ABSTRACT
Title of Document: THE EFFECTS OF CHANGES IN LAND
COVER AND LAND USE ON NUTRIENT
LOADINGS TO THE CHESAPEAKE BAY
USING FORECASTS OF URBANIZATION
Allen Derrick Roberts, Doctor of Philosophy,
2010
Directed By: Professor, Stephen D. Prince, Department of
Geography
This dissertation examined the effects of land cover and land use (LC/LU) change
on nutrient loadings (mass for a specified time) to the Chesapeake Bay, after future
projections of urbanization were applied. This was accomplished by quantifying the
comprehensive impacts of landscape on nutrients throughout the watershed. In order to
quantify forecasted impacts of future development and LC/LU change, the current (2000)
effects of landscape composition and configuration on total nitrogen (TN) and total
phosphorus (TP) were examined. The effects of cover types were examined not only at
catchment scales, but within riparian stream buffer to quantify the effects of spatial
arrangement. Using the SPAtially Referenced Regressions On Watershed Attributes
(SPARROW) model, several compositional and configurational metrics at both scales
were significantly (p value ? 0.05) correlated to nutrient genesis and transport and helped
estimate loadings to the Chesapeake Bay with slightly better accuracy and precision.
Remotely sensed forecasts of future (2030) urbanization were integrated into
SPARROW using these metrics to project TN and TP loadings into the future. After
estimation of these metrics and other LC/LU-based sources, it was found that overall
nutrient transport to the Chesapeake Bay will decrease due to agricultural land losses and
fertilizer reductions. Although point and non-point source urban loadings increased in
the watershed, these gains were not enough to negate decreased agricultural impacts.
In catchments forecasted to undergo urban sprawl conditions by 2030, the
response of TN locally generated within catchments varied. The forecasted placement of
smaller patches of development within agricultural lands of higher nutrient production
was correlated to projected losses. However, shifting forecasted growth onto or adjacent
to existing development, not agricultural lands, resulted in projected gains. This
indicated the importance of forecasted spatial arrangement to projected TN runoff from
the watershed.
In conclusion, comprehensive landscape analysis resulted in differences in
simulations of current and future nutrient loadings to the Chesapeake Bay, as a result of
urbanization and LC/LU change. With eutrophication from excess nutrients being the
primary challenge to the estuary, information gained from the estimation of these effects
could improve the future management and regulation of the Chesapeake Bay.
THE EFFECTS OF CHANGES IN LAND COVER AND LAND USE ON
NUTRIENT LOADINGS TO THE CHESAPEAKE BAY USING FORECASTS OF
URBANIZATION
By
Allen Derrick Roberts
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
2010
Advisory Committee:
Professor Stephen D. Prince, PhD, Chair
Professor Ralph Dubayah, PhD
Professor In-Young Yeo, PhD
Professor Brian Needelman, PhD
Gregory Schwarz, PhD
? Copyright by
Allen Derrick Roberts
2010
ii
Dedication
This dissertation is dedicated to my mother who passed away back in 1999.
Without her wisdom, knowledge, and drive for me to succeed, I would not be in the
place where I am in today. Additionally, I would like to dedicate this work to my
wife who influenced me to go to the University of Maryland and receive all of the
training needed to complete not only my dissertation, but a life-long dream. To the
end, thank you both for everything that you have given me. I will always love you
both and it is more than you will ever know.
iii
Acknowledgements
There are so many people whom I would like to thank in helping me with the
successful completion of this dissertation. First of all, to my advisor Dr. Stephen D.
Prince (Department of Geography), without your insight, knowledge, and thoughtful
suggestions regarding the areas of: the Chesapeake Bay; land cover/land use (LC/LU)
change, hydrology, modeling, remote sensing, and statistical analysis, this research
could not have been realistically accomplished. Furthermore, your in-depth analysis
and guidance in helping to nurture my formulation of research goals and technical
writing skills were invaluable. It so greatly appreciated.
To my University of Maryland committee members of: Dr. Dubayah; Dr.
Yeo; and Dr. Needelman, this research would of also not been possible without your
invaluable assessments and critiques needed to improve upon the findings and future
directions presented in this dissertation. In regards to Dr. Dubayah and Dr. Yeo (both
of the Department of Geography), your suggestions on statistical modeling and model
uncertainty were paramount to retrieving the results discovered by this research. In
regards to Dr. Needelman (Department of Environmental Science and Technology),
our conversations about the potential impacts of climate change and nutrient loadings
(mass for a specified time) from sources, such as combined animal feeding operations
(CAFOs) on this research were extremely helpful and provided insight about its
possible future direction.
To Dr. Gregory Schwarz, a special appreciation and gratitude on my behalf
towards you is warranted. Without your countless responses to my emails, meetings
with me at the United States Geological Survey (USGS) headquarters in Reston,
iv
Virginia, phone conversations, and invaluable insight and knowledge into the inner
workings of the SPAtially Referenced Regressions On Watershed Attributes
(SPARROW) model, this dissertation could have never been completed. Through
your mentoring and encouragement, learning and understanding the SPARROW
model was not only quite easier, but rather enjoyable. Thank you once again for
taking the time out of your busy schedule to accommodate all of my requests in
regards to the SPARROW model and my research. Additionally to John Brakebill
and Dr. Stephen Preston, also of the USGS, without your guidance, original datasets,
and SAS codes representing the Chesapeake Bay version of the SPARROW models,
this research would have not been possible. To the both of you, thank you very much
and it was truly the starting point for the modeling analysis conducted in this
research. Furthermore, I would also like to thank Dr. Richard Alexander and Dr.
Richard Smith of the USGS in Reston for helpful input regarding how to utilize the
SPARROW model more effectively in nutrient loading assessments for the Bay.
I would also like to thank Dr. Scott Goetz of the Woods Hole Research Center
in Falmouth, Massachusetts and Claire Jantz of Shippensburg University in
Shippensburg, Pennsylvania. The funding of this project was conducted in part under
a National Aeronautics and Space Administration (NASA) grant received by Dr.
Goetz that was interested in the outcomes of this research. Without access to
projected LC/LU datasets created by Dr. Goetz and Dr. Jantz for the Chesapeake Bay
watershed and their insight into how LC/LU should be applied in this current and
future modeling of nutrients to the Bay, this research could not have been conducted.
v
To my all of my Department of Geography cohorts, especially those who are
or were part of the Dr. Prince group, I would love to thank you for the memories,
laughter, and lifelong friendships that were developed here. So to Jyothy, Khaldoun,
Nancy, and Katie, I would like to thank you all personally for the academic
discussions and scholarly exchanges that we were able to bounce off one another for
the past few years. It was a learning experience that I will cherish. To Tao, Karl,
Karen, Sage, Evan, Kelley, Jennifer, Angira, Ritvik, Derrick and others, I wanted to
say thank you all for the hallway conversations, updates, and overall understanding of
this process that we all chose to endure for the sake of academic enlightenment. I
wish you all the best in life and hope through some venue, we can all stay connected.
To my family and friends, what more can I say but thank you understanding
about my desire to come back to school to finish up my doctorate. Although an easy
decision for me, I know that it was sometimes difficult on you all due to the time
constraints, dedication, and manpower that was required of me to complete my
dissertation. To my aunts (Maggie, Estella, Minnie, Anna, and Joyce), uncles (Billy,
Ben, Joe, and Lurie), cousins (Barry, Kim, Mary, Tiffany, Tonya, Tatiana, Jeffrey,
Tiffany and Tambra), family on my wife?s side (Mr. and Mrs. Johnson and Girard)
and friends (Mike, Marcus, Alex, Carl, James, and Shibu) your support has been vital
even though I have not been able to spend as much time with you all, as I would of
hoped. However, I still love you all and will always be grateful to know each of you.
Last, but never least, I want to thank my wife Kischa for the undying support,
love, and foundation that she has given me as we have journeyed on this path
together. Without her, even the thought of this endeavor could not have occurred.
vi
You will always have a special place in my heart for allowing me to undertake the
greatest challenge of life. This is the culmination of our hard work and we did it!
vii
Table of Contents
Dedication..................................................................................................................ii
Acknowledgments......................................................................................................iii
Table of Contents.......................................................................................................vi
List of Tables............................................................................................................. ix
List of Figures............................................................................................................xi
Chapter 1: Introduction.............................................................................................1
1.1 Background and nutrient history......................................................................1
1.2 Current urbanization in the Chesapeake Bay...................................................3
1.3 Forecasted urbanization in the Chesapeake Bay..............................................5
1.4 Regional implications of effects of currents and forecasted urbanization on
nutrients in the Chesapeake Bay......................................................................8
1.5 Priority questions regarding effects of current and forecasted urbanization on
nutrients in the Chesapeake Bay......................................................................11
1.6 Objectives.........................................................................................................12
1.7 The dissertation and its organization................................................................13
Chapter 2: Effects of urban and non-urban land cover on nitrogen and phosphorus
runoff to Chesapeake Bay.......................................................................16
2.1 Abstract............................................................................................................16
2.2 Introduction......................................................................................................17
2.3 Materials and methods..................................................................................... 22
2.3.1 Chesapeake Bay watershed land cover and land use data.......................22
2.3.2 Landscape metrics.....................................................................................23
2.3.3 The SPARROW model...............................................................................24
2.3.4 Model calibration......................................................................................27
2.4 Results..............................................................................................................29
2.4.1 TN model...................................................................................................29
2.5.1 TP model................................................................................................... 35
2.5 Discussion........................................................................................................40
2.5.1 Effects of non-urban and urban LC/LU on TN and TP runoff to the
Chesapeake Bay........................................................................................40
2.5.2 Comparisons with B & P models of the Chesapeake Bay watershed.......47
2.5.3 Comparisons between Chesapeake Bay HSPF and SPARROW TN
and TP models...........................................................................................52
2.6 Conclusions......................................................................................................53
Chapter 3: Effects of projected future urban land cover on nitrogen and phosphorus
runoff to Chesapeake Bay.......................................................................56
3.1 Abstract............................................................................................................56
3.2 Introduction......................................................................................................57
3.3 Materials and methods..................................................................................... 63
viii
3.3.1 Future Chesapeake Bay watershed land cover and land use................... 63
3.3.2 Landscape metrics and Chesapeake Bay SPARROW models...................64
3.3.3 Projected fertilizer and manure applications and point source
loadings......................................................................................................70
3.4 Results..............................................................................................................74
3.4.1 TN..............................................................................................................74
3.4.2 TP..............................................................................................................77
3.5 Discussions.......................................................................................................78
3.5.1 Agricultural land losses and reductions in total and agricultural non-
point source loadings...............................................................................78
3.5.2 Forests (and other non-agricultural and non-urban) land losses and
reductions in loadings...............................................................................81
3.5.3 Impervious surface area gains and changes in urban non-point
loadings.....................................................................................................83
3.5.4 Changes with point sources with urbanization.........................................86
3.5.5 Land-to-water delivery losses and reductions in non-urban, non-point
loadings.....................................................................................................87
3.6 Conclusions......................................................................................................90
Chapter 4: Effects of future urban sprawl on nitrogen runoff in subwatersheds
of Chesapeake Bay...................................................................................93
4.1 Abstract............................................................................................................93
4.2 Introduction......................................................................................................94
4.3 Materials and methods..................................................................................... 100
4.3.1 Categories of sprawling catchments.........................................................100
4.3.2 Results for the entire Chesapeake Bay watershed (Roberts and
Prince, 2010)..............................................................................................102
4.3.3 Yields used for 2030 TN runoff projections.............................................. 104
4.3.4 Sprawling catchments and parametric testing..........................................104
4.4 Results..............................................................................................................104
4.4.1 Categories of sprawling catchments.........................................................104
4.4.2 Sprawling catchments and SPARROW..................................................... 108
4.5 Discussion........................................................................................................109
4.5.1 Area-weighted mean urban (? 10% ISA) patch size.................................109
4.5.2 Applied fertilizer and manure applications.............................................. 112
4.5.3 Percentage of land on coastal plain......................................................... 118
4.5.4 Percentage of cropland.............................................................................121
4.5 Conclusions......................................................................................................124
Chapter 5: Lessons learned from current and projected future urban land cover in
Chesapeake Bay watershed?.................................................................126
5.1 Abstract???????????????????????????126
5.2 Introduction......................................................................................................127
5.3 Significance of findings for priority management questions?????.......132
5.4 Additional findings????????????.......................................... 147
5.5 Management and policy recommendations???????......................... 151
ix
5.5.1 The case for using more comprehensive LC/LU effects to estimate
nutrient loadings?????????????????????... 151
5.5.2 The case for targeting and prioritizing non-urban LC/LU for
urbanization???????????????????????.. 152
5.6 Future directions??????????????????.................... 154
x
List of Tables
Table 2.1 The sixteen Chesapeake Bay watershed cover classes created from the
2000 RESAC LC/LU, % ISA, and % TC maps??????........ 23
Table 2.2 Comparison of r2, RMSE, and AIC values between the Brakebill and
Preston, 2004 (1997 B & P) and the 2000 RESAC 31-1,000 m TN and
TP models????.......................................................................... 30
Table 2.3 Comparison of the significant (p value ? 0.05) landscape metrics in the
2000 RESAC 31 m TN and TP models??????????..... 31
Table 2.4 All significant (p value ? 0.05) variables in the 1997 B & P and 2000
RESAC TN models using 31-1,000 m riparian stream buffer widths
with model coefficients and p values (in parentheses). * Denotes all
significant landscape metrics. ONU = original land source area "not
utilized" as a result of being replaced with new surrogate landscape
source area metric in 2000 model runs. MNU = new landscape source
area and land-to-water delivery metrics "not utilized" in 1997 model
run. MIS = new landscape land-to-water metric found to be
"insignificant? (p value > 0.05) for that 2000 model run????...32
Table 2.5 Sensitivity analysis comparison of all significant (p value ? 0.05)
estimated variables between initial parametric and averaged
bootstrapped coefficient estimates in 2000 RESAC 31 m TN model.
* Denotes all landscape metrics found to be significant????....33
Table 2.6 All significant (p value ? 0.05) variables in the 1997 B & P and 2000
RESAC TP models using 31-1,000 m riparian stream buffer widths
with model coefficients and p values (in parentheses). * Denotes all
significant landscape metrics. ONU = original land source areas "not
utilized" as a result of being replaced with new surrogate landscape
source area metrics in 2000 model runs. MNU = new landscape
source area and land-to-water delivery metrics "not utilized" in 1997
model run????????????????..??????. 36
Table 2.7 Sensitivity analysis comparison of all significant (p value ? 0.05)
estimated variables between initial parametric and averaged
bootstrapped coefficient estimates in 2000 RESAC 31 m TP model.
* Denotes all landscape metrics found to be significant????....37
Table 3.1 The fourteen Chesapeake Bay watershed land cover classes
used??............................................................................................ 64
xi
Table 3.2 All significant (p value ? 0.05) variables in the 2000 RESAC 31 m TN
and TP SPARROW models (Roberts and Prince, 2010). RMSE = root
mean squared error. * Denotes all significant landscape metrics?. 68
Table 3.3 Comparison of 2000 and projected 2030 Chesapeake Bay watershed-
wide total discharged point source and applied fertilizer and manure
loadings (kg/yr), land-based source variables (ha), and land-to-water
delivery variables (%). * Denotes the averaged value of these land-
based source variables from all 2,339 catchments??.....................73
Table 3.4 Comparison of 2000 and projected 2030 Chesapeake Bay watershed-
wide total loadings (kg/yr) delivered to the estuary from all significant
sources and mean yield (kg/ha/yr) from all 2,339 catchments...........75
Table 3.5 Comparison of 2000 and projected 2030 Chesapeake Bay watershed-
wide total loadings (kg/yr) delivered to the estuary from all significant
sources and mean yield (kg/ha/yr) from all 2,339 catchments with
prediction errors for each year, range of 2030-2000 change, and range
of 2030-2000% change?????????????????.. 76
Table 4.1 Landscape metrics that were found to be significant (p value ? 0.05) in
the 2000 RESAC 31 meter SPARROW model????????. 99
Table 4.2 Variables in the 2000 RESAC 31 m SPARROW model found
significant (p value ? 0.05)????????????................ 103
Table 4.3 Comparison of the significant (p value ? 0.05) median changes from
2000 to 2030 for the RESAC 31 m SPARROW variables in the
catchments with losses and gains in locally generated TN yield.
* Denotes static hydrogeomorphic province variable that does not vary
from 2000 to 2030???????????????????..108
Table 4.4 Comparison of the median changes in locally generated applied
fertilizer and manure yield (kg/ha/yr) from 2000 to 2030 in catchments
with losses and gains in locally generated TN yield??????..109
Table 5.1 2000 watershed-wide compositions of significant land-to-water
delivery LC/LUs................................................................................ 134
Table 5.2 Comparison of 2000 and projected 2030 total loadings in kg/yr
delivered to the Chesapeake Bay from all significant sources in
RESAC 31 m models. * Denotes significant source variable in the
RESAC 31 m TN models that used the same inputs for 2000 and 2030
and was projected to decrease by 2030 due to conversions of
significant land-to-water delivery LC/LUs........................................145
xii
Table 5.3 Comparison of yield coefficient of determination (r2), and root mean
squared error (RMSE) between the 1997 Brakebill and Preston
(B & P) and 2000 RESAC 31 m SPARROW models....................... 148
Table 5.4 Comparison of 1997 B & P and 2000 RESAC 31 m SPARROW
model runs with the mean 1985-1994 and 2000 HSPF Phase 4.3 model
runs for total loadings in kg/yr delivered to the Chesapeake Bay..... 149
xiii
List of Figures
Figure 1.1 Example of how point and non-point N and P sources may be
transported to the Chesapeake Bay (Goldman, 2005)...................... 2
Figure 1.2 Current (2000) Chesapeake Bay watershed-wide maps of a) LC/LU, b)
% ISA, and c) % TC. In Figure 1.2a, URG =
Urban/residential/recreational grasses, EXT = Extractive, BAR =
Barren, DEF = Deciduous forest, EVF = Evergreen forest, MIX =
Mixed (deciduous-evergreen) forest, PAS = Pasture, CRO = Cropland,
NAT = ?Natural? grass, DWW = Deciduous wooded wetland, EVW =
Evergreen wooded wetland, and EMV = Emergent (sedge-herb)
wetland classification.........................................................................4
Figure 1.3 Schematic showing effects on the water balance of impervious cover
and surface runoff in the Chesapeake Bay when urbanization is
forecasted to increase over time (Federal Interagency Stream
Restoration Working Group, 1998)....................................................5
Figure 1.4 Projected (2030) Chesapeake Bay watershed-wide map of SLEUTH
estimated urban and non-urban LC/LU............................................. 7
Figure 2.1 The Chesapeake Bay watershed showing the locations of: a) Streams
and rivers draining the estuary and b) Urban centers located within its
boundaries..........................................................................................19
Figure 2.2 Illustration of a set of nested stream reaches and reservoir shorelines in
relation to monitoring stations (Alexander et al., 2002b). In model
calibration, reach i refers to any reach containing a monitoring station.
In model application, reach i refers to any reach for which a prediction
can be made........................................................................................25
Figure 2.3 Map of: a) 2,339 catchments used and b) Schematic diagram of an
example catchment area showing fixed 31, 62, 125, 250, 500, and
1,000 meters (m) riparian stream buffer width areas surrounding
stream reach and corresponding upstream and downstream runoff
paths................................................................................................... 27
Figure 2.4 Comparison between the natural log of observed versus predicted
loadings (kilograms per year (kg/yr)) in the: a) 2000 RESAC 31 m TN
model utilizing 87 Chesapeake Bay water quality N monitoring
stations and b) 2000 RESAC 31 m TP model utilizing 104 Chesapeake
Bay water quality P monitoring stations............................................30
xiv
Figure 2.5 Per catchment estimated TN yield (kg/ha/yr) map from the 2000
RESAC 31 m model of: a) Local generation and b) Delivery to the
Chesapeake Bay.................................................................................34
Figure 2.6 Per catchment estimated area-weighted mean urban (? 10% ISA) patch
size N yield (kg/ha/yr) map from the 2000 RESAC 31 m model of: a)
Local generation and b) Delivery to the Chesapeake Bay.................35
Figure 2.7 Per catchment estimated TP yield (kg/ha/yr) map from the 2000
RESAC 31 m model of: a) Local generation and b) Delivery to the
Chesapeake Bay.................................................................................38
Figure 2.8 Per catchment estimated area-weighted mean urban (? 10% ISA) patch
size P yield (kg/ha/yr) map from the 2000 RESAC 31 m model of: a)
Local generation and b) Delivery to the Chesapeake Bay.................38
Figure 2.9 Per catchment estimated area-weighted mean non-agricultural/non-
urban (NA/NU) patch size P yield (kg/ha/yr) map from the 2000
RESAC 31 m model of: a) Local generation and b) Delivery to the
Chesapeake Bay.................................................................................39
Figure 2.10 2,339 SPARROW catchment comparison of the natural log of the: a)
Significant (p value ? 0.05) 1997 NLCD source factor of total urban
land area (ha) used in the 1997 B & P TN and TP models versus the
significant source factor of 2000 area-weighted mean urban (? 10%
ISA) patch sizes (ha) used in the 2000 RESAC 31 m TN and TP
models and b) Significant 1997 NLCD source factor of non-
agricultural/non-urban land area (ha) used in the 1997 B & P TP model
versus the significant source factor of 2000 area-weighted mean non-
agricultural/non-urban patch sizes (ha) used in the 2000 RESAC 31 m
TP model............................................................................................41
Figure 2.11 Per catchment source factor metrics of: a) Area-weighted mean urban
(? 10% ISA) patch size (ha) in the 2000 RESAC 31 m TN and TP
models and b) Area-weighted mean non-agricultural/non-urban
(NA/NU) patch size (ha) in the 2000 RESAC 31 m TP model......... 42
Figure 2.12 Per catchment land-to-water delivery metrics of: a) Extractive land
(%), b) Area-weighted mean edge contrast of deciduous forest (%), c)
Cropland (%), and d) Evergreen forest within the riparian stream
buffer (%) in the 2000 RESAC 31 m TN model............................... 45
Figure 2.13 Per catchment land-to-water delivery metric of barren land within the
riparian stream buffer (%) in the 2000 RESAC 31 m TP model.......46
xv
Figure 2.14 Comparison of the breakdown of variance explained by variables in
the: a) 1997 B & P TN, b) 2000 RESAC 31 m TN, c) 1997 B & P TP,
and d) 2000 RESAC 31 m TP models. PS = point sources, AF =
applied fertilizer, AM = applied manure, AD = atmospheric
deposition, UL = urban land, CP = coastal plain, SS = small streams,
IS = intermediate streams, LS = large streams, R = reservoir, UV =
unexplained variance, AWMUPS = area-weighted mean urban (? 10%
ISA) patch size, PEL = percentage of extractive land, AWMECDF =
area-weighted mean edge contrast of deciduous forest, PCL =
percentage of cropland, PEFRSP = percentage of evergreen forest in
the riparian stream buffer, NANUL = non-agricultural/non-urban land,
and AWMNANUPS = area-weighted mean non-agricultural/non-urban
patch size............................................................................................48
Figure 2.15 Per catchment estimated difference between the 2000 RESAC 31 m
and 1997 B & P model in predicted yield (kg/ha/yr) delivered to the
Chesapeake Bay for: a) TN and b) TP...............................................51
Figure 2.16 Comparison of the natural log of the 1997 B & P and 2000 RESAC 31
m model estimated loadings (kg/yr) for the James, Patuxent, Potomac,
Rappahannock, Susquehanna, and the York Basins within the
Chesapeake Bay watershed for: a) TN) and b) TP. 1 = the James
River Basin, 2 = the Patuxent River Basin, 3 = the Potomac River
Basin, 4 = the Rappahannock River Basin, 5 = the Susquehanna River
Basin, and 6 = the York River Basin................................................. 51
Figure 2.17 Comparison of the natural log of the 2000 Phase 4.3 HSPF and 2000
RESAC 31 m model estimated loadings (kg/yr) for the James,
Patuxent, Potomac, Rappahannock, Susquehanna, and the York Basins
within the Chesapeake Bay watershed for: a) TN) and b) TP. 1 = the
James River Basin, 2 = the Patuxent River Basin, 3 = the Potomac
River Basin, 4 = the Rappahannock River Basin, 5 = the Susquehanna
River Basin, and 6 = the York River Basin....................................... 53
Figure 3.1 The Chesapeake Bay watershed showing the locations of: a) Streams
and rivers draining the estuary and b) Urban centers located within its
boundaries..........................................................................................59
Figure 3.2 2,339 catchments used in the: a) 2000 RESAC 31 m SPARROW
Chesapeake Bay TN and TP models with b) Schematic diagram of an
example catchment area showing fixed 31 m riparian stream buffer
width area surrounding stream reach and corresponding upstream and
downstream runoff paths....................................................................62
xvi
Figure 3.3 Per catchment estimated 2030-2000 difference maps of the total yield
in kg/ha/yr per year for: a) N and b) P delivered to the Chesapeake
Bay..................................................................................................... 77
Figure 3.4 Per catchment 2030-2000 difference maps of the area-weighted mean:
a) Non-agricultural/non-urban (NA/NU) and b) Urban (? 10% ISA)
patch size source metrics in ha...........................................................82
Figure 3.5 Per catchment estimated 2030-2000 difference map of the NA/NU
patch size yield in kg/ha/yr for P delivered to the Chesapeake
Bay..................................................................................................... 83
Figure 3.6 Per catchment estimated 2030-2000 difference maps of the area-
weighted mean urban (? 10% ISA) patch size yield in kg/ha/yr for: a)
N and b) P delivered to the Chesapeake Bay.....................................84
Figure 3.7 Per catchment 2030-2000 difference maps of the: a) Percentage of
extractive land, b) Area-weighted mean edge contrast of deciduous
forest land (%), c) Percentage of cropland, and d) Percentage of
evergreen forest land in the 31 m riparian stream buffer land-to-water
delivery metrics for the RESAC TN model.......................................89
Figure 3.8 Per catchment 2030-2000 difference map of the percentage of barren
land in the 31 m riparian stream buffer land-to-water delivery metric
for the RESAC TP model.................................................................. 90
Figure 4.1 The Chesapeake Bay watershed showing the locations of: a) Streams
and rivers draining the estuary and b) Urban centers located within its
boundaries..........................................................................................96
Figure 4.2 2,339 catchments used in the: a) 2000 RESAC 31 m SPARROW
Chesapeake TN model with b) Schematic diagram of an example
catchment area showing fixed 31 m riparian stream buffer width area
surrounding stream reach and corresponding upstream and
downstream runoff paths....................................................................98
Figure 4.3 Map of the Chesapeake Bay watershed locations of the: a) 157
catchments showing decreases in area-weighted mean urban (? 10%
ISA patch size with b) 106 catchments indicating losses and c) 51
catchments indicating gains in locally generated TN yield from 2000
to 2030............................................................................................... 101
xvii
Figure 4.4 Per catchment 2030-2000 change maps of: a) Losses in locally
generated TN yield (kg/ha/yr) with b) Decreases in the area-weighted
mean urban (? 10% ISA) patch size (ha), and c-d) Corresponding
frequency distributions of the number of catchments falling in each
category. For Figure 4.4c, category classes are: 1 = < -1.8, 2 = -1.8 - -
1.6, 3 = -1.6 - -1.4, 4 = -1.4 - -1.2, 5 = -1.2 - -1.0, 6 = -1.0 - -0.8, 7 = -
0.8 - -0.6, 8 = -0.6 - -0.4, 9 = -0.4 - -0.2, and 10 = -0.2 - 0.0 kg/ha/yr.
For Figure 4d, category classes are 1 = < -27, 2 = -27 - -24, 3 = -24 - -
21, 4 = -21 - -18, 5 = -18 - -15, 6 = -15 - -12, 7 = -12 - -9, 8 = -9 - -6, 9
= -6 - -3, and 10 = -3 - 0 ha................................................................106
Figure 4.5 Per catchment 2030-2000 change maps of: a) Gains in locally generated
TN yield (kg/ha/yr) with b) Decreases in the area-weighted mean urban
(? 10% ISA) patch size (ha) and c-d) Corresponding frequency
distributions of the number of catchments falling in each category. For
Figure 4.5c, category classes are: 1 = 0.0 -0.2, 2 = 0.2 -0.4, 3 = 0.4 - 0.6,
4 = 0.6 - 0.8, 5 = 0.8 - 1.0, 6 = 1.0 - 1.2, 7 = 1.2 - 1.4, 8 = 1.4 - 1.6, 9 =
1.6 - 1.8, and 10 = > 1.8 kg/ha/yr. For Figure 4.5d, category classes are
1 = < -27, 2 = -27 - -24, 3 = -24 - -21, 4 = -21 - -18, 5 = -18 - -15, 6 = -
15 - -12, 7 = -12 - -9, 8 = -9 - -6, 9 = -6 - -3, and 10 = -3 - 0 ha........107
Figure 4.6 Per catchment 2030-2000 change map of: a) Losses in locally
generated area-weighted mean urban (? 10% ISA) patch size yield
(kg/ha/yr) with losses in locally generated TN yield and b)
Corresponding frequency distribution of the number of catchments
falling in each category. Category classes are: 1 = < -0.09, 2 = -0.09 -
-0.08, 3 = -0.08 - -0.07, 4 = -0.07 - -0.06, 5 = -0.06 - -0.05, 6 = -0.05 -
-0.04, 7 = -0.04 - -0.03, 8 = -0.03 - -0.02, 9 = -0.02 - -0.01, and 10 = -
0.01 - 0.00 kg/ha/yr............................................................................111
Figure 4.7 Per catchment 2030-2000 change map of: a) Losses in locally
generated area-weighted mean urban (? 10% ISA) patch size yield
(kg/ha/yr) with gains in locally generated TN yield and b)
Corresponding frequency distribution of the number of catchments
falling in each category. Category classes are: 1 = < -0.09, 2 = -0.09 -
-0.08, 3 = -0.08 - -0.07, 4 = -0.07 - -0.06, 5 = -0.06 - -0.05, 6 = -0.05 -
-0.04, 7 = -0.04 - -0.03, 8 = -0.03 - -0.02, 9 = -0.02 - -0.01, and 10 = -
0.01 - 0.00 kg/ha/yr............................................................................112
Figure 4.8 Per catchment 2030-2000 change maps of: a) Application and b) Local
generation of fertilizer N yield (kg/ha/yr) with losses in locally
generated TN yield and c-d) Corresponding frequency distributions of
the number of catchments falling in each category. Category classes
are: 1 = < -30.0, 2 = -30.0 - -26.5, 3 = -26.5 - -23.0, 4 = -23.0 - -19.5, 5
= -19.5 - -16.0, 6 = -16.0 - -12.5, 7 = -12.5 - 9.0, 8 = -9.0 - -5.5, 9 = -
5.5 - -2.0, and 10 = > -2.0 kg/ha/yr.......,,,..........................................114
xviii
Figure 4.9 Per catchment 2030-2000 change maps of: a) Application and b) Local
generation of manure N yield (kg/ha/yr) with losses in locally
generated TN yield and c-d) Corresponding frequency distributions of
the number of catchments falling in each category. Category classes
are: 1 = < 2, 2 = 2 - 5, 3 = 5 - 8, 4 = 8 - 11, 5 = 11 - 14, 6 = 14 - 17, 7
= 17 - 20, 8 = 20 - 23, 9 = 23 - 26, and 10 = > 26
kg/ha/yr.............................................................................................. 115
Figure 4.10 Per catchment 2030-2000 change maps of: a) Application and b) Local
generation of fertilizer N yield (kg/ha/yr) with gains in locally
generated TN yield and c-d) Corresponding frequency distributions of
the number of catchments falling in each category. Category classes
are: 1 = < -4.8, 2 = -4.8 - -4.2, 3 = -4.2 - -3.6, 4 = -3.6 - -3.0, 5 = -3.0 -
-2.4, 6 = -2.4 - -1.8, 7 = -1.8 - 1.2, 8 = -1.2 - -0.6, 9 = -0.6 - -0.0, and
10 = > 0.0 kg/ha/yr.............................................................................117
Figure 4.11 Per catchment 2030-2000 change maps of: a) Application and b) Local
generation of manure N yield (kg/ha/yr) with gains in locally
generated TN yield and c-d) Corresponding frequency distributions of
the number of catchments falling in each category. Category classes
are: 1 = < 0.0, 2 = 0.0 - 0.5, 3 = 0.5 - 1.0, 4 = 1.0 - 1.5, 5 = 1.5 - 2.0, 6
= 2.0 - 2.5, 7 = 2.5 - 3.0, 8 = 3.0 - 3.5, 9 = 3.5 - 4.0, and 10 = > 4.0
kg/ha/yr.............................................................................................. 118
Figure 4.12 Per catchment map of: a) Percentage of land on the coastal plain (%)
with losses in locally generated TN yield and b) Corresponding
frequency distribution of the number of catchments falling in each
category. Category classes are: 1 = 0 - 10, 2 = 10 - 20, 3 = 20 - 30, 4 =
30 - 40, 5 = 40 - 50, 6 = 50 - 60, 7 = 60 - 70, 8 = 70 - 80, 9 = 80 - 90,
and 10 = 90 - 100%............................................................................120
Figure 4.13 Per catchment map of: a) Percentage of land on the coastal plain (%)
with gains in locally generated TN yield and b) Corresponding
frequency distribution of the number of catchments falling in each
category. Category classes are: 1 = 0 - 10, 2 = 10 - 20, 3 = 20 - 30, 4 =
30 - 40, 5 = 40 - 50, 6 = 50 - 60, 7 = 60 - 70, 8 = 70 - 80, 9 = 80 - 90,
and 10 = 90 - 100%............................................................................121
Figure 4.14 Per catchment 2030-2000 change map of: a) Decreases in percentage
of cropland (%) with losses in locally generated TN yield and b)
Corresponding frequency distribution of the number of catchments
falling in each category. Category classes are: 1 = < -0.9, 2 = -0.9 - -
0.8, 3 = -0.8 - -0.7, 4 = -0.7 - -0.6, 5 = -0.6 - -0.5, 6 = -0.5 - -0.4, 7 = -
0.4 - -0.3, 8 = -0.3 - -0.2, 9 = -0.2 - -0.1, and 10 = -0.1 - 0.0 %........ 123
xix
Figure 4.15 Per catchment 2030-2000 change map of: a) Decreases in percentage
of cropland (%) with gains in locally generated TN yield and b)
Corresponding frequency distribution of the number of catchments
falling in each category. Category classes are: 1 = < -0.9, 2 = -0.9 - -
0.8, 3 = -0.8 - -0.7, 4 = -0.7 - -0.6, 5 = -0.6 - -0.5, 6 = -0.5 - -0.4, 7 = -
0.4 - -0.3, 8 = -0.3 - -0.2, 9 = -0.2 - -0.1, and 10 = -0.1 -
0.0 %.................................................................................................. 124
Figure 5.1 The Chesapeake Bay watershed showing the locations of: a) Streams
and rivers draining the estuary and b) Urban centers located within its
boundaries..........................................................................................128
Figure 5.2 2,339 catchments used in the 2000 and 2030 runs of the: a) RESAC 31
m Chesapeake Bay SPARROW models with b) Schematic diagram of
an example catchment area showing fixed 31 m riparian stream buffer
width area surrounding stream reach and corresponding upstream and
downstream runoff paths....................................................................130
Figure 5.3 Per catchment source factor metric of: a) Area-weighted mean non-
agricultural/non-urban (NA/NU) patch size in ha with delivery of b) P
yield to Chesapeake Bay in kg/ha/yr in the 2000 RESAC 31 m TP
model..................................................................................................133
Figure 5.4 Per catchment land-to-water delivery metrics of: a) Extractive land
(%), b) Cropland (%), and c) Evergreen forest within the riparian
stream buffer (%) in the 2000 RESAC 31 m TN model and d) Barren
land within the riparian stream buffer (%) in the 2000 RESAC 31 m
TP model............................................................................................135
Figure 5.5 Per catchment source factor metrics of: a) Area-weighted mean urban
(? 10% ISA) patch size (ha) with delivery of b) N and c) P yield to
Chesapeake Bay in kg/ha/yr in the 2000 RESAC 31 m models........138
Figure 5.6 Per catchment 2030-2000 differences of the: a) Area-weighted mean
urban (? 10% ISA) patch size source metric in ha and b) Cropland (%)
projected to have the greatest impacts on catchments with the highest
2000 delivery of c) TN and d) TP yield to Chesapeake Bay in kg/ha/yr
by 2030.............................................................................................. 142
Figure 5.7 Per catchment 2030-2000 differences of the land-to-water delivery
metrics of: a) Extractive land (%) and b) Barren land within the
riparian stream buffer (%) in the RESAC 31 m models.................... 143
xx
Figure 5.8 Per catchment delivery of: a) TN and b) TP in yield to Chesapeake
Bay in kg/ha/yr in the 2030 RESAC 31 m models with 2030-2000
differences of delivery of: c) TN and d) TP yield to Chesapeake Bay in
kg/ha/yr in the RESAC 31 m models.................................................146
1
Chapter 1: Introduction
1.1 Background and nutrient history
The Chesapeake Bay is the largest estuary system occurring within the United
States (U.S.). Its watershed has an area of 166,534 km2 (64,299 mi2) covering the mid-
Atlantic region of the eastern seaboard. The watershed drains land masses in six states
(New York (NY), Pennsylvania (PA), Delaware (DE), Maryland (MD), West Virginia
(WV) and Virginia (VA)) and the District of Columbia (DC). The watershed also has the
largest ratio of land-to-water volume of any estuary in the U.S., making runoff from the
terrestrial landscape extremely important to the overall, ecological functioning of the
estuary (Shuyler et al., 1995).
Once considered a natural treasure due to its rich ecological diversity, recreational
uses, and numerous fishing industries that it supported, the Chesapeake Bay has been in
steady decline for over 400 years. The primary cause of this decline was the first
significant, anthropogenic-driven transformations in land cover and land use (LC/LU) by
European settlement that removed forests in favor of agricultural and urban cover types
in the watershed in the 17th and 18th centuries (Boesch, 2006). Since this initial influx of
population into the watershed, LC/LU changes have contributed greatly to problems of:
increased sedimentation, turbidity (water opaqueness), nutrients, eutrophication (nutrient
enrichment), hypoxia (low levels of dissolved oxygen (DO)), and anoxia (devoid of DO)
(Bratton et al., 2003). Furthermore, these changes have lead to a lowering in the
population of submerged aquatic vegetation (SAV), numerous aquatic species, and the
overall sustainability of the watershed and Chesapeake Bay itself (Bratton et al., 2003).
2
Environmental and ecological concerns resulting from nutrient runoff associated
with agriculture and increased urbanization have been cited as driving factors threatening
the long-term sustainability of the Chesapeake Bay (Burke et al., 2000; Paolisso and
Maloney, 2000). The two nutrients of greatest concern to degraded water quality and
sustainability are nitrogen (N) and phosphorus (P). These nutrients enter into the
Chesapeake Bay primarily through six major basins (the James, Patuxent, Potomac,
Rappahannock, Susquehanna, and York) and dozens of other smaller rivers and streams
at an estimated freshwater inflow rate of 8.7 x 106 m3/h (Sims and Coale, 2002). The
nutrients are transported to the estuary from point sources that include commercial,
industrial, and municipal wastewater discharges directly into streams and from non-point
(diffuse) sources that include atmospheric deposition, fertilizer and manure applications,
and impervious surfaces (Alexander et al., 2000b; and Castro and Driscoll, 2002) (Figure
1.1). The introduction of nutrients may have significant impacts on their biogeochemical
cycling not only at the local and regional scale, but quite possibly on a global level.
