extensively adapted to late exponential 
phase growth. Before the inoculum was transferred into the respirometer vessels, it 
 
 98 
 
was harvested by centrifugation (12 000 x g for 10 min), freed of residual glycerol 
and naphthalene by resuspension and washing (three times) in MSNS. The inoculum 
for the ?intrinsic biokinetics? of Uper-1 in the ISFC was aseptically collected from 
the ISFC effluent during various experimental phases of selected bioremediation 
simulation experiments. 
The same basic procedure was followed in all of the respirometry studies, 
whatever the source of the inoculum. First, the respirometer vessels and Teflon-
coated stir bars were washed with Alconox solution. Then, each respirometer vessel 
was filled with 555 mL artificial groundwater (MSNS), leaving 550 mL after 
autoclaving. This operation was calibrated in triplicate before the preparation. The 
respirometer vessels were then autoclaved at 121?C for 15 minutes.  
A naphthalene solution of 10,000 mg/L in DMF was prepared in advance. An 
aliquot of 1 ml of this naphthalene in DMF solution was taken by a syringe with 
Teflon filter. A small amount of the solution was pushed through to rinse the filter. 
Then a volume of 0.55 ml of the filter-sterilized stock solution was spiked into each 
respirometer vessel. The nominal target concentration was 10 mg/L in the 550 ml 
solutions, which was confirmed by analyzing the samples taken at this point for 
naphthalene by either gas chromatography (GC) or spectrofluophotometry. The 
theoretical concentration of COD in 10 mg/L naphthalene solution is 30 mg/L. In 
order to ensure that substantial microbial growth occurred during the biokinetic assay, 
the COD of the seed culture was measured and the volume of seed culture added into 
the vessels was adjusted as appropriate to give a 
0
0
X
S
ratio either >> 20 or << 0.025 
to meet the requirement of desired biokinetic parameter assay. 
 
 99 
 
After inoculation, the autoclaved CO
2
 traps were filled with about 5 ml KOH 
solution (45% by weight), which had been filtered-sterilized (0.2 ?m filter). The 
KOH traps were then aseptically inserted into the prepared respirometer vessels and 
capped snugly. The vessels were placed on the Challenge MS58-300 magnetic stirrer 
base in the constant temperature water batch (24?C). The stirring speed was adjusted 
so as to produce a vortex that extended from the gas-liquid interface down to the 
stirring bar. A sterile 20-gage needle was inserted through the septum momentarily to 
vent the vessel in order to equalize the pressure prior to the test. The operation was 
repeated to ensure the pressure inside all the vessels that were connected to flow cells 
had equilibrated with the pressure outside of vessels. The sterile needles attached to 
the tubing connected to each of the oxygen flow cells were inserted into the septa of 
the vessels at an angle of 45?. The needle tips were carefully adjusted to ensure free 
flow of oxygen by preventing the needle tip from contacting the vessel wall or the 
KOH trap holders. The oxygen cylinder was then opened and the oxygen flow rate 
was set up at approximately 3 bubbles /second. 
After confirming that the software was set up properly, a trial run was started 
to test the communication between the vessels and respirometer. A 10-ml syringe and 
20-gage needle were used to withdraw headspace gas from each respirometer vessel 
until one or two bubbles passed through the respective flow cell. The counts were 
checked to make sure that they were registered on the computer monitor. At this 
point, the trial data were deleted by resetting the system counters and timers and data 
acquisition was initiated. Subsequently, the oxygen uptake rate and oxygen uptake 
accumulation were followed during the experimental run. 
 
 100 
 
Replicate vessels containing the inoculum vessels were set up along with the 
duplicate control vessels for each respirometry experiment. The preliminary 
respirometry experiments were run with inoculum from the batch primary culture. In 
addition, experiments were also performed using as an inoculum from the cells in 
effluent sample collected during Scenario #1 Phase I and Phase II simulated 
bioremediation experiments, as well as the cells in effluent samples collected during 
Scenario #2 Phase I simulated bioremediation experiments.  
The parameters in the nonlinear Monod rate expressions were estimated by 
nonlinear methods. The algorithms in Microsoft Excel (Version 2002) add-in Solver 
were first utilized to solve the non-linear optimization of the sum of squared errors 
(SSE),  which  was evaluated by taking a measured oxygen uptake value, subtracting 
the model predicted value from it, squaring the difference, and adding that value to a 
running total (Smets et al. 2001). However, it was difficult to obtain a reasonable K
s
 
value based on this non-linear optimization method, which usually resulted in the 
fitted K
s
 equaling the upper constraint. Therefore, ultimately a FORTRAN program 
?Nvolma?, which was obtained from Prof. Barth F. Smets (Technical University of 
Denmark) and originally developed at Clemson University, was used to perform the 
biokinetic parameter estimation (Appendix 3). This program fits a curve to batch 
oxygen uptake data using Monod or Andrews kinetics to describe the removal of a 
non-volatile substrate.  The best fit values for the parameters ?
m
 and K
s
 are found 
using a least sum of squares estimator to determine the best fit for a given set of 
experimental data. The complex search routine of Box (Kuester and Mize 1973) is 
used to find the best fit. Initial estimates of ?
m
 and K
s
 for the fitting program were 
 
 101 
 
obtained from the oxygen uptake data by following the procedures of Brown et al. 
(1990). 
5.3.5 Sorption/Desorption Equilibrium and Kinetics 
To accomplish Scenario #3, the sorptive characteristics of the two-layered 
porous media used in Scenario #1 and Scenario #2 needed to be increased so that the 
meso-scale desorption rate was the overall rate controlling mass-transfer process. The 
overall goal in creating Scenario #3 was to develop a synthetic system that modeled a 
real subsurface system, yet possessed defined quantitative physicochemical 
parameters. Obviously, the physicochemical properties of the model system should be 
similar to actual subsurface systems so that the results can be translated into realistic 
systems. There were two major concerns with the synthetic system. First, the 
modified porous media system needed to possess a strong sorptive capacity so that 
the sorption/desorption would limit the overall bioremediation rate. Second, it was 
important that the modification to the porous media not significantly alter the 
hydraulic conductivities of the system, nor alter the amount or the nature of the 
surface area available for microbial attachment and growth for Scenario #3 compared 
to Scenario #1 and 2.  
The original idea was to increase the sorptive capacity of the porous media by 
using the protocol of Szecsody and Bales (1989) to create a thin, patchy surface 
coating of organosilanes, including aliphatic-chain groups (C
1
, C
8
, and C
18
) and an 
aliphatic chain terminating in a phenyl group, on the silica sand. However, the idea 
was abandoned because of the health hazards and potential for laboratory damage 
posed by organosilanes (J. Szecsody, personal communication).  
 
 102 
 
Three other alternatives were evaluated in batch sorption experiments, which 
fell into two categories of synthetic sorbent systems: (1) modification of the sand 
surface to increase the sorption capacity, and (2) mixing sand with a certain amount 
of a synthetic sorbent. Two of the alternatives tested for the preparation of a synthetic 
model soil were in the first category--2,2,4,6,6,8,8-heptamethylnonane (HMN) and 
surfactant coated sands-- and one was in the second category-- mixing sand with 
Amberlite Resins. The investigation of these three options is described in detail in the 
following subsections.  
In these experimental studies, all of the apparatus used consisted of 
borosilicate glass where possible, or Teflon, which demonstrated negligible sorption 
of naphthalene. These experiments consisted of two general types: rate investigation 
and equilibrium isotherm studies. The rate studies were designed to assess the effect 
of time on the partitioning of naphthalene and sorption/desorption kinetics.  The 
equilibrium isotherm studies were designed to evaluate the partitioning coefficient of 
naphthalene onto the modified synthetic sorbents. In order to select a synthetic 
sorbent with a strong sorptive capacity, as characterized by the partitioning 
coefficient,
d
K , sorption isotherm experiments were generally conducted first, 
followed by the rate studies, if sorbent was considered promising. 
5.3.5.1 Unmodified Sand Sorption 
As the first step in the sorption kinetics study, the sorption isotherm of 
naphthalene on unmodified Mystic White II coarse sand was investigated. 
Naphthalene solutions of 5, 4, 3, 2, 1 and 0.5 mg/L were prepared and added to 40 ml 
amber vials along with 1 g of sand. All the vials were then tumbled on a rotary shaker 
 
 103 
 
(RKVS, Appropriate Technical Resources, Inc., Laurel, MD USA) 124 hours to 
ensure that the aqueous-solid equilibrium was reached. The aqueous equilibrium 
naphthalene concentrations were determined using the spectrofluorophotometer. The 
sorbed naphthalene was obtained by the concentration difference method (Burris and 
Antworth 1992). 
5.3.5.2 HMN-Coated sand sorption 
The first option evaluated was to increase the sand sorptive capacity by 
coating HMN onto the surface of the sand, and thereby simulate the sorption of PAHs 
onto organic matter in the subsurface. HMN was selected because of its high fluidity, 
which allowed a very homogeneous coating on the sand (Appert-Collin et al. 1999). 
To prepare the sand, it was coated with 0.5% HMN (weight fraction) and then air 
dried for 48 hours before use. 
The preliminary naphthalene sorption isotherm was accomplished by placing 
1 g of HMN-Coated sand in each of twenty 40-ml borosilicate vials. Then 40 ml of 
naphthalene solution in MSNS with varying initial concentrations (0.5 mg/L, 1 mg/L, 
2 mg/L, 3 mg/L, 5 mg/L, 10 mg/L, 12.5 mg/L, 15 mg/L and 20 mg/L) was added to 
duplicate sets of vials. In addition, duplicate blank samples were prepared without 
HMN-Coated sand to determine the naphthalene loss due to sorption to the vials or 
vaporization. All the vials were then tumbled on a end-over-end shaker for 48 hrs to 
minimize mass transfer limitations. All steps were performed aseptically. After 48 
hrs, 2 ml of equilibrium solution was taken from each vial and subjected to liquid-
liquid extraction and analysis by gas chromatography (GC) as described in the 
Analytical Methods Section. The aqueous phase concentrations measured were then 
used to calculate the solid-phase concentrations by difference.  The equilibrium solid-
 
 104 
 
phase concentration data were presented as a function of the equilibrium liquid-phase 
concentrations to analyze the sorption isotherm. 
5.3.5.3 Surfactant-coated sand sorption 
The second option tested, the use of cationic surfactants to enhance the 
sorptive capacity of the sand, was investigated very carefully. The surfactant-coated 
sand sorption study was conducted with the help of James Stagge, an undergraduate 
Honors student in Civil and Environmental Department, University of Maryland. 
Cationic surfactants have been used successfully by other researchers to enhance 
organic contaminant sorption onto sand, e.g., (Brown and Burris 1996; Sheng et al. 
1998; Wang et al. 2001). In particular, the cationic surfactants cetylpyridinium 
chloride (CPC) and cetyltrimethylammonium bromide (CTAB or HDTMA) have 
been the most widely used cationic surfactants for this purpose. Therefore, they were 
studied for the feasibility of their use in this research. The CPC was chosen over 
CTAB because of its greater affinity for natural materials (Kibbey and Hayes 1993), 
although the latter has been used more often in the surfactant research field.  
There were three steps in the evaluation of sorption of naphthalene onto CPC-
coated sand. First, it was necessary to evaluate the equilibrium time for CPC sorption 
onto the sand. This was accomplished by mixing a group of 28 samples for different 
time intervals. Each sample contained 40 ml of 30 ?M CPC solution and 1.2 g of 
Mystic White II coarse sand in 40 ml amber vials. The pH and ionic strength in these 
samples were varied. Specifically, in addition to a blank, for every mixing time 
intervals, there were four subsets of samples: Groups 1 and 2 had pH=7 and NaCl = 
0.0 and 0.05 M, respectively, and Groups 3 and 4 had pH=9 and CaCl
2
 = 0.0 and 0.05 
 
 105 
 
M, respectively.  The vials were mixed on the end-to-end shaker for 4, 8, 12, 24, 36, 
and 48 hours. At each sampling point, four samples, one from each group, were 
sacrificed and the final CPC concentrations were determined using spectrophotometry 
(Shimadzu, UV160A). Sorbed CPC concentrations were determined by difference. It 
was shown that CPC sorption onto sand reached equilibrium within 14 hours. 
Therefore, a mixing time of 24 hours was used subsequently to ensure the sorption 
equilibrium of CPC onto the sand. A pH value of 9 was experimentally determined to 
be the most appropriate for the CPC coating. 
Second, a CPC sorption isotherm experiment was conducted to determine how 
much CPC could be coated onto sand and the best experimental conditions for 
preparing the CPC-coated sand. Specifically, the preliminary CPC sorption isotherm 
experiment was carried out using duplicate vials containing 1g of Mystic White II 
coarse sand and 170, 102, 34, 23.8, 17, 10.2, 3.4, and 1.7 mg/L CPC solution 
prepared in DI water, with the pH adjusted to 9 using 0.1 N NaOH. The samples were 
then rotated at 25?C for 48 h to reach equilibrium. After equilibration, the samples 
were centrifuged, filtered (Acrodisc 0.2 ?m PTFE filter), and the concentrations of 
CPC in samples were measured. The sorbed concentration was determined by 
difference. 
Finally, naphthalene sorption onto CPC-coated sand was studied to determine 
whether the sorption/desorption process could limit the overall rate of bioremediation 
when naphthalene sorbed to the coated sand. The CPC-coated sand for the 
naphthalene sorption test was prepared in a 1L bottle, using 18 g of Mystic White II 
coarse sand with a CPC solution of 34 mg/L =
4
101
?
?  M (as discussed further below, 
 
 106 
 
it was determined that the critical micelle concentration (CMC) for CPC for the 
conditions used in this study was between 
4
101
?
?  and 
4
103
?
? M) and pH value of 9. 
The mixture was mixed on a horizontal shaker table for 48 hours, at which time a 
sample was taken to determine the final aqueous concentration and, thereby, sorbed 
CPC concentration. The sand sample was then drained and oven dried at 110 ?C for 3 
days. One gram samples of the dried CPC-coated sand were added to 40 ml amber 
vials, followed by addition of naphthalene solutions with the same concentrations as 
in the sorption onto unmodified sand experiment (Section 5.3.5.1). Vials were rotated 
end-over-end for 124 hours, centrifuged (1500 rpm for 10 min), and the aqueous 
phase filtered (Acrodisc 0.2 ?m PTFE filter) and analyzed by spectrophotometry. The 
sorbed naphthalene concentration was determined by difference.  
5.3.5.4 Mixing sands with Amberlite resins 
The third option considered for modifying the sorptive properties of the 
porous media sand involved mixing the porous media sand with a relatively small 
amount of a commercially available Amberlite resin. Amberlite resins have been 
shown to be a good sorption sink based on the partitioning coefficient (Guerin and 
Boyd 1997; Mulder et al. 2000). Initially, XAD-2 (Supelco, Bellefonte, PA, USA) 
was chosen because it can be used to simulate both the strong sorption of organic 
matter and intraparticle diffusion-limited sorption (Mulder et al. 2000). Importantly 
for this project, Guerin and Boyd (1997) demonstrated that reversible sorption of 
naphthalene to XAD-2 resin allowed it to be desorbed and mineralized by two 
microbial strains. Thus, use of XAD-2 was a promising candidate for Scenario #3, 
especially if enhanced desorption could be achieved with the utilization of (bio) 
 
 107 
 
surfactant, which, would in turn, enhance the biodegradation. The physical properties 
of XAD-2 are presented in Table 5.6.  
 
Table 5.6 Physical properties of Amberlite resins XAD-2 and XAD-7 
Name Average pore 
diameter (? ) 
Specific surface 
area (m
2
/g) 
Skeletal density 
(kg/m
3
) 
Porosity  
XAD-2 90 300 640 0.41 
XAD-7 90 450 1240 0.55 
 
Preliminary sorption isotherm experiments were conducted to observe the 
sorption behavior of naphthalene with XAD-2 in the simulated groundwater system. 
Ten naphthalene in MSNS solutions with concentrations of 0.5 mg/L, 1 mg/L, 2 
mg/L, 3 mg/L, 5 mg/L, 10 mg/L, 12.5 mg/L, 17.5 mg/L and 20 mg/L, were prepared. 
First, 5.0 mg autoclaved XAD-2 was placed in each of the 40-ml borosilicate amber 
vials using tweezers. Next, 40 ml of each naphthalene solution was introduced into 
duplicate vials. All vials were then capped and shaken at an average room 
temperature of 24?C for a sufficiently long period (approximately 136 hours) to 
ensure that sorption equilibrium was attained. The analysis procedures are the same 
as those described for the previous sorption isotherm experimental investigations. 
As discussed below, the sorption capacity of XAD-2 was demonstrated to be 
very promising for Scenario #3. However, there was one concern with regard to the 
application of XAD-2: the skeleton density. The density of XAD-2 is significantly 
lower than that of either of the two types of sands used in the ISFC (Table 5.6), which 
at a minimum may lead to a problem of the XAD-2 floating out of the sand and might 
result in a change in the physical properties of simulated aquifer. Therefore, XAD-7 
 
 108 
 
was chosen as the next alternative to evaluate because it had a similar sorption 
capacity as XAD-2, yet more attractive physical properties, in particular a skeletal 
density of 1240 kg/m
3
, which was more similar to the particle densities of the sands 
used in this research. Thus, compared to XAD-2, it was more promising that the 
mixing of XAD-7 resin with sand would not affect the flow cell hydraulics 
significantly. Therefore, further efforts were focused on the investigating 
sorption/desorption of XAD-7.  
The sorption isotherm experiments of XAD-7 were conducted following the 
same procedures as described above for XAD-2. Specifically, two sorption isotherms 
for naphthalene sorption onto XAD-7 were investigated, one in which the XAD-7 
was mixed with fine sand and the other in which the XAD-7 was mixed with the 
coarse sand, which how XAD-7 was eventually incorporated Scenario #3, as 
discussed below. To do this, 30 mg XAD-7 were mixed 270 mg of either fine sand or 
coarse sand, which resulted in the synthetic model soil having an Amber resin content 
of 10% by weight.   
Based on the determined partition coefficient, XAD-7 also has a strong 
affinity for sorbing naphthalene from solution. Thus, it was decided that the XAD-
7/coarse sand mixture would be selected for meeting the desired criteria in creating 
the increased sorptive characteristics. Therefore, further sorption/desorption kinetics 
experiments were performed using XAD-7. The kinetic experiments were carried out 
in completely mixed batch reactors (CMBRs) following the bottle-point rate 
procedure described by Weber and Miller (1988).  
 
 109 
 
The bottle-point rate experiments were performed by preparing 28 identical 
40-ml borosilicate amber vials, each containing 40 ml of naphthalene solution 
(nominal concentration = 5 mg/L) and 30 mg XAD-7. Once set up, the vials were 
shaken on a end-to-end shaker to maintain completely mixed conditions. 
Subsequently, duplicate reactors and a solution blank were removed from the shaker 
at selected intervals (ranging from 1h to 192 h). The samples from these vials were 
then analyzed for the naphthalene aqueous concentration and the data were used to 
obtain serial data for sorbed naphthalene as a function of time by difference. The data 
collected from these bottle-point rate studies were used to evaluate the rate of 
sorption and the time required to reach the equilibrium. 
Scenario #3 represented a system in which the meso-scale desorption rate was 
to be the overall rate limiting process. It was important to investigate the desorption 
rate of the XAD-7/sand mixture to ensure that the slow desorption rate was limiting 
the overall rate of bioremediation. Therefore, a bottle-rate experiment was also 
conducted to evaluate the desorption kinetics. First, twenty identical 40-ml 
borosilicate amber vials were prepared, each containing 40 ml of naphthalene solution 
(C
0
 = 20 mg/L) and 30 mg XAD-7. The vials were then shaken on a rotary shaker for 
48 hours to ensure that the equilibrium was reached. Second, the naphthalene 
concentrations were measured to determine the amount of naphthalene sorbed onto 
XAD-7 in each vial. Third, the naphthalene solutions were decanted, being careful 
not to lose any XAD-7. Fourth, the vials were filled with 40 ml of fresh MSNS, 
which was free of naphthalene. Finally, the vials were shaken on the rotary shaker 
and duplicate vials were removed from the shaker at selected intervals (intervals 
 
 110 
 
ranging 1h to 52 h). Naphthalene aqueous concentrations were analyzed and the data 
were used to obtain serial rate data for desorption as a function of time. 
As discussed in the sorption/desorption literature review, many models have 
been developed to describe sorption/desorption kinetics of with porous media. A 
dual-resistance model proposed by Crittenden et al. (1986), the batch pore surface 
diffusion model (BPSDM) for describing solute transport in a soil system, was 
utilized to obtain the sorption/desorption kinetic parameters. The model incorporates 
the following mechanisms: (1) axial dispersion and diffusions; (2) film transfer 
resistance to mass transport from the mobile to the stationary phase; (3) local sorption 
equilibrium; and (4) surface and pore diffusion as intraparticle mass transport 
mechanisms. Thus, it incorporates film, pore and surface diffusion mass transfer 
coefficients.The BPSDM model was obtained from Dr. D.W. Hand (Michigan 
Technological University, Houghton, MI).  
In Scenario #3 Phase IIB, an engineered perturbation was needed to increase 
the limting desorption rate of naphthalene and, thereby, increase the biodegradation 
rate. Based on a review of the literature, a biosurfactant was chosen for this purpose 
over a synthetic surfactant. Biosurfactants are advantageous compared to synthetic 
compounds for use in soil remediation because biosurfactants are natural compounds 
that will have a low environmental effect. Thus, complete removal after treatment 
may not be necessary. In addition, in situ production may also be possible, e.g., 
(Desai and Banat 1997). To be effective for enhancing the removal of sorbed 
contaminants, a biosurfactant should have high solubilizing properties, should be 
 
 111 
 
water soluble, and should not strongly adsorb to soil. On the basis of these criteria, a 
rhamnolipid biosurfactant was selected for the study. 
In particular, a rhamnolipid biosurfactant was selected because: (1) it is a 
glycolipid, which is the most commonly isolated type of biosurfactant; and (2) 
members of the genus Pseudomonas, which produce various rhamnolipids, are 
common soil microorganisms (Zhang and Miller 1992). The specific rhamnolipid 
selected was the Jeneil Biosurfactant Company (Sankville, WI) product JBR425. The 
two major rhamnolipid components present in the rhamnolipid solution are 
monorhamnolipid (?  -L-rhamnopyranosyl-? -hydroxydecanoyl-? -
hydroxydecanoate) and dirhamnolipid (2-O-? -L-Rhamnopyranosyl-?  -L-
rhamnopyranosyl-? -hydroxydecanoyl-? -hydroxydecanoate). The mono- to di- 
rhamnolipid ratio is of 1:1, with molecular weights of 504 and 650, respectively. The 
critical micelle concentration of this biosurfactant is 0.037 mM (Gu and Chang 2001). 
Due to the time limitations, the investigation of enhanced desorption of 
naphthalene from XAD-7 by biosurfactant was not completed during this work. 
Therefore, that study will be performed by Ms. Eunyoung Hong. 
5.4 Bioremediation Simulation Experiments 
 As discussed in Chapter 4, three different bioremediation scenarios were 
simulated in the ISFC to represent different combinations of natural rate-limiting 
phenomena in subsurface systems. Each scenario was conducted in two phases: Phase 
I, intrinsic biodegradation, and Phase II, engineered bioremediation, with selected 
system perturbations, as discussed further below. However, after performing the 
point-source concentration tracer studies and before inoculating the flow cell and 
 
 112 
 
conducting the bioremediation simulation experiments, several control experiments 
were first done in the ISFC to demonstrate that effects observed in subsequent 
experiments after inoculation were due to microbial activity, not physicochemical 
changes in the system. The control experiments were originally designed to replicate 
the Scenario #1 experiments, but without microbial activity. Although the design of 
the Scenario #1 experiments was altered from a continuous injection of naphthalene 
in MSNS in the control experiments to a pulse of contaminated artificial groundwater 
in the actual Scenario #1 bioremediation simulation experiments that were performed, 
several key pieces of information were still obtained. In particular, these experiments 
were useful for developing sampling techniques, evaluating interlayer dispersivity, 
and demonstrating the effect of the Phase II system perturbation in the absence of 
microbial activity in Scenario #1. In the following subsections, the control 
experiments are described, followed by a description of the general procedures 
followed during the bioremediation simulation experiments and explanation of the 
methodology used for simulation Scenarios #1, 2, and 3. 
5.4.1 Control Experiments 
5.4.1.1 Phase I: Intrinsic Control Experiments 
The control experiment for Scenario #1 was conducted for three purposes: (1) 
to develop and refine the techniques to be used in the full simulated bioremediation 
experiments; (2) to determine the interlayer dispersivity because of the importance of 
understanding processes near interfaces in the subsurface environment; and (3) to 
evaluate the effects of naphthalene mass loss in the ISFC via volatilization, or 
sorption onto the silicone caulking. 
 
 113 
 
In the Phase I control experiments, the ISFC was filled with filter-sterilized 
MSNS before starting. Subsequently, filter-sterilized MSNS with naphthalene (~5 
mg/L) was pumped into the ISFC continuously at a rate of 6 ml/min. The general 
microbial inhibitor, diluted acid-free formaldehyde in PIPES buffer solution, was 
added at a concentration of 0.04% (by weight) to inhibit microbial growth (Brock 
1977; Trevor 1996). The flow rate was monitored and adjusted appropriately 
throughout the experiment to ensure that it did not fluctuate too much either due to 
the Teflon pump tubing wearing out or accidental mistakes in setting up the 
Masterflex
?
 L/S
?
 PTFE-tubing pump. In only one case was it necessary to adjust the 
pump back to 6 ml/min due to a slight slowing down because of worn out tubing. 
During this experiment, and all of the other ISFC experiments, the movement 
of the contaminant plume was followed by monitoring the naphthalene in the pore 
water and effluent. In the case of the control experiments, monitoring was continued 
until a steady state condition was reached with respect to naphthalene plume 
development. The pore water samples were used to obtain breakthrough curve data at 
selected locations. Therefore, it was necessary to develop sampling techniques and 
locations. 
Designing the sampling approach in the ISFC experiments in general was very 
challenging because of the small volume of water present within the flow cell 
combined with the minimal volumetric fluxes moving through the flow cell. Thus,  
withdrawal of significant quantities of water was impossible without significantly 
affecting the flow field (Silliman et al. 1998). In a compromise between minimizing 
flow disturbance and having sufficient sample for analysis, the sampling volume was 
 
 114 
 
selected to be 1ml, which required some adjustment in analytical protocols (e.g., the 
naphthalene analysis using GC, see below). Samples were withdrawn from sampling 
ports with a syringe pump (Harvard Apparatus, Model number 55-2226) using gas-
tight glass syringes attached to the Mininert valve of the ports via a Teflon tubing set 
(1/16" OD).  The pore-water sampling rates were set at 250 ?L/min and 50 ?L/min 
for the high-K layer and low-K layer, respectively. The samples from both influent 
and effluent sampling ports were taken manually, using a gas-tight glass syringe 
(Hamilton Series 1000 Gastight Syringe, Fisher Scientific) at a rate that was as slow 
as possible. These sampling flow rates were chosen carefully to be less than 5% of the 
bulk flow rate in order to be small enough to not interfere with the bulk flow in the 
flow cell. These sampling techniques provided the capability for complete chemical 
analysis of the pore fluid withdrawn from the flow cell, but limited the number of 
samples that could be withdrawn both temporally and spatially based on minimizing 
the impact of the sampling on the flow field, which, in turn, made the arrangement of 
the sampling schedule challenging during ISFC experiments. 
In order to obtain the contaminant breakthrough curve data, the selection of 
sampling times and locations was of primary importance. Samples from the inlet flow 
line, the outlet flow line, and selected pore-water sampling ports were taken at 
intervals of hours early in the ISFC experiments and days at later times. The pore-
water sampling ports selected for monitoring during Phase I control experiment were 
A4, C4, E4, B9, C9 and D9. The sampling ports A4, C4, B9 and C9 were sampled 
more intensely than the others during the early period of each experiment, because the 
plume in the high-K layer developed much faster than that in the low-K layer. On the 
 
 115 
 
other hand, the sampling ports E2 and D9 were monitored more intensely during the 
later period of each experiment because at that point the plume was mainly 
developing in the low-K layer.  The effluent sampling port was sampled at the same 
time interval as the influent sampling port, which was taken every time the pore-water 
samples were withdrawn. 
The interlayer dispersivity was obtained by calibrating the control 
experimental breakthrough curves from D9. The experimental results were simulated 
using RT3D in VMOD and the interlayer dispersivity was adjusted by trial-and-error 
to fit the data of the initial mass pulse from interlayer mass transfer. 
5.4.1.2 Phase IIA: Nutrients Addition Control Experiments 
As part of the control experiments for Scenario #1, a control perturbation 
experiment was also conduced. Numerous studies have been published on the effects 
on pollutant biodegradation rates in response to nutrient supplements. The results 
have been controversial, with some studies reporting they observed stimulated 
degradation with the addition of nitrogen and/or phosphorous (Alexander 1994), 
while others observed that the degradation rate was unaffected by the added nutrients 
or may actually have declined relative to controls (Johnson and Scow 1999; Manilal 
and Alexander 1991). Despite the continuing debate over the effect of nutrient 
supplements for bioremediation in the field, the addition of excess N and/or P has 
been a common remedial strategy. For this reason, in Phase IIA of all the 
experimental scenarios, the first ?engineered? perturbation received was the addition 
of N and P in excess of the stoichiometric requirement. Similar to many field 
observations, this perturbation was not expected to be effective in the experimental 
scenarios, because another factor other than nutrient levels was ultimately most 
 
 116 
 
limiting the biodegradation rate. Therefore, the effects of addition of excess N and/or 
P were investigated in a control experiment, too. To perform this component of the 
control experiment, 6 times the stoichiometric requirement (Equation 5.1) for N and P 
were applied in the simulated groundwater. This control experiment was performed 
once, following the same protocol as described in the Phase I control experiment. 
5.4.1.3 Phase IIB: Engineered Control Experiments 
In the second engineered perturbation experiment of Scenario #1, Phase IIB, 
advection was increased to increase transverse dispersion, being careful not to 
increase advection so much that it started to limit the biodegradation rate by dilution. 
Therefore, the effect of increasing flow rate was examined in the control experiment 
as well. In order to determine the appropriate flow rate that would have a significant 
effect on transverse dispersion, but at the same time be experimentally feasible, a 
simulation was run using RT3D VMOD. Based on the modeling simulation, the flow 
rate was doubled to meet the above criteria. This control experiment with the 
increased flow rate was performed once for 24 hours, following the general 
procedures of Phase I control experiment. 
5.4.2 General Bioremediation Simulation Experimental Procedures 
5.4.2.1 ISFC preparation and Inoculation 
When the ISFC was originally set up before the control experiments, all 
components of the influent feed line ? tubings, connections, and feed bottles ? were 
sterilized by autoclaving. All feed solutions were then filter (Membrane filters, 0.25 
?m, Schleicher & Schuell, Dassel Germany) sterilized during addition to the feed 
bottle. In addition, before the start of bioremediation simulation experiments, the 
 
 117 
 
ISFC, pump, tubing and connection parts were treated with 3 pore volume of the 
formaldehyde solution (0.04% by weight) in PIPES buffer solution to inhibit 
microbial activity (Brock 1977; Trevor 1996). The ISFC was then allowed to sit 
stagnant for 3 days to assure complete microbial inhibition. To confirm complete 
microbial inhibition, the microbial populations in the flow cell effluent were 
estimated using the heterotrophic plate count (HPC) method. Subsequently, the 
system was rinsed with 3 pore volumes of filter sterilized MSNS to prepare for 
inoculation. Solutions, glassware and other materials used in the ISFC bioremediation 
simulation experiments were also sterilized by autoclaving for 20 min.  
After flushing with filter-sterilized MSNS, the ISFC was inoculated with the 
chosen microorganism Uper-1. The 30L (~ 2 pore volumes) of inoculated MSNS 
were then pumped through the ISFC for inoculation. To maximize microbial 
attachment, the ISFC was subsequently allowed to stand for 24 hours before the 
initiation of experiment. Following that, samples were taken from A2 and E2 for HPC 
to determine the biomass concentration in aqueous phase. 
The inoculation of ISFC had to be performed a second time later during the 
simulated bioremediation experiments. Specifically, following the excess addition of 
N and P in the Phase IIA experiments in Scenario #1 and #2, algae growth occurred 
in the high-K layer and the interface between the high-K and low-K layers, which 
affected the ISFC physicochemical and biological conditions. As the result, the ISFC 
was treated using a disinfectant prepared from household bleach and Alconox 
detergent (Barkley and Richardson 1994). The household bleach was diluted 100 
times, containing 525 mg/L of available chlorine, with 0.7% of Alconox detergent 
 
 118 
 
added. One pore volume of prepared disinfectant solution was first pumped through 
to disinfect the ISFC. Then, 3 pore volumes of filter-sterilized MSNS were pumped 
into the ISFC to flush out the residual disinfectant solution. Following that, 3 pore 
volumes of formaldehyde (4%) (Trevor 1996) were pumped through the ISFC to 
ensure complete microbial inhibition. Finally, the ISFC was rinsed with 3 pore 
volumes of filter-sterilized MSNS to prepare for the inoculation. The ISFC was 
inoculated again following the same procedure as described above. 
5.4.2.2 General ISFC Operation 
Although the experimental conditions were different for each bioremediation 
simulation scenario, the general sampling plan and data analysis were the same. To 
quantify the processes controlling bioremediation and the process interactions, the 
nonreactive tracer (Br), naphthalene and dissolved oxygen (DO) concentration were 
monitored within the high-K layer (A2 and B7), within the low-K layer (D1, E2 and 
D7), at the sand layer interface (C2 and C7), and in the influent and effluent. Samples 
of ~ 1 ml were withdrawn from these sampling ports following the same procedures 
as described in the control experiments. The aqueous microbial numbers were also 
quantified using HPC within both layers (A2 and E2) as well between the sand layer 
interface (C2) and in the effluent. Although most of the microbial biomass was 
probably associated with the porous media, changes in the attached biomass should 
be reflected in changes in the microbial biomass in solution. 
A temporal moment data analysis was utilized in this research for interpreting 
the breakthrough curve data from the bioremediation simulation experiments, e.g., 
(Kent et al. 1994). The zeroth moment is defined as the area under the breakthrough 
curve, which was calculated utilizing the linear trapezoidal integration. In this 
 
 119 
 
research, the zeroth moment analysis was performed using the areas under the 
naphthalene and Br breakthrough curves, A
naph
 and A
Br 
, respectively, after different 
horizontal flow distances, at selected sampling ports. The ratio of A
naph
/A
Br
 was 
utilized for comparison between experimental Phase I and Phase II because the effect 
of dispersion on the naphthalene concentration could be eliminated by normalizing to 
A
Br
. In addition, the average pore-water velocity at each port was determined by 
dividing the Br travel distance by the estimated average travel time.  
As described further below, bioremediation Scenarios #1 and #2 were 
simulated in the ISFC. The experiments were broken down into two phases, similar to 
the control experiments. Phase I represented intrinsic biodegradation under natural 
conditions, as described in each scenario. Each scenario was designed such that a 
particular process (e.g., a physical-chemical mass transfer process or biokinetics) was 
the overall rate-limiting process controlling the overall bioremediation rate. The 
results of the Phase I experiments were used to evaluate if the limitations on 
biodegradation under intrinsic conditions were what was expected. Specifically, the 
experimental results from the ISFC were modeled using RT3D in VMOD and the 
independent parameter estimates to confirm that the parameter estimates successfully 
described the system. Then, the parameter estimates were used in the quantitative 
framework to identify the rate-limiting process. The Phase II experiments then 
represented perturbations to the system as part of an engineered bioremediation. The 
successful perturbation was expected to be one selected based on the overall rate-
limiting phenomenon and how the remedial strategy affected it. 
 
