  
 
 
 
 
 
ABSTRACT 
 
 
 
 
Title of Document: VARIABILITY OF TERRESTRIAL CARBON 
CYCLE AND ITS INTERACTION WITH 
CLIMATE UNDER GLOBAL WARMING 
  
 Haifeng Qian, Doctor of Philosophy, 2008 
Directed By: Professor Ning Zeng 
Department of Atmospheric and Oceanic Science
 
 
Land-atmosphere carbon exchange makes a significant contribution to the 
variability of atmospheric CO
2
 concentration on time scales of seasons to centuries. 
In this thesis, a terrestrial vegetation and carbon model, VEgetation-Global-
Atmosphere-Soil (VEGAS), is used to study the interactions between the terrestrial 
carbon cycle and climate over a wide-range of temporal and spatial scales.  
 The VEGAS model was first evaluated by comparison with FLUXNET 
observations. One primary focus of the thesis was to investigate the interannual 
variability of terrestrial carbon cycle related to climate variations, in particular to El 
Ni?o-Southern Oscillation (ENSO). Our analysis indicates that VEGAS can properly 
capture the response of terrestrial carbon cycle to ENSO: suppression of vegetative 
activity coupled with enhancement of soil decomposition, due to predominant warmer 
and drier climate patterns over tropical land associated with El Ni?o. The combined 
affect of these forcings causes substantial carbon flux into the atmosphere. A unique 
  
aspect of this work is to quantify the direct and indirect effects of soil wetness 
vegetation activities and consequently on land-atmosphere carbon fluxes. Besides this 
canonic dominance of the tropical response to ENSO, our modeling study simulated a 
large carbon flux from the northern mid-latitudes, triggered by the 1998-2002 drought 
and warming in the region. Our modeling indicates that this drought could be 
responsible for the abnormally high increase in atmospheric CO
2
 growth rate (2 
ppm/yr) during 2002-2003. 
 We then investigated the carbon cycle-climate feedback in the 21
st
 century. A 
modest feedback was identified, and the result was incorporated into the Coupled 
Carbon Cycle Climate Model Inter-comparison Project (C4MIP). Using the fully 
coupled carbon cycle-climate simulations from C4MIP, we examined the carbon 
uptake in the Northern High Latitudes poleward of 60?N (NHL) in the 21
st
 century. 
C4MIP model results project that the NHL will be a carbon sink by 2100, as CO
2
 
fertilization and warming stimulate vegetation growth, canceling the effect of 
enhancement of soil decomposition by warming. However, such competing 
mechanisms may lead to a switch of NHL from a net carbon sink to source after 
2100. All these effects are enhanced as a result of positive carbon cycle-climate 
feedbacks. 
 
 
 
  
 
 
 
VARIBILITY OF TERRESTRIAL CARBON CYCLE AND ITS INTERACTION 
WITH CLIMATE UNDER GLOBAL WARMING  
 
 
 
By 
 
 
Haifeng Qian 
 
 
 
 
 
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  
2008 
 
 
 
 
 
 
 
 
 
Advisory Committee: 
Professor Ning Zeng, Chair 
Professor Eugenia Kalnay 
Dr. G. James Collatz 
Professor Raghu Murtugudde 
Professor Christopher Justice 
 
 
 
 
 
 
 
 
  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
? Copyright by 
Haifeng Qian 
2008 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 ii 
 
Preface 
This dissertation has a seven-chapter layout. Chapter 1: Introduction; Chapter 
2: Evaluation of VEGAS against Observations; Chapter 3: Response of the Terrestrial 
Carbon Cycle to the El Ni?o-Southern Oscillation; Chapter 4: Impact of 1998?2002 
Midlatitude Drought and Warming on Terrestrial Ecosystem and the Global Carbon 
Cycle; Chapter 5: How Strong is Carbon Cycle-Climate Feedback under Global 
Warming? Chapter 6: Enhanced Terrestrial Carbon uptake in the Northern High 
Latitudes under Climate Change in the 21
st
 Century from C4MIP Models Projections; 
Chapter 7: Conclusion and Future Directions. Finally, Appendices containing the 
description of VEGAS and a glossary of the thesis follow. In the interest of future 
usability and readability of this dissertation, each of the main chapters may include its 
own introductory remarks, literature survey, data and methodology descriptions, and 
a short summary. 
Some chapters of this dissertation have been individually submitted, or 
published in scientific journals: 
 ? Qian, H., R. Joseph, and N. Zeng, 2008: Enhanced terrestrial carbon uptake in the 
Northern High Latitudes in the 21
st
 century from the C4MIP model projections. 
Submitted to Global Change Biol. 
 ? Qian, H., R. Joseph, and N. Zeng, 2008: Response of the terrestrial carbon cycle 
to the El Ni?o-Southern Oscillation, Tellus B, doi: 10.1111/j.1600-0889.2008.00 
360.x. 
 ?  Zeng, N., H. Qian, C. R?denbeck, and M. Heimann, 2005: Impact of 1998-2002 
Midlatitude drought and warming on terrestrial ecosystem and the global carbon 
cycle. Geophys. Res. Lett. 32, L2270910.1029/2005GL024607. 
 
 
 iii 
 
 ?  Zeng, N., H. Qian, E. Munoz, and R. Iacono, 2004: How strong is carbon cycle-
climate feedback under global warming? Geophys. Res. Lett. 31, L20203, doi:10. 
1029/2004GL020904. 
 
 
 
 iv 
 
 
Acknowledgements 
First of all, I would like to express my gratitude to my advisor Professor Ning 
Zeng for his valuable support, guidance, and patience over years. His optimism, 
patience and scientific supervision have fostered my enthusiasm for the research 
presented here. I would like to thank my committee members: Professor Eugenia 
Kalnay, Raghu Murtugudde, Christopher Justice, and Dr. G. James Collatz. I owe a 
great deal of gratitude to Dr. Jin-ho Yoon, Dr. Renu Joseph, and Brian Cook, with 
whom I have had very productive scientific discussions and collaborations.  
I want to extend thanks to all members of the Department of Atmospheric and 
Oceanic Science. Special thank goes to the staff in our department and the 
International Education Service of University of Maryland for providing help and 
solving my VISA related problems when I was delayed in China for half a year. I am 
also thankful to my friends at University of Maryland for the learning and personal 
experiences we have shared during all this time, especially to my classmates: Dr. 
Junjie Liu, Dr. Hong Li, Emily Becker etc. for our enjoyable lunchtime discussions, 
and Bin Guan, Can Li, Feng Niu etc. for their helpful discussions on coursework 
initially and research afterwards.  
Finally, I dedicate this dissertation to my beloved parents and my wife Fang 
Zhang for everything they did for me throughout the years. It is their love, support 
and encouragement that provided me with the opportunity to pursue higher education 
and shape my life here. 
 
 
 v 
 
Table of Contents 
 
 
Preface........................................................................................................................... ii 
Acknowledgements...................................................................................................... iv 
Table of Contents.......................................................................................................... v 
List of Tables .............................................................................................................. vii 
List of Figures............................................................................................................viii 
Chapter 1: Introduction................................................................................................. 1 
1.1 Importance of terrestrial carbon cycle in climate systems.................................. 1 
1.2 The interaction between the terrestrial carbon cycle and climate....................... 5 
1.2.1 Modeling the seasonal cycles of terrestrial carbon ecosystem .................... 6 
1.2.2 Modeling the interannual variability of terrestrial carbon cycle.................. 8 
1.2.3 Modeling the terrestrial carbon uptake under global warming.................. 11 
Chapter 2: Evaluation of VEGAS against Observations ............................................ 15 
2.1 Introduction....................................................................................................... 15 
2.2 Model description and setup ............................................................................. 16 
2.3 Seasonal and interannual variability of terrestrial carbon fluxes in the Amazon 
Basin???????? ....................................................................................... 20 
2.4 Seasonal and interannual variability of terrestrial carbon fluxes of boreal forest 
in Canada ................................................................................................................ 24 
2.5 Evaluation of the interannual variability of VEGAS LAI with NDVI ............. 25 
2.6 Summary........................................................................................................... 26 
Chapter 3: Response of the Terrestrial Carbon Cycle to the El Ni?o-Southern 
Oscillation................................................................................................................... 28 
3.1 Introduction....................................................................................................... 28 
3.2 Interannual variability of terrestrial carbon flux............................................... 32 
3.3 Robust features of the terrestrial response to ENSO ........................................ 35 
3.4 Sensitivity Simulations to quantify the impact of climatic factors................... 44 
3.5 Conclusion and discussion................................................................................ 52 
Chapter 4: Impact of 1998?2002 Midlatitude Drought and Warming on Terrestrial 
Ecosystem and the Global Carbon Cycle.................................................................... 55 
4.1 Introduction....................................................................................................... 55 
4.2 Drought and CO
2
............................................................................................... 57 
4.3 Regional contributions and mechanisms .......................................................... 59 
4.4 Conclusion and discussion................................................................................ 64 
Chapter 5: How Strong is Carbon Cycle-Climate Feedback under Global Warming?
..................................................................................................................................... 66 
5.1 Introduction....................................................................................................... 66 
5.2 Model and methodology ................................................................................... 66 
5.3 Carbon cycle-Climate feedback........................................................................ 70 
5.4 Uncertainties in land carbon response .............................................................. 72 
5.5 Discussion and conclusion................................................................................ 75 
Chapter 6: Enhanced Terrestrial Carbon uptake in the Northern High Latitudes under 
Climate Change in the 21
st
 Century from C4MIP Models Projections ...................... 79 
6.1 Introduction....................................................................................................... 79 
 
 
 vi 
 
6.2 Data and methodology ...................................................................................... 84 
6.3 NHL terrestrial carbon storage in the 21
st
 century............................................ 86 
6.4 CO
2
 fertilization and climate change effects in NHL carbon cycle-climate 
coupling................................................................................................................... 96 
6.5 Discussion and conclusion.............................................................................. 100 
Chapter 7: Conclusion and Future Directions........................................................... 105 
7.1 Summary......................................................................................................... 105 
7.1.1 Evaluation of VEGAS.............................................................................. 105 
7.1.2 Terrestrial carbon cycle in response to ENSO......................................... 106 
7.1.3 Impact of Midlatitude drought on the terrestrial ecosystem .................... 109 
7.1.4 Carbon cycle-climate feedback................................................................ 110 
7.1.5 Future carbon uptake in Northern High Latitudes ................................... 111 
7.2 Future directions ............................................................................................. 113 
7.2.1 Multi-decadal variability of terrestrial carbon ......................................... 113 
7.2.2 Soil moisture effect on the vegetation growth and soil respiration ......... 114 
7.2.3 Improvement of VEGAS ......................................................................... 115 
Appendices................................................................................................................ 116 
A.1 Vegetation carbon dynamics in VEGAS ....................................................... 116 
A.2 Soil carbon dynamics in VEGAS................................................................... 121 
A.3 Fire module in VEGAS.................................................................................. 123 
A.4 Carbon fluxes at steady stage in VEGAS ...................................................... 125 
A.5 Variables used in the carbon research community......................................... 129 
Glossary .................................................................................................................... 130 
Bibliography ............................................................................................................. 133 
 
 
 
 
This Table of Contents is automatically generated by MS Word, linked to the 
Heading formats used within the Chapter text. 
 
 
 
 
 
 
 
 
 vii 
 
List of Tables 
Table 1.1 The global carbon budget (adapted from House et al., 2003). Positive 
values represent atmospheric CO
2
 increase (or ocean/land sources); 
negative numbers represent atmospheric CO
2
 decrease (ocean/land 
sinks). Unit in PgC/yr. ...................................................................................3 
Table 3.1 Changes in physical climate forcing (temperature/precipitation) in the 
sensitivity simulations. The same seasonal climatology of radiation, 
humidity and wind speed as well as constant CO
2
 level for VEGAS, 
were used in all experiments listed. .............................................................46 
Table 3.2 Mean of composite global and tropical carbon fluxes during the peak 
response to ENSO (shaded area in Figure 3.8 corresponding to the 6-
month average from April to September of ?yr +1?) for VEGAS 
standard, ?P-only?, ?T-only?, ?Swet-fixed? experiments and CO
2 
growth rate. Units in PgC/yr. .......................................................................47 
Table 5.1 Differences of the atmospheric CO
2
 (ppm) and surface temperature (?C) 
changes from 1860 to 2100 between the coupled run and the uncoupled 
runs from UMD, Hadley Centre and IPSL. .................................................71 
Table 5.2 Change in the carbon pools by 2100 from 1860 in three coupled 
carbon-climate models:UMD, Hadley Centre, and IPSL. Units in PgC......72 
Table 5.3 Three sensitivity experiments with different model parameterizations. .....76 
Table 5.4 Difference in total land carbon pool (PgC) between coupled and 
uncoupled runs at year 2100 for various sensitivity simulations of 
UMD, Hadley Centre, and IPSL..................................................................76 
Table 6.1 Major characteristics of the C4MIP models components (adapted from 
Friedlingtein et al., 2006).............................................................................85 
Table A.1 Vegetation parameters used in VEGAS for 4 different Plant Functional 
Types (PFTs), namely, broadleaf tree, needle leaf tree, cold grass and 
warm grass. In the following chart, PFT1 means broadleaf tree, PFT2 
for needle leaf tree, PFT3 for cold grass, and PFT4 for warm grass. ........119 
Table A.2 Soil parameters used in VEGAS for three soil carbon pools: fast soil, 
intermediate soil and slow soil carbon pool...............................................122 
Table A.3 Parameters used in the fire module in VEGAS........................................125 
 
 
 
 viii 
 
List of Figures 
Figure 1.1 Atmospheric carbon dioxide monthly mean mixing ratios in Mauna 
Loa Observatory. Data prior to 1974 is from the Scripps Institution of 
Oceanography (SIO, blue), data since 1974 is from the National 
Oceanic and Atmospheric Administration (NOAA, red) (source from 
www.cmdl.noaa.gov). The annual cycle of CO
2
 concentration is shown 
at bottom right for reference, which is based on the period of 1960-2005 
after long-term increasing trend is removed. .................................................1 
Figure 1.2 Global carbon cycle diagram. The illustration above shows total 
amounts of stored carbon in black and annual carbon fluxes in purple. 
Unit for carbon storage is PgC and PgC/yr for fluxes. Source from 
http://earthobservatory.nasa.gov....................................................................2 
Figure 1.3 Time-series indicating the correspondence of the CO
2
 growth rate 
(gCO
2
) at Mauna Loa Observatory, Hawaii with the ENSO signal 
(Multivariate ENSO Index: MEI; units dimensionless). The seasonal 
cycle has been removed with a filter of 12-month running mean for 
both. The exception during 1992-1993 is caused by the eruption of 
Mount Pinatubo, which is discussed in the text.............................................4 
Figure 1.4 A conceptual diagram shows the competition between the vegetation 
growth and soil respiration, which determines whether and when the 
land will become a carbon sink or source in the future. Increasing CO
2
 
and temperature labeled in X-axis is only for reference purpose. .................5 
Figure 2.1 Concept of the VEgetation-Global-Atmosphere-Soil (VEGAS) model....18 
Figure 2.2 Seasonal cycles of carbon fluxes from VEGAS simulation and on-site 
observation in Tapajos National Forest (Km67: 2.8?S, 55.0?W), Brazil. 
(a) Land-atmosphere carbon flux (NEE) from VEGAS (climatology of 
1991-2004), FLUXNET (climatology of 2002 -2004), and inversion 
(climatology of 1992 -2001); (b) GPP and R
e
 from VEGAS with 
FLUXNET. Unit in KgC/m
2
/yr....................................................................21 
Figure 2.3 Seasonal cycle of land-atmosphere carbon flux from FLUXNET and 
two vegetation and carbon models: TEM (dotted curve) and IBIS 
(dashed curve) at Tapajos National Forest (adapted from Saleska et al., 
2003). ...........................................................................................................22 
Figure 2.4 Anomaly of land-atmosphere carbon flux from VEGAS and 
FLUXNET after season cycle is removed in Tapajos National Forest; 
The inversion result is not plotted here because its simulation time 
coverage ends 2001. Unit in KgC/m
2
/yr......................................................23 
Figure 2.5 Same as Figure 2.2 but for Old Black Spruce, Manitoba, Canada 
(55.9?N, 98.5?W)..........................................................................................24 
Figure 2.6 Same as Figure 2.4 but for Old Black Spruce, Manitoba, Canada 
(55.9?N, 98.5?W)..........................................................................................25 
Figure 2.7 Evaluation of VEGAS: the spatial pattern of correlation between 
VEGAS simulated LAI and the satellite observed NDVI between 1981 
and 2004. A contour is used for shaded areas with value smaller than -
 
 
 ix 
 
0.15. Value of 0.33 is statistically significant at the 90% level of 
Student?s t-test. ............................................................................................26 
Figure 3.1 Time-series indicating the correspondence of the CO
2
 growth rate 
(gCO
2
) at Mauna Loa, Hawaii with global total land-atmosphere carbon 
flux simulated by VEGAS and their lag with the Multivariate ENSO 
Index (MEI; units dimensionless), which is shifted up by 3 units. The 
seasonal cycle has been removed with a filter of 12-month running 
mean. The right top panel is the lagged correlation of MEI, F
ta
 and 
atmospheric CO
2
 growth rate with MEI where the X-axis indicates the 
time lag in months. Value of 0.33 is statistically significant at the 90% 
level of Student?s t-test. ...............................................................................29 
Figure 3.2 Interannual variability of land-atmosphere carbon fluxes (F
ta
) by 
VEGAS and inversion from R?denbeck et al. (2003) (lines for the 
inversion modeling with the CO
2
 measurements from 11, 16, 19, 26 and 
35 sites) for various regions: (a) Global; (b) Tropics between 22.5?S 
and 22.5?N; (c) Northern Hemispheric extratropics between 22.5?N and 
90?N. Seasonal climatology is calculated based on the 1981-1999 time 
period for VEGAS and the inversion model with 11 observational sites; 
The climatology of the other inversion simulations are calculated with 
their respective time coverage. The filled triangles represent the ENSO 
events which have been selected in the composite analysis that follows 
in the text, with red triangles for El Ni?o events and green for La Ni?a 
events. The individual correlation of the land-atmosphere carbon fluxes 
between VEGAS and the inversion simulation with different number of 
stations is indicated in the 3 panels. The figure legend refers to the 
number of stations used in the inversion modeling calculations. All 
these correlation values are statistically significant at the 90% level of 
Student?s t-test. ............................................................................................34 
Figure 3.3 Composites of the carbon fluxes and the climatic field anomalies 
during the growth, mature, and decaying phases of ENSO: The left (a, 
d, g), central (b, e, h) and right (c, f, i) panels are the global, the tropical 
(22.5?S to 22.5?N) and Northern Hemispheric extratropics (22.5?N to 
90?N) composites of the terrestrial fields. The top panels (a, b, c) are the 
terrestrial land-atmosphere carbon flux anomalies of VEGAS (F
ta
) and 
inversion model, the atmospheric CO
2
 growth rate (gCO
2
) and the 
reference MEI; The middle panels (d, e, f) are the corresponding Net 
Primary Production (NPP), and heterotrophic transpiration (R
h
), and the 
bottom panels (g, h, i) are the observed temperature, precipitation and 
modeled soil wetness (Swet). Units for carbon fluxes and CO
2
 growth 
rate are PgC/yr; mm/day  for precipitation, and ?C for temperature, and 
MEI and soil wetness are non-dimensional. The notation of ?yr 0?, ?yr 
+1?, ?yr +2? is the same as in Rassmusson and Carpenter (1982) in 
which ?yr 0? is the mature phase of ENSO. ................................................38 
Figure 3.4 Lead-lag correlations of tropical terrestrial carbon fluxes and climatic 
fields with MEI. The observed CO
2
 growth rate (gCO
2
), temperature, 
precipitation, modeled NPP, Rh, LAI, F
ta
 and soil wetness are indicated; 
 
 
 x 
 
the solid magenta line is the autocorrelation of MEI with itself. 
Negative values in x-axis represent the number of months the 
corresponding variables lead the MEI and the positive values denote the 
number of months the variables lag by. Arrows indicate the peak 
response of each variable. A correlation value of 0.33 is statistically 
significant at the 90% level of Student?s t-test. ...........................................39 
Figure 3.5 Spatial patterns of the anomalies of composite variables with 12-
month averaging centered at the 6th month after MEI peak (as in Figure 
3.3a) for: (a) Land-atmosphere carbon flux by VEGAS (F
ta
) in unit of 
gC/m2/yr; (b) Land-atmosphere carbon flux by the inversion 
(gC/m2/yr). (c) LAI by VEGAS; (d) Satellite observed NDVI. (e) Net 
Primary Production (NPP) in gC/m2/yr; (f) Precipitation in mm/day; (g) 
Heterotrophic Respiration (R
h
) in gC/m2/yr; (h) Surface air temperature 
in ?C. Note: the shading for NPP is opposite to R
h
 and F
ta
..........................43 
Figure 3.6 Time-series of the anomalies of modeled carbon flux due to biomass 
burning and total carbon flux in the tropics. The top right panel is the 
ENSO composites of these two variables, indicating the contribution of 
biomass burning to total carbon flux anomalies. .........................................44 
Figure 3.7 Tropical composites for (a) El Ni?o events only: 1982-83, 1986-87, 
1994-95 and 1997-98; (b) La Ni?a events only: 1984-85, 1988-89, 
1995-96, 1998-99. The selected variables are the same as Figure 3.3b.......44 
Figure 3.8 ENSO-composites of carbon fluxes for the sensitivity simulations: 
(a)VEGAS F
ta
 response in the tropics for the standard, ?P-only?, ?T-
only? and ?Swet-fixed? sensitivity simulations; (b) the same as (a) 
except for NPP; (c) the same as (a) except for heterotrophic respiration. 
The shaded period of Apr.-Sep. of ?yr +1? is the maximum regime of 
the response to the ENSO cycle. The carbon fluxes averaged in this 
regime are indicated in Table 3.2. Units in PgC/yr......................................47 
Figure 3.9 A conceptual diagram of the mechanism of tropical terrestrial carbon 
response to the ENSO cycle. The percentages refer to the contribution 
to F
ta
 variation (100%), with carbon flux anomalies in parenthesis 
(PgC/yr). These carbon fluxes are estimated from Table 3.2. The 
climate fields, such as precipitation, temperature and soil wetness are 
shown as rectangles and the corresponding NPP/Rh/ F
ta
 as rounded 
rectangle. The dashed arrow indicates the indirect effect of temperature 
on NPP through soil wetness. The solid lines in black from temperature 
are the direct effects of temperature on soil decomposition and on NPP. 
All processes are contributable to F
ta 
in the same direction (for example, 
increase of soil decomposition and decrease of vegetation growth 
during El Ni?o) ............................................................................................51 
Figure 4.1 (a) Growth rate of atmospheric CO
2
 observed at Mauna Loa, Hawaii 
from 1980 to 2003; a 12 month running mean (green) was used to 
remove the seasonal cycle, and 24 month running mean (red) was used 
to emphasize the lower frequency variability. (b) VEGAS simulated 
total land-atmosphere carbon flux (black), compared to Mauna Loa CO
2
 