Figure 1.1: Example of how point and non-point N and P sources may be transported to
the Chesapeake Bay (Goldman, 2005).
3
1.2 Current urbanization in the Chesapeake Bay
Since the end of World War II, urbanization and rapid population growth in the
watershed have significantly increased the rate of genesis and delivery of nutrients to the
Chesapeake Bay from stormwater runoff and point-source discharges (Sims and Coale,
2002). In fact, studies have indicated that the current rates of N and P reaching the Bay
are 5-8 and 13-24 times their pre-colonial rates, respectively, (Boynton et al., 1995).
These increases are primarily due to the conversion of pervious non-urban cover types
that include wetlands and forests to impervious cover types indicative of urban
development, such as residential and commercial uses. Additionally, over the last couple
of decades, an even greater rate of the transference of non-urban LC/LUs by urbanization
has been indicated throughout the watershed (Goetz et al., 2004a-b; Jantz et al., 2005).
The LC/LU transformations can be directly correlated with the estimated doubling of
population from 8 million to approximately 16 million over this time period (McConnell,
1995). One estimate of current LC/LU indicates that the watershed is comprised of: 58%
forest, 33% agriculture, 8% urban, and 1% barren and wetlands (Sims and Coale, 2002).
Whereas estimates associated with population are not always easily quantified,
urban and non-urban LC/LUs can be measured with more certainty using remotely sensed
data derived from satellite imagery. In the watershed, LC/LU has been previously (1992)
and currently (2001) mapped using the National Land Cover Dataset (NLCD) created
from the Multi-Resolution Land Characteristic (MRLC) program (Homer et al., 2004).
However, current (2000) LC/LU has been mapped even more comprehensively using
LANDSAT data collected by the Regional Earth Science Applications Center (RESAC).
RESAC products that are available for analysis include watershed-wide LC/LU, percent
4
impervious surface area (% ISA), and percent tree cover (% TC) maps (Goetz et al.,
2003, 2004a, b; Jantz et al., 2005) (Figure 1.2a-c).
URG
EXT
BAR
DEF
EVF
MIF
PAS
CRO
NAT
DWW
EWW
EMW
0 40 80 120 16020
Kilometers ?
a)
Urban (> or = 10% ISA)
Non-Urban (< 10% ISA)
0 40 80 120 16020
Kilometers ?
b)
Non-Forest (< or = to 50% TC)
Forest (> 50% TC)
0 40 80 120 16020
Kilometers ?
c)
Figure 1.2: Current (2000) Chesapeake Bay watershed-wide maps of a) LC/LU, b) %
ISA, and c) % TC. In Figure 1.2a, URG = Urban/residential/recreational grasses, EXT =
Extractive, BAR = Barren, DEF = Deciduous forest, EVF = Evergreen forest, MIX =
Mixed (deciduous-evergreen) forest, PAS = Pasture, CRO = Cropland, NAT = ?Natural?
grass, DWW = Deciduous wooded wetland, EVW = Evergreen wooded wetland, and
EMV = Emergent (sedge-herb) wetland classification.
5
1.3 Forecasted urbanization in the Chesapeake Bay
By 2020, the population in the watershed is estimated to increase by
approximately 20% over levels estimated in the mid 1990s (McConnell, 1995). As
population is forecasted to increase over this time period and beyond, urbanized areas are
likely to continue to expand throughout the developed and undeveloped regions of the
watershed. With this expansion of urban growth, impervious surface areas capable of the
concentrated buildup of non-point N and P that can runoff to streams during storm events
will also continue to rise (Figure 1.3).
Figure 1.3: Schematic showing effects on the water balance of impervious cover and
surface runoff in the Chesapeake Bay when urbanization is forecasted to increase over
time (Federal Interagency Stream Restoration Working Group, 1998).
6
Projections of future urbanization have also been completed and incorporated
within studies of the Chesapeake Bay watershed in an effort to quantify the effects of this
increase in estimated impervious cover. The studies were conducted only for smaller
areas and were used to correlate this forecasted LC/LU change to a variety of
environmental processes. In a near comprehensive examination of the watershed (the
New York portion was excluded), Wickham et al. (2002) used non-linear regression
techniques to forecast urbanization to quantify changes in N and P export. In the
Patuxent Basin, Costanza et al. (2002) used an economic land-use model to project
development to determine its impacts on N and P export and net primary production
(NPP). For the headwaters of the York Basin, Im et al. (2003) used the continuation of
historical LC/LU change patterns provided by Caroline County?s (VA) comprehensive
development plan to evaluate forecasted urbanization on N and P export, peak flow,
sediment mass, and total runoff volume. Finally, in the Susquehanna Basin tributaries of
southeastern PA, Chang (2004) also used Lancaster County Planning Commission
projections to forecast growth to 2030 to quantify N and P export responses.
Although it has been shown that other methods have been used for forecasting
future urbanization to quantify environmental impacts in Chesapeake Bay subwatersheds,
another method has been developed with the distinct advantage of modeling these
estimated changes at the scale of the entire watershed with spatially-explicit forecasts of
the locations of development and measures of uncertainty (Jantz et al., in press). The
predictions were provided by a model, which is referred to as Slope, Land use,
Exclusion, Urban extent, Transportation, Hillshade or (SLEUTH), a cellular automata
simulation that represents urban growth/expansion processes in a spatially-explicit, two-
7
dimensional grid (Clarke et al., 1997). Similarly to the other methods described earlier,
SLEUTH has also been applied at much smaller regional scales in the watershed to
predict urbanization (Clarke and Gaydos, 1998; Carlson, 2004).
A large regional application of this model in the watershed, that consists of a
23,700 km2 section of the Baltimore-Washington metropolitan region (approximately
15% of the total watershed area), was used to forecast urbanization to 2030 (Jantz et al.,
2004; Claggett et al., 2005; Jantz and Goetz, 2005). This application described projected
urbanization under three distinct scenarios of development that included: 1) current trend,
2) managed/smart growth, and 3) ecological sustainable (Jantz et al., 2004). In another
watershed-wide application of SLEUTH, low-density development indicative of
continued urban sprawl has been modeled for the entire Chesapeake Bay watershed
(Jantz et al., in press) (Figure 1.4).
SLEUTH 2030 Urban
SLEUTH 2030 Non-Urban
0 40 80 120 16020
Kilometers ?
Figure 1.4: Projected (2030) Chesapeake Bay watershed-wide map of SLEUTH
estimated urban and non-urban LC/LU.
8
1.4 Regional implications of effects of current and forecasted urbanization on
nutrients in the Chesapeake Bay
According to Boesch et al. (2001), the multistate effort to restore the Chesapeake
Bay ecosystem by reducing the inputs of nutrients that stimulate organic over-enrichment
is one of the world?s most ambitious attempts at large-scale ecosystem restoration. By
the late 1990s, the Susquehanna Basin (the largest tributary to the estuary) was estimated
to deliver over 1 x 109 metric tons of sediment, 5 x 104 tons of N, and 3 x 103 tons of P to
the Chesapeake Bay annually (Preston and Brakebill, 1999; Langland and Cronin, 2003;
Goetz et al., 2004a; United States Environmental Protection Agency, 2009a). As an N-
limited system in which its excess quantities control the trophic condition of the estuary,
the reduction in N and surplus P delivered to its tributaries is paramount to reversing the
trends of eutrophication and related ecological and environmental declines seen in the
Chesapeake Bay (Weller, 2003). Thus, in an effort to mitigate these aforementioned
regional impacts, the Chesapeake Bay Agreement was enacted in 1987 to reduce
controllable nutrient levels 40% by 2000 from an initial 1985 baseline (Chesapeake
Executive Council, 1987). Additionally, new amendments to the agreement in 2000
have called for the removal of all nutrient and sediment-related water quality issues
impairing the Chesapeake Bay and its removal from the list of impaired waters under the
Clean Water Acts of 1972 , 1977, and 1990 (Wang et al., 2006).
Of the watershed traits such as: soil properties; vegetative cover; moisture
conditions; size; shape; topography; orientation; geology; and channel characteristics that
impact nutrient genesis and transport, cultural practices, such as LC/LU, may be the most
pivotal (Wolfe, 2001). Throughout the Chesapeake Bay watershed, increasing human-
9
populations have caused LC/LU changes from forest to agriculture to urban uses that
have resulted in greater freshwater flows and increases in N and P loadings (mass for a
specified time) from subwatersheds to the estuary (Fisher et al., 2006). Although these
LC/LU changes have increased inputs into the Chesapeake Bay in a widespread manner,
the pattern of response has been not been universal (Fisher et al., 2006). This fluctuation
in the pattern of response to nutrient genesis and transport is based upon the variability in
the properties of composition and configuration associated with these LC/LU changes.
In regards to LC/LU composition, current estimates indicate that the greatest
(about 40%) contributor of excess nutrients delivered to the Chesapeake Bay are through
agricultural practices, such as the non-points source delivery of applied fertilizer and
manure (Goetz et al., 2004a). The next largest contributor is from urban-induced point
discharges that account for 23% of the N and 34% of the P measured in the estuary
(Boesch et al., 2001). Additionally, atmospheric deposition accounts for 11% of the N
and 6% of the P measured in the estuary (Boesch et al., 2001). In conjunction with the
losses from non-point urban uses (commercial, industrial, residential, and transportation)
that account for 9% and 8% of the N and P, respectively, in the Chesapeake Bay (Boesch
et al., 2001), composition of LC/LU is a primary factor controlling regional nutrient
conditions reported in the estuary. Thus, with the anticipated increase of urban uses in
the watershed, distinct changes in the amount of N and P reaching the estuary are
expected to occur. However, the composition of these and other LC/LUs capable of
impacting N and P are not evenly distributed watershed-wide and are likely to continue to
cause variations in their long-term response.
10
Whereas the widespread pattern of uneven N and P responses to variability in
composition is more discernible in the watershed, the responses to LC/LU configuration
have only been extensively studied at much smaller, localized scales. Configuration can
be envisioned as the arrangement of LC/LU at the catchment, basin, or regional
watershed scale, but has been only quantified at the riparian stream buffer scale in
smaller regional studies. In a small tributary located on the Chesapeake Bay's western
shore, deciduous forest in the buffer adjacent to croplands removed over 80% of the
nitrate (NO3) and phosphate (PO4) in overland runoff and nearly 85% of the NO3 in
shallow subsurface runoff (Correll et al., 1992; Jordan et al., 1993; Boesch et al., 2001).
In localized watersheds in central MD, it was determined that urban riparian stream
buffer zones had a lower potential to remove N via denitrification than forested riparian
stream buffer zones (Groffman et al., 2002; Goetz et al., 2003). However, in other
smaller watersheds on the eastern shore of MD, some forested riparian stream buffers
were shown to increase runoff to streams (Norton and Fisher, 2000). Although an
attempt has been made to quantify the impacts of riparian LC/LU at more of a regional
scale (Baker et al., 2006), the comprehensive configurational effects of watershed-wide
riparian and catchment LC/LU on nutrient genesis and transport has yet to be completed.
Yet, similarly to composition, the uneven distribution of configuration throughout the
watershed should also cause discrepancies in the response of N and P.
By using the watershed-wide maps of current development and non-urban LC/LU
with forecasted development, the comprehensive effects of landscape properties on
present and projected nutrient loadings to the Chesapeake Bay could be quantified. This
is an important management and research priority that would give insight into the
11
potential correlations between the terrestrial and aquatic environments that were
previously unknown at regional scales. Additionally, the potential findings here may also
generate greater insight into the mechanisms and processes controlling biogeochemical
cycling at local and global scales. Finally, if these potential findings were to lead to
similar findings substantiated in other watersheds, a new approach to the management
and restoration of nutrient-enriched ecosystems could be introduced.
1.5 Priority questions regarding effects of current and forecasted urbanization on
nutrients in the Chesapeake Bay
Due to enactment of the Chesapeake Bay Agreements and enforcement of the
Clean Water Acts, new research depicting how forecasts of future urbanization and the
resulting LC/LU will affect total nitrogen (TN) and total phosphorus (TP) loadings to the
Chesapeake Bay is needed. This is in response to local, state, and federal government
entities endeavoring to meet these mutually agreed upon regional and federally-binding
nutrient reduction mandates. With that in mind, several priority management questions
regarding quantifying the effects of current and forecasted urbanization on nutrient
loadings include:
? What LC/LUs at catchment and riparian stream buffer scales are significant in the
current (2000) non-point (diffuse) sources and delivery of TN and TP loadings
estimated in the Chesapeake Bay watershed?
? What are the causes of the observed effects of LC/LUs at catchment and riparian
stream buffer scales on the current (2000) non-point (diffuse) sources and
delivery of TN and TP loadings estimated in the Chesapeake Bay watershed?
12
? What specific types of LC/LUs lost to future urbanization will have the greatest
impacts on projected TN and TP loadings estimated to the Chesapeake Bay?
? What changes in TN and TP loadings will occur in the Chesapeake Bay between
2000 and the future (2030) under the urbanization scenario applied here?
These priority management questions are addressed in this dissertation by
combining satellite-based, remotely sensed LC/LU data with a watershed-wide,
nutrient loading simulation approach. Studies in this dissertation concentrate on how
current and forecasted changes in landscape composition and configuration can be
used to quantify their comprehensive effects on nutrient loadings to the Chesapeake
Bay. This region is chosen for this dissertation due to its varied LC/LU types, the
intensity of anticipated changes, the availability of key data, and its potential for
application to reducing eutrophication that is now and has been the top priority for the
management, policy-making and restorative efforts of the estuary for nearly three
decades (Malone et al., 1993; Boesch et al., 2001).
1.6 Objectives
The specific objectives of this dissertation were to:
1. Quantify the generation and transport of nutrient loadings to the
Chesapeake Bay using the current (2000), spatially-explicit (30 m
resolution) distribution of urban and non-urban cover types,
2. Characterize changes in the generation and transport of nutrient loadings
to the Chesapeake Bay using the future (2030), spatially-explicit (30 m
resolution) distribution of urban and non-urban cover types, and
13
3. Compare and evaluate the characteristics of sprawling catchments
projected to have losses with those projected to have gains in TN by 2030
to assess potential changes in LC/LU that may lead to new management
actions in the Chesapeake Bay further reducing eutrophication.
1.7 The dissertation and its organization
In this chapter (Chapter 1), a brief overview of the nutrient background of the
Chesapeake Bay is presented, in regards to how current and forecasted urbanization and
resulting LC/LU changes could affect TN and TP loadings. The implications of how
forecasted changes in urbanization could potentially impact these biogeochemical cycles
regionally and how evaluating the comprehensive effects of landscape composition and
configuration could help the overall quantification of TN and TP loadings to the estuary
are also briefly described. The information here provides the framework for the research
presented in this dissertation.
Chapter 2 describes a new methodology to quantify the comprehensive effects of
current (2000) urban and non-urban LC/LU on nutrient loadings to the entire Chesapeake
Bay. This is accomplished using the RESAC watershed-wide maps of LC/LU, % ISA,
and % TC (Goetz et al., 2003, 2004a, b; Jantz et al., 2005) integrated within the
Chesapeake Bay Version 3.0 TN and TP SPAtially Referenced Regressions On
Watershed Attributes (SPARROW) models (Brakebill and Preston, 2004). The discovery
and evaluation of newly, significant (p value ? 0.05), compositional and configurational
landscape metrics representing non-point source and land-to-water delivery variables to
the Chesapeake Bay are presented in new RESAC SPARROW models. Additionally, the
14
spatially-explicit distribution of these landscape metrics and the estimates of current TN
and TP loadings to the Chesapeake Bay are also presented. Finally, comparisons of the
new RESAC SPARROW models with the original Chesapeake Bay Version 3.0
SPARROW models and the Phase 4.3 Hydrologic Simulation Program-FORTRAN
(HSPF) models are presented in regards to loadings to the Chesapeake Bay, loadings
within the six largest basins draining the estuary, and the accuracy and precision of
estimates.
Chapter 3 discusses a new approach to determining the forecasted effects of
future (2030) urbanization on nutrient loadings to the Chesapeake Bay using the new
RESAC SPARROW models developed in Chapter 2. The spatially-explicit, future
changes in watershed-wide urbanization are derived from the SLEUTH forecasts and the
RESAC LC/LU map and then integrated within the SPARROW models using the
significant landscape metrics. Other LC/LU-correlated, significant variables in the
SPARROW models are also projected into the future in an effort to more accurately
portray estimated TN and TP loadings. The changes in TN and TP loadings from current
to projected estimates for the entire Chesapeake Bay and the six largest basins are
discussed. In addition, the changes to the significant landscape and non-landscape metric
variables leading to these results are also evaluated and presented.
Chapter 4 documents the application of the TN results concluded in Chapter 3
to catchments only forecasting an increase in low-intensity development (urban sprawl),
but with differing responses of either a gain or loss in these projected loadings. The
sprawling catchments are divided into two groups according to their projected TN
loading response and compared for significant statistical differences. The statistical
15
comparisons are based upon the changes or quantities of all significant variables found in
the RESAC SPARROW model from the current to projected simulation. The statistical
comparisons are evaluated and all landscape and non-landscape metric variables
indicated with significant statistical differences are presented. The spatial arrangement of
urban sprawl, in regards to the replacement of non-urban LC/LUs that would lead to
gains or losses in projected TN loadings is also discussed.
Finally, Chapter 5 presents the conclusions and the management and policy
recommendations developed from the results presented in Chapters 2-4. This
dissertation commences with a discussion of directions for future research.
16
Chapter 2: Effects of urban and non-urban land cover on nitrogen
and phosphorus runoff to Chesapeake Bay1
2.1 Abstract
The aim of this study was to determine the effects of catchment and riparian
stream buffer-wide urban and non-urban land cover and land use (LC/LU) on total
nitrogen (TN) and total phosphorus (TP) runoff to the Chesapeake Bay. The effects of
the composition and configuration of LC/LU patches were explored in particular. A
hybrid statistical-process model, the SPAtially Referenced Regression On Watershed
Attributes (SPARROW), was calibrated with year 1997 watershed-wide, average annual
TN and TP discharges to Chesapeake Bay. Two variables were predicted: 1) yield per
unit watershed area and 2) mass delivered to the estuary. The 166,534 km2 watershed
was divided into 2,339 catchments averaging 71 km2. LC/LU was described using
sixteen classes applied to both the catchments and also to riparian stream buffers alone.
Seven distinct landscape metrics were evaluated. In all, 167 (TN) and 168 (TP) LC/LU
class metric combinations were tested in each model calibration run. Runs were made
with LC/LU in six fixed riparian buffer widths (31, 62, 125, 250, 500, and 1,000 meters
(m)) and entire catchments. The significance of the non-point source type (land cover,
manure and fertilizer application, and atmospheric deposition) and factors affecting land-
1 The material in Chapter 2 was previously published. Roberts, A. D., and Prince, S. D., 2010. Effects of
urban and non-urban land cover on nitrogen and phosphorus runoff to Chesapeake Bay. Ecological
Indicators 10(2): 459-474.
17
to-water delivery (physiographic province and natural or artificial land surfaces) was
assessed. The model with a 31 m riparian stream buffer width accounted for the highest
variance of mean annual TN (r2 =0.9366) and TP (r2 =0.7503) yield (mass for a specified
time normalized by drainage area). TN and TP loadings (mass for a specified time)
entering the Chesapeake Bay were estimated to be 1.449 x 108 and 5.367 x 106 kg/yr,
respectively. Five of the 167 TN and three of the 168 TP landscape metrics were shown
to be significant (p value ? 0.05) either for non-point sources or land-to-water delivery
variables. This is the first demonstration of the significance of riparian LC/LU and
landscape metrics on water quality simulation in a watershed as large as the Chesapeake
Bay. Land cover metrics can therefore be expected to improve the accuracy and
precision of estimated TN and TP annual loadings to the Chesapeake Bay and may also
suggest changes in land management that may be beneficial in control of nutrient runoff
to the Chesapeake Bay and similar watersheds elsewhere.
2.2 Introduction
Land cover and land use (LC/LU) and its changes have large effects on water
quality of streams, rivers, lakes, and estuaries. Urbanization is a pervasive form of
LC/LU alteration that is rapidly growing (Paul and Meyer, 2001). This involves
conversion of croplands, forests, grasslands, pastures, wetlands, and other cover types to
residential and transportation and also commercial and industrial uses, thereby increasing
the areas of impervious surfaces (Tsegaye et al., 2006). Impervious surfaces are
quantifiable indicators that correlate very closely with increases in non-point (diffuse)
sources of polluted runoff which degrades the quality of aquatic resources (Arnold and
18
Gibbons, 1996). When combined with other anthropogenic and natural processes,
landscape variables affect non-point nitrogen (N) and phosphorus (P) transport from land
to the receiving water bodies and can contribute to eutrophication (nutrient enrichment)
leading to poor water quality. Thus, LC/LU are critical properties that affect waterway
pollution.
One such region where LC/LU changes are said to have affected regional water
quality is the Chesapeake Bay watershed (Figure 2.1a). The Chesapeake Bay is the
largest estuary in the United States (US); its watershed (166,534 km2) encompasses
portions of six states: New York (NY), Pennsylvania (PA), Delaware (DE), Maryland
(MD), West Virginia (WV), and Virginia (VA)) and the District of Columbia (DC).
Once considered a natural treasure due to its rich wildlife habitat and seafood industries,
the estuary has been in steady decline starting with the colonial landscape transformations
in the mid 1600s. By the 1990s, the human population of the watershed was
approximately 16 million (McConnell 1995), concentrated in fast-growing urban
corridors (Figure 2.1b). With the increase in population and the built environment,
LC/LU modifications within the watershed have contributed to sedimentation, turbidity,
eutrophication, and hypoxia, consequently reducing submerged aquatic vegetation (SAV)
and affecting many other aspects of the aquatic ecosystem (Breitburg 1992; Hassett et al.,
2005). As stricter regulations involving point source discharges, fertilizer and manure
applications, and fossil fuel emissions that lead to atmospheric deposition are enacted,
spatial information on how landscape properties affect regional nutrient runoff is needed
to meet reduction goals.
19
PA
VA
NY
WV
MD NJ
NC
OH
DE
DC
JAM E
S R
SUS
QUEHAN N A R
PO
TO
MA
C R
RA
PP
AH
ANN
OCK
R
PA
TU
XE
NT
R
MA
TTA
PO
NI RJA
ME
S R
SU
SQ
UE
HA
NN
A R
PAT
UX
EN
T R
PAM
UNK
EY
R
a)
?
0 40 80 120 16020
Kilometers
Baltimore, MD
Washington, DC--MD--VA
Richmond, VA
Harrisburg, PA
Norfolk--Virginia Beach--Newport Ne
Scranton--Wilkes-Barre, PA
York, PALancaster, PA
Lynchburg, VA
Binghamton, NY
Petersburg, VA
Cumberland, MD--WV
Elmira, NY
Altoona, PA
Frederick, MD
Fredericksburg, VA
Annapolis, MD
Hagerstown, MD--PA--WV
Charlottesville, VA
Williamsport, PA
State College, PAPA
VA
NY
WV
NJ
NC
OH
DEMD
b)
?
0 40 80 120 16020
Kilometers
Figure 2.1: The Chesapeake Bay watershed showing the locations of: a) Streams and
rivers draining the estuary and b) Urban centers located within its boundaries.
Remotely sensed LC/LU data have recently become available for landscape
analysis of the Chesapeake Bay watershed, including the Regional Earth Science
Application Center?s (RESAC) LC/LU, percent impervious surface area (% ISA), and
percent tree cover (% TC) maps for 2000 (Goetz et al., 2003, 2004a, b; Jantz et al.,
2005). Other remotely sensed datasets include the National Land Cover Dataset (NLCD)
developed by the Multi-Resolution Land Characteristics (MRLC) (Homer et al., 2004)
consortium for 1992 and 2001 which cover the entire conterminous United States. In
particular, watershed-wide % ISA may yield important new information in linking the
effects of urbanized areas to the estuary's water quality to compliment previous findings,
within the watershed's smaller tributaries, correlating ISA to changes in stream hydrology
(Carlson and Arthur, 2000; Jennings and Jarnagin, 2002; Dougherty et al., 2004).
20
Water quality models, such as the Hydrologic Simulation Program-FORTRAN
(HSPF) (Bicknell et al., 1996, 2001) and the SPAtially Referenced Regressions On
Watershed Attributes (SPARROW) (Smith et al., 1997), have been applied to the
Chesapeake Bay watershed (Preston and Brakebill, 1999; Linker et al., 2000; Brakebill et
al., 2001; Brakebill and Preston, 2004; Goetz et al., 2004a). The HSPF model has often
been used by management and regulatory entities, such as the Chesapeake Bay Program
(CBP) and the United States Environmental Protection Agency (USEPA). HSPF is a
process-based deterministic model that simulates nutrient loadings (mass) to the tidal
tributaries (Linker et al., 2000). Complex process models of this type, however, require
extensive temporally variable data (such as hourly rainfall, temperature, wind, and
evapotranspiration) and detailed calibration. These requirements, generally limit
application to a few watersheds (Alexander et al., 2002a).
The SPARROW model developed by the United States Geological Survey
(USGS) has been used to estimate stream export and improve interpretability of model
parameters (Alexander et al., 2002a). SPARROW utilizes a hybrid-statistical process
structure that implements deterministic functions with spatially-distributed components,
thus accounting for the dendritic features of watersheds (Alexander et al., 2002b). The
model addresses many shortcomings of purely statistical or regression-based models by
incorporating deterministic components of nutrient transport that includes flow paths,
first-order loss functions, and mass-balance constraints (Alexander et al., 2002b).
Furthermore, unlike HSPF, SPARROW provides robust measures of uncertainty.
SPARROW was designed to reduce problems associated with data interpretation caused
by sparse stream sampling measurement networks, network sampling biases, and basin
21
heterogeneity. SPARROW has been applied at national (Smith et al., 1997; Smith et al.,
2003; Alexander et al., 2004), regional (Alexander et al., 2000; Moore et al., 2004), and
even localized watershed scales (Alexander et al., 2002a, b; McMahon et al., 2003).
Previous versions of SPARROW for the Chesapeake Bay watershed (Preston and
Brakebill, 1999; Brakebill et al., 2001; Brakebill and Preston, 2004) have found no
correlation between LC/LU and land-to-water delivery of non-point N and P. In these
versions, however, LC/LU types were represented only as total areas of each cover type
and did not take into account their configuration, or the possibility of differences in the
role of LC/LU types in the riparian zones alone. Developed urban land cover in riparian
zones have been shown to increase non-point N losses to streams (Groffman et al., 2002),
unlike non-developed (forested) zones (Goetz et al., 2003).
Previous research has found that landscape metrics applied in catchment nutrient
export models can improve nutrient predictions over those that use total LC/LU areal
extent (Carle et al., 2005). Landscape metrics describe the spatial structure of patches,
the cover classes of patches, and patch mosaics, and provide other measures of
composition and configuration (Leitao et al., 2006). Landscape composition is the
variety and abundance of patch types without regard to their spatial character or
arrangement, whereas configuration quantifies spatial character and arrangement,
position, and orientation of landscape elements. Although landscape metrics have been
shown to improve correlations between the land surface and nutrient loading dynamics at
catchment and riparian scales in subwatersheds of the Chesapeake Bay (Jones et al.,
2001; King et al., 2005; Baker et al., 2006), there has not been a comprehensive analysis
of the entire watershed.
22
Thus, the overall purpose of this study was to use the SPARROW model with
improved LC/LU maps from remotely-sensed data to determine the effects of LC/LU on
TN and TP runoff from entire catchments and from the riparian stream buffer alone.
Furthermore, the effects of landscape composition and configuration on runoff were
explored.
2.3 Materials and methods
2.3.1 Chesapeake Bay watershed land cover and land use data
Three RESAC maps of LC/LU of the entire watershed were used, all at 30m
resolution. The first map had 18 distinct LC/LU types (Goetz et al., 2004a, b; Jantz et
al., 2005); the second depicted % ISA (Goetz et al., 2004a, b; Jantz et al., 2005); the
third was of % TC (Goetz et al., 2003, 2004a).
The watershed-wide % ISA map was partitioned into urban (? 10% ISA, class 1)
and non-urban (< 10% ISA, class 2). Non-urban areas were coded as the appropriate
LC/LU class. Twelve LC/LU classes were used: urban/residential/recreational grasses;
extractive land; barren land; deciduous forest; evergreen forest; mixed (deciduous-
evergreen) forest; pasture/hay; cropland; natural grass; deciduous wooded wetland;
evergreen wooded wetland; and emergent (sedge-herb) wetland. The four built classes
(low, medium, and high intensity developed, and transportation) were included in the
urban grouping. The two remaining non-urban RESAC LC/LU classes: open water (not
applicable) and mixed wetland (negligible areal coverage) were omitted. The % TC map
was partitioned into non-forested (? 50% TC) and forested (> 50% TC) classes. Overall,
a total of sixteen classes were created (Table 2.1).
23
2000 land cover class Class number
Urban (? 10% ISA) 1
Non-urban (< 10% ISA) 2
Urban/residential/recreational grasses 3
Extractive 4
Barren 5
Deciduous forest 6
Evergreen forest 7
Mixed (deciduous-evergreen) forest 8
Pasture/hay 9
Croplands 10
Natural grass 11
Deciduous wooded wetland 12
Evergreen wooded wetland 13
Emergent (sedge-herb) wetland 14
Non-forested (? 50% TC) 15
Forested (> 50% TC) 16
Table 2.1: The sixteen Chesapeake Bay watershed cover classes created from the 2000
RESAC LC/LU, % ISA, and % TC maps.
2.3.2 Landscape metrics
Seven landscape metrics, previously shown to be significant indicators of
downstream water impairment in catchments within the northeastern United States
(Leitao et al., 2006), were used. These seven landscape metrics were calculated and
applied to the sixteen LC/LU classes. Five metrics (1-3, 6, 7) measure landscape
configuration and two (4, 5) composition. The complete definitions for all seven metrics
are given in Leitao et al. (2006).
1) Contagion quantifies the degree to which LC/LU types were clumped as
opposed to dispersed in many smaller fragments
2) Area-weighted mean radius of gyration (distance) measures connectivity by
correlation length. Correlation length is the average distance one might traverse across a
map from a random starting point and moving in a random direction while remaining in
the same LC/LU (Keitt et al., 1997). Larger values of area-weighted mean radius of
gyration indicate more connected landscapes.
24
3) Patch number measures total LC/LU fragmentation by the total number of
patches of a particular LC/LU.
4) Percentage of the landscape area composed of a specified LC/LU.
5) Area-weighted mean patch size quantifies the sum, across all patches of a
particular LC/LU, of patch area multiplied by proportional abundance of the patch. Since
this metric weights each patch by its size relative to the total area of that particular
LC/LU, larger patches exert greater influence than smaller patches, reducing the effects
of extremely small patches.
6) Area-weighted mean edge contrast (percentage) quantifies the amount of
contrast between adjacent LC/LU patches. In this application, contrast is defined as
physical characteristics of differing cover types influencing nutrient transport. The
metric quantifies the functional edge based on predetermined contrast weights assigned to
pairwise comparisons of LC/LU types giving greater influence to larger patches.
7) Area-weighted mean Euclidean nearest neighbor distance quantifies the
shortest distance from one patch to the next patch of the same LC/LU type, weighted in
favor of larger patch sizes.
2.3.3 The SPARROW model
SPARROW estimates TN and TP loadings (mass for a specified time) and yields
(mass for a specified time normalized for drainage area) from spatially-referenced
watershed networks using source, land-to-water delivery, and stream and reservoir decay
variables. The model is most frequently parameterized using average watershed-wide
conditions for one specific year that represents a "snapshot" in time and is therefore not
event-based (unlike HSPF). However, this allows for SPARROW to predict loadings and
25
yields at a substantially higher number of points throughout these networks than HSPF is
capable of by using linked nested stream reaches and their contributing catchment areas
that are much smaller in magnitude. Catchments surrounding these nested reaches are
denoted as J(i) and are the set of all reaches upstream that include reach i, except for
those containing, or upstream of, monitoring stations of reach i (Figure 2.2) (Alexander et
al., 2002b). Unlike many other source-transport watershed models, SPARROW can
simulate large watersheds using land-to-water delivery variables and simplified, yet
process-based, descriptions of the sources (Schwarz et al., 2006).
Figure 2.2: Illustration of a set of nested stream reaches and reservoir shorelines in
relation to monitoring stations (Alexander et al., 2002b). In model calibration, reach i
refers to any reach containing a monitoring station. In model application, reach i refers to
any reach for which a prediction can be made.
Both point and non-point source variables resulting in river discharges into the
receiving water body (in this case the Chesapeake Bay) are included. Land-to-water
delivery variables describe properties of the landscape relating climatic and other natural
and human-induced surface processes affecting non-point N and P transport to streams.
Stream decay is described by first-order losses of TN and TP loadings along stream
26
channels, whereas reservoir decay is described by attenuation factors that influence TN
and TP losses through large lakes and reservoirs. SPARROW uses nonlinear least
squares regression to determine which variables are significant (p value ? 0.05) at the
(Chesapeake Bay) watershed stations.
To assess model robustness, SPARROW utilizes a sensitivity analysis with Monte
Carlo data resampling bootstrapping methods. The analysis calculates whether any
coefficient has the wrong sign (Smith et al., 1997) using the mean coefficient value
(bootstrap estimate), confidence interval, and probability. Coefficients for significant
variables are generated in 200 separate model runs by resampling (with replacement)
from the original data.
The structure of Version 3.0 Chesapeake Bay model (Brakebill and Preston,
2004), referred to as B & P, was used so that the overall topology of the stream network
and B & P non-LC/LU variables, such as point sources, could be used. Variables not
related to landscape properties were the most current (1997) watershed-wide estimated
datasets available. In addition, this version most closely approximated the year (2000) of
the RESAC landscape maps. The difference in this present application from B & P was
that landscape metrics were added by replacing catchment-wide non-point nutrient land
sources and creating new catchment and riparian stream buffer-wide land-to-water
delivery variables in order to determine the total significance of LC/LU composition and
configuration. Riparian stream buffers are defined here as fixed, transitional areas
between terrestrial landscapes and stream reaches created from the linked, spatially-
referenced watershed network. In Version 3.0, 2,339 separate catchments averaging 71
27
km2 were modeled (Figure 2.3a) using the enhanced river reach file (E3RF1) (Brakebill
and Preston, 2004).
The TN and TP models were calibrated with observations at 87 and 104 stream
loading sites, respectively, collected in 1997 by federal and state agencies. Stream
nutrient loadings were calculated from the data using a log-linear regression model know
as ESTIMATOR (Cohn et al., 1989), which uses the 1950-2000 averages of daily stream
discharge, specifying 1997 as the trend component (Brakebill and Preston, 2004).
PA
VA
NY
WV
MD NJ
NC
OH
DE
DC
?
0 40 80 120 16020
Kilometers
a)
Example Catchment
Runoff from Catchment
Enters Stream
1000 m Buffer
500 m Buffer
250 m Buffer
125 m Buffer
62 m Buffer
31 m Buffer
Upstream Catchments
Carrying N and P Loads
Upstream Plus Locally Generated
Runoff, Less Instream Losses
Leaves Catchment
b)
?
0 1
Kilometer
i
i
i
ii
i
Figure 2.3: Map of: a) 2,339 catchments used and b) Schematic diagram of an example
catchment area showing fixed 31, 62, 125, 250, 500, and 1,000 meters (m) riparian
stream buffer width areas surrounding stream reach and corresponding upstream and
downstream runoff paths.
2.3.4 Model calibration
Using the 2,339 modeled catchments and their associated stream reaches, fixed
riparian stream buffers of 31, 62, 125, 250, 500, and 1,000 m (Figure 2.3b) were specified
28
for a total of 4,678 land areas per model. A model run consisted of analyzing 2000 land
cover classes within each of the 2,339 catchments with the only differences being which
land cover area within one of the six fixed riparian stream buffer areas was also analyzed.
Using FRAGSTATS (McGarigal et al., 2002), five of the seven metrics (area-weighted
mean radius of gyration, patch number, percentage of landscape, area-weighted mean
edge contrast, and area-weighted mean Euclidean nearest neighbor distance) were created
from the sixteen cover classes at both catchment and riparian stream buffer width scales,
adding 160 new variables per run evaluated for land-to-water delivery significance.
The sixth metric (contagion) was calculated for land cover classes in 2000 at the
catchment and the six riparian stream buffer widths. The values were calculated per map,
not using LC/LU data cross-referenced between the other two maps. Contagion was
calculated at the landscape map level between the cover classes of: 1) urban (class 1) and
non-urban (class 2) in the % ISA map, 2) the twelve remaining non-urban classes (classes
3-14) in the LC/LU map, and 3) non-forested (class 15) and forested (class 16) from the
% TC map at both catchment and riparian stream buffer width scales to add six more
variables. By adding in the six contagion variables, a total of 166 new variables were
evaluated for each TN and TP model run.
Finally, the 1997 B & P model urban (TN and TP models) and non-
agricultural/non-urban (TP model only) areas were replaced by the area-weighted mean
patch size for 2000 urban and non-agricultural/non-urban land. Non-agricultural/non-
urban land area in the B & P TP model was modified here for new TP models by
formation of area-weighted mean non-agricultural/non-urban patch sizes obtained by
subtracting area-weighted mean cropland patch sizes (class 10) from area-weighted mean
29
non-urban patch sizes (class 2). By adding catchment-wide area-weighted mean urban (?
10% ISA) patch sizes (TN and TP models), along with area-weighted mean non-
agricultural/non-urban patch sizes (TP model only) as non-point source regions, the final
number of variables increased to 167 and 168 for each TN and TP model run,
respectively.
The area-weighted mean patch size metric for the sixteen landscape classes was
not evaluated for land-to-water delivery significance since it could not be normalized for
catchment and riparian stream buffer width area. Any type of land-to-water delivery
variable directly correlated with catchment and riparian stream buffer width size would
tend to confuse the relationship between scale-dependent, non-point source variables and
non-point N and P stream loadings.
2.4 Results
2.4.1 TN mode1
Of the six models that used riparian stream buffers, the 31 m buffer out-performed
all other simulations (yield coefficient of determination (r2) 0.9366; root mean squared
error (RMSE) 0.2407%) (Table 2.2). This was also indicated by the smallest Akaike
Information Criteria (AIC) value (-2.7020) (Table 2.2) that indicated a better model and
evidence for its selection over the other simulations (Luke, 2004). The 31 m model
explained nearly 94% of variations in the mean annual TN yield resulting in a RMSE of
24%. The r2 between the natural log of the observed and predicted loadings at the 87
Chesapeake Bay TN monitoring stations was 0.9896 (Figure 2.4a). Buffers of 62-1,000
m had lower yield r2 and higher RMSE values, but the same significant coefficients.