 120 
 
5.4.3 Scenario #1: Dispersion-Limited Biodegradation 
The experimental conditions for Scenario #1 were selected to simulate a field 
scenario with relatively fast biokinetics in porous media with low sorption, making 
macro-scale transverse vertical dispersion from the fast-K layer to the slow-K layer 
the overall rate-limiting process.  Therefore, this scenario represents a key macro-
scale interfacial process that can significantly impact in situ bioremediation, i.e., the 
hydraulic mixing at the interface of a dissolved contaminant plume and the "clean" 
groundwater, which is induced by subsurface hydraulic conductivity heterogeneities. 
To experimentally set up this scenario, the two-layer, low sorptive capacity porous 
media and microorganism Uper-1 were used as described in Sections 5.2.6 and 5.1.3, 
respectively. 
5.4.3.1 Phase I: Intrinsic Biodegradation 
Phase I represented intrinsic biodegradation under natural conditions for 
which vertical transverse dispersion was the overall rate-limiting process. The MSNS 
with naphthalene (~ 10 mg/L) and Br (~ 120 mg/L) was pumped into the ISFC for 48 
hours at a flow rate of 3 ml/min to simulate a pulse input of contaminant. This was 
followed by an input of MSNS alone at the same flow rate until the end of the 
experiment. The flow rate used in these experiments was half of that in control 
experiment, because: (1) it reduced the amount of feeding solutions needed to run the 
experiments, and (2) it ensured that the flow rate increase in Phase IIB be significant 
enough to alleviate the transverse dispersion limitation. This experiment was 
performed in triplicate. However, Br was not added in the first experiment. Therefore, 
the Br data obtained from the second experiment was utilized in the moment analysis. 
As described in Section 5.4.2.1, the ISFC had to be disinfected and re-inoculated 
 
 121 
 
following the same procedures as during the initial inoculation. Therefore, to ensure 
the system was under the same conditions, the baseline Phase I experiment was 
performed a third time. 
In order to investigate the naphthalene plume development and distribution in 
the ISFC, a plume contour plot was desired. To obtain the data required for a 
complete plume contour plot, it was necessary to capture the plume contour in both 
layers, which meant that the selection of sampling ports and the sampling schedule 
had to be very accurate. If the sampling timing was too early, there was a possibility 
that the plume in the low-K layer had not developed sufficiently to reach the first 
column of sampling ports. However, if the sampling time was too late, it was possible 
that the plume front in the high-K layer could have exited the ISFC before sampling 
was completed.  With respect to the sampling ports, it was also important to choose 
the appropriate locations. On one hand, sampling as many ports as possible would 
have provided the most complete description of the plume contours. On the other 
hand, sampling many ports was not desirable because it could contribute to changes 
in the plume due to interferences with the simulated groundwater flow. Therefore, 
special attention was paid to the design of the sampling approach for the plume 
contour. Based on practicing and preliminary predictions using model simulations in 
RT3D in VMOD , the following sampling ports- C1, D1, C2, E2, C5, C6, C7, C8, 
A10, C10, B11, and C11-were sampled in addition to those sampled for the 
breakthrough curve analysis at 48 hours, immediately after the pulse input was halted.  
This plume contour investigation was not performed in Phase I (1) because 
that was the first bioremediation simulation experiment performed, and it was not 
 
 122 
 
possible to complete the labor and time consuming plume contour sampling. Instead, 
the contour investigation was practiced in the experiment Phase I (2). However, 
because of the intense sampling schedule, it was very challenging to take samples 
from sufficient sampling ports at one ?snapshot? to produce the desired contour plot. 
Furthermore, the plume contour plot was only a qualitative technique to evaluate the 
experimental results. Therefore, this practice was abandoned in the later 
bioremediation simulation experiments.  
5.4.3.2 Phase IIA: Nutrient Enhanced Biodegradation 
The first perturbation represented the commonly applied engineered remedial 
practice in the field, in which N and P in excess of the stoichiometric requirement 
were added to enhance the biodegradation rate. Therefore, the experiment was 
repeated as described for Phase I except that 2.5 times of the N and P stoichiometric 
requirements for 10 mg/L of naphthalene was used in the input MSNS. The Phase IIA 
experiment was performed only once to minimize the algae problems caused by the 
excess addition of N and P nutrients. 
5.4.3.3 Phase IIB: Engineered Biodegradation 
The second perturbation represented a manipulation to the system as part of an 
engineered bioremediation that was selected based on the identification of rate-
limiting process using the system parameters confirmed in Phase I and the 
quantitative framework in Figure 4.3. As discussed below, it was determined based 
on the framework that dispersion was the overall rate-limiting process. Therefore, the 
flow rate was increased to a level of three times that in Phase I, to increase the 
advection rate and thereby the dispersion (Equation 5.5), especially the interlayer 
transverse dispersion between the high-K layer and low-K layer. It was expected that 
 
 123 
 
the increased transverse dispersion would increase the spreading of the naphthalene in 
the vertical direction and allow more mixing with the DO, thus increasing the 
biodegradation rate and the overall rate of bioremediation. In order to compare the 
results between Phase I and Phase II, the pulse input time was reduced to 16 hours to 
ensure that the naphthalene mass pumped into the ISFC was the same as in Phase I. 
This experiment was conducted twice to ensure the results were reproducible. 
5.4.4 Scenario #2: Biodegradation-Limited Biodegradation 
 Scenario #2 was designed to simulate a field scenario in which slow-
biokinetics would limit the overall bioremediation rate. Experimentally, the same 
porous media and sorptive characteristics were utilized in Scenario #2 as those of 
Scenario #1, i.e. layered heterogeneity and very limited sorption/desorption. In 
addition, the same microbial culture, Uper-1 was used. However, the biodegradation 
rate was inhibited so that the slower biokinetics limited the overall bioremediation 
rate.  
There were two criteria used in the selection of inhibition method: (1) the 
inhibition needed to be significant enough such that the biokinetics were the overall 
rate-limiting process; and (2) the bacteria needed to be able to resume the original 
activity once the inhibitor was removed from the ISFC to facilitate experiment 
replication. There are a wide range of inhibition methods available, as summarized by 
Brock (1977). In general, there are three categories of general inhibitors: (1) chemical 
agents, such as formaldehyde (4%) (Trevor 1996), mercuric bichloride (100 ?g/ml or 
10 ?g/ml), streptomycin sulfate (100 ?g/ml), sodium azide (1000 ?g/ml), novobiocin 
(100 ?g/ml) and hydrochloric acid (0.1N); (2) temperature as an inhibitor; and (3) 
 
 124 
 
water potential as an inhibitor, which includes a variety of solutes such as NaCl, 
alcohols, sugars, urea, etc. In all cases, it is very necessary to assume that the utilized 
inhibitor has no effect on the biological reactions. 
Different options for inhibiting the biodegradation rate were investigated 
using batch respirometry experiments. This work was done by Ms. Eunyoung Hong. 
Ultimately, it was decided to inhibit the microbes by applying a high salt 
concentration (NaCl) (Brock 1977). Thus, this scenario represented bioremediation of 
a saline, hydrocarbon contaminated subsurface. This is actually an important problem 
in many areas, e.g., due to oil exploration and processing (Rhykerd et al. 1995; Riis et 
al. 2003). If salinity levels are sufficiently high to inhibit biodegradation, one option 
for enhancing bioremediation was to add water to reduce the salinity. 
5.4.4.1 Phase I: Intrinsic Biodegradation 
Phase I represented intrinsic biodegradation in a high salinity, hydrocarbon 
contaminated subsurface system for which slow-biokinetics were the overall rate-
limiting process. . The MSNS with naphthalene (~ 10 mg/L), Br (~ 120 mg/L) and 
NaCl (2.5% by weight) was pumped into the ISFC for 48 hours at a flow rate of 3 
ml/min to simulate a pulse input of contaminant, which was followed by an input of 
MSNS with NaCl at the same flow rate until the end of the experiment. This 
experiment was performed in duplicate.  
5.4.4.2 Phase IIA: Nutrient Enhanced Biodegradation 
As in Scenario #1 Phase IIA, N and P in excess of the stoichiometric 
requirement were added in an effort to enhance the biodegradation rate, as commonly 
applied in field-scale remedial actions. Therefore, the experiment was repeated with 
 
 125 
 
the high salinity as in Phase I except that 2.5 times of the N and P stoichiometric 
requirements were used in the input MSNS.  
5.4.4.3 Phase IIB: Engineered Biodegradation 
As discussed below, in Scenario #2, it was determined based on the 
quantitative framework and system parameters that biodegradation was the overall 
rate-limiting process. Therefore, the biokinetics were increased to alleviate the 
limitation and, thus, enhance the overall rate of bioremediation. Experimentally this 
was accomplished by the removal of the high concentration of salt to speed up the 
biodegradation rate. Once the salt was eliminated from the MSNS feeding solution, 
the biomass theoretically resumed to their original activity and the engineered 
bioremediation experiment was conducted. Therefore, the experimental settings for 
Phase IIB were the same as those of Scenario #1 Phase I experiments and the results 
were compared with those from Scenario #1 Phase I. As discussed below, it was 
verified that the results from Scenario #2 Phase IIB were very similar to those 
obtained from the Scenario #1 Phase I bioremediation experiments. As a result, the 
engineered bioremediation simulation of Scenario #2 Phase IIB was only performed 
once and the duplicate experiments in Scenario #1 Phase I were utilized as the 
replicate experiments. 
5.4.5 Scenario #3: Sorption-Limited Biodegradation 
Scenario #3 represented a system in which the meso-scale desorption rate was 
the overall rate limiting process. Therefore, the influent solution, two-layered porous 
media and biokinetics was kept the same as in Scenario #1, but the sorptive 
characteristics of the porous media were altered. Specifically, the experiment was set 
 
 126 
 
up in such a way that a naphthalene-loaded XAD-7/coarse sand mixture was the 
contamination source, such that the aqueous phase concentration in equilibrium with 
the sorbed source was initially approximately 10 mg/L. In order to minimize the 
disturbance to the ISFC, this sorbed contaminant source zone was incorporated by 
installing a new stainless-steel screen in the influent clearwell, which was located 1 
cm away from the influent side wall, leaving ~ 4 cm of space between the two 
stainless-steel screens. The naphthalene-loaded XAD-7/coarse sand mixture was 
packed in the space between the two stainless-steel screens, following the same wet 
packing method as used in the construction of the ISFC. The MSNS was then pumped 
into the ISFC and the plume development was monitored, as described above.  
A moment analysis was to be performed for this scenario as well. However, to 
obtain the non-reactive tracer data, a separate control experiment was to be conducted 
with sodium azide as a microbial inhibitor so that naphthalene desorbing from the 
source zone could be used as the tracer. This was necessary because Br could not be 
loaded onto XAD-7/coarse sand mixture in the same way as naphthalene. Due to time 
constraints, these experiments will be completed by Ms. Eunyoung Hong and are not 
reported on further here. 
 
 127 
 
5.5 Analytical Methods 
5.5.1 Naphthalene Analysis 
5.5.1.1 Gas Chromatography 
5.5.1.1.1 Liquid-Liquid Hexane Extraction 
 
A liquid-liquid hexane extraction method was used to prepare aqueous 
samples for naphthalene analysis by gas chromatography (GC). GC grade hexane 
(Fisher Scientific, 99%) was used as the extraction solvent.  An internal standard was 
used for the GC analysis in order to reduce possible errors in the analysis resulting 
from manual injection of the samples.  Specifically, acenaphthene (Acros Chemicals, 
99.9%) in methanol was used as the internal standard.   
For each analysis, 2 mL of aqueous samples was transferred into 8-mL amber 
vials with screw caps and Teflon-lined septa (Fisher Scientific)  in order to prevent 
possible photolytic decomposition of naphthalene, which was light sensitive (APHA, 
1995).  After sample transfer, 10 ?L of a 500 mg/L acenaphthene in methanol internal 
standard solution was spiked into the sample inserting a Hamilton Gastight syringe 
into the sample vial through the septum in the vial cap.  After addition of internal 
standard, the vial was shaken by hand vigorously for 1 minute to mix the solutions.  
Following mixing, 1 mL of hexane was injected into the vial. During the injection, a 
22-gauge needle was also inserted through the septum of the vial to relieve any 
pressure that may build up due to the hexane injection. The vial was then shaken 
vigorously again for 10 minutes using a mechanical horizontal shaker (Eperback 
Corporation, Ann Arbor, Michigan, U.S.A.).  After shaking, the sample vial sat 
 
 128 
 
stagnant for 10 minutes to allow the two liquid phases to separate.  Subsequently, the 
GC measurement was performed on the extract within 3 days after extraction, as 
described below.  When the samples could not be analyzed immediately, they were 
stored in a refrigerator at 4?C to reduce any chance of volatilization and biological 
degradation that might occur before analysis (APHA et al. 1995) 
5.5.1.1.2 Solid-Liquid Hexane Extraction 
 
 A solid-liquid hexane extraction method was used to analyze naphthalene 
concentration that was sorbed on the solid phase of interest, e.g., XAD-2, XAD-7 or 
sand. As with the liquid-liquid extraction, GC grade hexane (Fisher Scientific, 99%) 
was used as the extraction solvent.   
For solid-phase sample extraction, the wet weight of the solid phase sample 
was measured, followed by transfer into an 8-ml amber vial that had been autoclaved. 
Next, 8 mL of hexane was added to each vial, after which the vials were shaken for 
45 minutes using a rotating shaker. Unlike the operation in liquid-liquid extraction 
method, no internal standard was used. After shaking, the hexane extract was directly 
injected into the GC to measure the naphthalene concentration, as described below. 
As with the aqueous samples, the GC measurement was performed within 3 days after 
extraction unless the samples could not be analyzed immediately, in which case they 
were stored in a refrigerator at 4?C to reduce any chance of volatilization and 
biological degradation that might occur before analysis. 
The moisture content of solid sample was determined at the same time as the 
naphthalene analysis by drying a subsample of the wet solid phase material in a 
 
 129 
 
104?C oven until a constant weight was reached (e.g., 24 hours was found to be 
sufficient for XAD-2).  The weight difference between the wet and dried subsamples 
was used to calculate the moisture content.  In addition, the dry weight of solid 
sample was later used to calculate the sorbed naphthalene concentration.   
5.5.1.1.3 Gas Chromatography Analysis 
 
Analysis of the naphthalene concentration in the hexane extracts was 
performed using a Hewlett Packard (HP) gas chromatograph (GC) Model 6890 with a 
flame ionization detector (FID).  The software Chemstation (version 6.03) was used 
for analysis and integration of the output signal from the GC.  The column installed 
was the HP-5, which is a cross-linked 5% phenyl/methyl siloxane column with a 
length of 30 m, an internal diameter of 0.32 mm, and a film thickness of 0.25 ?m.  
The GC system and operating conditions used in analysis of naphthalene and 
acenaphthene were the same as used by Seagren and Moore (2003), which were 
adapted from a HP method (David et al. 1993) for polynuclear aromatic hydrocarbon 
(PAH) analysis in hexane.  A summary of the GC operating conditions for the method 
is shown in Table 5.7.  The mode of operation used was splitless injection and the 
septum purge time was 0.75 minutes.  With these operating conditions, the retention 
times for naphthalene and acenaphthene were 6.1 and 7.9 minutes respectively. 
Therefore, all runs were stopped at 9 minutes.
 
 130 
 
Table 5.7 Operation parameters for GC analysis 
Carrier Gas Helium 
Septum Purge  60 mL/min 
Column Flow rate 8.2 mL/min 
Makeup Gas (Nitrogen) 45 mL/min 
Flame Oxidant (Air) 400 mL/min 
Gas flow rates 
Flame Combustible (Hydrogen) 40 mL/min 
Initial Temperature 
30 ?C 
Initial Hold Time 2 min 
Temperature Gradient 
25?C 
Hold Time 1 min 
Oven 
Final Temperature 
205 ?C 
Injector 
Temperature 
250 ?C 
Detector 
Temperature 
300 ?C 
 
 131 
 
A sample volume of 1 ?L was manually injected using a 10 ?L Hewlett 
Packard analytical syringe.  Before injection of each sample, the syringe was rinsed 
with the sample 3 times.  After each injection, the syringe was washed with hexane 3 
times.  All injections were made in duplicate for each sample to ensure good 
reproducibility. 
This GC analysis was calibrated with a 1mg/L naphthalene standard sample 
on daily basis. A new calibration curve was prepared with both standard solutions of 
naphthalene and acenaphthene in hexane when the GC conditions were changed due 
to septa change, gas change, or any other possible maintenance.  The calibration 
curves were obtained from the relationship between the peak area from the GC 
response and the concentration of the standard solution.  The standard solutions of 
naphthalene in hexane were obtained by first making a standard 50 mg/L stock 
solution of naphthalene-in-hexane, which was then serially diluted to concentrations 
of 25, 10, 5, 1, and 0.1 mg/L. The standard solutions of acenaphthene were prepared 
in the same manner. These standard solutions were stored in 8-mL amber vials with 
Teflon septa, and kept at 4?C when not in use in order to reduce any volatilization or 
biodegradation that may occur during storage. To produce the standard curves, three 
1-?L injections were made from each standard vial using Hewlett Packard standard 
10-?L syringes. The average, standard deviation, and coefficient of variation of the 
peak areas for the triplicate injections of both naphthalene- and acenaphthene-in-
hexane as a function of concentration were obtained, and calibration curves for peak 
area versus concentration of the standard solutions were produced using an absolute 
 
 132 
 
least square linear regression. The concentrations of samples were then obtained by 
inputting the peak area to the linear regression equation from the calibration curve. 
By using the internal standard in the aqueous samples, it was possible to 
calculate the aqueous naphthalene concentration using measurements that were 
known accurately (e.g., V
is 
, the volume of internal standard; V
sam 
, the volume of 
aqueous sample), or for which the uncertainty could be quantified (e.g., NAPH
hex
 and 
ACEN
hex
 , the concentrations of naphthalene and acenaphthene measured by GC in 
the hexane extract). Thus, the aqueous concentration of naphthalene NAPH
aq
 could be 
calculated with the following equation (Seagren and Moore 2003): 
is
sam
is
hex
hex
aq
ACEN
V
V
ACEN
NAPH
NAPH
?
?
?
?
?
?
?
?
=    (5.10) 
where ACEN
is 
= the known concentration of the internal standard solution, 500 mg/L.  
 A MDL study was performed for the aqueous concentration of naphthalene 
analysis using GC. It was determined the MDL for naphthalene analysis by GC 
method was 0.1 mg/L. 
For the solid-phase samples, the naphthalene concentration determined in the 
hexane extract by GC analysis was multiplied by the volume of hexane, to give the 
mass of naphthalene in the known volume of hexane that was used for extraction. 
Then, the solid phase naphthalene concentration on the solid sample (mg 
naphthalene/dry g solid sample) was determined by dividing the mass of naphthalene 
extracted by the dry weight of solid sample. 
 
 133 
 
5.5.1.2 Spectrofluophotometry 
A spectrofluorophotometer (Shimadzu, RF5301 PC) method was utilized an 
alternative to measure the naphthalene concentration. In the naphthalene analysis, the 
excitation wavelength and emission wavelength utilized were 273 nm and 336 nm, 
respectively.  
2 ml sample was transferred into the cuvette (Shimadzu, cell fluorimeter, 10 
mm, Silica). The volume of 2 ml was experimentally determined to be large enough 
not to interfere with the fluorescein or naphthalene analysis, yet small enough not to 
disturb the flow in the ISFC while taking samples from it. The cuvette was put into 
the holder and the reading of concentration was recorded. After each sample 
measurement, the cuvette was washed with DI water 3 times, followed by washing 
with acetone once. A Vakuwash? cell washer (Bel-art Products, Pequannock, NJ 
USA) was used to accelerate the drying process of cuvette. The naphthalene analysis 
was conducted as soon as the samples were collected to minimize the loss due to 
vaporization.  
Naphthalene intensity readings were converted to naphthalene concentration 
using standard curves. To prepare a standard curve, a naphthalene stock solution of 
20 mg/L was made first, and standards at 15, 12.5, 10, 7.5, 5, 2.5, 1, 0.5, 0.1 mg/L 
were made by serial dilution. A multi-point working curve was utilized in the 
standard curve preparation, which was quantified by a third order of curve equation. 
The detection limit for the aqueous naphthalene analysis was reported to be 0.4 ppb 
(Famisan and Brusseau 2003; Mulder et al. 2000). 
It is possible that intermediates formed during naphthalene biodegradation at 
occurred in the bioremediation simulation experiments might have interfered with the 
 
 134 
 
naphthalene fluorescence measurements. However, on several occasions, naphthalene 
measurements of effluent samples using the spectrofluophotometry method were 
compared to the measurements using the GC and found to be comparable. 
5.5.2 Non Reactive Tracer Analysis 
5.5.2.1 Fluorescein 
The spectrofluorophotometer (Shimadzu, RF5301 PC) method was utilized to 
measure the tracer fluorescein concentration. To measure the fluorescein, the 
excitation wavelength and emission wavelength were set up at 440 nm and 514 nm, 
respectively. The fluorescein samples were analyzed immediately after each tracer 
experiment due to its sensitivity to light. Fluorescein intensity readings were 
converted to fluorescein concentration using standard curves. To prepare a standard 
curve, a fluorescein stock solution of 50 mg/L was made first, and standards at 20, 10, 
5, 2, 0.5, 0.1, 0.05 and 0.01 mg/L were made by serial dilution. The specific operation 
and calibration procedures were followed as described in Section 5.5.1.2. 
5.5.2.2 Bromide (Br) 
5.5.2.2.1 Bromide Probe 
 
Bromide concentration in pore water samples obtained from bioremediation 
simulation experiments was measured using an Orion 94-35 Bromide Electrode 
(Thermo Electron Corporation, Beverly, MA), connected to an ion meter Orion 
Model 520A.  The Orion double junction reference electrode 90-02 was utilized along 
with Orion 94-35. Filling solutions were added to inner chamber and outer chamber 
each time before using the electrode. A 5 M NaNO
3
 solution was prepared as the 
 
 135 
 
ionic strength adjustor (ISA) to ensure all samples having a constant background 
ionic strength. 
Pore water samples taken from ISFC, which were analyzed for naphthalene 
concentration using a spectrofluorophotometer immediately after taking the samples, 
as described above, were subsequently poured from the cuvette back to the amber vial 
for storage in the refrigerator. Then, after the completion of each bioremediation 
simulation experiment, all the pore water samples were analyzed for the Br 
concentration. To perform the analysis, a 1 ml sample was taken using pipette, 
transferred into a 50-ml volumetric flask and then diluted to 50 ml using DI water. 
The diluted sample was next transferred into a 100-ml beaker, and 0.5 ml ISA 
solution added to adjust the ionic strength. The diluted solution was stirred 
thoroughly at a moderate speed. The electrodes were rinsed with DI water, blotted dry 
and placed into the beaker containing the diluted sample. When a stable reading was 
displayed, the mV value was recorded. 
mV readings were converted to Br concentration in the diluted samples using  
standard curves. To make standard curves, a 1000 mg/L KBr standard solution was 
made by dissolving 1.488 g of KBr in 1L of DI water. Standards at 100, 75, 50, 25, 
10, 5, 0.5 mg/L were made by serial dilution. Several standard curves of mV versus 
logBr concentration were prepared and the linear regression was performed to fit 
these data. In Scenario #2, a high salinity was utilized to inhibit the biodegradation. 
Different standard curves were prepared with the employed high salinity in standard 
solutions following the above procedures.  The bromide concentration in pore water 
sample was then calculated by multiply the diluted sample concentration with the 
 
 136 
 
dilution factor. The detection limit for the bromide probe was reported 0.4 ppm from 
the manufacture?s manual. 
5.5.2.2.2 Ion Chromatography 
 
 Bromide samples from tracer studies were originally analyzed using a Dionex 
DX-100 ion chromatograph with a Dionex AS4 column. The samples were filtered 
with 0.2 ?m syringe filter. A solution of 1.2 mM sodium carbonate/2.8 mM sodium 
bicarbonate was employed as the eluent. The concentration of Br was determined 
against the standard curve prepared with KBr (Fisher Scientific) in DI water. The 
standard solutions were made using the technique as described in the above section.  
5.5.3 CPC Analysis 
In the surfactant-coated sand study, the CPC aqueous samples were analyzed 
using UV-vis Spectrophotometry. Specifically, a spectrophotometer (Shimadzu, 
UV160U) was used at the wavelength of 258 nm with 2 ml aqueous samples for 
analysis. The CPC samples were diluted appropriately to ensure the diluted 
concentration fall into the concentration range of standard solutions before the 
analysis. The absorbance readings were converted to CPC concentrations in the 
diluted samples using standard curves. To make standard curves, a CPC standard 
stock solution of 0.2 mM was made. Standards at 0.1, 0.05 and 0.01 mM were made 
by serial dilution. The linear regression was performed to fit the standard curve. 
5.5.4 Chemical Oxygen Demand (COD) Test 
In the biodegradation kinetic study, the initial and final substrate 
concentration and biomass concentration were express as COD. Therefore, COD tests 
 
 137 
 
were conducted before the respirometry experiments to determine the ratio of 
0
0
X
S
. 
The dichromate reactor digestion method was used, with Hach?s COD digestion 
regent vials (Sample concentration of 0~1500 mg/L).  The vials contain the pre-
measured reagents, including catalysts and chloride compensator. The COD reactor 
was turned on in advance to preheat to 150 ?C.  Two ml of inoculum sample was 
pipetted into vials. The outside of COD vials was rinsed with deionized water and 
wiped clean with a paper towel. A blank was prepared by substituting 2.00 ml 
deionized water for the sample. The vials were then incubated in the reactor for 2 
hours to assure complete digestion. The reactor was turned off and the vials were 
cooled. The measurement was made either with the spectrophotometer at 620 nm or 
by titration. 
COD calibration curves were prepared by digestion of dilutions from a 1000-
mg/L stock solution of potassium acid phthalate (KHP). The stock solution was made 
by diluting 850 mg of dried (105?C, overnight) KHP in 1000 mL deionized water. 
COD analyses were performed following procedures from manufacture instructions 
(Hach Chemical Co., Loveland, CO). The selected COD vials are accurate over a 
range of 0-1500 mg COD/L with a typical method detection limit of 12 mg COD/L 
provided by the supplier?s manual. 
5.5.5 Dissolved Oxygen 
Dissolved oxygen (DO) was determined utilizing an oxygen microelectrode 
(Microelectrodes, Inc., Bedford, NH USA), connected to an OM-4 oxygen meter. The 
measurement was taken during the bioremediation simulation experiments by 
 
 138 
 
opening the Minert valve attached on the sampling port and inserting the tubing 
connected to the microelectrode flow cell into the outlet of the valve. About 0.1 ml 
pore water samples were then withdrawn into the flow-through chamber using a 1mL 
syringe. The oxygen percentage was recorded by the OM-4 meter and later converted 
into mg/L using the following equation: 
 
32000)100/%()760/)760(()414.22/( ????= rpaS  (5.11) 
where S = solubility of gas in mg per liter; a = absorption coefficient of gas at 
temperature; p = vapor pressure of water at temperature; and r% = actually reading in 
percent of oxygen. 
The microelectrode was calibrated by a two-point calibrating using ambient 
air as well as in a sample with zero DO. The calibration with ambient air was 
performed whenever the microelectrode was used during the experiments. The 
calibration with a sample of zero DO was carried out once a week. The sample with 
zero DO was prepared by adding excess sodium sulfite, Na
2
SO
3
, and a trace of cobalt 
chloride, CoCl
2
, to bring DO to zero (APHA et al. 1995). The relative accuracy of 
microelectrode was reported ?0.04 mg/L at 24?C by the manufacture.  
5.5.6 Heterotrophic Plate Count (HPC) 
During bioremediation experiments, the microbial populations in the porous 
media were estimated using the heterotrophic plate count(s) (HPC) to monitor the 
microbial number in the pore water, which, in turn, indicated the microbial biomass 
associated with porous media. The experimental procedures used in this method was 
adapted from Standard Method 9215 (APHA et al. 1995).   
 
 139 
 
Petri dishes for the HPC were prepared using sterilized disposable petri dishes 
(VWR, No. 25384-070) and commercially available R2A agar (Difco, No. DF1826-
17). Nine grams of R2A agar were added to 500 mL of deionized water and the 
solution was mixed and heated for complete dissolution.  The R2A medium was 
autoclaved for 15 minutes at 121?C and cooled to about 45?C to 50?C for the proper 
amount of surface moisture and easy handling.  Then, the agar medium was 
aseptically poured into the bottom of the sterile plates.  Once the agar solidified, the 
plates were placed in a sealed bag and stored at 4?C in the refrigerator. The plates 
were predried in an incubator at 30?C for 24 hours before use to provide sufficient 
water loss from the surface of the agar.  Throughout the plate preparation procedure 
and all following steps, aseptic techniques were used, and all of the equipment used 
(e.g., test tubes with caps, dilution bottles, pipette tips, and filter papers) were 
autoclaved before use. 
Dilution water for diluting the aqueous samples was made with phosphate and 
magnesium chloride stock solutions (APHA et al. 1995).  The stock phosphate buffer 
solution was made by dissolving 34.0 g of KH
2
PO
4
 in 500 mL deionized water, 
adjusting the pH to 7.2 ? 0.5 using 1 N NaOH, and diluting to 1 L with deionized 
water. The MgCl
2
 stock solution was prepared by dissolving 81.1 g of MgCl
2
?6H
2
O in 
deionized water and diluting to 1 L.  The final dilution water was made by adding 
1.25 mL stock phosphate solution and 5.0 mL MgCl
2
 stock solution and diluting to 1 
L distilled water.   
One mL of aqueous sample and 9 mL of dilution water were aseptically 
placed into a sterile dilution test tube and shaken for 30 seconds vigorously using a 
 
 140 
 
mini vortexer (VWR Scientific Products), establishing a 10
-1
 dilution. The sample 
solution was further diluted by transferring 0.1 mL of sample from the first test tube 
into a second test tube containing 9.9 mL of dilution water, to produce a 10
-3
 solution. 
The same procedures were followed to obtain a series of test tubes with a range of 
sample dilutions.  Subsequently, taking 1 mL or 0.1 mL of each suspension and 
transferring it to the agar plate established the desired dilution, or a one order of 
magnitude lower dilution, respectively.  For example, if 1 mL and 0.1 mL of solution 
from the 10
-3
 dilution test tube are introduced to plates, the dilutions are 10
-3
 and 10
-4
, 
respectively.  The series of dilution steps performed is summarized in Figure 5.7. 
After aseptically pipetting 1 mL or 0.1 mL of solution from each test tube 
onto the agar plates, the solution was distributed over the agar surface using a flame-
sterilized bent glass rod while rotating the plate. The R2A agar plates for each 
dilution were prepared in duplicate.  Subsequently, the edges of R2A agar plates were 
wrapped with Parafilm, and the plates were inverted and incubated at ~30?C for 48 
hours.  After incubation, plates having 30 to 300 colonies were considered in 
determining the plate count (APHA et al., 1995). All colonies in plates with 
appropriate range were counted.  When counting was delayed, the plates were 
temporarily stored at 4?C.  If the number of colonies per plate far exceeded 300, the 
result was recorded as "too numerous to count" (TNTC).  The bacterial count per 
milliliter (colony-forming units (CFU)/mL) was computed by following equation: 
 
 
 
mLdishinsampleofvolumeactual
countedcolonies
mLCFU
,
/ =
(5.12)
 
 141 
 
 
 
 
 
 
 
 
 
Figure 5.7 Sample dilution steps used for microbial plate counts. 
 
0.1 mL
Sample 9 mL 
dilution 
water
9.9 mL 
dilution 
water
9.9 mL 
dilution 
water
1 mL 
1 mL 1 mL 1 mL 
0.1 mL 0.1 mL
0.1 mL
0.1 mL
10
-1
 10
-2
10
-3
10
-4
10
-5
 10
-6
10
-1
 10
-3
10
-5
 
 
 142 
 
Chapter 6 Results and Discussion 
 This chapter presents and discusses the results of the integrated modeling and 
experimental evaluation of the research hypothesis and quantitative framework. First, 
the results of parameter estimation experiments are presented. Next, the results of the 
control experiments are described and analyzed. Finally, the results from the Scenario 
#1 and Scenario #2 bioremediation simulation experiments are reported and 
evaluated. For each scenario, the Phase I data and parameter estimates are used to 
demonstrate the application of the quantitative framework to predict the rate-limiting 
process. Then a qualitative and quantitative comparison of the Phase I and Phase II 
data is used to investigate whether the appropriate perturbation to the system 
predicted based on the quantitative framework had the expected effect. 
6.1 Parameter Estimation Experiments 
6.1.1 Hydraulic Conductivity 
The hydraulic conductivities determined for the Mystic White II Sand and 
Filtersil Sand using the constant head method and Equation 5.2 are given in Table 
6.1. The order of magnitude of these K-values is reasonable based on representative 
values for different sands (Domenico and Schwartz 1998), which range from 9?10
-7
- 
6?10
-3
 m/s for coarse sand to 2?10
-7
- 2?10
-4
 m/s for fine sand. Furthermore, the 
hydraulic conductivity in the high-K layer (Mystic White II) sand was approximately 
7 times larger than in the low-K layer (Filtersil) sand, which was close to the goal of a 
factor 10 times difference.
 
 143 
 
Table 6.1 Physical properties for Mystic White II and Filtersil sands 
 
Properties 
Sand 
d
10
 (mm)
1
 K  (m/s)
2
 
b
? (kg/m
3
) 
P
? (kg/m
3
) ?  
Mystic White II 0.8 
0.0040 
? 0.0001 
1415 2655 0.47 
Filtersil 0.15 
0.00065  
? 0.00008 
1813 2788 0.35 
1 
Source: manufacture, 
2 
Average? standard deviation for duplicate measurements. 
6.1.2 Porous Media Density and Porosity 
 The results of the bulk density,
b
? , and particle density, 
p
? , are summarized 
in Table 6.1. The particle density for Mystic White II sand reported by the supplier is 
2650 kg/m
3
, which is essentially the same as determined in this work and indicates 
that the techniques utilized in this research were able to accurately estimate the 
particle densities.  The bulk density results presented were obtained during the 
packing of the ISFC. The order of magnitude of the bulk densities is reasonable 
compared to values reported in the literature for sands of a similar size (Pearce et al. 
1994; Whelan et al. 1994). The porosity data that were obtained using the particle and 
bulk density values in Equation 5.3 are also presented in Table 6.1. These values fall 
into the reported porosity range of 0.31-0.46 for coarse sand and 0.26-0.53 for fine 
sand (Davis 1969; Schwartz and Zhang 2002), which further verifies that the 
techniques used to determine the bulk and particle densities were appropriate. 
 