growth rate (green, labeled on the right) and ENSO signal (the negative 
 
 
 xi 
 
southern oscillation index: -SOI, in purple; mb labeled on the left). (c) 
VEGAS simulated global (black), tropics (20?S-20?N, in green) and 
Northern Hemisphere Midlatitude (20?N-50?N, in red). (d) Same as in 
(c) but from the atmospheric inversion of R?denbeck et al. (2003). 
Seasonal cycle has been removed from all figures except where 
otherwise noted. Shading is for the June 1998?May 2002 period during 
which Northern Hemisphere Midlatitude released anomalously large 
amount of CO
2
, modifying the normally tropically dominated ENSO 
response both in terms of amplitude and phasing........................................56 
Figure 4.2 Anomalies for the period June 1998?May 2002 relative to the 
climatology of 1980?2003 for: (a) precipitation (Xie and Arkin, 1996) 
normalized by local standard deviation, (b) VEGAS modeled land-
atmosphere CO
2
 flux (kgC /m2/yr)..............................................................57 
Figure 4.3 Anomalies of the period June 1998?May 2002 for (a) VEGAS 
modeled Leaf Area Index (LAI); (b) observed normalized difference 
vegetation index (NDVI); (c) modeled Net Primary Production (NPP, 
kgC/m
2
/yr); (d) land-atmosphere flux from inversion of R?denbeck et 
al. (2003) with 11 CO
2
 stations (kgC/m
2
/yr) for 1998?2001; (e) 
modeled soil respiration (R
h
, kgC/m
2
/yr); (f) observed surface air 
temperature (?C)...........................................................................................60 
Figure 4.4 Observed precipitation (normalized by standard deviation; green) and 
temperature (red; not labeled: the range from minimum to maximum is 
1.6 ?C), and VEGAS modeled land-atmosphere carbon flux (black) for 
(a) Northern Hemisphere Midlatitude (20?N?50?N); (b) North America 
20?N?50?N; (c) Eurasia 20?N?50?N. Also plotted in (d) is carbon flux 
for the same three regions from the inversion. (e) Modeled Northern 
Hemisphere Midlatitude carbon flux from two sensitivity experiments 
using ?P-only? or ?T-only? as forcing.........................................................63 
Figure 5.1 A conceptual diagram of the Earth system, divided into components 
consisting of physical climate and carbon cycle. The arrows indicate the 
major physical and biological processes, including those affecting 
carbon. The arrows in red are associated with the human activities for 
the carbon cycle. In UMD, the physical climate component includes an 
atmospheric model QTCM, a slab mixed-layer ocean model and the 
SLand unit, which handles processes such as soil moisture storage, but 
not carbon. VEGAS and a box ocean carbon model serve as the carbon 
cycle components.........................................................................................68 
Figure 5.2 Model experiments: (a) Coupled: the physical climate and carbon 
cycle model are fully coupled from 1791 to 2100, with no other 
changing external forcing except for the anthropogenic CO
2
 emissions. 
(b) Uncoupled: the carbon model sees a nearly constant climate without 
global warming, but carbon components are fully interactive including 
CO
2
 fertilization and emission. Then an additional simulation where the 
CO
2
 is from uncoupled run was used to force the physical climate 
model to calculate the surface temperature increase, similar to 
 
 
 xii 
 
conventional GCM global warming simulations. The external 
anthropogenic emission used IPCC-SRES A1B scenario. ..........................69 
Figure 5.3 (a) Atmospheric CO
2
 (ppm) and (b) surface temperature change (?C) 
from 1860 to 2100, simulated by the fully coupled carbon cycle-climate 
model (in red), with constant (pre-industrial) climate (in blue), 
compared to observations (in black). The surface temperature curve 
labeled in uncoupled was obtained by an additional simulation where 
the CO
2
 from uncoupled run was used to force the physical climate 
model, similar to conventional GCM global warming simulations.............70 
Figure 5.4 Cumulative vegetation and soil carbon changes in PgC since 1860 for 
the fully coupled run (red) and uncoupled run (blue) from the present 
model (UMD, upper panels), the Hadley Centre (middle panels), and 
IPSL (lower panels). ....................................................................................73 
Figure 5.5 Spatial distribution of total land carbon (vegetation + soil) change for 
the UMD coupled and uncoupled runs. These are the differences 
between the last 30 years (2071?2100) and the first 30 years (1860?
1889), showing different behavior at high latitude and mid-low latitude 
regions. Units in kgC/m
2
..............................................................................75 
Figure 5.6 Total land carbon change (PgC) since 1860 in four fully coupled runs 
of UMD: the standard run (same run as in Figure 5.3), and three 
sensitivity experiments: strong CO
2
 fertilization effect, single soil pool, 
and high soil decomposition rate dependence on temperature (Q
10
 = 2.2 
for all soil layers, blue). ...............................................................................77 
Figure 6.1 Boreal vegetation distribution in the northern high latitudes (modified 
from Montaigne, 2002). The area poleward of 60?N (circled in dark 
brown) is defined as Northern High Latitudes (NHL) in this study. ...........81 
Figure 6.2 Simulated carbon storage for the pre-industrial era in the vegetation 
and soil by the C4MIP models and model-mean for: (a) Global total; (b) 
NHL. The stored carbon is calculated based on 1860-1870 for all 
models except 1870-1880 for LLNL, and 1901-1910 for CLIMBER-
LPJ. Soil carbon is indicated in brown, and vegetation in dark green. 
Most of the carbon is locked in the soil particularly in the NHL. Units 
of storage are in PgC....................................................................................88 
Figure 6.3 Temperature increase from 1860 to 2100 projected by the fully 
coupled simulations from C4MIP models for global mean (blue 
diamond) and NHL (red circle) except for LLNL which is from 1870 to 
2100, and CLIMBER-LPJ from 1901 to 2100. The multi-model mean is 
plotted. It is evident that significantly more intense warming is expected 
in the NHL, compared to global mean. The UMD model is an exception 
since it does not include a snow/ice albedo scheme. ...................................89 
Figure 6.4 Changes of NEP and total land carbon in the NHL during 1860-2100 
in C4MIP coupled simulations. (a) NEP; (b) Total land carbon change. 
Colored lines in the (a) (b) are for individual C4MIP models; (c) (d) the 
multi-model mean (red solid line) and the 1-?  spread (blue shading) 
are indicated. The change in total land carbon is indicated relative to 
1860, except for LLNL and CLIMBER-LPJ where it is relative to 1870 
 
 
 xiii 
 
and 1901 respectively. A 6-year running mean filter is applied to all the 
curves. Unit for NEP is PgC/yr and for land carbon change is PgC............91 
Figure 6.5 Changes of vegetation and soil carbon in the NHL in C4MIP couple 
simulations: (a, c) Vegetation carbon; (b, d) Soil carbon. Colored lines 
in the (a) and (b) are for individual C4MIP models; (c) and (d) show the 
multi-model mean (red solid line) and the 1-?  spread (blue shading) 
are indicated. All changes are relative to the year of 1860 except for 
LLNL with 1870, and CLIMBER-LPJ with 1901. A 6-year running 
mean filter is applied to all the curves. Units are in PgC.............................92 
Figure 6.6 The multi-model mean and 1-? spread of carbon fluxes in vegetation 
in the NHL in the C4MIP coupled simulations for: (a) Net Primary 
Product; (b) Vegetation turnover; (c) Vegetation carbon growth rate. 
The changes are relative to the values in the late 18
th
 century; a 6-year 
running mean filter is applied to all the curves. Units are PgC/yr for all 
three panels. .................................................................................................94 
Figure 6.7 Same as Figure 6.6 but for the changes of carbon fluxes in soil for: (a) 
Heterotrophic Respiration; (b) Soil carbon growth rate. Units are 
PgC/yr. .........................................................................................................95 
Figure 6.8 Changes in carbon storage during the 1860-2100 period due to carbon 
cycle-climate feedback (coupled?uncoupled simulation) for the 11 
C4MIP members for the globe (a, c, e) and the NHL (b, d, f) for: (a, b) 
Total land carbon. (c, d) Vegetation carbon; (e, f) Soil carbon; All 
values are relative to the year of 1860 except for LLNL and CLIMBER-
LPJ, which have the beginning years of 1860 and 1901 respectively. A 
6-year running mean filter is applied to all the curves.................................98 
Figure 6.9 The sensitivity of NPP to CO
2
 fertilization and temperature change for 
the whole globe (a, c) and the NHL (b, d) during 1860-2100 for the 11 
C4MIP members. The NPP sensitivity to CO
2
 fertilization (a, b) is 
calculated directly from the C4MIP uncoupled simulations, in which the 
constant climate is prescribed while CO
2
 increases for photosynthesis. 
Then NPP sensitivity to CO
2
 fertilization at 2 x CO
2
 is used in the 
coupled simulations to isolate the NPP sensitivity to temperature 
increase. The NPP sensitivity to temperature (b, d) thus is not the direct 
NPP from C4MIP coupled simulation. It has been corrected using the 
sensitivity of NPP to CO
2
 fertilization based on uncoupled simulation. 
For details of this calculation, refer to Friedlingstein et al. (2006). It 
indicates that NPP sensitivity to temperature change in the NHL is 
different from the one of global scale. The panels on the left (a, c) are 
the same as Figure 3a, b in Friedlingstein et al. (2006), and are used 
here only for comparison to the NHL result. ...............................................99 
Figure 6.10 The conceptual diagram to show effects CO
2
 fertilization and intense 
warming in NHL on the changes of carbon fluxes and storages by 2100 
from C4MIP coupled simulation. The multi-model mean and one 
standard deviation are provided to indicate the relative magnitudes of 
these changes. All changes are relative to 1860. Unit for carbon flux is 
PgC/yr and PgC for carbon storage. ..........................................................103 
 
 
 xiv 
 
Figure 7.1 (a) Land-atmosphere carbon flux of 1901-2004 from VEGAS offline 
simulation; (b) Pacific Decadal Oscillation (PDO) Index of 1901-2004. 
6-year running mean applied to smooth the curves to extract the low 
frequency signal (red). ...............................................................................114 
Figure A.1 The temperature dependent function )(T? for 4 PFTs in VEGAS. (a) 
Warm tree and cold tree; (b) Warm grass (C4) and cold grass (C3). ........120 
Figure A.2 The light dependent ),,( HLAIPAR?  on the radiation in VEGAS........120 
Figure A.3 The CO
2
 dependent growth factor )(
2
CO?  for C3 (warm/cold tree 
and cold grass) and C4 (warm grass) in VEGAS. .....................................121 
Figure A.4 Soil wetness effect on the soil respiration in VEGAS with different 
topography. The topography information is scaled to the topography 
gradient in VEGAS and doesn?t change in modeling with time. The plot 
indicates that at a higher soil wetness condition, the respiration tends to 
be reduced somewhat (e.g., wetland) by the topographic effects. .............122 
Figure A.5 Temperature effect Q
10
 on the respiration in VEGAS. Q
10
 =2.2 for 
leaf and wood, 1.5 for the root and fast soil, 1.35 for the intermediate 
soil and 1.105 for the slow soil carbon pool. .............................................123 
Figure A.6 (a) Annual mean of carbon fluxes (PgC/yr) in VEGAS. The figure 
indicates the magnitude of the individual carbon fluxes in VEGAS. The 
values in red in parenthesis are calculated at steady state of offline 
VEGAS simulation. The arrows in colors are used to indicate different 
biological processes. (b) Annual mean of GPP, Ra and Rh (PgC/yr) and 
individual carbon pool (PgC) of vegetation and soil in VEGAS. The 
physical meaning of these variables in this figure with respect to their 
names in Appendices can be found in the following table: .......................127 
 
 
 1 
 
Chapter 1: Introduction 
 
1.1 Importance of terrestrial carbon cycle in climate systems 
Carbon cycle is one of the most important biogeochemical cycles in the 
climate system because carbon dioxide (CO
2
) is a principal greenhouse gas that 
contributes significantly to global warming. Since the beginning of the industrial era, 
around 1750, the CO
2
 concentration in the atmosphere has risen, at an increasing rate, 
from around 280 ppm to nearly 387 ppm in 2007 at Mauna Loa Observatory (MLO) 
(Figure 1.1). Ice core record reveals that current CO
2
 level
 
is already 27% higher than 
its highest recorded level during the past 650,000 years before the Industrial 
Revolution (Siegenthaler et al., 2005; Spahni et al., 2005). Over the 20
th
 century the 
global average surface temperature resultantly has risen by 0.65?0.2?C (IPCC, 2007). 
 
Figure 1.1 Atmospheric carbon dioxide monthly mean mixing ratios in Mauna Loa Observatory. 
Data prior to 1974 is from the Scripps Institution of Oceanography (SIO, blue), data since 1974 is 
from the National Oceanic and Atmospheric Administration (NOAA, red) (source from 
www.cmdl.noaa.gov). The annual cycle of CO
2
 concentration is shown at bottom right for 
reference, which is based on the period of 1960-2005 after long-term increasing trend is removed. 
 
 
 2 
 
The amount of carbon dioxide in the atmosphere is highly dependent on the 
carbon exchange between the atmosphere and ecosystem, anthropogenic fossil fuel 
emission and land use change (Figure 1.2). Although recent estimates of fossil fuel 
burning and atmospheric CO
2
 concentration are quite precise, there is a discrepancy 
in estimates of land/ocean uptake and land use change (Thompson et al., 1996; 
Rayner et al., 1999; Bousquet et al., 2000; McGuire et al., 2001; Gurney et al., 2002; 
Le Quere et al., 2002; DeFries et al., 2002; Houghton, 2003a, b; R?denbeck et al., 
2003; Baker et al., 2006). For instance, one recent estimate (House et al., 2003) 
indicated that in the 1990s the ocean and land carbon fluxes were ?2.1 ? 0.7 PgC/yr 
and ?1.0 ? 0.8 PgC/yr respectively, compared with ?1.7 ? 0.5 PgC/yr for ocean and 
?1.4 ? 0.7 PgC/yr for land according to Prentice et al. (2001) (Table 1.1). 
 
Figure 1.2 Global carbon cycle diagram. The illustration above shows total amounts of stored 
carbon in black and annual carbon fluxes in purple. Unit for carbon storage is PgC and PgC/yr for 
fluxes. Source from http://earthobservatory.nasa.gov. 
 
 
 3 
 
Table 1.1 The global carbon budget (adapted from House et al., 2003). Positive values represent 
atmospheric CO
2
 increase (or ocean/land sources); negative numbers represent atmospheric CO
2
 
decrease (ocean/land sinks). Unit in PgC/yr. 
IPCC
1 
Update 
 
1980s 1990s 1980s 1990s 
Atmospheric increase 3.3 ? 0.1 3.2 ? 0.1   
Emissions (fossil fuel, 
cement) 
5.4 ? 0.3 6.3 ? 0.4   
Ocean?atmosphere flux ?1.9 ? 0.6 ?1.7 ? 0.5 ?1.8 ? 0.8 ?2.1 ? 0.7 
Land?atmosphere flux
2
 ?0.2 ? 0.7 ?1.4 ? 0.7 ?0.3 ? 0.9 ?1.0 ? 0.8 
?Land-use change
3
 1.7 (0.6 to 2.5) Incomplete 0.9 to 2.8 1.4 to 3.0 
?Residual terrestrial sink ?1.9 (?3.8 to 0.3) Incomplete 
 
?4.0 to ?0.3 ?4.8 to ?1.6 
1 
Prentice et al., 2001, IPCC Third Assessment Report. 
2 
The net land atmosphere flux consists of emissions due to land-use change as estimated by models, and sinks due 
to other processes, calculated as a residual. 
3 
The IPCC estimated range for the land use change flux is based on the full range of Houghton?s book-keeping 
model approach (Houghton, 1999; Houghton et al., 1999; Houghton et al., 2000) and the Carbon Cycle Model 
Linkage Project (CCMLP) (McGuire et al., 2001). In the update, the 1980s estimate is the full range of Houghton 
updated (Houghton, 2003a) and CCMLP; while the 1990s flux is based on Houghton (2003a) only as the CCMLP 
analysis stopped at 1995. 
 
 
Furthermore, year-to-year variations of the atmospheric CO
2
 growth rate 
(gCO
2
: time derivative of the CO
2
 concentration of Mauna Loa Observatory in this 
thesis) are still imperfectly understood. Such interannual variability of CO
2
 growth 
rate can be up to 4-5 PgC/yr during El Ni?o years (Figure 1.3). Because fossil fuel 
emissions and land use change increase slowly with small year-to-year variation, the 
interannual variability of CO
2
 growth rate in the atmosphere lies in the changing 
absorption capacity of the land and ocean and is related to climate variation. Bousquet 
et al. (2000) applied an inversion model to 20 years of atmospheric CO
2
 
measurements and reported that global terrestrial carbon flux was approximately 
twice as variable as ocean fluxes between 1980 and 1998, and that the tropical land 
contributes most of the interannual changes in global carbon balance.  
 
 
 4 
 
 
Figure 1.3 Time-series indicating the correspondence of the CO
2
 growth rate (gCO
2
) at Mauna 
Loa Observatory, Hawaii with the ENSO signal (Multivariate ENSO Index: MEI; units 
dimensionless). The seasonal cycle has been removed with a filter of 12-month running mean for 
both. The exception during 1992-1993 is caused by the eruption of Mount Pinatubo, which is 
discussed in the text. 
 
These studies of global carbon cycle on interannual and decadal timescale 
have advanced our understanding of terrestrial ecosystem in response to climate 
change (Keeling et al., 1985; Thompson et al., 1996; Bousquet et al., 2000; Gurney et 
al., 2002; Defries et al., 2002; Houghton, 2003b; R?denbeck et al., 2003). Some 
important unanswered questions are: How will the terrestrial biosphere respond to 
global warming (Figure 1.4), and how will this affect the climate in the future? Fully 
coupled three-dimensional carbon cycle-climate models recently have been used to 
study the interaction between the global carbon cycle and the climate (Cox et al., 
2000; Friedlingstein et al., 2001; Joos et al., 2001; Zeng et al., 2004; Matthews et al., 
2005; Friedlingstein et al., 2006;). Based on simulations from 11 modeling groups in 
the Coupled Carbon Cycle Climate Model Inter-comparison Project (C4MIP), 
Friedlingstein et al. (2006) concluded that there was a positive carbon cycle-climate 
feedback in all C4MIP models, albeit of different magnitudes. Due to this positive 
feedback, there will be an additional 20-200 ppm CO
2
 in the atmosphere, leading to 
 
 
 5 
 
temperature 0.1-1.5?C warmer by 2100. Most C4MIP models simulate a robust 
reduction of global terrestrial carbon uptake by 2100, dominated by the tropics. This 
may cause the tropical forest vulnerable to decay under future global warming, such 
as a dieback of Amazon rainforest (e.g., Cox et al., 2000; Betts et al., 2004; Cox et al., 
2004). All of these studies disclose the crucial role of terrestrial ecosystems in the 
climate system. Understanding the uptake of atmospheric CO
2
 by the land surface is 
of significance in projecting future climate. 
 
Figure 1.4 A conceptual diagram shows the competition between the vegetation growth and soil 
respiration, which determines whether and when the land will become a carbon sink or source in 
the future. Increasing CO
2
 and temperature labeled in X-axis is only for reference purpose. 
 
1.2 The interaction between the terrestrial carbon cycle and climate 
 
The terrestrial biosphere gains carbon from the atmosphere through 
photosynthesis and loses it primarily through respiration (autotrophic and 
heterotrophic). This land-atmosphere carbon exchange is, in principle, governed by 
the prevailing physical (incoming solar radiation), climatic (temperature and 
 
 
 6 
 
precipitation), and edaphic (e.g., nutrients and soil texture) conditions leading to the 
notion of a biome equilibrium distribution (Budyko, 1974). On the other hand, 
changes in the vegetation type, structure and physiology of terrestrial ecosystems can 
feedback to climate via changes in the partitioning of energy between latent and 
sensible heat, albedo, and roughness, etc. 
In this thesis, I will present a study of the interactions between the terrestrial 
carbon ecosystem and climate over a wide-range of temporal and spatial scales, based 
on a terrestrial vegetation carbon model, VEgetation-Global-Atmosphere-Soil 
(VEGAS) (Zeng, 2003; Zeng et al. 2005a). Several sensitivity simulations have been 
designed to diagnose and understand the variability of the carbon cycle at different 
temporal scales. A focus is the response of the terrestrial carbon ecosystem to climate 
change, particularly to El Ni?o-Southern Oscillation (ENSO) and Midlatitude 
droughts. This helps to understand how terrestrial carbon cycle influences the 
atmospheric CO
2
 growth rate. The carbon cycle-climate feedback and carbon uptake 
by the Northern High Latitudes terrestrial biosphere in the 21
st
 century also have been 
studied. 
1.2.1 Modeling the seasonal cycles of terrestrial carbon ecosystem  
Modeling a ?correct? seasonal cycle is critical for terrestrial vegetation and 
carbon cycle model to quantify the phenology and biochemical processes. Heimann et 
al. (1998) evaluated 6 terrestrial carbon cycle models on their capability to capture the 
seasonal cycle of atmospheric CO
2
. Their results showed that, in the tropics, the 
prognostic models generally tended to over-predict the net seasonal exchanges of 
carbon fluxes and had stronger seasonal cycles than indicated from observations. 
 