30
Model run Model yield r2 Model RMSE (%) Model AIC
1997 B & P TN 0.9073 0.2834 -2.4140
2000 RESAC 31 m TN 0.9366 0.2407 -2.7020
2000 RESAC 62 m TN 0.9332 0.2454 -2.6727
2000 RESAC 125 m TN 0.9332 0.2454 -2.6727
2000 RESAC 250 m TN 0.9332 0.2454 -2.6727
2000 RESAC 500 m TN 0.9332 0.2454 -2.6727
2000 RESAC 1000 m TN 0.9332 0.2454 -2.6727
1997 B & P TP 0.7413 0.3257 -2.1610
2000 RESAC 31 m TP 0.7503 0.3126 -2.2345
2000 RESAC 62 m TP 0.7457 0.3246 -2.1591
2000 RESAC 125 m TP 0.7353 0.3329 -2.1086
2000 RESAC 250 m TP 0.7262 0.3368 -2.0853
2000 RESAC 500 m TP 0.7220 0.3393 -2.0705
2000 RESAC 1000 m TP 0.7209 0.3400 -2.0664
Table 2.2: Comparison of r2, RMSE, and AIC values between the Brakebill and Preston
(2004) (1997 B & P) and the 2000 RESAC 31-1,000 m TN and TP models.
a)  b)
Total Nitrogen (TN)
y = 0.9837x + 0.2106
R2 = 0.9896
4
6
8
10
12
14
16
18
20
4 6 8 10 12 14 16 18 20
LN of 1997 Observed Load (kg/yr)
LN
of
20
00
P
re
dic
ted
Lo
ad
(k
g/y
r)
Total Phosphorus (TP)
y = 0.9745x + 0.2509
R2 = 0.9733
4
6
8
10
12
14
16
18
20
4 6 8 10 12 14 16 18 20
LN of 1997 Observed Load (kg/yr)
LN
of
20
00
P
re
dic
ted
Lo
ad
(k
g/y
r)
Figure 2.4: Comparison between the natural log of observed versus predicted loadings
(kilograms per year (kg/yr)) in the: a) 2000 RESAC 31 m TN model utilizing 87
Chesapeake Bay water quality N monitoring stations and b) 2000 RESAC 31 m TP
model utilizing 104 Chesapeake Bay water quality P monitoring stations.
Of the 167 landscape variables tested for either source or land-to-water delivery
potential in the 31 m TN model, only five were significantly related to non-point N
sources or delivery to streams. The five metrics were: area-weighted mean urban (? 10%
ISA) patch size (source), percentage of extractive land (land-to-water delivery), area-
weighted mean edge contrast of deciduous forest (land-to-water delivery), percentage of
cropland (land-to-water delivery), and percentage of evergreen forest within the riparian
31
stream buffer (land-to-water delivery) (Table 2.3). The only difference between the 31 m
model and the other models was the highly significant (p value = 4.87 x 10-2) effect of
percentage of evergreen forest at this buffer width (Table 2.4). The percentage of
evergreen forest in models of buffer widths of 62-1000 m was omitted as a land-to-water
delivery variable (p-value > 0.05). Sensitivity analysis results comparing initial
parametric with a final set of averaged bootstrapped coefficient estimates showed good
agreement between the two for all significant variables (Table 2.5).
2000 RESAC 31 m model Significant (p value ? 0.05) landscape metric variables
Area-weighted mean urban (? 10% ISA) patch size
(source)
Percentage of cropland
(land-to-water delivery)
TN Percentage of extractive land
(land-to-water delivery)
Area-weighted mean edge contrast of deciduous forest
(land-to-water delivery)
Percentage of evergreen forest within the riparian stream buffer
(land-to-water delivery)
Area-weighted mean non-agricultural/non-urban patch size
(source)
TP Percentage of barren land within the riparian stream buffer
(land-to-water delivery)
Area-weighted mean urban (? 10% ISA) patch size
(source)
Table 2.3: Comparison of the significant (p value ? 0.05) landscape metrics in the 2000
RESAC 31 m TN and TP models.
32
Variable
(units)
Model run
(Year)
B & P
(1997)
RESAC
(2000)
31 m 62 m 125 m 250 m 500 m 1000 m
Yield r2 0.9073 0.9366 0.9332 0.9332 0.9332 0.9332 0.9332
RMSE (%) 0.2834 0.2406 0.2454 0.2454 0.2454 0.2454 0.2454
AIC -2.4140 -2.7020 -2.6727 -2.6727 -2.6727 -2.6727 -2.6727
Model
category
Point sources
(kg/yr)
Nitrogen
source
1.530
(1.1 x 10-4)
1.173
(1.2 x 10-4)
1.169
(1.6 x 10-4)
1.169
(1.6 x 10-4)
1.169
(1.6 x 10-4)
1.169
(1.6 x 10-4)
1.169
(1.6 x 10-4)
Applied fertilizer
(kg/yr)
Nitrogen
source
0.294
(6.8 x 10-13)
0.175
(3.9 x 10-6)
0.194
(1.0 x 10-6)
0.194
(1.0 x 10-6)
0.194
(1.0 x 10-6)
0.194
(1.0 x 10-6)
0.194
(1.0 x 10-6)
Atmospheric
deposition
(kg/yr)
Nitrogen
source
0.215
(3.5 x 10-7)
0.492
(2.0 x 10-7)
0.476
(4.0 x 10-7)
0.476
(4.0 x 10-7)
0.476
(4.0 x 10-7)
0.476
(4.0 x 10-7)
0.476
(4.0 x 10-7)
Applied manure
(kg/yr)
Nitrogen
source
0.065
(7.3 x 10-3)
0.078
(7.7 x 10-4)
0.087
(5.1 x 10-4)
0.087
(5.1 x 10-4)
0.087
(5.1 x 10-4)
0.087
(5.1 x 10-4)
0.087
(5.1 x 10-4)
Urban land
(kg/ha/yr)
Nitrogen
source
9.157
(8.7 x 10-6)
ONU ONU ONU ONU ONU ONU
* Area-weighted mean
urban (? 10% ISA)
patch size
(kg/ha/yr)
Nitrogen
source
MNU 24.885
(7.2 x 10-7)
23.200
(1.2 x 10-6)
23.200
(1.2 x 10-6)
23.200
(1.2 x 10-6)
23.200
(1.2 x 10-6)
23.200
(1.2 x 10-6)
Percentage of Coastal
Plain
(%)
Landscape
delivery
-0.735
(1.6 x 10-7)
-0.729
(4.1 x 10-8)
-0.679
(1.9 x 10-7)
-0.679
(1.9 x 10-7)
-0.679
(1.9 x 10-7)
-0.679
(1.9 x 10-7)
-0.679
(1.9 x 10-7)
* Percentage of
extractive land
(%)
Landscape
delivery
MNU 0.270
(7.4 x 10-4)
0.283
(6.5 x 10-4)
0.283
(6.5 x 10-4)
0.283
(6.5 x 10-4)
0.283
(6.5 x 10-4)
0.283
(6.5 x 10-4)
* Area-weighted mean
edge contrast of
deciduous forest
(%)
Landscape
delivery
MNU 0.014
(7.2 x 10-3)
0.011
(3.1 x 10-2)
0.011
(3.1 x 10-2)
0.011
(3.1 x 10-2)
0.011
(3.1 x 10-2)
0.011
(3.1 x 10-2)
* Percentage of
cropland
(%)
Landscape
delivery
MNU 0.021
(1.1 x 10-4)
0.020
(2.0 x 10-4)
0.020
(2.0 x 10-4)
0.020
(2.0 x 10-4)
0.020
(2.0 x 10-4)
0.020
(2.0 x 10-4)
* Percentage of
evergreen forest
within the riparian
stream buffer
(%)
Landscape
delivery
MNU 0.013
(4.7 x 10-2)
MIS MIS MIS MIS MIS
Small
streams
(m/day)
Stream
decay
0.375
(1.8 x 10-2)
0.249
(5.5 x 10-2)
0.299
(2.4 x 10-2)
0.299
(2.4 x 10-2)
0.299
(2.4 x 10-2)
0.299
(2.4 x 10-2)
0.299
(2.4 x 10-2)
Intermediate
streams
(m/day)
Stream
decay
0.135
(2.2 x 10-1)
0.090
(3.2 x 10-1)
0.088
(3.4 x 10-1)
0.088
(3.4 x 10-1)
0.088
(3.4 x 10-1)
0.088
(3.4 x 10-1)
0.088
(3.4 x 10-1)
Large
streams
(m/day)
Stream
decay
0.031
(5.4 x 10-1)
0.030
(4.8 x 10-1)
0.028
(5.1 x 10-1)
0.028
(5.1 x 10-1)
0.028
(5.1 x 10-1)
0.028
(5.1 x 10-1)
0.028
(5.1 x 10-1)
Reservoir
(m/yr)
Reservoir
decay
19.036
(2.8 x 10-2)
14.224
(2.3 x 10-2)
14.466
(2.3 x 10-2)
14.466
(2.3 x 10-2)
14.466
(2.3 x 10-2)
14.466
(2.3 x 10-2)
14.466
(2.3 x 10-2)
Table 2.4: All significant (p value ? 0.05) variables in the 1997 B & P and 2000 RESAC
TN models using 31-1,000 m riparian stream buffer widths with model coefficients and p
values (in parentheses). * Denotes all significant landscape metrics. ONU = original
land source area "not utilized" as a result of being replaced with new surrogate landscape
source area metric in 2000 model runs. MNU = new landscape source area and land-to-
water delivery metrics "not utilized" in 1997 model run. MIS = new landscape land-to-
water metric found to be "insignificant" (p value > 0.05) for that 2000 model run.
33
Variables
(units)
Model run
(Year)
RESAC
(2000)
31 M
initial
parametric
31 M
averaged
bootstrapped
Model
category
Point sources
(kg/yr)
Nitrogen
source
1.173 1.189
Applied fertilizer
(kg/yr)
Nitrogen
source
0.175 0.170
Atmospheric
deposition
(kg/yr)
Nitrogen
source
0.492 0.508
Applied manure
(kg/yr)
Nitrogen
source
0.078 0.080
* Area-weighted
mean urban (? 10%
ISA) patch size
(kg/ha/yr)
Nitrogen
source
24.885 28.555
Percentage of
coastal plain
(%)
Landscape
delivery
-0.729 -0.757
* Percentage of
extractive land
(%)
Landscape
delivery
0.270 0.280
* Area-weighted
mean edge contrast
of deciduous forest
(%)
Landscape
delivery
0.014 0.015
* Percentage of
cropland
(%)
Landscape
delivery
0.021 0.021
* Percentage of
evergreen forest
within the riparian
stream buffer
(%)
Landscape
delivery
0.013 0.019
Small
streams
(m/day)
Stream
decay
0.249 0.261
Intermediate
streams
(m/day)
Stream
decay
0.090 0.103
Large
streams
(m/day)
Stream
decay
0.030 0.037
Reservoir
(m/yr)
Reservoir
decay
14.224 13.669
Table 2.5: Sensitivity analysis comparison of all significant (p value ? 0.05) estimated
variables between initial parametric and averaged bootstrapped coefficient estimates in
2000 RESAC 31 m TN model. * Denotes all landscape metrics found to be significant.
Of the 2,339 catchments in the 31 m model, the largest TN yields (> 18 kilograms
per hectare per year (kg/ha/yr)) were identified in: the lower Susquehanna Basin
containing cities such as Harrisburg, Lancaster, and York (PA); along tributaries of the
34
middle Potomac Basin in north central MD; the eastern shore of MD; and southwestern
DE (Figure 2.5a). Smaller areas of high locally generated TN yields were also found
near the urban areas of DC and Richmond (VA). Although much attenuation occurred
after accounting for stream and reservoir loss processes, there were only local differences
in the amount of largest TN yields (> 18 kg/ha/yr) delivered to the estuary (Figure 2.5b).
The highest locally generated urban patch size N yields (> 0.99 kg/ha/yr) per catchment
were associated with the major urban centers of: DC; Baltimore (MD); Scranton/Wilkes-
Barre, Harrisburg, Lancaster, and York (PA); Richmond, Petersburg, and Norfolk-
Virginia Beach-Newport News (VA); and Elmira and Binghamton (NY) (Figure 2.6a).
Stream and reservoir decay reduced area-weighted mean urban patch size N yield per
catchment. However, largest yields (> 0.99 kg/ha/yr) to the estuary were still shown to
originate from these areas (Figure 2.6b).
2000 Local Generation of TN Yield (kg/ha/yr)
0 - 2
2 - 4
4 - 6
6 - 8
8 - 10
10 - 12
12 - 14
14 - 16
16 - 18
> 18
a)
?
0 40 80 120 16020
Kilometers
2000 Delivery of TN Yield (kg/ha/yr)
0 - 2
2 - 4
4 - 6
6 - 8
8 - 10
10 - 12
12 - 14
14 - 16
16 - 18
> 18
b)
?
0 40 80 120 16020
Kilometers
Figure 2.5: Per catchment estimated TN yield (kg/ha/yr) map from the 2000 RESAC 31
m model of: a) Local generation and b) Delivery to the Chesapeake Bay.
35
2000 Local Generation of Urban N Yield (kg/ha/yr)
0.00 - 0.11
0.11 - 0.22
0.22 - 0.33
0.33 - 0.44
0.44 - 0.55
0.55 - 0.66
0.66 - 0.77
0.77 - 0.88
0.88 - 0.99
> 0.99
a)
?
0 40 80 120 16020
Kilometers
2000 Delivery of Urban N Yield (kg/ha/yr)
0.00 - 0.11
0.11 - 0.22
0.22 - 0.33
0.33 - 0.44
0.44 - 0.55
0.55 - 0.66
0.66 - 0.77
0.77 - 0.88
0.88 - 0.99
> 0.99
b)
?
0 40 80 120 16020
Kilometers
Figure 2.6: Per catchment estimated area-weighted mean urban (? 10% ISA) patch size
N yield (kg/ha/yr) map from the 2000 RESAC 31 m model of: a) Local generation and b)
Delivery to the Chesapeake Bay.
2.4.2 TP mode1
As in the case of the TN models, restriction of metrics to the 31 m riparian stream
buffer gave the best results. This model also had the highest yield r2 (0.7503) and lowest
RMSE (0.3216) and AIC (-2.234) values (Table 2.2). The 31 m model explained about
75% of variation in the mean annual TP yield, a small improvement over the 62 m model.
The r2 between the natural log of the observed and predicted loadings at the 104 TP
monitoring stations was 0.9733 (Figure 2.4b). Model significance decreased with each of
the wider buffers.
36
Variables
(units)
Model run
(Year)
B & P
(1997)
RESAC
(2000)
31 m 62 m 125 m 250 m 500 m 1000 m
Yield r2 0.7413 0.7503 0.7457 0.7353 0.7262 0.7220 0.7209
RMSE (%) 0.3257 0.3216 0.3246 0.3330 0.3367 0.3394 0.3400
AIC -2.1610 -2.2345 -2.1591 -2.1086 -2.0853 -2.0705 -2.0664
Model
category
Point sources
(kg/yr)
Phosphorus
source
0.673
(4.7 x 10-6)
0.738
(1.3 x 10-6)
0.743
(1.4 x 10-6)
0.710
(5.4 x 10-6)
0.734
(3.4 x 10-6)
0.724
(4.5 x 10-6)
0.725
(4.5 x 10-6)
Applied fertilizer
(kg/yr)
Phosphorus
source
0.016
(3.3 x 10-4)
0.016
(1.1 x 10-3)
0.016
(1.1 x 10-3)
0.019
(3.4 x 10-4)
0.016
(1.9 x 10-3)
0.016
(1.8 x 10-3)
0.016
(1.9 x 10-3)
Applied manure
(kg/yr)
Phosphorus
source
0.007
(4.5 x 10-2)
0.008
(3.0 x 10-2)
0.008
(3.5 x 10-2)
0.009
(2.9 x 10-2)
0.008
(3.6 x 10-2)
0.008
(3.9 x 10-2)
0.008
(3.8 x 10-2)
Non-agricultural/non-
urban land
(kg/ha/yr)
Phosphorus
source
0.093
(9.0 x 10-13)
ONU ONU ONU ONU ONU ONU
* Area-weighted
mean non-
agricultural/non-
urban patch size
(kg/ha/yr)
Phosphorus
source
MNU 0.110
(5.8 x 10-13)
0.110
(1.1 x 10-12)
0.094
(2.5 x 10-11)
0.105
(6.6 x 10-12)
0.104
(7.6 x 10-12)
0.102
(6.6 x 10-12)
Urban land
(kg/ha/yr)
Phosphorus
source
0.442
(7.8 x 10-5)
ONU ONU ONU ONU ONU ONU
* Area-weighted
mean urban (? 10%
ISA) patch size
(kg/ha/yr)
Phosphorus
source
MNU 0.921
(1.0 x 10-4)
0.893
(1.3 x 10-4)
0.975
(1.1 x 10-4)
0.873
(2.3 x 10-4)
0.861
(2.8 x 10-4)
0.867
(2.8 x 10-4)
* Percentage of
barren land within the
riparian stream buffer
(%)
Landscape
delivery
MNU 0.281
(1.2 x 10-6)
0.279
(6.1 x 10-5)
0.213
(1.8 x 10-3)
0.167
(7.0 x 10-3)
0.132
(1.6 x 10-2)
0.114
(2.0 x 10-2)
Small
streams
(m/day)
Stream
decay
-0.260
(4.9 x 10-2)
-0.198
(1.3 x 10-1)
-0.199
(1.4 x 10-1)
-0.210
(1.3 x 10-1)
-0.194
(1.7 x 10-1)
-0.191
(1.9 x 10-1)
-0.184
(2.1 x 10-1)
Intermediate
streams
(m/day)
Stream
decay
0.028
(7.9 x 10-1)
0.150
(1.9 x 10-1)
0.144
(2.1 x 10-1)
0.137
(2.4 x 10-1)
0.095
(4.1 x 10-1)
0.082
(4.8 x 10-1)
0.071
(5.3 x 10-1)
Large
streams
(m/day)
Stream
decay
0.039
(4.6 x 10-1)
0.034
(5.2 x 10-1)
0.035
(5.1 x 10-1)
0.053
(3.5 x 10-1)
0.037
(5.0 x 10-1)
0.037
(5.1 x 10-1)
0.038
(5.0 x 10-1)
Reservoir
(m/yr)
Reservoir
decay
17.004
(5.4 x 10-2)
19.019
(5.7 x 10-2)
18.473
(6.3 x 10-2)
17.669
(7.5 x 10-2)
19.611
(6.1 x 10-2)
19.965
(5.9 x 10-2)
19.700
(5.8 x 10-2)
Table 2.6: All significant (p value ? 0.05) variables in the 1997 B & P and 2000 RESAC
TP models using 31-1,000 m riparian stream buffer widths with model coefficients and p
values (in parentheses). * Denotes all significant landscape metrics. ONU = original
land source areas "not utilized" as a result of being replaced with new surrogate
landscape source area metrics in 2000 model runs. MNU = new landscape source area
and land-to-water delivery metrics "not utilized" in 1997 model run.
Of the 168 landscape metric variables evaluated in the 31 m model, only three
were significantly related to non-point P sources or delivery processes (Table 2.3).
Within the riparian stream buffer, the percentage of barren land was the only significant
37
land-to-water delivery metric (Table 2.6). However, significance of this variable fell as
the width of the riparian stream buffer increased. Comparisons between the initial
parametric and the final set of averaged bootstrapped coefficient estimates indicated little
to no deviation for all significant variables (Table 2.7).
Variables
(units)
Model run
(Year)
RESAC
(2000)
31 M
Initial
Parametric
31 M
Averaged
Bootstrapped
Model
category
Point sources
(kg/yr)
Phosphorus
source
0.738 0.738
Applied fertilizer
(kg/yr)
Phosphorus
source
0.016 0.016
Applied manure
(kg/yr)
Phosphorus
source
0.008 0.008
* Area-weighted
mean non-
agricultural/non-
urban patch size
(kg/ha/yr)
Phosphorus
source
0.110 0.110
* Area-weighted
mean urban (? 10%
ISA) patch size
(kg/ha/yr)
Phosphorus
source
0.921 1.139
* Percentage of
barren land within
the riparian stream
buffer
(%)
Landscape
delivery
0.281 0.277
Small
streams
(m/day)
Stream
decay
-0.198 -0.231
Intermediate
streams
(m/day)
Stream
decay
0.150 0.138
Large
streams
(m/day)
Stream
decay
0.034 0.044
Reservoir
(m/yr)
Reservoir
decay
19.019 20.529
Table 2.7: Sensitivity analysis comparison of all significant (p value ? 0.05) estimated
variables between initial parametric and averaged bootstrapped coefficient estimates in
2000 RESAC 31 m TP model. * Denotes all landscape metrics found to be significant.
38
2000 Local Generation of TP Yield (kg/ha/yr)
0.00 - 0.11
0.11 - 0.22
0.22 - 0.33
0.33 - 0.44
0.44 - 0.55
0.55 - 0.66
0.66 - 0.77
0.77 - 0.88
0.88 - 0.99
> 0.99
a)
?
0 40 80 120 16020
Kilometers
2000 Delivery of TP Yield (kg/ha/yr)
0.00 - 0.11
0.11 - 0.22
0.22 - 0.33
0.33 - 0.44
0.44 - 0.55
0.55 - 0.66
0.66 - 0.77
0.77 - 0.88
0.88 - 0.99
> 0.99
b)
?
0 40 80 120 16020
Kilometers
Figure 2.7: Per catchment estimated TP yield (kg/ha/yr) map from the 2000 RESAC 31
m model of: a) Local generation and b) Delivery to the Chesapeake Bay.
2000 Local Generation of Urban P Yield (kg/ha/yr)
0.00 - 0.05
0.05 - 0.10
0.10 - 0.15
0.15 - 0.20
0.20 - 0.25
0.25 - 0.30
0.30 - 0.35
0.30 - 0.40
0.40 - 0.45
> 0.45
a)
?
0 40 80 120 16020
Kilometers
2000 Delivery of Urban P Yield (kg/ha/yr)
0.00 - 0.05
0.05 - 0.10
0.10 - 0.15
0.15 - 0.20
0.20 - 0.25
0.25 - 0.30
0.30 - 0.35
0.35 - 0.40
0.40 - 0.45
> 0.45
b)
?
0 40 80 120 16020
Kilometers
Figure 2.8: Per catchment estimated area-weighted mean urban (? 10% ISA) patch size
P yield (kg/ha/yr) map from the 2000 RESAC 31 m model of: a) Local generation and b)
Delivery to the Chesapeake Bay.
39
2000 Local Generation of NA/NU P Yield (kg/ha/yr)
0.00 - 0.03
0.03 - 0.06
0.06 - 0.09
0.09 - 0.12
0.12 - 0.15
0.15 - 0.18
0.18 - 0.21
0.21 - 0.24
0.24 - 0.27
> 0.27
a)
?
0 40 80 120 16020
Kilometers
2000 Delivery of NA/NU P Yield (kg/ha/yr)
0.00 - 0.03
0.03 - 0.06
0.06 - 0.09
0.09 - 0.12
0.12 - 0.15
0.15 - 0.18
0.18 - 0.21
0.21 - 0.24
0.24 - 0.27
> 0.27
b)
?
0 40 80 120 16020
Kilometers
Figure 2.9: Per catchment estimated area-weighted mean non-agricultural/non-urban
(NA/NU) patch size P yield (kg/ha/yr) map from the 2000 RESAC 31 m model of: a)
Local generation and b) Delivery to the Chesapeake Bay estuary.
A map of locally generated TP yields showed that highest yields (> 0.99 kg/ha/yr)
were found in: the central and lower Susquehanna Basin near Scranton/Wilkes Barre,
Harrisburg, Lancaster, and York (PA); central MD; and Richmond and Petersburg (VA)
in the middle-to-lower James Basin (Figure 2.7a). After accounting for stream and
reservoir losses, areas with the highest TP yields (> 0.99 kg/ha/yr) delivered to the
estuary were significantly reduced, with the exception of near Lancaster (PA) (Figure
2.7b). A map of locally generated area-weighted mean urban (? 10% ISA) patch size P
yields per catchment showed that higher values (> 0.45 kg/ha/yr) were estimated in:
Harrisburg (PA); DC; Baltimore (MD); and Richmond and Norfolk-Virginia Beach-
Newport News, (VA) (Figure 2. 8a). Highest delivered yields (> 0.45 kg/ha/yr) were also
found in these urban municipalities (Figure 2.8b). Finally, although lower in magnitude,
40
the greatest locally generated area-weighted mean non-agricultural/non-urban patch size
P yields per catchment (> 0.27 kg/ha/yr) were in northern and central VA (Figure 2. 9a).
The largest delivered area-weighted mean non-agricultural/non-urban patch size P yields
(> 0.27 kg/ha/yr) were also from these areas (Figure 2.9b).
2.5 Discussion
2.5.1 Effects of non-urban and urban LC/LU on TN and TP runoff to the
Chesapeake Bay
The natural log of the 2000 area-weighted mean urban (? 10% ISA) patch sizes
derived from the RESAC map was found to be well correlated with the 1997 NLCD
urban, as used by B & P, which suggests that the new parameterization did not change the
fundamental structure of B & P, and so the use here of non-LC/LU elements of B & P
was justified (Figure 2.10a). Numerous studies (Bannerman et al., 1993; Rushton, 2001;
Sonada et al., 2001; Stow et al., 2001; Line et al., 2002; Shinya et al., 2003; Coulter et
al., 2004; Groffman et al., 2004; Law et al., 2004; Osmond and Hardy, 2004; Caccia and
Boyer, 2005; Erickson et al., 2005; Wakida and Lerner, 2005, 2006; Williams et al.,
2005; Gilbert and Clausen, 2006) have found that non-point N and P generated from low,
medium, and high intensity developed, as well as transportation LC/LU classes are linked
to stream loadings. The natural log of the 1997 non-agricultural/non-urban land used by
B & P was also well correlated with the 2000 area-weighted mean non-agricultural/non-
urban patch size derived from RESAC in the 31 m model (Figure 2.10b).
41
a)  b)
y = 0.4886x - 40.178
R2 = 0.7908
-3
4.5
12
-3 4.5 12
LN of 1997 NLCD Urban Land (ha)
LN
of
20
00
M
ea
nU
rba
nP
atc
hS
ize
(h
a)
(
y = 1.0326x + 54.317
R2 = 0.9885
-3
4.5
12
-3 4.5 12
LN of 1997 NLCD NA/NU Land (ha)
LN
of
20
00
M
ea
nN
A/
NU
P
atc
hS
ize
(h
a)
(
Figure 2.10: 2,339 SPARROW catchment comparison of the natural log of the: a)
Significant (p value ? 0.05) 1997 NLCD source factor of total urban land area (ha) used
in the 1997 B & P TN and TP models versus the significant source factor of 2000 area-
weighted mean urban (? 10% ISA) patch sizes (ha) used in the 2000 RESAC 31 m TN
and TP models and b) Significant 1997 NLCD source factor of non-agricultural/non-
urban land area (ha) used in the 1997 B & P TP model versus the significant source factor
of 2000 area-weighted mean non-agricultural/non-urban patch sizes (ha) used in the 2000
RESAC 31 m TP model.
Nearly 25 kg of non-point N in the 31 m TN model, as compared to close to 1 kg
of non-point P in the 31 m TP model, were estimated to be generated and measured from
streams draining the estuary from each ha of area-weighted mean urban patch (? 10%
ISA) size annually (Table 2.4; Table 2.6). The 31 m TN model coefficient value was
well within the expected range of export values for all urban land use yields (3-40
kg/ha/yr) (Schwarz et al., 2006). The largest area-weighted mean urban (? 10% ISA)
patch sizes (>270 ha) were identified in: DC; Baltimore (MD); Scranton/Wilkes-Barre,
Harrisburg, Lancaster, and York (PA); Richmond, Petersburg and Norfolk-Virginia
Beach-Newport News (VA); and Elmira and Binghamton (NY) (Figure 2.11a).
Just over 0.1 kg of non-point P in the 31 m model was estimated to be generated
and measured from streams draining the Chesapeake Bay from each ha of area-weighted
mean non-agricultural/non-urban patch size annually (Table 2.6). The highest area-
42
weighted mean non-agricultural/non-urban patch sizes per catchment (> 27,000 ha) were
located in PA and western VA (Figure 2. 11b).
2000 Mean Urban Patch Size (ha)
0 - 30
30 - 60
60 - 90
90 - 120
120 - 150
150 - 180
180 - 210
210 - 240
240 - 270
> 270
a)
?
0 40 80 120 16020
Kilometers
2000 Mean NA/NU Patch Size (ha)
0 - 3,000
3,000 - 6,000
6,000- 9,000
9,000 - 12,000
12,000 - 15,000
15,000 - 18,000
18,000 - 21,000
21,000 - 24,000
24,000 - 27,000
> 27,000
b)
?
0 40 80 120 16020
Kilometers
Figure 2.11: Per catchment source factor metrics of: a) Area-weighted mean urban (?
10% ISA) patch size (ha) in the 2000 RESAC 31 m TN and TP models and b) Area-
weighted mean non-agricultural/non-urban (NA/NU) patch size (ha) in the 2000 RESAC
31 m TP model.
For every one percent of extractive land composition in a catchment, a 0.27%
increase in delivery from all non-point N sources to streams draining the estuary was
estimated. Characteristics associated with extractive land may help explain this. Low
infiltration capacities tied to decreased hydraulic conductivity (the ability of a porous
medium to transmit a specific fluid under a unit hydraulic gradient) (Ward and Trimble,
2004) of mine soils that increase overland runoff have been found throughout the
watershed's PA tributaries (Ritter and Gardner, 1993; Guebert and Gardner, 2001).
Furthermore N-fixing trees, such as black locust (Robinia pseudoacacia), used in mine
43
spoil reclamation in the Susquehanna Basin (Bruns, 2005) may also increase non-point N
stream delivery. The highest percentages of extractive land per catchment (> 4.5%) were
found in central PA, western MD, and northeastern WV (Figure 2.12a).
Area-weighted mean edge contrast of deciduous forest measures the contrast
between the eleven non-urban LC/LU classes and deciduous forest in the LC/LU map.
Greatest differences in this metric were between deciduous forest and
urban/residential/recreational grasses, extractive, barren, pasture/hay, croplands, and
natural grass. The greater the incidence of dissimilar LC/LU classes configured around
deciduous forest, the higher the area-weighted mean edge contrast. This metric is also
related to forest fragmentation. For each one percent of the area-weighted mean edge
contrast of deciduous forest in a catchment, delivery from all non-point N sources to
streams draining the estuary was estimated to increase by 0.014%. Fragmentation-related
increases in runoff based upon observed lower hydraulic conductivities in nearby non-
forested LC/LU classes, as compared to forested LC/LU classes, have been found
elsewhere (Chandler and Walter, 1998; Giambelluca, 2002; Ziegler et al., 2004a, 2006,
2007; Zimmermann et al., 2006). The highest area-weighted mean edge contrasts of
deciduous forest per catchment (> 63%) were found in southeastern PA and northern MD
(Figure 2.12b).
For every one percent of cropland composition in a catchment, delivery from all
non-point N sources to streams draining the estuary was estimated to increase by 0.021%.
This may be a result of decreased hydraulic conductivity and increased runoff due to
cultivation practices that may have contributed to applied fertilizer, manure and other
non-point N delivery from these areas. Field compaction from tillage and machinery
44
alone may promote surface sealing and overland runoff (Logsdon and Jaynes, 1996). The
highest percentages of cropland per catchment (> 45%) were in central and southeastern
PA, the eastern shore of MD, and southwestern DE (Figure 2.12c).
For every one percent of evergreen forest composition within riparian stream
buffers, delivery from all non-point N sources to streams draining the estuary was
estimated to increase by 0.013%. Increased overland and shallow subsurface N runoff
correlated with greater evergreen forest has been found elsewhere, albeit in small
catchments, resulting from decreased soil hydraulic conductivities (Allan et al., 1993;
Allan and Roulet, 1994; and Wetzel, 2003). A similar conclusion was also reached in
North Carolina's (US) piedmont region where a greater proportion of forest in some
riparian buffers increased N loadings to streams due to increased overland and shallow
subsurface runoff linked to decreased soil hydraulic conductivity in these buffers
(Verchot et al., 1997). Highest percentages of evergreen forest within riparian stream
buffers per catchment (> 22.5%) were in central PA, central and western VA, and south
central NY (Figure 2.12d).
45
2000 Extractive Land (%)
0.0 - 0.5
0.5 - 1.0
1.0 - 1.5
1.5 - 2.0
2.0 - 2.5
2.5 - 3.0
3.0 - 3.5
3.5 - 4.0
4.0 - 4.5
> 4.5
a)
?
0 40 80 120 16020
Kilometers
2000 Mean Edge Contrast of Deciduous Forest (%)
0 - 7
7 - 14
14 - 21
21 - 28
28 - 35
35 - 42
42 - 49
49 - 56
56 - 63
> 63
b)
?
0 40 80 120 16020
Kilometers
2000 Cropland (%)
0 - 5
5 - 10
10 - 15
15 - 20
20 - 25
25 - 30
30 - 35
35 - 40
40 - 45
> 45
c)
?
0 40 80 120 16020
Kilometers
2000 Evergreen Forest In Buffer (%)
0.0 - 2.5
2.5 - 5.0
5.0 - 7.5
7.5 - 10.0
10.0 - 12.5
12.5 - 15.0
15.0 - 17.5
17.5 - 20.0
20.0 - 22.5
> 22.5
d)
?
0 40 80 120 16020
Kilometers
Figure 2.12: Per catchment land-to-water delivery metrics of: a) Extractive land (%), b)
Area-weighted mean edge contrast of deciduous forest (%), c) Cropland (%), and d)
Evergreen forest within the riparian stream buffer (%) in the 2000 RESAC 31 m TN
model.
46
2000 Barren Land In Buffer (%)
0.00 - 0.25
0.25 - 0.50
0.50 - 0.75
0.75 - 1.00
1.00 - 1.25
1.25 - 1.50
1.50 - 1.75
1.75 - 2.00
2.00 - 2.25
> 2.25
?
0 40 80 120 16020
Kilometers
Figure 2.13: Per catchment land-to-water delivery metric of barren land within the
riparian stream buffer (%) in the 2000 RESAC 31 m TP model.
Finally, for every one percent of barren land composition within the riparian
stream buffer, delivery from all non-point P sources to streams draining the estuary was
estimated to increase by 0.281%. In other studies elsewhere, higher overland runoff
related to greater compositions of barren land with lower hydraulic conductivity has been
found at catchment scales (Ziegler and Giambelluca, 1997, 1998; Ziegler et al., 2001,
2004b; Assouline and Mualem, 2002; Perkins et al., 2007). In addition, a recent study
conducted in Mississippi Basin (US) tributaries within the states of Arkansas and
Missouri determined that compositions of this same landscape metric found significant in
this study (percent barren land in the riparian stream buffer) were correlated to increase P
in overland flow delivered to streams (Lopez et al, 2008). However in that study,
percent barren land in the riparian stream buffer was found to be significant at a fixed 120
m width. The highest percentages of barren land (> 2.25%) in riparian stream buffers
47
were in southeastern PA, central and the eastern shore of MD, southwestern DE, and
central to southeastern VA (Figure 2.13).
2.5.2 Comparisons with B & P models of the Chesapeake Bay watershed
Since both the 31 m TN and TP SPARROW models and the B & P models are
driven by the same non-LC/LU data, predictions of loadings and yields were similar.
However, loading and yield allocations within similar catchments varied between the 31
m and B & P models. Utilizing an F-test, the inclusion of the new additional explanatory
landscape metrics in the 31 m TN model indicated a significant (p value ? 0.05)
difference from the B & P TN model. Additionally, a t-test indicated that barren land in
the 31 m TP model was highly significance (p value ? 0.05) and provided evidence that
the two TP models were statistically different. Thus, with the inclusion of these new
metrics, initial biases found in the B & P models of their original explanatory variables
not accounting for the entire variability in the observed, watershed-wide nutrient loading
station estimates were slightly reduced (Figure 2.14a-d). This slight increase in the
estimated accuracy between the observed and predicted measurements at stream loading
sites was also indicated by the increased yield r2 and decreased RMSE and AIC values
from the 31 m to the B & P TN and TP models (Table 2.2).
48
a)  b)
Breakdown of Variance Explained by Variables in 1997 B & P
TN Model
3.93%
2.62%
49.85%
9.27%1.28%
0.07%
0.42%
0.46%
2.98%
19.16%
9.96%
PS (kg/yr)
AF (kg/yr)
AM (kg/yr)
AD (kg/yr)
UL (kg/ha/yr)
CP (%)
SS (m/day)
IS (m/day)
LS (m/day)
R (m/yr)
UV
Breakdown of Variance Explained by Variables in 2000 RESAC
31 m TN Model
3.91%
2.62%
49.85%
10.13%
6.34%
19.16%
2.38%
1.34%
0.53%
1.76%
0.91%
0.06%
0.51%
0.21%
0.29%
PS (kg/yr)
AF (kg/yr)
AM (kg/yr)
AD (kg/yr)
AWMUPS (kg/ha/yr)
CP (%)
PEL (%)
AWMECDF (%)
PCL (%)
PEFRSB (%)
SS (m/day)
IS (m/day)
LS (m/day)
R (m/yr)
UV
c)  d)
Breakdown of Variance Explained by Variables in 1997 B & P
TP Model
25.06%
34.60%
1.33%
5.13%
7.99%
0.01%
0.00%
0.00%
0.01%
25.87% PS (kg/yr)
AF (kg/yr)
AM (kg/yr)
NANUL (kg/ha/yr)
UL (kg/ha/yr)
SS (m/day)
IS (m/day)
LS (m/day)
R (m/yr)
UV
Breakdown of Variance Explained by Variables in 2000 RESAC
31 m TP Model
25.38%
32.31%
1.35%
6.41%
8.09%
1.48%
0.00%
0.00%
0.00%
0.01%
24.97% PS (kg/yr)
AF (kg/yr)
AM (kg/yr)
AWMNANUPS (kg/ha/yr)
AWMUPS (kg/ha/yr)
PBLRSB (%)
SS (m/day)
IS (m/day)
LS (m/day)
R (m/yr)
UV
Figure 2.14: Comparison of the breakdown of variance explained by variables in the: a)
1997 B & P TN, b) 2000 RESAC 31 m TN, c) 1997 B & P TP, and d) 2000 RESAC 31
m TP models. PS = point sources, AF = applied fertilizer, AM = applied manure, AD =
atmospheric deposition, UL = urban land, CP = coastal plain, SS = small streams, IS =
intermediate streams, LS = large streams, R = reservoir, UV = unexplained variance,
AWMUPS = area-weighted mean urban (? 10% ISA) patch size, PEL = percentage of
extractive land, AWMECDF = area-weighted mean edge contrast of deciduous forest,
PCL = percentage of cropland, PEFRSP = percentage of evergreen forest in the riparian
stream buffer, NANUL = non-agricultural/non-urban land, and AWMNANUPS = area-
weighted mean non-agricultural/non-urban patch size.
49
The TN annual loadings (kg/yr) estimated to enter the Chesapeake Bay from the
31 m model were 1.449 x 108, as compared to 1.480 x 108 generated from the B & P
model, approximately 98% of the B & P model value. The mean annual TN yield
(kg/ha/yr) per catchment draining directly to the Chesapeake Bay from the 31 m model
was about 55.03, as compared to 62.58 from the B & P model. The largest increases in
delivered yield (> 3.00 kg/ha/yr) to the estuary per catchment from the new model, as
compared to the B & P model, were found near: central NY, central and southeastern PA,
and northeastern MD in the upper, middle, and lower Susquehanna Basin; Cumberland
(MD), northeastern WV, and northern VA in the upper-to-middle Potomac Basin; the
eastern shore of MD; and DE (Figure 2.15a). The largest decreases in delivered yield (>
3.00 kg/ha/yr) to the estuary per catchment from the 31 m model, as compared to the B &
P model, were found near: DC in the lower Potomac Basin and Richmond and Petersburg
(VA) in the middle-to-lower James Basin (Figure 2.15a).