 144 
 
6.1.3 Dispersivity 
As discussed in Section 5.3.3, the tracer study was conducted first with Br and 
then changed to fluorescein because of analytical difficulties. However, due to the 
sensitivity of fluorescein to light, there were some concerns as to whether the 
fluorescein would behave conservatively in the ISFC. Therefore, the breakthrough 
from a fluorescein tracer experiment was initially compared with that from a Br tracer 
experiment (Figure 6.1). Based on this comparison, it was confirmed that the 
techniques used in these tracer studies ensured that fluorescein was conservative. 
The point-source tracer (fluorescein) experimental results are illustrated in 
Figures 6.2 and 6.3 for the high-K layer and low-K layer, respectively. The absolute 
least-square regressions for each concentration breakthrough curve are also included, 
which were obtained by using the FORTRAN program, ?trafit3d?, with the best fit 
estimates for the
x
v and 
x
D values. The optimization routines always converged to the 
same solution from different initial 
x
v  and
x
D  guesses, indicating that the obtained 
parameters were unique. A visual appraisal of the experimental data and the 
regression curves reveals a close match and indicates that the test data could be 
closely approximated by the FORTRAN program. The best-fit estimates for the 
parameter 
x
v  and 
x
D  obtained using the non-linear regression are listed in Table 6.2, 
along with the values for 
x
? , which were calculated by ignoring the molecular 
diffusion and assuming 
x
x
x
v
D
=?
. The average longitudinal dispersivities calculated 
from the replicate experiments and used in RT3D and VMOD are presented in Table 
6.3. 
 
 145 
 
 
012
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
 
 
 Br
 Fluorescein
C/
C
'
Time (hr)
 
Figure 6.1 Comparison between breakthrough curves at sampling port B3 for the Br 
and fluorescein tracer experiments.
 
 146 
 
0123
0.0
0.2
0.4
0.6
0.8
1.0 (a)
 Experimental data
 Modeling fit data
 
 
C/
C'
Time (hours)
 
0123
0.0
0.2
0.4
0.6
0.8
1.0
(b)
 Experimental data
 Modeling fit data
 
 
C/
C'
Time (hours)
 
012
0.0
0.2
0.4
0.6
0.8
1.0 (c)
 Experimental data
 Modeling fit data
 
 
C/
C'
Time (hours)
 
Figure 6.2 Breakthrough curves from continuous point source fluorescein 
experimental data along with the model fitting data in the high-K layer. Graphs (a), 
(b) and (c) are data from triplicate experiments. 
 
 147 
 
0246810
0.0
0.2
0.4
0.6
0.8
1.0 (a)
 Experimental data
 Modeling fit data
 
 
C/
C
'
Time (hours)
 
 
0246810
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(b)
 Experimental data
 Modeling fit data
 
 
C/
C
'
Time (hours)
 
 
Figure 6.3 Breakthrough curves from continuous point source fluorescein 
experimental data along with the model fitting data in the low-K layer. Graphs (a), 
(b), (c) and (d) are data from replicate experiments. 
 
 
 148 
 
0246810
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(c)
 Experimental data
 Modeling fit data
 
 
C/
C
'
Time (hours)
 
 
 
0246810
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(d)
 
 
 Experimental data
 Model fit data
C/
C
'
Time (hours)
 
 
Figure 6.3 Continued 
 
 149 
 
Table 6.2 FORTRAN ?trafit3d? calibration results for the longitudinal dispersion 
 Velocity (cm/hr)
1
 
x
D  (cm
2
/hr)
1
 
x
?  (cm)
2
 
High-K (a) 43.23 9.34 0. 22 
High-K (b) 39.69 2.72 0. 07 
High-K (c) 43.58 7.53 0. 17 
Low-K (a) 3.23 0.23 0. 07 
Low-K (b) 3.26 0.18 0. 06 
Low-K (c) 3.05 0.11 0. 04 
Low-K (d) 3.04 0.26 0.08 
1
 Fitted value. 
2
 Calculated value. 
 
Table 6.3 Summary of non-reactive tracer results 
 
x
?  (m)
 1
 
z
?  (m)
 1
 
x
? /
z
?  
High-K Layer (0.8mm) 0.0015? 0.00070 7.6E-5? 9.7E-6 20 
Low-K Layer (0.15mm) 0.0006? 0.0002 0.002? 0.001 0.3 
1 
Average ?  standard deviation for three measurements in the high-K layer and four 
measurements in the low-K layer. 
 
Once 
x
v  and
x
D were estimated, the assumption that the experimental flow 
rates were large enough for
x
D  to be dominated by mechanical dispersion was 
confirmed in the two layers. First, based on Equation 5.7 and assuming an effective 
diffusion coefficient for fluorescein of 0.0104 cm
2
/hr (Seagren 1994), the relative 
error of over estimation introduced by ignoring the molecular diffusion effect was 
approximately 2.9 % in the high-K layer and 1.8 % for the low-K layer. Second, 
 
 150 
 
based on Rumer?s (1962) study 
x
D  is dominated by mechanical dispersion for 
Reynolds number 
3
10Re
?
> (where 
?
50
Re
dv
x
= , with 
50
d =50% grain size, and 
? =kinematic viscosity). Using? = sm /10003.1
26?
?  at 20?C (Metcalf & Eddy Inc. 
1991), 
50
d =1.2mm and 
x
v = sm /102.1
4?
? (Table 6.2), and 
50
d =0.3mm 
and
x
v = sm /1074.8
6?
? (Table 6.2), for Mystic White II sand in high-K layer and 
Filtersil sand in low-K layer, respectively, the Re numbers equals 0.14 for Mystic 
White II sand and 
3
106.2
?
? for the Filtersil sand. Both of these values are greater 
than 10
-3
 and also indicate that the approximation 
xxx
vD ?= is appropriate. 
The vertical transverse dispersion coefficient,
z
D , was calculated using 
Equation 5.4. The tracer injection concentration,
0
C , tracer injection rate, Q, and the 
steady-state tracer concentration, 'C , used in the equation are summarized in Table 
6.4. The steady-state tracer concentration at the sampling location, 'C , was calculated 
by averaging the sample concentrations after they remain relatively constant (See 
Figure 6.2 and 6.3). When there was a question whether or not a sample concentration 
should be included in average, the Q test (Fritz and Schenk 1979), which uses the 
range to determine if a questionable result should be rejected for small sample 
numbers (n=3-10), was used with a 95% confidence to decide whether the value 
should be used in the average or not. The final value required for Equation 5.4, the 
distance between the tip of the injection needle and the tip of the sampling needle, x, 
was estimated to be 20.2 cm. The obtained 
z
D and 
z
? values are presented in Table 
6.4 as well. 
z
? was calculated by assuming the effect of molecular diffusion was 
 
 151 
 
Table 6.4 Vertical transverse dispersivities calculation summarization table 
Tracer 
Fluorescein 
Injection flow 
rate, Q 
(ml/min)
 1
 
Injection 
concentration, 
0
C  
(mg/L)
 1
 
Steady-state 
concentration, 'C  
(mg/L)
 2
 
Velocity,
x
v  
(cm/hr)
3
 
z
D  
(cm
2
/hr)
 2
 
z
?  
(cm)
 2
 
High-K (a) 0.05 10 
 
0.32 43.23 0.3696 
 
0.0086 
High-K (b) 0.05 10 
 
0.35 39.69 0.3400 
 
0.0076 
High-K (c) 0.05 10 
 
0.41 43.58 0.2933 
 
0.0067 
Low-K (a) 0.05 10 
 
0.09 3.23 1.2877 
 
0.4 
 
Low-K (b) 0.05 14 
 
0.63 
 
3.26 0.2648 
 
0.08 
 
Low-K (c) 0.05 22 
 
0.72 
 
3.05 0.3657 
 
0.12 
 
Low-K (d) 0.05 22 
 
0.48 
 
3.04 0.5475 
 
0.18 
 
1
 Experimentally determined, 
2
 Calculated value.
 
 152 
 
negligible, so 
x
z
z
v
D
=? . Using Equation 5.5, and applying the effective molecular 
diffusion coefficient for fluorescein in porous media, *D =0.0104 cm
2
/hr (Seagren 
1994), the relative error of over estimation introduced by ignoring the molecular 
diffusion effect was approximately 8% in the high-K layer and less than 2% for the 
low-K layer. The average vertical transverse dispersivities based on the replicate 
experiment, and later used in the RT3D/VMOD model predications, are summarized 
in Table 6.3. 
Values of dispersivity generally appear to be dependent on the techniques 
utilized in testing or observation (Gelhar et al. 1992). The dispersivity values have 
been determined in a number of studies, e.g., (Gelhar et al. 1992; Pearce et al. 1994; 
Robbins 1989; Seagren et al. 1994; Szecsody et al. 1994; Voudrias and Yeh 1994). 
These studies provided a range of longitudinal dispersivity values from 10
-4
 m to 10
-3
 
m. A rule of thumb is that 
x
? should be of the same order of magnitude as the mean 
grain size (Rumer 1962). This rule holds for the experimental values obtained in this 
research, as shown in Table 6.3. The values of transverse dispersivities from the 
above literature studies ranged from 10
-5
 m to 10
-4
 m, although most were between 
1?10
-5
and 8?10
-5
m. The transverse dispersivity in the high-K layer is reasonable 
compared to the literature values. However, the dispersivity value in the low-K layer 
is relatively high in this study, because 
x
? is generally one order of magnitude higher 
than
z
? , and in this layer
z
x
?
?
<1. A similar observation was made for the fine silica 
sand by Szecsody et al., (1994), who obtained an
z
x
?
?
=3. It is speculated that the 
 
 153 
 
loosely packed fine sand in the low-K layer in the vertical direction may contribute to 
the high 
Z
? value. Other references who packed sand tanks in a similar manner, e.g., 
(Pearce et al. 1994; Voudrias and Yeh 1994; Whelan et al. 1994), also had relatively 
high
z
? value, although not as high as found in this work. Nonetheless, these 
experimental values are reproducible and describe the ISFC simulated aquifer well, as 
demonstrated in the Section 6.2 Bioremediation Simulation Experiments. Therefore, 
they were utilized in the quantitative framework to define the rate-limiting process for 
the designed bioremediation simulation experiments. 
6.1.4 Biodegradation Kinetics 
 As described in Section 5.4.3, careful manipulation of the batch respirometry 
experimental conditions, such as initial 
0
0
X
S
ratio, was necessary in order to reduce 
the correlations between Monod parameters and allow for the estimation of unique 
parameters with the lowest degree of uncertainty. The initial substrate concentration 
S
0
, initial biomass concentration X
0
, and 
0
0
X
S
 are listed in Table 6.5 for the three 
kinetics experiments used in the data analysis of this work. These three experimental 
runs utilized three different inocula as follows: (1) an inoculum from the Uper-1 
primary culture, (2) an inoculum aseptically collected from the ISFC effluent during 
the Scenario #1 Phase I experiment, and (3) an inoculum from the Uper-1 primary 
 
 154 
 
Table 6.5 Biokinetic parameters for respirometry experiments 
Cell Sources Primary culture 
in MSNS  
(1) 
Scenario #1 
Phase I effluent 
(2) 
Primary culture in 
MSNS with high 
salinity (3) 
Initial substrate, S
0
 
1
 
(mg/L as COD) 
30 30 30 
Initial biomass, X
0
 
2
 
(mg/L as COD) 
0.6 0.5 320 
0
0
X
S
 
50 60 0.09 
Yield Coefficient, Y 
 (mg biomass / mg 
naphthalene) 
3
 
1.12 0.34 1.27 
max
?  (/hr), initial estimate 
0.050 0.050 0.004 
max
?  (1/hr), best-fit 
0.015 0.060 0.001 
S
K (mg/L), initial estimate 
4.0 5.3 2.0 
S
K (mg/L), best-fit 
5.0 8.0 1.7 
1 
Based on measured naphthalene concentration and 3 mg COD/mg naphthalene. 
2 
Measured COD. 
3 
Calculated using Equation 5.9.
 
 155 
 
culture in MSNS with high salinity to represent the slow biokinetics in Scenario #2 
Phase I. The initial 
0
0
X
S
ratios for experimental runs (1) and (2) were set up to be 50 
and 60, respectively, to obtain the ?intrinsic? kinetics, which represented the 
maximum capabilities of the Uper-1 primary culture and the culture in the ISFC 
effluent. For experimental run (3), the 
0
0
X
S
ratio of 0.09 was achieved, indicating 
the biokinetics obtained were neither intrinsic nor extant (Grady Jr. et al. 1996). 
Unfortunately, this experiment, which was performed in MSNS with high salinity 
(5% w/v as used in Scenario #2 Phase I), was the only data set available for 
quantifying the effect of high salinity on the biokinetics of Uper-1. Respirometry 
experiments using an inoculum collected from the ISFC effluent during Scenario #2 
Phase I, which had the high salinity, were conducted under intrinsic conditions as 
well, just as in Scenario #1 Phase II. However, although there was a long lag phase, 
apparently due to the high salinity, the culture always eventually exhibited biokinetics 
comparable to the uninhibited cells in the Scenario #1 Phase I effluent. These 
biokinetics were inconsistent with Uper-1?s performance in the ISFC with high 
salinity. Therefore, the slow biokinetics were estimated using the inoculum from the 
Uper-1 primary culture inhibited with high salinity experiment (Run 3). Finally, a 
respirometry experiment using an inoculum collected from Scenario #1 Phase II was 
also conducted once, but resulted in an invalid data set based on the technique for 
initial parameter estimates described below. Therefore, only the above mentioned 
three experimental results are reported. 
 
 156 
 
Before performing the nonlinear parameter estimation using the FORTRAN 
program ?nvolma?, some initial work was required. First, before the raw data could 
be used for parameter estimation, some clean-up of the data was required. 
Specifically, in the plot of cumulative oxygen uptake, there were instances in which 
lags existed before initiation of oxygen uptake. The points recorded during the lag 
phase were removed from the data set because the models employed do not account 
for a lag phase.  
Second, initial estimates of the yield coefficient, Y, maximum specific growth 
rate,
m
? , and half-saturation constant, 
s
K , were required. The initial estimates of the 
yield coefficients were calculated from the oxygen consumption at the plateau of the 
accumulated oxygen uptake profile (e.g., Figure 6.5) by using Equation 5.9 and are 
summarized in Table 6.5 for the respirometry experiments with the three different 
sources of initial inoculum. Based on the work of Brown, et al. (1990), these initial 
estimates of the yield coefficients were just used as the final parameter values.  
 Although the determination of the final 
max
? and 
S
K values were 
accomplished by nonlinear parameter estimation using the FORTRAN routine, good 
initial guesses were necessary to make the search progress more efficient and to 
perform the search in the appropriate range in order to obtain the best fit. Thus, it was 
desirable to have a reliable technique to make initial estimates, which could also be 
used to examine whether the data sets were satisfactory for parameter estimation. 
Therefore, the techniques utilized by Brown et al. (1990) were employed in this 
study, which proved to be very valuable during initial parameter estimation, as 
described below. 
 
 157 
 
 In the initial estimate technique employed, it is assumed that the biomass 
concentration in a batch reactor increases at a rate that is first order with respect to 
biomass concentration during growth conditions in which endogenous metabolism 
and biomass decay may be neglected, which holds for this respirometry study. Thus, a 
plot of dtdX / vs. X will have a constant positive slope equal to the specific growth 
rate, ? , until something becomes limiting in a reactor (Kono and Asai 1969). In 
addition, for a batch reactor, dtdX / is proportional to the rate of oxygen 
uptake, dtdO
u
/ , whereas X is proportional to the cumulative oxygen uptake, 
u
O (Dand et al. 1989). Thus, a plot of dtdO
u
/vs. 
u
O  will exhibit the same 
characteristics as a plot of dtdX / vs. X, with a slope equal to? . Therefore, a routine 
was created in a Microsoft Excel spreadsheet that used the cumulative oxygen uptake 
data to calculate the rate of oxygen uptake as a function of time, tO
u
?? /, for 
successive data points. Then those data were used to make a plot a four point running 
average of dtdO
u
/ versus the cumulative oxygen uptake, 
u
O (Brown et al. 1990). A 
typical plot for a valid data set (Run 2) is shown in Figure 6.4. 
 
 158 
 
-2 0 2 4 6 8 101214161820
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
 
 
dO
u/dt[(
m
g
/L)/h]
Oxygen uptate (mg/L)
 
Figure 6.4 Typical plot of the rate of oxygen uptake vs. the cumulative oxygen uptake 
for a valid data set (Run 2) (an inoculum collected from the ISFC effluent during 
Scenario #1 Phase I).
 
 159 
 
 In the Monod kinetics, the values of ? can then be obtained from the 
calculated oxygen update rates tO
u
?? / using (Brown et al. 1990): 
tup
tu
OYYYX
dtdO
)()1)(/(
)/(
0
+??
=?    (6.1) 
The calculated ? values were then plotted as a function of oxygen uptake. Based on 
that point, the initial estimate of 
max
? was easily obtained as the maximum observed 
value of ? . Because 
S
K  is the substrate concentration corresponding to ? = 
0.5
max
? , an initial estimate can obtained from the equation (Brown et al. 1990): 
YY
O
SK
p
uh
s
??
?=
1
0
    (6.2) 
where 
uh
O is the oxygen uptake associated with ? = 0.5
max
? . The initial estimates for 
max
? and 
S
K made based on this approach are summarized in Table 6.5, along with 
the final parameter estimation results. These biokinetic parameters were obtained by 
fitting a single oxygen cumulative curve, although 2 or 3 vessels were set up for each 
run. The model results using the best fit parameters are demonstrated in Figure 6.5, 
along with the laboratory cumulative oxygen uptake profile. 
 
 160 
 
0 25 50 75 100
0
10
20
30
(b)
 
 
 experimental data
 model simulation
Accum
u
la
ted
 O
x
y
g
e
n
 Co
ncen
tratio
n (m
g
/
L)
Time (hours)
 
0 50 100
0
2
4
6
8
10
12
(c)
 
 
 experimental data
 model simulation
Accu
m
u
l
a
ted
 O
x
y
g
e
n
 Co
ncen
t
r
a
t
io
n (m
g/L
)
Time (hours)
 
Figure 6.5 Cumulative oxygen uptake profiles during respirometry batch growth with 
Uper-1 from different cell sources: (a) an inoculum from the Uper-1 primary culture, 
(b) an inoculum aseptically collected from the ISFC effluent during the Scenario #1 
Phase I experiment, and (c) an inoculum from the Uper-1 primary culture in MSNS 
with high salinity. 
 
 
 161 
 
 
0 5 10 15 20
0
1
2
3
4
5
6
7
8
(a)
 
 
 experimental data
 model simulation
Accu
m
u
l
a
te
d Ox
y
g
en
 Con
c
e
n
trati
o
n
 (m
g/L
)
Time (hours)
 
Figure 6.5 Continued 
 
 
 
 162 
 
The calculated yield coefficients in Table 6.5 are consistent with yield values 
reported in the literature for naphthalene, which range from 0.25 -1.2 (Ghoshal and 
Luthy 1998; Volkering et al. 1993; Woozinski and Johnson 1968). However, it was 
surprising that the yield coefficient with the high salinity inhibition (1.26) was higher 
than the one obtained without inhibition (0.34). This might be due to the source of 
inoculum, because the Y values obtained using the primary culture (with or without 
inhibition) are similar to each other, whereas the value calculated using inoculum 
from Scenario #1 Phase I effluent is much lower. Alternatively, it could be that the 
nutrient levels were reduced in the effluent sample, which reduced the yield. The 
final
max
? estimation of 0.06 hr
-1
 (1.44 d
-1
) obtained using the inoculum collected from 
Scenario #1 Phase I effluent is comparable to the literature values of 0.7-1.2 d
-1
 by 
Goshal et al. (1996), 1.6 d
-1
 by Goshal and Luthy (1998)and 7.9 d
-1
 by Volkering et 
al. (1993). The 
max
? value obtained from primary culture with high salinity for 
Scenario #2 is fairly reasonable considering the inhibitory effect of the high salinity, 
which is expected to slow down the specific growth rate. Interestingly, the intrinsic 
biokinetics, i.e., the maximum capability of the isolated culture growing aerobically 
on naphthalene measured with the Uper-1 primary culture was estimated to be only 
0.015 hr
-1
 (0.36 d
-1
), which is lower than the value obtained from the Scenario #1 
Phase I effluent experiments. Nonetheless, this value was not utilized in the 
evaluation of the quantitative framework. The 
S
K values are in the same order of 
magnitude as reported in the literature for naphthalene, which range from 0.02-3.4 
mg/L (Ghoshal and Luthy 1998; Ghoshal et al. 1996; Volkering et al. 1993). 
Therefore, the kinetic coefficients appear reasonable and the values obtained with the 
 
 163 
 
Scenario #1 Phase I effluent and the primary culture in MSNS with high salinity were 
used in the dimensionless number analysis. 
6.1.5 Sorption/Desorption Equilibrium and Kinetics 
6.1.5.1 Unmodified Sand Sorption 
The naphthalene sorption isotherm for the unmodified coarse sand provided a 
baseline to compare with the sorption capacity of modified sands in the later 
experiments. The equilibrium solid-phase concentration data as a function of the 
equilibrium liquid-phase concentrations are presented in the naphthalene sorption 
isotherm plot in Figure 6.6. The partitioning coefficient, K
d
, was obtained by fitting 
the experimental data with the linear sorption isotherm model. As expected given the 
relatively low organic content of the coarse sand, the unmodified coarse sand shows 
relatively low sorption capacity, as verified by the partitioning coefficient of 0.0031 
L/g sand. It was assumed that the partitioning coefficient for the unmodified fine sand 
was comparable considering the similar properties for both sands, such as the 
negligible organic content. Given the low sorption levels on the unmodified coarse 
sand, the sorption/desorption kinetics were not determined, but were assumed to be 
very fast. 
6.1.5.2 HMN-Coated Sand Sorption 
The results of the HMN-coated sand sorption isotherm experiments are shown 
in Figure 6.7, and suggest a linear relationship between the experimental sorbed and 
aqueous phase equilibrium concentration data. Applying the linear isotherm model, 
the partitioning coefficients, K
d
, are 0.010 L/g and 0.016 L/g for the HMN-Coated 
Mystic White II coarse sand and Filtersil fine sand, respectively.  
 
 164 
 
0246810
0.00
0.01
0.02
0.03
Y = 0.0031*X-0.0005
R = 0.9595
 
 
 Experimental data
 Linear fit data
Sorbed Co
ncentration (m
g/
g
 
sand)
Equilibrium Concentration (mg/L)
 
Figure 6.6 Naphthalene sorption isotherm onto unmodified coarse sand. 
 
 
 165 
 
012345
0.00
0.01
0.02
0.03
0.04
0.05
(a)
Y = 0.00956*X-0.00212
R = 0.9915
 
 
 Experimental Data
 Linear Fit Data
Sorbed Co
ncentration (m
g/
g
 
sand)
Equilibrium Concentration (mg/L)
 
 
0123
0.00
0.01
0.02
0.03
0.04 (b)
Y = 0.01623*X-0.00292
R = 0.9974
 
 
 Experimental Data
 Linear Fit Data
Sorbed Co
ncentration (m
g/
g
 
sand)
Equilibrium Concentration (mg/L)
 
 
Figure 6.7 Naphthalene sorption isotherms onto HMN-Coated sands: (a) HMN-
Coated Mystic White II coarse sand, and (b) HMN-Coated Filtersil fine sand. 
 
 166 
 
The partitioning coefficients for HMN-coated sands are fairly high compared 
to the values of 
5
108
?
?  and
4
107.1
?
?  L/g reported for two types of sands by Appert-
Collin et al., (1999). Compared to the sorption capacity of unmodified sand, the 
increases in the partitioning coefficients with the HMN coating are 3.2 and 5.2 times 
for the Mystic White II and Filtersil sands, respectively (assuming =
d
K 0.0031 L/g 
for both sands). Although the increase in the partitioning coefficients is fairly 
significant due to the very low sorption capacity of the unmodified sands, most of the 
naphthalene still remained in the aqueous phase, as demonstrated by the equilibrium 
aqueous naphthalene concentrations. As discussed in Section 5.3.5, one of the major 
concerns regarding the modified porous media system was that the modified system 
should possess a strong sorptive capacity, yet slow desorption rate, so that the 
sorption/desorption rate would limit the overall bioremediation rate. Therefore, there 
was concern that the sorption capacity of the HMN-coated sands did not increase 
enough compared to the unmodified sand to meet the above requirement. 
Furthermore, Appert-Collin et al., (1999) showed that the desorption rate of 
naphthalene from HMN-coated sand was relatively fast, which did not meet the 
requirement that the desorption rate should be the overall rate-limiting process for in 
situ bioremediation. Therefore, the idea of coating sand with HMN was abandoned 
for the utilization in this project, and no sorption/desorption kinetic experiments were 
performed. 
 
 
 167 
 
6.1.5.3 Surfactant (CPC)-coated sand sorption 
The major factor that affects the capacity of sand to sorb a surfactant coating 
is the cation exchange capacity (CEC). Although the CEC value for Mystic White II 
coarse sand was determined with Agri Analysis, Inc., it was measured again when the 
CEC value for the Filtersil sand needed to be determined. The CEC value for the 
Mystic White II and the Filtersil sands were determined to be 0.1-0.3 meq/g by the 
Maryland Cooperative Extension Soil Testing Lab (University of Maryland, College 
Park, MD USA). These low values suggest that this approach may be challenging, 
although others have reported success applying surfactant coatings to sandy aquifer 
materials, e.g.,(Kibbey and Hayes 1993). 
In the investigation of CPC-coated sand sorption, the operational conditions 
that promoted the best surfactant coating results were determined and found to be 
pH=9, when the aqueous phase was DI water. It was also shown that CPC sorption 
onto sand reached equilibrium within 14 hours. Therefore, a mixing time of 24 hours 
was used subsequently to ensure the sorption equilibrium of CPC onto the sand.  
Critical micelle concentration (CMC) is another key factor that affects the 
coating of CPC on the sand. Kibbey and Hayes (1993) reported the CMC for CPC to 
be 
5
104.4
?
?  M at 0.1 M ionic strength and 25?C. However, in general, the CMC 
values are subject to large changes depending on factors such as temperature, ionic 
strength and pH. It was estimated that the CMC for CPC for the conditions used in 
this study was between 
4
101
?
?  and 
4
103
?
? M by examining the CPC aqueous 
solutions. When the CPC concentrations were greater than 
4
101
?
? M, no CPC was 
sorbed on the sand, suggesting the formation of micelles, which interfered with the 
 
 168 
 
coating of CPC and measurement of CPC by spectrophotometer. Therefore, the CPC 
sorption isotherm was conducted with CPC concentrations lower than 
4
101
?
? M. The 
CPC partitioning coefficient was determined to be 0.02 L/g under the experimental 
conditions, which is much lower than the value of 16 L/g reported by (Kibbey and 
Hayes 1993) for CPC-coated silica, probably due to the low CEC value of the coarse 
sand used in this study. 
The naphthalene sorption isotherm onto the CPC-coated Mystic White II 
coarse sand is shown in Figure 6.8. Again the data appeared to be linear and a linear 
isotherm model was applied. The partitioning coefficient (K
d
) was determined to be  
0.0064 L/g. This represents an increase in the value of twice over the unmodified 
sand, but it is still relatively low, suggesting very little naphthalene sorbed on the 
CPC-coated sand, consistent with the limited amount of CPC coating on the sand.  As 
with the HMN-Coated sands, a synthetic model soil with such limited sorption 
capacity could not sorb enough naphthalene to meet the requirement of Scenario #3, 
which disqualified the CPC-coated sand as a candidate for the synthetic model soil. 
Therefore, no further study was conducted with CPC-coated Filtersil sand because of 
its even lower CEC, and no sorption/desorption kinetics experiments were performed. 
 
 169 
 
024681012
0.00
0.02
0.04
0.06
0.08
Y = 0.0064X-0.0094
R = 0.8565
 
 
 Experiment
 Linear Fit
Sorbed concentrat
i
on (m
g/g)
Equilibrium concentration (mg/L)
 
Figure 6.8 Naphthalene sorption isotherms onto CPC-coated Mystic White II sand. 
 
 170 
 
6.1.5.4 Mixing sands with Amberlite Resins 
The naphthalene sorption onto Amberlite resin XAD-2 was investigated first 
and the resulting isotherm data are presented in Figure 6.9. These data showed that 
XAD-2 had a significant sorption capacity for naphthalene. A comparison of sorption 
isotherm models revealed that the XAD-2 data could be described using the 
Freundlich equation. Therefore, the best fit of the Freundlich equation to the 
experimental data by the linear regression of log (Q) versus log (C) is shown along 
with the experimental data in Figure 6.9, and the resulting Freundlich equation 
constants are given in Table 6.6. Guerin and Boyd (1997) reported a K
d
 value of 13.6 
L/g for the naphthalene sorption isotherm on XAD-2, which is in the same order of 
magnitude of the value obtained in this study, although smaller. Nonetheless, both the 
partitioning coefficients suggest that the sorption capacity of XAD-2 for naphthalene 
is very large. 
 
Table 6.6 Freundlich constants for sorption of naphthalene from XAD-2 
Material K
f
 (mg
1-1/n
?L
n
?g
-1
) n 
XAD-2 56 0.46 
 
 171 
 
-8 -6 -4 -2 0 2
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
Q=56C
0.46
R
2
=0.9582
 
 
Ln(Q
)
Ln(C)
 
Figure 6.9 Naphthalene sorption isotherm onto XAD-2. 
 
 172 
 
Although it was demonstrated that XAD-2 has a large sorption capacity for 
naphthalene, there was concern that the naphthalene desorption rate from XAD-2 
would be extremely slow (based on an approximate evaluation of XAD-2 properties 
such as pore size and polarity), which would make the experimental operation in 
Scenario #3 infeasible if it took too long to monitor the desorption process. In 
addition, as described previously, in Section 5.3.5, another major concern about the 
synthetic sorption system is that the modification to the porous media should not 
significantly alter the hydraulic conductivities of the two-layer ISFC system, nor 
should they alter the amount or the nature of the surface area available for microbial 
attachment and growth. However, the skeleton density of XAD-2 is 640 kg/m
3
, which 
is significantly lower than the particle densities of both Mystic White II coarse sand 
and Filtersil fine sand (Table 6.1). As a result, there was some concern regarding the 
potential for floating of XAD-2 in the synthetic sorption system. Therefore, XAD-7 
(Alltech, USA) was chosen as an alternative to evaluate. Although it has lower 
sorption capacity for naphthalene compared to XAD-2, the partitioning coefficient 
was expected to sufficiently high to ensure that enough naphthalene could be sorbed 
on the XAD-7. In addition, XAD-7 possesses more attractive physical properties, 
such as a density of 1240 kg/m
3
, which is more similar to the sands used in ISFC. 
Therefore, it was decided that, compared to XAD-2, the mixing of XAD-7 resin with 
sand was less likely to affect the hydraulics in the ISFC significantly. Therefore, all 
further efforts were focused on the sorption/desorption of XAD-7.  
The experimental data for the naphthalene sorption isotherm using XAD-7 is 
presented in Figure 6.10, and the naphthalene sorption isotherms onto mixtures of 
 
 173 
 
XAD-7 with Mystic White II coarse sand and the Filtersil fine sand are presented in 
Figure 6.11 and 6.12, respectively. In all cases, the data were fit with the linear 
isotherm model. Two observations can be made with regard to the naphthalene 
sorption isotherms on XAD-7. One, the partition coefficient of naphthalene on XAD-
7 was 3.9 L/g, based on the linear sorption model, which was much higher than the 
value reported by Mulder et al. (2000). In their study, the naphthalene sorption 
isotherm on XAD-7 was described by the Freundlich isotherm model, with the K
d
 and 
n values of 1.93 L
3n
/g
n
 and 0.69, respectively. This difference is speculated to be 
possibly due to the fact that in Mulder et al. (2000) study, the XAD-7 was sieved and 
the resulting size fractions were utilized, whereas the XAD-7 was utilized without 
any sieving pretreatment in this study. Two, as summarized in Table 6.7, the partition 
coefficients for the mixtures of XAD-7 with Mystic White II coarse sand and Filtersil 
fine are less than for XAD-7 alone, which suggests that the presence of the sand 
interferes with the contact between the naphthalene and XAD-7, resulting in the 
decreased sorption capacity. 
 
 
 174 
 
012345
0
5
10
15
20
25
Y = 3.904*X+2.847
R = 0.976
 
 
 Experimental Data
 Linear Fit
S
o
r
b
e
d
 Concentrat
i
on (m
g/g)
Equilibrium Concentration (mg/L)
 
 
Figure 6.10 Naphthalene sorption isotherm onto XAD-7.
 
 175 
 
01234567
0
5
10
15
20
25
(a)
Y = 3.0830*X + 2.9799
R = 0.9723
 
 
 Experiment data
 Linear fit
Sorbed Concentration
 
(m
g/g X
A
D-7)
Equilibrium Concentration (mg/L)
 
 
01234567
0.0
0.5
1.0
1.5
2.0
2.5
(b)
Y = 0.30822*X+0.29807
R = 0.97229
 
 
 Experimental data
 Linear Fit
Sorbed Concentration
 
(m
g/g m
i
xture)
Equilibrium Concentration (mg/L)
 
 
Figure 6.11 Naphthalene sorption isotherms onto the mixture of Mystic White II 
coarse sand and XAD-7 (10% w/w of XAD-7): (a) Sorbed concentration based on the 
XAD-7 mass; and (b) Sorbed concentration based on the mass of the mixture of 
XAD-7 with Mystic White II coarse sand. 
 
 176 
 
0123456
0
5
10
15
20
25
(a)
Y = 3.1565*X+0.3766
R = 0.9727
 
 
 Experimental Data
 Linear Fit
Sorbed Concentration
 
(m
g/g X
A
D-7)
Equilibrium Concentration (mg/L)
 
 
0123456
0.0
0.5
1.0
1.5
2.0
2.5
(b)
Y = 0.31569*X+0.03764
R = 0.97272
 
 
 Experimental Data
 Linear Fit
Sorbed Concentration
 
(m
g/g m
i
xture)
Equilibrium Concentration (mg/L)
 
 
Figure 6.12 Naphthalene sorption isotherms onto the mixture of Filtersil fine sand and 
XAD-7 (10% w/w of XAD-7): (a) Sorbed concentration based on the XAD-7 mass; 
and (b) Sorbed concentration based on the mass of the mixture of XAD-7 with 
Filtersil fine sand.
 