 
 7 
 
Wittenberg et al. (1998) suggested that this was possibly due to the influence of 
biomass burning on the seasonal CO
2
 signal as observed at monitoring stations. 
Cramer et al. (1999) pointed out large differences in the simulated seasonal changes 
among models, both globally and locally. Presumably the differences are due to 
distinct representation in the models. Randerson et al. (1999) found that an increase in 
early season ecosystem uptake could explain recent changes in the seasonal cycle of 
atmospheric CO
2
 at high northern latitudes. This was in line with what Keeling et al. 
(1996) anticipated - since1960s the biosphere activity in the northern latitudes may 
have changed strongly, producing a longer vegetation growing period. Myneni et al. 
(1997) confirmed this with Advanced Very High Resolution Radiometer (AVHRR) 
satellite records. With modeling developments, current terrestrial vegetation and 
carbon models generally agree on the phase of the terrestrial seasonal cycle in the 
high northern latitudes (Nemry et al., 1999; Cramer et al., 1999; Zhuang et al., 2001; 
Dargaville et al., 2002). 
However, models and observations exhibited oppositely phased seasonal 
cycles for tree growth at measurement site in the Amazon basin (e.g., Saleska et al., 
2003). Saleska et al. (2003) proposed that seasonally moist tropical evergreen forests 
might have evolved two adaptive mechanisms in an environment with strong seasonal 
variations of light and water: deep roots for access to water in deep soils and leaf 
phenology for access to light. Xiao et al. (2005) simulated high GPP (Gross Primary 
Productivity) in the late dry seasons in Amazon basin, consistent with the estimated 
GPP from the CO
2
 eddy flux tower there. Huete et al. (2006) found MODIS 
Enhanced Vegetation Index (EVI, an index of canopy photosynthetic capacity) 
 
 
 8 
 
increased by 25% with sunlight during the dry season across Amazon forests. 
Oliveira et al. (2005) suggested that in addition to soil water uptake by deep root, the 
process of hydraulic redistribution could be another contributing factor explaining the 
absence of plant water stress during drought. Their proposed mechanism for this 
redistribution is the nocturnal transfer of water by roots from moist to dry regions of 
the soil profile.  
Eddy covariance observation stations, which characterize fluxes and energy 
exchange at the surface and provide useful parameters to global and regional climate 
modelers, are still limited. Although the micro-meteorological flux measurements at 
FLUXNET sites are building up coverage with more sites across the globe in North, 
Central and South America, Europe, Scandinavia, Siberia, Asia, and Africa, they are 
still very sparse, especially in tropical forests. Observations and recent improvements 
of carbon cycle modeling have advanced our understanding of the seasonal dynamics 
of the terrestrial carbon cycle, particularly in different regions, such as the tropical 
and boreal forest. In Chapter 2, I will discuss the model comparison of VEGAS 
against FLUXNET observations. A brief description of physical processes of VEGAS 
will be also provided firstly in that chapter. 
1.2.2 Modeling the interannual variability of terrestrial carbon cycle 
The records of atmospheric CO
2
 at Mauna Loa Observatory (MLO) since 
1958 indicate that besides the seasonal cycle, substantial interannual variability of 
atmospheric CO
2
 is superimposed on the ongoing increasing atmospheric CO
2
 
concentration. The association between CO
2
 growth rate and ENSO (Figure 1.3) was 
initially reported in the 1970?s and has been confirmed by recent studies (Bacastow, 
 
 
 9 
 
1976; Keeling and Revelle, 1985; Braswell et al., 1997; Rayner et al., 1999; Jones et 
al., 2001; Zeng et al., 2005a). It was noticed that during El Ni?o (La Ni?a) events, the 
atmospheric CO
2
 growth rate increased (decreased) at Mauna Loa Observatory with a 
5-month lag of ENSO peak. 
 Different climatic responses have been suggested in previous studies to 
explain the ENSO-related terrestrial carbon cycle variation. For example, 
Kindermann et al. (1996) suggested that the temperature dependence of Net Primary 
Production (NPP, a variable usually used to indicate vegetation growth) is the most 
important factor in determining land-atmosphere carbon flux. However, precipitation 
has been suggested alternatively as the dominant factor for variation of terrestrial 
carbon cycle (Tian et al., 1998; Zeng et al., 2005a). Nemani et al. (2003) and Ichii et 
al. (2005) indicated that in tropical terrestrial ecosystems, variations in solar radiation 
and, to a lesser extent, temperature and precipitation, explained most interannual 
variation in the Gross Primary Production (GPP). Hashimoto et al. (2004) confirmed 
the dependence of global heterotrophic respiration and fire carbon fluxes on 
interannual temperature variability and strongly suggested that ENSO-mediated NPP 
variability influenced the atmospheric CO
2
 growth rate. Besides these direct 
biological responses, biomass burning due to climate change might as well play an 
important role on the variation of total land-atmosphere carbon flux (Page et al., 
2002; Langenfelds et al., 2002; Van der Werf et al., 2004). Van der Werf et al. (2004) 
estimated that during the 1997-98 El Ni?o, the anomalous carbon emission due to 
fires were 2.1 ? 0.8 PgC, or 66 ? 24% of the atmospheric CO
2
 growth rate anomaly in 
that period.  
 
 
 10 
 
These studies emphasized various factors related to the interannual variability 
of the terrestrial carbon flux; they have provided insight and opportunities to explore 
the underlying physical and biological mechanisms of the interannual variability of 
atmospheric CO
2
 growth rate. Most of the studies cited above, however, have 
primarily focused on the effect of the climatic factors on photosynthetic processes 
(GPP, NPP), or the land-atmosphere carbon flux with respect to biomass burning. To 
date, few studies have been conducted to understand the underlying physical 
processes through which soil moisture affects the interannual variability of the 
terrestrial carbon flux. For instance, temperature has a direct impact on vegetation 
photosynthesis and soil respiration On the other hand; the temperature regulates 
evapotranspiration, and thus influences the variation of soil moisture, which is an 
important factor for vegetation growth and soil respiration. Through this process, the 
temperature thus has an indirect impact on the vegetation growth and soil respiration. 
This indirect influence has not yet been studied.  
In Chapter 3, I will present an investigation of the response of the terrestrial 
carbon ecosystem to ENSO for the period of 1980-2004, based on the simulations of 
VEGAS. A detailed investigation of soil moisture effect on terrestrial carbon 
exchange is of interest. Besides the canonic tropical response to ENSO, in Chapter 4, 
a modeling study of the impact of the 1998-2002 Midlatitude drought on terrestrial 
ecosystem and the global carbon cycle will serve as a clue regarding why the 
atmospheric CO
2 
growth rate was over 2 ppm/yr during 2002?2003, an unusual 2-
year increase like no other during the course of the Mauna Loa Observatory station 
record. 
 
 
 11 
 
1.2.3 Modeling the terrestrial carbon uptake under global warming 
The future CO
2
 concentration in the atmosphere is not straightforward to 
predict since it depends critically on the capacity of the land and ocean carbon 
absorption, which, in turn, is sensitive to the future climate change. If the Earth 
ecosystem reduces the capacity to take up the anthropogenic emissions, more CO
2
 
will stay in the atmosphere, accelerating global warming. This will trigger a further 
robust reduction of carbon uptake by ecosystems, and provide a positive feedback to 
atmospheric CO
2 
increase. Understanding the carbon cycle-climate feedback is thus 
of importance for the projection of future climate.  
Early general circulation models generally excluded the interactions between 
climate and the biosphere, using static vegetation distributions and CO
2
 
concentrations from simple carbon-cycle models that did not include climate change. 
Betts et al. (1997) attempted to quantify the effects of both physiological and 
structural vegetation feedbacks on a doubled-CO
2
 climate, with a general circulation 
model iteratively coupled to an equilibrium vegetation model. Using a terrestrial 
biogeochemical model, forced by simulations of transient climate change with a 
general circulation model, Cao et al. (1998) predicted that terrestrial ecosystem 
carbon fluxes responded to and strongly influence the atmospheric CO
2
 increase and 
climate change. Cox et al. (2000) used a fully coupled three-dimensional carbon-
climate model and reported that carbon-cycle feedbacks could significantly accelerate 
climate change during the 21
th
 century. In their study, the atmospheric CO
2
 
concentration could be 250 ppm higher by 2100 due to a positive carbon-climate 
feedback, causing an additional 1.5?C warming. The Amazon rainforest was even 
 
 
 12 
 
projected to have a dieback towards the end of this century. However, the 
experiments by Institute Pierre Simon Laplace (IPSL) (Friedlingstein et al., 2001; 
Dufresne et al., 2002; Berthelot et al., 2002) suggested that the amplitude of positive 
feedback was three times smaller than that simulated by Cox et al. (2000). Fung et al. 
(2005) indicated that the amplification of climate change by the additional CO
2
 could 
be small at the end of the 21
st
 century. Quantifying and predicting this carbon cycle-
climate feedback is thus extremely difficult because of the limited understanding of 
the processes by which carbon and the associated nutrients are transformed or 
recycled within ecosystems, and exchanged with the overlying atmosphere (Heimann 
et al., 2008). 
During the past decades, Northern High Latitudes (NHL: poleward of 60?N) 
have witnessed dramatic changes, where the annual average temperatures increased 
by 1-2?C in northern Eurasia and northwestern North America (Arctic Climate Impact 
Assessment, 2005), much larger than the of 0.65?0.2?C increase of global average 
surface temperature over the 20
th
 century (IPCC, 2007). Accompanying the 
accelerating climate changes in the NHL, a vegetation ?greening? trend has been 
observed in the boreal forests with satellite data and phenology studies (Keeling et al., 
1996; Myneni et al., 1997; Zhou et al., 2001; Tucker et al., 2001; Lucht et al., 2002). 
The enhancement in photosynthetic activity, associated with the persistent increase in 
the length of the growing season, may be leading to long-term increase in carbon 
storage and changes in vegetation cover, which in turn affects the climate system. 
However, warming accelerates decomposition of dead organic matter, thus losing soil 
carbon to the atmosphere. Reichstein et al. (2007) analyzed the European eddy 
 
 
 13 
 
covariance fluxes and reported that water availability was a significant modulator of 
Net Ecosystem Productivity (NEP) on these sites, while the multivariate effect of 
mean annual temperature was small and not significant. However, Piao et al. (2008) 
found that there was a net carbon dioxide loss of northern ecosystems in response to 
autumn warming, because the increase in respiration was greater than photosynthesis 
during autumn warming. NHL contains a large amount of carbon in frozen soil, 
which is vulnerable to decay under global warming (Davidson et al., 2006). To date, 
such processes in particular are poorly understood (Melillo et al., 2002; Bond-
Lamberty et al., 2004; Lawrence and Slater, 2005; Bronson et al., 2008). The 
competition between the absorption of carbon by boreal forests and the release of 
carbon from soil therefore is important for future carbon-cycle climate feedback and 
the degree of climate change. Previous projection of carbon uptake in the NHL under 
global warming scenario was mostly based on offline simulations of vegetation-
carbon models, which were forced by the IPCC climate projection (McGuire et al., 
2000; Cramer et al., 2001; Schaphoff et al., 2006; Lucht el al. 2006; Sitch et al., 
2007). However, fewer studies have considered the carbon cycle-climate coupling in 
the NHL (McGuire et al., 2006). With the positive carbon cycle-climate feedback 
included, the NHL may undergo more intense warming than expected. Consequently, 
the frozen soil may emit carbon and NHL can become a carbon source of the 
atmosphere.  
In Chapter 5, using a fully coupled carbon cycle?climate model (UMD), we 
have investigated the carbon cycle-climate feedback on the future climate system. 
 
 
 14 
 
Chapter 6 will focus on the study of carbon uptake in NHL terrestrial biosphere under 
climate change in the 21
st 
century, based on simulations of C4MIP.  
 
 
 15 
 
  
Chapter 2: Evaluation of VEGAS against Observations  
 
2.1 Introduction 
This chapter provides a description of VEGAS and focuses on the evaluation 
of the model against the observations. This helps to measure the ability of VEGAS in 
capturing some of the physical processes of the biosphere, and also gives some 
insight into the further improvement for the VEGAS model. A detailed description of 
the processes in VEGAS can be found in the Appendices. Seasonal and interannual 
variability of modeled carbon fluxes from VEGAS have been compared across many 
sites of FLUXNET in North and South America. Two suites of comparisons will be 
discussed here: Station Km67 in Tapajos National Forest in Brazil, and Old Black 
Spruce in Canada. A lot of attention has been paid to the Amazon forest basin 
recently (e.g., Cox et al., 2000; Friedlingstein et al., 2001; Zeng et al., 2004; Cook et 
al., 2008). The Amazon forest plays an important role in the global carbon cycle 
because of deforestation, and could be subject to a recession in the future. In Chapter 
3, we will look into the response of the terrestrial carbon ecosystem to ENSO, 
highlighting the importance of the variability of the tropical terrestrial ecosystem. The 
Old Black Spruce site in Canada is a typical boreal forest type, which is an important 
carbon reservoir. Accompanying the accelerating climate changes during the past 
decades, a ?greening? trend with vegetation growth has been observed in the boreal 
forests (Keeling et al., 1996; Myneni et al., 1997; Zhou et al., 2001; Tucker et al., 
2001; Lucht et al., 2002); however, the frozen soil carbon in the NHL is vulnerable to 
 
 
 16 
 
decay due to warming there, leading NHL to be a significant carbon source in the 
future. In Chapter 6, we will investigate the future carbon uptake of the NHL in the 
21
th
 century. After the comparison with the FLUXNET at the local scale, we then 
provide an independent evaluation of the interannual variability of VEGAS against 
observations by comparing its simulated Leaf Area Index (LAI) to observed NDVI. 
2.2 Model description and setup 
Figure 2.1 is a conceptual diagram indicating the important processes of 
terrestrial carbon cycle in VEGAS (see details of its dynamics in Appendices). 
Briefly, VEGAS simulates vegetation growth and competition among four different 
Plant Functional Types (PFTs): broadleaf tree, needle leaf tree, cold grass and warm 
grass (Zeng, 2003; Zeng et al., 2005a). In VEGAS, the photosynthetic pathways for 
these 4 types are distinguished as C3 (the first three PFTs above) or C4 (warm grass). 
The difference between C3 and C4 is the way in which they accept CO
2
 from the air. 
C3 plants use the enzyme ribulosodiphosphatcarboxylase while C4 plants use 
phosphoenolpyruvatcarboxylase. C3 plants include more than 95 percent of the plant 
species on earth, including crops such as wheat, barley, potatoes and sugar beet. C4 
plants also include crop plants, such as corn and sugar cane. Phenology (or the life 
cycle of plants) is simulated dynamically as the balance between growth and 
respiration/turnover. Competition between growths for the different PFTs is 
determined by climatic constraints and resource allocation strategy such as 
temperature tolerance and height. The relative competitive advantage then determines 
fractional coverage of each PFT with possibility of coexistence. Accompanying the 
vegetation dynamics is a full terrestrial carbon ecosystem, starting from 
 
 
 17 
 
photosynthetic carbon assimilation in the leaves and the allocation of this carbon into 
three vegetation carbon pools: leaf, root and wood. After accounting for respiration, 
the biomass turnover from these three vegetation carbon pools cascades into a fast 
soil carbon pool, an intermediate and finally a slow soil pool. Temperature and 
moisture dependent decomposition, as well as occurrence of fires, of these fuel loads 
returns carbon back into the atmosphere, thus closing the terrestrial carbon cycle. In 
VEGAS, the carbon flux associated with biomass burning of vegetation is included in 
the respiration of vegetation and the rest is included in soil decomposition. The key 
carbon flux outputs from VEGAS are related as follows: 
       NPP = GPP ?R
a
      (2.1) 
                   NEP = NPP ?R
h
      (2.) 
F
ta 
= ?NEP      (2.3) 
Here GPP, Gross Primary Production is the total amount of carbon fixed by 
the photosynthesis of the terrestrial ecosystem from the atmosphere. NPP denotes Net 
Primary Production, which is the amount of carbon fixed after subtracting the 
respiration of vegetation, called Autotrophic Respiration (R
a
), from GPP. The 
difference between Heterotrophic Respiration (R
h
), which is the carbon released due 
to decomposition of soil organic matter, and NPP, is the Net Ecosystem Production 
(NEP) of the terrestrial ecosystem. F
ta
 is the land-atmosphere carbon flux, otherwise 
referred to as the Net Ecosystem Exchange (NEE). 
VEGAS was coupled with a simple land surface model SLand (Zeng et al., 
1999). SLand is an intermediate-complexity model which consists of two soil layers 
of temperature and one layer for soil moisture. It includes subgrid-scale 
 
 
 18 
 
parameterizations of the major processes of evapotranspiration, interception loss, 
surface and subsurface runoff. The validation of modeled soil moisture by SLand on 
seasonal, interannual and longer timescales can be found in Zeng et al. (2008). In this 
paper, basin-scale terrestrial water storage for the Amazon and Mississippi, diagnosed 
using the Precipitation-Evaporation-and-Runoff (PER) method from SLand, was 
compared with those from other land surface models, the reanalysis, and NASA?s 
Gravity Recovery and Climate Experiment (GRACE) satellite gravity data. The 
results indicate that SLand is reliable on seasonal and interannual time scales. In this 
thesis, soil wetness (Swet) is used to indicate the relative soil water saturation. It is 
defined as a ratio of modeled soil moisture (mm) to the maximum value of 500 mm. 
Swet varies from 0 to 1.  
 
Figure 2.1 Concept of the VEgetation-Global-Atmosphere-Soil (VEGAS) model. 
 
 
 19 
 
In this chapter and Chapter 3 and 4, an offline simulation, which is forced by 
observed climate variables, is used to evaluate VEGAS and investigate the variability 
of the terrestrial carbon ecosystem in response to climate variation. The physical 
climate forcing for this offline simulation includes surface temperature, precipitation, 
atmospheric humidity, radiation forcing, and surface wind. The driving data of 
precipitation for VEGAS was a combination of the Climate Research Unit (CRU: 
New et al., 1999; Mitchell and Jones, 2005) dataset for the period of 1901?1979, and 
the Xie and Arkin (1996) dataset of 1980-2006 (which has been adjusted with the 
1981-2000 climatology of CRU dataset). The surface air temperature driving data was 
from the National Aeronautics and Space Administration (NASA) Goddard Institute 
for Space Studies (GISS) by Hansen et al. (1999) and CRU. We created a new 
temperature dataset using the anomaly of GISS temperature and CRU climatology of 
1961-1990, in light of the climatology of precipitation from CRU. Seasonal 
climatology of radiation, humidity and wind speed were used to eliminate the 
potential CO
2
 variability related to these climate fields. Humidity and wind speed are 
important terms in the land surface energy budget; however, their variations have 
second order effects on vegetation growth and soil decomposition. There is a large 
discrepancy as to the role of radiation in the terrestrial carbon flux, even the sign of 
the effect is uncertain. For instance, reduced direct solar radiation during wet periods 
would reduce photosynthesis (e.g., Knorr, 2000) while increased diffuse light under 
cloudy conditions (e.g., Gu et al., 2003) increases photosynthesis leading to a carbon 
uptake. We thus chose to use the climatology of light, humidity and wind speed, in 
order to focus on the effects of the interannual variability of precipitation and 
 
 
 20 
 
temperature in this study. The atmospheric CO
2
 concentration was kept constant at 
pre-industrial level of 280 ppm since the year-to-year variation of CO
2
 level has a 
small impact on photosynthesis. The advantage of this offline simulation is that it 
gives prominence to the terrestrial responses to observed climate more directly than a 
fully coupled ecosystem model with model can, by eliminating model biases which 
would in turn affect the carbon model. The resolution of VEGAS is 1.0? x 1.0?. The 
steady state of model was reached by repeatedly using 1901 climate forcing. At this 
state, the global total GPP is 124 PgC/yr with NPP of 61 PgC/yr, and vegetation and 
soil carbon pools are 641 PgC, 1848 PgC, respectively. These are within the range of 
observationally based estimates (Schlesinger, 1991). This state was then used as the 
initial condition for the 1901-2006 run. We compare the results to an inversion model 
result by R?denbeck et al. (2003), which estimated CO
2
 fluxes by using a time-
dependent Bayesian inversion technique, based on 11, 16, 19, 26 and 35 observation 
sites of atmospheric CO
2
 concentration data from NOAA Climate Monitoring and 
Diagnostics Laboratory (CMDL) and a global atmospheric tracer transport model. 
2.3 Seasonal and interannual variability of terrestrial carbon fluxes 
in the Amazon Basin 
 
Station Km67 of FLUXNET is located in Tapajos National Forest in Brazil 
(2.8?S, 55.0?W). This important site is one of the few ones in the region to monitor 
and understand the characteristics of primary forest in Amazon basin. VEGAS 
modeled Net Ecosystem Production (NEE), Gross Primary Production (GPP), and 
Total Respiration (R
e
=R
a
+R
h
) are compared with those from the observation. 
 
 
 21 
 
Temperature and precipitation forcing VEGAS have been compared with those from 
the FLUXNET site and found to be in general agreement with them.  
Though the magnitude of VEGAS NEE and the FLUXNET observations are 
similar, it is notable that the phase of the seasonal cycle is opposite (Figure 2.2a). The 
inversion result by R?denbeck et al. (2003) shows a similar seasonal cycle to 
VEGAS. During the wet season (Jan-Aug), VEGAS simulates that land absorbs 
carbon because the modeled photosynthesis (GPP) is larger than the total respiration 
(R
e
) (Figure 2.2b), and results in negative NEE, implying carbon uptake by the land. 
During the dry season, vegetation activity is suppressed a lot by water supply in 
VEGAS, while respiration increases a bit with high temperature. As a result, a net 
loss of carbon is seen in the dry season. However, the observation indicates that high 
respiration can be maintained in the wet cool season and the vegetation growth is not 
suppressed in dry season. 
 
 
Figure 2.2 Seasonal cycles of carbon fluxes from VEGAS simulation and on-site observation in 
Tapajos National Forest (Km67: 2.8?S, 55.0?W), Brazil. (a) Land-atmosphere carbon flux (NEE) 
from VEGAS (climatology of 1991-2004), FLUXNET (climatology of 2002 -2004), and 
inversion (climatology of 1992 -2001); (b) GPP and R
e
 from VEGAS with FLUXNET. Unit in 
KgC/m
2
/yr. 
 
 
 
 
 22 
 
 
 
Figure 2.3 Seasonal cycle of land-atmosphere carbon flux from FLUXNET and two vegetation 
and carbon models: TEM (dotted curve) and IBIS (dashed curve) at Tapajos National Forest 
(adapted from Saleska et al., 2003).  
 