The B & P and 31 m models showed close agreement in coefficients for LC/LU
sources (urban land and area-weighted mean urban (? 10% ISA) patch size), other
sources (point, fertilizer and manure applications, and atmospheric N deposition), land-
to-water delivery (coastal plain), stream decay (small, intermediate, and large), and
reservoir decay (Table 2. 4). A comparison of the natural log of the 1997 B & P and
2000 31 m TN model loadings (kg/yr) in the six largest basins (James (27,019 km2),
Patuxent (2,479 km2), Potomac (38,000 km2), Rappahannock (7,405 km2), Susquehanna
(71,225 km2), and York (6,915 km2)) had an r2 value of 0.9984 (Figure 2.16a). The York
Basin is formed by the confluence of the Mattaponi and Pamunkey Basins in southeastern
VA.
50
The TP annual loadings (kg/yr) estimated to reach the Chesapeake Bay from the
new model were 5.367 x 106 versus 5.210 x 106 kg/yr estimated to reach the estuary from
the B & P model, approximately 3% higher than the predicted loadings in the B & P
model. The mean annual TP yield (kg/ha/yr) per catchment draining directly to the
Chesapeake Bay from the 31 m model was about 2.38, as compared to 2.14 from the B &
P model. The largest increases in delivered yield (> 0.08 kg/ha/yr) to the Chesapeake
Bay per catchment from the 31 m model to the B & P model were near: Lancaster and
Harrisburg (PA) in the lower Susquehanna Basin; Baltimore (MD); Frederick and
southern MD, DC, and northern VA in the middle-to-lower Potomac Basin;
Fredericksburg (VA) in the middle Rappahannock Basin; central VA in the upper York
Basin; Norfolk-Virginia Beach-Newport News (VA) in the lower James Basin; the
eastern shores of MD and VA; and DE (Figure 2.15b). The largest decreases in delivered
yield (> 0.08 kg/ha/yr) to the estuary per catchment from the 31 m model to the B & P
model were near: Baltimore (MD) and Lynchburg and Norfolk-Virginia Beach-Newport
News (VA) in the upper-to-lower James Basin (Figure 2.15b).
Similarly to the TN model, the comparisons of the significant coefficients of
LC/LU sources (urban land, area-weighted mean urban (? 10% ISA) patch size, non-
agricultural/non-urban land, and area-weighted mean non-agricultural/non-urban patch
size), other sources (point, fertilizer and manure applications), stream decay (small,
intermediate, and large), and reservoir decay between the B & P and 31 m model were in
close agreement (Table 2. 6). The comparison of the natural log of the 1997 B & P and
2000 31 m TP model loadings for the six largest basins in the watershed indicated an r2
value of 0.9991 (Figure 2.16b).
51
2000-1997 Delivery of TN Yield (kg/ha/yr)
< -3.00
-3.00 - -2.25
-2.25 - -1.50
-1.50 - -0.75
-0.75 - 0.00
0.00 - 0.75
0.75 - 1.50
1.50 - 2.25
2.25 - 3.00
> 3.00
a)
?
0 40 80 120 16020
Kilometers
2000-1997 Delivery of TP Yield (kg/ha/yr)
< -0.08
-0.08 - -0.06
-0.06 - -0.04
-0.04 - -0.02
-0.02 - 0.00
0.00 - 0.02
0.02 - 0.04
0.04 - 0.06
0.06 - 0.08
> 0.08
b)
?
0 40 80 120 16020
Kilometers
Figure 2.15: Per catchment estimated difference between the 2000 RESAC 31 m and
1997 B & P model in predicted yield (kg/ha/yr) delivered to the Chesapeake Bay for: a)
TN and b) TP.
a)  b)
Total Nitrogen (TN)
5
1
3
6
4
2
y = 1.0049x - 0.1632
R2 = 0.9984
4
6
8
10
12
14
16
18
20
4 6 8 10 12 14 16 18 20
LN of 1997 B & P Model Load (kg/yr)
LN
of
20
00
31
m
M
od
el
Lo
ad
(k
g/y
r)
Total Phosphorus (TP)
2
4
6
31
5
y = 1.0006x + 0.0103
R2 = 0.9991
4
6
8
10
12
14
16
18
20
4 6 8 10 12 14 16 18 20
LN of 1997 B & P Model Load (kg/yr)
LN
of
20
00
31
m
M
od
el
Lo
ad
(k
g/y
r)
Figure 2.16: Comparison of the natural log of the 1997 B & P and 2000 RESAC 31 m
model estimated loadings (kg/yr) for the James, Patuxent, Potomac, Rappahannock,
Susquehanna, and the York Basins within the Chesapeake Bay watershed for: a) TN) and
b) TP. 1 = the James River Basin, 2 = the Patuxent River Basin, 3 = the Potomac River
Basin, 4 = the Rappahannock River Basin, 5 = the Susquehanna River Basin, and 6 = the
York River Basin.
52
2.5.3 Comparisons between the Chesapeake Bay HSPF and SPARROW TN and TP
models
The new TN and TP runoff model results were compared with the results of the
Chesapeake Bay HSPF Phase 4.3 simulation parameterized upon 64 to 94 basins used
from 1985-2000 (Wang et al., 2001; Chesapeake Bay Program, 2007). Chesapeake Bay
HSPF modeling has been in progress since 1982 as a management tool for the estuary's
restoration. In comparison with the mean 1985-1994 TN loading of 1.420 x 108 kg/yr
estimated to enter the Chesapeake Bay from the Phase 4.3 HSPF simulation (Wang et al.,
2001), the 2000 31 m TN model loadings were approximately 102% of the HSPF value.
The 2000 31 m TN model loadings were nearly 112% of the 2000 Phase 4.3 HSPF TN
model value of 1.292 x 108 kg/yr (Chesapeake Bay Program, 2007). The comparison
between the natural log of the estimated 2000 Phase 4.3 HSPF and the 31 m model TN
loadings of the six largest basins in the watershed indicated a r2 of 0.9883 (Figure 2.17a).
The close agreement between the SPARROW and the HSPF estimates suggest that the
inclusion of new compositional and configuration representations of catchment and
riparian stream buffer-wide urban and non-urban LC/LU in the SPARROW models could
improve the precision of annual Chesapeake Bay TN runoff simulation..
The mean 1985-1994 estimated HSPF Phase 4.3 Bay TP loading was 9.991 x 106
kg/yr (Wang et al., 2001). The 2000 31 m TP model results were only 54% of this HSPF
value. Furthermore, the 2000 31 m TP model loadings were approximately 62% of the
estimated 2000 Phase 4.3 HSPF Bay TP loadings of 8.673 x 106 kg/yr (Chesapeake Bay
Program, 2007). A comparison between the natural log of the TP loadings for the 2000
Phase 4.3 HSPF and the SPARROW model in the six largest basins in the watershed also
indicated a high r2 of 0.8692 (Figure 2.17b). Although 2000 was 16% below the normal
53
long-term mean Chesapeake Bay watershed-wide stream runoff (United States
Geological Survey, 2001), these findings still suggest an underprediction of TP in
SPARROW. However, the inclusion of new compositional metrics of watershed and
riparian stream buffer-wide urban and non-urban LC/LU within SPARROW slightly
improved the precision of predicted annual Chesapeake Bay-wide TP runoff against
HSPF estimates.
Total Nitrogen (TN)
2
4
6
3
1
5
y = 1.2078x - 3.6588
R2 = 0.9883
4
6
8
10
12
14
16
18
20
4 6 8 10 12 14 16 18 20
LN of 2000 Phase 4.3 HSPF Model Load (kg/yr)
LN
of
20
00
31
m
M
od
el
Lo
ad
(k
g/y
r)
Total Phosphorus (TP)
5
1
3
6
4
2
y = 0.9715x - 0.2988
R2 = 0.8692
4
6
8
10
12
14
16
18
20
4 6 8 10 12 14 16 18 20
LN of 2000 Phase 4.3 HSPF Model Load (kg/yr)
LN
of
20
00
31
m
M
od
el
Lo
ad
(k
g/y
r)
Figure 2.17: Comparison of the natural log of the 2000 Phase 4.3 HSPF and 2000
RESAC 31 m model estimated loadings (kg/yr) for the James, Patuxent, Potomac,
Rappahannock, Susquehanna, and the York Basins within the Chesapeake Bay watershed
for: a) TN) and b) TP. 1 = the James River Basin, 2 = the Patuxent River Basin, 3 = the
Potomac River Basin, 4 = the Rappahannock River Basin, 5 = the Susquehanna River
Basin, and 6 = the York River Basin.
2.6 Conclusions
This paper investigated the effect of LC/LU on simulations of regional nutrient
loading to the Chesapeake Bay estuary using a modification of the USGS Version 3.0
Chesapeake Bay 1997 SPARROW (Brakebill and Preston, 2004 (B & P)) TN and TP
models with the same catchments, but new compositional and configurational landscape
metrics and a consideration of riparian stream buffers for non-point source and land-to-
54
water delivery significance.. New watershed wide maps of LC/LU, % ISA, and % TC for
2000 were used. It was concluded that the new models improved the predictive ability of
SPARROW and the precision of simulated annual TN and TP loadings reaching the
estuary, as compared with HSPF estimates, indicating that these two approaches may
now be more complimentary.
Catchment-wide compositional landscape metrics of area-weighted mean urban (?
10% ISA) and non-agricultural/non-urban patch sizes significantly depicted the
generation of non-point N and P land sources eventually reaching the Chesapeake Bay.
Whereas, compositional and configurational landscape metrics in catchments (percentage
of extractive land, area-weighted mean edge contrast of deciduous forest, and percentage
of cropland) that are thought to be associated with water quality at localized scales, were
shown here to be significantly correlated with the delivery of non-point N loadings to
larger, nested river networks draining the estuary.
Additionally, at the localized scale, riparian stream buffers are thought to
attenuate nutrients eventually reaching the stream channel. However, at the scale of the
entire watershed, the demonstration here of the significant effects of LC/LU on TN and
TP runoff extends to riparian stream buffer compositional percentages of evergreen forest
and barren land increasing delivery of non-point N and P, respectively, to the entire
Chesapeake Bay. Hence, these representations of LC/LU at catchment and riparian
stream buffer width scales should be used in future data-driven water quality models
representative of the entire Chesapeake Bay TN and TP watershed runoff. The increased
transport of non-point N and P from all compositional and configurational land-to-water
delivery landscape metrics to Chesapeake Bay streams at both catchment and riparian
55
stream buffer-wide scales suggest the vital role that overland and shallow subsurface flow
processes may play in enhanced nutrient transport, as a result of decreased soil hydraulic
conductivity linked to these and adjacent cover types.
In regards to the greatest reduction of non-point N transported to the Chesapeake
Bay, future land management should be focused on reducing the highest percentages of
extractive land in catchments located in central PA, western MD, and northeastern WV.
Likewise, in terms of decreasing non-point P transported to streams draining the estuary,
projected land management strategies should involve reducing percentages of barren land
in riparian stream buffers within southeastern PA, central and the eastern shore of MD,
southwestern DE, and central to southeastern VA. Thus, these and the rest of our
findings are relevant to land managers, planners, lawmakers, and other stakeholders in
making better-informed landscape decisions involving not only the overall nutrient health
of Chesapeake Bay, but in similar watersheds elsewhere.
56
Chapter 3: Effects of projected future urban land cover on nitrogen
and phosphorus runoff to Chesapeake Bay2
3.1 Abstract
This paper examined the effects of simulated land cover and land use (LC/LU)
change from 2000 to 2030 on nutrient loadings to the Chesapeake Bay. The SPAtially
Referenced Regression On Watershed Attributes (SPARROW) model was used with
anticipated watershed-wide LC/LU change from a growth forecast model that provides
spatially-explicit probabilities of conversion to impervious surface. The total nitrogen
(TN) and total phosphorus (TP) loadings estimated to enter the Chesapeake Bay were
reduced by 20% and 19%, respectively. In general, as development replaced other
LC/LUs from 2000 to 2030, TN and TP runoff was significantly reduced by losses of
non-point, non-urban source loadings, yields, and land-to-water delivery. The simulation
results suggest future changes in landscape composition and configuration at catchment
and riparian stream buffer width scales could lower TN and TP runoff to the estuary.
2 The material in Chapter 3 was previously published. Roberts, A. D., Prince, S. D., Jantz, C. A., and
Goetz, S. J., 2009. Effects of projected future urban land cover on nitrogen and phosphorus runoff to the
Chesapeake Bay. Ecological Engineering 35(12): 1758-1772.
57
3.2 Introduction
Urbanization is a primary form of land cover and land use (LC/LU) change that is
accelerating and has significant influence on watershed-wide environmental conditions.
Urbanization converts croplands, forests, grasslands, pastures, wetlands, and other cover
types to, in particular, residential and transportation, but also commercial and industrial
uses, increasing significantly areas of impervious surfaces (Tsegaye et al., 2006).
Globally, as population increases and shifts from rural areas to cities, urban expansion is
inevitable. Furthermore, within the United States, population and its associated
development is growing twice as fast in coastal areas as compared to inland areas
(Bartlett et al., 2000; Conway and Lathrop, 2005). As urban land cover continues to
increase, the incidence of non-point (diffuse) source nutrients, such as nitrogen (N) and
phosphorus (P), in streams from impervious cover can be expected to rise significantly.
These nutrients travel from land surfaces to streams as eroded organic and dissolved
inorganic species via overland, shallow interflow, and even baseflow runoff processes.
Excess nutrients are the main causes of eutrophication, hypoxia, and anoxia in rivers,
estuaries, and coastal oceans (Paerl, 2006). Thus, the impacts on nutrient loading (mass
for a specified time) estimates within rivers, estuaries, and coastal oceans of projected
future urban growth are of interest. Some studies have examined scenarios of future
urbanization on non-point source N and P in smaller watershed regions (Tsihrintzis et al.,
1996; Bhaduri et al., 2000; Costanza et al., 2002; Chang, 2004; Filoso et al., 2004; Tang
et al., 2005; Wang et al., 2005), however larger regions have not been studied
thoroughly, yet significant impacts on regional, national, and even global nutrient
loadings can be expected.
58
To quantify the potential future nutrient loadings of a significant larger region, the
Chesapeake Bay watershed was examined. The Chesapeake Bay is the largest estuary in
the United States, with a watershed (166,534 km2) encompassing portions of six states
(New York (NY), Pennsylvania (PA), Delaware (DE), Maryland (MD), West Virginia
(WV), and Virginia (VA)) and the District of Columbia (DC) (Figure 3.1a-b). The
Chesapeake Bay also has the highest watershed land area per volume of water of any
estuary in the United States, making runoff from the land surface critically important in
determining the nutrient status of the estuary (Shuyler et al., 1995). While human-
induced LC/LU transformations that lead to increases in N and P first appeared in the
watershed in the mid-1600s (Boesch, 2006), this rate increased after the end of World
War II in the late 1940s when the Chesapeake Bay population was still under 8 million
(McConnell, 1995). By 2000, the watershed?s population was approximately 15.7
million (Chesapeake Bay Program, 2008a), with expectations of close to 20 million by
2030 (Chesapeake Bay Program, 2008b). Clearly this increase will further drive human-
induced LC/LU changes in the form of urbanization. Increases in nutrient delivery to the
estuary resulting from these population and consequent LC/LU changes have been the
primary focus of research and policy efforts relating to restoring the Chesapeake Bay
through the Federal Water Pollution Control Act (FWPCA) of 1972 and the Clean Water
Act (CWA) of 1977 (Morgan and Owens, 2001). Thus, watershed-wide examinations of
projected future urbanization on delivered N and P loadings to the estuary are warranted.
59
PA
VA
NY
WV
MD NJ
NC
OH
DE
DC
JAM E
S R
SUS
QUEHAN N A R
PO
TO
MA
C R
RA
PP
AH
ANN
OCK
R
PA
TU
XE
NT
R
MA
TTA
PO
NI RJA
ME
S R
SU
SQ
UE
HA
NN
A R
PAT
UX
EN
T R
PAM
UNK
EY
R
a)
?
0 40 80 120 16020
Kilometers
Baltimore, MD
Washington, DC--MD--VA
Richmond, VA
Harrisburg, PA
Norfolk--Virginia Beach--Newport Ne
Scranton--Wilkes-Barre, PA
York, PALancaster, PA
Lynchburg, VA
Binghamton, NY
Petersburg, VA
Cumberland, MD--WV
Elmira, NY
Altoona, PA
Frederick, MD
Fredericksburg, VA
Annapolis, MD
Hagerstown, MD--PA--WV
Charlottesville, VA
Williamsport, PA
State College, PAPA
VA
NY
WV
NJ
NC
OH
DEMD
b)
?
0 40 80 120 16020
Kilometers
Figure 3.1: The Chesapeake Bay watershed showing the locations of: a) Streams and
rivers draining the estuary and b) Urban centers located within its boundaries.
The spatial pattern of urban development in the Chesapeake Bay watershed is
increasingly taking the form of low-density, decentralized residential and commercial
development (Jantz et al., 2004). Previous research has indicated that between 1970 and
2000, lot sizes throughout the watershed increased by 60% and that the average home
size also increased from 1500 to 2265 ft2 (Chesapeake Bay Program, 2008c). These
trends are expected to continue over time (Chesapeake Bay Program, 2008c). It is
reasonable to assume that water quality and aquatic habitats in the watershed will decline
due to this urbanization, but low density development may have less effect than earlier,
more concentrated expansion. Projection of the current trend of growth to 2030 may
provide a better insight into the probable effects on the Chesapeake Bay nutrient status.
Watershed-wide, spatially-explicit, predictions of urbanization with 30 m
resolution have been modeled by Jantz et al. (in press) using the Slope, Land use,
60
Exclusion, Urban extent, Transportation, and Hillshade (SLEUTH) urban growth model.
SLEUTH is a cellular automaton, pattern-extrapolation model calibrated using urban
development patterns in the past and forecasts of these patterns into the future (Jantz and
Goetz, 2005). This version of SLEUTH was developed from the Clarke urban growth
(Clarke et al., 1997) and land cover change models (United States Geological Survey,
2008a). SLEUTH has been applied to model urban growth in numerous areas (Clarke
and Gaydos, 1998; Silva and Clarke, 2002; Arthur-Hartranft et al., 2003; Herold et al.,
2003; Yang and Lo, 2003; Dietzel and Clarke, 2004; Solecki and Oliveri, 2004; Xian and
Crane, 2005; Xian et al., 2005). In addition to the Jantz et al. (in press) recent application
of SLEUTH to the entire Chesapeake Bay watershed, the model has been previously
applied (although to smaller regions) in the watershed near Baltimore (MD), DC, and
State College (PA) (Clarke and Gaydos, 1998; Carlson, 2004; Jantz et al., 2004; Claggett
et al., 2005).
Increases in impervious surface areas may contribute more non-point N and P to
the Chesapeake Bay as a result of increases in: leaf litter, vehicle emissions, residential
and roadside landscaping (fertilizers), urban wildlife and pets, construction, and
infrastructure (Minton, 2002).
The projected effects of watershed-wide urban growth and consequent effects on
LC/LU and its changes on nutrient loadings were estimated using the United States
Geological Survey's (USGS) SPAtially Referenced Regressions On Watershed
Attributes (SPARROW) model (Smith et al., 1997; Schwarz et al., 2006). SPARROW
estimates total nitrogen (TN) and total phosphorus (TP) runoff from watersheds of
various sizes by statistical functions that relate upstream point and non-point sources,
61
land-to-water delivery variables, and stream and reservoir nutrient attenuation (loss)
processes that change TN and TP loadings as they travel downstream through nested
river channel and reservoir networks. Quantities of contaminants in streams may be
expressed as either loadings or yields (mass loading normalized by drainage area). Land-
to-water delivery variables describe properties of the landscape relating climatic, natural,
and human-induced surface processes affecting non-point N and P transport to streams.
The hybrid statistical-process structure of the model allows for the
implementation of deterministic functions (such as first-order stream loss functions) with
spatially-distributed components (such as sources and land-to-water delivery variables
within stream networks) to account for the dendritic nature of watersheds (Alexander et
al., 2002b). SPARROW is based upon an average year and does not account for
individual storms (that is, it is not event-based). The model simulates average annual
discharge in terms of loadings and yields locally generated and delivered to the estuary,
using just those input variables that have significant effects (p value ? 0.05). Local
generation is the amount of TN or TP generated from within each catchment independent
of any upstream loadings or yields, whereas delivered refers to the amount of TN or TP
reaching the estuary after accounting for any upstream loadings and yields and also in-
stream and reservoir losses. The model is applied to each of 2,339 catchments of the
Chesapeake Bay watershed (Figure 3.2a) using a watershed map by Brakebill and Preston
(2004). The model has been applied to the Chesapeake Bay watershed using the National
Land Cover Dataset (NLCD) (Preston and Brakebill, 1999; Brakebill et al., 2001;
Brakebill and Preston, 2004) and by Roberts and Prince (2010) incorporating the
Regional Earth Science Application Center?s (RESAC) remotely sensed LC/LU and
62
percent impervious surface area (% ISA) maps for 2000 (Goetz et al., 2003, 2004a, b;
Jantz et al., 2005).
PA
VA
NY
WV
MD NJ
NC
OH
DE
DC
?
0 40 80 120 16020
Kilometers
a)
Example Catchment
Runoff from Catchment
Enters Stream
31 m Buffer
Upstream Catchments
Carrying N Loads
Upstream Plus Locally Generated
Runoff, Less Instream Losses
Leaves Catchment
b)
?
0 1
Kilometer
i
i
i
ii
i
Figure 3.2: 2,339 catchments used in the: a) 2000 RESAC 31 m SPARROW
Chesapeake Bay TN and TP models with b) Schematic diagram of an example catchment
area showing fixed 31 m riparian stream buffer width area surrounding stream reach and
corresponding upstream and downstream runoff paths.
Roberts and Prince (2010) used entire catchment and 31 m riparian stream buffer
landscape metrics to specify non-point sources and land-to-water delivery variables that
affect TN and TP loadings to the Chesapeake Bay. Landscape metrics describe the
spatial structure of patches, the cover classes of patches, and patch mosaics, thus
providing measures of composition (the variety and abundance of patch types) and
configuration (spatial character and arrangement, position, and orientation of landscape
elements (Leitao et al., 2006)). The use of landscape metrics, in conjunction with
projected future urban growth, has previously indicated that spatial alterations in LC/LU
63
affect predicted N and P loadings in streams throughout a significant portion of
Chesapeake Bay watershed (Wickham et al., 2002), although a holistic watershed
approach was not implemented. Even in remote reaches of the watershed, locally-
dependent land development decisions that lead to more urban growth can adversely
affect downstream loadings to the estuary. Thus, holistic watershed management is
needed to bridge this gap between land use planning and comprehensive natural resource
management (Conway and Lathrop, 2005). Modeling is needed to integrate the local
catchment level to impacts on the entire Chesapeake Bay watershed.
Thus, the overall purpose of this study was to estimate future TN and TP runoff to
the Chesapeake Bay, using SPARROW models, with maps of projected future
urbanization.
3.3 Materials and methods
3.3.1 Future Chesapeake Bay watershed land cover and land use
Maps of the projected development in 2030 were derived from SLEUTH model
runs for the Chesapeake Bay watershed (Jantz et al., in press). The model was calibrated
based upon data delivered to the Chesapeake Bay Program that was stratified by sub-
region and two dates of impervious cover. For more information about the parameters
and validation exercises used to calibrate these SLEUTH model runs, please consult Jantz
et al. (in press).
Results from these SLEUTH model runs were then used to make projections of
the probabilities of development in 2030 by acting as surrogates for impervious surface
area (ISA) at the 30 x 30 m (900 m2) pixel scale used to define urban areas in 2000 for
64
the entire watershed. The probabilities were converted to a binary pixel classification of
0 that equaled projected non-urban areas and 1 that equaled projected urban areas by
2030. These new classes were then used to represent urban (? 10% ISA) (class 1) and
non-urban (< 10% ISA) (class 2) areas characterized in 2000 by Roberts and Prince
(2010). Class 2 pixels were assumed to remain in their existing twelve LC/LU classes, as
given by the 2000 RESAC LC/LU map (Goetz et al., 2004a, b; Jantz et al., 2005). All
fourteen LC/LU used in this study are given in Table 3.1. Complete accounts of the
LC/LU mapping methodologies are given in Goetz et al. (2004b) and Jantz et al. (2005).
Ten 2030 SLEUTH map runs were averaged to create mean values of 2030 LC/LU.
2000 and 2030 land cover classes Class number
Urban (? 10% ISA) 1
Non-urban (< 10% ISA) 2
Urban/residential/recreational grasses 3
Extractive 4
Barren 5
Deciduous forest 6
Evergreen forest 7
Mixed (deciduous-evergreen) forest 8
Pasture/hay 9
Croplands 10
Natural grass 11
Deciduous wooded wetland 12
Evergreen wooded wetland 13
Emergent (sedge-herb) wetland 14
Table 3.1: The fourteen Chesapeake Bay watershed land cover classes used.
3.3.2 Landscape metrics and Chesapeake Bay SPARROW models
The models used here addressed some shortcomings of previous Chesapeake Bay
TN and TP SPARROW models (Preston and Brakebill, 1999; Brakebill et al., 2001;
Brakebill and Preston, 2004), including: adding a relationship between LC/LU
composition and the land-to-water delivery of non-point N and P; consideration of
landscape configuration; and other spatial configuration factors such as LC/LU in
65
riparian zones. Roberts and Prince (2010) found that LC/LU composition and
configuration in catchments and composition in riparian stream buffers improved the
accuracy and precision of TN and TP loadings estimates for 2000. A complete account
of the definitions, techniques, and other methods used regarding the incorporation of
landscape metrics and the 2000 SPARROW model calibrations is given in Roberts and
Prince (2010).
In the 2000 RESAC 31 m models, five metrics (1-3, 6,7) measuring landscape
configuration and two metrics (4-5) measuring composition (Leitao et al., 2006) were
initially evaluated in SPARROW for each catchment to quantify the effects of non-urban
and urban land cover on current TN and TP runoff to the Chesapeake Bay. The
following summaries indicate the properties that each metric measures. Complete
definitions for all seven metrics are given in Leitao et al. (2006).
1) Contagion quantifies the degree to which LC/LU types were clumped in larger
patches as opposed to dispersed in many smaller fragments.
2) Area-weighted mean radius of gyration measures connectivity using
correlation length. This is the average distance one might traverse across a map from a
random starting point and moving in a random direction while remaining in the patch.
3) Patch number indicates total number of patches of a particular LC/LU.
4) Percentage of the landscape area composed of a specified LC/LU.
5) Area-weighted mean patch size quantifies the sum, across all patches of a
particular LC/LU, of patch area multiplied by proportional abundance of the patch.
66
6) Area-weighted mean edge contrast quantifies the amount of contrast between
adjacent LC/LU patches. In this application, contrast is defined as physical
characteristics of differing cover types that may influence nutrient transport and delivery.
7) Area-weighted mean Euclidean nearest neighbor distance quantifies the
shortest distance from one patch to the next patch of the identical LC/LU type.
All area-weighted mean landscape metrics weight each patch by its size relative to the
total area of that specific LC/LU, meaning that larger patches will exert greater influence
than smaller patches and they are insensitive to extremely small patches.
Of the seven landscape metrics analyzed, only three (4-6) were found to be
significant in the 2000 models (Table 3.2) and so the others were excluded. Thus, the
SPARROW models were based upon landscape metrics for LC/LU in whole catchments
and in just a 31 m riparian stream buffer. Riparian stream buffers are defined here as
fixed, transitional areas between terrestrial landscapes and stream reaches created from
the linked, spatially-referenced watershed network (Figure 3.2b). Non-point sources and
land-to-water delivery variables from the models calibrated with 2000 data (Roberts and
Prince, 2010) were used to model future Chesapeake Bay TN and TP loadings using
maps of LC/LU classes in 2030.
The predictor variables (Table 3.2) used in the 2000 models are described below;
for further details see Roberts and Prince (2010). In the TN model, for each kilogram
(kg) of: 1) point sources discharged, 2) fertilizer and 3) manure applied to agricultural
land (crop and pasture) and 4) atmospheric N deposited on the watershed, approximately
1.2, 0.2, 0.1, and 0.5 kg of N, respectively, were estimated in Chesapeake Bay streams
annually. For each hectare (ha) of: 5) area-weighted mean urban (? 10% ISA) patch size,
67
nearly 25 kg of N was estimated in Chesapeake Bay streams annually. Urban patch size
was used to estimate the buildup and washoff of N to sewers from urban non-point
sources. For each percent (%) of: 6) land on the coastal plain, non-point N delivered to
streams was estimated to decrease by 0.73% annually. Unlike the coastal plain, for each
% of: 7) extractive, 8) cropland, and 9) area-weighted mean edge contrast of deciduous
forest, non-point N delivered to streams was estimated to increase by 0.27, 0.021, and
0.014% annually, respectively. Area-weighted mean edge contrast of deciduous forest
was used to measure the N transport differences between eleven non-urban classes and
deciduous forest in the 2000 RESAC LC/LU map. Greatest differences between
deciduous forest were with: urban/residential/recreational grasses, extractive, barren,
pasture/hay, croplands, and natural grass. For each % of: 10) evergreen forest in the
riparian stream buffer, non-point N delivered to streams was estimated to increase by
0.013% annually. In the case of all four of these land-to-water delivery metrics, these
variables estimate land properties (reduced hydraulic conductivities), associated with
these or surrounding cover types at the soil and shallow subsurface scales, that increased
N-enriched overland and shallow subsurface runoff. For every meter (m) traveled in: 11)
small (mean flow ? 100 cubic feet per second (ft3/s)), 12) intermediate (mean flow >100
and ? 500 ft3/s), and 13) large (mean flow > 500 ft3/s) streams per day, about 25, 9, and
3% of the instream N was estimated to be lost, respectively. Finally, for any: 14)
reservoir in the watershed, N was removed from the water column at an average settling
velocity of over 14 m/yr.
68
Model component
(units)
TN model
Yield r2 = 0.9366
RMSE = 0.2406
TP model
Yield r2 = 0.7503
RMSE = 0.3216
Variable Coefficient
Value
P value Variable Coefficient
value
P value
Source
(1 = kg/yr, 2 = kg/ha/yr)
Point1 1.173 1.2 x 10-4 Point1 0.738 1.3 x 10-6
Applied fertilizer1 0.175 3.9 x 10-6 Applied fertilizer1 0.016 1.1 x 10-3
Atmospheric
deposition1
0.492 2.0 x 10-7 Applied manure1 0.008 3.0 x 10-2
Applied manure1 0.078 7.7 x 10-4 * Area-weighted mean
non-agricultural/non-
urban patch size2
0.110 5.8 x 10-13
* Area-weighted mean
urban (? 10% ISA) patch
size2
24.885 7.2 x 10-7 * Area-weighted mean
urban (? 10% ISA) patch
size2
0.921 1.0 x 10-4
Landscape delivery
(%)
Percentage of coastal
plain
-0.729 4.1 x 10-8 * Percentage of barren
land within the riparian
stream buffer
0.281 1.2 x 10-6
* Percentage of
extractive land
0.270 7.4 x 10-4
* Area-weighted mean
edge contrast of
deciduous forest
0.014 7.2 x 10-3
* Percentage of cropland 0.021 1.1 x 10-4
* Percentage of
evergreen forest within
the riparian stream
buffer
0.013 4.7 x 10-2
Stream decay
(m/day)
Small streams 0.249 5.5 x 10-2 Small streams -0.198 1.3 x 10-1
Intermediate streams 0.090 3.2 x 10-1 Intermediate streams 0.150 1.9 x 10-1
Large streams 0.030 4.8 x 10-1 Large streams 0.034 5.2 x 10-1
Reservoir decay
(m/yr)
Reservoir 14.224 2.3 x 10-2 Reservoir 19.019 5.7 x 10-2
Table 3.2: All significant (p value ? 0.05) variables in the 2000 RESAC 31 m TN and
TP SPARROW models (Roberts and Prince, 2010). RMSE = root mean squared error. *
Denotes all significant landscape metrics.
In the TP model, for each kg of: 1) point sources discharged, 2) fertilizer and 3) manure
applied to agricultural land, approximately 0.7, 0.02, and 0.01 kg of P, respectively, were
estimated in Chesapeake Bay streams annually. For each ha of: 4) area-weighted mean
non-agricultural/non-urban patch size, over 0.1 kg of P was estimated in Chesapeake Bay
streams annually. This metric represented patches of mainly forest that exported P to
streams as smaller predominant quantities of dissolved inorganic P via groundwater
(baseflow) and shallow subsurface discharges. It was created by subtracting area-
69
weighted mean cropland patch sizes from area-weighted mean non-urban patch sizes.
For each ha of: 5) area-weighted mean urban (? 10% ISA) patch size, approximately 1
kg of P was estimated in Chesapeake Bay streams annually. For each % of: 6) barren
land in the riparian stream buffer, non-point P delivered to streams was estimated to
increase by 0.281% annually. This metric also represented reductions in hydraulic
conductivity at the soil surface that increased P-enriched overland runoff. For every m
traveled in: 7) small streams per day, an increase of about 20% P was estimated to occur,
thus indicating small streams were a source by acting as a mechanism to erode P-enriched
sediments. For each m traveled in: 8) intermediate and 9) large streams per day, about 15
and 3% of the instream P was estimated to be lost, respectively. Finally, for any: 10)
reservoir in the watershed, P was removed from the water column at an average settling
velocity of over 19 m/yr.
Although some stream and reservoir decay coefficients have p values > 0.05 that
indicate these variables are statistically insignificant, these variables are still included in
model calibrations on the grounds of being mechanistically significant within the
SPARROW model structure.
In all, five TN and three TP landscape metrics were significant non-point sources
or land-to-water delivery variables in the model (Table 3.2) (Roberts and Prince, 2010).
The predicted values of these metrics in 2030 were calculated using forecasted 2030
LC/LU data (Table 3.1). ISA was modeled using SLEUTH while the other LC/LU
classes in 2030 were estimated using the 2000 RESAC LC/LU map for all areas that were
not predicted to become ISA.
70
3.3.3 Projected fertilizer and manure applications and point source loadings
The values of those source loading variables that were significant in the TN and
TP models that are subject to LC/LU change, such as fertilizer and manure applications
and point source loadings, were projected forward to 2030. The 2000 atmospheric
deposition of N was unchanged since the aim of this work was to determine the effects of
urbanization on source loadings and because of the difficulty in forecasting a variable
that is largely determined by policy, regulation, and legal changes.
The 2030 annual commercial fertilizer and manure loadings were only considered
for pasture and row-crop (croplands), as in the simulations by Brakebill and Preston (B &
P) (2004). This was done to ensure consistency with B & P, since SPARROW is subject
to changes in model structure; for example the number of source or land-to-water
delivery variables found to be significant may change if the models are calibrated with
new datasets. Fertilizer and manure application rates used in the Phase 5.0 Hydrologic
Simulation Program FORTRAN (HSPF) Chesapeake Bay model (Chesapeake
Community Modeling Program, 2008) were used. These data provided fertilizer and
manure loading rates in terms of several chemical forms of N and P, applied to cropland
and pasture within 1,000 watershed segments. Fertilizer was defined as applications of
ammonia-N (NH3N) and/or nitrate-N (NO3N) for N and phosphate-P (PO4P) for P.
Manure was defined as applications of ammonia-N (NH3N), nitrate-N (NO3N), and/or
organic N for N and phosphate-P (PO4P) and/or organic P for P. The data also included
several management strategies (high till, low till, no till, and nutrient management) used
on cropland and pasture for each month in the year. From these data, annual mean
applied rates of N fertilizer were calculated; 28.02 and 15.83 kg/ha/yr, for cropland and
71
pasture, respectively, whereas P fertilizer had rates of 17.19 and 5.16 kg/ha/yr for these
same cover types. Annual mean applied manure rates for N manure were 9.79 and 52.74
kg/ha/yr, for cropland and pasture. Finally, P manure rates used for cropland and pasture
were 5.56 and 23.77 kg/ha/yr.
Using the 2030 areas of cropland and pasture in all of the 2,339 model
catchments, 2030 annual fertilizer and manure application loadings were tabulated. To
determine the total 2030 fertilizer and manure application loadings of N and P, cropland
and pasture quantities per catchment were combined. The 2000 and projected 2030
watershed-wide estimates of fertilizer and manure applications are given in Table 3.3.
Population throughout the watershed in 2030 was predicted for each catchment
using an empirical correlation of population with SLEUTH ISA output transformed to
housing density (United States Geological Survey, 2008b). For 2000, a non-urban land
density of 0.0615 housing units/acre and an urban land density of 2.1 housing units/acre
were used.
log(population density) = 3.18 + log(housing density) (1)
Utilizing Equation 1, these housing densities lead to population densities of 93 people per
square mile for non-urban land and 3,178 people per square mile for urban land. The
2000 population of the Chesapeake Bay watershed was 15,710,840 (Chesapeake Bay
Program, 2008a) while the estimate using Equation 1 was 15,761,476 only 0.003% higher
than the official tally. To project population densities in 2030 based upon housing
densities, a non-urban land density of 0.0615 housing units/acre was once again used.
However, a lower housing density of a 1.5 housing units/acre replaced the 2000 urban
land value to represent the effect of continued increases in area occupied by each
72
dwelling. These housing densities lead to population densities of 93 people per square
mile for non-urban land and 2,270 people per square mile for urban land in 2030. Thus,
the population estimated for 2030 was 19,761,581, quite similar to the near 20 million
estimate made for the catchment as a whole by the Chesapeake Bay Program (2008b).
Equation 1 was preferred, rather than whole-watershed estimates of future population,
since the modeling was based on populations in each of the 2,339 catchments, not a
single, Chesapeake Bay watershed-wide projection.
Recent wastewater treatment plant (WWTP) estimates show that, on average, 2.72
and 0.16 kg/yr of N and P are discharged per person into the watershed (Cummins,
2004). Using the projected urban population gains, in conjunction with these discharge
values, provided estimates for 2030 N and P point source loadings (kg/yr) from each
catchment with municipal WWTPs discharging into Chesapeake Bay waterways, as of
2000. For the 2,339 catchments, all estimated increases in point discharges were then
assigned to their nearest WWTP for 2030 projections. A comparison of 2000 to the
projected 2030 estimated increase in the point source N and P loadings discharged to
streams draining the Chesapeake Bay is shown in Table 3. 3.
73
TN model TP model
Variable 2000 2030 2030-2000
change
2030-2000
% change
Variable 2000 2030 2030-2000
change
2030-2000
% change
Point 3.6605 x 107 4.8122 x 107 +1.1517 x 107 +31.46 Point 2.6955 x 106 3.3730 x 106 +6.7750 x 105 +25.13
Applied fertilizer 1.9346 x 108 7.6587 x 107 -1.1687 x 108 -60.41 Applied fertilizer 6.6409 x 107 3.6437 x 107 -2.9972 x 107 -45.13
Applied manure 8.7020 x 107 1.4673 x 108 +5.9710 x 107 +68.62 Applied manure 7.3470 x 107 6.7662 x 107 -5.8080 x 106 -7.91
* Area-weighted
mean urban (?
10% ISA) patch
size
109 182 73 +66.97 * Area-weighted
mean non-
agricultural/
non-urban patch
size
6,629 2,398 -4,231 -63.83
* Area-weighted
mean urban (?