 177 
 
 Table 6.7 Partition coefficient of naphthalene onto XAD-7 
Sample Partition coefficient, 
d
K  (L/g XAD-7) 
XAD-7 3.904 
Mixture of XAD-7 (10% w/w) with 
Mystic White coarse sand 
3.083 
Mixture of XAD-7 (10% w/w) with 
Filtersil fine sand 
3.157 
 
Once the properties describing the sorption equilibrium for naphthalene on 
XAD-7 had been quantified, the next step was to quantify the sorption/desorption 
kinetics. The sorption/desorption kinetic parameters were estimated using a nonlinear 
minimization procedure with the dual resistance CMBR model (Equation 3.8). 
Figures 6.13 and 6.14 present the experimental results of the sorption and desorption 
rate studies, respectively, along with the model nonlinear regression. These results 
indicated that the sorption/desorption process exhibited a phased behavior, 
proceeding with a rapid initial rate that was followed by a slower rate, which required 
as long as several days to approach equilibrium. The sensitivity analysis performed 
for the dual resistance model demonstrated that the film transfer rate may be 
estimated from the sorption rate at early values of time, while surface diffusivity 
usually controls the rate of sorption after long periods of time (Mulder et al. 2001). 
This observation was applicable to the stirred CMBR data of this study
 
 178 
 
0 1000 2000 3000 4000
0.0
0.2
0.4
0.6
0.8
1.0
 
 
 Experimental Data
 Dual Resistance Model Data
F
r
acton of
 Naph
t
h
alene
 
Unsorbed
Time (Minutes)
 
Figure 6.13 Naphthalene sorption kinetics onto a mixture of coarse sand 
XAD-7. 
0 1000 2000 3000 4000
0.0
0.2
0.4
0.6
0.8
1.0
 
 
 Experimental Data
 Dual Resistance Model Data
F
r
actio
n
 
o
f
 naphtha
lene und
esorbed
Time (Minutes)
 
 
Figure 6.14 Naphthalene desorption kinetics from a mixture of preloaded 
coarse sand and XAD-7. 
 
 
 179 
 
In the sorption kinetic study, the best fit model parameters for the dual-
resistance CMBR model were 9?10
-4
 cm/s and 1.3?10
-8
 cm
2
/s for the film transfer 
coefficient and intraparticle diffusion coefficient, respectively, whereas in the 
desorption study, these two values for the film transfer coefficient and intraparticle 
diffusion coefficient were 2?10
-4
 cm/s and 1.3?10
-8
 cm
2
/s, respectively. These 
values were obtained by fitting the CMBR model with both kinetic experimental data. 
The sorption rate model provided reasonable reproduction of the observed laboratory 
sorption data (Figure 6.13). However, the desorption rate model prediction did not 
reproduce the observed experimental data well at the later stage of the desorption 
process (Figure 6.14). Based on these experimental observations, it was speculated 
that there was a fraction of residual naphthalene that either desorbed only at an 
extremely slow rate, or did not desorb from XAD-7 at all, indicating that a fraction of 
naphthalene sorbed onto XAD-7 was irreversible.  
As discussed in Section 3.2, at high Biot numbers ( 1>>Bi ), film diffusion 
limitations can be neglected and intraparticle mass transfer is the rate-limiting step. In 
this study, the surface- and pore- based Biot number is 5.4?10
5
, which indicates that 
the intraparticle mass transfer is the rate-limiting process for the naphthalene 
sorption/desorption process. Therefore, the intraparticle diffusion coefficient was 
estimated to be the representative kinetic rate parameter with a value of 1.3?10
-8
 
cm
2
/s, which would be adjusted by the radius of the particle, R, to be utilized in the 
quantitative framework for the simulated bioremediation experiment Scenario #3. The 
intraparticle effective diffusion coefficient of 1.3?10
-8
 cm
2
/s determined in this work 
is about 400 times smaller than the value of 5.29?10
-6
 cm
2
/s reported by Mulder et al. 
 
 180 
 
(2000). Again, as discussed above with regard to the discrepancies in sorption 
isotherm results between this experiment and Mulder et al. (2000), the different 
XAD-7 particle size fractions may have also contributed to the observed kinetic 
differences. Furthermore, there were differences in the rotational speeds in the two 
studies, which affect the Reynolds numbers, and thereby affect the 
sorption/desorption kinetics (Mulder et al. 2000). 
 
 181 
 
6.2 Bioremediation Simulation Control Experiments 
These control experiments were originally designed to replicate the conditions 
in the Scenario #1 bioremediation simulation experiments, but without biological 
activities. However, the control experiments were performed with a step input of a 
naphthalene plume, as was the original experimental plan, whereas the actual 
Scenario #1 experiments were conducted with a pulse input of a naphthalene plume to 
enhance the importance of dispersion. Nevertheless, although not exact replicates of 
the Scenario #1 conditions, the control experiments were still useful and informative 
as described below.  
6.2.1 Phase I: Intrinsic Control Experiments 
The Phase I intrinsic control experiments were conducted twice. The 
representative operational flow rate and naphthalene concentration obtained from the 
influent sampling port as a function of time are presented in Figure 6.15 (a) and (b), 
respectively. It is clearly shown that the flow rate was consistent during the 
experiment. The naphthalene concentration was slightly higher (~5.6 mg/L) than the 
desired concentration of 5 mg/L at the early stage of the experiment, and then 
dropped to ~ 4.3 mg/L, presumably due to the volatilization into the headspace in the 
feeding bottle. The naphthalene concentration was quickly adjusted back to 5 mg/L 
by adding more fresh naphthalene solution of ~ 6 mg/L into the feeding bottle. 
Nonetheless, the average naphthalene sample concentration in the influent was 5.0 
mg/L during the Phase I experiment. 
 
 182 
 
0 100 200 300
3
4
5
6
7
8
(a)
 
 
Flow r
a
t
e
 
(m
l/m
i
n)
Time (hours)
 
0 100 200
2
3
4
5
6
7
(b)
 
 
Naph
t
h
alene
 Conce
n
t
r
ati
on 
(
m
g/L)
Time (hours)
 
 
Figure 6.15 Operational conditions for Phase I control experiment: (a) flow rate; and 
(b) naphthalene concentration from the influent sampling port. 
 
 183 
 
Representative Phase I breakthrough curves for naphthalene, normalized by its 
average sample concentrations in the influent (5.0 mg/L), are presented as a function 
of time for sampling ports A4,  C4, E4, B9, C9 and D9 in Figure 6.16. These data are 
typical of the results obtained during the Phase I ISFC control experiments. The 
RT3D in VMOD simulation results for the normalized naphthalene concentrations are 
presented along with the experimental data, with the naphthalene concentrations 
normalized by average influent concentration as well. The physical and chemical 
kinetic parameters used in the RT3D simulation are summarized in Table 6.8. No 
biokinetic reaction was simulated in the RT3D numeric transport engine for these 
experiments, which were performed before inoculation of the ISFC. 
As shown in the breakthrough curves from A4 and B9 (Figure 6.16 (a) and 
(b)) in the high-K layer, the naphthalene concentration arrived and reached the steady 
state in a very early stage of the experiment due to the fast velocity. In comparison, 
naphthalene concentration arrived at the steady state much later in the low-K layer 
because of the much slower flow rate (Figure 6.16 (e) and (f)). The experimental 
naphthalene concentration was slightly lower than the model simulated concentration 
in the high-K layer sampling ports (Figure 6.16 (a) and (b)), possibly due to the minor 
volatilization of naphthalene into the unsaturated zone and the overlying headspace in 
the ISFC, although the concentration did eventually reach the predicted value. The 
?tailing? in the breakthrough curves at A4 and B9 has been commonly observed by 
others, e.g., (Miller and Weber 1988). On the other hand at E4, the experimental 
concentration never quite reached the predicted concentration, probably due to the 
sorption to the silicone caulking. 
 
 184 
 
0 100 200 300
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(a)
A4 - High-K layer
 
 
 experimental data
 model simulation
N
o
rmal
iz
ed
 C
o
n
c
en
t
r
a
t
i
o
n
Time (hours)
 
0 100 200 300
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(b)
B9 - High-K layer
 
 
 experimental data
 model simulation
N
o
r
m
a
liz
e
d
 Co
nc
en
t
r
at
ion
Time (hours)
 
0 100 200 300
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(c)
C4 - Interface
 
 
 experimental data
 model simulation
Nor
m
al
i
z
ed
 C
onc
en
t
r
atio
n
Time (hours)
 
Figure 6.16 Example breakthrough curves from the Phase I control experiment for 
sampling ports: (a) A4, (b) B9, (c) C4 (d) C9, (e) E4 and (f) D9. Each graph shows 
the experimental naphthalene concentration along with the RT3D model prediction.  
 
 185 
 
0 100 200 300
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(d)
C9 - Interface
 
 
 experimental data
 model simulation
Nor
m
ali
z
e
d
 Con
c
en
t
r
at
ion
Time (hours)
 
0 100 200 300
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(e) E4 - Low-K layer
 
 
 experimental data
 model simulation
Nor
m
ali
z
e
d
 Con
c
en
t
r
at
ion
Time (hours)
 
0 100 200 300
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(f)
D9 - Low-K layer
 
 
 experimental data
 model simulation
Nor
m
ali
z
e
d
 Con
c
en
t
r
at
ion
Time (hours)
 
Figure 6.16 Continued 
 
 
 186 
 
Table 6.8 Hydro-geological parameters utilized in RT3D simulation  
Hydro-geological parameters 
Term Coarse Layer Fine Layer 
Hydraulic Conductivity, 
x
K  (m/s) 
0.004 0.00065 
Longitudinal Dispersivity, 
x
?  (m)              
0.0015 0.0006 
Vertical Transverse Dispersivity, 
z
? (m) 
7.6 E-5 0.002 
Effective Diffusion Coefficient, *D  (m
2
/hr)
1
 6.5 E-7 5.8 E-7 
Interlayer dispersivity, '
z
?  (m)
 2
 
0.05 
1
m
DD ?=* , where the aqueous diffusion coefficient
m
D is calculated using the 
equation by Wilke and Chang (Welty et al. 1984), ? is estimated as 
3
1
?? = (Schwartz 
and Zhang 2002), 
2 
Used for model prediction at sampling port C4 and C9. 
 
As was observed by Szecsody et al. (1994) in their laboratory with stratified 
porous media, the data presented in Figure 6.16 (f) for sampling ports D9 demonstrate 
that  the interlayer mass transfer produced two stages of tracer breakthrough: an 
initial pulse resulting from transverse mass transfer from the high-K layer, followed 
by a mass pulse from the advective-dispersive movement in the low-K layer. The 
interlayer transverse dispersivity near the interface, '
Z
? , was initially modeled using 
RT3D/VMOD with the dispersivities determined within each layer. Then it was 
adjusted by trial and error based on a visual appraisal to obtain the best fit the data of 
initial mass pulse from interlayer mass transfer. Based on the calibration, the 
interlayer mass transfer coefficient was estimated to be 0.05 m. This is an order of 
magnitude larger than the interlayer dispersivity value of 0.003 m estimated by 
Szecsody et al. (1994). Nevertheless, the naphthalene broke through more slowly at 
 
 187 
 
sampling port C4 and C9 than was predicted by the model simulation using the 
estimated '
Z
?  value (Figure 6.16 (c) and (d)), suggesting that the interlayer 
dispersion coefficient fitted using the port D9 data might actually be underestimated. 
Szecsody et al. (1994) discussed several possible explanations for the 
observation of increased dispersivity at the interface in a low contrast (high-K/low-
K=6) system. A possible reason is that the interface may be a transitional zone, and as 
fluid travels back and forth across the interface, tortuosity increases and 
x
? and 
z
? become somewhat larger. Some experimental artifacts may also contribute to the 
increased interlayer dispersivity, including: (1) irregular packing at the interface, (2) 
flow not being parallel to layers, or (3) acquisition of samples from near the interface. 
However, all these three factors were carefully considered either when packing the 
ISFC or taking samples during bioremediation simulated experiments to minimize 
these effects on the simulated aquifer system. Therefore, it was speculated that the 
interface being a transitional zone, rather than the experimental artifacts, might 
contribute to the increased interlayer dispersivity in this work. 
6.2.2 Phase IIA: Nutrients Enhanced Control Experiments 
The Phase IIA control experiment was performed two times following the 
same procedures as in Phase I, but with the input of additional N and P to the MSNS 
media. To achieve this, the N and P content of the MSNS media was increase to 5 
times of the original concentration, resulting in N and P addition equal to 6 times the 
stoichiometric requirement for 5 mg/L naphthalene (Equation 5.1). The results of a 
representative nutrient addition experiment are presented in Figure 6.17 for sampling 
ports B9, C9 and D9. The Phase I naphthalene concentration data are included as well 
 
 188 
 
for comparison purposes. Clearly, the nutrient addition did not have any effect on the 
concentration of naphthalene in the ISFC, as expected. Indeed, the results from Phase 
I and IIA are very similar and indicative of the reproducibility of these experiments. 
6.2.3 Phase IIB: Engineered Control Experiments 
In Phase IIB of the control experiments, the flow rate was doubled to increase 
the dispersion rate and simulate the alternative perturbation to be used in Scenario #1. 
The data from this experiment are also presented in Figure 6.17 for sampling ports 
B9, C9 and D9. These data show no significant effect on the steady-state naphthalene 
concentration in response to the increased advection and discussion, as was expected 
once the plume front had passed through the tank, because there was no 
biodegradation occurring in the system. 
6.2.4 Summary and Conclusion 
The independently determined hydrogeologic parameters ( K and ? ) 
summarized in Table 6.8 and used in RT3D/VMOD were able to accurately describe 
the experimental data and verified the use of these parameters for modeling of 
bioremediation simulation experiments. There were some discrepancies, but these 
could be explained. 
The data demonstrated the interlayer mass transfer phenomena and were used 
to fit the interlayer mass transfer coefficient, '
z
? . There was no biodegradation of 
naphthalene in the ISFC prior to inoculation, and, as expected, the naphthalene 
transport in the ISFC was not changed when an excess of N and P were added or 
when the flow rate was increased. 
 
 189 
 
0 100 200 300
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(a)
B9 - High-K layer
 
 
 Phase I
 Phase IIA
 Phase IIB
Nor
m
al
i
z
ed
 C
onc
en
t
r
atio
n
Time (hours)
 
0 100 200 300
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(b)
C9 - Interface
 
 
 Phase I
 Phase IIA
 Phase IIB
N
o
rm
al
i
z
ed Co
nce
n
tra
t
i
o
n
Time (hours)
 
0 100 200 300
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(c)
D9 - Low-K layer
 
 
 Phase I
 Phase IIA
 Phase IIB
Nor
m
al
i
z
ed
 C
onc
en
t
r
atio
n
Time (hours)
 
Figure 6.17 Example breakthrough curves from Phase II control experiment for 
sampling ports: (a) B9, (b) C9 and (d) D9. Each graph shows the naphthalene data 
from Phase IIA, Phase IIB and Phase I for comparison.
 
 190 
 
6.3 Bioremediation Simulation Experiments, Scenario #1: Dispersion-Limited 
Scenario 
As discussed in Section 5.4.3, Scenario #1 Phase I was performed to simulate 
the intrinsic biodegradation under natural conditions for which vertical transverse 
dispersion was the overall rate-limiting process. In Phase IIA, the first perturbation 
received by the system was adding N and P in excess of the stoichiometric 
requirements in an attempt to enhance the biodegradation rate. In Phase IIB, the flow 
rate was increased to a level of three times that in Phase I, to increase the dispersion, 
especially the interlayer transverse dispersion between the high-K layer and low-K 
layer. The results from these simulated bioremediation experiments were presented 
and discussed as follows. 
6.3.1 Phase I: Intrinsic Biodegradation 
As described in Section 5.4.3.1, the naphthalene and Br in MSNS were 
pumped into the ISFC continuously for 48 hours, followed by MSNS only for the rest 
of experimental run. The Phase I experiment was performed twice (Phase I (1) and 
(2)). However, as described in Section 5.4.2.1, the ISFC was disinfected after the 
Scenario #1 experiments and Scenario #2 Phase IIA (Nutrients Enhanced 
Biodegradation), due to the algal growth that resulted from the addition of excess N 
and P. Therefore, the ISFC was inoculated again following the same inoculation 
procedure as utilized the first time. In order to verify that the results were 
reproducible after the reinoculation, the Phase I experiment was conducted one more 
time (Phase I (3)). This experiment was also the Scenario #2 Phase II experiment, 
 
 191 
 
because the same experimental conditions were required and employed. Therefore, 
the results of Scenario #1 Phase I and Scenario #2 Phase II were combined as 
replicate results in the following discussion. 
In order to compare the results between different phases and different 
scenarios, the Br, naphthalene and DO concentrations were normalized. The 
naphthalene and Br concentrations in the sampling ports and effluent were 
normalized by their average sample concentrations in the influent, whereas the DO 
concentrations in the sampling ports and effluent were normalized by dividing by the 
DO concentration in influent taken at the same sampling time to eliminate the effect 
of the temperature and atmospheric pressure. In the following results and discussion, 
the same normalization techniques were always utilized. 
The influent curves for the experimental Br, naphthalene and DO data, are 
shown in Figure 6.18 (a). The normalized effluent Br, naphthalene and DO data are 
presented in Figure 6.18 (b). For comparison purposes, the normalized Br and 
naphthalene data in the influent are also plotted in Figure 6.18 (b). As shown clearly 
in Figure 6.18 (a), the influent Br and naphthalene concentrations in the pulse input 
are fairly steady. With respect to the effluent data in Figure 6.18, the naphthalene 
concentration is much reduced compared to the Br concentration. The fact that the 
DO concentration in effluent was approximately 50% of that in the influent indicates 
that the naphthalene loss was due to the consumption of naphthalene via 
biodegradation. 
 
 192 
 
0 100 200
0
2
4
6
8
10
12
(a)
 Naphthalene
 DO
 Br
Time (hours)
Naph
t
h
alene
 o
r
 
DO
 Conce
n
tr
ati
on 
(
m
g/L)
0
100
200
Br
 Concentratio
n
 
(m
g/L)
 
 
0 100 200
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(b)
 
 
 Influent Naphthalene
 Influent Br
 Effluent Naphthalene
 Effluent Br
 Effluent Oxygen
Norm
a
liz
ed Concentration
Time (hours)
 
 
Figure 6.18 Breakthrough curves for influent and effluent from the Phase I (2): (a) 
Naphthalene, Br and DO concentration in the influent and (b) Normalized 
naphthalene and Br in the influent and effluent, and normalized DO in the effluent.
 
 193 
 
As discussed in Section 5.4.3.1, a ?snapshot? contour sampling round was 
also performed in Phase I (2). Therefore, a contour plot was obtained to provide for 
visual inspection and subjective comparison of the experimental results and RT3D 
model simulation. To accomplish this, the output file was first exported for the 
selected time period of 48 hours from RT3D, and manipulated using Surfer
?
 (Version 
6.0, developed by Golden software, Inc.), to create a plume contour plot based on the 
model data set. Then, the experimental data recorded for all the sampling port 
samples taken at 48 hours were incorporated into the plot for visual comparison. 
Specifically, the first step in creating a contour plot within Surfer
?
 using the 
model simulation data was to ?grid? the contour framework, which consisted of 
inputting the model grid spacing and dimensions into Surfer
?
 Shan and Stephens 
(1994) provided a recommended approach for the gridding process to ensure that the 
actual grid spacing from the numerical model was imported into Surfer
?
 and used for 
the subsequent contour development. Then, the contour plot was generated by 
interpolating the spatial distribution using the dataset of known points exported from 
the numerical model. The method used in this work was kriging, which is commonly 
used in contouring, e.g., (Anderson and Woessner 1992). The created contour plot for 
Phase I (2) at 48 hours is shown in Figure 6.19. The contour plot is presented with the 
longitudinal distance (0.0-1.2 m) along the x-axis and the transverse vertical distance 
(0.0-0.3 m) along the z-axis. 
 
 
 194 
 
 
 
 
Figure 6.19 Contour plot of naphthalene plume at 48 hours for Scenario #1 Phase I (2). 
 
 
 195 
 
Two observations can be made based on a visual comparison of the 
experimental data with the model simulation data. One, in general, the model 
simulation results matched the experimental data well, except for the data in the slow-
K layer, suggesting the hydro-geological and biological parameters utilized in RT3D 
model simulation were reasonable. Two, as shown in the plot, the contour lines along 
the interface between the two layers are spread out wider than within each layer, 
because the interlayer dispersion coefficient is greater than those within each layer. 
Representative Phase I breakthrough curves for Br, naphthalene and DO are 
presented as a function of time for sampling ports A2,  B7, C2, C7, D1 and E2 in 
Figure 6.20. These data are typical of the results obtained during the Phase I ISFC 
experiments for Scenario #1. The RT3D in VMOD simulation results for the 
normalized concentrations are presented along with the experimental data, with each 
species normalized in the same way as described for the experimental data. 
The hydro-geological parameters summarized in Table 6.8 were utilized in the 
RT3D simulation. In addition, the biological kinetic parameters employed are 
summarized in Table 6.9. All of the parameters were independently estimated as 
described earlier in this chapter, with the exception of the half-saturation constant for 
the electron acceptor for oxygen,
A
K , which was taken from the literature. 
 
 196 
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(a)A2 - High-K layer
 
 
 Experimental Naphthalene
 Model Naphthalene
 Experimental Br
 Model Br
 Experimental DO
Norm
a
liz
ed Concentration
Time (hours)
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(b)B7 - High-K layer
 
 
 Experimental Naphthalene
 Model Naphthalene
 Experimental Br
 Model Br
 Experimental DO
Norm
a
liz
ed Concentration
Time (hours)
 
 
Figure 6.20 Example breakthrough curves from the Phase I (2) experiment for 
sampling ports: (a) A2, (b) B7, (c) C2 and (d) C7, (e) D1 and (f) E2. Each graph 
shows Br tracer, naphthalene, and DO data, as well as the RT3D predictions for Br 
and naphthalene. 
 
 197 
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(c)C2 - Interface
 
 
 Experimental Naphthalene
 Model Naphthalene
 Experimental Br
 Model Br
 Experimental DO
Norm
a
liz
ed Concentration
Time (hours)
 
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(d)C7 - Interface
 
 
 Experimental Naphthalene
 Model Naphthalene
 Experimental Br
 Model Br
 Experimental DO
Norm
a
liz
ed Concentration
Time (hours)
 
 
Figure 6.20 Continued 
 
 198 
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
(e)D1 - Low-K layer
 
 
 Experimental Naphthalene
 Model Naphthalene
 Experimental Br
 Model Br
 Experimental DO
Norm
a
liz
ed Concentration
Time (hours)
 
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
(f)E2 - Low-K layer
 
 
 Experimental Naphthalene
 Model Naphthalene
 Experimental Br
 Model Br
 Experimental DO
Norm
a
liz
ed Concentration
Time (hours)
 
 
 
Figure 6.20 Continued 
 
 
 199 
 
Table 6.9 Biological parameters utilized in RT3D simulation for Scenario #1 
Biological parameters  
Model parameters Values 
Initial microbial concentration, M
0
 0.8 mg/L 
Measured initial oxygen background concentration, A
0
 3.5 mg/L 
Maximum specific substrate utilization rate, 
m
q  
0.18 /hr 
Monod half-saturation constant for naphthalene, 
S
K  
8 mg/L  
Monod half-saturation constant for oxygen, 
A
K  
0.1 mg/L
(1)
 
Biomass produced per unit of naphthalene utilized, 
SX
Y
/
 
0.34 
Oxygen used per unit of naphthalene utilized, 
SA
Y
/
 
3
(2)
 
(1)
 Literature value from Wilson et al. (1997). 
(2)
 Calculated value based on stoichiometry, ignoring biomass synthesis. 
 
The initial biomass concentration was estimated based on the HPC data. The 
HPC values of samples taken from the inoculum and sampling ports A2 and E2 after 
inoculation are 8.5 x 10
11
 and 8 x 10
8 
CFU/L, respectively, resulting in a ~ 99.9% of 
attachment of biomass onto the sands, which verifies the assumption that most 
bacteria are on the solid phase (Harvey et al. 1984). Based on the biomass 
concentration of 8 x 10
8 
CFU/L from sampling ports, and assuming an individual cell 
weight of 9.5 x 10 
-13 
g (Ghoshal and Luthy 1998), the initial aqueous microbial 
concentration in the ISFC was estimated to be approximately 0.8 mg/L on a pore 
volume basis, which was also equivalent to 0.27 mg dry wt/kg sand in the high-K 
layer and 0.15 mg dry wt/kg sand in the low-K layer, respectively, on a solid phase 
basis computed from 
b
M
?
?
0
(Clement et al. 1998). These values are similar to the 
median value of 0.2 mg dry wt/kg soil used by MacQuarrie and Sudicky (1990). The 
 
 200 
 
initial biomass concentration was based on the HPC obtained from samples taken 
immediately after inoculation. Therefore, strictly speaking, it is only appropriate for 
the first experiment run. However, it was used as an estimate for all experiments, 
because it was not possible to sample the solid phase during the experiments. This 
approximation appeared to be reasonable given that the aqueous phase microbial 
counts did not change appreciably during these experiments. 
Several general trends can be observed from/in the breakthrough curves from 
Figure 6.20. One, naphthalene and Br breakthrough occurred at an earlier stage of the 
experiment in the high-K layer, compared to the low-K layer, as expected (e.g., 
compare sampling ports A2 and E2). Further, the naphthalene/Br pulse arrived at 
sampling locations near/at interface earlier than at sampling locations ?deeper? in the 
low-K layer (e.g., compare sampling ports C2 and E2), due to the interlayer mass 
transfer. This phenomenon was also observed by Szecsody et al. (1994) in their 
laboratory study of the transport and biodegradation of quinoline in a two-
dimensional horizontally stratified porous media under dual substrate limitation. Two, 
at all sampling ports, the normalized naphthalene concentration is lower than that of 
Br due to the biodegradation of naphthalene. Correspondingly, as the plume passed a 
port, the oxygen concentration decreased rapidly, presumably due to the aerobic 
biodegradation of the naphthalene, reaching a minimum value at a time that 
corresponds to the time of the highest concentration of naphthalene passing the 
sampling port. Subsequently, the oxygen concentration increased as the naphthalene 
available for biodegradation decreased. Three, as the naphthalene plume moved 
through the ISFC, further loss due to biodegradation occurred. For example, the 
 
 201 
 
normalized naphthalene concentration at C7 is lower relative to that of Br compared 
to C2 due to the further biodegradation of naphthalene downgradient.  
The aqueous heterotrophic plate count data were relatively consistent with 
time at sampling ports A2, C2, E2 and in the effluent (Figure 6.21). However, the 
data did show that the microbial biomass tended to increase slightly when the 
naphthalene plume passed the port and decreased slowly as the naphthalene traveled 
further downgradient, e.g., see the sampling port A2 data especially.  
In general, the simulated naphthalene and Br results match the experimental 
study values well in the high- (e.g., A2 and B7) and low-K (e.g., D1 and E2) layers, 
which verifies the utilization of the independent transport and biodegradation 
parameter estimates. However, at the interface between high- and low-K layers, there 
are some discrepancies between the observed and simulated data in sampling port C2 
and C7, in particular for Br. As shown in C2 (Figure 6.19 (c)) and C7 (Figure 6.19 
(d)), the experimental Br data spread out wider than the RT3D model simulation Br 
data, which indicates an underestimation of interlayer dispersivity, as discussed 
earlier. Interestingly, the underestimation of the interlayer dispersivity had less of an 
effect on the naphthalene model predictions, which are much closer to the 
experimental data at ports C2 and C7 than for Br. As discussed further below, this can 
be explained by the fact that the naphthalene biokinetics were much faster than the 
dispersion rate, even at the interface; therefore, the naphthalene model predictions are 
dominated by the biodegradation rate, not the dispersion rate. Importantly, as 
observed by Szecsody et al. (1994), the early arrival of naphthalene near the interface 
due to the interlayer mass transfer resulted in biodegradation of naphthalene 
 
 202 
 
0 100 200 300
3
4
5
6
7
 
 
 Phase I A2
 Phase I C2
 Phase I E2
 Phase I Effluent
Heterotrophi
c plate coun
t
 Lo
g(
CFU/m
L
)
Time (hours)
 
Figure 6.21 Aqueous heterotrophic plate count data at sampling ports A2, C2, E2 and 
Effluent for Phase I (2). 
 
 203 
 
for a longer time than within the layers, e.g., compare C2 with A2 and E2, and C7 
with B7. 
To quantify the magnitude of naphthalene loss due to biodegradation, a 
moment analysis was performed as described in Chapter 5. The moment analysis 
results for the Phase I replicate data are presented in Table 6.11. First, velocities in 
each layer were computed from the individual breakthrough curves, using sampling 
port B7 for high-K layer and E2 for low-K layer. The calculated values vary slightly 
among different replicates, which could be due to the slow development of biomass 
distributed in the flow cell over time (Taylor and Jaffe 1990). Consistent with the K 
values, which varied by a factor of 7, velocities in the high-K layer were ~ 9 times 
greater than in the low-K layer. Second, the zeroth moment analysis was used to 
quantify naphthalene biodegradation. This analysis was necessary because examining 
biodegradation by just comparing the peak concentrations at different sampling ports, 
as was done above, is insufficient, due to the fact that dispersion will also reduce the 
peak concentration over time and distance even though the total mass remains the 
same. To eliminate this dispersion effect and quantify the loss due to biodegradation, 
the quantitative analysis was based on the areas under the Br and naphthalene 
breakthrough curves, A
Br
 and A
naph
, respectively, which represent the total mass of 
each species detected after different horizontal flow distances. Specifically, by using 
the ratio of A
naph
/A
Br
 for comparison within an experiment and between experimental 
Phases I and II, the effect of dispersion on naphthalene concentration can be 
eliminated by normalizing to A
Br
. Further, comparison from replicate experiments 
and between experimental Phases I and II is improved by normalizing to A
Br
 and 
 
 204 
 
eliminating effects like differences in the pulses, or advection and dispersion between 
experiments. 
The moment analysis was focused on sampling ports C2 and C7 because of 
the importance of interlayer mass transfer and the resulting increased biodegradation, 
as discussed above.  Several key observations can be made based on the moment 
analysis summarized in Table 6.10. One, significant biodegradation of naphthalene 
(43-76%) removal occurred in Phase I by the time the plume reached sampling port 
C2. The largest removal of naphthalene by C2 occurred in Phase I (1), which was the 
first experiment performed after inoculating the ISFC. Given that the inoculum was 
pumped from the inlet, it is speculated that there was a high biomass concentration 
between the inlet and C2. Two, the extent of biodegradation increased with increasing 
distance, as demonstrated by the decrease in value of A
naph
/A
Br
 from C2 to C7. For 
the triplicate experiments, the average naphthalene mass loss between C2 and C7 due 
to biodegradation (A
naph
/A
Br
 (C2) - A
naph
/A
Br
 (C7)) is 8%, which provides the baseline 
for comparison with the impact of the engineered biodegradation perturbations in 
Phase II. Although this value is relatively small, it was reproducible (Table 6.11). 
Finally, following the moment of the naphthalene plume further downgradient, almost 
all of the naphthalene was removed in the low-K layer via biodegradation by the time 
the plume exited the flow cell based on the fact that no naphthalene was detected at 
port D7, while the naphthalene removal in the both layers due to biodegradation, 
coupled with the dilution by the low-K layer flow resulted in a total removal across 
the tank of about 80% (1-A
naph
/A
Br
 (Effluent)).  
 
 205 
 
Table 6.10 Moment analysis summarization for Scenario #1 
High-K 
layer (m/d) 
Experiment Run 
Low-K layer 
(m/d) 
A
naph
/A
Br
 
(C2)
 
A
naph
/A
Br
 
(C7)
 
A
naph
/A
Br
(C2) 
- 
A
naph
/A
Br
(C7) 
A
naph
/A
Br
 
(Effluent) 
0.30 
Phase I (1) 
0.035 
0.24 0.18 0.06 0.16 
0.30 
Phase I (2) 
0.035 
0.40 0.32 0.09 0.21 
0.28 
Phase I (3) 
0.03 
0.57 0.48 0.09 0.20 
0.26 
Phase IIA 
0.029 
0.33 0.23 0.10 0.20 
0.75 
Phase IIB (1) 
0.17 
0.58 0.39 0.19 0.17 
0.75 
Phase IIB (2) 
0.17 
0.78 0.54 0.24 0.26 
 
 206 
 
To verify that dispersion was the overall rate-limiting process for the system 
under the Phase I conditions, the quantitative framework in Figure 4.3 was applied 
using the parameter estimates in Table 6.8 and 6.9, and other system parameters as 
summarized in Table 6.11. The calculated dimensionless numbers are presented in 
Table 6.12.  Note that the second rate comparison using the modified Sherwood No. 
2, Sh
2
?, could not be quantitatively made because there was minimal sorption of 
naphthalene to the sands. Therefore, the sorption mass transfer rate was assumed to 
be infinite compared to the transverse dispersion rate.  Based on the quantitative 
comparison in Table 6.13, dispersion was identified as the overall rate-limiting 
process. Therefore, based on the quantitative framework, it was a priori predicted that 
the best approach to stimulate biodegradation would be increasing advection and, 
thereby, increasing the overall rate-limiting mass transfer process, transverse 
mechanical dispersion. For this approach to work, it was important that vertical 
transverse dispersion be controlled by mechanical dispersion. For the high-K layer, 
*D
v
xz
?
=1.4 and for the low-K layer, 
*D
v
xz
?
=4.7; therefore mechanical 
dispersion is dominant in both layers and this approach should work.
 
 207 
 
Table 6.11 Parameters for quantitative framework evaluation for Scenario #1 
Parameters High-K layer Low-K layer 
Characteristic length, L, m 
(=vertical thickness of model aquifer) 
0.3 0.3 
Average pore water velocity, 
x
v , m/d 
0.293 0.033 
Vertical transverse dispersivity, 
Z
? , m 
7.6E-5 0.002 
Aqueous molecular diffusion coefficient, 
m
D , m
2
/d 
2E-5 2E-5 
Tortuosity of the medium, ?   0.78 0.70 
Effective molecular diffusion coefficient, 
D*, m
2
/d 
1
 
1.56E-5 1.4E-5 
Yield coefficient, 
sx
Y
/
,  
mg biomass /mg naphthalene 
2
 
0.34 0.34 
Maximum specific substrate utilization 
rate, 
s
x
Y
q
max
max
?
= , 1/d 
4.24 4.24 
Initial biomass concentration, M
0
,  
mg biomass/L 
0.8 0.8 
Initial naphthalene substrate concentration, 
S
0
, mg/L 
9.3 9.3 
1
m
DD ?=* , where
m
D is calculated using the equation by Wilke and Chang (Welty et al. 
1984), and ? is estimated as 
3
1
?? =  (Schwartz and Zhang 2002). 
2 
Conversion factors: 1 mg naphthalene is equivalent to 3mg naphthalene COD and 1 mg 
biomass is equivalent to 1.42 mg biomass COD based on the stoichiometric equation.
 