Saleska et al. (2003) compared the same FLUXNET observation with the 
modeling results from TEM (Terrestrial Ecosystem Model) and IBIS (Integrated 
Biosphere Simulator). These two terrestrial carbon models simulated a seasonal cycle 
of NEE similar to VEGAS (Figure 2.3). Saleska et al. (2003) suggested that the old-
grown trees in the Tapajos had access to deep soil water during the dry periods and 
show little evidence of water stress. Wood increment rates declined a little in the 
beginning of dry season, but maintained substantial increase just before the rains 
return, suggesting an adaptive mechanism of vegetation rather than a response to the 
seasonal climate forcing. This ?hydraulic? adaptive behavior of tropical big trees has 
not been considered in terrestrial carbon models before. In principle, terrestrial carbon 
models usually parameterize tropical vegetation dynamics based on the assumption 
that the tropics are a water-limited area. Furthermore, Saleska et al. (2003) proposed 
that respiration in soils, litter, and coarse wood debris is concentrated near the surface 
 
 
 23 
 
and inhibited by desiccation during the dry season. This implies that soil moisture, 
rather than temperature, may control soil decomposition.  
However, it is of interest that VEGAS modeled NEE anomaly generally 
agrees with the observations after the seasonal cycle is removed (Figure 2.4). Why 
does VEGAS simulate ?right? interannual variability of land-atmosphere carbon flux, 
while giving an ?out-of-phase? seasonal cycle against the observations? Further 
analysis of interannual variability of GPP and R
e
 indicates that in Tapajos the models 
and the observation may have different mechanisms of controlling the carbon uptake 
with respect to the climate variation on interannual timescales. VEGAS determines its 
interannual variability of NEE by NPP, while the observations attribute it to the 
respiration (not shown). However, sparse sites and short-term coverage of the 
observations in the tropical forest limit the model evaluation in the Amazon basin. 
The discrepancies between terrestrial carbon models (e.g., VEGAS, TEM and IBIS) 
and observations, though not resolved, have advanced our understanding of physical 
processes in the tropical forest area. Additional observation sites in the tropical forest 
and improvement of terrestrial carbon models are both called for. 
 
Figure 2.4 Anomaly of land-atmosphere carbon flux from VEGAS and FLUXNET after season 
cycle is removed in Tapajos National Forest; The inversion result is not plotted here because its 
simulation time coverage ends 2001. Unit in KgC/m
2
/yr. 
 
 
 24 
 
2.4 Seasonal and interannual variability of terrestrial carbon fluxes 
of boreal forest in Canada 
 
The station site in Old Black Spruce, Manitoba, Canada (55.9?N, 98.5?W) is 
covered with boreal evergreen forests and has records dating back nearly 10 years. In 
this site, vegetation begins to grow when the temperature returns above 0?C in 
VEGAS (GPP) (Figure 2. 5b). This matches the observations well. VEGAS tends to 
underestimate the amplitude of vegetation growth during summer time (75% of the 
observed GPP). After the summer season, VEGAS GPP and the observational GPP 
are comparable. Similar to GPP, VEGAS modeled R
e
 is also underestimated in 
summer season; but both agree with measurements in spring, fall and winter season. 
Although VEGAS GPP and R
e
 are both underestimated, their difference, NEE has a 
good agreement with the observational value in terms of seasonal phase (Figure 2.5a). 
The observations indicate that land takes up carbon from April to August, and the 
model extends this uptake to October but with low efficiency of carbon absorption. 
On the interannual time scale, there is an agreement of the NEE anomalies between 
VEGAS and the observations (Figure 2.6). The similarities of seasonality and 
interannual variability between VEGAS and the observations imply that VEGAS is 
able to reproduce the natural physical processes in the boreal forest. 
 
Figure 2.5 Same as Figure 2.2 but for Old Black Spruce, Manitoba, Canada (55.9?N, 98.5?W). 
 
 
 25 
 
 
Figure 2.6 Same as Figure 2.4 but for Old Black Spruce, Manitoba, Canada (55.9?N, 98.5?W). 
 
2.5 Evaluation of the interannual variability of VEGAS LAI with 
NDVI 
 
Here, an independent evaluation of the interannual variability of VEGAS was 
performed by comparing its simulated Leaf Area Index (LAI) to satellite observed 
Normalized Difference Vegetation Index (NDVI) (Figure 2.7). The high resolution 
1.0? x 1.0? satellite NDVI dataset of 1981-2004 comes from the Global Inventory 
Modeling and Mapping Studies (GIMMS) data set, which provides a 24-year satellite 
record of monthly changes in terrestrial vegetation (Tucker et al., 2001; Zhou et al., 
2001; http://gimms.gsfc.nasa.gov/). NDVI is the difference between the Advanced 
Very High Resolution Radiometer (AVHRR) near-infrared and visible bands divided 
by the sum of these two bands and can be used to quantify the density of plant growth 
on the Earth. Many studies have used NDVI as an indicator of vegetation growth to 
address its interannual-to-decadal variability in the northern hemisphere (Myneni et 
al., 1998; Zhou et al., 2001; Tucker et al., 2001). LAI is defined as the total leaf area 
per unit ground area. It is an important canopy parameter for ecosystem studies 
(Nemani and Running, 1989). In general, there is a close relationship between NDVI 
and LAI (Wardley and Curran, 1984). The spatial pattern of the temporal correlation 
between the filtered LAI simulated by VEGAS and the NDVI is indicated in Figure 
 
 
 26 
 
2.7 for the period of 1981-2004. The results indicate that the correlation is good in 
Northern and Southern Hemisphere extratropics, but poor in the tropics. As 
mentioned in the previous section, both the satellite observations and model 
simulation could have problems which could lead to the poor correlation in the 
tropics.  
 
Figure 2.7 Evaluation of VEGAS: the spatial pattern of correlation between VEGAS simulated 
LAI and the satellite observed NDVI between 1981 and 2004. A contour is used for shaded areas 
with value smaller than -0.15. Value of 0.33 is statistically significant at the 90% level of 
Student?s t-test. 
2.6 Summary  
In short, VEGAS is able to capture the interannual variation of the land-
atmosphere carbon flux against the observations in both Tapajos and Old Black 
Spruce. At the Tapajos site, VEGAS simulates an ?opposite? seasonal cycle of NEE 
as do other state-of-the-art vegetation models like TBM and IBIS. The reason is that 
the current terrestrial carbon models (VEGAS, TEM and IBIS) may not incorporate 
some important physical processes, such as a ?hydraulic? adaptive process for trees 
 
 
 27 
 
during the dry season as suggested by Saleska et al. (2003). But these processes are 
still poorly understood. The extent to which, soil moisture, temperature or both, 
control the heterotrophic respiration, is still subject to debate, and is in fact, even 
more poorly understood. The density of measurement stations in the tropics is still 
very sparse and the spatiotemporal patterns of carbon fluxes could vary with location 
or canopy type. The Tapajos site even only has a 3-year observational record. The 
similarity in the interannual variability of land-atmosphere carbon fluxes between 
VEGAS and observations is encouraging, and this provides an opportunity to 
investigate the response of terrestrial carbon ecosystem to climate variation (in 
Chapter 3, 4). In the boreal forest in North America, VEGAS is able to reproduce the 
vegetation growth, respiration, and land-atmosphere carbon flux seen at the station 
site. This agreement encourages the application of process-based modeling in 
studying the carbon absorption in the boreal forest. 
The evaluation of the interannual variability of VEGAS LAI against satellite 
observed NDVI indicates higher correlation in extratropics, but lower correlation in 
the tropics. Note that the satellite NDVI data may be of limited value due to cloud 
contamination and saturation of satellite signals (Knyazikhin et al., 1998; Myneni et 
al., 2002). 
 
 
 
 28 
 
 
Chapter 3: Response of the Terrestrial Carbon Cycle to the El 
Ni?o-Southern Oscillation 
 
3.1 Introduction 
An association between the CO
2
 growth rate and ENSO (Figure 3.1) was 
initially reported in the 1970?s and has been confirmed by recent statistical analysis 
(Bacastow, 1976; Keeling and Revelle, 1985; Braswell et al., 1997; Rayner et al., 
1999; Jones et al., 2001; Zeng et al., 2005a). Climate driven variations in the global 
carbon cycle have been attributed as the primary cause of the interannual variability 
of atmospheric CO
2
 growth rate after accounting for long-term trends of fossil fuel 
emission and land use change. The typical El Ni?o events are characterized by 
changes in atmospheric circulation and precipitation patterns (Ropelewski and 
Halpert, 1987) that give rise to warmer and drier conditions in the tropical land 
regions. Such a pervasive influence is likely to exert an impact on the land-
atmosphere carbon exchange. Those physical and biological responses of the 
terrestrial ecosystem to climate variation, especially to ENSO cycle, however, are not 
well known.  
Several studies have addressed the primary cause of ENSO-related terrestrial 
carbon cycle variation from different climatic points of view. For example, 
Kindermann et al. (1996) suggested that the temperature dependence of NPP is the 
most important factor in determining land-atmosphere carbon flux. However, 
precipitation has alternatively been suggested as the dominant factor for variation of 
 
 
 29 
 
terrestrial carbon cycle (Tian et al., 1998; Zeng et al., 2005a). Further, studies by 
Nemani et al. (2003) and Ichii et al. (2005) had indicated that in tropical terrestrial 
ecosystems, variations in solar radiation and, to a lesser extent, temperature and 
precipitation, explained most interannual variation in GPP. Besides these direct 
biological responses, fire due to climate change was also noted to play an important 
role in the variation of total land-atmosphere carbon flux (Page et al., 2002; 
Langenfelds et al., 2002; Hashimoto et al., 2004; Van der Werf et al., 2004). 
Hashimoto et al. (2004) confirmed the dependence of global fire carbon fluxes on 
interannual temperature variability. Van der Werf et al. (2004) estimated that during 
the 1997-98 El Ni?o, the anomalous carbon emission due to fire was 2.1 ? 0.8 PgC of 
carbon, or 66 ? 24% of the atmospheric CO
2
 growth rate anomaly in that period.  
 
Figure 3.1 Time-series indicating the correspondence of the CO
2
 growth rate (gCO
2
) at Mauna 
Loa, Hawaii with global total land-atmosphere carbon flux simulated by VEGAS and their lag 
with the Multivariate ENSO Index (MEI; units dimensionless), which is shifted up by 3 units. The 
seasonal cycle has been removed with a filter of 12-month running mean. The right top panel is 
the lagged correlation of MEI, F
ta
 and atmospheric CO
2
 growth rate with MEI where the X-axis 
indicates the time lag in months. Value of 0.33 is statistically significant at the 90% level of 
Student?s t-test.  
 
 
 30 
 
Zeng et al. (2005a) studied the interannual variability of the terrestrial carbon 
and have suggested that the tropical dominance is a result of a ?conspiracy? between 
climate and plant/soil physiology. During ENSO warm events, less precipitation and 
higher temperature in the tropical land tended to suppress vegetation growth and 
enhance soil decomposition, both contributing to more land carbon release to the 
atmosphere in the same direction. Though the results of these studies emphasizing the 
dominant factors influencing the interannual variability of the terrestrial carbon flux 
vary a lot, they provided insight and opportunities to explore the underlying physical 
and biological mechanisms of the interannual variability of atmospheric CO
2
 growth 
rate. Most of the studies cited above, however, have primarily focused on the effect of 
the climatic factors on photosynthetic processes (GPP, NPP). To date, few studies 
have been conducted on understanding the physical processes by which soil moisture 
affects the interannual variability of the terrestrial carbon flux. For instance, 
temperature has a direct impact on vegetation photosynthesis and soil decomposition. 
On the other hand, the temperature largely regulates the evapotranspiration, and thus 
influences the variation of soil moisture, which is an important factor for vegetation 
growth and soil decomposition. Through this, the temperature thus has an indirect 
impact on the vegetation growth and soil decomposition. This indirect influence has 
not yet been studied.  
The primary purpose of this present study is to provide insight into the 
mechanism of how ENSO-related precipitation and temperature variations regulate 
vegetation growth, soil respiration and land-atmosphere carbon flux. The following 
specific scientific questions will be addressed with relevant data sets: 
 
 
 31 
 
 z What are the robust spatial and temporal features of the terrestrial carbon 
ecosystem in response to ENSO in observations and can it be simulated 
by a process-based model?  
 z How do the variations of physical climate (e.g., temperature, 
precipitation, etc.) associated with ENSO govern terrestrial biophysical 
processes, such as the vegetation activity and soil decomposition? What 
are their relative contributions? Can the direct and indirect effects of 
temperature on the terrestrial carbon flux be quantified? 
To answer these questions, an offline simulation of VEGAS is used to 
investigate the response of the terrestrial carbon ecosystem to ENSO. The 
Multivariate ENSO Index (MEI) from NOAA is used as an indicator of ENSO signal. 
MEI is produced based on six observed variables over the tropical Pacific: sea-level 
pressure, zonal and meridional components of the surface wind, sea surface 
temperature, surface air temperature, and total cloudiness fraction of the sky (Wolter 
and Timlin, 1998; Markgraf and Diaz, 2000; http://www.cdc.noaa.gov/ENSO). The 
uninterrupted CO
2
 record for 1965-2006 at Mauna Loa Observatory (MLO) is used to 
calculate the atmospheric CO
2
 growth rate. Besides the inversion modeling of 
R?denbeck et al. (2003), a high resolution 1.0? x 1.0? NDVI dataset of 1981-2004 
was used. We will focus on the period 1980-2004 in view of the availability of 
observational NDVI dataset and the inversion model results. A caveat of the 
evaluation process is that the evaluating data themselves have limitations. For 
instance, inversion estimates themselves are based on transport models constrained by 
observational data. Gurney et al. (2003) have found significant differences in carbon 
 
 
 32 
 
flux between different inversion models. Similarly, in the tropics, the satellite 
produced NDVI data may be deemed suspect because of cloud contamination and 
saturation of NDVI (Knyazikhin et al., 1998; Myneni et al., 2002) as mentioned 
before. On the other hand, Saleska et al. (2007) have suggested that many terrestrial 
vegetation and carbon models do not have roots deep enough to capture growth 
during short-term droughts in the tropics, consistent with observations as discussed in 
Chapter 2. 
3.2 Interannual variability of terrestrial carbon flux 
Interannual fluctuations in Sea Surface Temperature (SST) in the tropical 
Eastern and Central Pacific are linked to global climate anomalies. For example, the 
changes of SSTs in the eastern tropical Pacific Ocean and associated deep convection 
perturb the Hadley circulation and produce a wavetrain starting with anomalous upper 
troposphere divergence in the tropics (Trenberth et al., 1998). Ropelewski and 
Halpert (1987) examined the link between ENSO events and regional precipitation 
patterns around the globe and noted that Northeastern South America from Brazil up 
to Venezuela had 16 dry episodes during the past 17 ENSO events. Others areas in 
the tropics, such as Indonesia, Australia, India etc, also showed a strong tendency to 
be dry during ENSO events. These dry conditions are usually followed by forest fires 
and decline of crop yield. Thus the climate variations associated with ENSO cycle 
tend to have a strong impact on the terrestrial biosphere in the lower latitudes, 
especially in the tropical forest areas. 
The global and regional F
ta
 simulated by VEGAS (The interannual variability 
of the time-series of VEGAS and the inversion model is calculated by removing the 
 
 
 33 
 
seasonal cycle and then filtering the data using a 12-month running mean to remove 
short term fluctuations.) was compared with the corresponding fluxes from the 
inversion model by R?denbeck et al. (2003), which conducted inversions using either 
inversion simulations with 11, 16, 19, 26 or 35 sites. We use the simulation of 11 sites 
for the following ENSO composite analysis because of its longest temporal coverage) 
(Figure 3.2). Overall, the interannual variability of F
ta
 by VEGAS and the inversion 
model show a good agreement globally with a correlation of 0.68 (11 sites). A major 
discrepancy between VEGAS and inversion is the period after the 1991 Mount 
Pinatubo eruption (Figure 3.2a). The relatively low atmospheric CO
2
 growth rate after 
Mount Pinatubo eruption has been discussed in the past but a consensus about the 
underlying mechanism has not yet been reached (Jones and Cox, 2001; Lucht et al., 
2002; Gu et al., 2003; Angert et al., 2004). The climatic suppression of respiration by 
the observed cooling after the eruption, and the enhancement of productivity by 
changes in diffuse light due to aerosols, have been suggested as possible reasons for 
this discrepancy. Most of the contribution to the global flux comes from the tropics 
(Figure 3.2b). The consistency of VEGAS and the inversion model in the tropics is 
reflected in the correlation value of 0.55 (11 sites). In contrast to the tropics, the 
interannual variation in the extratropics is smaller. The terrestrial carbon flux from 
the Northern Hemisphere extratropics by VEGAS lies within the range of the 
inversion results (Figure 3.2c).  
 
 
 34 
 
 
Figure 3.2 Interannual variability of land-atmosphere carbon fluxes (F
ta
) by VEGAS and 
inversion from R?denbeck et al. (2003) (lines for the inversion modeling with the CO
2
 
measurements from 11, 16, 19, 26 and 35 sites) for various regions: (a) Global; (b) Tropics 
between 22.5?S and 22.5?N; (c) Northern Hemispheric extratropics between 22.5?N and 90?N. 
Seasonal climatology is calculated based on the 1981-1999 time period for VEGAS and the 
inversion model with 11 observational sites; The climatology of the other inversion simulations 
are calculated with their respective time coverage. The filled triangles represent the ENSO events 
which have been selected in the composite analysis that follows in the text, with red triangles for 
El Ni?o events and green for La Ni?a events. The individual correlation of the land-atmosphere 
carbon fluxes between VEGAS and the inversion simulation with different number of stations is 
indicated in the 3 panels. The figure legend refers to the number of stations used in the inversion 
modeling calculations. All these correlation values are statistically significant at the 90% level of 
Student?s t-test. 
 
 
 
 35 
 
3.3 Robust features of the terrestrial response to ENSO 
Composite analysis is a useful method in climate studies to highlight the 
canonical features of a given type of variability without focusing on the differences 
between individual events (Rasmusson and Carpenter, 1982). Here similar events are 
averaged to emphasize the general aspects related to the growth and decay of the 
event. We used this method to investigate the robust response of the terrestrial 
ecosystem and associated climate variables during the ENSO cycle. El Ni?o and La 
Ni?a events used in this analysis have values that exceed a standard deviation of 0.75 
based on the MEI. These events are consistent with the analysis of Trenberth (1997). 
El Ni?o occurred in 1982-83, 1986-1987, 1991-92, 1993, 1994-95 and 1997-98 and 
La Ni?a occurred in 1984-85, 1988-89, 1995-96, and 1998-99. Because of the 1991 
Pinatubo eruption, 1991-92 and 1993 were not included in the following composite 
analysis. La Ni?a events were essentially treated as negative El Ni?o events, and 
combined to investigate the terrestrial response of ENSO. The composite analysis is 
similar to that of Rasmusson and Carpenter (1982). To highlight features related to 
the interannual variability, the composites used in this study were made after the 
seasonal cycle was removed and a filter was used to smooth the data. Figure 3.3 
shows a composite of the 2.5-year time evolution of ENSO events of the 
observational climate fields and simulated terrestrial responses for the whole globe 
(left panel), the tropics (central panel) and the Northern Hemisphere extratropics (the 
right panel). The global total land-atmosphere carbon fluxes from VEGAS (F
ta
), the 
inversion model, and the atmospheric CO
2
 growth rate resemble each other closely 
during the initiation phase of ENSO and exhibit a 6-month lag with the MEI index 
 
 
 36 
 
(Figure 3.3a). The maturity of El Ni?o occurs in November of ?year 0? of the ENSO 
event and the peak of the inversion flux anomaly lags by 4-5 months, followed by F
ta
 
and CO
2
 growth rate by 1-2 months later. The tropical land-atmosphere carbon flux 
anomalies have a similar magnitude and phase as the global fluxes, not only in 
VEGAS simulation but also in the inversion model (Figure 3.3b). In the tropics, the 
composite phase of the precipitation anomaly has nearly the same phase as MEI, but 
the response of modeled soil moisture is delayed because of its memory (Figure 
3.3h). The simulated vegetation activity (NPP) peaks 6 months after the MEI (Figure 
3.3e), more close to the phase of soil wetness rather than to precipitation. On the other 
hand, modeled soil decomposition R
h
 also peaks at about a 6-month lag with MEI. As 
a result, in the model, the decrease of NPP anomaly and increase of R
h
 anomaly 
contributes in the same direction and dramatically increase the F
ta
 anomaly during the 
ENSO warm events. In the middle and high latitudes of Northern Hemisphere, 
VEGAS F
ta
 anomaly phase is in general agreement with inversion flux model results. 
The response to ENSO is much weaker (Figure 3.3c). The variation of precipitation 
and temperature are in phase there, which is different from the out-of-phase 
relationship between precipitation and temperature in the tropics.  
To better understand the response of the terrestrial ecosystem to ENSO, lead-
lag correlations of tropical temperature, precipitation, simulated soil wetness and CO
2
 
growth rate, as well as terrestrial carbon fluxes with MEI during 1980-2004 were 
calculated (Figure 3.4). In this analysis, only the seasonal cycle was removed from 
data and no smoothing of the data was performed. This highlights the findings in the 
central panels of Figure 3.3 and gives the exact phase relationship. The peak of the 
 
 
 37 
 
atmospheric CO
2
 growth lags MEI about 5-6 months (R=0.45), similar to Zeng et al. 
(2005a). In the tropics, observed precipitation closely follows the MEI peak with 
about 1-month lag (R=-0.83), while temperature lags by 5-6 months (R=0.43). These 
climate patterns were suggested to be associated with the changes in atmospheric 
circulation during ENSO event (Ropelewski and Halpert, 1987). Because of the soil 
memory of water recharge, modeled soil wetness has a 5-month lag. Modeled NPP 
lags MEI by about 6 months (R=-0.8), with a 11-month lag for LAI by VEGAS, 
which suggests that vegetation does not show an immediate response to precipitation 
but rather responds to soil wetness more closely. On the other hand, R
h
 follows 
observed temperature closely and peaks with about a 6-month lag with the MEI. 
Thus, NPP and R
h
 simultaneously contribute to the F
ta
 lag of 6 months with the MEI 
(R=0.83) in the tropics in VEGAS. The ?perfect? out of phase relationship of NPP 
and R
h
 in the tropics is determined by the anti-correlation of precipitation and 
temperature pattern associated with ENSO events. 
 