10% ISA) patch
size
109 182 73 +66.97
Percentage of
extractive land
0.21 0.16 -0.05 -23.81 Percentage of
barren land
within the
riparian stream
buffer
0.43 0.35 -0.08 -18.60
Area-weighted
mean edge
contrast of
deciduous forest
31.84 28.34 -3.50 -10.99
Percentage of
cropland
10.05 8.68 -1.37 -13.63
Percentage of
evergreen forest
within the
riparian stream
buffer
5.66 5.31 -0.34 -6.18
Table 3.3: Comparison of 2000 and projected 2030 Chesapeake Bay watershed-wide total discharged point source and applied
fertilizer and manure loadings (kg/yr), land-based source variables (ha), and land-to-water delivery variables (%). * Denotes the
averaged value of these land-based source variables from all 2,339 catchments.
74
3.4 Results
3.4.1 TN
TN annual loadings projected to be delivered to the Chesapeake Bay by 2030
were 1.171 x 108 kg/yr, as compared to 1.449 x 108 kg/yr estimated in 2000 (Roberts and
Prince, 2010), about 19% less than the 2000 quantity (Table 3.4). Using the root mean
squared error (RMSE) of the TN model (0.2406) (Table 3.2), uncertainties of the model
predictions were obtained and indicated that this projected 19% reduction in total
loadings was well within the range of change that predicted total loadings could of
decreased by as much as over 50% or increased by upwards of 32% between 2000 and
2030 (Table 3.5). The highest increases in projected TN yield (> 4 kg/ha/yr) were
predicted near: Harrisburg, Lancaster and York (PA); the northern and the eastern shore
of MD; DE; and central VA (Figure 3.3a). Catchments with the largest decreases in
projected TN yield (> 4 kg/ha/yr) were predicted near Baltimore (MD) and DC (Figure
3.3a). A comparison of the six largest basins - the James (27,019 km2), Patuxent (2,479
km2), Potomac (38,000 km2), Rappahannock (7,405 km2), Susquehanna (71,225 km2),
and York (6,915 km2 basin formed by the confluence of the Mattaponi and Pamunkey in
southeastern VA) from 2000 to 2030 indicated that annual TN loadings in three of these
basins (the James, Patuxent, and Rappahannock) were predicted to increased, while the
others decreased. In all six basins, the overall annual TN loadings were predicted to
decrease by nearly 17% from 1.1001 x 108 in 2000 to 9.1776 x 107 by 2030 and were
close to the Chesapeake-wide projected decline of 19%.
75
TN model TP model
Variable 2000 2030 2030-2000
change
2030-2000
% change
Variable 2000 2030 2030-2000
change
2030-2000
% change
Point 4.4105 x 107 5.2200 x 107 +8.0950 x 106 +18.35 Point 1.9665 x 106 2.3677 x 106 +4.0120 x 105 +20.40
Applied fertilizer 4.8395 x 107 1.5615 x 107 -3.2780 x 107 -67.73 Applied fertilizer 1.0226 x 106 5.4169 x 105 -4.8091 x 105 -47.03
Applied manure 9.7110 x 106 1.2783 x 107 +3.0720 x 106 +31.63 Applied manure 5.2862 x 105 4.8027 x 105 -4.8350 x 104 -9.15
Atmospheric
deposition
3.5803 x 107 2.7355 x 107 -8.4480 x 106 -23.60 Area-weighted
mean non-
agricultural/
non-urban patch
size
1.5772 x 106 5.0174 x 105 -1.0755 x 106 -68.19
Area-weighted
mean urban (?
10% ISA) patch
size
6.9311 x 106 9.1806 x 106 +2.2495 x 106 +32.46 Area-weighted
mean urban (?
10% ISA) patch
size
2.7222 x 105 4.0387 x 105 +1.3165 x 105 +48.36
Total 1.4495 x 108 1.1713 x 108 -2.7820 x 107 -19.19 Total 5.3671 x 106 4.2953 x 106 -1.0718 x 106 -19.97
Point 16.43 16.98 +0.55 +3.35 Point 0.79 0.82 +0.03 +3.80
Applied fertilizer 3.84 1.01 -2.83 -73.70 Applied fertilizer 0.08 0.04 -0.04 -50.00
Applied manure 0.42 0.75 +0.33 +78.57 Applied manure 0.03 0.03 +0.00 +0.00
Atmospheric
deposition
2.39 1.91 -0.48 -20.08 Area-weighted
mean non-
agricultural/non-
urban patch size
0.11 0.04 -0.07 -63.64
Area-weighted
mean urban (?
10% ISA) patch
size
0.85 0.84 -0.01 -1.18 Area-weighted
mean urban (?
10% ISA) patch
size
0.03 0.04 +0.01 +33.33
Total 23.93 21.49 -2.44 -10.20 Total 1.04 0.97 -0.07 -6.73
Table 3.4: Comparison of 2000 and projected 2030 Chesapeake Bay watershed-wide total loadings (kg/yr) delivered to the estuary
from all significant sources and mean yield (kg/ha/yr) from all 2,339 catchments.
76
TN model TP model
Variable 2000 2030 2030-2000
change
2030-
2000 %
Change
Variable 2000 2030 2030-2000
change
2030-
2000 %
ChangePoint 4.4105 x 107
?1.0612 x107
5.2200 x 107
?1.2559 x107
-1.5076 x 107 -
+3.1266 x 107
-27.55 -
+93.35
Point 1.9665 x 106
?6.3243 x105
2.3677 x 106
?5.6991 x105
-9.9268 x 105 -
+1.7951 x106
-38.20 -
+134.56
Applied fertilizer 4.8395 x 107
?1.1644 x107
1.5615 x 107
?3.7570 x106
-4.8181 x 107 -
-1.7379 x 107
-80.25 -
-47.29
Applied fertilizer 1.0226 x 106
?3.2887 x105
5.4169 x 105
?1.3038 x105
-9.8399 x 105 -
+2.2166 x 104
-72.81 -
+3.20
Applied manure 9.7110 x 106
?2.3365 x106
1.2783 x 107
?3.0756 x106
-2.3401 x 106 -
+8.4841 x106
-19.42 -
+115.05
Applied manure 5.2862 x 105
?1.7000 x105
4.8027 x 105
?1.1560 x105
-3.7281 x 105 -
+2.7611 x 105
-53.64 -
+76.99
Atmospheric
deposition
3.5803 x 107
?8.6142 x106
2.7355 x 107
?6.5816 x106
-2.3644 x 107 -
+6.7478 x106
-53.23 -
+24.82
Area-weighted
mean non-
agricultural/
non-urban
patch size
1.5772 x 106
?5.0723 x105
5.0174 x 105
?1.2077 x105
-1.7440 x 106 -
-4.0687 x 105
-83.67 -
-38.03
Area-weighted
mean urban (?
10% ISA) patch
size
6.9311 x 106
?1.6676 x106
9.1806 x 106
?2.2089 x106
-1.6270 x 106 -
+6.1260 x106
-18.92 -
+116.39
Area-weighted
mean urban (?
10% ISA) patch
size
2.7222 x 105
?8.7546 x104
4.0387 x 105
?9.7212 x104
-8.5781 x 104 -
+3.4908 x 105
-23.84 -
+189.03
Total 1.4495 x 108
?3.4875 x107
1.1713 x 108
?2.8181 x107
-9.0876 x 107 -
+3.5236 x107
-50.54 -
+32.01
Total 5.3671 x 106
?1.7261 x106
4.2953 x 106
?1.0339 x106
-4.1792 x 106 -
+2.0356 x 106
-58.92 -
+55.91
Point 16.43
?3.95
16.98
?4.09
-7.49 -
+8.59
-36.74 -
+68.83
Point 0.79
?0.25
0.82
?0.20
-0.48 -
+0.54
-46.72 -
+102.21
Applied fertilizer 3.84
?0.92
1.01
?0.24
-4.00 -
-1.66
-83.90 -
-57.03
Applied fertilizer 0.08
?0.03
0.04
?0.01
-0.08 -
+0.00
-74.33 -
-2.59
Applied manure 0.42
?0.10
0.75
?0.18
+0.05 -
+0.61
+9.31 -
+191.72
Applied manure 0.03
?0.01
0.03
?0.01
-0.20 -
+0.20
-48.67 -
+94.81
Atmospheric
deposition
2.39
?0.58
1.91
?0.46
-1.51 -
+0.56
-51.08 -
+30.56
Area-weighted
mean non-
agricultural/non
-urban patch
size
0.11
?0.04
0.04
?0.01
-0.12 -
-0.02
-81.33 -
-29.16
Area-weighted
mean urban (?
10% ISA) patch
size
0.85
?0.20
0.84
?0.20
-0.42 -
+0.40
-39.51 -
+61.44
Area-weighted
mean urban (?
10% ISA)patch
size
0.03
?0.01
0.04
?0.01
-0.01 -
+0.03
-31.56 -
+159.75
Total 23.93
?5.76
21.49
?5.17
-13.37 -
+8.49
-45.03 -
+46.71
Total 1.04
?0.33
0.97
?0.23
-0.72 -
+0.58
-52.12 -
+81.70
Table 3.5: Comparison of 2000 and projected 2030 Chesapeake Bay watershed-wide total loadings (kg/yr) delivered to the estuary
from all significant sources and mean yield (kg/ha/yr) from all 2,339 catchments with prediction errors for each year, range of 2030-
2000 change, and range of 2030-2000% change.
77
2030-2000 Delivery of TN Yield (kg/ha/yr)
< -4
-4 - -3
-3 - -2
-2 - -1
-1 - 0
0 - 1
1 - 2
2 - 3
3 - 4
> 4
a)
?
0 40 80 120 16020
Kilometers
2030-2000 Delivery of TP Yield (kg/ha/yr)
< -0.20
-0.20 - -0.15
-0.15 - -0.10
-0.10 - -0.05
-0.05 - 0.00
0.00 - 0.05
0.05 - 0.10
0.10 - 0.15
0.15 - 0.20
> 0.20
b)
?
0 40 80 120 16020
Kilometers
Figure 3.3: Per catchment estimated 2030-2000 difference maps of the total yield in
kg/ha/yr per year for: a) N and b) P delivered to the Chesapeake Bay.
3.4.2 TP
2030 TP annual loadings projected to reach the Chesapeake Bay estuary were
20% lower (4.295 x 106 kg/yr) than the 5.367 x 106 kg/yr predicted to enter the estuary in
2000 (Roberts and Prince, 2010) (Table 3.4). Based upon the RMSE for the TP model
(0.3216) (Table 3.2), predictions of model uncertainty for the change in TP loadings
indicated that this 20% reduction value was also well within the range of change that
projected total loadings could of decreased by as much as 59% or increased by 56%
between 2000 and 2030 (Table 3.5). Catchments with the highest increases in projected
TP yield (> 0.20 kg/ha/yr) were also near: Harrisburg, Lancaster, and York (PA);
northern and the eastern shore of MD; DE; and central VA (Figure 3.3b). The largest
decreases in projected TP yield (> 0.20 kg/ha/yr) were predicted to occur in the same
regions as the largest decreases in TN yield (Figure 3.3b). The annual TP loadings
78
comparison of the watershed's six largest basins from 2000 to 2030 was also projected to
show an overall decline of approximately 21%, with TP loadings decreasing in all
individual basins.
3.5 Discussion
3.5.1 Agricultural land losses and reductions in total and agricultural non-point
loadings
Overall, the 2030 SPARROW modeled results showed that the predicted
conversion of agricultural land to urban uses throughout the Chesapeake Bay watershed
can be expected to result in significant reductions in delivered TN and TP. Agriculture is
currently the single largest contributor of Chesapeake Bay nutrient pollution, representing
39% and 49% of its N and P loadings (Sims and Coale, 2002). Furthermore, since World
War II, the geographic intensification of the use of commercial chemical fertilizers and
animal agriculture within regions, such as the lower Susquehanna Basin in southeastern
PA the eastern shore of MD, and DE, have increased agricultural nutrient runoff to the
Chesapeake Bay by substantially increasing N and P available for non-point source
runoff to streams. Thus, as a result of the predicted conversion of agricultural (crop and
pasture) land and the lower estimated rates of N and P fertilizer loadings to be applied to
these remaining lands by 2030, the projected overall reduction in TN and TP seen here
result from smaller quantities of fertilizer loadings delivered to Chesapeake Bay streams.
Lower applications of fertilizers represented smaller loadings of eroded organic and
dissolved inorganic N and P species available on the land surface and in the shallow
subsurface for non-point source delivery to the Chesapeake Bay when transported to
streams via overland, shallow interflow, and even baseflow runoff processes.
79
With the recent adoption of agricultural best management practices (BMPs), such
as conservation-tillage and off-season (winter) cover crop conservation programs, total
applications of commercial fertilizer are expected to decline (Sims and Coale, 2002).
Applications of manure are expected to increase or stay similar to 2000 quantities to
provide for crop nutrient needs. Both of these trends are incorporated in the projections
of total fertilizer and manure loadings applied for 2030 (Table 3.3) and help explain the
overall, estimated 2030 nutrient reduction trends.
Watershed-wide agricultural land was predicted to decrease from approximately
25% in 2000 to 22% in 2030, with the greatest decreases (> 9%) predicted to occur in the
most intensely farmed catchments of the lower Susquehanna Basin of southeastern PA,
the eastern shore of MD, and DE. This finding has great significance since these
catchments were also found in the results for 2000 to produce disproportionately the
highest TN (> 18 kg/ha/yr) and TP (> 0.99 kg/ha/yr) delivered yields to the Chesapeake
Bay (Roberts and Prince, 2010). This was mainly a product of the substantial,
agricultural, non-point source losses associated with the highest applications rates of
fertilizer and manure that occurred throughout the watershed. However, by 2030, the
mean delivered yield from these highest producing catchments in 2000 was projected to
decrease nearly 11% and 1% for TN and TP, respectively. The declines in mean TN and
TP yield seen in these 2000 highest producing catchments were correlated with the
anticipated, substantial decreases in applied fertilizer N and P.
The results reported here are similar to those of other studies of the effects of
development on future nutrient loadings in smaller watersheds. The increase in TN of
between 0.13-0.21% for the Saint Louis, Missouri region estimated by Wang et al. (2005)
80
from 2005-2030 was an effect of their projected extreme urbanization event. Similarly,
Tang et al. (2005) evaluated non-point source nutrient loading differences in north-
central Michigan from 1978-2040 with predicted urbanization and determined that after
development, TN and TP losses would also only slightly increase (< 3%). As in the
previous study, Tang et al. (2005) projected an extreme increase in urban land from 4.2 to
11.5%, nearly 300%, as compared to <200% in the present study. Thus, in both these
studies, the small increases on TN and TP runoff were obtained with overwhelming gains
in impervious-based runoff to streams. Without these very high changes in
imperviousness, the reduction in non-point losses of fertilizer and manure would of lead
to declines in TN and TP loadings. Notwithstanding, even though there were gains in
non-point N and/or P due to urbanization, much greater nutrient increases were limited by
the conversion of agricultural lands that have higher nutrient contribution than urban uses
(Tang et al., 2005; Wang et al., 2005).
Bhaduri et al. (2000) utilized a land use simulator to estimate impervious surface
growth near Indianapolis, Indiana from 1973-1991 and found that, over this period, an
18% increase in impervious areas resulted in a 15% decrease in non-point source N and P
loadings. The decline in non-point N and P loadings was directly attributed to losses of
agricultural lands. Furthermore, in Miami, Florida, Tsihrintzis et al. (1996) found that a
specific agricultural land area would have significant non-point source N and P
reductions of 54% and 35%, respectively, if it was entirely converted to urban use. In
both of these studies, these reductions were primarily due to the decreases in fertilizer
use.
81
By substantially lowering the amount of estimated, delivered fertilizers and
manure through land conversion in those catchments that contributed the most TN and
TP, such as those in the lower Susquehanna Basin of southeastern PA, a greater
proportional effect on Chesapeake Bay water quality than elsewhere can be expected. In
the Susquehanna Basin alone, TN and TP fell by about 33 and 30%, respectively. This
finding is quite significant since the Susquehanna Basin contributes about 50% of water
that enters the Chesapeake Bay annually (Susquehanna River Basin Commission, 2008),
thus even moderate percentage reductions in loadings would significantly reduce TN and
TP entering the estuary. Chang (2004) used a land use change model of proposed
development from the late 1990s to 2030 in several lower Susquehanna Basin catchments
and also found that agricultural land conversion to urban uses decreased overall, non-
point source P loadings. These trends in smaller parts of the watershed support the
findings reported here, that, for the entire watershed, the mean delivered yield will
decrease by about 10% and 7% for TN and TP from 2000 to 2030 (Table 3.4).
3.5.2 Forest (and other non-agricultural and non-urban) land losses and reductions
in non-point loadings
Conversion of forest and other non-agricultural and non-urban land areas to
development was also found to correlate with the predicted reduction in TP by 2030. P
export from these primarily forested regions of the watershed was shown by the yield of
area-weighted mean non-agricultural/non-urban patches that indicated larger patches
have greater influences on non-point source generation (Roberts and Prince, 2010). By
2030, these large, contiguous patches that contributed dissolved inorganic P to streams
were predicted to be substantially smaller as a result of development that stopped these
infiltration-based P runoff processes. This was seen in the averaged area-weighted mean
82
non-agricultural/non-urban patch size that decreased nearly 64% from 2000 to 2030
(Table 3.3), in step with a 67% reduction in its mean delivered yield (Table 3.4), and the
over 19% decline in overall TP delivered to the Chesapeake Bay (Table 3.4). The
catchments with the largest losses in area weighted mean non-agricultural/non-urban
patch size (>36,000 ha) were predicted to occur throughout the watershed (Figure 3.4a).
Catchments with the highest decreases in delivered P yield (> 0.27 kg/ha/yr) from this
source were predicted to occur only near central VA (Figure 3.5).
2030-2000 Mean NA/NU Patch Size (ha)
< -36,000
-36,000 - -32,000
-32,000 - -28,000
-28,000 - -24,000
-24,000 - -20,000
-20,000 - -16,000
-16,000 - -12,000
-12,000 - -8,000
-8,000 - -4,000
-4,000 - 0
a)
?
0 40 80 120 16020
Kilometers
2030-2000 Mean Urban Patch Size (ha)
< -1,300
-1,300 - -975
-975 - -650
-650 - -325
-325 - 0
0 - 325
325 - 650
650 - 975
975 - 1,300
> 1,300
b)
?
0 40 80 120 16020
Kilometers
Figure 3.4: Per catchment 2030-2000 difference maps of the area-weighted mean: a)
Non-agricultural/non-urban (NA/NU) and b) Urban (? 10% ISA) patch size source
metrics in ha.
83
2030-2000 Delivery of Mean NA/NU P Yield (kg/ha/yr)
< -0.27
-0.27 - -0.24
-0.24 - -0.21
-0.21 - -0.18
-0.18 - -0.15
-0.15 - -0.12
-0.12 - -0.09
-0.09 - -0.06
-0.06 - -0.03
-0.03 - 0.00
?
0 40 80 120 16020
Kilometers
Figure 3.5: Per catchment estimated 2030-2000 difference map of the NA/NU patch size
yield in kg/ha/yr for P delivered to the Chesapeake Bay.
3.5.3 Impervious surface area gains and changes in urban non-point loadings
As a direct result of increases in urban non-point source loadings, TN was
projected to increase in the range of 1-33% from 2000 to 2030 in the James, Patuxent,
and Rappahannock Basins. This result is similar to that of Costanza et al. (2002) for the
Patuxent, who compared mean delivered TN concentration to the estuary in 1997 with a
future "buildout" scenario and found an overall 14% gain. The catchments with the
greatest increases in non-point urban N yields (> 2.00 kg/ha/yr) were near: Baltimore,
Cumberland, and Frederick (MD); northeastern WV; DC; and Richmond, Norfolk-
Virginia Beach-Newport News, and Lynchburg (VA) (Figure 3.6a). Catchments with the
largest increases in non-point urban P yields (> 0.08 kg/ha/yr) were projected to occur in
the same regions as urban N with the exceptions of: Cumberland and Frederick (MD) and
northwestern WV (Figure 3.6b). Catchments with the greatest decreases in non-point N
84
and P yields (> 2.00 and > 0.08 kg/ha/yr, respectively) were projected to occur only near
DC (Figure 3.6b).
2030-2000 Delivery of Mean Urban N Yield (kg/ha/yr)
< -2.0
-2.0 - -1.5
-1.5 - -1.0
-1.0 - -0.5
-0.5 - 0.0
0.0 - 0.5
0.5 - 1.0
1.0 - 1.5
1.5 - 2.0
> 2.0
a)
?
0 40 80 120 16020
Kilometers
2030-2000 Delivery of Mean Urban P Yield (kg/ha/yr)
< -0.08
-0.08 - -0.06
-0.06 - -0.04
-0.04 - -0.02
-0.02 - 0.00
0.00 - 0.02
0.02 - 0.04
0.04 - 0.06
0.06 - 0.08
> 0.08
b)
?
0 40 80 120 16020
Kilometers
Figure 3.6: Per catchment estimated 2030-2000 difference maps of the area-weighted
mean urban (? 10% ISA) patch size yield in kg/ha/yr for: a) N and b) P delivered to the
Chesapeake Bay.
Surprisingly however, from 2000 to 2030, the mean non-point urban N and P
yields delivered to the Chesapeake Bay for all 2,339 catchments were projected to remain
virtually the same, decreasing and increasing by only 0.01 kg/ha/yr, respectively (Table
3.4). Non-point urban yields were modeled using area-weighted mean urban (? 10%
ISA) patches so that larger patches of development, as opposed to the same area in
smaller patches, had greater influences on non-point N and P generation (Roberts and
Prince, 2010).
85
The results of the simulation of non-point N and P yields to streams by 2030 are
counter-intuitive since the increase in urbanization might have been expected to increase
the runoff of nutrients. The expectation of an increase in pollution by urbanization is also
suggested by the SLEUTH projections that non-urban LC/LUs will be converted to
development from 7% in 2000 to 13% in 2030. The expectation is yet further reinforced
since the averaged area-weighted mean urban (? 10% ISA) patch size for all 2,339
catchments was predicted to increase by 67% from 2000 to 2030 (Table 3.3).
Catchments with the largest increase in area weighted mean urban (? 10% ISA) patch
size (>1,300 ha) were predicted to occur near: Harrisburg (PA); Baltimore and
Hagerstown (MD); DC; and Richmond, Norfolk-Virginia Beach-Newport News, and
Lynchburg (VA) (Figure 3.4b). Catchments with the highest decreases in this patch size
(>1,300 ha) were predicted only near DC (Figure 3.4b).
However, TN and TP loadings in streams and reservoirs are attenuated by
processes that include denitrification under anaerobic conditions, biological uptake by
stream organisms, and sedimentation onto stream and reservoir floors (Alexander et al.,
2000). Thus, any increase in non-point N and P discharge from impervious surfaces may
be diminished by cumulative downstream water attenuation processes. The greatest
increases in imperviousness were predicted to occur in catchments within the small and
intermediate stream categories that were estimated to attenuate the highest percentages of
instream N loadings throughout the watershed. In the case of P, however, no attenuation
occurs in small streams and may explain the overall slight increase in mean delivered P
yield projected to the Chesapeake Bay by 2030.
86
Filoso et al. (2004) found that after the projected conversion of 44% forested land
to urban uses from 1991 to 2101 in the Ipswich Basin of Massachusetts, gains in
ammonium-N concentrations were trivial (0.2-0.5 ?M). The lack of effect on N could
have been a result of ammonia volatilization and/or ammonium sorption to sediments
within the stream channel (Filoso et al., 2004). Thus, the expected increases in non-point
N, as a result of urbanization, were reduced in part due to significant stream attenuation.
Only in catchments with unusually high rates of estimated urbanization were the
projected non-point urban N gains large enough to overwhelm downstream attenuation.
Thus, expected substantial gains in mean non-point urban yield delivered to the
Chesapeake Bay from 2000 to 2030 did not occur.
3.5.4 Changes in point sources with urbanization
Gains in population and urban land throughout the watershed must lead to an
increase in point source loadings, as well as changes in non-point sources discussed
above. Using the models, it was shown that projected increases in delivered point source
loadings from 2000 to 2030 would offset the TN and TP loading losses to the Chesapeake
Bay caused by non-urban, non-point source reductions. The delivered N and P point
loadings to the estuary were projected to increase over 18 and 20%, respectively, between
2000 and 2030 (Table 3.4). Similarly to the non-point loadings, the only attenuation
processes reducing point N and P loadings delivered to the estuary occurred in streams
and reservoirs presumably via denitrification, biological uptake, sedimentation, etc.
The negation of the reductions caused by reduction in agriculture, as a result of
increases in point sources, must be qualified by recognition of the uncertainties in
estimation of future population size and geographical distribution. Furthermore, the
87
projected 2030 discharge additions were based on WWTP estimates of domestic effluent
discharge that do not take into account gains or losses from industrial or commercial
point sources. Neither do these loadings take into account the future locations of
WWTPs, nor any future advances in effluent removal technology that can be expected to
decrease these loadings substantially. All of these are limitations of the point loadings
with the result that they cannot be quantified with great accuracy.
3.5.5 Land-to-water delivery losses and reductions in non-urban, non-point loadings
The results indicated that reductions in land-to-water delivery variables, resulting
from the conversion of non-urban LC/LUs to development, also contributed to the
decreases in delivered TN and TP to the Chesapeake Bay. Roberts and Prince (2010)
showed that several landscape metrics were significantly related to increased non-point
delivery to Chesapeake Bay streams. For N these were: percentage of extractive land,
percentage of cropland, area-weighted mean edge contrast of deciduous forest, and
percentage of evergreen forest within the riparian stream buffer and, for P, percentage of
barren land within the riparian stream buffer.
The significance of these findings is that changes in the spatial composition and
configuration of LC/LU can be expected to provide a means of reduction in land-to-water
delivery. Increases in non-point N and P transport by landscape compositional and
configuration changes might be caused by processes such as reduced soil hydraulic
conductivity properties and increases in overland and shallow subsurface flow paths
(Roberts and Prince, 2010). Overland and shallow subsurface flows are favored by
compacted or saturated non-urban surfaces often found associated with extractive lands.
Croplands, evergreen forests, and barren lands have been shown to have this effect.
88
Stormflow in some Chesapeake Bay tributaries has been suggested to provide pathways
for non-point N to reach streams, allowing surface flow to traverse forested riparian
buffers (Norton and Fisher, 2000). Changes of LC/LU associated with development may
eliminate some surfaces with limited hydraulic conductivity, such as barren land, and
contribute to reduction of non-point N and P transport.
The catchments in which significant changes in factors that affect surface and
shallow subsurface flow were not evenly distributed across the Chesapeake Bay
watershed. The largest decreases in percentage of extractive land (> 0.27%) were
predicted in: central and east-central PA; western and central MD; northeastern WV; DC;
and northern VA (Figure 3.7a). Catchments with the greatest decreases in percentage of
cropland (> 9%) were predicted in: southern PA; the eastern shore of MD; and DE
(Figure 3.7b). The catchments with the greatest decreases in area-weighted mean edge
contrast of deciduous forest (> 2.0%) were predicted throughout the watershed, whereas
the greatest increases in area-weighted mean edge contrasts of deciduous forest (> 2.0%)
were predicted near: east-central MD and Norfolk-Virginia Beach-Newport News (VA)
(Figure 3.7c). Catchments with the largest decreases in percentage of evergreen forest
within the riparian stream buffer (> 1.35%) were predicted watershed-wide (Figure 3.7d).
Finally, the catchments with the greatest decreases in the percentage of barren land within
the riparian stream buffer (> 0.45%) were predicted near: Lancaster (PA); Baltimore and
Frederick (MD); DC; and Norfolk-Virginia Beach-Newport News (VA) (Figure 3.8).
89
2030-2000 Extractive Land (%)
< -0.27
-0.27 - -0.24
-0.24 - -0.21
-0.21 - -0.18
-0.18 - -0.15
-0.15 - -0.12
-0.12 - -0.09
-0.09 - -0.06
-0.06 - -0.03
-0.03 - 0.00
a)
?
0 40 80 120 16020
Kilometers
2030-2000 Cropland (%)
< -9
-9 - -8
-8 - -7
-7 - -6
-6 - -5
-5 - -4
-4 - -3
-3 - -2
-2 - -1
-1 - 0
b)
?
0 40 80 120 16020
Kilometers
2030-2000 Mean Edge Contrast of Deciduous Forest (%)
< -2.0
-2.0 - -1.5
-1.5 - -1.0
-1.0 - -0.5
-0.5 - 0.0
0.0 - 0.5
0.5 - 1.0
1.0 - 1.5
1.5 - 2.0
> 2.0
c)
?
0 40 80 120 16020
Kilometers
2030-2000 Evergreen Forest in Buffer (%)
< -1.35
-1.35 - -1.20
-1.20 - -1.05
-1.05 - -0.90
-0.90 - -0.75
-0.75 - -0.60
-0.60 - -0.45
-0.45 - -0.30
-0.30 - -0.15
-0.15 - 0.00
d)
?
0 40 80 120 16020
Kilometers
Figure 3.7: Per catchment 2030-2000 difference maps of the: a) Percentage of extractive
land, b) Area-weighted mean edge contrast of deciduous forest land (%), c) Percentage of
cropland, and d) Percentage of evergreen forest land in the 31 m riparian stream buffer
land-to-water delivery metrics for the RESAC TN model.
90
2030-2000 Barren Land in Buffer (%)
< -0.45
-0.45 - -0.40
-0.40 - -0.35
-0.35 - -0.30
-0.30 - -0.25
-0.25 - -0.20
-0.20 - -0.15
-0.15 - -0.10
-0.10 - -0.05
-0.05 - 0.00
?
0 40 80 120 16020
Kilometers
Figure 3.8: Per catchment 2030-2000 difference map of the percentage of barren land in
the 31 m riparian stream buffer land-to-water delivery metric for the RESAC TP model.
3.6 Conclusions
Previously, the quantification of the impacts of projected future urbanization on
nutrient loading estimates has been limited to smaller watersheds and impacts on
regional, national, and even global nutrient loadings have been deduced from the results
of local studies. To substantiate the expectations based on the small catchments,
substantially larger watersheds regions that drain into large estuaries and coastal oceans
need to be examined from the detailed small catchment scale aggregated to the larger
catchments in which they occur. The present study of the potential future sources and
transport of TN and TP using projections of urbanization in the Chesapeake Bay
watershed is an attempt to undertake such an assessment. The effects of LC/LU change
from 2000 to 2030 in the Chesapeake Bay watershed, in particular as a result of forecast
91
population increases and consequent increases in urbanization, was modeled and the
effects on nutrient loadings to the estuary assessed.
There was an estimated 19% and 20% reduction in overall delivered TN and TP
to the Chesapeake Bay. Although substantial increases in development-induced, point
source N and P loadings were apparent watershed-wide, the estimated conversion of
agricultural lands leading to declines in delivered fertilizer loadings to streams was the
primary reason for the overall reductions in TN and TP delivery to the estuary that were
simulated to occur from 2000 to 2030. In contrast to the non-point source changes, the
projected increases in point source N and P loadings are necessarily imprecise because
future improvements in effluent removal technologies, future WWTP locations that could
alter their watershed-wide distribution, and possible gains or losses in industrial and
commercial sources cannot be predicted. Increases in impervious surfaces associated
with urbanization that would otherwise have increased the mean, delivered non-point
urban N and P yields for all catchments from 2000 to 2030 were negated due to
downstream water attenuation processes decreasing delivered TN and TP to the
Chesapeake Bay from all sources.
The relative magnitude of TN and TP contributions by point sources, fertilizers,
and other non-point sources to future nutrient loadings depends on the land cover mosiacs
of a watershed (Anbumozhi et al., 2005), as well as total area. The results suggest that
lowering area-weighted mean patch sizes within catchments of significant non-point
LC/LU sources could reduce future TN and TP loadings to the Chesapeake Bay. This is
especially true of mean urban patch sizes, where decreases in its total area would be
likely to limit its yields by reducing larger impervious surface areas capable of capturing
92
greater quantities of non-point N and P that could be delivered to streams. In addition,
limiting urban growth to the replacement of agricultural source lands and other non-urban
cover types associated with large land-to-water delivery of non-point N and P to
Chesapeake Bay streams would be particularly effective.
Thus, to minimize projected TN and TP loadings to the Chesapeake Bay, the
estimated, spatial distribution of catchment and riparian stream buffer-wide LC/LU
should be examined and evaluated in the future prior to development. This would allow
for the impacts of the compositional and configurational landscape properties
demonstrated here to be quantified and taken into account before any potential
management actions are considered.
93
Chapter 4: Effects of future urban sprawl on nitrogen runoff in
subwatersheds of Chesapeake Bay3
4.1 Abstract
The effects of sprawl on non-point N runoff from the Chesapeake Bay watershed
was studied using the SPAtially Referenced Regressions On Watershed Attributes
(SPARROW) model, a nutrient runoff model, applied to maps of impervious surface in
2000 and modeled impervious surface in 2030. Projected catchment-wide losses and
gains in TN runoff were calculated for 2,339 catchments. Results suggested leapfrog
growth that was predicted to occur in smaller, detached patches on crop and pasture will
lower catchment total nitrogen (TN) runoff from 2000 to 2030. This was in contrast to
gains in TN runoff that were predicted to occur with infill and peripheral growth,
predominantly in cover types other than crop and pasture. Furthermore, development in
smaller, detached patches would also reduce urban, non-point N runoff in catchments
over this time period. These results suggest that the strategic placement of leapfrog
development may be capable of reducing future TN runoff to the Chesapeake Bay and in
similar watersheds elsewhere.
3 The material in Chapter 4 is currently in review. Roberts, A. D., Prince, S. D., Jantz, C. A., and Goetz, S.
J., Submitted for publication. Effects of future urban sprawl on nitrogen runoff in subwatersheds of
Chesapeake Bay. Ecological Engineering.
94
4.2 Introduction
Total nitrogen (TN) runoff from the land is the primary causative factor in the
formation of eutrophic, hypoxic, and anoxic conditions in estuaries and coastal oceans
(Paerl, 2006). Thus, any land cover and land use (LC/LU) change through urbanization,
including the form referred to as sprawl, that converts non-urban cover types to
development will most certainly alter nitrogen (N) runoff to streams (Tang et al., 2003).
According to Gilliam (2002), urban sprawl can be defined as a form of urbanization
distinguished by leapfrog patterns of development, commercial strip development, low-
density land uses, large expanses of separated single land-uses, automobile dominance,
and a minimum of public open space. LC/LU change from non-urban to urban has been
shown to increase N runoff significantly (Carpenter et al., 1998). Long-term water
quantity is reduced by sprawl through the channeling of rainwater to drains and the
reduction in the volume of water that infiltrates, percolates, and recharges soil profiles
and aquifers, although water quality only generally declines (Tu et al., 2007).
Urban sprawl can lead to an increase in TN caused by increases in impervious
cover that allow for the buildup and consequential washoff of N to sewers from urban
non-point sources that include vehicle emissions, lawn fertilizers, and pet wastes
(Babbitt, 2007). Sprawl can also lead to TN gains through increases in point source
discharges associated with increased N release from wastewater treatment plants
(WWTPs) because of population gains. As the population expands and moves from
centralized, higher-density, developed areas to form lower-density, suburban areas, the
effects of sprawl may decrease TN if growth is arranged to replace higher yielding (mass
for a specified time normalized by drainage area) cover types, such as cropland or
95
pasture. In contrast, sprawl effects may increase TN if growth is configured to replace
lower yielding cover types, such as forest or wetlands. Thus, the impacts of the spatial
patterns of future sprawl on projected TN loadings (mass for a specified time) in rivers,
estuaries, and coastal oceans are of interest.
The correlation between urban sprawl and reduced water quality is not universal.
While nitrate (NO3-) and TN have been found to be positively correlated to urban and
agricultural area in eastern Massachusetts watersheds (Williams et al., 2005; Tu et al.,
2007), the Waquoit Bay watersheds surrounding Cape Cod, Massachusetts (Bowen and
Valiela, 2001), and several South Carolina coastal streams (Tufford et al., 2003), other
dissolved organic nitrogen (DON) species and ammonia (NH3) were negatively
correlated to sprawl in these same eastern Massachusetts and South Carolina coastal
stream studies.
The effects of anticipated sprawl on TN in smaller watersheds in Michigan (Tang
et al., 2005), Missouri (Wang et al., 2005), Sweden (Jansson and Colding, 2007), and
Georgia (Dale et al., 2008) have all indicated an increase in loadings over time.
However, it is not clear in large, nested basins if future sprawl could produce losses or
gains in projected TN loadings, based upon where growth may occur in the watershed.
Although briefly examined by Roberts et al., (2009), this has not been studied
exclusively, yet it may have significant management implications.
Here, the Chesapeake Bay watershed, a large, nested basin, was examined to
determine which characteristics associated with future sprawl could affect TN loading
losses or gains. The Chesapeake Bay is the largest estuarine system in the United States
(166,534 km2 watershed) draining portions of six states (New York (NY), Pennsylvania
96
(PA), Delaware (DE), Maryland (MD), West Virginia (WV), and Virginia (VA)) and the
District of Columbia (DC) that include numerous urban centers (Figure 4.1a-b). The
substantial population increase and intensification of agriculture leading to increases in
application of chemical fertilizer after the end of World War II has increased TN loadings
to the detriment of the overall nutrient health of the estuary (Sims and Coale, 2002). As
agricultural non-point runoff, N-based fertilizers can be transported to streams in
enriched sediments and DON via surface overland flow, shallow subsurface interflow,
and subsurface percolation (Pellerin et al., 2006). Additionally, a significant portion of
these TN loadings are generated from impervious cover-based non-point runoff and point
discharges from wastewater treatment plants (WWTP), industrial, and commercial
sources directly into Chesapeake Bay streams.
PA
VA
NY
WV
MD NJ
NC
OH
DE
DC
JAM E
S R
SUS
QUEHAN N A R
PO
TO
MA
C R
RA
PP
AH
ANN
OCK
R
PA
TU
XE
NT
R
MA
TTA
PO
NI RJA
ME
S R
SU
SQ
UE
HA
NN
A R
PAT
UX
EN
T R
PAM
UNK
EY
R
a)
?
0 40 80 120 16020
Kilometers
Baltimore, MD
Washington, DC--MD--VA
Richmond, VA
Harrisburg, PA
Norfolk--Virginia Beach--Newport Ne
Scranton--Wilkes-Barre, PA
York, PALancaster, PA
Lynchburg, VA
Binghamton, NY
Petersburg, VA
Cumberland, MD--WV
Elmira, NY
Altoona, PA
Frederick, MD
Fredericksburg, VA
Annapolis, MD
Hagerstown, MD--PA--WV
Charlottesville, VA
Williamsport, PA
State College, PAPA
VA
NY
WV
NJ
NC
OH
DEMD
b)
?
0 40 80 120 16020
Kilometers
Figure 4.1: The Chesapeake Bay watershed showing the locations of: a) Streams and
rivers draining the estuary and b) Urban centers located within its boundaries.
97
Since the early 1980s, federal, regional, and state entities have made significant
efforts in research and policy initiatives aimed at restoring the estuary to pre-World War
II conditions. These include: the United States Environmental Protection Agency
(USEPA), Chesapeake Bay Program (CBP), National Oceanic and Atmospheric
Administration (NOAA), United States Geological Survey (USGS), states boarding the
Chesapeake Bay (MD and VA) and DC, and various other local jurisdictions in the
watershed. The estuary is N-limited, with a nutrient input ratio of TN to total phosphorus
(TP) well above the Redfield ratio of 16:1 for the normal production of phytoplankton
(Weller, 2003). This is caused by an excess of available N rather than deficiencies of TP
(Weller, 2003). Thus, a comparison of subcatchments with high and low TN runoff is
relevant to understanding and management of TN loadings to the Chesapeake Bay.