 208 
 
Table 6.12  Quantitative framework used to identify the rate-limiting process 
Symbol  High-K Layer Low-K Layer 
Pe
t
 (Transverse Peclet No.) 
= 
z
x
D
Lv
 
Pe
t
 = 2325 >>1: 
transverse dispersion 
limits 
Pe
t
 = 124>>1: transverse 
dispersion limits 
Sh
2
? (Modified Sherwood No. 2) 
= 
z
ls
D
LK
2
 
Sh
2
?=?>>1: transverse dispersion limits 
Da
6
 (Damk?hler No. 6) 
= 
z
DS
LMq
0
2
0max
 
Da
6
 = 875 >>1: 
transverse dispersion 
limits 
Da
6
 = 410>>1: transverse 
dispersion limits 
 
6.3.2 Phase IIA: Nutrient Enhanced Biodegradation 
Before the predicted appropriate perturbation was investigated, the commonly 
practiced approach of attempting to enhance in situ biodegradation in the field via 
addition of excess of N and P was tested. To achieve this, the N and P content of the 
MSNS media was increase to 5 times of the original concentration, resulting in N and 
P additions equal to 3 times the stoichiometric requirement for 10 mg/L naphthalene. 
In all other respects, the experimental conditions were the same as in Phase I. The 
influent curves for the experimental Br, naphthalene and DO data, are shown in 
Figure 6.22 (a). The normalized effluent Br, naphthalene and DO data are presented 
in Figure 6.22 (b). For comparison purposes, the normalized Br and naphthalene data 
in the influent are also plotted in Figure 6.22 (b). In this experiment, at one point the 
naphthalene concentration in the pulse input dropped to about 80% of the average 
concentration, which was adjusted quickly back to the desired concentration (~10 
 
 209 
 
mg/L) by feeding more naphthalene solution with higher concentrations (~12 mg/L). 
The normalized Br concentration, on the other hand, was very steady in the pulse  
input. As in Phase I, the coupled losses of naphthalene and DO across the ISFC 
indicate biodegradation occurred in the tank.  
The results of N and P addition experiment, which was performed once, are 
presented in Figure 6.23 in the form of breakthrough curves at ports B7, E2, C2 and 
C7. For comparison purposes, the Phase I naphthalene data are include as well. In 
addition, the HPC data for Phase IIA are presented in Figure 6.24. The aqueous 
heterotrophic plate count data were relatively consistent with time at sampling ports 
A2, C2, E2 and in the effluent, as in the Phase I (2) experiment. The microbial 
biomass tended to decrease at the end of experiment for all 4 sampling ports, 
indicating less naphthalene was available. 
 
 210 
 
0 100 200
0
2
4
6
8
10
12 (a)
 Naphthalene
 DO
 Br
Time (hours)
Naph
t
h
alene
 o
r
 
DO
 Conce
n
tr
ati
on 
(
m
g/L)
0
50
100
150
200
Br
 Concentratio
n
 
(m
g/L)
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
1.2 (b)
 
 
 Influent Naphthalene
 Influent Br
 Effluent Naphthalene
 Effluent Br
 Effluent DO
Norm
a
liz
ed Concentration
Time (hours)
 
Figure 6.22 Breakthrough curves for influent and effluent from the Phase IIA: (a) 
Naphthalene, Br and DO concentration in the influent and (b) Normalized 
naphthalene and Br in the influent and effluent, and normalized DO in the effluent.
 
 211 
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
(a)
B7 - High-K layer
 
 
 Phase II A Naphthalene
 Phase II A Br
 Phase II A Oxygen
 Phase I Naphthalene
Norm
a
liz
ed Concentration
Time (hours)
 
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0 (b)
E2 - Low-K layer
 
 
 Phase II A Naphthalene
 Phase II A Br
 Phase II A Oxygen
 Phase I Naphthalene
Norm
a
liz
ed Concentration
Time (hours)
 
Figure 6.23 Example breakthrough curves from Phase IIA experiment for sampling 
ports: (a) B7, (b) E2, (c) C2 and (d) C7. Each graph shows Br tracer data, naphthalene 
data, dissolved oxygen (DO) data and naphthalene data from Phase I (2) for 
comparison. 
 
 212 
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0 (c)
C2 - Interface
 
 
 Phase II A Naphthalene
 Phase II A Br
 Phase II A Oxygen
 Phase I Naphthalene
Norm
a
liz
ed Concentration
Time (hours)
 
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
(d) C7 - Interface
 
 
 Phase II A Naphthalene
 Phase II A Br
 Phase II A Oxygen
 Phase I Naphthalene
Norm
a
liz
ed Concentration
Time (hours)
 
 
Figure 6.23 Continued 
 
 213 
 
0 100 200 300
4
5
6
7
 
 
 A2
 C2
 E2
 Effluent
Het
e
r
o
t
r
oph
ic P
l
ate Cou
n
t 
l
og(CFU/m
l
)
Time (hours)
 
Figure 6.24 Aqueous heterotrophic plate count data at sampling ports A2, C2, E2 and 
Effluent for Phase IIA 
 
 
 214 
 
With regard to the concentration trends for Br, naphthalene and DO, they are 
very similar to those in Phase I experiments. As in Phase I, the naphthalene 
concentration was reduced relative to Br, and the DO concentration decreased when 
the naphthalene broke through. The naphthalene concentration was slightly lower in 
Phase IIA than that in the Phase I, as expected considering that the N and P 
concentrations in the original MSNS artificial groundwater are 60% the 
stoichiometric amount required (Equation 5.1) for the biodegradation of naphthalene 
with a nominal concentration of 10 mg/L. However, the overall amount of 
biodegradation was very similar to Phase I, as demonstrated by the moment analysis 
in Table 6.10. Although the Phase IIA experiment was only performed once, the 
naphthalene mass loss due to biodegradation during the transport between C2 and C7 
was 10%, which is very close to the average naphthalene loss of 8% in the Phase I 
replicates. This outcome is not surprising because the rate-limiting process in the 
scenario is dispersion, as determined by the quantitative framework, not biokinetics 
due to the nutrients limitation. Therefore, a more appropriate approach is needed to 
alleviate the overall rate-limiting process so as to enhance the overall bioremediation 
rate. This was investigated in Phase IIB. 
 
 215 
 
6.3.3 Phase IIB: Engineered Biodegradation 
As discussed above, the engineered perturbation that was expected to be 
successful for this scenario was increasing the groundwater flow rate so as to increase 
the rate of transverse mixing between the two layers, alleviating the overall-rate 
limiting process of transverse dispersion, and thereby stimulating contaminant 
degradation and growth. Therefore, the Phase IIB experiments were conducted in the 
same way as in Phase I, except that (1) the bulk flow rate was increased 3 times, 
resulting in an ~ 2.6 increase in
x
v  (Table 6.10); and (2) the pulse input time was 
reduced to 16 hours. The influent naphthalene, Br, and DO are presented in Figure 
6.25 (a), along with the normalized naphthalene and Br data for influent and effluent, 
and normalized DO for the effluent only in Figure 6.25 (b), for the Phase IIB (2) 
experiment. In this case, the influent naphthalene and Br were relatively consistent 
during the pulse input. The effluent Br concentration was reduced compared to Phase 
I, due to the increased dispersion.  Qualitatively, the normalized effluent DO and 
naphthalene concentrations appear to be reduced compared to Phase I, suggesting 
increased biodegradation across the tank.  
Examples of the Phase IIB (2) normalized breakthrough curve data are 
presented in Figure 6.26 for sampling ports B7, C2, C7, E2 and D7. The Phase I (2) 
naphthalene breakthrough data are also presented as a point of comparison. In 
addition, the HPC data for Phase IIB are presented in Figure 6.27.  
 
 216 
 
0 100 200
0
2
4
6
8
10
12 (a)
 Naphthalene
 DO
 Br
Time (hours)
Naph
t
h
alene
 o
r
 
DO
 Conce
n
tr
ati
on 
(
m
g/L)
0
50
100
150
200
Br
 Concentratio
n
 
(m
g/L)
 
-20 0 20 40 60 80 100 120
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(b)
 
 
 Influent Naphthalene
 Influent Br
 Effluent Naphthalene
 Effluent Br
 Effluent Oxygen
Norm
a
liz
ed Concentration
Time (hours)
 
Figure 6.25 Breakthrough curves for influent and effluent from the Phase IIB (2): (a) 
Naphthalene, Br and DO concentration in the influent and (b) Normalized 
naphthalene and Br in the influent and effluent, and normalized DO in the effluent. 
 
 217 
 
0 50 100 150 200 250
0.0
0.2
0.4
0.6
0.8
1.0
(a)
B7 - High-K layer
 
 
 Phase II B Naphthalene
 Phase II B Br
 Phase II B Oxygen
 Phase I Naphthalene
Norm
a
liz
ed Concentration
Time (hours)
 
0 50 100 150 200
0.0
0.2
0.4
0.6
0.8
1.0
(b)
 Phase II B Naphthalene
 Phase II B Br
 Phase II B Oxygen
 Phase I Naphthalene
C2 - Interface
 
 
Norm
a
liz
ed Concentration
Time (hours)
 
Figure 6.26 Example breakthrough curves from Phase IIB (2) experiment for 
sampling ports: (a) B7, (b) E2, (c) C2, (d) C7 and (e) D7. Each graph shows Br tracer 
data, naphthalene data and DO data along with naphthalene data from Phase I for 
comparison.
 
 218 
 
0 50 100 150 200 250
0.0
0.2
0.4
0.6
0.8
(c)
 Phase II B Naphthalene
 Phase II B Br
 Phase II B Oxygen
 Phase I Naphthalene
C7 - Interface
 
 
Norm
aliz
e
d
 Concent
rat
i
on
Time (hours)
 
0 50 100 150 200 250
0.0
0.2
0.4
0.6
0.8
1.0
(d)
E2 - Low-K layer
 
 
 Phase II B Naphthalene
 Phase II B Br
 Phase II B Oxygen
 Phase I Naphthalene
Norm
aliz
e
d
 Concent
rat
i
on
Time (hours)
 
0 50 100 150 200 250
0.0
0.2
0.4
0.6
0.8
(e)D7 - Low-K layer
 
 
 Phase II B Naphthalene
 Phase II B Br
 Phase II B Oxygen
Norm
aliz
e
d
 Concent
rat
i
on
Time (hours)
 
Figure 6.26 Continued 
 
 219 
 
-20 0 20 40 60 80 100
4
5
6
7
HPC
 
 
 A2
 C2
 E2
 Effluent
Het
e
r
o
t
r
oph
ic P
l
ate Cou
n
t 
l
og(CFU/m
l
)
Time (hours)
 
Figure 6.27 Aqueous heterotrophic plate count data at sampling ports A2, C2, E2 and 
Effluent for Phase IIB (2).
 
 220 
 
These data illustrate the general effects of increased advection and the 
resulting increased longitudinal, transverse and interlayer dispersion on the system. 
One, due to the increased flow rate, the naphthalene and Br broke through earlier than 
in the Phase I experiment, as expected. This can clearly be seen by comparing the 
Phase I and Phase IIB naphthalene or Br breakthrough curves. Two, as a result of 
increased interlayer dispersion, there are actually two breakthroughs shown at the 
layer interface sampling ports C2 and C7: an initial mass pulse of naphthalene, Br and 
possibly oxygen due to the interlayer dispersion from the high-K layer, and a second 
mass peak resulting from advection in the low-K layer. This effect was also observed 
at sampling port D7 (Figure 6.26 (e)). No naphthalene was detected at sampling port 
in D7 in Phase I, because the naphthalene in the low-K layer was completely 
degraded before it reached the D7 sampling port. However, due to the increased 
advection in the low-K layer as well as increased dispersion from the high-K layer 
into the low-K layer, two breakthroughs occurred in D7 as well. 
Following similar trends as in Phase I, the aqueous heterotrophic plate count 
data were relatively consistent with time at sampling ports A2, C2, E2 and in the 
effluent (Figure 6.27). However, the data did show that the microbial biomass tended 
to increase slightly when the naphthalene plume passed the port and decreased slowly 
as the naphthalene traveled further downgradient, which was illustrated in all 4 
sampling ports. 
It is speculated that the increased dispersion at the edge of the plume within 
the layers, and at the interface of the different K layers, increased the microbial 
biomass access to the DO and naphthalene, resulting in more biodegradation. For 
 
 221 
 
example, as the plume moved through the high-K layer, there would have been 
increased mixing with the DO containing water in the low-K ahead of the plume in 
the that layer. This is demonstrated by the moment analysis in Table 6.10, which 
shows the naphthalene loss between C2 and C7 due to biodegradation was ?2.7 times 
greater in Phase IIB compared to Phase I, representing a significant enhancement in 
biodegradation. The magnitude of the increase in biodegradation between C2 and C7 
is equivalent in magnitude to the increase in 
x
v . That correspondence is consistent 
with the hypothesis that the increase in biodegradation was due to the increased 
supply and mixing of DO. Specifically, the flux of oxygen into the ISFC is quantified 
by ?
x
Cv , which would increase proportional to the increase in 
x
v . Similarly, the 
increase in mechanical dispersion (
xz
v? ) would also be proportional to the increase 
in 
x
v . However, the naphthalene mass loss by port C2 was reduced compared to 
Phase I, presumably due to the reduced contact time between the plume and the 
biomass in that part of the tank. 
However, it is important to note that although the increase in advection 
resulted in an increase in naphthalene plume degradation, the plume did travel further 
downgradient in a given time frame than in Phase I, as discussed above. For example, 
as shown in Figure 6.26 (b), it took 25 hours for naphthalene to break through in 
Sampling port E2 in Phase IIB (increased flow rate), while it took 90 hours for the 
naphthalene to do so in Phase I. Therefore, as discussed by MacQuarrie and Sudicky 
(1990), if the position of the naphthalene plume is more important than the time 
required for substantial mass loss, then the remediation of naphthalene with high-
velocity flow may be of more concern, even though it experiences faster mass loss. 
 
 222 
 
The general observations made in this research are consistent with those of 
Johnson (2004), who performed a model simulation of a dispersion-limited Scenario 
similar to that investigated in this Scenario #1, although under different initial 
conditions. In Johnson?s (2004) simulation, dispersion was identified as the rate-
limiting process for the baseline Phase I, based on application of the quantitative 
framework in Figure 4.4. The enhanced biodegradation of Phase II was implemented 
by numerically increasing the advection rate about 6 times. As in this work, through 
an increase in the rate of advection, the transverse dispersion and, thus, the mixing 
between the dissolved contaminant plume and ?clean? groundwater was increased 
and the overall bioremediation rate was enhanced. This was verified by a total mass 
balance estimation for the domain, which quantified an overall greater decrease in the 
total mass for enhanced biodegradation (Phase IIB), with greater advection and, 
therefore, increased dispersion. However, it is important to note that although the 
naphthalene plume degraded at a more rapid rate with increased advection, it did 
travel further in a given time frame, as observed experimentally in this work. 
Therefore, as discussed above, if the plume location is more important in a given 
situation than the time needed for substantial mass loss, increasing advection rate 
might be a concern (MacQuarrie and Sudicky 1990). 
 
 223 
 
6.3.4 Summary and Conclusion 
The hydraulic mixing at the interface of the dissolved contaminant plume and 
?clean? groundwater that is induced by subsurface hydraulic conductivity 
heterogeneities, is a key interfacial process that has been observed to impact in situ 
bioremediation (Barker et al. 1987; MacIntyre et al. 1993; Molz and Widdowson 
1988; Sutton and Barker 1985). Further, laboratory (Szecsody et al. 1994) and 
modeling (MacQuarrie and Sudicky 1990; Odencrantz 1991; Wood et al. 1994; Yang 
et al. 1994) studies have demonstrated increased microbial activity and, thus, 
enhanced biodegradation near the two-layer interface where hydraulic mixing 
between waters carrying different substrates occurs due to vertical transverse 
dispersion. Therefore, in Scenario 1, the experimental conditions were established 
such that the dimensionless numbers were as follows, 1>>
T
Pe , 1'
2
>>Sh and 
1
6
>>Da . Correspondingly, the vertical transverse dispersion was successfully 
identified experimentally as the overall rate-limiting process based on the quantitative 
framework. Based on this analysis, increased advection, and thereby increased 
vertical transverse dispersion was selected as the appropriate system perturbation and 
was observed to increase the biodegradation of naphthalene by ~ 2.7 times. 
Specifically, the increased flow rate (3 times), increased the vertical transverse 
dispersion, which probably resulted in a greater mixing of electron donor 
(naphthalene) and electron acceptor (oxygen) across the interface and, thus, an 
enhanced amount and rate of biodegradation.
 
 224 
 
6.4 Bioremediation Simulation Experiments, Scenario #2: Biodegradation-Limited 
Scenario 
As discussed in Section 5.4.4, Scenario #2 Phase I was performed to simulate 
a field scenario in which slow-biokinetics would limit the overall bioremediation rate. 
As in Scenario #1, N and P in excess of the stoichiometric requirement were added to 
enhance the biodegradation rate in Phase IIA. In Phase IIB, the biokinetics were 
increased by the removal of the high concentration of salt to alleviate the limitation, 
and, thus, enhance the overall biodegradation rate. In this Scenario, the excess of 
nutrient addition experiment (Phase IIA) was performed first, which resulted in the 
above mentioned algal problem. Therefore, the ISFC was disinfected and reinoculated 
right after the Phase IIA experiment. Following the reinoculation, the Phase IIB 
experiment was conducted to ensure the reproducibility of the ISFC compared to 
Scenario #1 Phase I experiments. Finally, the duplicated Phase I experiments were 
performed with high salinity. The results from these simulated bioremediation 
experiments are presented and discussed below.  
6.4.1 Phase I: Intrinsic Biodegradation 
As in Phase I of Scenario #1, the naphthalene and Br pulse was pumped into 
the ISFC for 48 hours, but in this case the MSNS had a high salinity as did the MSNS  
pumped in after the pulse. The influent curves for the experimental Br, naphthalene 
and DO data, are shown in Figure 6.18 (a) for Phase I (2), and the normalized Br and 
naphthalene data for the influent and effluent sampling ports, along with the DO data 
for the effluent, are shown in Figure 6.28 (b). 
 
 225 
 
 
0 100 200 300
0
2
4
6
8
10
12 (a)
 Naphthalene
 DO
 Br
Time (hours)
Naph
t
h
alene
 o
r
 
DO
 Conce
n
tr
ati
on 
(
m
g/L)
0
100
200
Br
 Concentratio
n
 
(m
g/L)
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(b)
 
 
 Influent Naphthalene
 Influent Br
 Effluent Naphthalene
 Effluent Br
 Effluent Oxygen
Norm
a
liz
ed Concentration
Time (hours)
 
Figure 6.28 Breakthrough curves for influent and effluent from the Phase I (2): (a) 
Naphthalene, Br and DO concentration in the influent and (b) Normalized 
naphthalene and Br in the influent and effluent, and normalized DO in the effluent.
 
 226 
 
The influent Br and naphthalene concentrations in the pulse were relatively 
consistent. Comparing the influent and effluent in Phase I of Scenario #2 to the 
difference between those curves in Phase I (2) of Scenario #1 (Figure 6.18), it is clear 
that the amount of biodegradation was significantly reduced by the salinity. However, 
the DO in the effluent was still quite reduced. 
Representative Phase I (2) breakthrough curves for Br, naphthalene and DO, 
are presented as a function of time for sampling ports B7, C2, C7, D1 and E2 in 
Figure 6.29. The RT3D in VMOD simulation results for the normalized naphthalene 
and Br concentrations are presented along with the experimental data as well. The 
hydro-geological parameters and the biological kinetic parameters employed in RT3D 
simulations are summarized in Tables 6.9 and 6.13, respectively. 
 
Table 6.13 Parameters utilized in RT3D simulation for Scenario #2 
Biological parameters  
Model parameters Values 
Initial microbial concentration, M
0
 0.8 mg/L 
Measured initial oxygen background concentration, A
0
 3.5 mg/L 
Maximum specific substrate utilization rate, 
m
q  
7.89E-4/hr 
Monod half-saturation constant for naphthalene, 
S
K  
1.7 mg/L  
Monod half-saturation constant for oxygen, 
A
K  
0.1 mg/L
(1)
 
Biomass produced per unit of naphthalene utilized, 
SX
Y
/
 
1.27 
Oxygen used per unit of naphthalene utilized, 
SA
Y
/
 
3
(2)
 
(1)
 Literature value from Wilson et al. (1997). 
(2)
 Calculated value based on stoichiometry, ignoring biomass synthesis. 
 
 
 227 
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
(a)
B7 - High-K Layer
 
 
 Experimental Naphthalene
 Model Naphthalene
 Experimental Br
 Model Br
 Experimental DO
Norm
a
liz
ed Concentration
Time (hours)
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
(b)
C2 - Interface
 
 
 Experimental Naphthalene
 Model Naphthalene
 Experimental Br
 Model Br
 Experimental DONorm
a
liz
ed Concentration
Time (hours)
 
Figure 6.29 Example breakthrough curves from the Phase I (2) experiment for 
sampling ports: (a) B7, (b) C2, (c) C7, (d) D1 and (e) E2. Each graph shows Br tracer, 
naphthalene, RT3D fitting data for Br and naphthalene, and DO data.
 
 228 
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
(c) C7 - Interface
 
 
 Experimental Naphthalene
 Model Naphthalene
 Experimental Br
 Model Br
 Experimental DO
Norm
aliz
e
d
 Concent
rat
i
on
Time (hours)
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
(d) D1 - Low-K layer
 
 
 Experimental Naphthalene
 Model Naphthalene
 Experimental Br
 Model Br
 Experimental DO
Norm
aliz
e
d
 Concent
rat
i
on
Time (hours)
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(e) E2 - Low-K layer
 
 
 Experimental Naphthalene
 Model Naphthalene
 Experimental Br
 Model Br
 Experimental DO
Norm
aliz
e
d
 Concent
rat
i
on
Time (hours)
 
 
Figure 6.29 Continued 
 
 
 229 
 
Several general observations were made based on the breakthrough curves 
presented in Figure 6.29. One, naphthalene and Br breakthrough occurred at an earlier 
stage of the experiment in the high-K layer, compared to the low-K layer, as expected 
and observed in Scenario #1 (e.g., compare port B7 and E2). Two, at all sampling 
ports, but in particular in the high-K layer (Sampling port B7) and at the interface 
between the two layers (Sampling ports C2 and C7), the normalized naphthalene 
concentration is only slightly lower than that of Br due to the very slow, inhibited 
biodegradation of naphthalene. Three, the normalized naphthalene concentration is 
similar to that of Br at C2 and C7, indicating little biodegradation of naphthalene 
occurred moving downgradient.  Importantly, as the plume passed a port, the oxygen 
concentrations for B7 and E2 fluctuated, while those for C2, C7 and D1, decreased 
while the naphthalene plume passed the port, presumably due to the biodegradative 
consumption of the naphthalene, suggesting that the little biodegradation that did 
occur was primarily at or near to the interface between the two layers where 
naphthalene and oxygen mixed well, which was demonstrated in Scenario #1.  
The aqueous heterotrophic plate count data were relatively consistent at 
sampling ports A2, C2, E2 and Effluent (Figure 6.30). However, the biomass 
concentration tended to decrease at all sampled ports, consistent with the inhibited 
microbial activity and, thus, decreased biomass, which further supports the 
assumption that a relatively small amount of biodegradation of naphthalene occurred 
for Phase I.  
 
 230 
 
0 100 200 300
3
4
5
6
7
8
 
 
 A2
 C2
 E2
 EFFLUNET
Heterotrophic 
Plate Count LO
G
(
CFU/m
L
)
Time (hours)
 
 
Figure 6.30 Aqueous heterotrophic plate count data at sampling ports A2, C2, E2 and 
Effluent for Phase I (2).
 
 231 
 
The simulated naphthalene and Br results using RT3D/VMOD describe 
general trends in the experimental data for the high- and low-K layers. However, 
there are some discrepancies between the observed and simulated data. First, the 
naphthalene and Br pulses broke through earlier in the experiment than predicted by 
the RT3D simulation at sampling ports B7 and C7 (Figure 6.29 (a) and (c)). Two 
factors were speculated to have contributed to this observation: (1) more water was 
forced to flow through rows B and C because of clogging of sampling port A2 that 
occurred when starting the experimental operation for Scenario #2, and thus the flow 
rate at rows B and C was increased, which in turn resulted in the earlier breakthrough 
of Br and naphthalene plume; and (2) the hydro-geological parameters were evolving 
slowly during the experimental operation due to the biomass development in the ISFC 
(Taylor and Jaffe 1990). Second, as shown in C2 and C7 (Figure 6.29 (b) and (c)), the 
experimental Br and naphthalene spread out wider than the RT3D model simulation 
data, indicating an underestimation of interlayer dispersivity, as observed in the 
control and Scenario #1 experiments. Note that in this experiment, due to the reduced 
biokinetics, the effect of dispersion on Br and naphthalene was similar, unlike 
Scenario #1 Phase I in which the biokinetics effects dominated over dispersion for 
naphthalene. Finally, the peak Br and naphthalene breakthrough concentration data 
from experiments are almost all greater than those values obtained from RT3D model 
simulation. This suggests in the case of naphthalene an underestimation of the 
inhibition effect of high salinity on the biodegradation rate (i.e., an overestimation of 
the rate of biodegradation). Furthermore, based on the Br data, it suggests an 
overestimation of dispersion in general, except at sampling port B7 in the high-K 
 
 232 
 
layer. Nonetheless, given that the independent parameter estimates utilized in RT3D 
model simulation resulted in a reasonable general description of the experimental 
trends, they were used in the quantitative framework evaluation.  
The magnitude of naphthalene loss due to biodegradation for the Phase I data 
was quantified by the moment analysis, the results of which are presented in Table 
6.14. Several key statements can be made based on the moment analysis results. One, 
as occurred in Scenario #1, the velocities computed from individual breakthrough 
curves varied slightly among different replicates, probably due to the slow 
development of biomass distributed in the flow cell (Taylor and Jaffe 1990). Two, 
much less biodegradation of naphthalene occurred in these experiments by the time 
the pulse reached sampling port C2 (14-18%) compared to in Phase I of Scenario #1 
(43-76%). Three, the extent of biodegradation increased only slightly with increasing 
distance as demonstrated by the decrease in value of A
naph
/A
Br
 from C2 to C7, which 
verifies the inhibited biodegradation. The average naphthalene mass loss between C2 
and C7 due to biodegradation (A
naph
/A
Br
 (C2) - A
naph
/A
Br
 (C7)) is only 3% for the 
duplicate experiments, which provides the baseline for comparison with the impact of 
the engineered biodegradation perturbations in Phase II. Finally, following the 
moment of the naphthalene plume further downgradient, an average of only 48% of 
the naphthalene was removed via biodegradation by the time the plume exited the 
flow cell, with an average of 52% of naphthalene remaining in the effluent. In 
comparison, there was a total removal by biodegradation of ~80% in Phase I of 
Scenario #1. Clearly, these data indicate that  biodegradation was inhibited under the 
Scenario #2 Phase I conditions
 
 233 
 
Table 6.14 Moment analysis summarization for Scenario #2 
High-K 
layer (m/d) 
 
Low-K layer 
(m/d) 
A
naph
/A
Br
 
(C2)
 
A
naph
/A
Br
 
(C7)
 
A
naph
/A
Br
(
C2) - 
A
naph
/A
Br
(
C7) 
A
naph
/A
Br
 
(Effluent) 
0.285 
Phase I (1) 
0.045 
0.86 0.83 0.03 0.57 
0.30 
Phase I (2) 
0.045 
0.82 0.79 0.03 0.47 
0.293 
Phase IIA 
0.04 
0.66 0.61 0.05 0.33 
0.30 
Phase IIB (1) 
1
 
0.035 
0.24 0.18 0.06 0.16 
0.30 
Phase IIB (2)
 2
 
0.035 
0.40 0.32 0.09 0.21 
0.28 
Phase IIB (3)
 3
 
0.03 
0.57 0.48 0.09 0.20 
1 
Same as Scenario #1 Phase I (1) experiment, 
2
 Same as Scenario #1 Phase I (2) experiment, 
3
 Same as Scenario #1 Phase I (3) experiment. 
 
 
 234 
 
Following the approach of Scenario #1, to verify that slow biodegradation was 
the overall rate-limiting process for the system, the quantitative framework in Figure 
4.4 was applied using the parameter values in Table 6.8 and 6.14, and other system 
parameters as summarized in Table 6.15. The calculated dimensionless numbers are 
presented in Table 6.16. Again as discussed in Scenario #1, the sorption mass transfer 
rate was assumed to be infinite compared to the transverse dispersion rate. Based on 
the quantitative comparison in Table 6.16, biokinetics and transverse vertical 
dispersion were identified as the overall rate-limiting processes. Therefore, it was 
decided to stimulate overall bioremediation by increasing the slow biokinetics 
(MacQuarrie and Sudicky 1990), given that the effect of increasing dispersion had 
already been investigated in Scenario #1. 
 
 235 
 
Table 6.15 Parameters for quantitative framework evaluation for Scenario #2 
Parameters High-K layer Low-K layer 
Characteristic length, L, m 
(=Vertical thickness of model aquifer) 
0.3 0.3 
Average pore water velocity, 
x
v , m/d 
0.293 0.033 
Vertical transverse dispersivity, 
Z
? , m 
7.6E-5 0.002 
Aqueous molecular diffusion coefficient, 
m
D , m
2
/d 
2E-5 2E-5 
Tortuosity of the medium, ?   0.78 0.70 
Effective molecular diffusion coefficient, 
D*, m
2
/d 
1
 
1.56E-5 1.4E-5 
Yield coefficient, 
sx
Y
/
,  
mg biomass /mg naphthalene 
2
 
1.27 1.27 
Maximum specific substrate utilization 
rate, 
s
x
Y
q
max
max
?
= , 1/d 
0.019 0.019 
Initial biomass concentration, M
0
,  
mg biomass/L 
0.8 0.8 
Initial naphthalene substrate concentration, 
S
0
, mg/L 
10.0 10.0 
1
m
DD ?=* , where
m
D is calculated using the equation by Wilke and Chang (Welty et al. 
1984), ? is estimated as 
3
1
?? =  (Schwartz and Zhang 2002). 
2 
Conversion factors: 1 mg naphthalene is equivalent to 3mg naphthalene COD and 1 mg 
biomass is equivalent to 1.42 mg biomass COD based on the stoichiometric equation. 
 
 236 
 
Table 6.16  Quantitative framework used to identify the rate-limiting process 
Symbol  High-K Layer Low-K Layer 
Pe
t
 (Transverse Peclet No.) 
= 
z
x
D
Lv
 
Pe
t
 = 2325 >>1: 
transverse dispersion 
limits 
Pe
t
 = 124 >>1: transverse 
dispersion limits 
Sh
2
? (Modified Sherwood No. 2) 
= 
z
ls
D
LK
2
 
Sh
2
?=?>>1: transverse dispersion limits 
Da
6
 (Damk?hler No. 6) 
= 
z
DS
LMq
0
2
0max
 
Da
6
 = 0.2<3.6<5: both 
dispersion and 
biokinetics limits 
Da
6
 = 0.2<1.8<5: both 
dispersion and biokinetics 
limits 
 
6.4.2 Phase IIA: Nutrient Enhanced Biodegradation 
Before the predicted appropriate perturbation was investigated, the addition of 
excess of N and P was tested in Phase IIA, which was performed once. As in Scenario 
#1, the N and P content of the MSNS was increased to 3 times the stoichiometric 
requirement for 10 mg/L naphthalene. The influent data and the normalized influent 
and effluent data are given in Figure 6.31 (a) and (b) respectively. The resulting 
breakthrough curves for Br, naphthalene and DO are presented in Figure 6.32. For 
comparison purposes, the Phase I (2) naphthalene data are plotted in Figure 6.32 as 
well. There are both similarities and differences between the results of Phase I and 
Phase IIA. With regard to the similarities, the general concentration trends for Br, 
naphthalene and DO, for Phase IIA, are very similar to those in Phase I experiments: 
the naphthalene concentration was reduced somewhat relative to Br, and the DO 
concentration decreased when the naphthalene broke through. 
 
 237 
 
0 100 200
0
2
4
6
8
10
12
(a)
 Naphthalene
 DO
 Br
Time (hours)
Naph
t
h
alene
 o
r
 
DO
 Conce
n
tr
ati
on 
(
m
g/L)
0
100
200
Br
 Concentratio
n
 
(m
g/L)
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(b)
 
 
 Phase II A Influent Naphthalene
 Phase II A Influent Br
 Phase II A Effluent Naphthalene
 Phase II A Effluent Br
 Phase II A Effluent Oxygen
 Phase I(2) Effluent Naphthalene
Norm
a
liz
ed Concentration
Time (hours)
 
Figure 6.31 Breakthrough curves for influent and effluent from the Phase IIA: (a) 
Naphthalene, Br and DO concentration in the influent and (b) Normalized 
naphthalene and Br in the influent and effluent, and normalized DO in the effluent. 
.
 
 238 
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
(a)B7 - High-K layer
 
 
 Phase II A Naphthalene
 Phase II A Br
 Phase II A Oxygen
 Phase I Naphthalene
Norm
a
liz
ed Concentration
Time (hours)
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
(b)
C2 - Interface
 
 
 Phase II A Naphthalene
 Phase II A Br
 Phase II A Oxygen
 Phase I Naphthalene
Norm
a
liz
ed Concentration
Time (hours)
 
 
Figure 6.32 Example breakthrough curves from Phase IIA experiment for sampling 
ports: (a) B7, (b) C2, (c) C7, (d) D1 and (e) E2. Each graph shows Br tracer data, 
naphthalene data, DO data and naphthalene data from Phase I (2) for comparison. 
 
 239 
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
(c)C7 - Interface
 
 
 Phase II A Naphthalene
 Phase II A Br
 Phase II A Oxygen
 Phase I Naphthalene
Norm
aliz
e
d
 Concent
rat
i
on
Time (hours)
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
(d) D1 - Low-K layer
 
 
 Experimental Naphthalene
 Model Naphthalene
 Experimental Br
 Model Br
 Experimental DO
Norm
aliz
e
d
 Concent
rat
i
on
Time (hours)
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
(e)E2 - Low-K layer
 
 
 Phase II A Naphthalene
 Phase II A Br
 Phase II A Oxygen
 Phase I Naphthalene
Norm
aliz
e
d
 Concent
rat
i
on
Time (hours)
 
 
Figure 6.32 Continued 
 
 240 
 
With respect to the differences, when compared to the values in Phase I 
experiments, although the naphthalene concentrations were similar in the two sets of 
experiments at sampling ports C2 and C7, they were lower in Phase IIA at sampling 
B7 (high-K layer) and higher in Phase IIA at D1 and E2 (low-K layer). This 
phenomenon was speculated to be due to differences that resulted from replicate 
operations. As described in Section 5.1.3, there were clogging problems after the run 
of Phase IIA of Scenario #2 due to the algae growth in the ISFC, which resulted from 
the addition of excess N and P. As a result, the simulated aquifer in the ISFC was 
disinfected and reinoculated again, which may have contributed to some of the 
differences that occurred between the Phase IIA and the Phase I experiments of 
Scenario #2, although they were performed under the same experimental conditions 
except for the nutrient concentrations. However, although the naphthalene removal 
was not greater at all sampling ports, the overall amount of bioremediation was 
improved in Phase IIA compared to Phase I, as demonstrated by the moment analysis 
in Table 6.14. Although the Phase IIA experiment was only performed once, the 
naphthalene mass loss due to biodegradation by the time the pulse reached sampling 
port C2 increased from less than 20% in Phase I to over 30% in Phase IIA. In 
addition, during the transport between C2 and C7 the loss due to biodegradation was 
5%, which represented an increase of 2% compared to the Phase I average 
naphthalene loss. Furthermore, the naphthalene remaining in the effluent was reduced 
to 33%, indicating an increase of 19% in naphthalene removal compared to Phase I. 
The outcome of increased biodegradation is not surprising considering that the N and 
P concentrations in the original MSNS artificial groundwater were 60% of the 
 
 241 
 
stoichiometric amount required for the biodegradation of naphthalene with a 
concentration of 10 mg/L. Thus, it is speculated that under the biokinetics limited 
conditions, as determined by the quantitative framework, the added N and P nutrients 
enhanced the inhibited biodegradation and resulted in a moderate increase in the 
biokinetics rate. However, the removal of naphthalene due to biodegradation in Phase 
IIA did not reach that observed in Phase I of Scenario #1 (the same conditions 
without high salinity), because the slow biokinetics was also a result of the high salts 
concentration, not only due to the nutrients limitation. Therefore, it was expected the 
overall rate of bioremediation could be further enhanced if an appropriate approach 
was implemented to alleviate the rate-limiting process of slow biokinetics, which was 
investigated in Phase IIB. 
The trend in the aqueous heterotrophic plate count concentrations at sampling 
ports A2, C2 and Effluent were relatively consistent (Figure 6.33). In all cases, the 
HPC tended to decrease until the peak of the naphthalene pulse had passed, after 
which the cell counts increased again. This was especially accentuated with sampling 
port E2 (Figure 6.33). The HPC concentrations at sampling port E2 decreased rapidly, 
apparently due to the shortage of naphthalene substrate and then increased slowly 
once the naphthalene arrived at the sampling port.  
 