 
 38 
 
 
Figure 3.3 Composites of the carbon fluxes and the climatic field anomalies during the growth, 
mature, and decaying phases of ENSO: The left (a, d, g), central (b, e, h) and right (c, f, i) panels 
are the global, the tropical (22.5?S to 22.5?N) and Northern Hemispheric extratropics (22.5?N to 
90?N) composites of the terrestrial fields. The top panels (a, b, c) are the terrestrial land-
atmosphere carbon flux anomalies of VEGAS (F
ta
) and inversion model, the atmospheric CO
2
 
growth rate (gCO
2
) and the reference MEI; The middle panels (d, e, f) are the corresponding Net 
Primary Production (NPP), and heterotrophic transpiration (R
h
), and the bottom panels (g, h, i) are 
the observed temperature, precipitation and modeled soil wetness (Swet). Units for carbon fluxes 
and CO
2
 growth rate are PgC/yr; mm/day  for precipitation, and ?C for temperature, and MEI and 
soil wetness are non-dimensional. The notation of ?yr 0?, ?yr +1?, ?yr +2? is the same as in 
Rassmusson and Carpenter (1982) in which ?yr 0? is the mature phase of ENSO. 
 
 
 39 
 
 
Figure 3.4 Lead-lag correlations of tropical terrestrial carbon fluxes and climatic fields with MEI. 
The observed CO
2
 growth rate (gCO
2
), temperature, precipitation, modeled NPP, Rh, LAI, F
ta
 and 
soil wetness are indicated; the solid magenta line is the autocorrelation of MEI with itself. 
Negative values in x-axis represent the number of months the corresponding variables lead the 
MEI and the positive values denote the number of months the variables lag by. Arrows indicate 
the peak response of each variable. A correlation value of 0.33 is statistically significant at the 
90% level of Student?s t-test. 
 
 Spatial patterns of the response of VEGAS 6 months after the peak of the 
ENSO phase are shown in Figure 3.5a, c, e, g. The tropical regions have robust 
positive anomalies of land-atmosphere carbon fluxes. The largest positive signals are 
located in eastern Brazil, and equatorial and southern Africa, consistent with the 
results of Jones et al. (2001). Southeast Asia and the east coast of Australia have 
sparse positive signals. Overall, the inversion model results (Figure 3.5b) and those of 
VEGAS agree well in the tropics though there are discrepancies in the extratropics. 
Modeled LAI from VEGAS and NDVI still show global agreement except in the 
tropics (Figure 3.5c, d). The regions associated with large positive anomalies of F
ta
 
 
 
 40 
 
regions consistently have large negative NPP anomalies in VEGAS (Figure 3.5e) and 
large positive R
h
 anomalies (Figure 3.5g) in the tropics, such as in the Amazon basin, 
consistent with the area averaged values in Figure 3.4b. These are also regions with 
correspondingly warm and dry conditions with negative signals of soil wetness 
(Figure 3.5g, h and Figure 3.4h). Zeng et al. (1999) proposed dynamical and 
hydrological mechanisms for anti-correlation between precipitation and temperature 
anomalies in the tropics. Decrease (increase) in precipitation tends to make the land 
surface dryer (wetter) and with less (more) evapotranspiration, less (more) 
evaporative cooling and then cause higher (lower) temperatures. Increase in 
absorption of solar radiation due to the absence of cloud during warm ENSO events is 
also likely to be a candidate for land surface warming. 
In the Northern Hemisphere extratropics, temperature anomaly is positively 
correlated with precipitation anomaly. The response of vegetation to ENSO in the 
extratropics is small (as seen in Figure 3.2 and Figure 3.3) and does not correspond to 
the 6-month lag response of vegetation in the tropics. However, interesting features 
emerge in Figure 3.5. North America exhibits a zonal dipole mode of F
ta
 anomalies. 
In the southwest U.S. the negative F
ta
 anomaly results from NPP positive anomalies 
surpassing R
h
 positive anomalies in strength. In contrast, the F
ta
 anomaly in the 
Northeast is dominated by negative NPP anomaly and positive R
h
 anomaly. Over the 
Eurasian continent, weak positive signals exist in East Asia, Northwest China, South 
Europe and Northeast Russia. In western Asia the large negative F
ta
 anomaly is 
attributable to soil moisture, which stimulates the vegetation activities and leads to 
large positive NPP anomaly. The scenario in other regions is more complex. Other 
 
 
 41 
 
modes of variability like the Arctic Oscillation (Buermann et al., 2003) may influence 
the high latitudes. It is interesting to note that the spatial pattern of NDVI anomalies 
generally match LAI anomaly in the middle and high latitudes except in Northeast of 
North America, which demonstrates the capability of VEGAS to generally reproduce 
the variability of vegetation activities.  
The carbon flux associated biomass burning is an important fraction of the 
total terrestrial carbon flux (Page et al., 2002; Langenfelds et al., 2002; Hashimoto et 
al., 2004; Van der Werf et al., 2004). In VEGAS, the carbon flux associated with 
biomass burning of vegetation and fast soil pool is included in the respiration of 
vegetation (R
a
) and in soil decomposition (R
h
). When fire occurs in VEGAS, leaves 
burn completely, while in reality only part of the wood gets burnt and the rest is 
lumped into the soil carbon pool and biomass burning of surface soil. Our simulation 
shows that the biomass burning accounts for about 25% of the total flux anomalies in 
the tropics during ENSO events (Figure 3.6), consistent with Zeng et al. (2005a). 
About 83% of the fire-related carbon release is included in the R
a
 and the remaining 
17% is included in R
h
. 
The composite analysis used above has treated La Ni?a events as the linearly 
?opposite? phase of El Ni?o events to capture the common features in the response of 
vegetation to the ENSO cycle and increase the number of events for statistical 
stability. To highlight the differences in the responses between the two phases of the 
ENSO, composites of El Ni?o and La Ni?a events are separately plotted in Figure 
3.7, for the atmospheric growth rate of CO
2
, the land-atmosphere carbon fluxes of 
VEGAS and inversion model. First of all, the land-atmosphere carbon fluxes of 
 
 
 42 
 
VEGAS and inversion, CO
2
 growth rate agree in phase both in El Ni?o and La Ni?a 
events (Figure 3.7a, b), respectively. However, the phase evolutions during El Ni?o 
and La Ni?a do differ. After La Ni?a matures, MEI decreases slowly and takes more 
than one and half years to decay. Though the lag time of F
ta
 anomaly increases with 
respect to MEI in La Ni?a cycles is at most 2 months more than for El Ni?o cycles, 
the maturity of La Ni?a carbon flux responses last many months longer. By 
examining the evolution of the two individual El Ni?o events from the 1980s, we find 
that El Ni?o events normally decayed quickly in the second year while La Ni?a 
events normally took longer to decay. For example, 1982 and 1997 events decayed 
quickly in July/August of the second year. This quick decay of El Ni?o normally 
accompanies a quick change of climate which determines the terrestrial carbon 
response. Similarly, during La Ni?a periods, the longer persistence of F
ta
, inversion-
calculated flux, and CO
2
 growth rate anomaly is the result of the slower decay of La 
Ni?a cycle. Not surprisingly, the later maturity of global F
ta
 anomaly in the overall 
ENSO composite comes from the contribution of the La Ni?a events to the 
composite. This is also true for the inversion fluxes and the CO
2
 growth rate 
anomalies. Although there are some phase differences of variables between El Ni?o 
and La Ni?a composites, the overall ENSO composite still highlights the most 
important features of terrestrial responses to ENSO cycle, while using all the data at 
hand.  
 
 
 43 
 
 
Figure 3.5 Spatial patterns of the anomalies of composite variables with 12-month averaging 
centered at the 6th month after MEI peak (as in Figure 3.3a) for: (a) Land-atmosphere carbon flux 
by VEGAS (F
ta
) in unit of gC/m2/yr; (b) Land-atmosphere carbon flux by the inversion 
(gC/m2/yr). (c) LAI by VEGAS; (d) Satellite observed NDVI. (e) Net Primary Production (NPP) 
in gC/m2/yr; (f) Precipitation in mm/day; (g) Heterotrophic Respiration (R
h
) in gC/m2/yr; (h) 
Surface air temperature in ?C. Note: the shading for NPP is opposite to R
h
 and F
ta
. 
 
 
 44 
 
 
Figure 3.6 Time-series of the anomalies of modeled carbon flux due to biomass burning and total 
carbon flux in the tropics. The top right panel is the ENSO composites of these two variables, 
indicating the contribution of biomass burning to total carbon flux anomalies. 
 
 
Figure 3.7 Tropical composites for (a) El Ni?o events only: 1982-83, 1986-87, 1994-95 and 
1997-98; (b) La Ni?a events only: 1984-85, 1988-89, 1995-96, 1998-99. The selected variables 
are the same as Figure 3.3b. 
 
3.4 Sensitivity Simulations to quantify the impact of climatic factors 
Physical climatic factors like precipitation, temperature and soil moisture, etc. 
govern the terrestrial carbon cycle in the model. To elucidate and quantify the effects 
of individual climate forcing, we carefully designed 3 sensitivity simulations and 
compared them to the standard simulation (Table 3.1). The standard simulation is the 
 
 
 45 
 
one in which temperature and precipitation are given from observations as discussed 
in the previous section. A ?climatology? simulation in which the precipitation and 
temperature climatology is fixed provides input to one of the 3 sensitivity 
simulations. The sensitivity simulations are as follows: (1) ?P-only? experiment, 
which uses the observed precipitation while using the climatology of temperature; (2) 
?T-only? experiment, which uses the observed temperature while using precipitation 
climatology; (3) ?Swet-fixed? experiment, which is like ?T-only?, except soil 
moisture is also from climatology, not simulated by SLand model. Specifically, this 
soil moisture climatology comes from the ?climatology? simulation described earlier. 
So, vegetation photosynthesis will see the variation of temperature; while the soil 
moisture changes due to evapotranspiration by temperature are excluded from 
vegetation photosynthesis in (3). This ?Swet-fixed? simulation thus informs us of the 
indirect temperature effect on NPP through soil moisture, in contrast to the ?T-only? 
simulation which includes both the direct effect of temperature and the indirect effect 
of temperature through soil moisture on NPP.  
ENSO composites of these sensitivity simulations help quantify the effects of 
temperature, precipitation, and soil wetness on NPP/ R
h
 / F
ta
 and are shown in Figure 
3.8 for the tropics. The period of maximum response (April to September of the ?year 
+1?) of the ENSO event is highlighted and the averaged values of those periods are 
calculated (Table 3.2). The percentage contribution of each climatic factor on F
ta
 is 
also provided (Figure 3.9). The period selected is based on the composite analysis and 
the lead-lag correlation in section 3.2.  
 
 
 46 
 
Table 3.1 Changes in physical climate forcing (temperature/precipitation) in the sensitivity 
simulations. The same seasonal climatology of radiation, humidity and wind speed as well as 
constant CO
2
 level for VEGAS, were used in all experiments listed. 
Experiment Physical climate forcing 
Climatology Climatology of temperature and precipitation, and soil moisture is 
simulated by SLand. 
 
Standard Observed temperature and precipitation, and soil moisture is simulated by 
SLand. 
 
P-only 
 
Observed precipitation while using the climatology of temperature, and soil 
moisture is simulated by SLand. 
 
T-only 
 
Observed temperature while using precipitation climatology and soil 
moisture is simulated by SLand. 
 
Swet-fixed 
 
In addition to ?T-only?, soil moisture is provided with the climatology from 
the ?climatology? experiment rather than modeled by SLand in ?T-only?. 
 
 
The mean global total of the F
ta
 anomaly for the selected period in the 
standard run is 1.68 PgC/yr, which is a little larger than that of the inversion model 
flux with 1.36 PgC/yr
 
(Table 3.2). Both of them are larger than CO
2
 growth rate 
which is 1.22 PgC/yr. This difference can be attributable to the net oceanic carbon 
uptake from the atmosphere during El Ni?o event and is within the range of previous 
estimates (Keeling and Revelle, 1985; Winguth et al., 1994; Feely et al., 2002; 
Bousquet et al., 2000, Plattner et al. 2002).  
 
 
 
 47 
 
 
Figure 3.8 ENSO-composites of carbon fluxes for the sensitivity simulations: (a)VEGAS F
ta
 
response in the tropics for the standard, ?P-only?, ?T-only? and ?Swet-fixed? sensitivity 
simulations; (b) the same as (a) except for NPP; (c) the same as (a) except for heterotrophic 
respiration. The shaded period of Apr.-Sep. of ?yr +1? is the maximum regime of the response to 
the ENSO cycle. The carbon fluxes averaged in this regime are indicated in Table 3.2. Units in 
PgC/yr. 
 
Table 3.2 Mean of composite global and tropical carbon fluxes during the peak response to 
ENSO (shaded area in Figure 3.8 corresponding to the 6-month average from April to September 
of ?yr +1?) for VEGAS standard, ?P-only?, ?T-only?, ?Swet-fixed? experiments and CO
2 
growth 
rate. Units in PgC/yr. 
Region R?denbeck
a
 Standard P-only T-only Swet-fixed 
 
RF 
F
ta
 (NPP/R
h
) F
ta
 (NPP/R
h
) F
ta
 (NPP/R
h
) F
ta
 (NPP/R
h
) 
Global 
1.36 1.68(-0.99/0.69) 0.74(-0.65/0.09) 0.95(-0.34/0.60) 0.79(-0.17/0.62) 
Tropics 
1.19 1.77(-1.21/0.56) 1.00(-0.86/0.14) 0.79(-0.34/0.44) 0.59(-0.13/0.46) 
gCO
2
  
1.22 
a
 R?denbeck et al. (2003) inversion modeling with 11 station measurements. 
 
 
 48 
 
The tropical land-atmosphere carbon flux dominates the global total both in 
VEGAS and in the inversion model. However, in the standard VEGAS simulation, 
the extratropics acts as a relatively small carbon sink, while the inversion model 
indicates it as a small carbon source. In the tropics, out of the 1.77 PgC/yr
 
F
ta
 
variation, 1.21 PgC/yr comes from the suppression of vegetation (NPP) and the rest 
0.56 PgC/yr
 
from soil decomposition (Figure 3.8b, c). Our sensitivity simulations 
show a linear relationship of the effects of anomalous temperature and precipitation, 
on the terrestrial carbon cycle, particularly at the peak of the response. In reality, such 
effects, however, are extremely complicated and not well known. One possible reason 
is that the interannual variability can be considered as a small perturbation to the 
system (VEGAS), which itself is highly non-linear. Assuming an approximately 
linear relationship we can estimate that precipitation contributes about 1.0 PgC/yr, 
while temperature provides about 0.79 PgC/yr of the F
ta
 variation in the standard 
simulation. In the tropics, the ?P-only? sensitivity experiment produced a change of -
0.86 PgC/yr
 
for the NPP anomaly, while soil decomposition decreases dramatically to 
0.14 PgC/yr
 
from 0.56 PgC/yr during this period. In the ?T-only? simulation, the soil 
decomposition is remarkably enhanced due to warming, while less reduction of 
vegetation activity is seen. Spatially, the distribution of the regions with remarkable 
reduction of NPP in ?P-only? is similar to the one in standard run, albeit with a 
smaller intensity (figures not shown). Compared to standard run, the regions with 
positive signal of soil decomposition shrink to encompass only Northern South 
America and Center Africa in the ?P-only? run. The remarkable enhancements of soil 
decomposition due to warming and dramatic vegetation suppression due to drought 
 
 
 49 
 
lie in the nature of the tropics, which are highly-decomposed and water-limited on the 
interannual timescale. 
The absorption of water for vegetation depends on the water pressure level 
difference between the soil moisture and vegetation root. Thus vegetation activity 
closely follows soil moisture, rather than immediately responding to precipitation. 
However, soil moisture can be influenced by temperature through evapotranspiration. 
Thus precipitation affects NPP mainly through soil wetness, while temperature can 
influence NPP variation through soil moisture. In the ?Swet-fixed? simulation, only -
0.13 PgC/yr
 
of NPP occurs compared to the -0.34 PgC/yr
 
reduction in the ?T-only? 
simulation, which reflect the indirect effect of temperature on the NPP through soil 
moisture. Spatially, the central Amazon basin and South Africa have much less 
vegetation suppression in the ?Swet-fixed? run (spatial pattern not shown). The 
?Swet-fixed? sensitivity experiment indicates clearly that the indirect effect of 
temperature on the vegetation photosynthesis through soil wetness is significant, even 
a little larger than the direct temperature effect on the photosynthetic processes in the 
tropics in model. This is interesting because the tropics always has favorable 
temperature conditions for vegetation year-round, while the change in water supply 
during the ENSO period can drastically affect the vegetation growth.  
Figrue 3.10 is a schematic diagram of the percentage of relative contribution 
of the individual and combined effects of precipitation and temperature on the 
variation of F
ta
 in the tropics during ENSO. Values used to calculate the percentages 
were obtained from Table 3.2 which correspond to a 6-month average (April to 
September) at the peak of the ENSO response. Some are obtained as a direct result of 
 
 
 50 
 
the simulations, while the rest are obtained as residuals. The tropical ENSO response 
of the total carbon flux into the atmosphere is a result of 68% of photosynthesis and 
32% of soil decomposition as obtained from the standard simulation. We estimated 
that out of the total land-atmosphere carbon flux, about 25% of that is due to biomass 
burning, which is accounted in the vegetation and soil respiration portion. 
Precipitation is the major contributor to photosynthesis variability while temperature 
controls soil decomposition. In terms relative to the variation of F
ta
, precipitation 
contributes about 56% of the variation mainly through soil wetness in photosynthesis, 
while temperature accounts for 44%, as seen from the total flux of the ?P-only? and 
?T-only? experiments. The total effect of temperature arises from direct and indirect 
effects. The direct effect is through a 26% influence on soil respiration and a 7% 
influence on respiration of vegetation as obtained from the ?Swet-fixed experiment?. 
The remaining contribution of temperature is through its indirect effect by which 
photosynthesis is affected by temperature changes affecting evaporation and thus 
changing soil moisture. This indirect effect of temperature is 12% of the total 
variation and is obtained as the difference between the ?Swet-fixed? and ?T-only? 
experiment. This indirect effect is highly complicated and has not been studied in the 
past. The total effect of soil wetness on the total flux into the atmosphere is obtained 
as a residual of the remaining terms, and 61% affects photosynthesis, while 6% 
affects soil decomposition. 
 
 
 
 
 
 
 
 
 
 51 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Figure 3.9 A conceptual diagram of the mechanism of tropical terrestrial carbon response to the 
ENSO cycle. The percentages refer to the contribution to F
ta
 variation (100%), with carbon flux 
anomalies in parenthesis (PgC/yr). These carbon fluxes are estimated from Table 3.2. The climate 
fields, such as precipitation, temperature and soil wetness are shown as rectangles and the 
corresponding NPP/Rh/ F
ta
 as rounded rectangle. The dashed arrow indicates the indirect effect of 
temperature on NPP through soil wetness. The solid lines in black from temperature are the direct 
effects of temperature on soil decomposition and on NPP. All processes are contributable to F
ta 
in 
the same direction (for example, increase of soil decomposition and decrease of vegetation growth 
during El Ni?o) 
 
 
 
 52 
 
3.5 Conclusion and discussion 
In this study, we have used VEGAS to investigate the response of the 
terrestrial carbon ecosystem to the physical climate variations and to isolate those 
effects, especially the ones associated with ENSO. The results of the model are first 
evaluated against observations to verify its capability in capturing the interannual 
variability. The simulated global total land-atmosphere carbon flux (F
ta
) in our study 
agrees well with the results of inversion modeling based on the global observations 
on interannual timescales, especially in the tropics. Both of them are in agreement 
with the atmospheric CO
2
 growth rate variation. The land-atmosphere carbon flux in 
the tropics has a similar magnitude as the global total, confirming the dominant role 
of the tropics on interannual timescale. This underscores the importance of extending 
the observational network in the tropics in the future. 
Our analysis of ENSO composite and lead-lag correlation indicate that in the 
tropics VEGAS simulates a robust signal of F
ta
 with a 6-month lag of the ENSO 
cycle. This is largely caused by the suppression of vegetation and enhancement of soil 
decomposition, as suggested by the ?conspiracy? theory of Zeng et al. (2005a). 
Though the anomalous precipitation lags the ENSO cycle by 1-2 months, the 
vegetation-dependent soil moisture takes about 5 months to respond due to soil 
memory. Less precipitation, with higher temperature drains the soil moisture and 
results in drought condition unfavorable for vegetation activities. This results in a 6-
month lag of the NPP to ENSO. During El Ni?o, large areas of vegetation 
suppression occur in Brazil, equatorial Africa and South Africa. Simultaneously, 
higher temperature directly enhances soil decomposition in northern Brazil, equatorial 
 
 
 53 
 
Africa. As a result, reduction of NPP and enhancement of R
h
 contributes in same 
direction and provide a large amount of carbon release to the atmosphere. The 
opposite is true for La Ni?a, although the time-dependence of the lags differs from El 
Ni?o. 
Sensitivity simulations were performed to elucidate and quantify the effects of 
precipitation and temperature on the tropical terrestrial ecosystem. During ENSO 
events, the precipitation and temperature have very different mechanisms in 
regulating the above variation of vegetation growth and soil decomposition. A near 
linear relationship of terrestrial carbon response to the anomalous precipitation and 
temperature was found in model. In total, precipitation variation contributes to about 
56% of F
ta
 variation, mostly through soil wetness in the photosynthesis process, while 
temperature accounts for 44%, which includes 25% from direct effect on the soil 
decomposition, 7% from its direct effect on vegetation respiration and 12% from its 
indirect effect on the photosynthesis through soil wetness. Such an indirect effect is 
significant, and has not been studied in the past. As a result, in the tropics vegetation 
activity suppression (effects on NPP) contributes 68% of variation of F
ta
, with the 
remaining 32% coming from soil decomposition under a warmer and drier condition. 
Out of the total land-atmosphere carbon flux, about 25% is due to biomass burning, 
which is included in the vegetation and soil respiration. 
Due to spatial cancellation and weak ENSO teleconnections, the response of 
the terrestrial system in middle and high latitudes is less robust and more complex. 
The interannual variability of the land-atmosphere carbon flux in the extratropics is 
complicated by other modes of climate variability, such as the Pacific North 
 
 
 54 
 
American pattern of variability, the Atlantic Oscillation and Northern Annular Mode 
(Potter et al., 2003; Buermann et al., 2003; Russell and Wallace, 2004). Moreover, the 
role of solar radiation in control of vegetation activity is still in debate and has not 
been addressed in this study (Gu et al., 2003; Angert et al., 2004).  
The results in this study are based on a single vegetation model and hence 
could be model dependent. Yet it has been useful in addressing the canonical 
response of the terrestrial carbon flux to ENSO. The linear response observed in the 
tropics helped quantify the mechanisms through which precipitation and temperature 
affect the carbon flux directly and indirectly, thus highlighting the importance of soil 
moisture on vegetation. Improved observations and advanced modeling studies are 
required to quantify exactly the terrestrial carbon cycle under changing climate. 
 