Anticipated urban growth to 2030 has shown that expected sprawl will not be
spatially-uniform throughout the watershed (Jantz et al., 2004, in press). Simulations of
future, urban land cover derived from the Slope, Land use, Exclusion, Urban extent,
Transportation, and Hillshade (SLEUTH) urban growth model (Jantz et al., 2004, in
press) at 30 m resolution show a continued trend from the last several decades of
significant increases in developed land that is detached and distributed further away from
existing and new primary urban centers, that is, sprawl. With sprawl predicted to occur
across non-urban lands of varying N generation and transport abilities, its spatial
arrangement is quite likely to be an important indicator of projected TN.
98
PA
VA
NY
WV
MD NJ
NC
OH
DE
DC
?
0 40 80 120 16020
Kilometers
a)
Example Catchment
Runoff from Catchment
Enters Stream
31 m Buffer
Upstream Catchments
Carrying N Loads
Upstream Plus Locally Generated
Runoff, Less Instream Losses
Leaves Catchment
b)
?
0 1
Kilometer
i
i
i
ii
i
Figure 4.2: 2,339 catchments used in the: a) 2000 RESAC 31 m SPARROW
Chesapeake Bay TN model with b) Schematic diagram of an example catchment area
showing fixed 31 m riparian stream buffer width area surrounding stream reach and
corresponding upstream and downstream runoff paths.
Using watershed-wide 2000 non-urban and urban cover types derived from the
Regional Earth Science Application Center?s (RESAC) LC/LU and percent impervious
surface area (% ISA) maps (Goetz et al., 2004a, b; Jantz et al., 2005), Roberts and Prince
(2010) modeled current TN loadings to the Chesapeake Bay. The model was a
recalibrated version of the 1997 USGS Version 3.0 SPAtially Referenced Regressions
On Watershed Attributes (SPARROW) model (Brakebill and Preston, 2004). 2,339
catchments draining to the estuary (Figure 4.2a) were modeled. Unlike the Brakebill and
Preston (B & P) model, the RESAC version of SPARROW added compositional and
configurational landscape metrics. In addition, metrics were applied within a 31 m
riparian stream buffer (Figure 4.2b). These metrics were shown to have significant
99
effects on non-point sources and land-to-water delivery variables affecting TN loadings
to the estuary (Table 4.1). The recalibrated model was used to estimate the effects of the
anticipated expansion and its spatial distribution of development from 2000 to 2030.
Landscape metric variables Effect
Source Land-to-water delivery
Area-weighted mean urban (? 10% ISA) patch size X
Percentage of cropland X
Percentage of extractive land X
Area-weighted mean edge contrast of deciduous forest X
Percentage of evergreen forest In the riparian stream buffer X
Table 4.1: Landscape metrics that were found to be significant (p value ? 0.05) in the
2000 RESAC 31 meter SPARROW model.
Of the 5 landscape metrics tested by Roberts and Prince (2010), area-weighted
mean urban (? 10% ISA) patch size was found to have the strongest relationship to TN
runoff from 2000 to 2030. The term ? 10% ISA described watershed-wide urbanization
as any LANDSAT pixel with the proportion of its 30 m x 30 m (900 m2) area greater than
or equal to 10% impervious surface area in 2000 (Goetz et al., 2004a, b; Jantz et al.,
2005). Projected urban areas by SLEUTH model runs (Jantz et al., in press) were then
used as surrogates to represent 2000 urban areas derived from this ? 10% ISA threshold
for 2030 estimates. The area-weighted mean urban (? 10% ISA) patch size metric
measures one aspect of landscape composition and quantifies the sum, across all patches
of urban, of patch area multiplied by proportional abundance of the patch. Thus, larger
patches exert greater influence than smaller patches. The calibrated 2000 model found
that per ha of area-weighted mean urban (? 10% ISA) patch size increased non-point N
by about 25 kg to the Chesapeake Bay annually (Roberts and Prince, 2010).
By 2030, if growth occurs by creation of smaller patches of development
unattached to existing infrastructure or added to smaller patches of urban land, area-
100
weighted mean urban (? 10% ISA) patch size will decrease from 2000 values. With
potential losses in urban non-point N yields, TN losses would also be possible. However,
reductions in TN loadings would depend upon the current LC/LU of non-urban lands to
be converted, since they have different impacts on catchment TN runoff.
In contrast, 2030 growth will increase discharged point sources that could cause
increases in TN runoff in catchments. If gains in point sources are greater than non-point
N losses, overall increases in TN would occur. Therefore, discharges of point sources for
2030 were estimated from projected gains in urban population (Roberts et al., 2009).
Thus, the overall purpose of this study was to compare the characteristics of
future sprawling catchments that are projected to have losses with those expected to gain
in TN loadings.
4.3 Materials and methods
4.3.1 Categories of sprawling catchments
Between 2000 and 2030, the RESAC 31 m SPARROW model (Roberts and
Prince, 2010) indicated that 157 catchments out of 2,339 would decrease in area-
weighted mean urban (? 10% ISA) patch size (Figure 4.3a). Here, the 157 catchments
were divided into those in which the comparison of 2000 with 2030 ISA had an increase
with those that had a decrease in locally generated TN yield. There were 106 catchments
that were modeled to decreased (Figure 4.3b) and 51 catchments (Figure 4.3c) to increase
in locally generated TN yield. In the SPARROW model, local generation of TN
indicated the yields and loadings generated from within a catchment independent of any
upstream loadings.
101
?
0 40 80 120 16020
Kilometers
a)
?
0 40 80 120 16020
Kilometers
b)
?
0 40 80 120 16020
Kilometers
c)
Figure 4.3: Map of the Chesapeake Bay watershed locations of the: a) 157 catchments
showing decreases in area-weighted mean urban (? 10% ISA) patch size with b) 106
catchments indicating losses and c) 51 catchments indicating gains in locally generated
TN yield from 2000 to 2030.
102
In addition to the area-weighted mean urban (? 10% ISA) patch size, ten other
variables (Table 4.2) that were significant in the 2000 model (Roberts and Prince, 2010)
were analyzed in the two categories of catchments. The only other significant variable
that was not evaluated was an increase in atmospheric deposition of N from 2000 to
2030, owing to the difficulty in forecasting a variable that is largely determined by
policy, regulation, and legal changes. A complete description of these model variables
are given by Roberts and Prince (2010) and Roberts et al. (2009).
4.3.2 Results for the entire Chesapeake Bay watershed (Roberts and Prince, 2010)
For each kilogram (kg) of: point sources 1) discharged, 2) fertilizer and 3) manure
applied to agricultural land (crop and pasture) and atmospheric N deposited on the
watershed, approximately 1.2, 0.2, 0.1, and 0.5 kg of N, respectively, were estimated in
Chesapeake Bay streams. For every percent (%) of: 4) land on the coastal plain, non-
point N delivered to streams was estimated to decrease by 0.73%. In contrast, for each %
of: 5) extractive, 6) cropland, and 7) area-weighted mean edge contrast of deciduous
forest, non-point N delivered to streams was estimated to increase by 0.27, 0.021, and
0.014%, respectively. Area-weighted mean edge contrast of deciduous forest was used to
measure the contrast between eleven non-urban classes and deciduous forest in the 2000
RESAC LC/LU map. Contrast here was defined as the physical characteristics of
differing non-urban cover types influencing N transport. Greatest contrast differences for
deciduous forest were with: urban/residential/recreational grasses, extractive, barren,
pasture/hay, croplands, and natural grass. For each % of: 8) evergreen forest in the
riparian stream buffer, non-point N delivered to streams was estimated to increase by
0.013%. For every meter (m) traveled in small (mean flow ? 100 cubic feet per second
103
(ft3/s)), intermediate (mean flow >100 and ? 500 ft3/s), and large (mean flow > 500 ft3/s)
streams per day, about 25, 9, and 3% of the instream N was estimated to be lost to
processes, such as denitrification, biological uptake, and sedimentation, respectively.
Thus, each catchment has only one: 9) mean flow value used to determine the estimated
% of N lost based upon the categories just outlined. Finally, for any: 10) reservoir in the
watershed, N was removed from the water column at an average settling velocity of over
14 m/yr.
Model component
(units)
Variable Coefficient
value
Nitrogen source
(1 = kg/yr, 2 = kg/ha/yr)
Point sources1 1.173
Applied fertilizer1 0.175
Atmospheric
deposition1
0.492
Applied manure1 0.078
Area-weighted mean
urban (? 10% ISA) patch
size2
24.885
Landscape delivery
(%)
Percentage of coastal
plain
-0.729
Percentage of extractive
land
0.270
Area-weighted mean
edge contrast of
deciduous forest
0.014
Percentage of cropland 0.021
Percentage of evergreen
forest in the riparian
stream buffer
0.013
Stream decay
(m/day)
Small streams 0.249
Intermediate streams 0.090
Large Streams 0.030
Reservoir decay
(m/yr)
Reservoir 14.224
Table 4.2: Variables in the 2000 RESAC 31 m TN SPARROW model found significant
(p value ? 0.05).
104
4.3.3 Yields used for 2030 TN runoff projections
For 2030 estimates, fertilizer rates for cropland and pasture were taken to be
28.02 and 15.83 kg/yr, respectively, and manure at 9.79 and 52.74 kg/yr (Roberts et al.,
2009). Furthermore, for each person added to the urban population, it was estimated that
an additional 2.72 kg/yr of N would be added to streams from the nearest WWTP
servicing these areas in 2000 via increased point discharges (Roberts et al., 2009).
4.3.4 Sprawling catchments and parametric testing
Normality evaluations of: point source discharges; applied fertilizer; applied
manure; area-weighted mean urban (? 10% ISA) patch size; percentage of land on the
coastal plain; percentage of extractive land; area-weighted mean edge contrast of
deciduous forest; percentage of cropland; percentage of evergreen forest in the riparian
stream buffer; stream size; and reservoirs were conducted using the Shapiro-Wilk,
Anderson-Darling, Lillefors, and Jargue-Bera tests (Thode, 2002). After it was
determined that all datasets were non-normal and that the mean differences in these
variables could not be examined using parametric tests, median differences were
evaluated with the non-parametric Mann-Whitney U test (Kvam and Vidakovic, 2007).
Only significant (p value ? 0.05) median differences between each of the variables in the
two categories of catchments are reported.
4.4 Results
4.4.1 Categories of sprawling catchments
105
In catchments with projected TN yield losses from 2000 to 2030, the median
reduction was estimated at -1.10 kg/ha/yr. In the model, the highest decreases in
projected TN yield (> 1.8 kg/ha/yr) were predicted to occur near: DC; Baltimore and
Frederick (MD); Harrisburg, (PA); and Richmond (VA) (Figure 4. 4a, c). In these
catchments, the median decrease in the extent of area-weighted mean urban (? 10% ISA)
patch size was -2.89 ha. The largest decreases in area-weighted mean urban (? 10% ISA)
patch size (> 27 ha) were predicted to occur near: DC; Scranton/Wilkes-Barre,
Harrisburg, and Williamsport (PA); and Richmond (VA) (Figure 4. 4b, d).
In contrast, in catchments with projected TN yield gains, equivalent values were
+0.53 kg/ha/yr and -1.52 ha. The highest increases in projected TN yield (> 1.8 kg/ha/yr)
were predicted to occur near: DC; Hagerstown (MD); and Scranton/Wilkes-Barre,
Harrisburg, and Williamsport (PA) (Figure 4.5a, c). The largest decreases in area-
weighted mean urban (? 10% ISA) patch size (> 27 ha) will be located in the same
regions (Figure 4.5b, d).
106
2030-2000 Local Generation of TN Yield (kg/ha/yr)
< -1.8
-1.8 - -1.6
-1.6 - -1.4
-1.4 - -1.2
-1.2 - -1.0
-1.0 - -0.8
-0.8 - -0.6
-0.6 - -0.4
-0.4 - -0.2
-0.2 - 0.0
a)
?
0 40 80 120 16020
Kilometers
2030-2000 Mean Urban Patch Size (ha)
< -27
-27 - -24
-24 - -21
-21 - -18
-18 - -15
-15 - -12
-12 - -9
-9 - -6
-6 - -3
-3 - 0
b)
?
0 40 80 120 16020
Kilometers
c)  d)
2030-2000 Local Generation of TN Yield (kg/ha/yr)
12
7
39
10
2 5 4
12
5
10
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
Category
2030-2000 Mean Urban Patch Size (ha)
10
60
15
5
2 1 3 1 4
5
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
Category
Figure 4.4: Per catchment 2030-2000 change maps of: a) Losses in locally generated
TN yield (kg/ha/yr) with b) Decreases in the area-weighted mean urban (? 10% ISA)
patch size (ha), and c-d) Corresponding frequency distributions of the number of
catchments falling in each category. For Figure 4.4c, category classes are: 1 = < -1.8, 2
= -1.8 - -1.6, 3 = -1.6 - -1.4, 4 = -1.4 - -1.2, 5 = -1.2 - -1.0, 6 = -1.0 - -0.8, 7 = -0.8 - -0.6,
8 = -0.6 - -0.4, 9 = -0.4 - -0.2, and 10 = -0.2 - 0.0 kg/ha/yr. For Figure 4d, category
classes are 1 = < -27, 2 = -27 - -24, 3 = -24 - -21, 4 = -21 - -18, 5 = -18 - -15, 6 = -15 - -
12, 7 = -12 - -9, 8 = -9 - -6, 9 = -6 - -3, and 10 = -3 - 0 ha.
107
2030-2000 Local Generation of TN Yield (kg/ha/yr)
0.0 - 0.2
0.2 - 0.4
0.4 - 0.6
0.6 - 0.8
0.8 - 1.0
1.0 - 1.2
1.2 - 1.4
1.4 - 1.6
1.6 - 1.8
> 1.8
a)
?
0 40 80 120 16020
Kilometers
2030-2000 Mean Urban Patch Size (ha)
< -27
-27 - -24
-24 - -21
-21 - -18
-18 - -15
-15 - -12
-12 - -9
-9 - -6
-6 - -3
-3 - 0
b)
?
0 40 80 120 16020
Kilometers
c)  d)
2030-2000 Local Generation of TN Yield (kg/ha/yr)
34124
7
1
15
11
3
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
Category
2030-2000 Mean Urban Patch Size (ha)
12
26
4 4
0 0
3 1 1 0
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
Category
Figure 4.5: Per catchment 2030-2000 change maps of: a) Gains in locally generated TN
yield (kg/ha/yr) with b) Decreases in the area-weighted mean urban (? 10% ISA) patch
size (ha) and c-d) Corresponding frequency distributions of the number of catchments
falling in each category. For Figure 4.5c, category classes are: 1 = 0.0 -0.2, 2 = 0.2 -0.4, 3
= 0.4 - 0.6, 4 = 0.6 -0.8, 5 = 0.8 - 1.0, 6 = 1.0 - 1.2, 7 = 1.2 - 1.4, 8 = 1.4 - 1.6, 9 = 1.6 -
1.8, and 10 = > 1.8 kg/ha/yr. For Figure 4.5d, category classes are 1 = < -27, 2 = -27 - -
24, 3 = -24 - -21, 4 = -21 - -18, 5 = -18 - -15, 6 = -15 - -12, 7 = -12 - -9, 8 = -9 - -6, 9 = -6 -
-3, and 10 = -3 - 0 ha.
108
4.4.2 Sprawling catchments and SPARROW
Of the eleven variables tested, five were found to be significant (Table 4.3).
These were: area-weighted mean urban (? 10% ISA) patch size, applied fertilizer, applied
manure, percentage of coastal plain, and percentage of cropland.
Significant SPARROW variables Catchments with losses in local
generation of TN yield
Catchments with gains in local
generation of TN yield
Area-weighted mean urban (? 10%
ISA) patch size
-0.03 kg/ha/yr -0.01 kg/ha/yr
Applied fertilizer -37.48 kg/ha/yr -2.28 kg/ha/yr
* Percentage of land on coastal plain 87% 0%
Percentage of cropland -0.72% -0.25%
Applied manure +28.23 kg/ha/yr +31.70 kg/ha/yr
Table 4.3: Comparison of the significant (p value ? 0.05) median changes from 2000 to
2030 for the RESAC 31 m SPARROW variables in the catchments with losses and gains
in locally generated TN yield. * Denotes static hydrogeomorphic province variable that
does not vary from 2000 to 2030.
The median area-weighted mean urban (? 10% ISA) patch size N yield loss was
estimated to be -0.03 kg/ha/yr in catchments with projected TN yield losses (Table 4.3).
In contrast, the median area-weighted mean urban (? 10% ISA) patch size yield loss was
estimated to be only -0.01 kg/ha/yr in catchments with projected gains (Table 4.3). In
TN yield loss catchments, applied fertilizer yield decreased substantially (-37.48
kg/ha/yr), whereas applied manure yield increased by + 28.23 kg/ha/yr (Table 4.3). For
catchments with TN yield gains, applied fertilizer decreased only -2.28 kg/ha/yr, in
contrast to the sharp increase in applied manure yield of +31.70 kg/yr (Table 4.3). The
results of the changes in locally generated applied fertilizer and manure yield in the two
categories of catchments are given in Table 4.4.
109
Local generation Catchments with losses in local
generation of TN yield
Catchments with gains in local
generation of TN yield
Applied fertilizer -5.96 kg/ha/yr -0.30 kg/ha/yr
Applied manure +1.73 kg/ha/yr +2.90 kg/ha/yr
Table 4.4: Comparison of the median changes in locally generated applied fertilizer and
manure yield (kg/ha/yr) from 2000 to 2030 in catchments with losses and gains in locally
generated TN yield.
In catchments with projected TN yield losses, the median percentage on the
coastal plain was 87%, as compared to 0% in catchments with gains (Table 4. 3). The
median percentage of cropland decreased by -0.72% in catchments with projected losses,
but decreased by only -0.25% in catchments with gains (Table 4.3). Thus, the conversion
of cropland to urban was nearly 3:1 in catchments with projected losses.
4.5 Discussion
4.5.1 Area-weighted mean urban (? 10% ISA) patch size
The finding of a greater (-0.03 kg/ha/yr) yield loss attributed to urban patch size
in catchments with projected TN losses is interesting since this could be indicative of
variations in urban non-point N losses dependent upon where growth will occur.
Catchments with projected TN yield losses are predicted to have smaller patches of
growth on currently undeveloped land that is unattached to existing urban areas. This
type of sprawl is referred to as "leapfrog" growth. With the introduction of isolated
patches, urban non-point N loadings may fall catchment-wide in spite of an increase in
development, since these patches represent smaller areas and runoff volumes.
Additionally, smaller patches may help to disperse concentrated, urban non-point
sources, such as lawn fertilizer, pet waste, vehicles, etc; among more stream networks
110
leading to attenuation of loadings. Lawn fertilizer rates can be even higher and more
concentrated than rates applied to some crops (Bowen and Valiela, 2001). The highest
decreases in area-weighted mean urban (? 10% ISA) patch size yield (> 0.09 kg/ha/yr) in
catchments with projected TN yield losses were predicted to occur near: DC; Annapolis,
Baltimore, Frederick, and Hagerstown (MD); Scranton/Wilkes-Barre and Harrisburg
(PA); and Richmond (VA) (Figure 4.6a-b).
In contrast, the finding of patch size related yield loss (-0.01 kg/ha/yr) in
catchments projecting TN gains is interesting since it suggests that the spatial pattern of
future sprawl will occur around smaller areas of existing urban land. This type of sprawl
is referred to as "infill and peripheral" growth. Although this type of growth in
catchments with TN gains still led to a median decrease in patch size, there would be
smaller declines in urban non-point N runoff. In contrast to leapfrog growth, higher
concentrations of non-point N would reach these streams. The largest decreases in area-
weighted mean urban (? 10% ISA) patch size yield (> 0.09 kg/ha/yr) in catchments with
projected TN yield gains were predicted to occur near: DC; Hagerstown (MD); and
Harrisburg and Williamsport (PA) (Figure 4.7a-b).
111
2030-2000 Local Generation of Urban N Yield (kg/ha/yr)
< -0.09
-0.09 - -0.08
-0.08 - -0.07
-0.07 - -0.06
-0.06 - -0.05
-0.05 - -0.04
-0.04 - -0.03
-0.03 - -0.02
-0.02 - -0.01
-0.01 - 0.00
a)
?
0 40 80 120 16020
Kilometers
b)
2030-2000 Local Generation of Mean Urban N Yield (kg/ha/yr)
7
46
34
5
1 2 1 3
4 3
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
Category
Figure 4.6: Per catchment 2030-2000 change map of: a) Losses in locally generated
area-weighted mean urban (? 10% ISA) patch size yield (kg/ha/yr) with losses in locally
generated TN yield and b) Corresponding frequency distribution of the number of
catchments falling in each category. Category classes are: 1 = < -0.09, 2 = -0.09 - -0.08,
3 = -0.08 - -0.07, 4 = -0.07 - -0.06, 5 = -0.06 - -0.05, 6 = -0.05 - -0.04, 7 = -0.04 - -0.03, 8
= -0.03 - -0.02, 9 = -0.02 - -0.01, and 10 = -0.01 - 0.00 kg/ha/yr.
112
2030-2000 Local Generation of Urban N Yield (kg/ha/yr)
< -0.09
-0.09 - -0.08
-0.08 - -0.07
-0.07 - -0.06
-0.06 - -0.05
-0.05 - -0.04
-0.04 - -0.03
-0.03 - -0.02
-0.02 - -0.01
-0.01 - 0.00
a)
?
0 40 80 120 16020
Kilometers
b)
2030-2000 Local Generation of Mean Urban N Yield (kg/ha/yr)
100101 3
6
36
3
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
Category
Figure 4.7: Per catchment 2030-2000 change map of: a) Losses in locally generated
area-weighted mean urban (? 10% ISA) patch size yield (kg/ha/yr) with gains in locally
generated TN yield and b) Corresponding frequency distribution of the number of
catchments falling in each category. Category classes are: 1 = < -0.09, 2 = -0.09 - -0.08,
3 = -0.08 - -0.07, 4 = -0.07 - -0.06, 5 = -0.06 - -0.05, 6 = -0.05 - -0.04, 7 = -0.04 - -0.03, 8
= -0.03 - -0.02, 9 = -0.02 - -0.01, and 10 = -0.01 - 0.00 kg/ha/yr.
4.5.2 Applied fertilizer and manure applications
In catchments with projected TN yield losses, the net application in yield of
fertilizer and manure was -9.25 kg/ha/yr less (the projected -37.48 kg/ha/yr decrease in
113
applied fertilizer yield added to the projected +28.23 kg/ha/yr increase in applied manure
yield (Table 4. 3)), thus non-point source N fell. Overall, lower quantities of fertilizer
and manure were predicted to be transported to streams, as seen by the net loss of -4.23
kg/ha/yr (the projected -5.96 kg/ha/yr decrease in locally generated fertilizer yield added
to the projected +1.73 kg/ha/yr increase in locally generated manure yield (Table 4.4))
from these sources. The findings also suggest that because future sprawl will occur in
these catchments, possibly as leapfrog growth on cropland and pasture patches, overall
TN decreases. The highest decreases in applied fertilizer yield (> 30 kg/ha/yr) were
predicted to occur near: Baltimore and Frederick (MD) and Richmond and Norfolk-
Virginia Beach-Newport News (VA) (Figure 4.8a, c). The largest decreases in locally
generated fertilizer yield (> 30 kg/ha/yr) were predicted to occur only in small
catchments near Baltimore (MD) (Figure 4.8b, d). The highest increases in applied
manure yield (> 26 kg/ha/yr) were predicted to occur throughout the watershed (Figure
4.9a, c). However, the largest increases in locally generated manure yield (> 26 kg/ha/yr)
were also predicted to occur only in small catchments near Baltimore (MD) (Figure 4.9b,
d).
114
2030-2000 Application of Fertilizer N Yield (kg/ha/yr)
< -30.0
-30.0 - -26.5
-26.5 - -23.0
-23.0 - -19.5
-19.5 - -16.0
-16.0 - -12.5
-12.5 - -9.0
-9.0 - -5.5
-5.5 - -2.0
> -2.0
a)
?
0 40 80 120 16020
Kilometers
2030-2000 Local Generation of Fertilizer N Yield (kg/ha/yr)
< -30.0
-30.0 - -26.5
-26.5 - -23.0
-23.0 - -19.5
-19.5 - -16.0
-16.0 - -12.5
-12.5 - -9.0
-9.0 - -5.5
-5.5 - -2.0
> -2.0
b)
?
0 40 80 120 16020
Kilometers
c)  d)
2030-2000 Application of Fertilizer N Yield (kg/ha/yr)
4
8
67
34 4 4 4 3 5
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
Category
2030-2000 Local Generation of Fertilizer N Yield (kg/ha/yr)
30
18
5
31
0 1
3 2 2
14
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
Category
Figure 4.8: Per catchment 2030-2000 change maps of: a) Application and b) Local
generation of fertilizer N yield (kg/ha/yr) with losses in locally generated TN yield and c-
d) Corresponding frequency distributions of the number of catchments falling in each
category. Category classes are: 1 = < -30.0, 2 = -30.0 - -26.5, 3 = -26.5 - -23.0, 4 = -23.0
- -19.5, 5 = -19.5 - -16.0, 6 = -16.0 - -12.5, 7 = -12.5 - 9.0, 8 = -9.0 - -5.5, 9 = -5.5 - -2.0,
and 10 = > -2.0 kg/ha/yr.
115
2030-2000 Application of Manure N Yield (kg/ha/yr)
< 2
2 - 5
5 - 8
8 - 11
11 - 14
14 - 17
17 - 20
20 - 23
23 - 26
> 26
a)
?
0 40 80 120 16020
Kilometers
2030-2000 Local Generation of Manure N Yield (kg/ha/yr)
< 2
2 - 5
5 - 8
8 - 11
11 - 14
14 - 17
17 - 20
20 - 23
23 - 26
> 26
b)
?
0 40 80 120 16020
Kilometers
c)  d)
2030-2000 Application of Manure N Yield (kg/ha/yr)
11
61
10
5
2 3 2 3
5 4
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
Category
2030-2000 Local Generation of Manure N Yield (kg/ha/yr)
0010
7
30
0
66
11
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
Category
Figure 4.9: Per catchment 2030-2000 change maps of: a) Application and b) Local
generation of manure N yield (kg/ha/yr) with losses in locally generated TN yield and c-
d) Corresponding frequency distributions of the number of catchments falling in each
category. Category classes are: 1 = < 2, 2 = 2 - 5, 3 = 5 - 8, 4 = 8 - 11, 5 = 11 - 14, 6 =
14 - 17, 7 = 17 - 20, 8 = 20 - 23, 9 = 23 - 26, and 10 = > 26 kg/ha/yr.
In contrast, in catchments projecting TN yield gains, the net application of
fertilizer and manure yield was higher by +29.42 kg/ha/yr (the projected -2.28 kg/ha/yr
decrease in applied fertilizer yield added to the projected +31.70 kg/ha/yr increase in
116
applied manure yield (Table 4.3)). This was seen by the net gain in median yield within
streams of +2.60 kg/ha/yr (the projected -0.30 kg/ha/yr decrease in locally generated
fertilizer yield added to the projected +2.90 kg/ha/yr increase in locally generated manure
yield (Table 4.4)) attributed to these sources. Future sprawl here suggests that infill and
peripheral growth will occur where agricultural land conversions will be less likely. This
will be especially important in the case of pasture and its higher rate of manure
applications by 2030. The highest decreases in applied fertilizer yield (> 4.8 kg/ha/yr)
were predicted to occur near: DC; Hagerstown (MD); Harrisburg, Williamsport, State
College, and Altoona (PA); and Richmond (VA) (Figure 4.10a,c). The largest decreases
in locally generated fertilizer yield (> 4.8 kg/ha/yr) were predicted to occur near these
same regions (Figure 4.10b,d). The highest increases in applied manure yield (> 4.0
kg/ha/yr) will be scattered throughout the watershed (Figure 4.11a,c). The largest
increases in locally generated manure yield (> 4.0 kg/ha/yr) will be near: DC and
Scranton/Wilkes-Barre, Williamsport, State College, and Altoona (PA) (Figure 4. 11b,d).
117
2030-2000 Application of Fertilizer N Yield (kg/ha/yr)
< -4.8
-4.8 - -4.2
-4.2 - -3.6
-3.6 - -3.0
-3.0 - -2.4
-2.4 - -1.8
-1.8 - -1.2
-1.2 - -0.5
-0.6 - 0.0
> 0
a)
?
0 40 80 120 16020
Kilometers
2030-2000 Local Generation of Fertilizer N Yield (kg/ha/yr)
< -4.8
-4.8 - -4.2
-4.2 - -3.6
-3.6 - -3.0
-3.0 - -2.4
-2.4 - -1.8
-1.8 - -1.2
-1.2 - -0.6
-0.6 - 0.0
> 0.0
b)
?
0 40 80 120 16020
Kilometers
c)  d)
2030-2000 Application of Fertilizer N Yield (kg/ha/yr)
2
17
20
3
0 0
4
1 2 2
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
Category
2030-2000 Local Generation of Fertilizer N Yield (kg/ha/yr)
10
18
5 6
0
3 2
0
3 4
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
Category
Figure 4.10: Per catchment 2030-2000 change maps of: a) Application and b) Local
generation of fertilizer N yield (kg/ha/yr) with gains in locally generated TN yield and c-
d) Corresponding frequency distributions of the number of catchments falling in each
category. Category classes are: 1 = < -4.8, 2 = -4.8 - -4.2, 3 = -4.2 - -3.6, 4 = -3.6 - -3.0,
5 = -3.0 - -2.4, 6 = -2.4 - -1.8, 7 = -1.8 - 1.2, 8 = -1.2 - -0.6, 9 = -0.6 - -0.0, and 10 = > 0.0
kg/ha/yr.
118
2030-2000 Application of Manure N Yield (kg/ha/yr)
< 0
0.0 - 0.5
0.5 - 1.0
1.0 - 1.5
1.5 - 2.0
2.0 - 2.5
2.5 - 3.0
3.0 - 3.5
3.5 - 4.0
> 4.0
a)
?
0 40 80 120 16020
Kilometers
2030-2000 Local Generation of Manure N Yield (kg/ha/yr)
< 0.0
0.0 - 0.5
0.5 - 1.0
1.0 - 1.5
1.5 - 2.0
2.0 - 2.5
2.5 - 3.0
3.0 - 3.5
3.5 - 4.0
> 4.0
b)
?
0 40 80 120 16020
Kilometers
c)  d)
2030-2000 Application of Manure N Yield (kg/ha/yr)
0
47
3
01 0 0 0 0 0
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
Category
2030-2000 Local Generation of Manure N Yield (kg/ha/yr)
9
3
8
3
01
11
3
76
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
Category
Figure 4.11: Per catchment 2030-2000 change maps of: a) Application and b) Local
generation of manure N yield (kg/ha/yr) with gains in locally generated TN yield and c-d)
Corresponding frequency distributions of the number of catchments falling in each
category. Category classes are: 1 = < 0.0, 2 = 0.0 - 0.5, 3 = 0.5 - 1.0, 4 = 1.0 - 1.5, 5 =
1.5 - 2.0, 6 = 2.0 - 2.5, 7 = 2.5 - 3.0, 8 = 3.0 - 3.5, 9 = 3.5 - 4.0, and 10 = > 4.0 kg/ha/yr.
4.5.3 Percentage of land on coastal plain
Non-point N transport to Chesapeake Bay streams was less the more land on the
coastal plain. This could be a result of DON (NO3-) transported in surface overland flow
and shallow subsurface interflow being able to infiltrate deeper into soil profiles (Preston
119
and Brakebill, 1999). At these deeper subsurface depths, bacterial-induced denitrification
processes may occur under reduced (anoxic) conditions when in the presence of riparian
vegetation (Jordan et al., 1997; Willems et al., 1997; Cornwell et al., 1999; Hayashi and
Rosenberry, 2002). Increased infiltration and percolation has been linked with greater
hydraulic conductivity correlated to higher macropores found in coastal plain soils of the
Chesapeake Bay watershed (Willems et al., 1997; Wynn et al., 2000).
The finding that catchments with a loss in TN consisted of more coastal plain
suggested that leapfrog growth will occur primarily in areas not directly benefiting from
these larger regions of N attenuation. Thus, infill and peripheral growth will occur
predominantly in areas where larger regions of potential non-point attenuation would
have been more advantageous. The highest percentages of land on the coastal plain (90-
100%) in catchments with projected losses in TN are near: DC; Baltimore and Annapolis
(MD); and Richmond and Norfolk-Virginia Beach-Newport News (VA) (Figure 4.12a-b),
as compared to only small catchments near DC and Baltimore (MD) in those with
projected gains in TN (Figure 4.13a-b).
120
Coastal Plain (%)
0 - 10
10 - 20
20 - 30
30 - 40
40 - 50
50 - 60
60 - 70
70 - 80
80 - 90
90 - 100
a)
?
0 40 80 120 16020
Kilometers
b)
Coastal Plain (%)
11
45
34
6
0 1
3 2 2 2
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
Category
Figure 4.12: Per catchment map of: a) Percentage of land on the coastal plain (%) with
losses in locally generated TN yield and b) Corresponding frequency distribution of the
number of catchments falling in each category. Category classes are: 1 = 0 - 10, 2 = 10 -
20, 3 = 20 - 30, 4 = 30 - 40, 5 = 40 - 50, 6 = 50 - 60, 7 = 60 - 70, 8 = 70 - 80, 9 = 80 - 90,
and 10 = 90 - 100%.
121
Coastal Plain (%)
0 - 10
10 - 20
20 - 30
30 - 40
40 - 50
50 - 60
60 - 70
70 - 80
80 - 90
90 - 100
a)
?
0 40 80 120 16020
Kilometers
b)
Coastal Plain (%)
3
6
38
10 0 1 1 0 1
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
Category
Figure 4.13: Per catchment map of: a) Percentage of land on the coastal plain (%) with
gains in locally generated TN yield and b) Corresponding frequency distribution of the
number of catchments falling in each category. Category classes are: 1 = 0 - 10, 2 = 10 -
20, 3 = 20 - 30, 4 = 30 - 40, 5 = 40 - 50, 6 = 50 - 60, 7 = 60 - 70, 8 = 70 - 80, 9 = 80 - 90,
and 10 = 90 - 100%.
4.5.4 Percentage of Cropland
In catchments with projected TN losses, the higher amounts of cropland that
would be converted to urban would also lower the quantities of fertilizer and manure
122
being applied to the surface and thus, available for potential non-point leaching and
transport. It was estimated that cropland increased non-point N transport to streams
(Roberts and Prince, 2010). This has been attributed to decreases in hydraulic
conductivity from cultivation practices, such as tillage and machinery field compaction,
thus promoting surface sealing and overland runoff (Logsdon and Jaynes, 1996).
Catchments with TN losses also represented a transformation from cropland and
its higher N impacts to urban uses with lower N impacts. In another study of future
sprawl in the watershed's Patuxent Basin (Costanza et al., 2002), it was concluded that
there is a considerable gain in water quality by the conversion of agricultural land to
residential uses. In catchments with projected TN losses, the highest decreases in
percentage of cropland (> 0.9%) were predicted to occur scattered throughout the
watershed (Figure 4.14a-b).
In contrast, in catchments with projected gains in TN, it was suggested that even
with reduced, urban non-point impacts, the remaining agricultural non-point effects
would be sufficient to increase overall loadings. This was based in part on the lower
conversion of cropland that would occur under infill and peripheral growth conditions by
2030. In catchments with projected gains, the highest decreases in percentage of
cropland (> 0.9 %) were also predicted to occur scattered throughout the watershed-wide
(Figure 4.15a-b).
123
2030-2000 Cropland (%)
< -0.9
-0.9 - -0.8
-0.8 - -0.7
-0.7 - -0.6
-0.6 - -0.5
-0.5 - -0.4
-0.4 - -0.3
-0.3 - -0.2
-0.2 - -0.1
-0.1 - 0.0
a)
?
0 40 80 120 16020
Kilometers
b)
2030-2000 Cropland (%)
7
13
42
9
5
8 8 6
3 5
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
Category
Figure 4.14: Per catchment 2030-2000 change map of: a) Decreases in percentage of
cropland (%) with losses in locally generated TN yield and b) Corresponding frequency
distribution of the number of catchments falling in each category. Category classes are: 1
= < -0.9, 2 = -0.9 - -0.8, 3 = -0.8 - -0.7, 4 = -0.7 - -0.6, 5 = -0.6 - -0.5, 6 = -0.5 - -0.4, 7 = -
0.4 - -0.3, 8 = -0.3 - -0.2, 9 = -0.2 - -0.1, and 10 = -0.1 - 0.0 %.
124
2030-2000 Cropland (%)
< -0.9
-0.9 - -0.8
-0.8 - -0.7
-0.7 - -0.6
-0.6 - -0.5
-0.5 - -0.4
-0.4 - -0.3
-0.3 - -0.2
-0.2 - -0.1
-0.1 - 0.0
a)
?
0 40 80 120 16020
Kilometers
b)
2030-2000 Cropland (%)
6
16
9 7
0 2 1
4 3 3
0
10
20
30
40
50
60
70
1 2 3 4 5 6 7 8 9 10
Category
Figure 4.15: Per catchment 2030-2000 change map of: a) Decreases in percentage of
cropland (%) with gains in locally generated TN yield and b) Corresponding frequency
distribution of the number of catchments falling in each category. Category classes are: 1
= < -0.9, 2 = -0.9 - -0.8, 3 = -0.8 - -0.7, 4 = -0.7 - -0.6, 5 = -0.6 - -0.5, 6 = -0.5 - -0.4, 7 = -
0.4 - -0.3, 8 = -0.3 - -0.2, 9 = -0.2 - -0.1, and 10 = -0.1 - 0.0 %.
4.6 Conclusions
To decrease projected TN yields in sprawling catchments, it is suggested that
whenever possible, management should confine urbanization to smaller patches in
125
cropland and pasture. This leapfrog growth could reduce the non-point agricultural N
effects from fertilizer and manure applied to these cover types. This type of sprawl aids
in the transformation of land to urban uses with an overall lower N impact. Additionally,
leapfrog growth may lead to higher attenuation of urban non-point loadings by dispersing
these concentrated sources over a greater number of stream networks.
Although only five of the eleven SPARROW variables had significant changes or
differences between the catchment categories, all variables should be examined. On a
catchment-by-catchment basis, other variables may help explain projected TN losses or
gains. An example of this is point sources that may cause increases in catchments with
projected TN gains. This scenario was realized in a study of anticipated sprawl in
Sweden that concluded point sources were the cause of the overall gain in future TN to
the Baltic Sea (Jansson and Colding, 2007). Also, Bowen and Valiela (2001) concluded
that a factor of 17 increase in delivered point sources was a primary cause for the
substantial, overall TN gain to estuaries near Cape Cod, Massachusetts from 1938 to
1990.
In conclusion, the results of this study further confirmed our earlier demonstration
that landscape pattern has a significant effect on TN runoff. Also, it suggested
management practices and policies that could be enacted for sustainable future growth. It
was concluded that significant rates of urbanization can occur, while still improving the
overall nutrient health of the Chesapeake Bay, findings that could be beneficial to
watersheds projected to undergoing similar, developmental, transformations elsewhere.