 
 242 
 
0 100 200 300
1
2
3
4
5
6
7
 
 
 Phase IIA A2
 Phase IIA C2
 Phase IIA E2
 Phase IIA Effluent
Heterotrophi
c P
l
a
t
e
 
Count LO
G
(
CPU/m
L
)
Time (hours)
 
Figure 6.33 Aqueous HPC at sampling ports A2, C2, E2 and Effluent for Phase IIA. 
 
 243 
 
6.4.3 Phase IIB: Engineered Biodegradation 
As discussed above, an engineered perturbation that was expected to be 
successful for Scenario #2 was increasing the biokinetics so as to alleviate the 
overall-rate limiting process and stimulate the biodegradation rate. Therefore, all 
experiments were conducted under the same conditions as Phase I except that the 
high salinity was removed. In practical terms, this change could represent the 
reduction of salinity below inhibitory levels by addition of water (Riis et al. 2003). 
Because the conditions for Phase I of Scenario #2 without salinity are the same as 
Scenario #1 Phase I, two of those experiments (Scenario #1 Phase I (1) and (2)) were 
used as replicates for comparison to the Scenario #2 Phase IIB (3) experiment, which 
was performed after the reinoculation, yet before the salt addition. 
The influent data and the normalized influent and effluent data for Phase IIB 
(2) (also Phase I (2) of Scenario #1) are shown in Figure 6.34 (a) and (b), 
respectively. The influent Br and naphthalene concentrations were relatively steady, 
with some slight fluctuations during the pulse input. The effluent naphthalene 
concentrations were reduced compared to the Phase I (2) data presented, which 
coupled with the reduced DO in the effluent indicates increased naphthalene 
biodegradation had occurred. The aqueous heterotrophic plate count data were 
relatively consistent with time at sampling ports A2, C2, E2 and in the effluent 
(Figure 6.17). 
Examples of the Phase IIB (2) breakthrough curve data are presented in Figure 
6.35 for sampling ports B7, C2, C7, D1 and E2, which illustrate the general effects of 
increased biokinetics effects on the system. For comparison purposes, the Phase I (2) 
 
 244 
 
naphthalene breakthrough data are shown as well. Several observations can be made 
by comparing the Phase IIB and Phase I data. One, with the exception of D1, the 
naphthalene and Br concentrations broke through later in the Phase IIB experiment 
than in Phase I (2), especially at ports C7 and E2. This was probably because of the 
biomass development in the ISFC and the differences that resulted from reinoculation 
of ISFC, as discussed in Phase IIA. Two, due to the increased biokinetics, the 
normalized naphthalene concentration in Phase IIB was much lower than that of 
Phase I. Importantly, the effect of increased biokinetics on the overall bioremediation 
is quantified by the moment analysis in Table 6.14. This is seen in the increased loss 
of naphthalene due to biodegradation by the time at which the naphthalene pulse 
reached sampling port C2 (43-76% removal in Phase IIB compared to 34% in Phase 
IIA and 14-18% in Phase I). In addition, the naphthalene loss between C2 and C7 due 
to biodegradation was ?2.7 times greater in Phase IIB compared to Phase I and 1.6 
times greater than Phase IIA, representing a significant enhancement in 
biodegradation. This result is further verified by the moment analysis in the effluent, 
which had only an average of 19% of naphthalene remaining in the effluent when the 
plume exited the ISFC in Phase IIB, while there was 52% of naphthalene remaining 
in the effluent in Phase I, and 33% in Phase IIA. 
 
 245 
 
0 100 200
0
2
4
6
8
10
12
(a)
 Naphthalene
 DO
 Br
Time (hours)
Naph
t
h
alene
 o
r
 
DO
 Conce
n
tr
ati
on 
(
m
g/L)
0
100
200
Br
 Concentratio
n
 
(m
g/L)
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(b)
 
 
 Phase IIB Influent Naphthalene
 Phase IIB Influent Naphthalene
 Phase IIB Effluent Naphthalene
 Phase IIB Effluent Br
 Phase IIB Effluent Oxygen
 Phase I(2) Effluent Naphthalene
Norm
a
liz
ed Concentration
Time (hours)
 
Figure 6.34 Breakthrough curves for influent and effluent from the Phase IIB (2): (a) 
Naphthalene, Br and DO concentration in the influent and (b) Normalized 
naphthalene and Br in the influent and effluent, and normalized DO in the effluent. 
 
 246 
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(a)B7 - High-K layer
 
 
 Phase IIB Naphthalene
 Phase IIB Br
 Phase IIB Oxygen
 Phase I Naphthalene
Norm
a
liz
ed Concentration
Time (hours)
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(b)C2 - Interface
 
 
 Phase IIB Naphthalene
 Phase IIB Br
 Phase IIB Oxygen
 Phase I Naphthalene
Norm
a
liz
ed Concentration
Time (hours)
 
Figure 6.35 Example breakthrough curves from Phase IIB (2) experiment for 
sampling ports: (a) B7, (b) C2, (c) C7, (d) D1, (e) E2 and (f) Effluent. Each graph 
shows Br tracer data, naphthalene data and DO data along with naphthalene data from 
Phase I (2) for comparison. 
 
 247 
 
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(c)
C7 - Interface
 
 
 Phase IIB Naphthalene
 Phase IIB Br
 Phase IIB Oxygen
 Phase I Naphthalene
Norm
aliz
e
d
 Concent
rat
i
on
Time (hours)
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
(d)D1 - Low-K layer
 
 
 Phase IIB Experimental Naphthalene
 Phase IIB Experimental Br
 Phase IIB Experimental DO
 Phase I Experimental Naphthalene
Norm
al
iz
ed Concent
r
a
t
i
on
Time (hours)
0 100 200 300
0.0
0.2
0.4
0.6
0.8
1.0
(e)
E2 - Low-K layer
 
 
 Phase IIB Naphthalene
 Phase IIB Br
 Phase IIB DO
 Phase I Naphthalene
Normaliz
ed Concentrat
i
on
Time (hours)
 
Figure 6.35 Continued
 
 248 
 
 The Phase IIB (3) was the only replicate of Phase IIB performed after 
reinoculation the ISFC. The data sets for Phase IIB (3) are presented in Figure 6.36, 
6.37 and 6.38 to illustrate that the experimental results were comparable to those 
obtained before reinoculation. As in Phase IIB (2), the influent data, normalized 
influent and effluent data, breakthrough curves for sampling ports B7, C2, C7, D1, E2 
and Effluents, along with the HPC data for sampling ports A2, C2, E2 and Effluents 
are shown. In general, similar breakthrough trends were observed in Phase IIB (3), as 
in Phase IIB (2).  However, there were some discrepancies between Phase IIB (2) and 
Phase IIB (3) when compared to the results from Phase I (2). First, there was more 
oxygen loss in most sampling ports due to the naphthalene biodegradation 
consumption in Phase IIB (3) than in Phase IIB (2). The oxygen concentration in 
Phase IIB (3) dropped to ~ 40% of the influent concentration, but the decrease in 
Phase IIB (2) only reached ~ 50-60% of the influent oxygen concentration. This was 
especially accentuated with sampling port E2 (Figure 6.37 (f)), at which the oxygen 
concentration decreased to 0 mg/L. This observation further supported the enhanced 
biodegradation because of the removal of salt.  Interestingly, both the Br and 
naphthalene broke through much later in Phase IIB (3) than in Phase I (2) at sampling 
ports from low-K layer (sampling ports D1 and E2), especially at sampling port E2, 
for some unknown reasons . This observation only occurred in Phase IIB (3) once 
during all these experiments. Despite the above discussed discrepancies, the same 
conclusions could be made based on the moment analysis (Table 6.14). 
 
 
 249 
 
 
0 100 200 300 400 500
0
2
4
6
8
10
12
(a)
 Naphthalene
 DO
 Br
Time (hours)
Naph
t
h
alene
 o
r
 
DO
 Conce
n
tr
ati
on 
(
m
g/L)
0
100
200
Br
 Concentratio
n
 
(m
g/L)
 
0 100 200 300 400 500
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(b)
 
 
 Influent Naphthalene
 Influent Br
 Effluent Naphthalene
 Effluent Br
 Effluent Oxygen
Norm
a
liz
ed Concentration
Time (hours)
 
Figure 6.36 Breakthrough curves for influent and effluent from the Phase IIB (3): (a) 
Naphthalene, Br and DO concentration in the influent and (b) Normalized 
naphthalene and Br in the influent and effluent, and normalized DO in the effluent 
 
 250 
 
0 100 200 300 400 500
0.0
0.2
0.4
0.6
0.8
1.0
(a)B7 - High-K layer
 
 
 Phase IIB Naphthalene
 Phase IIB Br
 Phase IIB Oxygen
 Phase I Naphthalene
Norm
a
liz
ed Concentration
Time (hours)
 
0 100 200 300 400 500
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(b)C2 - Interface
 
 
 Phase IIB Naphthalene
 Phase IIB Br
 Phase IIB Oxygen
 Phase I Naphthalene
Norm
a
liz
ed Concentration
Time (hours)
 
Figure 6.37 Example breakthrough curves from Phase IIB (3) experiment for 
sampling ports: (a) B7, (b) C2, (c) C7, (d) D1, (e) E2 and (f) Effluent. Each graph 
shows Br tracer data, naphthalene data and DO data along with naphthalene data from 
Phase I (2) for comparison. 
 
 
 251 
 
0 100 200 300 400 500
0.0
0.2
0.4
0.6
0.8
1.0
(c)C7 - Interface
 
 
 Phase IIB Naphthalene
 Phase IIB Br
 Phase IIB Oxygen
 Phase I Naphthalene
Norm
aliz
e
d
 Concent
rat
i
on
Time (hours)
0 100 200 300 400 500
0.0
0.2
0.4
0.6
0.8
1.0
(d)D1 - Low-K layer
 
 
 Phase IIB Experimental Naphthalene
 Phase IIB Experimental Br
 Phase IIB Experimental DO
 Phase I Experimental Naphthalene
Norm
al
iz
ed Concent
rat
i
on
Time (hours)
0 100 200 300 400 500
0.0
0.2
0.4
0.6
0.8
1.0
1.2
(e)E2 - Low-K layer
 
 
 Phase IIB Naphthalene
 Phase IIB Br
 Phase IIB DO
 Phase I Naphthalene
Normaliz
ed Concentrat
i
on
Time (hours)
 
Figure 6.37 Continued
 
 252 
 
0 100 200 300 400
3
4
5
6
7
 
 
 Phase I A2
 Phase I C2
 Phase I E2
 Phase I Effluent
Heterotrophic 
p
l
ate counti
ng Log(CF
U/
m
L
)
Time (hours)
 
 
Figure 6.38 Aqueous heterotrophic plate count data at sampling ports A2, C2, E2 and 
Effluent for Phase IIB (3).
 
 253 
 
The general observations made in this research are consistent with those of 
Johnson (2004), who performed a model simulation of a biokinetics-limited Scenario 
similar to that investigated in this work. In Johnson?s (2004) simulation, the biokinetics 
limited case of Phase I was simulated by reduced kinetics under denitrifying conditions 
and the enhanced biodegradation of Phase II was implemented by greater biokinetics 
under aerobic conditions. As in this work, there was a high-K layer overlying a low-K 
layer and a pulse of naphthalene entered the domain. Specifically, the overall 
biodegradation was enhanced due to the increased biokinetics, i.e., the greater specific 
substrate utilization rate. This was verified by  a total mass balance estimation in the 
domain, which quantified that the initial removal rate of naphthalene was significantly 
greater early in the model simulation 0.980 mg/d of enhanced biodegradation (Phase IIB) 
compared to in the baseline biodegradation (Phase I) (0.492 mg/d). This observation was 
explained by the initial proximity of the electron donor substrate (naphthalene) to the 
electron acceptor along the injection front and, thus, the increased utilization rate in the 
enhanced phase (Phase IIB). Furthermore, the enhanced degradation was confirmed by a 
review of the electron acceptor mass data. The electron acceptor (nitrate) under the slow 
biokinetics (Phase I) increased due to the mass loading from the injection front, although 
experiencing a mass loss due to the advection and dispersion out of the domain. 
However, the electron acceptor mass data (oxygen) under the fast biokinetics conditions 
(Phase IIB) decreased significantly due to the greater consumption of naphthalene 
biodegradation.  
Finally, as noted in Section 6.4.1, based on the quantitative framework analysis, 
both biokinetics and vertical transverse dispersion were predicted to be the rate-limiting 
 
 254 
 
processes under the conditions of Scenario #2 Phase I. This conclusion is supported the 
results of Scenario #2 combined with the results of Scenario #1. In Scenario #2, removing 
the biokinetics limitation due to the high salinity resulted in an increase in the removal of 
naphthalene via biodegradation between C2 and C7 from 3% removal in Phase I to 8% 
removal in Phase IIB. The conditions of Phase IIB of Scenario #2 were equivalent to 
those of Scenario #1 Phase I. Alleviating the dispersion limitation in Scenario #1 by 
increasing the advection and, thereby, dispersion, resulted in an increase in the removal 
of naphthalene via biodegradation between ports C2 and C7 of 8% in Phase I to 
approximately 22 % in Phase II. Therefore, these data demonstrate successful application 
of the quantitative framework in a situation where more than process was overall rate-
limiting.  
6.4.4 Summary and Conclusion 
In general, it has been stated that the biodegradation kinetics of easily degraded 
organic contaminants do not control the mass loss rates observed in groundwater 
bioremediation (Bosma et al. 1997). Rather, it is often the rate of advective transport and 
dispersive mixing (macro-scale) of the contaminants and other substrates (e.g., DO) and 
the sorption (meso-scale) of the organics that have a greater influence on the actual mass 
loss by biodegradation (MacQuarrie and Sudicky 1990). However, the conditions in this 
scenario represented a case with slow biokinetics, e.g., due to inhibitory conditions or 
difficult-to-biodegrade organic contaminants. For the experimental conditions of 
Scenario #2 Phase I, with high salinity, the slow biokinetics were successfully identified 
as the one of the overall rate-limiting processes, based on the system parameter estimates 
and application of the quantitative framework of dimensionless numbers. Removal of the 
 
 255 
 
high salinity in Phase IIB illustrated that the kinetics of biodegradation have a strong 
influence on the mass loss of the naphthalene plume, as demonstrated by a mass loss 
increase of 2.7 times with the increased biokinetics. The addition of limiting nutrients (N 
and P) in Phase IIA had made a moderate positive effect on biodegradation compared to 
in Phase I, consistent with the determination that biokinetics were one of the overall rate-
limiting processes. 
 
 256 
 
Chapter 7  Conclusions and Recommendations 
7.1 Summary and Conclusions 
This research investigated the impact of substrate heterogeneity and the relevant 
hydro-geochemical interfaces on in situ bioremediation. The scales of heterogeneities and 
associated interfacial processes were utilized as an organizational tool, and dimensionless 
numbers were used to capture the complexity of the interactions between these 
subsurface physical, chemical and biological processes. It was hypothesized that a 
quantitative framework using the scales of heterogeneities as an organizing principle and 
based on a set of dimensionless numbers could be developed to define the limits of in situ 
bioremediation. The overall goal of this study was to refine and evaluate such a 
quantitative framework under different simulated scenarios relevant to the subsurface. 
To accomplish the overall goal, the developed quantitative framework was 
evaluated under three scenarios relevant to the subsurface: a macro-scale dispersion 
limited scenario, a micro-scale biokinetics limited scenario and a meso-scale 
sorption/desorption limited scenario. In each scenario, the quantitative framework was 
used to predict the overall limiting process and what engineering actions, if any, would 
positively impact the in situ biodegradation rates. 
This integrated modeling and experimental investigation had three major specific 
objectives and corresponding activities. First, the quantitative framework based on the 
systematic comparison of a set of dimensionless numbers was developed to compare the 
phenomena occurring at different scales and to predict the overall rate-limiting process 
for a given simulated scenarios. Second, the two-layered ISFC was built to represent a 
realistic, yet experimentally tractable natural subsurface system for evaluating the 
 
 257 
 
physical, chemical, and microbiological processes occurring at heterogeneous interfaces. 
Three, the bioremediation simulation experiments were conducted for three specifically 
selected scenarios. The mathematical model, RT3D/VMOD, was chosen to verify the 
independently estimated hydro-geological and biological parameters and to simulate the 
experimentally investigated scenarios in the ISFC.  
A critical component of the experimental evaluation of the quantitative 
framework was the independent parameter estimation experiments that were conducted to 
measure the kinetic parameters describing groundwater flow and contaminant fate and 
transport in the simulated subsurface system of the ISFC. Specifically, the hydraulic 
conductivity, dispersivity within each layer and between the two layers, 
sorption/desorption kinetics and biodegradation kinetics parameters were estimated. The 
independently determined hydrogeologic parameters summarized when used in 
RT3D/VMOD were able to accurately describe the experimental data obtained in the 
Control Bioremediation Simulation Experiments (Section 6.2.1), verifying the use of 
these parameters for modeling of the bioremediation simulation experiments. Although 
there were some discrepancies between the experimental and modeling data, these could 
be explained. The control experimental data also demonstrated the interlayer mass 
transfer phenomena and were used to fit the interlayer mass transfer coefficient, 
z
'? , 
which signifies the importance of interlayer mass transfer that resulted from 
the K heterogeneities. The control experiment also demonstrated there was no 
biodegradation of naphthalene in the ISFC prior to inoculation and the naphthalene 
transport in the ISFC responded as expected when excess of N and P were added or when 
the flow rate was increased. 
 
 258 
 
The first bioremediation simulation experiment, Scenario #1 simulated 
dispersion-limited in situ bioremediation. The hydraulic mixing at the interface of the 
dissolved contaminant plume and ?clean? groundwater that is induced by subsurface 
hydraulic conductivity heterogeneities is a key interfacial process that has been observed 
to impact in situ bioremediation (Barker et al. 1987; MacIntyre et al. 1993; Molz and 
Widdowson 1988; Sutton and Barker 1985). Further, laboratory (MacQuarrie and 
Sudicky 1990; Szecsody et al. 1994) and modeling (MacQuarrie and Sudicky 1990; 
Odencrantz 1991; Wood et al. 1994; Yang et al. 1994) studies have demonstrated 
increased microbial activity and, thus, enhanced biodegradation near the two-layer 
interface where hydraulic mixing between waters carrying different substrates occurs due 
to vertical transverse dispersion. Therefore, in Scenario 1, the experimental conditions 
were established such that the dimensionless numbers were as follows, 1>>
T
Pe , 
1'
2
>>Sh and 1
6
>>Da . Correspondingly, the vertical transverse dispersion was 
successfully identified experimentally as the overall rate-limiting process based on the 
quantitative framework. Based on this analysis, increased advection, and thereby 
increased vertical transverse dispersion was selected as the appropriate system 
perturbation and was observed to increase the biodegradation of naphthalene by ~ 2.7 
times. Specifically, the increased flow rate (3 times), increased the vertical transverse 
dispersion, which probably resulted in a greater mixing of electron donor (naphthalene) 
and electron acceptor (oxygen) across the interface and, thus, an enhanced amount and 
rate of biodegradation. 
The second bioremediation simulation experiment represented biokinetics-limited 
in situ bioremediation. Although it is often the rate of advective transport and dispersive 
 
 259 
 
mixing (macro-scale) of the contaminants and other substrates (e.g., DO) and the sorption 
(meso-scale) of the organics that have a greater influence on the actual mass loss by 
biodegradation (MacQuarrie and Sudicky 1990), as demonstrated in Scenario #1, the 
conditions in Scenario #2 represented a case with slow biokinetics, e.g., due to inhibitory 
conditions or difficult-to-biodegrade organic contaminants. For the experimental 
conditions of Scenario #2 Phase I, with high salinity, the dimensionless numbers were 
such that 1>>
T
Pe , 1'
2
>>Sh and 52.0
6
<< Da , indicating that both the dispersion and 
the slow biokinetics were the overall rate-limiting processes. The engineered remedial 
strategy to alleviate the limits of dispersion was investigated in Scenario #1. Therefore, 
the engineered perturbation to alleviate the slow biokinetics was studied in Scenario #2. 
Removal of the high salinity in Phase IIB of Scenario #2 illustrated that the kinetics of 
biodegradation have a strong influence on the mass loss of the naphthalene plume, as 
demonstrated by a mass loss increase of 2.7 times with the increased biokinetics. The 
addition of limiting nutrients (N and P) in Phase IIA had made a moderate positive effect 
on biodegradation compared to in Phase I, consistent with the determination that 
biokinetics were the overall rate-limiting process. 
Finally, in conclusion, this research evaluated the developed quantitative 
framework and confirmed via an integrated modeling and experimental investigation that 
it could be used successfully to predict the rate-limiting process for in situ bioremediation 
and to aid in selecting appropriate engineered perturbation(s) to enhance the rate of 
biodegradation. This represents the first step toward the development of a useful 
quantitative decision-making tool for practitioners in the field that does not require 
sophisticated numerical modeling tools but can be used to make decisions based on the 
 
 260 
 
systematic comparison of dimensionless numbers calculated using  parameters obtained 
in field and laboratory testing. 
7.2 Recommendations for Future Work 
As discussed above, the developed quantitative framework was successfully 
evaluated in the laboratory in the ISFC for two different scenarios: a macro-scale 
dispersion-limited scenario, and a micro-scale biokinetics-limited scenario. However, due 
to time limitation, the proposed third sorption/desorption-limited scenario was not 
accomplished completely, with only the sorption/desorption kinetics parameter 
estimation experiments and model predictions implemented so far. Therefore, given the 
importance of sorption limitations on bioavailability in the field, it is recommended that 
the unfinished Scenario #3 be completed to further evaluate the application of the 
quantitative framework under different conditions. This experimental investigation is 
currently in progress, with the work primarily performed by Ms. Eunyoung Hong, with 
some assistance from the author in setting up the experiments and analyzing samples. It is 
also recommended that the quantitative framework be evaluated in the laboratory using 
the ISFC under other key conditions that are relevant in situ bioremediation of the natural 
subsurface, in particular when NAPL dissolution is the overall rate-limiting process 
(Seagren et al. 1994), or when oxygen supply to the saturated zone is the rate-controlling 
factor (Borden and Bedient 1986; Borden et al. 1986). 
With respect to the laboratory-scale evaluation of the quantitative framework, one 
of the key aspects of the testing is the parameter estimation. Two of the parameter 
estimation techniques that needed be improved are the biokinetics and dispersivity 
determinations. As discussed in Section 5.3.4, Monod kinetic parameters estimated using 
 
 261 
 
non-steady-state batch analysis are subject to large uncertainties. These uncertainties are 
a function of the initial experimental conditions, the values of the parameters, the type 
and magnitude of the measurement errors, and the number of samples (Liu and Zachara, 
2001). Careful manipulation of experimental conditions was required to reduce the 
correlations between Monod parameters allowing for the estimation of unique parameters 
with the lowest degree of uncertainty. The estimation of unique biokinetics parameters in 
this study was made challenging due to the above mentioned uncertainties. For example, 
only few sets of respirometry experimental results proved to be valid, based on the initial 
estimation techniques used by Brown, et al. (1990). Thus, more replicate experiments 
need to be performed and more efforts are required for better manipulation of the 
experimental conditions for the respirometry studies and the model fitting techniques in 
order to obtain unique biokinetics parameters. 
In addition, as indicated by the results of the Bioremediation Simulation 
Experiments, Section 6.2-6.4, the hydro-geological parameters probably evolved slowly 
in the ISFC due to the slow development of biomass distributed in the flow cell over time 
(Taylor and Jaffe 1990). This may have contributed to some discrepancies between the 
model simulations and experimental data, which, in turn, could negatively affect the 
evaluation of the quantitative framework. Therefore, it is suggested that a tracer study 
(dispersivity within each layer) and a control experiment (dispersivity between two 
layers) be performed at appropriate time intervals in between scenarios to investigate the 
effect of biomass distribution on these parameters. 
The final step in the evaluation of the application of the quantitative framework is 
to test its application in the field. Of course, the field scale is expected to be more 
 
 262 
 
complex than the ISFC experiments of this work. Furthermore, given the heterogeneities 
in the field and possible changes in mass-transfer and biokinetic parameters over time, 
the assessment of the rate-limiting phenomenon at field scale will need to be checked 
spatially and temporally (Ramaswami and Luthy 1997; Sturman et al. 1995). One way to 
perform a field-scale evaluation is to examine well-studied sites, such as the Canadian 
Forces Base (CFB) Borden sites, e.g., (Schirmer and Barker 1998; Schirmer et al. 1999; 
Schirmer et al. 2000), that are well documented with sufficient data, including the hydro-
geological and biokinetics parameters needed to evaluate the quantitative framework. In 
addition, research is currently underway at field site investigation to evaluate the 
quantitative framework at the field scale. That study is being performed by Matt 
Wheaton, a M.S. student in the Civil and Environmental Engineering Department, 
University of Maryland, College Park. Successful implementation of this study at the 
field scale would contribute greatly to the evaluation of the quantitative framework. 
Testing of the quantitative framework using compiled field data, represents another step 
toward the development of a useful quantitative tool for defining in the field when in situ 
bioremediation will work, and when an engineered or intrinsic approach is best. 
 
 263 
 
Appendix 
Appendix 1: Materials and Parts for ISFC  
Materials: 
1. Glass plates (untempered): ?" think,  
2 plates "
4
1
52"
4
3
16 ? ; 
2 plates "
4
3
16"5 ? ; 
1 plate "
4
1
52"5 ? ; 
1 plate "
4
1
53"5 ? ; 
Cart material: 
2. 4 Thermoplastic rubber wheels, cushion-load cold forged caster swivel, 6" ? 2"; 
McMaster-Carr Supply Company, Part number: 2703T61 
3. 4 Brakes for 6" Diameter thermoplastic rubber wheel, 
McMaster-Carr Supply Company, Part number: 2703T33 
Screen material:  
4. Type 316 Stainless Steel Woven Wire Cloth 150 ?150 Mesh, .0026" Wire Dia, 12" ? 
24" Sheet; 
McMaster-Carr Supply Company, Part number: 9319T52 
5. Type 316 Stain Steel Sheet With #2B Finish .075" Thick, 12" ? 24";  
McMaster-Carr Supply Company, Part number: 88885K25 
6. 40 Machine screw nuts (Hex, 18-8 stainless-steel), 6-32 coarse; 
 
 264 
 
Physics Shop, University of Maryland, College Park 
Tubing and Fittings: 
7. 1 Bulkhead Unions (Teflon PFA), Tubing O.D.: "
8
3
;  
Cole-Parmer Instrument Company, Catalog No: A-06482-34 
8. 2 White Teflon Flat Washer, "
2
1
screw size, 0.500" ID, 1.003" OD, 0.057" thick; 
MacMaster-Carr Supply Company, Part number: 95630 A248 
9. Teflon PTFE tubing, I.D.: "
64
19
, O.D.: "
8
3
;  
Cole-Parmer Instrument Company, Catalog No: A-06407-46 
10. 1 Union Reducing, "
4
1
"
8
3
? ; 
Cole-Parmer Instrument Company, Catalog No: A-06373-72 
11. Teflon PTFE tubing, I.D.: "
16
3
, O.D.: "
4
1
; 
Cole-Parmer Instrument Company, Catalog No: A-06407-44 
Sampling port: 
12. 1 Branch Tee, Teflon PFA, Male "
8
1
"
4
1
? ; 
Cole-Parmer Instrument Company, Catalog No: A-06374-52 
13. 1 Mininert Valve "
8
1
Female NPT to Mininert Septum, 
Alltech Associate, Inc., Part Number: 631204  
14. Pump, Mflex, L/S, PTFE-Tubing, 115 V,  
Cole-Parmer Instrument Company, Catalog No: A-77912-00 
 
 265 
 
15. 3 Valves (PTFE teflon), Tubing O.D.: "
4
1
; 
Cole-Parmer Instrument Company, Catalog No: A-06373-55 
16. 3 Bulkhead Unions (Teflon PFA), Tubing O.D.: "
8
1
; 
Cole-Parmer Instrument Company, Catalog No: A-06357-32 
17. 2 White Teflon Flat Washer, "
8
3
screw size, 0.406" ID, 1.00" OD, 0.057" thick; 
MacMaster-Carr Supply Company, Part number: 95630 A248 
18. 1 Union crosses (Teflon PTFE), Tubing O.D.: "
4
1
;  
Cole-Parmer Instrument Company, Catalog No: A-06399-11 
19. 3 Male pipe adapter, Teflon PFA, "
4
1
"
4
1
? ; 
Cole-Parmer Instrument Company, Catalog No: A-06374-03 
20. Headtank material: Extruded acrylic sheet 12" x 24", 0.230" thick, 
McMaster-Carr Supply Company, Part number: 8589K82 
21. Silicone caulking 
Aquarium Sealant 100% Silicone, Home Depot, College Park, Maryland, U.S.A 
Sampling Parts: 
22. Laboratory Pipetting Needles, Stainless steel hub with 304 stainless steel tubing, 20 
Gauge, O.D.: 0.035", I.D.: 0.023", 4" long; 
Popper and Sons, Cat. No. for VWR: 20068-696 
23. Mininert syringe valves, push-button operations; To fit any standard luer tip syringes; 
Alltech Associate, Inc., Part No.: 6540 
 
 266 
 
Appendix 2 FORTRAN Program ?trafit3d? 
      PROGRAM trafit3d 
      USE numerical_libraries 
THIS IS A FORTRAN PROGRAM THAT CALCULATES THE SUM OF THE SQUARES 
OF EITHER THE ABSOLUTE OR RELATIVE RESIDUALS BETWEEN THE 
NORMALIZED EXPERIMENTAL CONSERVATIVE TRACER DATA AND THE 
NORMALIZED FLUX- AVERAGED CONCENTRATION CALCULATED USING THE 
CONTINUOUS POINT SOURCE MODEL AT STEADY STATE AS DESCRIBED BY 
ROBBINS (1989) AND THE BEST FIT HYDRODYNAMIC DISPERSION COEFFICIENT, 
DH, AND AVERAGE PORE WATER VELOCITY, V (WHICH IS USED TO CALCULATE 
THE BEST FIT PORO-SITY).  THE BEST FIT PARAMETERS ARE OBTAINED USING A 
MODIFIED LEVENBERG-MARQUARDT METHOD TO MINIMIZE THE SUMS OF THE 
SQUARES OF THE RESIDUALS BETWEEN OBSERVED AND CALCULATED 
CONCENTRATIONS.  THE EXPER- IMENTAL DATA ARE FROM A SAND FLOW TANK 
WITH A SQUARE CROSS-SECT. 
 THE MAIN PROGRAM CALLS FIVE SUBROUTINES: 
INPUT:  READS INPUT FROM A DATA FILE CALLED INDAT1D. 
DUNLSF: AN IMSL SUBROUTINE THAT SOLVES A NONLINEAR LEAST SQUARES 
PROBLEM USING A MODIFIED LEVENBERG-MARQUARDT ALGORITHM AND A 
FINITE-DIFFERENCE JACOBIAN. 
FCN: CALCULATES EITHER THE ABSOLUTE OR RELATIVE RESIDUAL VECTOR. 
EXER:   CALCULATES EXP(A) ERFC(B) 
VARIABLES: 
AREA - COLUMN CROSS SECTIONAL AREA (CM^2). 
BSTPOR - BEST FIT FOR POROSITY; CALCULATED FROM VELOCITY BEST FIT. 
CEX(I) - NORMALIZED EXPERIMENTAL EFFLUENT CONCENTRATION AT EACH 
SAMPLING TIME. 
CMOD(I) - CALCULATED NORMALIZED FLUX-AVERAGED EFFLUENT 
CONCENTRATION AT EACH SAMPLING TIME, USING OPTIMUM FIT PARAMETER 
VALUES. 
DATPTS - NUMBER OF EXPERIMENTAL OBSERVATIONS. 
VD(I) - ESTIMATED VALUES FOR AVERAGE PORE WATER VELOCITY AND 
HYDRODYNAMIC DISPERSION COEFFICIENT, RESPECTIVELY. 
FJAC(I,J) - FINITE DIFFERENCE APPROXIMATE JACOBIAN AT SOLUTION. 
FLOW - EXPERIMENTALLY DETERMINED BULK FLOW RATE (ML/HR). 
FSCALE(I) - DIAGONAL SCALING MATRIX FOR FUNCTION. 
LENGTH - LENGTH OF COLUMN (CM);(DISTANCE BETWEEN SAMPLING PORTS). 
N - NUMBER OF PARAMETERS TO BE ESTIMATED. 
PGUESS(I) - INITIAL GUESS FOR AVERAGE PORE WATER VELOCITY (CM/HR) 
AND HYDRODYNAMIC DISPERSION COEFFICIENT (CM^2/HR) , RESPECTIVELY. 
PSCALE(I) - DIAGONAL SCALING MATRIX FOR VARIABLES. 
RESID(I) - ABSOLUTE OR RELATIVE RESIDUALS. 
RPARAM(I) - PARAMETER VECTOR FOR OPTIMIZATION SUBROUTINE. 
SUMSQ - SUM OF SQUARES OF ABSOLUTE OR RELATIVE RESIDUALS. 
TIME(I) - TEST SAMLING TIMES (HR). 
DECLARATION OF VARIABLES 
 