 
 
 
 55 
 
 
Chapter 4: Impact of 1998?2002 Midlatitude Drought and 
Warming on Terrestrial Ecosystem and the Global Carbon Cycle 
 
4.1 Introduction 
The interannual variability in the atmospheric CO
2
 growth rate is dominated 
by tropical land response to ENSO due to the constructive plant and soil physiology 
affected by patially coherent tropical climate anomalies (Bousquet et al., 2000; Zeng 
et al., 2005a; Qian et al., 2008). Changes in the carbon sources and sinks in 
Midlatitude regions, on the other hand, often tend to cancel each other, so the overall 
variability is weakly correlated with ENSO and it contributes only to a small fraction 
of the interannual variability in atmospheric CO
2
. 
A growth rate of over 2 ppm/yr during 2002?2003 was unprecedented from 
the MLO station record (Jones and Cox, 2005). Although the yearly growth rate was 
somewhat smaller than during the short-lived 1997?98 El Ni?o event, the 2001?2003 
growth was sustained longer so that the bi-yearly peak is highest ever when the data 
is smoothed with 24 month running mean (Figure 4.1a). In the meantime, the 1998?
2002 drought spanned much of the Northern Hemisphere Midlatitudes, with great 
severity in the regions of the western US, southern Europe, Southwest Asia, eastern 
Asia and Siberia (Waple et al., 2002; Hoerling and Kumar, 2003) (Figure 4.2a), with 
wide spread impacts such as wildfires in the western US.  
In this section, we examine VEGAS offline simulation results to establish the 
relationship between the drought and the anomalous behavior in the carbon cycle, 
 
 
 56 
 
which serves as the clue for determining the causes of such an anomalously higher 
CO
2
 growth rate during 2002?2003. Atmospheric CO
2
 inversion model results 
(R?denbeck et al., 2003), and satellite observed vegetation index NDVI (Tucker et al., 
2005) are used to compare with the VEGAS simulation. 
 
Figure 4.1 (a) Growth rate of atmospheric CO
2
 observed at Mauna Loa, Hawaii from 1980 to 
2003; a 12 month running mean (green) was used to remove the seasonal cycle, and 24 month 
running mean (red) was used to emphasize the lower frequency variability. (b) VEGAS simulated 
total land-atmosphere carbon flux (black), compared to Mauna Loa CO
2
 growth rate (green, 
labeled on the right) and ENSO signal (the negative southern oscillation index: -SOI, in purple; 
mb labeled on the left). (c) VEGAS simulated global (black), tropics (20?S-20?N, in green) and 
Northern Hemisphere Midlatitude (20?N-50?N, in red). (d) Same as in (c) but from the 
atmospheric inversion of R?denbeck et al. (2003). Seasonal cycle has been removed from all 
figures except where otherwise noted. Shading is for the June 1998?May 2002 period during 
which Northern Hemisphere Midlatitude released anomalously large amount of CO
2
, modifying 
the normally tropically dominated ENSO response both in terms of amplitude and phasing. 
 
 
 57 
 
 
Figure 4.2 Anomalies for the period June 1998?May 2002 relative to the climatology of 1980?
2003 for: (a) precipitation (Xie and Arkin, 1996) normalized by local standard deviation, (b) 
VEGAS modeled land-atmosphere CO
2
 flux (kgC /m2/yr). 
 
4.2 Drought and CO
2
 
VEGAS modeled global land-atmosphere carbon flux variability shows a 
general agreement with MLO CO
2
 growth rate (Figure 4.1a, b), as well as with the 
global flux from an atmospheric inversion model study (R?denbeck et al., 2003) 
(Figure 4.1d). On regional scale, both tropical and Northern Hemisphere Midlatitude 
show overall agreement with inversion results. However, in the drought period of 
1998?2002, Midlatitude land regions are largely CO
2
 sources to the atmosphere 
(Figure 4.2b). The most pronounced source anomalies are in the western US, 
Southwest Asia and Northeast Asia, while only small sinks are found in Canada and 
northern Europe. Although the detailed spatial pattern of these anomalies varied 
somewhat over time (not shown), the persistence of this Midlatitude drought and 
 
 
 58 
 
carbon cycle response is striking, as during the same 4 year period, the tropics 
switched sign from the large 1997?98 El Ni?o to 1999?2001 La Ni?a and back to the 
modest 2002?2003 El Ni?o.  
Over most of the period, the Northern Hemisphere land between 20?N and 
50?N is a nearly constant source releasing 1.3 PgC/yr more than usual relative to the 
1980-98 average. The inversion for the same region (Figure 4.1d) shows similar trend, 
with an anomalous flux from 1998-2001 larger than any other time during the 22 year 
period of the inversion. In particular, carbon flux changed from 1980?98 mean sink 
of 0.7 PgC/yr to nearly neutral for 1998?2001, with excursions as source in 1998 and 
2001. Part of the difference between VEGAS and the inversion model may be that 
drought impact was partially alleviated by heavy land management such as irrigation 
in these regions. Since VEGAS did not include anthropogenic CO
2
 emission, volcanic 
eruptions, or oceanic uptake effects, the modeled total land-atmosphere flux is overall 
shifted compared to inversion so that only the relative changes are comparable.  
Another clue of the importance of the Midlatitude drought on the carbon cycle 
comes from the timing of the events. Typically, MLO CO
2
 growth rate correlates well 
with ENSO signal  with maximum correlation 0.59 at a lag of 5 months (here we use 
Southern Oscillation Index: SOI) (Zeng et al., 2005a, Qian et al., 2008). Though there 
is some difference between MEI and SOI (Wolter and Timlin, 1998), both of them are 
good ENSO phase indices with respect to the relationship to CO
2
 growth rate. This 
lag is due to the delayed response in hydrology and biological activity to ENSO-
related climate anomalies in the tropics. During 1998?2003, this normality was 
disrupted by the drought-related Midlatitude carbon release so that the CO
2
 recovery 
 
 
 59 
 
to higher post-ENSO event values after the 1999?2001 La Ni?a arrived several 
months earlier than usual. For instance, there is an early rise in MLO CO
2
 growth rate 
in 2000, and the peak during 2002?2003 slightly leads SOI, rather than following the 
normal 5-month lag. If the Northern Hemisphere anomalies had been zero during 
1998?2002, the tropical anomaly alone would have produced a significantly smaller 
CO
2
 increase (Figure 4.1c), whereas observations shows a sustained and 
unprecedented 2?3 year period of fast growth in CO
2
.  
4.3 Regional contributions and mechanisms 
The overall agreement between model simulated land-atmosphere carbon flux 
and MLO CO
2
 growth rate suggests the usefulness of further analysis of regional 
characteristics and mechanisms. The modeled leaf area index (LAI; Figure 4.3a) 
follows precipitation closely with reduction in the western US and northern Mexico, 
Southwest Asia, and East Asia, while LAI increased in northeastern Canada and 
central Europe. The LAI increase in the latter two regions is partly due to warming 
(Figure 4.3f) because temperature is a limiting factor for growth, especially in 
northeastern Canada. The modeled LAI shows a generally good agreement with the 
satellite observed normalized difference vegetation index (NDVI), but some 
differences exist particularly in Southwest Asia where reduced NDVI is somewhat 
north of the modeled LAI. Because the spatial anomaly patterns varied over time 
(Lotsch et al., 2005), modeled LAI and NDVI may compare better on a year to year 
basis. More importantly, land use and management that was not included in the model 
likely have modified the natural response significantly. The general agreement in 
these regions is in some sense remarkable, suggesting natural climate variability 
 
 
 60 
 
nonetheless manifests itself prominently. 
 
 
Figure 4.3 Anomalies of the period June 1998?May 2002 for (a) VEGAS modeled Leaf Area 
Index (LAI); (b) observed normalized difference vegetation index (NDVI); (c) modeled Net 
Primary Production (NPP, kgC/m
2
/yr); (d) land-atmosphere flux from inversion of R?denbeck et 
al. (2003) with 11 CO
2
 stations (kgC/m
2
/yr) for 1998?2001; (e) modeled soil respiration (R
h
, 
kgC/m
2
/yr); (f) observed surface air temperature (?C). 
 
 
 61 
 
The simulated anomalies in NPP in the Northern Hemisphere Midlatitude 
largely follow the precipitation change. However, the spatial extent of regions with 
reduced NPP is larger while the area with positive NPP anomalies shrank (see Figure 
4.3c and 4.2a). This is mostly due to the enhanced autotrophic respiration in response 
to general warming over this period (Figure 4.3f). This tendency is further enhanced 
in the total land-atmosphere carbon flux (Figure 2b) because heterotrophic soil 
respiration (Figure 4.3e) also increased in response to the warming. As a result, the 
Northern Hemisphere Midlatitude was predominantly a CO
2
 source to the atmosphere 
during 1998 ?2002 with a spatial extent larger than the area affected by reduced 
precipitation.  
Regional patterns from the inversion (R?denbeck et al., 2003) (Figure 4.3d) 
indicate that from May 1998 to October 2001, most of North America is a carbon 
source especially in the US. There is a moderate carbon sink in central Europe, 
consistent with our forward model (Figure 4.2c). The inversion also shows a band of 
source areas covering southwest to central and East Asia, albeit very weak compared 
to modeled anomalies. The inversion may not resolve the sub-continental variations 
especially in Eurasia due to the lack of CO
2
 station there.  
From 1980 to 2003, VEGAS land-atmosphere carbon flux shows high 
correlation with precipitation (Figure 4.4a-c), and different regions differ in their 
detailed temporal evolutions. The drought in Eurasia (Figure 4.4c) started early and 
peaked in 1999?2000, with largest contribution from Southwest Asia for the first half, 
and significant contribution from Northeast Asia for the latter half (not shown). In 
contrast, drought in western US occurred between 2000 and 2003. Similar to the 
 
 
 62 
 
forward model, the atmospheric inversion shows an anomalous increase early in 1998 
from Eurasia and another increase from North America in 2000?2001. However, 
these two events are further apart in time in the inversion, so that the total flux has a 
minimum in 1999, while the forward model has a more sustained peak from 1998 to 
2002.  
Temperature also played an important role during 1998?2002 and in the 
overall evolution. For instance, the total Midlatitude carbon flux during 1998?2002 is 
0.7 PgC/yr
 
larger than the period 1984?1989, while precipitation is only modestly 
smaller (Figure 4.4a). This is partly because the recent drought hit the more sensitive 
semi-arid regions, and partly because the long-term warming trend leads to more 
respiration in the 1998?2002 period so that the net carbon flux is significantly higher. 
The importance of warming on carbon loss can be seen more clearly in two model 
sensitivity experiments in which either precipitation or temperature alone were used 
to force the physical land-surface and the carbon model (Figure 4.4e). During the 
1998?2002 period, temperature-induced carbon release had somewhat smaller peak 
amplitude but lasted longer than the precipitation-induced anomalies. Interestingly, 
the temperature effect was significantly smaller during previous periods such as 
1984?1989, another indication that the recent drought was unusual. The pathway by 
which warming influences carbon flux is through the direct effect of more respiration 
loss, and though an indirect effect as higher temperature leads to larger evaporation 
and also more severe drought (Wetherald and Manabe, 1995).  
 
 
 63 
 
 
Figure 4.4 Observed precipitation (normalized by standard deviation; green) and temperature 
(red; not labeled: the range from minimum to maximum is 1.6 ?C), and VEGAS modeled land-
atmosphere carbon flux (black) for (a) Northern Hemisphere Midlatitude (20?N?50?N); (b) North 
America 20?N?50?N; (c) Eurasia 20?N?50?N. Also plotted in (d) is carbon flux for the same three 
regions from the inversion. (e) Modeled Northern Hemisphere Midlatitude carbon flux from two 
sensitivity experiments using ?P-only? or ?T-only? as forcing. 
 
 
 64 
 
4.4 Conclusion and discussion 
While climate models generally predict Midlatitude drying under global 
warming (Wetherald and Manabe, 1995), the changes in precipitation patterns and 
subsequent terrestrial carbon response are highly uncertain (Zeng et al., 2004). Our 
results suggest that the 1998?2002 Midlatitude drought was a response to reduced 
precipitation and increased temperature, with widespread consequences to the 
terrestrial ecosystem and the global carbon cycle, highlighted by the anomalous 
increase in atmospheric CO
2
 growth rate in recent years (Jones and Cox, 2005). 
However, new record of CO
2
 emissions from fossil-fuel burning and industrial 
processes may cloud the contribution from the land-atmosphere carbon flux for the 
higher CO
2 
growth during 2002-2003. Raupach et al. (2007) found that CO
2
 
emissions have been accelerating at a global scale, with their growth rate increasing 
from 1.1%/yr for 1990?1999 to >3%/yr for 2000?2004. Further analysis to quantify 
the contribution of CO
2
 emission and biosphere in response to climate variation thus 
are needed to provide a close accounting for the recent acceleration of CO
2
 growth 
rate. 
However, this 1998?2002 Midlatitude drought is unusual. The spatial 
extension of it has been attributed to a synergy of surface temperature changes across 
the tropical oceans. In addition to the usual changes associated with a La Ni?a cold 
event in the eastern Pacific Ocean, this wide-spread drought was significantly 
influenced by simultaneous warming in the western Pacific and the Indian Ocean 
(Hoerling and Kumar, 2003). Although not a perfect analog of global warming, since 
such conditions may become more likely (Stott et al., 2004), this event provides a 
 
 
 65 
 
glimpse into the possible future carbon cycle response to climate change.  
 
 
 66 
 
 
Chapter 5: How Strong is Carbon Cycle-Climate Feedback 
under Global Warming? 
 
5.1 Introduction 
More than half of the anthropogenic CO
2
 emission has been taken up by sinks 
in the ocean and over land (Prentice et al., 2001). The magnitude of future climate 
change depends critically on the behavior of these carbon sinks. One major feedback 
involves the change in these carbon sinks in response to climate change such as 
changes in temperature and precipitation patterns. Coupled carbon-climate modeling 
from the Hadley Centre (Cox et al., 2000; Betts et al., 2004) and IPSL (Friedlingstein 
et al., 2001; Dufresne et al., 2002; Berthelot et al., 2002) which took into account 
such feedbacks, showed large uncertainties in the predicted strength of carbon-
climate feedback and its impact on climate prediction. For instance, changes of the 
terrestrial carbon pools in these two models differ not only in magnitude but also in 
the direction. Here we present results from a fully coupled carbon-climate model and 
discuss the similarities and additional differences from the above two predictions. 
5.2 Model and methodology  
In Chapters 2, 3 and 4, VEGAS was run in offline mode. In this chapter, we 
build an Earth system model. It includes two major components: physical climate and 
carbon cycle (Figure 5.1). The physical climate component of the model usually 
consists of atmosphere/land/ocean models and the carbon cycle component includes a 
land/ocean carbon cycle model. Here we focus on the carbon cycle, so the Earth 
 
 
 67 
 
system model does not have other important biogeochemical cycles, such as Nitrogen, 
Sulfur, etc. This Earth system model is named UMD in our study. In UMD, the 
physical climate component consists of a global version of the atmospheric model 
Quasi-equilibrium Tropical Circulation Model (QTCM) (Neelin and Zeng, 2000; 
Zeng et al., 2000) and a slab mixed-layer ocean model with Q-flux to represent the 
effects of ocean dynamics (Hansen et al., 1983). The mixed-layer ocean depth is the 
annual mean derived from Levitus et al. (2000). QTCM is of intermediate complexity 
suitable for the modeling of tropical climate and its variability. It occupies a niche 
among climate models between complex general circulation models and simple 
models. The primitive-equation-based dynamical framework is constructed using 
analytical solutions from the quasi-equilibrium convective parameterization as the 
first basis function of a Galerkin representation of vertical structure. A uniqueness of 
the QTCM is its balanced treatment of dynamics and physical parameterizations. It 
includes a linearized longwave radiation scheme, simple cloud prediction and 
shortwave radiation schemes (http://www.atmos.ucla.edu/~csi/qtcm_frm.html). The 
land surface model SLand is included in QTCM. UMD?s carbon cycle component 
includes VEGAS and a box ocean carbon model. The box ocean carbon model has 3 
ocean carbon pools with different turnover times: upper ocean carbon, deep ocean 
carbon, and sediment carbon. The ocean-atmosphere carbon flux exchange depends 
on the sea temperature. However, UMD doesn?t consider complicate oceanic 
biochemical processes which could affect carbon sequestration. 
 
 
 68 
 
 
Figure 5.1 A conceptual diagram of the Earth system, divided into components consisting of 
physical climate and carbon cycle. The arrows indicate the major physical and biological 
processes, including those affecting carbon. The arrows in red are associated with the human 
activities for the carbon cycle. In UMD, the physical climate component includes an atmospheric 
model QTCM, a slab mixed-layer ocean model and the SLand unit, which handles processes such 
as soil moisture storage, but not carbon. VEGAS and a box ocean carbon model serve as the 
carbon cycle components.  
 
Experiments conducted with UMD are indicated the in Figure 5.2. The fully 
coupled carbon-climate model (UMD) was run to a pre-industrial steady state (apart 
from high frequency internal variability) at year 1790. During this spinup process, the 
atmospheric CO
2
 was nudged to an observed value of 281 ppm so that the carbon 
pools and climate simulated are close to observations. The model was then run in a 
freely coupled mode from 1791 to 2100 (results analyzed for 1860?2100), with no 
other changing external forcing except for the anthropogenic CO
2
 emission, taken 
 
 
 69 
 
from the IPCC-SRES A1B scenario. This run is referred to as the coupled run. In 
order to delineate the effects of carbon-climate feedback, another run was conducted 
by using a constant CO
2
 level of 281 ppm in the longwave radiation module of the 
atmospheric model. Thus the carbon model sees a nearly constant climate without 
global warming, but carbon components are fully interactive including CO
2
 
fertilization effect and emission. This run is termed as the uncoupled run. Such an 
experiment has been referred to as an ?offline? simulation by Cox et al. (2000), or 
?prescribed climate? by Dufresne et al. (2002), and uncoupled by Friedlingstein et al. 
(2003). The difference between the coupled and the uncoupled run is an indicator of 
the strength of carbon-climate feedback. 
 
Figure 5.2 Model experiments: (a) Coupled: the physical climate and carbon cycle model are 
fully coupled from 1791 to 2100, with no other changing external forcing except for the 
anthropogenic CO
2
 emissions. (b) Uncoupled: the carbon model sees a nearly constant climate 
without global warming, but carbon components are fully interactive including CO
2
 fertilization 
and emission. Then an additional simulation where the CO
2
 is from uncoupled run was used to 
force the physical climate model to calculate the surface temperature increase, similar to 
conventional GCM global warming simulations. The external anthropogenic emission used IPCC-
SRES A1B scenario. 
 
 
 70 
 
5.3 Carbon cycle-Climate feedback 
By year 2000, the UMD coupled run simulated 15 ppm higher CO
2
 compared 
to the observations, while the uncoupled CO
2
 is very close to the observation (Figure 
5.3a). The surface temperature in the coupled run has risen by about 0.6?C, 
comparable to the observed overall warming (Figure 5.3b). Since the multi-decadal 
surface temperature changes in the 20
th
 century are likely affected by factors not 
considered here such as solar variability, aerosol and non-CO
2
 greenhouse gases 
(Stott et al., 2000; Jones et al., 2003b), such level of agreement indicates a reasonably 
good representation of the past climate changes. During this historical period, our 
uncoupled run simulates a cumulative land sink of 28 PgC and an ocean sink of 178 
PgC, which is much stronger than the 100 PgC of the Hadley model (Cox et al., 2000). 
This strong ocean uptake partly compensates for our weak land sink so that CO
2
 level 
in 2000 is only slightly higher than the observed value.  
 
 
Figure 5.3 (a) Atmospheric CO
2
 (ppm) and (b) surface temperature change (?C) from 1860 to 
2100, simulated by the fully coupled carbon cycle-climate model (in red), with constant (pre-
industrial) climate (in blue), compared to observations (in black). The surface temperature curve 
labeled in uncoupled was obtained by an additional simulation where the CO
2
 from uncoupled run 
was used to force the physical climate model, similar to conventional GCM global warming 
simulations. 
 
 
 71 
 
 
By 2100, atmospheric CO
2
 reached a level of 658 ppm in the uncoupled run, 
but the coupled run produced a CO
2
 level of 748 ppm, 90 ppm higher. As a result, the 
coupled run surface temperature is 0.6?C higher by 2100 (Table 5.1). These results 
indicate a positive feedback from the interactive response of carbon cycle to climate 
change, because the only difference in the two runs is the climate forcing for the 
carbon cycle, thus supporting the results from the Hadley Centre (Cox et al., 2000) 
and IPSL (Friedlingstein et al., 2001; Dufresne et al., 2002). However, the magnitude 
of the change differs significantly among the three models (Table 5.1). UMD model 
predicts about 90 ppm additional CO
2
 due to climate impact on the carbon cycle, 
which implies a 0.6?C additional warming, while the Hadley model shows 250 ppm 
more CO
2
 and 1.5?C additional warming. The IPSL model results are more similar to 
ours in terms of these changes, but other aspects differ greatly (see below).  
 
Table 5.1 Differences of the atmospheric CO
2
 (ppm) and surface temperature (?C) changes from 
1860 to 2100 between the coupled run and the uncoupled runs from UMD, Hadley Centre and 
IPSL. 
 