126
Chapter 5: Lessons learned from current and projected future urban
land cover in Chesapeake Bay watershed4
5.1 Abstract
This overview summarizes findings from recent studies of the Chesapeake
Bay that combined remotely sensed, land cover and land use (LC/LU) data within a
watershed-wide, nutrient loading (mass for a specified time) model to quantify effects of
current and projected future urban land cover on total nitrogen (TN) and total phosphorus
(TP) runoff. Important findings included: 1) slightly improved accuracy and precision of
current estimated nutrient loadings to the Chesapeake Bay when using landscape
configuration and riparian stream buffers, 2) overall nutrient loadings to the Chesapeake
Bay decreased due to agricultural land losses and reductions in fertilizers after projected
urbanization, and 3) the importance of spatial arrangement of forecasted urban sprawl to
projected TN response within catchments of the watershed. Using these and other key
findings, it is concluded that several management and policy recommendations could be
implemented in the watershed to help mitigate nutrient-related issues in the Chesapeake
Bay. Furthermore, if climate change can be evaluated in addition to this landscape
analysis, a more comprehensive understanding of the anthropogenic impacts on projected
watershed-wide nutrient loadings can potentially be quantified.
4 The material in Chapter 5 is currently in review. Roberts, A. D., Submitted for publication. Lessons
learned from current and projected urban land cover in Chesapeake Bay watershed. Environmental Science
and Policy.
127
5.2 Introduction
Since the introduction of European settlement into the Chesapeake Bay watershed
in the 17th and 18th centuries, land cover and land use (LC/LU) changes in the form of
deforestation that created agricultural and urban cover resulted in steadily declining water
quality (Boesch, 2006). Although initially detrimental to the overall functioning of this
ecosystem, widespread degradation of the Chesapeake Bay did not become apparent until
after World War II when rapid urbanization and population growth and the intense use of
commercial fertilizers watershed-wide significantly increased the genesis and delivery of
nutrients to the Chesapeake Bay (Sims and Coale, 2002). These excess nutrients in the
form of total nitrogen (TN) and total phosphorus (TP) have contributed greatly to
problems of: increased sedimentation, turbidity (water opaqueness), hypoxia (low levels
of dissolved oxygen (DO)), anoxia (devoid of DO), a lowering in submerged aquatic
flora and fauna populations, and a lowering in the overall sustainability of the watershed
and Chesapeake Bay itself (Breitburg 1992; Bratton et al., 2003; Hassett et al., 2005).
In addition to being the largest estuary in the United States (U.S.), the Chesapeake
Bay also has the largest ratio of land-to-water volume of any estuary in the world, making
runoff from the surface extremely critical to the overall health of this ecosystem (Shuyler
et al., 1995). The 166,534 km2 watershed drains portions of six states: (New York (NY),
Pennsylvania (PA), Delaware (DE), Maryland (MD), West Virginia (WV) and Virginia
(VA)) and the District of Columbia (DC) from six major basins (the James, Patuxent,
Potomac, Rappahannock, Susquehanna, and York) and other smaller rivers and streams at
an estimated freshwater inflow rate of 8.7 x 106 m3/h (Figure 5.1a) (Sims and Coale,
128
2002). The watershed also contains numerous large, intermediate, and small urban
centers that have been previously correlated to the estuary?s overall trophic status (Figure
5.1b) (Preston and Brakebill, 1999; Brakebill et al., 2001; Brakebill and Preston, 2004).
PA
VA
NY
WV
MD NJ
NC
OH
DE
DC
JAM E
S R
SUS
QUEHAN N A R
PO
TO
MA
C R
RA
PP
AH
ANN
OCK
R
PA
TU
XE
NT
R
MA
TTA
PO
NI RJA
ME
S R
SU
SQ
UE
HA
NN
A R
PAT
UX
EN
T R
PAM
UNK
EY
R
a)
?
0 40 80 120 16020
Kilometers
Baltimore, MD
Washington, DC--MD--VA
Richmond, VA
Harrisburg, PA
Norfolk--Virginia Beach--Newport Ne
Scranton--Wilkes-Barre, PA
York, PALancaster, PA
Lynchburg, VA
Binghamton, NY
Petersburg, VA
Cumberland, MD--WV
Elmira, NY
Altoona, PA
Frederick, MD
Fredericksburg, VA
Annapolis, MD
Hagerstown, MD--PA--WV
Charlottesville, VA
Williamsport, PA
State College, PAPA
VA
NY
WV
NJ
NC
OH
DEMD
b)
?
0 40 80 120 16020
Kilometers
Figure 5.1: The Chesapeake Bay watershed showing the locations of: a) Streams and
rivers draining the estuary and b) Urban centers located within its boundaries.
To comprehensively examine current (2000) urban and non-urban land cover in
the watershed, the Regional Earth Science Applications Center (RESAC) used
LANDSAT data to map LC/LU, percent impervious surface area (% ISA), and percent
tree cover (% TC) (Goetz et al., 2003, 2004a, b; Jantz et al., 2005). Additionally,
spatially-explicit forecasts of the 2030 locations of watershed-wide development with
measures of uncertainty were provided by the Slope, Land use, Exclusion, Urban extent,
Transportation, Hillshade (SLEUTH) model, a cellular automata model that represented
urban growth and expansion processes in a two-dimensional grid (Jantz et al., in press).
129
Using these land cover datasets, current (2000) and projected (2030) TN and TP
loadings (mass for a specified time) to the Chesapeake Bay were simulated using the
SPAtially Referenced Regressions On Watershed Attributes (SPARROW) model
(Roberts et al., 2009; Roberts and Prince, 2010). SPARROW estimates these nutrient
loadings and yields (mass for a specified time normalized for drainage area) from
spatially-referenced watershed reaches, catchments, and networks using point and non-
point source, land-to-water delivery, and stream and reservoir decay variables (Schwarz
et al., 2006). For these models, 2,339 catchments draining to the estuary were simulated
with a 31 m riparian stream buffer (Figure 5.2a-b) using significant (p value ? 0.05)
compositional (variety and abundance) and configurational (orientation and position)
metrics of land cover (Roberts et al., 2009; Roberts and Prince, 2010). Furthermore,
other land cover-based variables (such as point sources and applied fertilizer and manure)
were estimated to 2030 in an effort to more accurately portray overall projected loadings.
130
PA
VA
NY
WV
MD NJ
NC
OH
DE
DC
?
0 40 80 120 16020
Kilometers
a)
Example Catchment
Runoff from Catchment
Enters Stream
31 m Buffer
Upstream Catchments
Carrying N Loads
Upstream Plus Locally Generated
Runoff, Less Instream Losses
Leaves Catchment
b)
?
0 1
Kilometer
i
i
i
ii
i
Figure 5.2: 2,339 catchments used in the 2000 and 2030 runs of the: a) RESAC 31 m
Chesapeake Bay SPARROW models with b) Schematic diagram of an example
catchment area showing fixed 31 m riparian stream buffer width area surrounding stream
reach and corresponding upstream and downstream runoff paths.
The multistate effort to restore the Chesapeake Bay ecosystem to pre-colonial
times by reducing the inputs of nutrients leading to enrichment is one of the world?s most
ambitious attempts at large-scale ecosystem restoration (Boesch et al., 2001). A prime
example of this was demonstrated with the enactment of the Chesapeake Bay Agreement
in 1987 that endeavored to reduce controllable 1985 nutrient levels by 40% by 2000
(Chesapeake Executive Council, 1987). In addition, new 2000 amendments called for the
removal of all nutrient-related issues impairing the Chesapeake Bay and its removal from
the list of impaired waters under the Clean Water Acts (Wang et al., 2006).
As a result of these efforts, it is concluded that using current and projected
watershed-wide urban land cover to quantify comprehensive landscape effects on nutrient
131
loadings to the Chesapeake Bay is an important management and research priority that
may help local, state, and federal stakeholders meet these reduction mandates. This is
due to the findings summarized here that hint towards new correlations between the
terrestrial and aquatic environments that may lead to new approaches to the management
and restoration of nutrient-enriched ecosystems. Findings that not only address these
correlations at the regional scale of the Chesapeake Bay watershed, but may also generate
greater insight into mechanisms and processes controlling nutrient-related,
biogeochemical cycling at local and global scales.
Thus, using these findings, several priority research and management questions
are addressed in this overview and include:
? What LC/LUs at catchment and riparian stream buffer scales are significant
in the current (2000) non-point (diffuse) sources and delivery of TN and TP
loadings estimated in the Chesapeake Bay watershed?
? What are the causes of the observed effects of LC/LUs at catchment and
riparian stream buffer scales on the current (2000) non-point (diffuse)
sources and delivery of TN and TP loadings estimated in the Chesapeake Bay
watershed?
? What specific types of LC/LUs lost to future urbanization will have the
greatest impacts on projected TN and TP loadings estimated to the
Chesapeake Bay?
? What changes in TN and TP loadings will occur in the Chesapeake Bay
between 2000 and the future (2030) under the urbanization scenario applied
here?
132
Additional findings and management and policy recommendations developed from these
summarized results are also included within this overview. Finally, this overview
commences with a discussion of directions for future research.
5.3 Significance of findings for priority management questions
? What LC/LUs at catchment and riparian stream buffer scales are significant
in the current (2000) non-point (diffuse) sources and delivery of TN and TP
loadings estimated in the Chesapeake Bay watershed?
At the entire Chesapeake catchment scale, the LC/LU sources of non-point TN to
the estuary were mostly related to agriculture (cropland and pasture) with the exception
of urban (Roberts and Prince, 2010). In regards to TP loadings, all LC/LUs
(anthropogenic and non-anthropogenic) were non-point sources to the Chesapeake Bay at
the catchment scale (Roberts and Prince, 2010). The significant cover types for TP were
urban, non-agricultural/non-urban, and agricultural.
For each hectare (ha) of urban (? 10% ISA) that weighted each patch of this
LC/LU by its size relative to the total area of this LC/LU per catchment (area-weighted
mean urban (? 10% ISA) patch size), nearly 25 and 1 kilograms per year (kg/yr) of N and
P were estimated in Chesapeake Bay streams (Roberts and Prince, 2010). In terms of TN
and TP loadings from agriculture, the application of fertilizers and manure was the main
cause of the high loadings. For each kg of fertilizer and manure N applied to agricultural
lands, approximately 0.2 and 0.1 kg/yr of N were estimated in streams (Roberts and
Prince, 2010). Each kg of fertilizer and manure P applied to these lands yielded an
estimated 0.02 and 0.01 kg/yr of P in streams (Roberts and Prince, 2010).
133
In the case of TP, non-agricultural/non-urban lands that consisted primarily of
forests were a major source of P. This was due to large area-weighted non-
agricultural/non-urban patch sizes being distributed watershed-wide, even though
delivery in yield from this land cover was quite low (Figure 5.3a-b). For every ha of this
patch size, about 0.1 kg/yr of P was estimated to streams (Roberts and Prince, 2010).
2000 Mean NA/NU Patch Size (ha)
0 - 3,000
3,000 - 6,000
6,000- 9,000
9,000 - 12,000
12,000 - 15,000
15,000 - 18,000
18,000 - 21,000
21,000 - 24,000
24,000 - 27,000
> 27,000
a)
?
0 40 80 120 16020
Kilometers
2000 Delivery of NA/NU P Yield (kg/ha/yr)
0.00 - 0.03
0.03 - 0.06
0.06 - 0.09
0.09 - 0.12
0.12 - 0.15
0.15 - 0.18
0.18 - 0.21
0.21 - 0.24
0.24 - 0.27
> 0.27
b)
?
0 40 80 120 16020
Kilometers
Figure 5.3: Per catchment source factor metric of: a) Area-weighted mean non-
agricultural/non-urban (NA/NU) patch size in ha with delivery of b) P yield to
Chesapeake Bay in kg/ha/yr in the 2000 RESAC 31 m TP model.
Significant LC/LUs estimated to contribute to the delivery of non-point N to the
Chesapeake Bay included: extractive land and cropland at the catchment scale and
evergreen forest isolated to the 31 m riparian stream buffer (Roberts and Prince, 2010).
For each percent of extractive, cropland, and evergreen forest at these scales, N delivered
to streams from anon-point sources was estimated to increase by 0.27%, 0.021%, and
134
0.013%, respectively. Barren land isolated to the 31 m riparian stream buffer was the
only significant LC/LU correlated to P transport by increasing delivery of non-point P to
streams by 0.281% for each percent of its composition (Roberts and Prince, 2010). For
both nutrients, this result was the first demonstration of the significance of riparian
stream buffer LC/LU on transport at the scale of the entire watershed. With the exception
of cropland, an interesting finding was that all of these LC/LUs are ones that are not
commonly associated with the non-point source delivery of N and P to the estuary.
Additionally, these significant land-to-water delivery cover types occurred throughout the
watershed in smaller percentages (Table 5.1) (Goetz et al., 2004a, b; Jantz et al., 2005).
Significant land-to-water delivery LC/LUs 2000 watershed-wide composition (%)
Extractive land 0.21
Cropland 8.18
Evergreen forest within 31 m riparian stream buffer 0.06
Barren land within 31 m riparian stream buffer 0.003
Table 5.1: 2000 watershed-wide compositions of significant land-to-water delivery
LC/LUs.
Extractive land was located predominantly near central PA, western MD, and
northeastern WV (Figure 5.4a). Cropland was clustered near southern Pennsylvania, the
eastern shore of MD, and DE (Figure 5.4b). Although scattered throughout the
watershed, the highest concentrations of evergreen forest within the 31 m riparian stream
buffer were found in central PA, south central NY, and central and western VA (Figure
5.4c). Finally, barren land within the 31 m riparian stream buffer was predominantly
clustered near central VA and southeastern and south central PA (Figure 5.4d). Thus, at
the regional scale of the entire Chesapeake Bay, the high impact LC/LUs varied in
location.
135
2000 Extractive Land (%)
0.0 - 0.5
0.5 - 1.0
1.0 - 1.5
1.5 - 2.0
2.0 - 2.5
2.5 - 3.0
3.0 - 3.5
3.5 - 4.0
4.0 - 4.5
> 4.5
a)
?
0 40 80 120 16020
Kilometers
2000 Cropland (%)
0 - 5
5 - 10
10 - 15
15 - 20
20 - 25
25 - 30
30 - 35
35 - 40
40 - 45
> 45
b)
?
0 40 80 120 16020
Kilometers
2000 Evergreen Forest In Buffer (%)
0.0 - 2.5
2.5 - 5.0
5.0 - 7.5
7.5 - 10.0
10.0 - 12.5
12.5 - 15.0
15.0 - 17.5
17.5 - 20.0
20.0 - 22.5
> 22.5
c)
?
0 40 80 120 16020
Kilometers
2000 Barren Land In Buffer (%)
0.00 - 0.25
0.25 - 0.50
0.50 - 0.75
0.75 - 1.00
1.00 - 1.25
1.25 - 1.50
1.50 - 1.75
1.75 - 2.00
2.00 - 2.25
> 2.25
?
0 40 80 120 16020
Kilometers
d)
Figure 5.4: Per catchment land-to-water delivery metrics of: a) Extractive land (%), b)
Cropland (%), and c) Evergreen forest within the riparian stream buffer (%) in the 2000
RESAC 31 m TN model and d) Barren land within the riparian stream buffer (%) in the
2000 RESAC 31 m TP model.
136
Finally, another LC/LU correlated to the non-point transport of N was catchment-
wide area-weighted mean edge contrast of deciduous forest (Roberts and Prince, 2010).
However, unlike the other significant delivery cover types, this variable was based upon
landscape configuration. Deciduous forest fragmentation influencing nutrient runoff was
quantified by either adjacencies to other non-urban, natural LC/LUs (such as mixed forest
and wetland) or non-urban, anthropogenic-altered LC/LUs (such as extractive, cropland,
and pasture). For each percent gain in this variable (meaning increased overall
adjacencies to non urban, anthropogenic-altered LC/LUs at the catchment scale), delivery
of non-point N to Chesapeake Bay streams was estimated to increase by 0.014% (Roberts
and Prince, 2010).
? What are the causes of the observed effects of LC/LUs at catchment and
riparian stream buffer scales on the current (2000) non-point (diffuse)
sources and delivery of TN and TP loadings estimated in the Chesapeake Bay
watershed?
It was surmised that developed (urban) and agricultural LC/LUs yielded
significant N and P loadings, either through urban sources, such as vehicle emissions,
lawn fertilizers, and animal wastes or through agricultural sources, such as fertilizer and
manure applications (Roberts and Prince, 2010). The nutrient sources were actively
delivered to streams after precipitation and irrigation events. During these sources of
runoff water, urban non-point N and P were delivered to streams via developed patches
connected to sewers, culverts and stormwater drains (Figure 5.5a-c) (Roberts and Prince,
2010). Agricultural non-point N and P were delivered to streams via non-urban LC/LUs
that were previously mentioned and capable of transport. For both LC/LU sources, the
137
concentrated buildup of non-point N and P occurred primarily as a result of the
impervious components within these cover types. The penetration of N and P into deeper
soil profiles was prevented by paving (urban), certain tillage practices, (agriculture) and
use of heavy-machinery that compact the surface (agriculture) (Logsdon and Jaynes,
1996), and allowed for a buildup of these nutrients at the surface and shallow-subsurface
level. Thus, both LC/LUs were substantial sources of non-point N and P to the estuary.
138
2000 Mean Urban Patch Size (ha)
0 - 30
30 - 60
60 - 90
90 - 120
120 - 150
150 - 180
180 - 210
210 - 240
240 - 270
> 270
a)
?
0 40 80 120 16020
Kilometers
2000 Delivery of Urban N Yield (kg/ha/yr)
0.00 - 0.11
0.11 - 0.22
0.22 - 0.33
0.33 - 0.44
0.44 - 0.55
0.55 - 0.66
0.66 - 0.77
0.77 - 0.88
0.88 - 0.99
> 0.99
b)
?
0 40 80 120 16020
Kilometers
2000 Delivery of Urban P Yield (kg/ha/yr)
0.00 - 0.05
0.05 - 0.10
0.10 - 0.15
0.15 - 0.20
0.20 - 0.25
0.25 - 0.30
0.30 - 0.35
0.35 - 0.40
0.40 - 0.45
> 0.45
c)
?
0 40 80 120 16020
Kilometers
Figure 5.5: Per catchment source factor metrics of: a) Area-weighted mean urban (?
10% ISA) patch size (ha) with delivery of b) N and c) P yield to Chesapeake Bay in
kg/ha/yr in the 2000 RESAC 31 m models.
139
In regards to TP, non-agricultural/non-urban LC/LU was the background (natural)
source of P in the Chesapeake Bay watershed before modern anthropogenic influences
started (Roberts et al., 2009). It was concluded that P from this LC/LU grouping was not
the cause of current eutrophication in the Chesapeake Bay and its watershed. This P was
generated from the leaching of decomposed organic matter on the forest floor (such as
litter fall) and subsurface (such as rooting systems) to groundwater sources. The
groundwater sources of dissolved P can subsequently runoff to Chesapeake Bay streams,
over temporal scales that can range from months to decades. This naturally-occurring
process is responsible for the generation of background P in streams and has been found
in other studies of areas similar to those of the Chesapeake Bay watershed (Yanai, 1992).
Finally, extractive land, cropland, barren land, evergreen forests and non-urban,
anthropogenic-impacted LC/LUs surrounding deciduous forests were all cover types that
can increase non-point N and P delivery to the Chesapeake Bay probably because of their
diminished permeability (Roberts and Prince, 2010). In effect, they were non-urban
lands that mimicked the impervious properties of urban uses. By the reduction of
hydraulic conductivity that would allow for the initial infiltration and subsequent
percolation of runoff through soil pores into subsurface groundwater reserves, these cover
types encouraged overland and shallow subsurface runoff. Thus, the cover types were
capable of transporting dissolved and sediment-enriched nutrients to the estuary. Other
studies have noted this effect, especially in regards to barren lands (Ziegler and
Giambelluca, 1997, 1998; Ziegler et al., 2001, 2004; Assouline and Mualem, 2002;
Perkins et al., 2007; Lopez et al., 2008). The occurrence of this effect at the catchment
and riparian stream buffer scale throughout the entire Chesapeake Bay watershed was a
140
significant finding. Furthermore, forested riparian buffers previously thought to attenuate
non-point N delivery at local scales (Correll et al., 1992; Jordan et al., 1993; Boesch et
al., 2001) were found here to increase non-point N delivery at the entire watershed scale.
? What specific types of LC/LUs lost to future urbanization will have the
greatest impacts on projected TN and TP loadings estimated to the
Chesapeake Bay?
The LC/LU types that had the greatest impacts on projected nutrient loadings
estimated to the estuary were cropland and pasture (Roberts et al., 2009). This resulted
from their higher nutrient production, as opposed to those of developed uses. The
conversion of agricultural land to urban cover estimated to occur in the lower
Susquehanna Basin had the greatest proportional impact on projected TN and TP
loadings, since the river provides nearly half of the freshwater volume entering the
Chesapeake Bay (Figure 5.6a). Thus, any conversion here would be the key to reducing
nutrient-related issues in the estuary and its watershed. In the southeastern PA portion of
the Susquehanna alone, several catchments were forecasted to have > 9% of their existing
cropland converted to urban by 2030 (Figure 5.6b). Catchments that had the highest
delivered TN (> 18 kg/ha/yr) and TP yields (> 0.99 kg/ha/yr) to the Chesapeake Bay in
2000 (Figure 5.6c-d). Reduction in cropland in these catchments occurred with westward
expansion of suburbs from Philadelphia (PA), northern expansion of suburbs from
Baltimore (MD), and the growth of urbanized areas surrounding Harrisburg, York, and
Lancaster (PA) (Figure 5.6a). This same growth was predicted on the eastern shore of
MD and DE (Figure 5.6a). The forecasted growth surrounding the Salisbury (MD) area
141
illustrated how increased urbanization rates in areas of the highest yields in 2000 had the
greatest impacts on reducing TN and TP loadings to the Chesapeake Bay.
142
2030-2000 Mean Urban Patch Size (ha)
< -1,300
-1,300 - -975
-975 - -650
-650 - -325
-325 - 0
0 - 325
325 - 650
650 - 975
975 - 1,300
> 1,300
a)
?
0 40 80 120 16020
Kilometers
2030-2000 Cropland (%)
< -9
-9 - -8
-8 - -7
-7 - -6
-6 - -5
-5 - -4
-4 - -3
-3 - -2
-2 - -1
-1 - 0
b)
?
0 40 80 120 16020
Kilometers
2000 Delivery of TN Yield (kg/ha/yr)
0 - 2
2 - 4
4 - 6
6 - 8
8 - 10
10 - 12
12 - 14
14 - 16
16 - 18
> 18
c)
?
0 40 80 120 16020
Kilometers
2000 Delivery of TP Yield (kg/ha/yr)
0.00 - 0.11
0.11 - 0.22
0.22 - 0.33
0.33 - 0.44
0.44 - 0.55
0.55 - 0.66
0.66 - 0.77
0.77 - 0.88
0.88 - 0.99
> 0.99
d)
?
0 40 80 120 16020
Kilometers
Figure 5.6: Per catchment 2030-2000 differences of the: a) Area-weighted mean urban
(? 10% ISA) patch size source metric in ha and b) Cropland (%) projected to have the
greatest impacts on catchments with the highest 2000 delivery of c) TN and d) TP yield
to Chesapeake Bay in kg/ha/yr by 2030.
143
The conversion of extractive land (Figure 5.7a) to development (Figure 5.6a) was
predicted to have the greatest impact in catchments currently producing the highest TN
yields near Scranton-Wilkes Barre (PA) in the middle Susquehanna Basin and
Cumberland (MD) in the upper Potomac Basin (Figure 5.6c) (Roberts et al., 2009;
Roberts and Prince, 2010). In addition, the forecasted conversion of barren land within
the 31 m riparian stream buffer (Figure 5.7b) to urban (Figure 5.6a) was also predicted to
reduce the highest yielding TP catchments currently in the lower Susquehanna Basin near
Lancaster (PA) (Figure 5.6d) (Roberts et al., 2009; Roberts and Prince, 2010). In these
catchments, the conversion of these cover types reduced lands that were correlated to
increased transport of non-point N and P loadings to streams from overland runoff.
2030-2000 Extractive Land (%)
< -0.27
-0.27 - -0.24
-0.24 - -0.21
-0.21 - -0.18
-0.18 - -0.15
-0.15 - -0.12
-0.12 - -0.09
-0.09 - -0.06
-0.06 - -0.03
-0.03 - 0.00
a)
?
0 40 80 120 16020
Kilometers
2030-2000 Barren Land in Buffer (%)
< -0.45
-0.45 - -0.40
-0.40 - -0.35
-0.35 - -0.30
-0.30 - -0.25
-0.25 - -0.20
-0.20 - -0.15
-0.15 - -0.10
-0.10 - -0.05
-0.05 - 0.00
?
0 40 80 120 16020
Kilometers
b)
Figure 5.7: Per catchment 2030-2000 differences of the land-to-water delivery metrics
of: a) Extractive land (%) and b) Barren land within the riparian stream buffer (%) in the
RESAC 31 m models.
144
? What changes in TN and TP loadings will occur in the Chesapeake Bay
between 2000 and the future (2030) under the urbanization scenario applied
here?
After urbanization, TN and TP loadings delivered to the Chesapeake Bay by 2030
were projected to be reduced by approximately 19% and 20% (Table 5.2; Figure 5.8a-d)
(Roberts et al., 2009). Reductions in TN and TP loadings indicated the pivotal role of
decreases in the application of fertilizers and their non-point source runoff (Roberts et al.,
2009). Reductions in non-point source losses from fertilizers occurred not only through
the conversion of cropland and pasture to development, but the use of a smaller mean
application rate for agricultural lands that decreased from 39.78 and 13.65 kg/ha/yr for N
and P in 2000 to an estimated 15.75 and 7.49 kg/ha/yr for N and P in 2030.
145
Table 5.2: Comparison of 2000 and projected 2030 total loadings in kg/yr delivered to the Chesapeake Bay from all significant
sources in RESAC 31 m models. * Denotes significant source variable in the RESAC 31 m TN models that used the same inputs for
2000 and 2030 and was projected to decrease by 2030 due to conversions of significant land-to-water delivery LC/LUs.
TN model run TP model run
Variable 2000 2030 2030-2000
change
2030-2000
% change
Variable 2000 2030 2030-2000
change
2030-2000
% change
Point 4.4105 x 107 5.2200 x 107 +8.0950 x 106 +18.35 Point 1.9665 x 106 2.3677 x 106 +4.0120 x 105 +20.40
Applied fertilizer 4.8395 x 107 1.5615 x 107 -3.2780 x 107 -67.73 Applied fertilizer 1.0226 x 106 5.4169 x 105 -4.8091 x 105 -47.03
Applied manure 9.7110 x 106 1.2783 x 107 +3.0720 x 106 +31.63 Applied manure 5.2862 x 105 4.8027 x 105 -4.8350 x 104 -9.15
* Atmospheric
deposition
3.5803 x 107 2.7355 x 107 -8.4480 x 106 -23.60 Area-weighted
mean non-
agricultural/
non-urban patch
size
1.5772 x 106 5.0174 x 105 -1.0755 x 106 -68.19
Area-weighted
mean urban (?
10% ISA) patch
size
6.9311 x 106 9.1806 x 106 +2.2495 x 106 +32.46 Area-weighted
mean urban (?
10% ISA) patch
size
2.7222 x 105 4.0387 x 105 +1.3165 x 105 +48.36
Total 1.4495 x 108 1.1713 x 108 -2.7820 x 107 -19.19 Total 5.3671 x 106 4.2953 x 106 -1.0718 x 106 -19.97
146
2030 Delivery of TN Yield (kg/ha/yr)
0 - 2
2 - 4
4 - 6
6 - 8
8 - 10
10 - 12
12 - 14
14 - 16
16 - 18
> 18
a)
?
0 40 80 120 16020
Kilometers
2030 Delivery of TP Yield (kg/ha/yr)
0.00 - 0.11
0.11 - 0.22
0.22 - 0.33
0.33 - 0.44
0.44 - 0.55
0.55 - 0.66
0.66 - 0.77
0.77 - 0.88
0.88 - 0.99
> 0.99
b)
?
0 40 80 120 16020
Kilometers
2030-2000 Delivery of TN Yield (kg/ha/yr)
< -4
-4 - -3
-3 - -2
-2 - -1
-1 - 0
0 - 1
1 - 2
2 - 3
3 - 4
> 4
c)
?
0 40 80 120 16020
Kilometers
2030-2000 Delivery of TP Yield (kg/ha/yr)
< -0.20
-0.20 - -0.15
-0.15 - -0.10
-0.10 - -0.05
-0.05 - 0.00
0.00 - 0.05
0.05 - 0.10
0.10 - 0.15
0.15 - 0.20
> 0.20
d)
?
0 40 80 120 16020
Kilometers
Figure 5.8: Per catchment delivery of: a) TN and b) TP in yield to Chesapeake Bay in
kg/ha/yr in the 2030 RESAC 31 m models with 2030-2000 differences of delivery of: c)
TN and d) TP yield to Chesapeake Bay in kg/ha/yr in the RESAC 31 m models.
147
Increases in point and impervious-based, non-point source loadings from urban
lands can be expected to have lower overall N and P loadings than the cropland and
pasture they replace. Even with a projected population increase of four million that lead
to an estimated 18% and 20% gain in point N and P loadings and a 32% and 48% gain in
non-point N and P loadings (Table 5.2), total urban loading increases were not sufficient
to negate the reduction in losses from agricultural lands (Roberts et al., 2009). Recent
studies in smaller watersheds in the midwestern U.S. have also concluded that the
replacement of agricultural lands and their higher nutrient production with urban LC/LUs
of lower nutrient production would lower overall TN and TP (Tang et al., 2005; Wang et
al., 2005). The result reported here was the first demonstration that this phenomenon
would occur at the large regional scale, represented by the Chesapeake Bay watershed.
5.4 Additional Findings
The integration of compositional and configuration LC/LU metrics at catchment
and riparian stream buffer scale within the 2000 SPARROW models gave slightly more
accurate stream loading estimates when compared with 87 TN and 104 TP measurement
sites, as opposed to the last SPARROW models applied to the Chesapeake Bay in 1997
(Table 5.3) (Brakebill and Preston, 2004; Roberts and Prince, 2010). In addition, the use
of metrics at these scales in SPARROW yielded slightly more precise results when
compared with the 2000 Phase 4.3 Hydrologic Simulation Program-FORTRAN (HSPF)
model estimates for loadings to the estuary, also as opposed to the 1997 models (Table
5.4) (Roberts and Prince, 2010). Since the enactment of the Chesapeake Bay Agreement
in the 1980s, HSPF has been developed by the Chesapeake Bay Program (CBP) and
148
United States Environmental Protection Agency (USEPA) as the primary tool aiding in
federal and state management and regulatory decisions regarding the estuary (United
States Environmental Protection Agency, 2009). Without quantifying comprehensive
landscape and riparian stream buffer effects, the 1997 TN model overestimated 2000
HSPF loadings by 15%, as compared to 12% using the RESAC 31 m model (Table 5.4)
(Roberts and Prince, 2010). The 1997 TP model underestimated 2000 HSPF by 40%, as
compared to 38% using the RESAC 31 m model (Table 5.4) (Roberts and Prince, 2010).
Model run Model yield r2 Model RMSE (%)
1997 B & P TN 0.9073 0.2834
2000 RESAC 31 m TN 0.9366 0.2407
1997 B & P TP 0.7413 0.3257
2000 RESAC 31 m TP 0.7503 0.3126
Table 5.3: Comparison of yield coefficient of determination (r2), and root mean squared
error (RMSE) between the 1997 Brakebill and Preston (B & P) and 2000 RESAC 31 m
SPARROW models.
149
Table 5.4: Comparison of 1997 B & P and 2000 RESAC 31 m SPARROW model runs with the mean 1985-1994 and 2000 HSPF
Phase 4.3 model runs for total loadings in kg/yr delivered to the Chesapeake Bay.
Nutrient
(kg/yr)
1997
B & P
2000
RESAC 31 m
Mean 1985-1994
Phase 4.3 HSPF
2000
Phase 4.3 HSPF
1997
B & P ?
mean 1985-1994
Phase 4.3 HSPF
% change
2000
RESAC 31 m ?
mean 1985-1994
Phase 4.3 HSPF
% change
1997
B & P ?
2000
HSPF Phase 4.3
% change
2000
RESAC 31 m ?
2000
HSPF Phase 4.3
% change
TN 1.480 x 108 1.449 x 108 1.420 x 108 1.292 x 108 +4.23 +2.04 +14.55 +12.15
TP 5.210 x 106 5.367 x 106 9.991 x 106 8.673 x 106 -47.85 -46.28 -39.93 -38.12
150
A surprising finding was that catchments long distances away from the
Chesapeake Bay delivered significant amounts of TN and TP (Roberts and Prince, 2010).
Catchments in the headwaters of the Potomac Basin in western MD near Cumberland and
northeastern WV and in the middle Susquehanna Basin near Scranton/Wilkes Barre (PA)
were estimated to deliver > 18 kg/ha/yr of TN to the Chesapeake Bay (Figure 5.6c).
Similarly, catchments in the headwaters of the Susquehanna Basin near Elmira (NY), in
the middle Susquehanna Basin near Scranton/Wilkes-Barre (PA), and in the upper James
Basin near Charlottesville (VA) were estimated to deliver > 0.99 kg/ha/yr of TP to the
Chesapeake Bay (Figure 5.6d). Seemingly, once nutrients were able to leave these
catchments, most were delivered to the Chesapeake Bay. This indicated that larger
streams can be pipelines to the estuary with minimal attenuation by stream and reservoirs.
The estimated 2030 applications of manure to cropland and pasture may be
greater in regions of the Chesapeake Bay watershed with larger combined animal feeding
operations (CAFOs). According to the Gollehon et al. (2001), manure nutrient
production exceeded the assimilative capacity of the land in the watershed in south
central NY and southern and southeastern PA in the upper and middle Susquehanna
Basin, the eastern shore of MD, DE, and in west central VA in the upper James and
Potomac Basins. In these areas, a surplus of 2.0 to 5.1 x 105 tons of dry manure was
produced. Thus, the annual estimated mean rate of 9.79 and 52.74 kg/ha/yr manure N for
cropland and pasture and 5.56 and 23.77 kg/ha/yr manure P for cropland and pasture used
for 2030 projections (Roberts et al., 2009) may underestimate the true future values in
these regions after LC/LU changes. Additionally, throughout the rest of the watershed
151
with smaller or no CAFOs, estimated rates may be too high for 2030 after urbanization.
In this scenario, findings gathered and presented in Roberts et al. (2009) may be too low.
5.5 Management and Policy Recommendations
5.5.1 The case for using more comprehensive LC/LU effects to estimate nutrient
loadings
Effects of LC/LU composition on TN and TP loadings to the Chesapeake Bay
estuary are being used by local, state, and federal entities (Prince Georges County, MD
and the City of Bowie (MD), 2003; Commonwealth of Pennsylvania, 2003;
Commonwealth of Virginia, 2005; West Virginia Tributary Strategy Stakeholders
Working Group, 2005; and United States Geological Survey (Senus et al., 2005)).
However, findings presented by Roberts and Prince (2010) and Roberts et al. (2009)
indicated that analysis of LC/LU should also consider landscape configuration and spatial
arrangement to capture more comprehensively the effects of LC/LU. Roberts and Prince
(2010) indicated that the use of compositional and configurational metrics of LC/LU
slightly improved the accuracy between predicted and observed loadings throughout the
watershed and precision of nutrient loading estimates to the estuary (Tables 5.3-4).
At the local and state level, several mechanistically-based, nutrient loading
models that include the effects of LC/LU composition on nutrient loadings have been
implemented. These include HSPF within tributaries of the York Basin (Im et al., 2003),
the Generalized Watershed Loading Function (GWLF) on the eastern shore of MD (Lee
et al., 2000, 2001), the Soil Water Assessment Tool (SWAT) model in the
Rappahannock Basin (Meng et al., 2009), and SWAT and the ANNualized AGricultural
Non Point Source Pollution model (AnnAgNPS) on the eastern shore of MD (Ali et al.,
152
2007). At the federal level, the CBP and USEPA are using HSPF to manage the effects
of LC/LU change on the watershed-wide nutrient loadings (Chesapeake Community
Modeling Program, 2009). As continuous (hourly and/or daily) models for the long-term
simulation of nutrients, only LC/LU composition is implemented to calculate TN and TP
loadings in watersheds (Borah and Bera, 2003). By incorporating landscape
configuration and spatial arrangement at the catchment and riparian stream buffer scale,
improved agreement between all of these models may occur.
5.5.2 The case for targeting and prioritizing non-urban LC/LU for urbanization
As a management tool at localized scales, specific LC/LU types should be
targeted in the future for urbanization to reduce projected nutrient loadings to the
Chesapeake Bay. Even in remote reaches of the watershed, locally-dependent land
development can impact TN and TP loadings that reach the estuary. As previously
indicated, numerous catchments that were long distances away from the Chesapeake Bay
still delivered high TN and TP yields (Roberts and Prince, 2010). Thus, a second policy
recommendation is for municipalities at the local (cities and counties) level to inventory
and prioritize all non-urban LC/LUs for urban conversion prior to the start of future
development. Based on findings from Roberts and Prince (2010) and Roberts et al.
(2009, submitted) non-urban LC/LUs that should be targeted and prioritized for future
urbanization are cropland then pasture, followed by extractive and barren land.
Development of cropland may have dual impacts in regards to reducing nutrient
runoff. Results outlined in Roberts and Prince (2010) and Roberts et al. (2009,
submitted) demonstrated cropland?s unique function as: 1) a non-point N and P source
through the application of fertilizer and manure and as 2) a land-to-water delivery
153
variable for non-point N resulting from reduced hydraulic conductivity properties that
allowed for N in overland and shallow subsurface runoff to reach streams. Thus,
conversion of cropland may have multiple benefits in regards to reducing eutrophication
in the estuary.
Following cropland, pasture should be prioritized as the next LC/LU targeted for
urban conversion due to its vast potential for fertilizer and manure-based, non-point
source nutrient genesis (Roberts et al., 2009; Roberts and Prince 2010). Furthermore,
using the results from Roberts et al. (submitted), forecasted sprawl in catchments was
shown to be most beneficial (as judged by TN reductions) if allowed to replace cropland
and pasture in smaller developed patches that are unattached to existing urban. In
catchments with forecasted sprawl onto or near existing urban, gains in TN loading were
projected due to smaller reductions in cropland and pasture (Roberts et al., submitted)
Extractive land is the third LC/LU type that should be prioritized for development
due to its non-point N contribution to Chesapeake Bay from overland runoff processes
(Roberts and Prince, 2010). Barren land is the fourth LC/LU to be targeted for future
urbanization. Although barren land within the riparian stream buffer was a significant
land-to-water delivery variable for non-point P to Chesapeake Bay streams, it was not a
significant transport mechanism at the catchment scale (Roberts and Prince, 2010).
Additionally, studies have indicated that revegetating barren land can counteract its
runoff transporting potential (Ziegler and Giambelluca, 1998). To attenuate overland
runoff from riparian barren lands, management activities should focus on revegetation.
154
Thus, by prioritizing development initiatives to the conversion of targeted cover
types and the strategic placement of growth, a reduction in adverse impacts on projected
nutrient loadings to the estuary can be expected.