 267 
 
      DOUBLE PRECISION TIME (70), CEX(70), CMOD(70), RESID(70), VD(2), 
     *RPARAM (7), FJAC(70,2), PSCALE(2), PGUESS(2), FSCALE(70) 
      DOUBLE PRECISION EXER, SUMSQ, LENGTH, FLOW, AREA, BSTPOR, STDDEV 
      INTEGER SSE, DATPTS, I, LDFJAC, IPARAM(6), N 
      COMMON/OBS/LENGTH, TIME, CEX, CMOD, SSE 
      EXTERNAL FCN, INPUT 
C OPEN FILES 
      OPEN (5, FILE='INDAT1D', STATUS='OLD') 
      OPEN (6, FILE='BSTFIT1D', STATUS='UNKNOWN') 
C READ INPUT VARIABLES 
      CALL INPUT (LENGTH, FLOW, AREA, PGUESS, SSE, DATPTS, TIME, CEX) 
C INITIALIZE OPTIMIZATION SUBROUTINE ARGUMENTS 
      N=2 
      DO 90 I=1,N 
         PSCALE(I)=1.0 
   90 CONTINUE 
      DO 100 I=1,70 
         FSCALE(I)=1.0 
  100 CONTINUE 
      IPARAM(1)=0 
      LDFJAC=70 
C CALL THE OPTIMIZATION SUBROUTINE TO SOLVE FOR THE BEST-FIT 
PARAMETERS 
      CALL DUNLSF(FCN, DATPTS, N, PGUESS, PSCALE, FSCALE, IPARAM, 
     *RPARAM, VD, RESID, FJAC, LDFJAC) 
C CALCULATE THE SUM OF THE SQUARES OF THE RESIDUALS 
      SUMSQ=0.0D0 
      DO 120 I=1,DATPTS 
         SUMSQ=SUMSQ+RESID(I)**2 
  120 CONTINUE 
C CALCULATE THE STANDARD DEVIATION OF THE ABSOLUTE OR RELATIVE 
RESIDUALS 
      STDDEV=DSQRT(SUMSQ/DFLOAT(DATPTS-2)) 
C CALCULATE THE BEST FIT POROSITY FROM THE BEST FIT AVG. PORE 
WATER VELOCITY, FLOW, AND CROSS-SECTIONAL AREA 
      BSTPOR=FLOW/(AREA*VD(1)) 
C WRITE RESULTS TO OUTPUT FILE 
      WRITE (6, 130) 
  130 FORMAT (8X, 'THE BEST-FIT PARAMETERS ARE:') 
      WRITE (6, 140) VD(1), BSTPOR, VD(2) 
  140 FORMAT (10X, 'VELOCITY =', X, F6.3, 4X, 'POROSITY =', X, E12.3, 4X 
     *, 'DH =', X, F6.3,/) 
      IF(SSE.EQ.1) THEN 
      WRITE (6, 150) 
  150 FORMAT (10X, 'TIME', 8X, 'C/C''(exp)', 4X, 'C/C''(model)', 
     *4X, 'ABS RESIDUAL', /) 
         ELSE 
      WRITE (6, 160) 
  160 FORMAT (10X, 'TIME', 8X, 'C/C''(exp)', 4X, 'C/C''(model)', 
     *4X, 'REL RESIDUAL', /) 
      ENDIF 
 
 268 
 
      DO 200 I=1, DATPTS 
         WRITE (6, 190) TIME(I), CEX(I), CMOD(I), RESID(I) 
  190    FORMAT (8X, F6.2, 8X, F6.4, 9X, F6.4, 8X, E10.4) 
  200 CONTINUE 
      WRITE (6,*) 
      WRITE (6, 210) SUMSQ 
  210 FORMAT (18X, 'THE SUM-OF-SQUARED RESIDUALS IS: ', G10.4) 
      WRITE (6, 220) STDDEV 
  220 FORMAT (7X, 'THE STANDARD DEVIATION OF THE RESIDUALS IS: ', G10.4) 
      STOP 
      END 
      SUBROUTINE INPUT (LENGTH, FLOW, AREA, PGUESS, SSE, DATPTS, TIME, 
CEX) 
C THIS SUBROUTINE READS INPUT FROM A DATA FILE CALLED INDAT1D 
C IF SSE=1, THE ABSOLUTE LEAST SQUARES (ALS) CRITERION IS USED; 
C IF SSE=2, THE RELATIVE LEAST SQUARES (RLS) CRITERION IS USED. 
      DOUBLE PRECISION TIME(70), CEX(70), PGUESS(2) 
      DOUBLE PRECISION LENGTH, FLOW, AREA 
      INTEGER SSE, DATPTS, I 
      READ (5,*) LENGTH, FLOW, AREA, PGUESS(1), PGUESS(2) 
 write (6, *) length, flow, area, pguess(1), pguess(2) 
    2 FORMAT(F6.3) 
      READ (5,*) SSE, DATPTS 
 write (6, *) sse, datpts 
    4 FORMAT(I2) 
      DO 10 I=1, DATPTS 
         READ (5,6) TIME(I), CEX(I) 
    write(6,*) time(i), cex(i) 
    6 FORMAT(F7.4, X, F6.4) 
   10 CONTINUE 
      RETURN  
      END 
SUBROUTINE FCN(DATPTS, N, VD, RESID) 
C THIS SUBROUTINE COMPUTES THE ANALYTICAL SOLUTION TO THE NON-
REACTIVE SOLUTE TRANSPORT EQUATION AT STEADY STATE (ROBBINS 1989). 
THE POSITION (X), TIME (T), PORE WATER VELOCITY (VD(1)), AND THE 
DISPERSION COEFFICIENT (VD(2)) ARE INPUTS. THE OUTPUT IS THE 
NORMALIZED(OVER STEADY STATE) FLUX-AVERAGED CONCENTRATION (C/C') 
AT THE GIVEN POSITION AND TIME. THIS 
REQUIRES AN EXTERNAL FUNCTION EXER(A,B), WHICH COMPUTES THE VALUE 
OF EXP(A)*ERFC(B). THE SUBROUTINE USES THESE VALUES TO THEN CALCULATE 
EITHER THE ABSOLUTE OR RELATIVE RESIDUAL VECTOR. 
C DECLARE VARIABLES 
      DOUBLE PRECISION TIME(70), CEX(70), CMOD(70), VD(2), RESID(70), 
     *LENGTH 
      INTEGER DATPTS, N, SSE, I 
      COMMON/OBS/LENGTH, TIME, CEX, CMOD, SSE 
C DECLARE LOCAL VARIABLES 
      DOUBLE PRECISION A(70), EXER 
C COMPUTE THE VALUES OF THE LOCAL VARIABLES 
      DO 300 I=1,DATPTS 
 
 269 
 
      A(I)=(LENGTH-VD(1)*TIME(I))/(2.0D0*DSQRT(VD(2)*TIME(I))) 
  300 CONTINUE 
C COMPUTE THE NORMALIZED(STEADY ST.) FLUX-AVERAGED 
CONCENTRATION AT THIS REQUIRES AN EXTERNAL FUNCTION EXER(A,B) WHICH 
COMPUTES THE VALUE OF EXP(A)*ERFC(B). 
      DO 310 I=1,DATPTS 
      CMOD(I)=0.5D0*EXER(0.0D0,A(I)) 
  310 CONTINUE 
C CALCULATE EITHER THE ABSOLUTE OR RELATIVE RESIDUALS VECTOR 
      DO 320 I=1,DATPTS 
         RESID(I)=CEX(I)-CMOD(I) 
         IF(SSE.EQ.2) RESID(I)=RESID(I)/CEX(I) 
  320 CONTINUE 
      RETURN 
      END 
      DOUBLE PRECISION FUNCTION EXER(A,B) 
C THIS SUBROUTINE IS FROM VAN GENUCHTEN AND ALVES (1982) 
C PURPOSE: TO CALCULATE EXP(A)*ERFC(B) 
C DECLARE DUMMY VARIABLES 
      DOUBLE PRECISION A, B 
C DECLARE LOCAL VARIABLES 
      DOUBLE PRECISION C, X, T, Y 
      EXER=0.0D0 
      IF ((DABS(A).GT.170.).AND.B.LE.0.0) RETURN 
      IF (B.NE.0.0) GOTO 100 
      EXER=DEXP(A) 
      RETURN 
  100 C=A-B*B 
      IF ((DABS(C).GT.170.).AND.(B.GT.0.0)) RETURN 
      IF (C.LT.-170.) GOTO 130 
      X=DABS(B) 
      IF (X.GT.3.0) GOTO 110 
      T=1.0D0/(1.0D0+0.3275911D0*X) 
      Y=T*(0.2548296D0-T*(0.2844967D0-T*(1.421414D0-T*(1.453152D0- 
     *1.061405D0*T)))) 
      GOTO 120 
  110 Y=0.5641896D0/(X+0.5D0/(X+1.0D0/(X+1.5D0/(X+2.0D0/(X+2.5D0/(X+ 
     *1.0D0)))))) 
  120 EXER=Y*DEXP(C) 
  130 IF (B.LT.0.0) EXER=2.0*DEXP(A)-EXER 
      RETURN 
      END 
*******************************************************************
 
 270 
 
Appendix 3: FORTRAN Program ?nvolma? 
          PROGRAM MAIN 
* -------------------------------------------------------------------- 
*                       PROGRAM 'NVOLMA' 
* 
*       This program fits a curve to batch oxygen consumption data 
*       using Monod or Andrews kinetics to describe the removal of a non-volatile 
*       substrate.  The best values for the parameters Um, Ks, Ki, and 
*       X0, are found using a least sum of squares estimator to 
*       determine the best fit for a given set of experimental data. 
*       The complex search routine of Box (Kuester, J.L. and Mize, 
*       J.H., "Optimization Techniques with Fortran", McGraw-Hill, 
*       pp 368-385, 1973.) is used to find the best fit. 
* 
*       You have the choice of having the program find the initial 
*       cell concentration, X0 or having X0 stay at a fixed value 
*       which you input and therefore only search for Um and Ks, Ki. 
*       You also have the choice of fitting to the Monod or Andrews equation 
* 
*        This program is a modification of volma.for wich was written for 
*       volatile components and any updates to volma.for must be included 
*       in this program 
*       Version 3.1 
*       Last Edited 7 JULY 1994 
* 
* ------------------------------------------------------------------- 
 
        INTEGER I,J,II,SIGN,FINDX,INDX,tnpts,cma 
        REAL T(500),AOU(500),MUMUB,MUMLB,KSUB,KSLB,X0UB,X0LB,S0, 
     1       Y,YP,B,TN,PAOU(500),MUM,KS,X0,F(9),X(4,9),XN(4,9), 
     2       XC(4),FMAX,FMIN,XUB(4),XLB(4),Ki, KiLB, KiUB 
        COMMON /BLOCKA/ T,AOU,S0,Y,YP,B,TN,PAOU,IIMAX 
        COMMON /BLOCKB/ X,F,XUB,XLB,NN 
        COMMON /BLOCKC/ XN,XC 
        COMMON /BLOCKD/ FINDX,cma 
        CHARACTER W1*3,W2*2,W3*3,W4*2,W5*3,W6*3,W7*3,W8*4,W9*9, 
     1            W10*9,W11*10,W12*1,W13*2,W14*3,W15*2,W16*5, 
     2            W17*3,W18*2,W19*2,W20*2,W22*5, W23*5, W21*10 
        CHARACTER*15 DATAIN, VOLPARAMSIN,RESULTSOUT 
* ----------------------------------------------------------------- 
*                        INSTRUCTIONS 
* 
*       The batch oxygen consumption data will be read in from 
*       a separate file, call "DATAIN". Parameter estimates,  
*       limits, and fitting criteria are read in from a seperate 
*       file called "VOLPARAMSIN". 
* 
 
 271 
 
* 
* ------------------------------------------------------------------- 
        WRITE (6,*)'This program Copyright of ESE Department' 
        WRITE (6,*)'Clemson University, 1994' 
        WRITE (6,*)'ENTER THE NAME OF YOUR DATA INPUT FILE' 
        READ (5,'(A15)') DATAIN 
        WRITE (6,*) ' DATAIN= ',DATAIN 
        WRITE (6,*)'ENTER THE NAME OF YOUR PARAMETER INPUT FILE' 
        READ (5,'(A15)') VOLPARAMSIN 
        WRITE (6,*) 'VOLPARAMSIN=', VOLPARAMSIN 
        WRITE (6,*)'ENTER THE NAME OF YOUR RESULTS OUTPUT FILE' 
        READ (5,'(A15)') RESULTSOUT 
        WRITE (6,*)'RESULTSOUT=', RESULTSOUT 
 
 
        OPEN (UNIT=7,FILE= DATAIN, STATUS='OLD') 
        open (UNIT=10,FILE= VOLPARAMSIN, STATUS='OLD') 
        OPEN (UNIT=8,FILE = RESULTSOUT, STATUS='UNKNOWN') 
 
 
 
* ------------------------------------------------------------------- 
*                DEFINITIONS OF INPUT DATA FOR PARMS.IN 
* 
*       S0 = initial substrate conc. if all the substrate were in 
*            the liquid phase.          MUMLB = Um lower boundary 
*       Y  = bacterial cell yield       MUMUB = Um upper boundary 
*       YP = product yield coefficient  KSLB  = Ks lower boundary 
*       B  = decay coefficient          KSUB  = Ks upper boundary 
*       KiLB= Ki lower boundary         X0LB  = X0 lower boundary 
*       KiUB = Ki upper boundary        X0UB  = X0 upper boundary 
*       ALFA = best fitting criterion   X0 = initial biomass conc 
*       BETA = converge factor          KI = Andrews coeff 
*       ITK  = maximum iteration times  KS = Monod coeff 
*       SEED = for random No. generation, 0.0<SEED<1.0 
*       FINDX = 1 if you want the program to find X0, 0 otherwise 
*       CMA = 1 if you want the program to fit to Andrews eqn, 0 otherwise 
* ------------------------------------------------------------------ 
 
        read (10,*) S0, X0,Y, YP, B, MUM, KS, Ki 
        read (10,*) MUMLB, MUMUB, KSLB, KSUB, X0LB, X0UB,KiLB, KiUB 
        read (10,*) ALFA, BETA, ITK, SEED 
        read (10,*) FINDX, cma 
 
* ------------------------------------------------------------------ 
*           ADDITIONAL INPUT DATA DUE TO VOLATILE COMPOUND 
* 
*       HS = Dimensionless Henry's Law Constant for Substrate 
*   VL = Volume of Liquid, mL. 
*   VG = Volume of Gas, mL. 
*   aMW = Molecular weight of substrate. 
 
 272 
 
*   CODr = mg COD/mg Mass for substrate. 
*       TempC = Temperature of Experiment, Celcius. 
* ------------------------------------------------------------------ 
 
*       read (10,*) HS, VL, VG, aMW, CODr, TempC  
  
* ----------------------------------------------------------------- 
*  This section for input of experimental data. 
* ----------------------------------------------------------------- 
 
        II=1 
 203    READ(7,*,end=205) T(II),AOU(II) 
    II=II+1 
           GO TO 203 
  
 205    IIMAX=II-1 
 tnpts=iimax 
         
        IF (FINDX .EQ. 0) THEN 
           NN=3 
        ELSE 
           NN=4 
        END IF 
 
* ---------------------------------------------------------------- 
*       This part is for rearranging data 
* ---------------------------------------------------------------- 
 
        X(1,1)=MUM 
        X(2,1)=KS 
        X(3,1)=Ki 
        X(4,1)=X0 
        XUB(1)=MUMUB 
        XUB(2)=KSUB 
        XUB(3)=KiUB 
        XUB(4)=X0UB 
        XLB(1)=MUMLB 
        XLB(2)=KSLB 
        XLB(3)=KiLB 
        XLB(4)=X0LB 
* -------------------------------------------------------------- 
*       This part is for generating 8 new starting points 
* -------------------------------------------------------------- 
 
        RAND=SEED 
        DO 30 J=2,9 
           DO 20 I=1,NN 
           CALL RANDOM(RAND) 
           X(I,J)=XLB(I)+RAND*(XUB(I)-XLB(I)) 
20         CONTINUE 
           IF (cma .EQ. 0) THEN 
 
 273 
 
              X(4,J)=KI 
           END IF 
           IF (FINDX .EQ. 0) THEN 
              X(4,J)=X0 
           END IF 
30      CONTINUE 
 
* ------------------------------------------------------------------------ 
*       This part is for calculation the objective function 
* ------------------------------------------------------------------------ 
        W17='RSS' 
        W18='Um' 
        W19='Ks' 
        W20='X0' 
        W22='Ki' 
        WRITE(8,*) '         ** Starting Points **' 
 WRITE(8,*) '----------------------------------------------------' 
        WRITE(8,5) W17,W18,W19,W22,W20 
 
        DO 40 J=1,9 
           CALL RKT(X(1,J),X(2,J),X(3,J),X(4,J),F(J)) 
 
        WRITE(8,6) F(J),X(1,J),X(2,J),X(3,J),X(4,J) 
40      CONTINUE 
        WRITE(8,*) '         ' 
 
* ---------------------------------------------------------------------- 
*       This part is for executing of Box search routine 
* ---------------------------------------------------------------------- 
        FMIN=1.0 
 DELT=1.0 
 SIGN=1 
        ITR=0 
 
  206  IF (SIGN .EQ. 0 .OR. FMIN .LT. ALFA .OR. DELT .LT. BETA 
     1         .OR. ITR .EQ. ITK) GO TO 207 
        FMAX=AMAX1(F(1),F(2),F(3),F(4),F(5),F(6),F(7),F(8),F(9)) 
        DO 50 J=1,9 
           IF (F(J) .EQ. FMAX) THEN 
              INDX=J 
           END IF 
 
50      CONTINUE 
 
        IF (SIGN .EQ. 1) THEN 
           CALL EXCBOX(INDX,Y1,Y2,Y3,Y4,YF,SIGN) 
           X(1,INDX)=Y1 
           X(2,INDX)=Y2 
           X(3,INDX)=Y3 
           X(4,INDX)=Y4 
           F(INDX)=YF 
 
 274 
 
           FMIN=AMIN1(F(1),F(2),F(3),F(4),F(5),F(6),F(7),F(8),F(9)) 
           FMAX=AMAX1(F(1),F(2),F(3),F(4),F(5),F(6),F(7),F(8),F(9)) 
        END IF 
 
        DELT=FMAX-FMIN 
        ITR=ITR+1 
        GO TO 206 
       CONTINUE 
 
 207    IF (FMIN .LT. ALFA) THEN 
        WRITE (8,*) 'Curve fitting criterion alpha was met' 
        END IF 
        IF (DELT .LT. BETA) THEN 
        WRITE (8,*) 'Curve fitting criterion beta was met' 
        END IF 
        IF (ITR .EQ. ITK) THEN 
        WRITE (8,*) 'Did not converge within preset iteration steps' 
        END IF 
        IF(SIGN .EQ. 0) THEN 
        WRITE (8,*) 'No further search is possible.' 
        WRITE (8,*) 'Try a different start point or seed.' 
        END IF 
        WRITE(8,*) '    ' 
 
        DO 60 J=1,9 
           IF (F(J) .EQ. FMIN) THEN 
           INDX=J 
           END IF 
 60      CONTINUE 
 
* --------------------------------------------------------------- 
*       This part for output of curve fitting results 
* --------------------------------------------------------------- 
 
        W1='S0=' 
        W2='Y=' 
        W3='Yp=' 
        W4='b=' 
        W5='Um=' 
        W6='Ks=' 
        W7='X0=' 
        W23='Ki=' 
        W8='RSS=' 
        W9='RSS<' 
        W10='RSSmax - RSSmin=' 
        W11='Iteration=' 
        W12='T' 
        W13='AOU' 
        W14='PAOU' 
        W15='to' 
        W16='SEED=' 
 
 275 
 
 W21='# of data points =' 
 
        IF (FINDX .EQ. 0) THEN 
 
           WRITE(8,*) 'Xo was given by the user.' 
        else 
          WRITE(8,*) 'This program searched for Xo.' 
        END IF 
 
        If(cma.eq.0) then 
          write(8,*)'This program fit to the Monod Equation' 
        else 
          write(8,*)'This program fit to the Andrews Equation' 
        endif 
 
        WRITE(8,*) '-------------------------------------------' 
        WRITE(8,*) '  KNOWN PARAMETERS AND INITIAL CONDITIONS' 
        WRITE(8,*) '-------------------------------------------' 
        WRITE(8,*) '         ' 
 
        WRITE(8,*) '           ** Known Parameters **' 
 WRITE(8,*) '------------------------------------------------' 
        IF (FINDX .EQ. 0) THEN 
           WRITE(8,1) W1,S0,W2,Y,W3,YP,W4,B,W7,X0 
        ELSE 
           WRITE(8,1) W1,S0,W2,Y,W3,YP,W4,B 
        END IF 
        WRITE(8,4) W21,tnpts 
 
        WRITE(8,*) '        ' 
        WRITE(8,*) '           ** Boundary Conditions **' 
 WRITE(8,*) '------------------------------------------------' 
        WRITE(8,2) W5,MUMLB,W15,MUMUB 
        WRITE(8,2) W6,KSLB,W15,KSUB 
        write(8,2) W23,KiLB,W15,KiUB 
        IF (FINDX .EQ. 1) THEN 
           WRITE(8,2) W7,X0LB,W15,X0UB 
        END IF 
 
        WRITE(8,*) '           ** Searching Criterions **' 
 WRITE(8,*) '------------------------------------------------' 
        WRITE(8,31) W9,ALFA 
        WRITE(8,31) W10,BETA 
        WRITE(8,32) W11,ITK 
        WRITE(8,31) W16,SEED 
        WRITE(8,*) '        ' 
        WRITE(8,*) '             ** Best Fit Results **' 
 WRITE(8,*) '------------------------------------------------' 
        WRITE(8,32) W11,ITR 
        WRITE(8,31) W10,DELT 
        IF (FINDX .EQ. 0) THEN 
 
 276 
 
           WRITE(8,7) W8,F(INDX),W5,X(1,INDX),W6,X(2,INDX),W23,X(3,INDX) 
        ELSE 
           WRITE(8,7) W8,F(INDX),W5,X(1,INDX),W6,X(2,INDX), 
     1                W23,X(3,INDX),W7,X(4,INDX) 
        END IF 
        WRITE(8,*) '        ' 
 
        BASE=0.0 
 
c       Call the Runge Kutta a final time with the values of umax, Ks, 
c       Xo, and Ki that yield the lowest residual SS 
 
        CALL RKT(X(1,indx),X(2,indx),X(3,indx),X(4,indx),F(indx)) 
 
        WRITE(8,5) W12,W13,W14 
        DO 70 I=1,IIMAX 
          WRITE(8,8) T(I),AOU(I),PAOU(I) 
70       CONTINUE 
 
 
1       FORMAT (A5,E10.2,2X,A5,F7.4,2X,A5,F7.4,2X,A5,F7.4,2X, 
     1          A5,F10.2) 
2       FORMAT (2X,A5,F12.4,2X,A2,F12.2) 
3       FORMAT (2X,A10,F7.4,2X,A2,2X,F7.4) 
4       FORMAT (2X,A20,I7,2X,A2,2X,I7) 
5       FORMAT (3X,A8,3X,A8,3X,A8,3X,A8) 
6       FORMAT (E15.5,3F11.4) 
7       FORMAT (A5,E15.5,2X,A5,F7.4,2X,A5,F12.4,2X,A5,F12.4) 
8       FORMAT (4F11.4) 
9       FORMAT (2X,A10,E15.5) 
31      FORMAT (2X,A15,E16.2) 
32      FORMAT (2X,A15,8X,I5) 
c234567 
        STOP 
        END 
 
* --------------------------------------------------------------- 
*       Subroutine RKT calculating the 'rss' for every set of (Um, 
*       Ks,Ki).  A 4th order universal Runge-Kutta was written to 
*       generate theoretical curve 
* -------------------------------------------------------------- 
        SUBROUTINE RKT(MUM,KS,Ki,X0,RSS) 
        REAL AOU(500),MUM,KS,Ki,T(500),S0,B,TN,X0,YP,PAOU(500),R,RSS,Y, 
     1   coeff,sub1, sub2, sub3, sub4 
        DOUBLE PRECISION CVR(5),VR(5),H(3,5),CF(3) 
        INTEGER J,NE,N,IIMAX,cma,findx 
        COMMON /BLOCKA/ T,AOU,S0,Y,YP,B,TN,PAOU,IIMAX 
        COMMON /BLOCKD/ FINDX,cma 
 
        N=3 
        R=0.0 
 
 277 
 
        PAOU(1)=0.0 
 
           CVR(1)=S0 
           CVR(3)=0.0 
           CVR(2)=X0 
    TT=T(1) 
 
        DO 500 II=2,IIMAX 
            DO 550 JJ=1,10 
                VR(1)=CVR(1) 
                VR(2)=CVR(2) 
                VR(3)=CVR(3) 
                TT=T(II)-T(II-1) 
                TN=TT/10 
            DO 400 NE=1,N 
                H(NE,1)=0.0 
400         CONTINUE 
 
            DO 300 J=2,5 
 
                CK=0.5*INT((J-1)/2) 
 
                   SUB1=MUM*(VR(1)+CK*H(1,J-1))*(VR(2)+CK*H(2,J-1)) 
                   SUB2=KS+VR(1)+CK*H(1,J-1) 
     1              + real(cma)*(VR(1)+CK*H(1,J-1))**2./(Ki) 
                   coeff=  real(cma)*(VR(1)+CK*H(1,J-1))**2./(Ki) 
                   SUB3=KS*B*(VR(2)+CK*H(2,J-1)) 
                   SUB4=KS+VR(1)+CK*H(1,J-1) 
c                   write(6,*)'coeff=', coeff 
c                   write(6,*)'sub1=', sub1 
c                   write(6,*)'sub2=', sub2 
c                   write(6,*)'sub3=', sub3 
                   CF(1)=-SUB1/(Y*SUB2) 
                   CF(2)=(SUB1/SUB2)-(SUB3/SUB4) 
                   CF(3)=YP*SUB1/(Y*SUB2) 
c                   write (6,*) 'cf(1) =',cf(1) 
c                   write (6,*) 'cf(2) =',cf(2) 
c                   write (6,*) 'cf(3)=',cf(3) 
 
                DO 200 NE=1,N 
                H(NE,J)=TN*(CF(NE)) 
200             CONTINUE 
300             CONTINUE 
 
           DO 100 NE=1,N 
           CVR(NE)=VR(NE)+(H(NE,2)+2.0*H(NE,3)+2.0*H(NE,4)+H(NE,5))/6 
100        CONTINUE 
           IF (CVR(1) .LT. 0.0) THEN 
              CVR(1)=0.0 
           END IF 
 
 
 278 
 
 
550        CONTINUE 
 
           PAOU(II)=(S0-CVR(1))-(CVR(2)-X0)-CVR(3) 
c        write(8,*) PAOU(II) 
 
           R=R+(PAOU(II)-AOU(II))**2 
 
500     CONTINUE 
        RSS=R 
 
        RETURN 
        END 
 
* ------------------------------------------------------------------- 
*       Subroutine RANDOM for generating radom numbers. 
* ------------------------------------------------------------------- 
 
        SUBROUTINE RANDOM(SEED) 
        REAL A,B,C,SEED 
 
        A=65536 
        B=25173 
        C=13849 
        SEED=((A*SEED+B)/C)-INT((A*SEED+B)/C) 
        RETURN 
        END 
 
* ------------------------------------------------------------------- 
*       Subroutine EXCBOX for execute Box search routine 
* ------------------------------------------------------------------- 
 
        SUBROUTINE EXCBOX(INDX,Y1,Y2,Y3,Y4,YF,SIGN) 
        INTEGER SIGN,findx,cma 
        REAL XC(4),XCN,XN(4,9),X(4,9),F(9),FN(9),XUB(4),XLB(4) 
        COMMON /BLOCKB/ X,F,XUB,XLB,NN 
        COMMON /BLOCKC/ XN,XC 
        COMMON /BLOCKD/ FINDX,cma 
        FMAX=AMAX1(F(1),F(2),F(3),F(4),F(5),F(6),F(7),F(8),F(9)) 
        DO 1010 I=1,NN 
           CALL CENTRD(I,INDX,XCN) 
        XC(I)=XCN 
1010    CONTINUE 
        DO 1020 I=1,NN 
        XN(I,INDX)=XC(I)+1.3*(XC(I)-X(I,INDX)) 
        IF (XN(I,INDX) .GT. XUB(I)) THEN 
           XN(I,INDX)=XUB(I) 
        END IF 
        IF (XN(I,INDX) .LT. XLB(I)) THEN 
        XN(I,INDX)=XLB(I) 
        END IF 
 
 279 
 
1020    CONTINUE 
        IF (cma .EQ. 0) THEN 
           XN(3,INDX)=X(3,INDX) 
        end if 
        IF (FINDX .EQ. 0) THEN 
           XN(4,INDX)=X(4,INDX) 
        END IF 
        CALL RKT(XN(1,INDX),XN(2,INDX),XN(3,INDX),XN(4,INDX),FN(INDX)) 
        IF (FN(INDX) .GE. F(INDX)) THEN 
        CALL GOLDEN(INDX,XN(1,INDX),XN(2,INDX),XN(3,INDX),XN(4,INDX), 
     1              FN(INDX),SIGN) 
        END IF 
        Y1=XN(1,INDX) 
        Y2=XN(2,INDX) 
        Y3=XN(3,INDX) 
        Y4=XN(4,INDX) 
        YF=FN(INDX) 
 
        RETURN 
        END 
 
* ----------------- -------------------------------------------------- 
*       Soubroutine CENTRD for calculation of centroids of 
*       remaining points. 
* ------------------------------------------------------------------- 
 
        SUBROUTINE CENTRD(I,INDX,XCN) 
        INTEGER I,INDX 
        REAL X(4,9),SIGMA,XCN,XN(4,9),XC(4),F(9),XUB(4),XLB(4) 
        COMMON /BLOCKB/ X,F,XUB,XLB,NN 
        COMMON /BLOCKC/XN,XC 
        SIGMA=0.0 
        DO 2010 J=1,9 
        SIGMA=SIGMA+X(I,J) 
2010    CONTINUE 
        XCN=(SIGMA-X(I,INDX))/8 
        RETURN 
        END 
 
 
* ------------------------------------------------------------------- 
*       Subroutine GOLDEN for execute golden section search. 
* ------------------------------------------------------------------- 
 
        SUBROUTINE GOLDEN(INDX,Y1,Y2,Y3,Y4,YF,SIGN) 
        REAL Y1,Y2,Y3,Y4,YF,X(4,9),F(9),XUB(4),XLB(4),P(4,4),PF(4) 
        INTEGER SIGN,INDX 
        COMMON /BLOCKB/ X,F,XUB,XLB,NN 
 
        G=0.618 
        IT=1 
 
 280 
 
        P(1,1)=X(1,INDX) 
        P(2,1)=X(2,INDX) 
        P(3,1)=X(3,INDX) 
        P(4,1)=X(4,INDX) 
        PF(1)=F(INDX) 
        P(1,4)=Y1 
        P(2,4)=Y2 
        P(3,4)=Y3 
        P(4,4)=Y4 
        PF(4)=YF 
 
        DO 3010 I=1,4 
           P(I,2)=P(I,1)+(1-G)*(P(I,4)-P(I,1)) 
           P(I,3)=P(I,1)+G*(P(I,4)-P(I,1)) 
3010    CONTINUE 
        CALL RKT(P(1,2),P(2,2),P(3,2),P(4,2),PF(2)) 
        CALL RKT(P(1,3),P(2,3),P(3,3),P(4,3),PF(3)) 
 
        PFMIN=AMIN1(PF(1),PF(2),PF(3),PF(4)) 
        IF (PFMIN .LT. F(INDX)) THEN 
           SIGN=1 
           DO 3020 J=2,4 
              IF (PF(J) .EQ. PFMIN) THEN 
              JG=J 
              END IF 
3020    CONTINUE 
           Y1=P(1,JG) 
           Y2=P(2,JG) 
           Y3=P(3,JG) 
           Y4=P(4,JG) 
           YP=PF(JG) 
        ELSE 
209       IF (PFMIN .LT. F(INDX) .OR. IT .EQ. 10) GO TO 208 
              IT=IT+1 
              IF (PFMIN .EQ. PF(1)) THEN 
                 DO 3030 I=1,4 
                    P(I,4)=P(I,2) 
                    P(I,2)=P(I,1) 
                    P(I,1)=P(I,2)-(1-G)*(P(I,4)-P(I,2)) 
                    P(I,3)=P(I,1)+G*(P(I,4)-P(I,1)) 
3030             CONTINUE 
                 PF(4)=PF(2) 
                 PF(2)=PF(1) 
                 CALL RKT(P(1,1),P(2,1),P(3,1),P(4,1),PF(1)) 
                 CALL RKT(P(1,3),P(2,3),P(3,3),P(4,3),PF(3)) 
              END IF 
 
              IF (PFMIN .EQ. PF(2)) THEN 
                 DO 3040 I=1,4 
                    P(I,4)=P(I,3) 
                    P(I,3)=P(I,2) 
 
 281 
 
                    P(I,2)=P(I,1)+(1-G)*(P(I,4)-P(I,1)) 
3040             CONTINUE 
                 PF(4)=PF(3) 
                 PF(3)=PF(2) 
                 CALL RKT(P(1,2),P(2,2),P(3,2),P(4,2),PF(2)) 
              END IF 
 
              IF (PFMIN .EQ. PF (3)) THEN 
                 DO 3050 I=1,4 
                    P(I,1)=P(I,2) 
                    P(I,2)=P(I,3) 
                    P(I,3)=P(I,1)+G*(P(I,4)-P(I,1)) 
3050             CONTINUE 
                 PF(1)=PF(2) 
                 PF(2)=PF(3) 
                 CALL RKT(P(1,3),P(2,3),P(3,3),P(4,3),PF(3)) 
              END IF 
 
              IF (PFMIN .EQ. PF(4)) THEN 
                 DO 3060 I=1,4 
                    P(I,1)=P(I,3) 
                    P(I,3)=P(I,4) 
                    P(I,4)=P(I,3)+(1-G)*(P(I,3)-P(I,1)) 
                    P(I,2)=P(I,4)-G*(P(I,4)-P(I,1)) 
3060             CONTINUE 
                 PF(1)=PF(3) 
                 PF(3)=PF(4) 
                 CALL RKT(P(1,2),P(2,2),P(3,2),P(4,2),PF(2)) 
                 CALL RKT(P(1,4),P(2,4),P(3,4),P(4,4),PF(4)) 
              END IF 
            GO TO 209 
 208       CONTINUE 
 
              IF (IT .EQ. 10 .AND. PFMIN .GE. F(INDX)) THEN 
                 SIGN=0 
                 Y1=X(1,INDX) 
                 Y2=X(2,INDX) 
                 Y3=X(3,INDX) 
                 Y4=X(4,INDX) 
                 YF=F(INDX) 
 
              ELSE 
                 DO 3070 J=1,4 
                    IF (PF(J) .LT. F(INDX)) THEN 
                       JG=J 
                    END IF 
3070             CONTINUE 
                 Y1=P(1,JG) 
                 Y2=P(2,JG) 
                 Y3=P(3,JG) 
                 Y4=P(4,JG) 
 
 282 
 
                 YF=PF(JG) 
                 SIGN=1 
              END IF 
        END IF 
 
        RETURN 
        END 
 
 
 
 
 283 
 
Bibliography 
Alexander, M. (1994). Biodegradation and Bioremediation, Academic Press, San 
Diego. 
Alexander, M. (1995). "How toxic are toxic chemicals in soil?" Environmental 
Science & Technology, 29, 2713-2716. 
Anderson, M. P., and Woessner, W. W. (1992). Applied Groundwater Modeling: 
Simulation of Flow and Advective Transport, Academic Press, Boston, MA. 
Angley, J. T., Brusseau, M. L., Miller, W. L., and Delfino, J. J. (1992). 
"Nonequilibrium sorption and aerobic biodegradation of dissolved 
alkylbenzenes during transport in aquifer material:  column experiments and 
evaluation of a coupled-process model." Environ. Sci. Technol., 26(7), 1404-
1410. 
APHA, AWWA, and WEF. (1995). Standard Methods for the Examination of Water 
and Wastewater, American Public Health Association, American Water 
Works Association, and Water Environment Federation, Washington, D.C. 
Appert-Collin, J. C., Dridi-Dhaouadi, S., Simonnot, M. O., and Sardin, M. (1999). 
"Nonlinear sorption of naphthalene and phenanthrene during saturated 
transport in natural porous media." Physics and Chemistry of the Earth Part 
(B)- Hydrology Oceans and Atmosphere, 24(6), 543-548. 
ASTM-D2434. (2000). Standard Test Method for Permeability of Granular Soils 
(Constant Head), ASTM International. 
Bae, W., and Rittmann, B. E. (1996). "A structured model of dual-limitation 
kinetics." Biotechnology and Bioengineering, 49, 683-689. 
Barker, J. F., Patrick, G. C., and Major, D. (1987). "Natural attenuation of aromatic 
hydrocarbons in a shallow sand aquifer." Ground Wat. Monitor. Rev., 7(1), 
64-71. 
Barkley, W. E., and Richardson, J. H. (1994). Methods for General and Molecular 
Bacteriology, American Society for Microbiology, Washington, D. C. 
Bath, G. R., Hill, M. C., Illangasekare, T. H., and Rajaram, H. (2001). "Predictive 
modeling of flow and transport in a two-dimensional intermediate-scale, 
heterogeneous porous medium." Water Resources Research, 37(10), 2503-
2512. 
Baveye, P., and Valocchi, A. (1989). "An evaluation of mathematical-models of the 
transport of biologically reacting solutes in saturated soils and aquifers." 
Water Resources Research, 25(6), 1413-1421. 
 