?CO
2
 ?Ts 
UMD 
90 0.6 
Hadley 
250 1.5 
IPSL 
75 0.6 
 
Part of the differences are due to the different climate sensitivity to a given 
CO
2
 change. For instance, our coupled run has a warming of about 3?C at year 2100, 
while it is 5.5?C for the Hadley model and 3?C for the IPSL model. These numbers 
are compounded with the strength of the carbon cycle feedback. It should be noted 
that the IPCC-SRES-A2 scenario used by IPSL has about 300 PgC larger cumulative 
 
 
 72 
 
emissions than the IPCC-SRES-A1B scenario we used which is more similar to the 
IS92a used by Hadley. The difference in ocean carbon models are also partly 
responsible, especially given the simplicity of the box model we used. From 1860 to 
2100, ocean in our coupled run absorbs 867 PgC, about 49% of the cumulative 1780 
PgC emission, somewhat larger than the 700 PgC of IPSL model, and significantly 
larger than the 490 PgC of the Hadley model (Table 5.2). This is also in line with the 
weak land and strong ocean carbon sinks from 1860?2000. An elegant analysis by 
Friedlingstein et al. (2003) has shown that the major contribution comes from land in 
the Hadley and IPSL models. The main differences from the UMD model also appear 
to be on land, which is the focus of the rest of this chapter.  
Table 5.2 Change in the carbon pools by 2100 from 1860 in three coupled carbon-climate 
models:UMD, Hadley Centre, and IPSL. Units in PgC. 
Uncoupled Coupled Difference  
Vege Soil Land Ocean Vege Soil Land Ocean Vege Soil Land Ocean 
UMD 60 40 100 866 20 -40 -20 867 -40 -80 -120 1 
Hadley 220 410 630 370 60 -150 -90 490 -160 -560 -720 120 
IPSL 380 300 680 670 310 170 480 700 -70 -130 -200 30 
 
5.4 Uncertainties in land carbon response 
Figure 5.4 shows the evolution of the land carbon pool which consists of 
vegetation and soil carbon. In the uncoupled run, CO
2
 fertilization leads to an 
increase in the net primary production (NPP) and subsequent accumulation of carbon 
in vegetation biomass. This drives an increase in soil carbon as vegetation-to-soil 
turnover also increases. The total increase in land carbon from 1860 to 2100 is about 
100 PgC with 60 PgC from vegetation, 40 PgC from soil (Table 5.2).  
 
 
 73 
 
A remarkable reversal of soil carbon uptake is seen in the coupled run where 
the soil has released 40 PgC by 2100. Although vegetation carbon increased, its 
storage has saturated at 20 PgC in 2100. The smaller vegetation uptake compared to 
the uncoupled run is partly due to warming-enhanced maintenance cost (autotrophic 
respiration), and partly due to the change in precipitation pattern. This modest 
increase in vegetation carbon (therefore the turnover) is not enough to counteract the 
enhancement of respiration loss in soil at higher temperature. As a result, land has 
become a net CO
2
 source of 20 PgC to the atmosphere by the end of the simulation.  
 
 
 Figure 5.4 Cumulative vegetation and soil carbon changes in PgC since 1860 for the fully 
coupled run (red) and uncoupled run (blue) from the present model (UMD, upper panels), the 
Hadley Centre (middle panels), and IPSL (lower panels). 
 
 
 74 
 
Such a divergence in land carbon response with or without climate-carbon 
feedback (120 PgC difference in the two runs) is also seen in the Hadley Centre 
results, but their amplitude is 720 PgC, 6 times larger than ours (Table 5.2). The IPSL 
difference between the two runs (200 PgC) is more similar to ours, but they differ 
from ours and Hadley?s in that their vegetation growth enhancement significantly 
outcompetes the climate change impact on respiration rate, so that in the coupled run 
both soil and vegetation pools continue to increase over time. In coupled run both soil 
and vegetation pools continue to increase over time.  
The spatial pattern of the land carbon change (Figure 5.5) indicates that the 
UMD uncoupled run has an increase of carbon accumulation everywhere due to CO
2
 
fertilization effect. The coupled run shows intriguing spatial variations. At high 
latitude regions in Canada, Scandinavia and northern Siberia, the coupled run shows 
increase in carbon storage larger than the uncoupled run due to enhanced growth in 
these currently temperature-limited regions, which locally outcompetes the increased 
respiration loss at higher temperature. However, at middle and low latitudes, 
increased soil decomposition and autotrophic respiration at higher temperature 
dominate over the CO
2
 fertilization effect, leading to less carbon storage. This is 
somewhat complicated by the change in precipitation which tends to have high spatial 
variation. Increases in precipitation (not shown) are responsible for the carbon 
increase in regions such as parts of the northern Amazon and West Africa. Such a 
precipitation pattern is somewhat different from the Hadley Centre model where a 
perpetual El Ni?o-like state gave rise to a reduced rainfall and vegetation dieback in 
the Amazon (Betts et al., 2004). Because the UMD model does not have ocean 
 
 
 75 
 
dynamics, and climate models also differ widely in their regional responses, such 
regional comparisons should only be viewed with great caution.  
 
Figure 5.5 Spatial distribution of total land carbon (vegetation + soil) change for the UMD 
coupled and uncoupled runs. These are the differences between the last 30 years (2071?2100) and 
the first 30 years (1860?1889), showing different behavior at high latitude and mid-low latitude 
regions. Units in kgC/m
2
. 
 
5.5 Discussion and conclusion 
In order to understand the large differences among the three models, we 
conducted the following three sensitivity experiments with UMD (coupled and 
uncoupled runs for each individual experiment) with different model 
parameterizations from the standard run described above: 1. Stronger fertilization; 2. 
Single soil pool; 3. Higher soil decomposition (Table 5.3). The results from the 
coupled runs for these three experiments together with the standard run are shown in 
Figure 5.6. A stronger fertilization effect increases growth significantly so that land is 
still a sink of 150 PgC by 2100, unlike the 20 PgC source in the standard run. This 
source-to-sink reversal in the coupled UMD also makes the direction of the carbon 
changes the same as that of the coupled IPSL result. The coupled single soil pool 
 
 
 76 
 
experiment shows that land becomes a carbon source of 100 PgC, much larger than 
the standard run, but similar to the Hadley carbon release under global warming 
(Jones et al., 2003a). We were thus surprised not to see a similarly high level of 
sensitivity in the third experiment varying Q
10
 as we did with the previous two model 
modifications.  This is likely because the turnover time in these two pools, especially 
the slow soil pool, are comparable or longer than the 100 year time scale considered 
here for global warming, and thus their full potential of carbon release has not been 
realized by the year 2100. In contrast, the single soil pool experiment has a fast 
turnover time for all the soil carbon and thus a near-equilibrium response to the 
warming.  
 
Table 5.3 Three sensitivity experiments with different model parameterizations. 
Experiment Model parameterizations 
Stronger CO
2
 
fertilization 
 
Double the sensitivity of photosynthesis to CO
2
 in VEGAS. 
One soil pool 
Single soil pool by lumping the three soil carbon pools (fast, intermediate 
and slow) into one, with a turnover time of 25 years at 25?C, a value 
somewhat slower than the standard run?s fast soil pool, but much faster 
than the intermediate (80 years) and the slow soil pool (1000 years). 
 
Higher soil 
decomposition 
 
Higher soil decomposition rate dependence on temperature for the lower 
two soil layers (Q
10 
= 2.2). 
 
 
Table 5.4 Difference in total land carbon pool (PgC) between coupled and uncoupled runs at year 
2100 for various sensitivity simulations of UMD, Hadley Centre, and IPSL. 
UMD 
Standard 
UMD 
Strong 
Fertilization 
UMD 
One soil 
pool 
UMD 
All Q
10
=2.2 
 
Hadley 
 
IPSL 
 
-120 -80 -260 -130 -720 -200 
 
 
 77 
 
 
 
Figure 5.6 Total land carbon change (PgC) since 1860 in four fully coupled runs of UMD: the 
standard run (same run as in Figure 5.3), and three sensitivity experiments: strong CO
2
 
fertilization effect, single soil pool, and high soil decomposition rate dependence on temperature 
(Q
10
 = 2.2 for all soil layers, blue). 
 
Current observations have not converged on the strength of CO
2
 fertilization 
on a global scale (Field, 2001). While models often rely on strong CO
2
 fertilization to 
explain the ?missing carbon sink? of 1?2 PgC/yr
 
observed in the 1980?90s (Prentice 
et al., 2001), the weak dependence in the UMD model produces only a small land 
CO
2
 sink of 0.5 PgC/yr in the uncoupled run, 0.2 PgC/yr
 
in the coupled run (partly 
due to warming-induced soil carbon release). However, the UMD model, as well as 
those of Hadley and IPSL, do not consider effects such as land use change and fire 
suppression, which may be important contributors to be considered in future coupled 
modeling. In addition, the slower soil pools may not respond to global warming as 
fast, and their temperature dependence may not be as strong as the surface soil and 
litter. Our sensitivity experiments suggest a significant impact of these uncertainties.  
Although all models agree on a positive carbon cycle feedback as indicated by 
 
 
 78 
 
the negative values in Table 5.4, the predicted difference between coupled and 
uncoupled runs for our strong fertilization case is -80 PgC, compared to -120 PgC in 
the standard case. Both values are modest, but somewhat similar to IPSL (-200 PgC). 
The biggest difference in the modified coupled runs is the flux of -260 PgC for the 
one soil pool runs, more than double the flux from the standard case. This is a bit 
closer to, but still significantly smaller than the Hadley model (-720 PgC). These 
results suggest that different CO
2
 fertilization strengths may explain some part of the 
UMD-IPSL differences (and the UMD-Hadley difference in the uncoupled runs), and 
soil decomposition and turnover time may explain partly the UMD-Hadley 
differences in the coupled runs. Thus, in order to narrow down uncertainties in the 
prediction of future carbon and climate change in projects such as the Coupled 
Carbon Cycle Climate Model Intercomparison Project (C4MIP) (Fung et al., 2000), it 
we urgently need to improve our knowledge of some major issues in the carbon cycle, 
including the global strength of CO
2
 fertilization effect and the turnover and 
decomposition rates of the slower soil pools. 
 
 
 79 
 
 
Chapter 6: Enhanced Terrestrial Carbon uptake in the Northern 
High Latitudes under Climate Change in the 21
st
 Century from 
C4MIP Models Projections 
 
6.1 Introduction 
Carbon cycle is one of the most important biogeochemical cycles in the 
climate system because carbon dioxide is a principal greenhouse gas that contributes 
significantly to global warming (IPCC, 2007). The amount of carbon dioxide in the 
atmosphere is not only dependent on anthropogenic fossil fuel emission, but also on 
the exchange of fluxes of carbon between the atmosphere, land, and ocean. Carbon 
exchange between the atmosphere and ecosystem are in turn highly sensitive to 
climate change. Therefore, understanding the interaction between climate and carbon 
cycle is essential for the accurate projection of future climate change. 
The terrestrial ecosystem mainly gains carbon from the atmosphere through 
photosynthesis and loses it primarily through respiration (autotrophic and 
heterotrophic) (biomass burning is also an important non-biological contributor). 
During the past decades, Northern High Latitudes (NHL: poleward of 60?N) have 
witnessed dramatic changes where the annual average temperature increased by 1-2?C 
in northern Eurasia and northwestern North America (Arctic Climate Impact 
Assessment, 2005). This intense warming is much larger than the increase of global 
average surface temperature of 0.65?0.2?C over the 20
th
 century (IPCC, 2007). The 
snow/ice albedo reduction due to warming is traditionally believed to be one of the 
 
 
 80 
 
main amplifiers of greenhouse warming in the NHL (Holland and Bitz, 2003; Kaplan 
and New, 2006).  
This dramatic climate change affects the two dominant biomes of the NHL: 
the boreal forests and the tundra (Figure 6.1). Accompanying the accelerating climate 
change in the NHL, in the past decades, a ?greening? of vegetation has been observed 
in the boreal forests based on satellite data and phenology studies (Keeling et al., 
1996; Myneni et al., 1997; Zhou et al., 2001; Tucker et al., 2001; Lucht et al., 2002). 
Such a ?greening? has been associated with a 10 to 20-day lengthening of the 
growing season (Zhou et al., 2001; Tucker et al., 2001; Linderholm, 2006; Schwartz 
et al., 2006; Piao et al., 2006; Piao et al., 2007). The increase in photosynthetic 
activity due to the persistent increase in the length of the growing season may lead to 
long-term increases in carbon storage and changes in vegetation cover, which in turn 
affect the climate system. However, Piao et al. (2008) found that there is a net carbon 
dioxide loss of northern ecosystems in response to autumn warming. This is because 
the increase in respiration is greater than photosynthesis during autumn warming. On 
the other hand, frozen soils are prevalent in the tundra with a north-south gradient 
from continuous-to-discontinuous permafrost. With low ambient temperatures, 
waterlogged soils, and slow drainage, these ecosystems have been slowly 
accumulating a large amount of carbon (Sitch et al., 2007). The soil carbon becomes 
metabolized when the frozen soil begins to melt in response to rising temperatures. 
This leads to an increase of soil carbon released to the atmosphere. One estimate 
suggested that global warming could thaw 25% of the permafrost area by 2100, thus 
rendering about 100 PgC vulnerable to decay (Davidson et al., 2006). To date, such 
 
 
 81 
 
below-ground processes are poorly understood (Melillo et al., 2002; Bond-Lamberty 
et al., 2004; Eliasson et al., 2005; Lawrence and Slater, 2005; Bronson et al., 2008). 
The competition between absorption of carbon due to the greening of the boreal 
forests and the release of carbon from the soil due to the rising temperatures in the 
terrestrial biosphere in the NHL therefore will determine whether the NHL becomes a 
source or sink of carbon in the future. 
 
Figure 6.1 Boreal vegetation distribution in the northern high latitudes (modified from 
Montaigne, 2002). The area poleward of 60?N (circled in dark brown) is defined as Northern High 
Latitudes (NHL) in this study. 
 
Several studies have used offline simulations of vegetation-carbon models 
forced by the IPCC climate projection to investigate the future effects of CO
2
 
fertilization, climate change, land-use change, and nutrient limitation on the NHL 
ecosystem (McGuire et al., 2000; Cramer et al., 2001; Schaphoff et al., 2006; Sitch et 
 
 
 82 
 
al., 2007). However, few studies have considered the carbon cycle-climate coupling 
in the NHL (McGuire et al., 2006). Fully coupled three-dimensional carbon cycle-
climate models have been used to study the interaction between the global carbon 
cycle and climate (Cox et al., 2000; Friedlingstein et al., 2001; Joos et al., 2001; Zeng 
et al., 2004; Thompson et al., 2004; Matthews et al., 2005). Albeit uncertain in 
magnitude, most studies indicate a positive feedback for the carbon cycle. 
Quantifying and predicting this carbon cycle-climate feedback is difficult because of 
the limited understanding of the processes by which carbon and the associated 
nutrients are transformed or recycled within ecosystems, and then exchanged with the 
overlying atmosphere (Heimann et al., 2008). Large uncertainties from various model 
parameterizations and modeling protocols have highlighted an urgent need of model 
intercomparsions to understand the interaction between climate and carbon cycle.  
In the Coupled Carbon Cycle Climate Model Intercomparison Project 
(C4MIP), 11 climate modeling groups have used their fully-coupled carbon cycle-
climate models to investigate such interactions, with particular emphasis on the 
feedback processes involved (Friedlingstein et al., 2006). The modeling protocols 
used by all the modeling groups were: 1) The same anthropogenic emission of CO
2
 
Intergovernmental Panel on Climate Change (IPCC) Special Report on Emissions 
Scenarios (SRES) A2 scenario as the external forcing for each fully coupled carbon 
cycle-climate model; 2) Two separate simulations: a fully coupled interactive carbon 
cycle-climate run and a uncoupled run wherein a prescribed climate was used 
(radiative CO
2
 concentration was kept at pre-industrial level), while the vegetation 
and carbon model can see the increasing CO
2
 for photosynthesis. Carbon cycle-
 
 
 83 
 
climate feedback is then calculated as the difference between the two simulations. 
Friedlingstein et al. (2006) indicate a positive carbon cycle-climate feedback from all 
C4MIP models in the 21
st
 century, on a global scale. Due to this positive feedback, 
there is an additional 20-200 ppm CO
2
 in the atmosphere by 2100, leading to a 0.1-
1.5?C warming in the C4MIP models. A majority of the models indicated a robust 
reduction of terrestrial carbon uptake in the tropics due to such climate feedback 
(Figure 6.1 in Friedlingstein et al., 2006).  
Climate change during the past decades has been changing the physical, 
biological, and societal conditions in the NHL. When the interaction between the 
climate and carbon cycle is included in the climate system, the following questions 
are still open: 
 z Will the NHL terrestrial biosphere become a carbon sink or a source in 
the future? In particular how will the vegetation and soil carbon change in 
the NHL?  
 z How do the mechanisms involved in the carbon exchange between the 
vegetation and the soil affect the individual growth rate of their carbon 
storage by 2100? In particular what are the effects of CO
2
 fertilization and 
climate change (intense warming in NHL) on vegetation photosynthesis 
and soil decomposition?  
 z Is the future projection of terrestrial carbon change by fully coupled 
carbon cycle-climate simulation, consistent with the current observed 
?greening? and a northward shift of vegetation distributions over the past 
decades?  
 
 
 84 
 
In contrast to previous modeling studies in which projection of future climate 
solely drives vegetation and carbon models, in this study, we investigated the 
interaction between climate and carbon cycle in the NHL based on the simulations of 
fully coupled carbon cycle-climate C4MIP models, with particular emphasis on 
addressing the above scientific questions. Current NHL ?greening? and a northward 
shift of vegetation distributions have been reported by satellites and land based 
observations during the past decades. The change in the NHL ecosystem is thus of 
importance and the interaction between carbon cycle and climate on the future 
climate system is examined here. If this C4MIP modeling study can provide an 
insight to assessment of the carbon uptake under global warming, it will be 
encouraging and of great significance to apply fully coupled models in the projection 
of future climate in the NHL. 
6.2 Data and methodology 
The principal focus of our study is to investigate the terrestrial carbon storage 
change in the high latitude regions poleward of 60?N in coupled carbon cycle-climate 
systems. The data used in this study is based on the multi-model simulations provided 
by C4MIP (http://c4mip.lsce.ipsl.fr/), which is a joint project between the 
International Geosphere-Biosphere Program (IGBP) and the World Climate Research 
Program (WCRP). The use of dynamic carbon models in examining future climate 
has also been highlighted in the 4th Assessment Report of IPCC (IPCC, 2007). The 
terrestrial carbon model used in the 11 coupled carbon-climate models can be found 
in Table 6.1 from Friedlingstein et al. (2006), as well as for a detailed description of 
the models please refer to Friedlingstein et al. (2006). Of those models, HadCM3LC 
 
 
 85 
 
and UVic use the same land carbon component of Met Office Surface Exchange 
Scheme (MOSES)/ Top-Down Representation of Interactive Foliage and Flora 
including Dynamics (TRIFFID), while CLIMBER2-LJP and BERN-CC use the same 
Lund?Potsdam?Jena (LPJ) vegetation model. These common components provide a 
good opportunity to examine cases of carbon-climate coupling using the same carbon 
model but different OAGCMs. All the C4MIP models used the same anthropogenic 
fossil fuel emissions from Marland et al. (2005) for the pre-industrial period and the 
IPCC SRES A2 for the 2000-2100 period. The coupled simulation in the C4MIP 
models, has the climate and the carbon components iteracting freely thus representing 
a full carbon cycle-climate coupling. The uncoupled simulation treats CO
2
 as a non-
radiatively active gas (so the carbon cycle experiences no CO
2
-induced climate 
change), while the ecosystem still sees an increasing CO
2
. The difference between 
coupled and uncoupled runs is defined as carbon cycle-climate feedback.  
Table 6.1 Major characteristics of the C4MIP models components (adapted from Friedlingtein et 
al., 2006) 
Model 
 
Atmosphere Ocean Land Carbon DGVM*
? 
Ocean Carbon 
HadCM3LC 
 
HADCM3 (2.5? x 3.75?) 
 
(2.5? x 3.75?), flux adjustment MOSES/TRIFFID 
 
Yes 
 
HadOCC 
 
IPSL-CM2C 
 
LMD5  (5? x 4?) 
 
OPA-7, (2? x 2?),  no flux adjustment SLAVE 
 
No 
 
NPZD 
 
IPSL-CM4-
LOOP 
 
LMDZ-4 (3? x 3?) 
 
ORCA2, (2? x 2?),  no flux adjustment ORCHIDEE 
 
Not here 
 
PISCES 
 
CSM-1 
 
CCM3 T31 
 
NCOM (3.6? x  (0.8 -1.8?)) 
 
LSM, CASA 
 
No 
 
OCMIP-biotic 
 
MPI 
 
ECHAM5, T63 
 
MPI-OM, 1.5?,  
no flux adjustment 
JSBACH 
 
No 
 
HAMOCC5 
 
LLNL 
 
CCM3, (2.8? x 2.8?) 
 
POP (0.6? x 0.6?) IBIS,flux 
adjustment 
 
Yes 
 
OCMIP 
 
FRCGC 
 
CCSR/NIES/FRCGC 
T42(2.8? x 2.8?) 
COCO No flux adjustment, ((0.5?-1.4?) 
x 1.4?) 
Sim-CYCLE No NPZD 
 
UMD 
 
QTCM (5.6? x 3.7?) 
 
Slab mixed layer, 5.6?x 3.7? 
 
VEGAS 
 
Yes 
 
Three-box mode 
 
UVic-2.7 
 
EMBM (1.8? x 3.6?) 
 
Mom 2.2, 1.8? x 3.6?,  
no flux adjustment 
MOSES/TRIFFID 
 
Yes 
 
OCMIP Abiotic 
 
CLIMBER2
-LPJ 
 
2.5-D, (10? x 51?) 
statistical-dynamical 
Zonally averaged, 2.5?lat, 3 basins 
 
LPJ 
 
Yes 
 
NPZD 
 
BERN-CC 
 
EBM (2.5? x 3.75?) 
 