5.6 Future directions
One important variable that was not evaluated in these studies that could affect
urban-induced LC/LU change on nutrient loadings to the Chesapeake Bay was climate
change. Climate change, (such as increased temperature, precipitation, and
evapotranspiration) that is predicted for the region by the Intergovernmental Panel on
Climate Change (IPCC) (Goldman, 2006) has not been used in forecasts of nutrient
runoff, mainly because changes in these were insignificant annually (Preston and
Brakebill, 1999; Brakebill et al., 2001; Brakebill and Preston, 2004).. In these existing
studies using previous Chesapeake Bay SPARROW models, the authors found no
correlation between climatic variables and nutrient loadings using the current ranges of
conditions. This can be expected because of the low range of interannual variability in
precipitation and temperature throughout the extent of the watershed.
Nevertheless over the time span considered here (30 years), significant climate
changes are expected. Since precipitation is the primary mechanism transporting
nutrients from their source to streams in watersheds (Wolfe, 2001), the forecasted
increase in precipitation in the Chesapeake Bay watershed may lead to detectable changes
in nutrient runoff in this time frame. Alexander et al. (1996) suggested that with current
TN in the watershed, loadings to the estuary could increase with higher precipitation.
155
Thus, the addition of climate change to obtain a comprehensive forecast of
nutrient loadings to the estuary 30 years into the future is warranted. This integration of
climate change and LC/LU variables could be accomplished if climatic variables and
derivatives, such as precipitation, evaporation, temperature, or even soil moisture of the
magnitude anticipated are incorporated within the RESAC 31 m SPARROW models
presented in Roberts and Prince (2010) and Roberts et al., (2009, submitted). In addition
to annual effects, these variables should be evaluated with seasonal SPARROW models
to determine significance of seasonality changes.
Seasonal SPARROW models could separate nutrient loading responses in the
watershed based upon the four seasons (three month intervals) and could quantify the
impacts of climatic variables that could change watershed-wide, such as snow pack
(winter), snow melt and thaw (spring), convective thunderstorms (summer), and soil
moisture recharge (autumn). In the watershed, it was already determined that on average
the highest volume of runoff occurs during the spring months of March to May (Preston
and Brakebill, 1999). Changes in management practices applied to agricultural lands,
such as spring tillage to coincide with times of increased runoff, could also be evaluated
in these seasonal models. Furthermore, all other significant variables in these models that
are subject to changes could be collected and evaluated seasonally as well. Thus, through
the use of seasonal models, landscape and climate change could be potentially analyzed
in unison to quantify the comprehensive effects of anthropogenic transformation in the
Chesapeake Bay and its watershed for better management practices.
156
Bibliography
Ali, S., Yoon, K., Graff, C., McCarty, G., McConnell, L., Shirmohammadi, A., Hively,
D. W., and Sefton, K., 2007. Comparison of SWAT and AnnAGNPS applications to a
sub-watershed within the Chesapeake Bay watershed in Maryland. 4th International
SWAT Conference Paper.
Allan, C. J. Roulet, N. T., and Hill, A. R., 1993. The biogeochemistry of pristine,
headwater Precambrian Shield watersheds: An analysis of material transport in a
heterogeneous landscape. Biogeochemistry 22: 37-79.
Allan, C. J. and Roulet, N. T., 1994. Runoff generation in zero-order Precambrian
Shield catchments: The stormflow response of a heterogeneous landscape. Hydrological
Processes 8: 369-388.
Alexander, R. B., Murdoch, P. S., and Smith, R. A., 1996. Streamflow induced
variations in nitrate flux in tributaries to the Atlantic coastal zone. Biogeochemistry 33:
149-177.
Alexander, R. B., Smith, R. A., and Schwarz. G. E., 2000. Effect of stream channel size
on the delivery of nitrogen to the Gulf of Mexico. Nature 403: 758-761.
Alexander, R. B., Johnes, P. J., Boyer, E. W., and Smith, R. A., 2002a. A comparison of
models for estimating the riverine export of nitrogen from large watersheds.
Biogeochemistry 57-58: 295-339.
Alexander, R. B., Elliot, A. H., Shankar, U., and McBridge, G. B., 2002b. Estimating
the sources and transport of nutrients in the Waikato River Basin, New Zealand. Water
Resources Research 38(12): 1-23.
Alexander, R. B., Smith, R. A. and Schwarz, G. E., 2004. Estimates of diffuse
phosphorus sources in surface waters of the United States using a spatially referenced
watershed model. Water Science and Technology 49(3): 1-10.
Anbumozhi, V., Radhakrishnan, J., and Yamaji, E., 2005. Impact of riparian buffer
zones on water quality and associated management considerations. Ecological
Engineering 24: 517-523.
Arnold Jr., C. L. and Gibbons, C. J., 1996. Impervious surface coverage: The emergence
of a key environmental indicator. Journal of the American Planning Association 62(2):
243-258.
Arthur-Hartranft, S. T., Carlson, T. N., and Clarke, K. C., 2003. Satellite and ground-
based microclimate and hydrologic analyses coupled with a regional urban growth model.
Remote Sensing of Environment 86(3): 385-400.
157
Assouline, S. and Mualem, Y., 2002. Infiltration during soil sealing: The effect of areal
heterogeneity of soil hydraulic properties. Water Resources Research 38(12): 1-9.
Babbitt, B., 2007. The case for conditioning federal infrastructure investment on state
land and water planning. Journal of the American Planning Association 73: 146-148.
Baker, M. E., Weller, D. E., and Jordan, T. E., 2006. Improved methods for quantifying
potential nutrient interception by riparian buffers. Landscape Ecology 21: 1327-1345.
Bannerman, R. T., Owens, D. W., Dodds, R. B., and Hornewer, N. J.,. 1993. Sources of
pollutants in Wisconsin stormwater. Water Science and Technology 28(3-5): 241-259.
Bartlett, J. G., Mageean, D. M., and O'Connor, R. J., 2000. Residential expansion as a
continental threat to U.S. coastal ecosystems. Population Environment 21(5): 429-469.
Bhaduri, B., Harbor, J., Engel, B., and Grove, M., 2000. Assessing watershed-scale
long-term hydrologic impacts of land-use change using a GIS-NPS model.
Environmental Management 26(6): 643-658.
Bicknell, B. R., Donigian, Jr., A. S., Jobes, T. H., and Chinnaswamy, R. V., 1996.
Modeling nitrogen cycling and export in forested watersheds using HSPF. Prepared for
U.S. Geological Survey, Water Resources Division, Reston, VA, and U.S. Environmental
Protection Agency, National Environmental Research Laboratory, Ecosystems Research
Division, Athens, GA.
Bicknell, B. R., Imhoff, J. C., Kittle, Jr., J. L., Jobes, T. H., and Donigian, Jr, A. S.,
2001. Hydrological Simulation Program-Fortran (HSPF). User?s Manual for Release 12.
U.S. Environmental Protection Agency, National Environmental Research Laboratory,
Ecosystems Research Division, Athens, GA, in cooperation with U.S. Geological Survey,
Water Resources Division, Reston, VA.
Boesch, D. F., Brinsfield, R. B., and Magnien, R. E., 2001. Chesapeake Bay
eutrophication: Scientific understanding, ecosystem restoration, and challenges for
agriculture. Journal of Environmental Quality 30: 303-320.
Boesch, D. F., 2006. Scientific requirements for ecosystem-based management in the
restoration of Chesapeake Bay and Coastal Louisiana. Ecological Engineering 26: 6-26.
Borah, D. K., and Bera, M., 2003. Watershed-scale hydrologic and nonpoint-source
pollution models: Review of mathematical bases. Transactions of the ASAE 46(6):
1553-1566.
158
Bowen, J. L., and Valiela, I., 2001. The ecological effects of urbanization of coastal
watersheds: Historical increases in nitrogen loads and eutrophication of Waquoit Bay
estuaries. Canadian Journal of Fisheries and Aquatic Sciences 58: 1489-1500.
Brakebill, J. W., Preston, S. D., and Martucci, S. K., 2001. Digital data used to relate
nutrient inputs to water quality in the Chesapeake Bay, Version 2.0. U.S. Geological
Survey Open-File Report OFR-01-251.
Brakebill, J. W. and Preston, S. D., 2004. Digital data used to relate nutrient inputs to
water quality in the Chesapeake Bay, Version 3.0. U.S. Geological Survey Open-File
Report OFR-2004-1433.
Bratton, J. F., Colman, S. M. and Seal II, R. R., 2003. Eutrophication and carbon
sources in Chesapeake Bay over the last 2700 Years: Human impacts in context.
Geochimica et Cosmochimica Acta. 67: 3385-3402.
Breitburg, D. L., 1992. Episodic hypoxia in Chesapeake Bay: Interacting of
recruitment, behavior, and physical disturbance. Ecological Monographs 62(4): 525-
546.
Bruns, D. A., 2005. Macroinvertebrate response to land cover, habitat, and water
chemistry in a mining-impacted river ecosystem: A GIS watershed analysis. Aquatic
Sciences 67: 403-423.
Caccia, V. G. and Boyer, J. N., 2005. Spatial patterning of water quality in Biscayne
bay, Florida as a function of land use and water management. Marine Pollution Bulletin
50: 1416-1429.
Carle, M. V., Halpin, P. N., and Stow, C. A., 2005. Patterns of watershed urbanization
and impacts on water quality. Journal of the American Water Resources Association
41(3): 693-708.
Carlson, T. N. and Arthur, S. T., 2000. The impact of land use-land cover changes due
to urbanization on surface microclimate and hydrology: a satellite perspective. Global
and Planetary Change 25(1-2): 49-65.
Carlson, T. N., 2004. Analysis and prediction of surface runoff in an urbanizing
watershed using satellite imagery. Journal of the American Water Resources Association
40(4): 1087-1098.
Carpenter, S. R., Caraco, N. F., Correll, D. L., Howarth, R. W., Sharpley, A. N., and
Smith, V. H., 1998. Nonpoint pollution of surface waters with phosphorus and nitrogen.
Ecological Applications 8: 559-568.
159
Chandler, D. G. and Walter, M. F., 1998. Runoff responses among common land uses in
the uplands of Matalom, Leyte, Philippines. Transactions of the ASAE 41(6): 1635-
1641.
Chang, H., 2004. Water quality impacts of climate and land use changes in southeastern
Pennsylvania. The Professional Geographer 56(2): 240-257.
Chesapeake Bay Program, 2007.
ftp://ftp.chesapeakebay.net/Modeling/Change1985to2000.xls. Accessed December 7,
2007.
Chesapeake Bay Program, 2008a.
http://www.chesapeakebay.net/publication.aspx?publicationid=13112. Accessed on June
2, 2008.
Chesapeake Bay Program, 2008b.
http://www.chesapeakebay.net/populationgrowth.aspx?menuitem=14669. Accessed on
June 2, 2008.
Chesapeake Bay Program, 2008c.
http://www.chesapeakebay.net/developmentpressure.aspx?menuitem=19514. Accessed
on June 6, 2008.
Chesapeake Community Modeling Program, 2009.
http://ches.communitymodeling.org/models/CBPhase5/index.php. Accessed on October
6, 2009.
Chesapeake Community Modeling Program, 2008.
http://ches.communitymodeling.org//models/CBPhase5/datalibrary.php. Accessed on
June 6, 2008.
Chesapeake Executive Council. 1987. Chesapeake Bay Agreement. Annapolis,
Maryland.
Claggett, P. R., Jantz, C. A., Goetz, S. J., and Bisland, C., 2005. Assessing development
pressure in the Chesapeake Bay watershed: An evaluation of two land-use change
models. Environmental Monitoring and Assessment 94(1-3): 129-146.
Clarke, K. C., Hoppen, S., and Gaydos, L., 1997. A self-modifying cellular automaton
model of historical urbanization in the San Francisco Bay area. Environmental and
Planning B-Planning and Design 24: 247-261.
Clarke, K. C. and Gaydos, L. J., 1998. Loose-coupling a cellular automaton and GIS:
long-term urban growth prediction for San Francisco and Washington/Baltimore.
International Journal of Geographical Information Science 12(7): 699-714.
160
Cohn, T. A., Delong, L. L., Gilroy, E. J., Hirsch, R. M., and Wells, D. K., 1989.
Estimating constituent loads. Water Resources Research 25(5): 937-942.
Commonwealth of Pennsylvania. 2003. Pennsylvania?s Chesapeake Bay tributary
strategy goals for nutrient and sediment reduction and habitat restoration.
Commonwealth of Virginia. 2005. Chesapeake Bay nutrient and sediment reduction
tributary strategy.
Cornwell, J. C., Kemp, W. M., and Kana, T. M., 1999. Denitrification in coastal
ecosystems: Methods, environmental controls, and ecosystem level controls, a review.
Aquatic Ecology 33: 41-54.
Costanza, R., Voinov, A., Boumans, R., Maxwell, T., Villa, F., Wagner, L., and Voinov,
H., 2002. Integrated ecological economical modeling of the Patuxent River Watershed,
Maryland. Ecological Monographs 72(2): 203-231.
Conway, T. M. and Lathrop, R. G., 2005. Alternative land use regulations and
environmental impacts: assessing land use in an urbanizing watershed. Landscape and
Urban Planning 71: 1-15.
Correll, D. L., Jordan, T. E., and Weller, D. E., 1992. Nutrient flux in a landscape:
Effects of coastal land use and terrestrial community mosaic on nutrient transport to
coastal waters. Estuaries 15: 431-442.
Coulter, C. B., Kolka, R. K., and Thompson, J. A., 2004. Water quality in agricultural,
urban, and mixed land use watersheds. Journal of the American Water Resources
Association 40(6): 1593-1601.
Cummins, J., 2004. Think septics are always bad? Then you don't know sewage. Small
Flows Quarterly 5(4): 17.
Dale, V. H., Akhtar, F., Aldridge, M., Baskaran, L., Berry, M., Browne, M., Chang, M.,
Efroymson, R., Garten, Jr., C., Lingerfelt, E., and Stewart, C., 2008. Modeling the
effects of land use on the quality of water, air, noise, and habitat for a five-county region
in Georgia. Ecology and Society 13: 10.
Dietzel, C., and Clarke, K. C., 2004. Replication of spatio-temporal land use patterns at
three levels of aggregation by an urban cellular automata. Cellular Automata,
Proceedings Lecture Notes in Computer Science 3305:523-532.
Dougherty, M., Dymond, R. L., Goetz, S. J., Jantz, C. A., and Goulet, N., 2004.
Evaluation of impervious surface estimates in a rapidly urbanizing watershed.
Photogrammetric Engineering and Remote Sensing 70(11): 1275-1284.
161
Erickson, J. E., Cisar, J. L., Snyder, G. H., and Volin, J. C., 2005. Phosphorus and
potassium leaching under contrasting landscape models established on a sandy soil. Crop
Science 45: 546-552.
Federal Interagency Stream Restoration Working Group. 1998. Stream Corridor
Restoration: Principles, Processes, and Practices. GPO Item No. 0120-A; SuDocs No. A
57.6/2:EN 3/PT.653. ISBN-0-934213-59-3.
Filoso, S., Vallino, J., Hopkinson, C., Rastetter, E., and Claessens, L., 2004. Modeling
nitrogen transport in the Ipswich River Basin, Massachusetts, using a hydrological
simulation program in FORTRAN (HSPF). Journal of the American Water Resources
Association 40(5): 1365-1384.
Fisher, T. R., Hagy III, J. D., Boynton, W. R., and Williams, M. R., 2006. Cultural
eutrophication in the Choptank and Patuxent estuaries of Chesapeake Bay. Limnography
and Oceanography 51(1): 435-477.
Giambelluca, T. W., 2002. Hydrology of altered forest. Hydrological Processes 16:
1665-1669.
Gilbert, J. K., and Clausen, J. C., 2006. Stormwater runoff quality and quantity from
asphalt, paver, and crushed stone driveways in Connecticut. Water Research 40: 826-
832.
Gilliam, O., 2002. The Limitless City: A Primer on the Urban Sprawl Debate. Island
Press, Washington, District of Columbia, USA.
Goetz, S. J., Wright, R., Smith, A. J., Zinecker, E., and Schaub, E., 2003. Ikonos
imagery for resource management: tree cover, impervious surfaces and riparian buffer
analyses in the mid-Atlantic region. Remote Sensing of Environment 88: 195-208.
Goetz S. J., Jantz, C. A., Prince, S. D., Smith, A. J., Wright, R., and Varlyguin, D.,
2004a. Integrated analysis of ecosystem interactions with land use change: the
Chesapeake Bay watershed. Ecosystems and Land Use Change, pp. 263-275. Eds.
DeFries, R. S., Asner, G. P., and Houghton, R. A., Geophysical Monograph Series,
American Geophysical Union, Washington, D.C., USA.
Goetz, S. J., Varlyguin, D., Smith, A. J., Wright, R. K., Jantz, C. A., Tringe, J., Prince, S.
D., Mazzacato, M. E., and Melchoir, B., 2004b. Application of multitemporal Landsat
data to map and monitor land cover and land use change in the Chesapeake Bay
watershed. Analysis of Multi-temporal Remote Sensing Images, pp. 223-232. Eds.
Smits, P., and Bruzzone, L., World Scientific Publishers, Singapore, China.
Goldman, E., 2005. Nitrogen?s underground passage. Chesapeake Quarterly 4(1): 12-
14.
162
Goldman, E., 2006. Model forecasts for a warming watershed. Chesapeake Quarterly
5(3): 10.
Gollehon, N., Caswell, M., Ribaudo, Kellogg, R., Lander, C., and Letson. D., 2001.
Confined animal production and manure nutrients. AIB-771, United States Department
of Agriculture, Economic Research Service.
Groffman, P. M., Boulware, N. J., Zipper, W. C., Pouyat, R. C., Band, L. E., and
Colosimo, M. F., 2002. Soil nitrogen cycle processes in urban riparian zones.
Environmental Science and Technology 36: 4547-4552.
Groffman, P. M., Law, N. E., Belt, K. T., Band, L. E., and Fisher, G. T., 2004. Nitrogen
loads and retention in urban watershed ecosystems. Ecosystems 7: 393-403.
Guebert, M. D., and Gardner, T. W., 2001. Macropore flow on a reclaimed surface
mine: infiltration and hillslope hydrology. Geomorphology 39: 151-169.
Hassett, B., Palmer, M., Bernhardt, E., Smith, S., Carr, J., and Hart, D., 2005. Restoring
watersheds project by project: Trends in Chesapeake Bay tributary restoration. Frontiers
in Ecology and the Environment 3(5): 259-267.
Hayashi, M., and Rosenberry, D. O., 2002. Effects of groundwater exchange on the
hydrology and ecology of surface water. Ground Water 40: 309-316.
Herold, M., Goldstein, N. C., and Clarke, K. C., 2003. The spatiotemporal form of urban
growth: Measurement, analysis, and modeling. Remote Sensing of Environment 86(3):
286-302.
Homer, C., Huang, C., Yang, L., Wylie, B., and Coan, M., 2004. Development of a
2001 National Landcover Database for the United States. Photogrammetric Engineering
and Remote Sensing 70(7): 829-840.
Im, S., Brannan, K. M., and Mostaghimi, S., 2003. Simulating hydrologic and water
quality impacts of an urbanizing watershed. Journal of the American Water Resources
Association 39(6): 1465-1479.
Jansson, A., and Colding, J., 2007. Tradeoffs between environmental goals and urban
development: The case of nitrogen load from Stockholm County to the Baltic Sea.
Ambio 36: 650-656.
Jantz, C. A., Goetz, S. J., and Shelley, M. K., 2004. Using the SLEUTH urban growth
model to simulate the impacts of future policy scenario on urban land use in the
Baltimore-Washington metropolitan area. Environmental and Planning B-Planning and
Design 31(2): 251-271.
163
Jantz, C. A. and Goetz, S. J., 2005. Analysis of scale dependencies in an urban land-use-
change model. International Journal of Geographical Information Science 19(2): 217-
241.
Jantz, P., Goetz, S., and Jantz, C., 2005. Urbanization and loss of resources land in the
Chesapeake Bay watershed. Environmental Management 36(6): 805-825.
Jantz, C. A., Goetz, S. J., Donato, D., and Claggett, P., In press. Designing and
implementing a regional urban modeling system using the SLEUTH cellular urban
model. Computer, Environment, and Urban Systems.
Jennings, D. B., and Jarnagin, S. T., 2002. Changes in anthropogenic impervious
surfaces, precipitation, and daily streamflow discharge: A historical perspective in a mid-
Atlantic subwatershed. Landscape Ecology 17: 471-489.
Jones, K. B., Neale, A. C., Nash, M. S., Van Remortel, R. D., Wickham, J. D., Riitters,
K. H., and O?Neil, R. V., 2001. Predicting nutrient and sediment loadings to streams
from landscape metrics: A multiple study from the United States Mid-Atlantic region.
Landscape Ecology 16: 301-312.
Jordan, T. E., Correll, D. L., and Weller, D. E., 1993. Nutrient interception by a riparian
forest receiving inputs from croplands. Journal of Environmental Quality 26: 836-848.
Jordan, T. E., Correll, D. L., and Weller, D. E., 1997. Relating nutrient discharges from
watersheds to land use and stream variability. Water Resources Research 33: 2579-2590
Keitt, H., Urban, D. L., and Milne, B. T., 1997. Detecting critical scales in fragmented
landscapes. Conservation Ecology 1(1): 4. Available online:
www.consecol.org/vol1/iss/art4.
King, R. S., Baker, M. E., Whigham, D. F., Weller, D. E., Jordan, T. E., Kazyak, P. F.,
and Hurd, M. K., 2005. Spatial considerations for linking watershed land cover to
ecological indicators in streams. Ecological Applications 15(1): 137-153.
Kvam, P. H., and Vidakovic, B., 2007. Nonparametric Statistics with Applications to
Science and Engineering. Wiley, John & Sons, Inc., Hoboken, New Jersey, USA.
Langland, M., and Cronin, T., 2003. A summary report of sediment processes in
Chesapeake Bay and watershed. U.S. Geological Survey Water Resources Investigations
Report 03-4123.
Law, N. E., Band, L. E., and Grove, J. M., 2004. Nitrogen input from residential lawn
care practices in suburban watersheds in Baltimore County, MD. Journal of
Environmental Planning and Management 47(5): 737-755.
164
Lee, K. Y., Fisher, T. R., Jordan, T. E., Correll, D. L., and Weller, D. E., 2000.
Modeling the hydrochemistry of the Choptank River Basin using GWLF and Arc/Info: 1.
Model calibration and validation. Biogeochemistry 49(2): 143-173.
Lee, K. Y., Fisher, T. R., and Rochelle-Newall, E., 2001. Modeling the hydrochemistry
of the Choptank River Basin using GWLF and Arc/Info: 2. Model validation and
application. Biogeochemistry 56(3): 311-348.
Leitao, A. B., Miller, J., Ahern, J., and McGarigal, K.,. 2006. Measuring Landscapes: A
Planner?s Handbook. Island Press, Washington, District of Columbia, USA.
Line, D. E., White, N. M., Osmond, D. L., Jennings, G. D., and Mojonnier, C. B., 2002.
Pollutant export from various land uses in the Upper Neuse River basin. Water
Environment Research 74(1): 100-108.
Linker, L. C., Shenk, G. W., Dennis, R. L., and Sweeney, J. S., 2000. Cross-media
models of the Chesapeake Bay watershed and airshed. Water Quality and Ecosystems
Modeling 1(1-4): 91-122.
Logsdon, S. D., and Jaynes, D. B., 1996. Spatially variability of hydraulic conductivity
in a cultivated field at different times. Soil Science Society Journal of America 60: 703-
709.
Luke, D. A., 2004. Multilevel Modeling: Series: Quantitative Applications in the Social
Sciences. Sage Publications, Inc, Thousand Oaks, California, USA.
Lopez, R. C., Nash, M. S., Heggem, D. T., and Ebert, D. W., 2008. Watershed
vulnerability predictions for the Ozarks using landscape models. Journal of
Environmental Quality 37: 1769-1780.
Malone, T. C., Boynton, W., Horton, T., and Stevenson, C., 1993. Nutrient loading to
surface waters: Chesapeake case study. Keeping pace with science and engineering, pp.
8-38. Eds. Uman, M. F., National Academy Press, Washington, District of Columbia,
USA.
McConnell, R. L., 1995. The human population carrying capacity of the Chesapeake
Bay Watershed: A preliminary analysis. Population and Environment 16(4): 335-351.
McMahon, G., Alexander, R. B., and Qian, S., 2003. Support of total maximum daily
load programs using spatially referenced regression models. Journal of Water Resources
Planning and Management 129(4): 315-323.
McGarigal, K., Cushman, S. A., Neel, M. C., and Ene, E., 2002. FRAGSTATS: Spatial
Pattern Analysis Program for Categorical Maps. Computer software program produced
by the authors at the University of Massachusetts, Amherst. Available at the following
web site: www.umass.edu/landeco/research/fragstats/fragstats.html.
165
Meng, H., Sexton, A., Maddox, M., Sood, A., Murtugudde, R., Brown, C., and Ferraro,
R., 2009. Modeling the Chesapeake Bay watershed using SWAT. 5th International
SWAT Conference Paper.
Minton, G., 2002. Stormwater Treatment: Biological, Chemical and Engineering
Principles. Resource Planning Associates, Seattle, Washington, USA.
Moore, R.B., Johnston, C. M., Robinson, K. W., and Deacon, J. R., 2004. Estimation of
total nitrogen and phosphorus in New England streams using spatially referenced
regression models. U.S. Geological Survey Scientific Investigations Report 2004-5012
Morgan, C., and Owens, N., 2001. Benefits of water quality policies: The Chesapeake
Bay. Ecological Economics 39: 271-284.
Neff, R., Chang, H., Knight, C. G., Najjar, R. G., Yarnal. B., and Walker, H. A., 2000.
Impact of climate variation and change on Mid-Atlantic region hydrology and water
resources. Climate Change 14: 207-218.
Norton, M. M., and Fisher, T. R., 2000. The effects of forest on stream water quality in
two coastal plain watersheds of the Chesapeake Bay. Ecological Engineering 14: 337-
362.
Paerl, H. W., 2006. Assessing and managing nutrient-enhanced eutrophication in
estuarine and coastal waters; Interactive effects of human and climatic perturbations.
Ecological Engineering 26: 40-54.
Paul, M. J., and Meyer, J. L., 2001. Streams in the urban landscape. Annual Review of
Ecological Systems 32: 333-365.
Pellerin, B. A., Kaushal, S. S., and McDowell, W. H., 2006. Does anthropogenic
nitrogen enrichment increase organic nitrogen concentrations in runoff from forested and
human-dominated watersheds? Ecosystems 9: 852-864.
Perkins, D. B., Haws, N. W., Jawitz, J. W., Das, B. S., and Rao, P. S. C., 2007. Soil
hydraulic properties as ecological indicators in forested watersheds impacted by
mechanized military training. Ecological Indicators 7: 589-597.
Preston, S. D., and Brakebill, J. W., 1999. Application of spatially referenced regression
modeling for the evaluation of total nitrogen loading in the Chesapeake Bay Watershed.
U.S. Geological Survey Water-Resources Investigations Report 99-4054.
Prince George?s County, Maryland and City of Bowie, Maryland. 2003. Western
Branch watershed characterization: In support of Prince George?s County and the City of
Bowie?s watershed restoration action strategy for the Western Branch watershed.
166
Ritter, J. B., and Gardner, T. W., 1993. Hydrologic evolution of drainage basins
disturbed by surface mining, central Pennsylvania. Geological Society of America
Bulletin 105: 101-115.
Roberts, A. D., and Prince, S. D., 2010. Effects of urban and non-urban land cover on
nitrogen and phosphorus runoff to Chesapeake Bay. Ecological Indicators 10(2): 459-
474.
Roberts, A. D., Prince, S. D., Jantz, C. A., and Goetz, S. J., 2009. Effects of projected
future urban land cover on nitrogen and phosphorus runoff to the Chesapeake Bay.
Ecological Engineering 35(12): 1758-1772.
Roberts, A. D., Prince, S. D., Jantz, C. A., and Goetz, S. J., Submitted. Effects of future
urban sprawl on nitrogen runoff in subwatersheds of Chesapeake Bay. Ecological
Engineering.
Rushton, B. T., 2001. Low-impact parking lot design reduces runoff and pollutant loads.
Journal of Water Resources Planning and Management 127(3): 172-179.
Osmond, D. L., and Hardy, D. H.,. 2004. Landscape and watershed processes:
characterization of turf practices in five North Carolina communities. Journal of
Environmental Quality 33: 565-575.
Schwarz, G. E., Hoos, A. B., Alexander, R. B., and Smith, R. A., 2006. Techniques and
methods Book 6, Modeling techniques, Section B, Surface water, Chapter 3, The
SPARROW surface water-quality model-Theory, application, and user Documentation:
U.S. Geological Survey, Techniques and methods 6 B-3, XX p. + CD-ROM.
Senus, M. P., Langland, M. J., and Moyer, D. L., 2005. Nutrient and sediment
concentrations, loads, and trends in four nontidal tributaries in the Chesapeake Bay
watershed, 1997-2001. Scientific Investigations Report 2004-5125.
Shinya, M., Tsuruho, K., Konishi, T., and Ishikawa, M., 2003. Evaluation of factors
influencing diffusion of pollutant loads in urban highway runoff. Water Science and
Technology 47(7-8): 227-232.
Shuyler, L. R., Linker, L. C., and Walters, C. P., 1995. The Chesapeake Bay story: The
science behind the program. Water Science and Technology 31(8): 133-139.
Silva, E. A., and Clarke, K. C., 2002. Calibration of the SLEUTH urban growth model
for Lisbon and Porto, Portugal. Computers, Environment, and Urban Systems 26:525-
552.
Sims, J. T., and Coale, F. J., 2002. Solutions to nutrient management problems in the
Chesapeake Bay watershed, USA. Agriculture, Hydrology, and Water Quality, pp. 345-
371. Eds. P. M. Haygarth and S. C. Jarvis. CABI Publishing, Wallingford, Oxon, UK.
167
Smith, R. A., Schwarz, G. E., and Alexander, R. B., 1997. Regional interpretation of
water-quality monitoring data. Water Resources Research 33(12): 2781:2798.
Smith, R. A., Alexander, R. B., and Schwarz, G. E., 2003. Natural background
concentrations of nutrients in streams and rivers in the conterminous United States.
Environmental Science and Technology 37(14): 3039-3047.
Solecki, W. D., and Oliveri, C., 2004. Downscaling climate change scenarios in an
urban land use change model. Journal of Environmental Management 72(1-2): 105-115.
Sonada, K., Yeakley, J. A., and Walker, C. E., 2001. Near-stream land use effects on
streamwater nutrient distribution in an urbanizing watershed. Journal of the American
Water Resources Association 37(6): 1517-1532.
Stow, C. A., Borsuk, M. E., and Stanley, D. W., 2001. Long-term changes in watershed
nutrient inputs and riverine exports in the Neuse River, North Carolina. Water Resources
35(6): 1489-1499.
Susquehanna River Basin Commission, 2008. http://www.srbc.net/about/geninfo.htm.
Accessed on July 31, 2008.
Tang, Z., Engel, B. A., Pijanowski, B. C., and Lim, K. J., 2003. Assessing the impact of
urbanization on long term runoff and NPS pollution in Muskegon river watershed.
American Society of Agricultural Engineers Annual Meeting Paper 032195.
Tang, Z., Engel, B. A., Pijanowski, B. C., and Lim, K. J., 2005. Forecasting land use
change and its environmental impact at a watershed scale. Journal of Environmental
Management 76: 35-45.
Tsegaye, T., Sheepard, D., Islam, K. R., Johnson, A., and Tadesse, W., 2006.
Development of chemical index as a measure of in-stream water quality in response to
land-use and land-cover changes. Water, Air, and Soil Pollution 174: 161-179.
Tsihrintzis, V. A., Fuentes, H. R., and Gadipudi, R. K., 1996. Modeling prevention
alternatives for non-point source pollution at a wellfield in Florida. Water Resources
Bulletin 32(2): 317-331.
Tu, J., Xia, Z. G., Clarke, K. C., and Frei, A., 2007. Impact of urban sprawl on water
quality in eastern Massachusetts, USA. Environmental Management 40: 183-200.
Tufford, D. L., Samarghitan, C. L., McKeller, H. N., Porter, D. E., Hussey, J. R., 2003.
Impacts of urbanization on nutrient concentrations in small southeastern coastal streams.
Journal of the American Water Resources Association 39: 301-312.
168
United States Environmental Protection Agency, 2009a.
http://cfpub.epa.gov/eroe/index.cfm?fuseaction=detail.viewPDF&ch=47&lShowInd=0&s
ubtop=200&lv=list.listByChapter&r=188222. Accessed on October 20, 2009.
United States Environmental Protection Agency, 2009b.
http://www.epa.gov/nrmrl/pubs/600r05149/600r05149hspf.pdf. Accessed on October 21,
2009.
United States Geological Survey, 2001.
http://md.water.usgs.gov/publications/press_release/2001/2001-01-05.html. Accessed on
February 28, 2008.
United States Geological Survey, 2008a. http://www.ncgia.ucsb.edu/projects/gig/.
Accessed on June 2, 2008.
United States Geological Survey, 2008b. http://water.usgs.gov/nawqa/wri94-
4250/h2.html. Accessed on June 6, 2008.
Verchot, L. U., Franklin, E. C., and Gilliam, J. W., 1997. Nitrogen cycling in piedmont
vegetated filter zones: I. surface soil processes. Journal of Environmental Quality 26:
327-336.
Wakida, F. T., and Lerner, D. N., 2005. Non-agricultural sources of groundwater nitrate:
a review and case study. Water Research 39: 3-16.
Wakida, F. T., and Lerner, D. N., 2006. Potential nitrate leaching to groundwater from
home building. Hydrological Processes 20: 2077-2081.
Wang, P., Batiuk, R., Linker, L., and Shenk, G.,. 2001. Assessment of best management
practices for improvements of dissolved oxygen in Chesapeake Bay estuary. Water
Science and Technology 44(7): 173-180.
Wang, P., Linker, L. C., Batiuk, R., and Cerco, C., 2006. Surface analysis of
Chesapeake Bay water quality response to different nutrient and sediment loads. Journal
of Environmental Engineering 132(2): 377-383.
Wang, Y., Choi, W., and Deal, B. M., 2005. Long-term impacts of land-use change on
non-point source pollutant loads for the St. Louis metropolitan area, USA.
Environmental Management 35(2): 194-205.
Ward, A. D., and Trimble, S. W., 2004. Environmental Hydrology: Second Edition.
Lewis Publishers, Boca Raton, Florida, USA.
Weller, K., 2003. Once and future bay. Chesapeake Futures: Choices for the 21st
Century, pp. 125-151. Eds. Boesch, D. F., and Greer, J., Chesapeake Research
Consortium, Inc, Edgewater, Maryland, USA.
169
West Virginia Tributary Strategy Stakeholders Working Group. 2005. West Virginia?s
Potomac tributary strategy.
Wickham, J. D., O'Neill, R. V., Riitters, K. H., Smith, E. R., Wade, T. G., and Jones, K.
B., 2002. Geographic targeting of increases in nutrient export due to future urbanization.
Ecological Applications 12(1): 93-106.
Wetzel, K. F., 2003. Runoff production processes in small alpine catchments in the
unconsolidated Pleistocene sediments of the Lainbach area (Upper Bavaria).
Hydrological Processes 17: 2463-2483.
Williams, M., Hopkinson, C., Rastetter, E., Vallino, J., and Claessens, L., 2005.
Relationships of land use and stream solute concentrations in the Ipswich River Basin,
Northeastern, Massachusetts. Water, Air, and Soil Pollution 161: 55-74.
Willems, H. P. L., Rotelli, M. D., Berry, D. F., Smith, E. P., Jr., Reneau, R. B., and
Mostaghimi, S., 1997. Nitrate removal in riparian wetland soils: Effects of flow rate,
temperature, nitrate concentration, and soil depth. Water Research 31: 841-849.
Wolfe, M. L., 2001. Hydrology. Agricultural Nonpoint Source Pollution, pp. 1-28.
Eds. Ritter, W. F., and Shirmohammadi., A., Lewis Publishers, Boca Raton, Florida,
USA.
Woods Hole Research Center, 2009.
http://www.whrc.org/midatlantic/modeling_change/SLEUTH/sleuth_calibration.htm.
Accessed on August 4, 2009.
Wynn, T. M., Mostaghimi, S., Frazee, J. W., McClellan, P. W., Shaffer, R. M., and Aust,
W. M., 2000. Effects of forest harvesting best management practices on surface water
quality in the Virginia coastal plain. Transactions of the ASAE 43: 927-936.
Xian, G. and Crane, M., 2005. Assessments of urban growth in the Tampa Bay
watershed using remote sensing data. Remote Sensing of Environment 97(2): 203-215.
Xian, G., Crane, M., and Steinwald, D., 2005. Dynamic modeling of Tampa Bay urban
development using parallel computing. Computers and Geosciences 7:920-928.
Yanai, R. D., 1992. Phosphorus budget of a 70-year old northern hardwood forest.
Biogeochemistry 17(1): 1-22.
Yang, X. and Lo, C. P., 2003. Modeling urban growth and landscape changes in the
Atlanta metropolitan area. International Journal of Geographical Information Science
17(5): 463-488.
170
Ziegler, A. D., and Giambelluca, T. W., 1997. Importance of rural roads as source areas
for runoff in mountainous areas of northern Thailand. Journal of Hydrology 196: 204-
209.
Ziegler, A. D., and Giambelluca, T. W., 1998. Influence of revegetation efforts on
hydrologic response and erosion Kaho'olawe Island, Hawaii. Land Degradation and
Development 9: 189-206.
Ziegler, A. D., Giambelluca, T. W., Sutherland, R. A., Vana, T. T., and Nullet, M. A.,
2001. Horton overland flow contribution to runoff on unpaved mountain roads: A case
study in northern Thailand. Hydrological Processes 15: 3203-3208.
Ziegler, A. D., Giambelluca, T. W., Tran, L. T., Vana, T. T., Nullet, M. A., Fox, J., Duc
Vien, T., Pinthong, J., Maxwell, J. F., and Evett, S., 2004a. Hydrological consequences
of landscape fragmentation in northern Vietnam: Evidence of accelerated overland flow
generation. Journal of Hydrology 287: 124-146.
Ziegler, A. D., Giambelluca, T. W., Sutherland, R. A., Nullet, M. A., Yarnasarn, S.,
Pinthong, J., Preechapanya, P., and Jaiaree, S., 2004b. Towards understanding the
cumulative impacts of roads in upland agricultural watersheds of northern Thailand.
Agriculture, Ecosystems, and Environment 104: 145-158.
Ziegler, A. D., Tran, L. T., Giambelluca, T. W., Sidle, R., Sutherland, R. A., Nullet, M.
A., and Duc Vien, T., 2006. Effective slope lengths for buffering hillslope runoff in
fragmented landscapes in northern Vietnam. Forest Ecology and Management 224: 104-
118.
Ziegler, A. D., Giambelluca, T. W., Plondke, D., Leisz, S., Tran, L. T., Fox, J., Nullet,
M. A., Vogler, J. B., Troung, D. M., and Duc Vien, T., 2007. Hydrological
consequences of landscape fragmentation in northern Vietnam: Buffering of Hortonian
overland flow. Journal of Hydrology 337: 52-67.
Zimmermann, B., Elsenbeer, H., and De Moraes, J. M.,. 2006. The influence of land-use
changes on soil hydraulic properties: Implications for runoff generation. Forest Ecology
and Management 222: 29-28.