 284 
 
Beck, A. J., Wilson, S. C., Alcock, R. E., and Jones, K. C. (1995). "Kinetic 
constraints on the loss of organic chemicals from contaminated soils: 
Implications for soil-quality limits." CRC Critical Reviews in Environmental 
Science and Technology, 25, 1-43. 
Behrens, H. "Water tracer chemistry: a factor determining performance and analytics 
of tracers." 5th Int. Symp. on Undergound Water Tracing, Insitute of Geology 
and Mineral Exploration, Athens, Greece, 121-133. 
Blake, G. R., and Hartge, K. H. (1986). "Particle density." Methods of soil analysis, 
Part 1, Physical and mineralogical methods, Am. Soc. of Agronomy, 
Madison, Wis. 
Borden, R. C., and Bedient, P. B. (1986). "Transport of dissolved hydrocarbons 
influenced by oxygen-limited biodegradation. 1. Theoretical development." 
Water Resour. Res., 22(13), 1973-1982. 
Borden, R. C., Bedient, P. B., Lee, M. D., Ward, C. H., and Wilson, J. T. (1986). 
"Transport of dissolved hydrocarbons influenced by oxygen-limited 
biodegradation.  2.  Field application." Water Resour. Res., 22(13), 1983-
1990. 
Bosma, T. N. P., Middeldorp, P. J. M., Schraa, G., and Zehnder, A. J. B. (1997). 
"Mass transfer limitation of biotransformation: quantifying bioavailability." 
Environmental Science & Technology, 31, 248-252. 
Bouwer, E., Durant, N., Wilson, L., Zhang, W., and Cummingham, A. (1994). 
"Degradation of xenobiotic compounds in situ: Capabilities and limits." 
Microbiology Review, 15, 307-317. 
Brock, T. D. (1977). "The Poisoned Control in biogeochemical Investigations." 
Environmental Biogeochemistry and Geomicrobiology, Ann Arbor Science, 
Ann Arbor, MICH., 717-725. 
Brockman, F. J., and Murray, C. J. (1997). Microbiological heterogeneity in the 
terrestrial subsurface and approaches for its description, Lewis Publishers, 
Boca Raton, FL. 
Brown, M. J., and Burris, D. R. (1996). "Enhanced organic contaminant sorption on 
soil treated with cationic surfactants." Ground Water, 34(4), 734-744. 
Brown, S. C., Grady Jr., C. P. L., and Tabak, H. H. (1990). "Biodegradation kinetics 
of substituted phenolics: Demonstration of a protocol based on electrolytic 
respirometry." Water Research, 24(7), 853-861. 
Brusseau, M. L. (1991). "Application of a multi-process non-equilibrium sorption 
model to solute transport in a stratified porous-medium." Water Resources 
Research, 27(4), 585-595. 
 
 285 
 
Brusseau, M. L. (1992). "Transport of rate-limited sorbing solutes in heterogeneous 
porous media." Water Resources Research, 28, 2485-2497. 
Brusseau, M. L. (1994). "Transport of reactive contaminants in heterogeneous 
porous-media." Review of Geophysics, 32(3), 285-313. 
Brusseau, M. L., Jessup, R. E., and Rao, P. S. C. (1991). "Non-equilibrium sorption 
of organic chemicals: Elucidation of rate-limiting processes." Environmental 
Science & Technology, 25, 134-142. 
Brusseau, M. L., and Rao, P. S. C. (1989). "Sorption nonideality during organic 
contaminant transport in porous media." CRC Critical Reviews in 
Environmental Control, 19, 22-99. 
Brusseau, M. L., and Rao, P. S. C. (1991). "Influence of sorbate structure on non-
equilibrium sorption of organic-compounds." Environmental Science & 
Technology, 25(8), 1501-1506. 
Brusseau, M. L., and Zachara, J. M. (1993). "Transport of CO32- in a physically and 
chemically heterogeneous porous-medium." Environmental Science & 
Technology, 27(9), 1937-1939. 
Burris, D. R., and Antworth, C. (1992). "In situ modification of an aquifer material by 
a cationic surfactant to enhance retardation of organic contaminants." Journal 
of Contaminant Hydrology, 10, 325-337. 
Calvillo, Y. M., and Alexander, M. (1996). "Mechanism of microbial utilization of 
biphenyl sorbed to polyacrylic beads." Applied Microbial Biotechnology, 45, 
383-390. 
Carrera, J. (1993). "An overview of uncertainties in modeling groundwater solute 
transport." Journal of Contaminant Hydrology, 13(1-4), 23-48. 
Chang, H. T., and Rittmann, B. E. (1987). "Verification of the model of biofilm on 
activated carbon." Environmental Science & Technology, 21, 280-288. 
Clement, T. P. (1998). "RT3D Tutorials for GMS users." PNNL-11805, Pacific 
Northwest National Laboratory Report. 
Clement, T. P., Sun, Y., Hooker, B. S., and Petersen, J. N. (1998). "Modeling multi-
species reactive transport in ground water." Groundwater Monitoring and 
Remediation Journal, 18(2), 79-92. 
Dand, J. S., Harvey, D. M., Jabbagy, A., and Grady, C. P. L. J. (1989). "Evaluationof 
biodegradation kinetics with repirometric data." Research of Journal of the 
Water Pollution Control Federation, 61(11-12). 
 
 286 
 
David, F., Sandra, P., Stafford, S. S., and Slavica, B. (1993). "Analysis of polynuclear 
aromatic hydrocarbons using pressure-pulsed splitless injection." Application 
Note 228-224, Hewlett Packard Company, Germany. 
Davis, S. N. (1969). Flow through porous materials, Academic Press, New York. 
Desai, J. D., and Banat, I. M. (1997). "Microbial production of surfactants and their 
commercial potential." Microbiology and Molecular Biology Reviews, 61(1), 
47-64. 
Domenico, P. A., and Schwartz, F. W. (1998). Physical and Chemical Hydrogeology, 
John Wiley & Sons, New York, NY. 
Efroymson, R. A., and Alexander, M. (1991). "Biodegradation by an Arthrobacter 
species of hydrocarbons partitioned into an organic solvent." Applied 
Environmental Microbiology, 57, 1441-1447. 
Eguchi, M., Kitagawa, M., Suzuki, Y., Nakamuara, M., Kawai, T., Okamura, K., 
Sasaki, S., and Miyake, Y. (2001). "A field evaluation of in situ 
biodegradation of trichloroethylene through methane injection." Water 
Research, 35(9), 2145-2152. 
Falatko, D. M., and Novak, J. T. (1992). "Effects of biologically produced surfactants 
on the mobility and biodegradation of petroleum-hydrocarbons." Water 
Environment Research, 64(2), 163-169. 
Famisan, G. B., and Brusseau, M. L. (2003). "Biodegradation during contaminant 
transport in porous media: 6. Impact of sorption on coupled degradation-
transport behavior." Environmental Toxicology and Chemistry, 22(3), 510-
517. 
Fetter, C. W. (2001). Applied Hydrogeology, Prentice-Hall, Inc., Upper Saddle River, 
New Jersey. 
Feuerstein, D. L., and Selleck, R. E. (1963). "Fluorescent tracers for dispersion 
measurements." J. Sanit. Eng. Div., Proc. Am. Soc. Civ. Eng., 89(SA 4), 1-21. 
Fritz, J. S., and Schenk, G. H. (1979). Quantitative Analytical Chemistry, Allyn and 
Bacon, Inc., Boston, MA. 
Fry, V. A., and Istok, J. D. (1994). "Effects of Rate-limited Desorption on the 
Feasibility of in-situ Bioremediation." Water Resources Research, 30(8), 
2413-2422. 
Gelhar, L. W., Welty, C., and Rehfeldt, K. R. (1992). "A critical review of data on 
field-scale dispersion in aquifers." Water Resources Research, 28(7), 1955-
1974. 
 
 287 
 
Ghoshal, S., and Luthy, R. G. (1998). "Biodegradation Kinetics of Naphthalene in 
Nonaqueous Liquid-Water Mixed Batch Systems: Comparson of Model 
Predictions and Experimental Results." Biotechnology and Bioengineering, 
57(3), 356-366. 
Ghoshal, S., Ramaswami, A., and Luthy, R. G. (1996). "Biodegradation of 
naphthalene from coal tar and heptamethylnonane in mixed batch systems." 
Environmental Science & Technology, 30(4), 1282-1291. 
Grady Jr., C. P. L., Smets, B. F., and Barbeau, D. S. (1996). "Variability in kinetic 
parameter estimates:  A review of possible causes and a proposed 
terminology." Water Research, 30(3), 742-748. 
Grblic-Galic, D. (1988). "Anaerobic Microbial Transformation of Aromatic 
Hydrocarbons and the Relevance to Regeneration of Contaminated Ground 
Waste Aquifers," Stanford University. 
Gu, M. B., and Chang, S. T. (2001). "Soil biosensor for the detection of PAH toxicity 
using an immobilized recombinant bacterium and a biosurfactant." 
BIOSENSORS & BIOELECTRONICS, 16(9-12), 667-674. 
Guerin, W. F., and Boyd, S. A. (1992). "Differential bioavailability of soil sorbed 
naphthalene to 2 bacterial species." Applied Environmental Microbiology, 58, 
1142-1152. 
Guerin, W. F., and Boyd, S. A. (1997). "Bioavailability of naphthalene associated 
with natural and synthetic sorbents." Water Research, 31(6), 1504-1512. 
Haggerty, R., and Gorelick, S. M. (1994). "Design of multiple contaminant 
remediation: Sensitivity to rate-limited mass transfer." Water Resources 
Research, 30(2), 435-446. 
Harayama, S. (1997). "Polycyclic aromatic hydrocarbon bioremediation design." 
Current Opinion in Biotechnology, 8, 268-273. 
Harms, H., and Zehnder, A. J. B. (1995). "Bioavailability of sorbed 3-
chlorodibenzofuran." Applied Environmental Microbiology, 61, 27-33. 
Harvey, R. W., Smith, R. L., and George, L. (1984). "Effect of organic contaminant 
upon microbial distribution and heterotrophic uptake in a Cape Co. Mass. 
aquifer." Applied Environmental Microbiology, 48(5), 1197-1202. 
Head, I. M. (1998). "Bioremediation: towards a credible technology." Microbiology-
UK, 144(3-4), 599-608. 
Hornberger, G. M., Mills, A. L., and Herman, J. S. (1992). "Bacterial transport in 
porous media: Evaluation of a model using laboratory observations." Water 
Resources Research, 28(3), 915-938. 
 
 288 
 
Hughes, J. B., Beckles, D. M., Chandra, S. D., and Ward, C. H. (1997). "Utilization 
of bioremediation processes for the treatment of PAH-contaminated 
sediments." Journal of Industrial Microbiology and Biotechnology, 18, 152-
160. 
Johnson, C. R., and Scow, K. M. (1999). "Effect of nitrogen and phosphorus addition 
on phenanthrene biodegradation in four soils." Biodegradation, 10, 43-50. 
Johnson, J. A. (2004). "A quantitative framework for understanding the compex 
interactions of competing interfacial processes and in situa biodegradation," 
Master of Science, University of Maryland, College Park. 
Kent, D. B., Davis, J. A., Anderson, L. C. D., Rea, B. A., and Waite, T. D. (1994). 
"Transport of chromium and selenium in the suboxic zone of a shallow 
aquifer: Influence of redox and adsorption reactions." Water Resources 
Research, 30(4), 1099-1114. 
Kibbey, T., and Hayes, K. F. (1993). "Partitioning and UV absorption studies of 
phenanthrene on cationic surfactant-coated silica." Environmental Science & 
Technology, 27, 2168-2173. 
Kissel, J. C., McCarty, P. L., and Street, R. L. (1984). "Numerical-simulation of 
mixed-culture biofilm." Journal of Environmental Engineering-ASCE, 110(2), 
399-411. 
Klonis, N., and Sawyer, W. H. (1996). "Spectral properties of the prototropic forms 
of fluorescein in aqueous solution." J. Fluoresc., 6(3), 147-157. 
Klotz, D., Seiler, K. P., Moser, H., and Neumaier, F. (1980). "Dispersivity and 
velocity relationship from laboratory and field experiments." Journal of 
Hydrology, 45(3-4), 169-184. 
Kono, T., and Asai, T. (1969). "Kinetics of continuous cultivation." Biotechnology 
and Bioengineering, 11, 19-36. 
Kramer, W. H., and Hayes, T. J. (1987). "Water Soluble Phase of Gasoline:  Results 
of a Laboratory Mixing Experiment." New Jersey Geological Survey 
Technical Memorandum 87-5, New Jersey Department of Environmental 
Protection, Trenton, NJ. 
Kuester, J. L., and Mize, J. H. (1973). Optimization Techniques with Fortran, 
McGraw-Hill. 
Lapidus, L., and Amundson, N. R. (1952). "The rate-determining steps in radial 
adsorption analysis." Journal of Physical Chemistry, 56(3), 373-383. 
 
 289 
 
MacIntyre, W. G., Boggs, M., Antworth, C. P., and Stauffer, T. B. (1993). 
"Degradation kinetics of aromatic organic solutes introduced into a 
heterogeneous aquifer." Water Resour. Res., 29(12), 4045-4051. 
MacQuarrie, K. T. B., and Sudicky, E. A. (1990). "Simulation of biodegradable 
organic contaminants in groundwater. 2. Plume behavior in uniform and 
random flow fields." Water Resour. Res., 26(2), 223-239. 
Macrae, J. D., and Kenneth, J. H. (1998). "Biodegradation of polycyclic aromatic 
hydrocarbons (PAH) in marine sediment under denitrifying conditions." 
Water Science Technology, 38(11), 177-185. 
Major, D. W., McMaster, M. L., Cox, E. E., Edwards, E. A., Dworatzek, S. M., 
Hendrickson, E. R., Starr, M. G., Payne, J. A., and Buonamici, L. W. (2002). 
"Field demonstration of successful bioaugmentation to achieve dechlorination 
of tetrachloroethene to ethene." Environmental Science & Technology, 36(23), 
5106-5116. 
Manilal, V. B., and Alexander, M. (1991). "Factors affecting the microbial 
degradation of phenanthrene in soil." Applied Microbiology and 
Biotechnology, 35, 401-405. 
McCarty, P. L. (1987). "Bioengineering issues related to in-situ remediation of 
contaminated soils and groundwater." Environmental Biotechnology, G. S. 
Omenn, ed., Plenum Press, New York, 143-162. 
McMahon, P. B., and Chapelle, F. H. (1991). "Microbial production of organic acids 
in aquitard sediments and its role in aquifer geochemistry." Nature, 
349(6306), 233-235. 
Metcalf & Eddy Inc. (1991). Wastewater Engineering:  Treatment, Disposal, and 
Reuse, McGraw-Hill Inc., New York, NY. 
Mihelcic, J. R., and Luthy, R. G. (1991). "Sorption and microbial degradation of 
naphthalene in soil-water suspensions under denitrification conditions." 
Environmental Science & Technology, 25(1), 169-177. 
Miller, C. T. (1984). "Modeling of sorption and desorption phenomena for 
hydrophobic organic contaminants in saturated soil environments," The 
University of Michigan. 
Miller, C. T., and Weber, W. J. (1988). "Modeling the sorption of hydrophobic 
contaminant by aquifer material 2. column reactor systems." Water Research, 
22(465-474). 
Miller, M. E., and Alexander, M. (1991). "Kinetics of bacterial degradation of 
benzylamine in a montomorilllonite suspension." Environmental Science & 
Technology, 24, 240-245. 
 
 290 
 
Molz, F. J., and Widdowson, M. A. (1988). "Internal inconsistencies in disperision-
dominated models that incorporate chemical and microbial kinetics." Water 
Resources Research, 24, 615-619. 
Mukherji, S., Peters, C. A., and Weber Jr., W. J. (1997). "Mass transfer of 
polynuclear aromatic hydrocarbons from complex DNAPL mixtures." 
Environmental Science & Technology, 31(2), 416-423. 
Mulder, H., Breure, A. M., Andel, J. G. V., Grotenhuis, J. T. C., and Rulkens, W. H. 
(2000). "Effect of mass-transfer limitations on bioavailability of sorbed 
naphthalene in synthetic model soil matrices." Environmental Toxicology and 
Chemistry, 19(9), 2224-2234. 
Mulder, H., Breure, A. M., and Rulkens, W. H. (2001). "Application of a mechanistic 
desorption-biodegradation model to describe the behavior of polycyclic 
aromatic hydrocarbons in peat soil aggregates." Chemosphere, 42, 285-299. 
Murphy, E. M., Ginn, T. R., Chilakapati, A., Resch, C. T., Phillips, J. L., Wietsma, T. 
W., and Spadoni, C. M. (1997). "The influence of physical hetergeneity on 
microbial degradation and distribution in porous media." Water Resources 
Research, 33(5), 1087-1103. 
National Research Council. (1993). In situ Bioremediation: When Does it Work?, 
National Academy Press, Washington, D.C. 
Odencrantz, J. E. (1991). "Modeling the biodegradation kinetics of dissolved organic 
contaminants in a heterogeneous two-dimensional aquifer," University of 
Illinois at Urbana-Champaign. 
Ogram, A. V., Jessup, R. E., Qu, L. T., and Rao, P. S. C. (1985). "Effects of sorption 
on biological degradation rates of (2,4-dichlorophenoxy) acetic acid in soils." 
Applied and Environmental Microbiology, 49(3), 582-587. 
Oliviera, I. B., Demond, A. H., and Salehzadeh, A. (1996). "Packing of sands for the 
production of homogeneous porous media." SOIL SCIENCE SOCIETY OF 
AMERICA JOURNAL, 60(1), 49-53. 
Oya, S., and Valocchi, A. J. (1998). "Transport and biodegradation of solutes in 
stratified aquifers under enhanced in situ bioremediation conditions." Water 
Resources Research, 34(12), 3323-3334. 
Park, J. H., Zhao, X. D., and Voice, T. C. (2002). "Development of a kinetic basis for 
bioavailability of sorbed naphthalene in soil slurries." Water Research, 36(6), 
1620-1628. 
Pearce, A. E., Voudrias, E. A., and Whelan, M. P. (1994). "Dissolution of TCE and 
TCA pools in saturated subsurface systems." Journal of Environmental 
Engineering, 120(5), 1191-1206. 
 
 291 
 
Peters, C. A., and Luthy, R. G. (1993). "Coal tar dissolution in water miscible 
solvents:  Experimental evaluation." Environmental Science & Technology, 
27(13), 2831-2843. 
Peyton, B. M., Skeen, R. S., Hooker, B. S., Lundman, R. W., and Cunningham, A. B. 
(1995). "Evaluation of bacterial detachment rates in porous media." Applied 
Biochemistry and Biotechnology, 51, 785-797. 
Phanikumar, M. S., and McGuire, J. T. (2004). "A 3D partial-equilibrium model to 
simulate coupled hydrogeological, microbiological, and geochemical 
processes in subsurface systems." Geophysical Research Letters, 31(11), Art. 
No. L11503. 
Ramaswami, A., and Luthy, R. G. (1997). "Measuring and modeling physicochemical 
limitations to bioavailability and biodegradation." Manual of Environmental 
Microbiology, C. J. Hurst, G. R. Knudsen, M. J. McInerney, L. D. 
Stetzenbach, and M. V. Walter, eds., ASM Press, Washington DC, 721-729. 
Rhykerd, R. L., Weaver, R. W., and McInnes, K. J. (1995). "Influence of salinity on 
bioremediation of oil in soil." Environmental Pollution, 90(1), 127-130. 
Riis, V., Kleinsteuber, S., and Babel, W. (2003). "Influence of high salinities on the 
degradation of diesel fuel by bacterial consortia." Canadian Journal of 
MIcrobilogy, 49, 713-721. 
Rijnaarts, H. M., Bachmann, A., Jumelet, J. C., and Zehnder, A. J. B. (1990). "Effect 
of desorption and intraparticle mass transfer on the aerobic biominerilization 
of hexachlorocyclohexane in a contaminated calcareous soil." Environmental 
Science & Technology, 24, 1349-1354. 
Rittmann, B. E. (1993). "The significance of biofilms in porous media." Water 
Resources Research, 29(7), 2195-2202. 
Rittmann, B. E., and McCarty, P. L. (1980). "Model of steady-state-biofilm kinetics." 
Biotechnology and Bioengineering, 22(22), 2343-2357. 
Rittmann, B. E., Seagren, E. A., Wrenn, B. A., Valocchi, A. J., Ray, C., and Raskin, 
L. (1994). In situ bioremediation, Noyes Publishers, Park Ridge, NJ. 
Rittmann, B. E., Valocchi, A. J., Seagren, E. A., Ray, C., and Wrenn, B. (1992). "A 
critical review of in situ bioremediation." Gas Res. Inst., Top. Rep. 
Robbins, G. A. (1989). "Methods for determining transverse dispersion coefficients 
of porous media in laboratory column experiments." Water Resources 
Research, 25(6), 1249-1258. 
Rockne, K. J., Chee-Sanford, J. C., Sandford, R. A., Hedlund, B. P., Staley, J. T., and 
Strand, S. E. (2000). "Anaerobic naphthalene degradation by microbial pure 
 
 292 
 
cultures under nitrate-reducing conditions." Applied and Environmental 
Microbiology, 66(4), 1595-1601. 
Rockne, K. J., and Strand, S. E. (1998). "Biodegradation of bicyclic and polycyclic 
aromatic hydrocarbons in anaerobic enrichments." Environmental Science & 
Technology, 32, 2962-2967. 
Rockne, K. J., and Strand, S. E. (2001). "Anaerobic biodegradation of naphthalene, 
phenanthrene, and biphenyl by a denitrifying enrichment culture." Water 
Research, 35(1), 291-299. 
Rumer, R. R., Jr. (1962). "Longitudinal dispersion in stready and unsteady flow." J. 
Hydraulics Div., Proc. Amer. Soc. Civil Eng., 88, 147-172. 
Salanitro, J. P., Johnson, P. C., Spinnler, G. E., Maner, P. M., Wisniewski, H. L., and 
Bruce, C. (2000). "Field scale demonstration of enhanced MTBE 
bioremediation through aquifer bioaugmentation and oxygenation." 
Environmental Science & Technology, 34(19), 4152-4162. 
Schafer, W., and Kinzelbach, W. (1992). "Stochastic modeling of in situ 
bioremediation in heterogeneous aquifers." Journal of Contaminant 
Hydrology, 10(1), 47-73. 
Scheibenbogen, K., Zytner, R. G., Lee, H., and Trevors, J. T. (1994). "Enhanced 
removal of selected hydrocarbons from soil by Pseudomonas aeruginosa UG2 
biosurfactants and some chemical surfactants." Journal of Chemical 
Technology and Biotechnology, 59(1), 53-59. 
Schicke, C. (1996). "Innovative evaluation methods for bioremediation: flow-cell 
construction and tracer analyses," Northwestern University, Evanston, IL. 
Schirmer, M., and Barker, J. F. (1998). "A study of long-term MTBE attenuation in 
the Borden Aquifer, Ontario, Canada." Ground Wat. Monitor. Remed., 18(2), 
113-122. 
Schirmer, M., Butler, B. J., Barker, J. F., Church, C. D., and Schirmer, K. (1999). 
"Evaluation of biodegradation and dispersion as natural attenuation processes 
of MTBE and benzene at the Borden field site." Phys. Chem. Earth (B), 24(6), 
557-560. 
Schirmer, M., Molson, J. W., Frind, E. O., and Barker, J. F. (2000). "Biodegradation 
modeling of a dissolved gasoline plume applying independent laboratory and 
field parameters." Journal of Contaminant Hydrology, 46, 339-374. 
Schwartz, F. W., and Zhang, H. (2002). Foundamentals of Ground Water, John Wiley 
& Sons, Inc., New York. 
 
 293 
 
Seagren, E. A. (1994). "Quantitative evaluation of flushing and biodegradation for 
enhancing in situ dissolution of nonaqueous phase liquids," Ph.D. 
Dissertation, University of Illinois, Urbana-Champaign. 
Seagren, E. A., and Moore, T. O. (2003). "Nonaqueous phase liquid pool dissolution 
as a function of average pore water velocity." JOURNAL OF 
ENVIRONMENTAL ENGINEERING-ASCE, 129(9), 786-799. 
Seagren, E. A., Rittmann, B. E., and Valocchi, A. J. (1994). "Quantitative evaluation 
of the enhancement of NAPL-pool dissolution by flushing and 
biodegradation." Environmental Science & Technology, 28(5), 833-839. 
Shan, C., and Stephens, D. B. (1994). "Recommendations for usage of SURFER to 
gridding model results." Ground Water, 32(3), 503-506. 
Sheng, G. Y., X.R.Wang, Wu, S. N., and Boyd, S. A. (1998). "Enhanced sorption of 
organic contaminants by smectitic soils modified with a cationic surfactant." 
Journal of Environmental Quality, 27(4), 806-814. 
Silliman, S. E., Zheng, L., and Conwell, P. (1998). "The use of laboratory 
experiments for the study of conservative solute transport in heterogeneous 
porous media." Hydrogeology Journal, 6, 166-177. 
Sims, R. M., and Overcash, M. R. (1983). "Fate of polynuclear aromatic compounds 
(PNAs) in soil-plant systems." Residue Review, 88, 1-68. 
Singleton, I. (1994). "Microbial metabolism of Xenobiotics: fundamental and applied 
research." Journal Chemical Technology and Biotechnology, 59(1), 9-23. 
Smets, B. F., Chandran, K., and Riefler, R. G. (2001). "Estimation of biokinetics 
parameters. Section 3-3-2." Environmental Engineering Process Laboratory 
Manual, S. E. Powers, J. J. Biosgni, Jr., J. G. Burken, and K. Pagilla, eds. 
Smith, S. A., and Pretorius, W. A. (2002). "The conservative behavior of 
fluorescein." Water SA, 28(4), 403-406. 
Smith, S. C., Ainsworth, C., Traina, S. J., and Hicks, R. J. (1992). "Effect of sorption 
on the biodegradation of quinoline." Soil Science Society of America Jounral, 
56, 737-746. 
Speitel, G. E., and DiGiano, F. A. (1987). "The bioregeneration of GAC used to treat 
micropollutants." Journal American Water Works Association, 79, 64-73. 
Sturman, P. J., Stewart, P. S., Cunningham, A. B., Bouwer, E. J., and Wolfram, J. H. 
(1995). "Engineering scale-up in situ bioremediation processes: a review." 
Journal of Contaminant Hydrology, 19, 171-203. 
 
 294 
 
Sudicky, E. A., Gillham, R. W., and Frind, E. O. (1985). "Experimental investigation 
of solute transport in stratified porous media. 1. The nonreactive case." Water 
Resources Research, 21, 1035-1041. 
Sun, Y., Pertersen, J. N., and Clement, T. P. (1999). "Analytical solutions for multiple 
species reactive transport in multiple dimensions." Journal of Contaminant 
Hydrology, 35(4), 429-440. 
Sutherland, J. B., RafII, F., Khan, A. A., and Cerniglia, C. E. (1995). "Mechanisms of 
polycylic aromatic hydrocarbon degradation." Microbial Transformation and 
Degradation of Toxic Organic Chemicals, Wiley-Liss, Inc., 269-306. 
Sutton, P. A., and Barker, J. F. (1985). "Migration and attenuation of selected 
organics in a sandy aquifer-A natural gradient experiment." Ground Water, 
23(10-16). 
Szecsody, J. E., and Bales, R. C. (1989). "Sorption kinetics of low-molecular-weight 
hydrophobic organic compounds on surface-modified silica." Journal of 
Contaminant Hydrology, 4, 181-203. 
Szecsody, J. E., Brockman, F. J., Wood, B. D., Steile, G. P., and Truex, M. J. (1994). 
"Transport and biodegradation of quinoline in horizontally stratified porous 
media." Journal of Contaminant Hydrology, 15, 277-304. 
Taylor, S. W., and Jaffe, P. R. (1990). "Substrate and biomass transport in a porous 
medium." Water Resources Research, 26(2153-2159). 
Tchobanoglous, G., and Burton, F. (1991). Wastewater engineering: treatment, 
disposal, and reuse, Metcalf & Eddy, Irwin/Mcgraw-Hill. 
Thornton, S. F., Quigley, S., Spence, M. J., Banwart, S. A., Bottrell, S., and Lerner, 
D. N. (2001). "Processes controlling the distribution and natural attenuation of 
dissolved phenolic compounds in a deep sandstone aquifer." Journal of 
Contaminant Hydrology, 53(3-4), 233-267. 
Trevor, J. T. (1996). "Sterilization and inhibition of microbial activity in soil." 
Journal of Microbiological Methods, 26, 53-59. 
Volkering, F. (1996). "Bioavailability and biodegradation of polycyclic aromatic 
hydrocarbons," Ph.D thesis, Agricultural University of Wageningen, 
Wageningen, The Netherlands. 
Volkering, F., Breure, A. M., and Van Andel, J. G. (1993). "Effect of microorganism 
on the bioavailability and biodegradation of crystalline naphthalene." Applied 
Microbiology and Biotechnology, 40, 535-545. 
 
 295 
 
Voudrias, E. A., and Yeh, M. F. (1994). "Dissolution of a toluene pool under constant 
and variable hydraulic gradients with implications for aquifer remediation." 
Ground Water, 32(2), 305-311. 
Wang, Y. L., Banziger, J., Dubin, P. L., Filippelli, B., and Nuraje, N. (2001). 
"Adsorptive partitioning of an organic compound onto polyelectrolyte-
immobilized micelles on porous glass and sand." Environmental Science & 
Technology, 35, 2608-2611. 
Weber, W., Jr, McGinley, P., and Katz, L. (1991). "Sorption phenomena in 
subsurface systems: concepts, models and effects on contaminant fate and 
transport." Water Research, 25, 499-528. 
Weber, W. J., and Miller, C. T. (1988). "Modeling the sorption of hydrophobic 
contaminants by aquifer materials. 1. rates and equilibria." Water Research, 
22(4), 457-464. 
Weber, W. J., and Smith, E. H. (1987). "Simulation and design models for adsorption 
processes." Environmental Science & Technology, 21, 1040-1050. 
Weiner, R. W., Seagren, E., Arnosti, C., and Quintero, E. (1999). "Bacterial survival 
in biofilms: Probes for exopolysaccharide and its hydrolysis and measurement 
of interphase mass fluxes." Methods in Enzymology, 310, 403-426. 
Welty, J. R., Wicks, C. E., and Wilson, R. E. (1984). Fundamentals of Momentum, 
Heat, and Mass Transfer, John Wiley & Sons, Inc., New York, NY. 
Whelan, M. P., Voudrias, E. A., and Pearce, A. (1994). "DNAPL pool dissolution in 
saturated porous media: Procedure development and preliminary results." 
Journal of Contaminant Hydrology, 15(3), 223-237. 
Whitman, B. E., Lueking, D. R., and Mihelcic, J. R. (1998). "Naphthalene update by 
a Pseudomonas fluorescens isolate." Canadian Journal of Microbiology, 44, 
1086-1093. 
Willumsem, P. A., and Arvin, E. (1999). "Kinetics of degradation of surfactant-
solubilized fluoranthene by a Sphingomonas paucimobilis." Environmental 
Science & Technology, 33, 2571-2578. 
Wilson, L. P., D'Adamo, P. C., and Bouwer, E. J. (1997). "Bioremediation of BTEX, 
naphthalene, and phenanthrene in aquifer material using mixed oxygen/nitrate 
electron acceptor conditions." National Risk Management Research 
Laboratory, USEPA. 
Wood, B. D., Dawson, C. N., Szecsody, J. E., and Streile, G. P. (1994). "Modeling 
contaminant transport and biodegradation in a layered porous media system." 
Water Resources Research, 30, 1833-1845. 
 
 296 
 
Woozinski, R. S., and Johnson, M. V. (1968). "Yields of baterial cells from 
hydrocarbons." Applied Environmental Microbiology, 23, 1077-1081. 
Yang, X., Erickson, L. E., and Fan, L. T. (1994). "Dispersive-convective 
characteristics in biorestoration of contaminated soil with a heterogeneous 
formation." Journal of Hazardous Materials, 38(163-185). 
Zeng, Y. M. (2004). "Naphthalene Pathway Map." University of Minnesota. 
http://umbbd.ahc.umn.edu/naph/naph_map.html. 
Zhang, X., Sullivan, E. R., and Young, L. Y. (2000). "Evidence for aromatic ring 
reduction in the biodegradation pathway of carboxylated naphthalene by a 
sulfate reducing consortium." Biodegradation, 11, 117-124. 
Zhang, Y. M., and Miller, R. M. (1992). "Enhanced Octadecane Dispersion and 
Biodegradation by a Pesudomonas Rhamnolipid Surfactant (Biosurfactant)." 
APPLIED AND ENVIRONMENTAL MICROBIOLOGY, 58(10), 3276-3282. 
Zheng, C. (1990). "MT3D. A modular three-dimensional transport model for 
simulation of advection, dispersion and chemical reactions of contaminants in 
groundwater systems." U.S.E.P.A. report.