HILDA box-diffusion model 
 
LPJ 
 
Yes 
 
Perturbation  approach 
* DVGM: Dynamic Global Vegetation Model 
 
 
 86 
 
Currently, each modeling group provided a standard time series of total fossil 
fuel emissions, simulated global atmospheric CO
2
 concentration, surface temperature, 
terrestrial Net Ecosystem Product (NEP), ocean net CO
2
 flux, Net Primary Product 
(NPP), Heterotrophic respiration (R
h
), and vegetation and soil carbon pools storages 
for the total globe and band-averaged regions: 90?S-30?S, 30?S-0?S, 0?N-30?N, 30?N-
60?N, 60?N-90?N. Our study focuses on the terrestrial regions poleward of 60?N. In 
this study, the total global values are used only as a reference for the NHL 
comparisons. The results will be presented as the individual response of the 11 
models or as that of the ensemble mean in the text. 
6.3 NHL terrestrial carbon storage in the 21
st
 century 
The NHL terrestrial ecosystem stores a large amount of carbon in boreal forest 
and frozen soil. It is very important to understand the changes in the uptake of carbon 
under global change. Prior to investigating the 21
st
 century change of the terrestrial 
carbon storage in the NHL in the C4MIP models, the NHL vegetation and below 
ground soil carbon in the pre-industrial era in the individual models was examined 
(Figure 6.2). Figure 6.2a shows the global partitioning of carbon into land and soil. 
Globally, the C4MIP models simulate a mean total land carbon storage of 2106 PgC, 
with a range from 1348 PgC to 3089 PgC. The vegetation carbon pool contains 279-
937 PgC and the soil contains 999-2152 PgC. Thus, the C4MIP models indicate that 
the carbon in the soil is about 2.5 times that of vegetation carbon, highlighting the soil 
as a major carbon reservoir of land carbon. Unlike this general agreement on the 
global total of land carbon in C4MIP models, there is a large uncertainty in the carbon 
stored in the NHL (Figure 6.2b). For example, the mean indicates that the NHL 
 
 
 87 
 
contains about 15% of the total global terrestrial carbon. IPSL-CM4-LOOP, though, 
has just 65 PgC in NHL land ecosystem, only about 4% of global total, while 
CLIMER-LPJ has 568 PgC in NHL, about 21% of global total. UMD has the largest 
percentage, at 30%. While the NHL contains 15% of total global land carbon, the 
NHL vegetation holds only 5.8% of global total vegetation carbon. Most of the land 
carbon in NHL is sealed below the ground in frozen soil. The mean NHL soil carbon 
is about 9.4 times that of the NHL vegetation carbon. Most of the carbon being 
locked in the soil is due to the low ambient temperature below freezing and is 
consistent with previous estimates (Post et al., 1982; Chapin and Matthews, 1993, 
McGuire and Hobbie, 1997). The mean NHL soil holds about 18% of global total 
carbon, which is lower than the estimate by McGuire and Hobbie (1997). IPSL-
CM2C, IPSL-CM4-LOOP, CSM-1 and MPI even have fractions less than 10%. UMD 
has a percentage of 35%, close to the estimate by McGuire and Hobbie (1997). 
Examining the total global partitioning of carbon in pre-industrial era between above 
and below ground carbon, the models that share the same terrestrial carbon models 
show similarities (between CLIMBER2-LPJ and BERN-CC, and between 
HadCM3LC and UVic). However, the partitioning in the NHL in between those two 
carbon models was more different. This reflects the differences in the two vegetation 
models to differing climatic conditions. 
In the past two decades, winter in the Northern High Latitudes has 
experienced some of the most rapid changes on the Earth, warming almost 2.5 times 
faster than the global average rate (Liu et al., 2007). Amplification of greenhouse 
warming in the NHL have been traditionally attributed to snow/ice-albedo reduction, 
 
 
 88 
 
dynamical feedbacks associated with the poleward heat transport of the oceans and 
atmosphere (Holland and Bitz, 2003; Alexeev et al., 2005), and internal feedbacks 
associated with polar processes (Overland et al., 2004). 
Global total
0
1000
2000
3000
4000
BE
RN
-
C
C
CL
I
M
BE
R2
-
L
P
J
CCS
M
-
1
Ha
d
C
M
3
L
C
I
SPL
-
C
M
2
C
IP
S
L
-C
M
4
-L
O
O
P
MP
I
LLN
L
F
RCG
C
UVi
c
UM
D
M
ean
(a)
Pg
C
Vege C
Soil C
 
NHL  total
0
200
400
600
BE
R
N
-
C
C
CL
I
M
BE
R
2
-
L
P
J
CCS
M
-
1
Had
C
M
3
L
C
IS
P
L
-C
M
2
C
IP
S
L
-C
M
4
-L
O
O
P
MP
I
LLN
L
FR
C
G
C
UVi
c
UM
D
M
ean
(b)
PgC
Vege C
Soil C
 
 
Figure 6.2 Simulated carbon storage for the pre-industrial era in the vegetation and soil by the 
C4MIP models and model-mean for: (a) Global total; (b) NHL. The stored carbon is calculated 
based on 1860-1870 for all models except 1870-1880 for LLNL, and 1901-1910 for CLIMBER-
LPJ. Soil carbon is indicated in brown, and vegetation in dark green. Most of the carbon is locked 
in the soil particularly in the NHL. Units of storage are in PgC.  
 
 
 
 89 
 
The ocean-atmosphere GCMs indicate that the earth will have warmed by 2?C 
relative to pre-industrial temperatures between 2026 and 2060, at which stage the 
mean annual temperature over the Arctic (60?N?90?N) will have increased by 
between 3.2?C and 6.6?C (Kaplan and New, 2006). Consistent with these predictions, 
the C4MIP models project that by 2100 the mean surface temperature in the NHL will 
increase by 5.9?C, in contrast to 3.2?C for global average (Figure 6.3). Most models 
show intense warming in NHL, nearly twice the global warming by 2100. The 
exception is UMD, which does not include a snow/ice scheme at all. LLNL, MPI, 
FRCGC and IPSL-CM4-LOOP have up to 7-8?C warming in the NHL.  
0
3
6
9
BE
R
N
-
C
C
CL
I
M
BE
R2
-
LP
J
CCS
M
-
1
Ha
d
C
M
3
L
C
I
SPL
-
C
M2
C
I
PSL
-
C
M4
-
LO
O
P
MPI LLN
L
F
RCG
C
UV
i
c
UM
D
Me
a
n
Tem
p
erature
 incre
ase (
 
?
C)
Global
NHL
 
 
Figure 6.3 Temperature increase from 1860 to 2100 projected by the fully coupled simulations 
from C4MIP models for global mean (blue diamond) and NHL (red circle) except for LLNL 
which is from 1870 to 2100, and CLIMBER-LPJ from 1901 to 2100. The multi-model mean is 
plotted. It is evident that significantly more intense warming is expected in the NHL, compared to 
global mean. The UMD model is an exception since it does not include a snow/ice albedo scheme. 
 
 Though, the ?greening? and vegetation migration in NHL during the past 
decades have been highly correlated to warming (Keeling et al., 1996; Myneni et al., 
1997; Zhou et al., 2001; Tucker et al., 2001), the intense warming also accelerates the 
decay of the organic matter in the soil which will subsequently release CO
2
 to the 
 
 
 90 
 
atmosphere, especially in the NHL, where most of the land carbon is stored in the 
permafrost soil levels. The increase in vegetation growth due to CO
2
 fertilization and 
warming counteracts the soil decomposition by warming. This difference will 
determine the change of carbon stored in the terrestrial biosphere in the NHL. Piao et 
al. (2008) found that there is a trend of net carbon dioxide loss of northern ecosystems 
in response to autumn warming because of the greater increase in respiration. While 
such a study by itself does not provide information about how these two highly 
uncertain competing effects could behave under different future scenarios, it 
highlights the importance of the effect of temperature on the soil respiration in NHL 
under global warming. 
The C4MIP coupled simulations indicate that by 2100 the NHL is a carbon 
sink with a mean of 0.4PgC/yr (Figure 6.4a, c). With the exception of FRCGC, the 
remaining 10 models agree that the NHL ecosystem will absorb carbon from the 
atmosphere. The range of absorption varies from 0.1 PgC/yr to 1.6 PgC/yr. The inter-
model deviation also increases with the time (Figure 6.4c). The model-mean (Figure 
6.4c) indicates a tendency of saturation or even a decrease in the carbon uptake by 
2060. This saturation tendency is visible in the HadCM3LC, FRCGC and BERN-CC 
models. The integration of this land-atmosphere carbon exchange indicates the 
amount of carbon land will absorb. The mean uptake of terrestrial carbon by 2100 is 
40 PgC while the range is from 17 to 82 PgC in the C4MIP models (Figure 6.4b, d). 
LLNL is an exception with a strong carbon source during the first century which 
switches to a sink in the 1970?s and ends as the largest carbon sink with 1.6 PgC/yr 
by 2100. A possible reason could be a global mean surface temperature drift of -0.15 
 
 
 91 
 
K and a 15 % increase in ice area in the 1900?s, which causes a regional decrease in 
ecosystem productivity in LLNL (Bala et al., 2006). The differences in trajectories 
between the models that share the same vegetation models are as large as the 
remaining inter-model deviations themselves. So, no similarities are identifiable on a 
regionally averaged scale.  
 
 
Figure 6.4 Changes of NEP and total land carbon in the NHL during 1860-2100 in C4MIP 
coupled simulations. (a) NEP; (b) Total land carbon change. Colored lines in the (a) (b) are for 
individual C4MIP models; (c) (d) the multi-model mean (red solid line) and the 1-?  spread (blue 
shading) are indicated. The change in total land carbon is indicated relative to 1860, except for 
LLNL and CLIMBER-LPJ where it is relative to 1870 and 1901 respectively. A 6-year running 
mean filter is applied to all the curves. Unit for NEP is PgC/yr and for land carbon change is PgC. 
 
The changes in the vegetation and the soil carbon are indicated in Figure 6.5a 
and 6.5b. The mean of the C4MIP models will gain18 PgC in above ground 
vegetation and 22 PgC in soil by 2100 compared to the late 18
th
 century. The 
increment of vegetation carbon stored is a little smaller than the increase of soil 
carbon pool. However, this is a significant increase in the vegetation carbon. The 
 
 
 92 
 
NHL vegetation contains about 10% of NHL land carbon, but this carbon storage 
increases from 38 PgC in the 1860s to 56 PgC by 2100, about 47%. The soil carbon 
increases only 8%, from 276 PgC to 298 PgC from 1860-2100. Contrary to the 
estimates of Davidson et al. (2006), our analysis of the C4MIP models indicates that 
at least in the 21
st
 century, the soil in the NHL will not lose carbon to the atmosphere. 
Soil carbon, protected from release to the atmosphere by cold and waterlogged 
conditions in NHL, was thought to be highly susceptible to changes in temperature 
and permafrost thawing in the 21
st
 century. This dilemma calls for a quantitative 
analysis of the processes involved that suppresses the effect of the intense warming 
on soil decomposition in the NHL.  
 
 
Figure 6.5 Changes of vegetation and soil carbon in the NHL in C4MIP couple simulations: (a, c) 
Vegetation carbon; (b, d) Soil carbon. Colored lines in the (a) and (b) are for individual C4MIP 
models; (c) and (d) show the multi-model mean (red solid line) and the 1-?  spread (blue 
shading) are indicated. All changes are relative to the year of 1860 except for LLNL with 1870, 
and CLIMBER-LPJ with 1901. A 6-year running mean filter is applied to all the curves. Units are 
in PgC. 
 
 
 93 
 
To explore this, the carbon budget of the above-ground vegetation and soil are 
examined separately. Though NEP reflects changes of the ecosystem carbon 
exchange, NPP has been used as an indicator of the vegetation activity; these 
variables do not directly measure the change of vegetation and soil carbon storage. 
Examination of the mechanisms controlling the vegetation and soil carbon exchange 
processes involves the following variables: (1) Vegetation (biomass) turnover into 
soil; (2) Growth rate of vegetation carbon; (3) Growth rate of soil carbon. To date, 
little attention has been paid to vegetation turnover; however, it is an important bridge 
for vegetation and soil carbon exchange. The difference between NPP and vegetation 
turnover is defined as the net growth rate of vegetation carbon. The difference 
between the vegetation turnover and soil decomposition indicates the net growth of 
soil carbon.  
Most vegetation carbon models constrain the total global NPP within a 
contain range. However C4MIP models differ a lot in the magnitude of the NHL NPP. 
It varies from 1.1 PgC/yr in IPCC-CM4-LOOP to 6.8 PgC/yr in UMD in 1860, with 
the percentage dependence on the global total ranging from 2% to 13% (not shown). 
In the NHL, the mean of C4MIP models indicates that NPP increases from 3.3 PgC/yr 
to 6.3 PgC/yr by 2100, nearly 90% more of vegetation activity (Figure 6.6a). IPSL-
CM4-LOOP even simulates more than a 400% increase in the NHL NPP (not shown). 
This robust vegetation increase also reflects the importance of albedo feedback. 
About 2.7 PgC/yr of biomass will turn over into the soil in the NHL (Figure 6.6b). 
Higher biomass turnover is normally expected following more vegetation 
productivity. The difference of about 0.3 PgC/yr remains in the vegetation and is 
 
 
 94 
 
reflected as the net growth rate of vegetation carbon (Figure 6.6c). This is consistent 
with the sink of 18 PgC that will accumulate in the vegetation by 2100 as discussed 
before (Figure 6.5c). The assumption that the biomass turnover is mostly proportional 
to NPP is reasonable because it is evident that more productivity can produce more 
biomass turnover over long-time periods in observations.  
 
Figure 6.6 The multi-model mean and 1-? spread of carbon fluxes in vegetation in the NHL in the 
C4MIP coupled simulations for: (a) Net Primary Product; (b) Vegetation turnover; (c) Vegetation 
carbon growth rate. The changes are relative to the values in the late 18
th
 century; a 6-year 
running mean filter is applied to all the curves. Units are PgC/yr for all three panels. 
 
The growth of soil carbon content is determined as a difference between 
vegetation turnover and soil decomposition. The mean of C4MIP models for these 
two factors is shown in Figure 6.7. Below the ground, the NHL soil decomposition is 
 
 
 95 
 
greatly enhanced from 3.3 PgC/yr to 5.9PgC/yr by the intense warming (Figure 6.7a); 
however, soil decomposition is smaller than biomass turnover, and the residual is 
locked in the soil as the growth rate of soil carbon (Figure 6.7b). The C4MIP mean-
model shows that the growth rate of soil carbon increases up to 2060 after which it 
decreases until 2100. A possible mechanism is that the soil decomposition increases 
dramatically after 2060, and quickly catches up with biomass turnover input by 2100. 
As the warming intensifies, there is a possibility that the NHL soil may lose carbon 
after 2100 from C4MIP, rather than in the 21
st
 century in the previous estimate. Our 
study based on C4MIP model results indicates that the vegetation biomass turnover is 
of utmost significance in understanding the processes that control the changes of 
vegetation and soil carbon in the NHL. 
 
Figure 6.7 Same as Figure 6.6 but for the changes of carbon fluxes in soil for: (a) Heterotrophic 
Respiration; (b) Soil carbon growth rate. Units are PgC/yr. 
 
 
 
 
 96 
 
6.4 CO
2
 fertilization and climate change effects in NHL carbon 
cycle-climate coupling 
 
The terrestrial carbon sink in the 21
st
 century in the NHL, identified in the 
previous section using C4MIP coupled simulations, is attributable to CO
2
 fertilization 
and climate change. These two factors play an important role in controlling the 
carbon stored in the NHL terrestrial system (McGuire et al., 2000; Cramer et al., 2001; 
Schaphoff et al., 2006; Sitch et al., 2007) and will be the focus of this section. In 
addition to the coupled simulation, C4MIP provides uncoupled simulation, in which 
the models do not see the effect of change in temperature due to increase of CO
2
 
while allowing for CO
2
 fertilization with increasing CO
2
. The difference between the 
coupled and uncoupled runs is defined as the carbon cycle-climate feedback 
(Friedlingstein et al., 2006).  
This carbon cycle-climate feedback discussed by Friedlinstein et al. (2006) 
includes mostly the effect of climate change and a small additional effect because of 
CO
2
 fertilization, due to CO
2
 concentration from the coupled run being considerably 
higher than that of the uncoupled run by 2100
1
. The combination of these two effects 
indicates the carbon cycle-climate coupling. The changes of vegetation and soil 
carbon in the NHL, as well as that for the whole globe are indicated in Figure 6.8. 
Contrary to the negative climate feedback on the global scale in which the land 
reduces its capacity for the uptake of carbon (Figure 6.8a), most models simulate that 
land carbon uptake in the NHL will increase slightly in the future due to carbon 
cycle-climate coupling (Figure 6.8b). All models show a robust increase in the 
                                                 
1
 
1
In C4MIP ?uncoupled? runs, the climate is kept constant while the CO
2
 level for vegetation photosynthesis is 
still increasing. This enables use to easily calculate the terrestrial carbon sensitivity to CO
2
 fertilization without 
contamination from the climate change. 
 
 
 97 
 
vegetation carbon pool size (Figure 6.8d) in the NHL.  
Contrary to the robust soil carbon loss due to carbon cycle-climate feedback 
on the global scale in all C4MIP models (Figure 6.8e) because of enhanced soil 
decomposition, the NHL soil carbon response to climate feedback varied substantially 
in the models (Figure 6.8f). However, the robust increase of NHL vegetation carbon 
largely dominates the terrestrial carbon change. Some models simulate a reduction in 
soil carbon in the NHL, but the robust carbon increase from vegetation compensates 
for the carbon loss from soil in most models (apart from CLIMBER, which has a tiny 
net carbon loss, and FRCGC which has a large net carbon loss), resulting in a general 
net increase in the terrestrial pool in NHL. Further analysis indicated that this small 
carbon increase between coupled and uncoupled cases is caused by the higher 
increase of vegetation growth than soil decomposition due to carbon-climate feedback 
(not shown). Our analysis indicates that the global and the northern high latitude 
regions have quite different underlying mechanisms that control the terrestrial carbon 
storage changes. An interesting point that is elucidated from the coupled climate 
feedback analysis is that soil carbon plays a key role in the global carbon pool (Figure 
6.8e), while robust vegetation growth is the dominant factor in the high latitudes 
(Figure 6.8d) in the 21
st
 century. Though the magnitude of the NHL terrestrial carbon 
and its change are much smaller than that of the global scale, this is of great 
significance because potentially dramatic impacts of climate changes on the NHL 
terrestrial ecosystem.  
Besides CO
2
 fertilization, the NHL carbon cycle-climate coupling has a 
potential to stimulate the vegetation activity in the future due to the feedback of 
 
 
 98 
 
warming and higher CO
2
 concentrations (Figure 6.9b, d). This is in contrast to a 
negative effect of carbon cycle-climate feedback on vegetation activities on a global 
scale reported by Friedlingstein et al. 2006 (Figure 6.9a, b).  
 
Figure 6.8 Changes in carbon storage during the 1860-2100 period due to carbon cycle-climate 
feedback (coupled?uncoupled simulation) for the 11 C4MIP members for the globe (a, c, e) and 
the NHL (b, d, f) for: (a, b) Total land carbon. (c, d) Vegetation carbon; (e, f) Soil carbon; All 
values are relative to the year of 1860 except for LLNL and CLIMBER-LPJ, which have the 
beginning years of 1860 and 1901 respectively. A 6-year running mean filter is applied to all the 
curves. 
 
 
 99 
 
 
Figure 6.9 The sensitivity of NPP to CO
2
 fertilization and temperature change for the whole 
globe (a, c) and the NHL (b, d) during 1860-2100 for the 11 C4MIP members. The NPP 
sensitivity to CO
2
 fertilization (a, b) is calculated directly from the C4MIP uncoupled simulations, 
in which the constant climate is prescribed while CO
2
 increases for photosynthesis. Then NPP 
sensitivity to CO
2
 fertilization at 2 x CO
2
 is used in the coupled simulations to isolate the NPP 
sensitivity to temperature increase. The NPP sensitivity to temperature (b, d) thus is not the direct 
NPP from C4MIP coupled simulation. It has been corrected using the sensitivity of NPP to CO
2
 
fertilization based on uncoupled simulation. For details of this calculation, refer to Friedlingstein 
et al. (2006). It indicates that NPP sensitivity to temperature change in the NHL is different from 
the one of global scale. The panels on the left (a, c) are the same as Figure 3a, b in Friedlingstein 
et al. (2006), and are used here only for comparison to the NHL result. 
 
 
 
 100 
 
6.5 Discussion and conclusion 
The terrestrial ecosystem in NHL covers 13% of the total global land area, 
mainly with boreal forests and the tundra. The boreal forest in Northern Hemisphere 
makes up one-third of the global woodland (Montaigne, 2002). With low ambient 
temperatures, waterlogged soils, and slow drainage, the NHL ecosystem has been 
slowly accumulating large amounts of carbon in the soil in the past. The Arctic tundra 
contains 30-40% of terrestrial soil carbon in the permafrost and peatland at an annual 
mean surface temperature of -10?C (Schimel, 1995, Melillo et al., 1995; McGuire and 
Hobbie, 1997, New et al., 1999). The climate change in the past decades in the NHL 
has changed the spatio-temporal patterns of terrestrial ecosystem. In the next century, 
coniferous forests are projected to expand northward and the area of the present 
tundra coverage could be halved (e.g., see Arctic Climate Impact Assessment, 2005). 
Changes and potential future changes in the carbon storage and uptake in the high 
latitudes are important and this work underscores the importance of understanding the 
underlying mechanisms.  
For this purpose, a suite of 11 fully coupled carbon climate simulations from 
C4MIP were studied to investigate the NHL carbon uptake through 2100. Unlike 
many modeling studies using climate projections to force carbon cycle models offline 
(e.g., McGuire et al., 2000; Cramer et al., 2001), the C4MIP coupled simulations 
consider the interactions between the carbon cycle and the climate. In the pre-
industrial period, the mean of the C4MIP models indicates that the NHL soil 
contained about 18% of global total soil carbon, while the vegetation contained only 
5.8% of its global vegetation carbon.  
 
 
 101 
 
By the end of the 21st century, the mean of the 11 models suggests a 5.9?C 
warming in the NHL, in contrast to a 3.2?C increase for the global mean temperature. 
The NHL terrestrial carbon changes under the intense high latitude warming and 
increasing CO
2
 in the 21st century can be summarized in a schematic (Fig. 6.10). Our 
analysis based on C4MIP simulations suggest that the NHL will be a carbon sink with 
a mean of 0.4 PgC yr-1 through 2100, with an increase of about 40 PgC in land 
carbon in the C4MIP coupled simulations. Of this, the vegetation carbon increases by 
about 18 PgC, and the soil carbon increases by about 22 PgC. Though the magnitude 
of the increment in vegetation carbon is a little smaller, it is robust with a 47% 
increase, and is of significance to the NHL land carbon budget in the future. In 
contrast, the increase in soil carbon amounts to only an 8% rise over pre-industrial 
levels .  
Our analysis of the C4MIP models confirms the enhanced vegetation growth 
ongoing in recent decades as well as projected under future climate change, as 
described by many previous studies. However, it also reveals a rarely highlighted 
mechanism, namely, the increased vegetation turnover input to soil carbon that drives 
an increase in the soil carbon pool. This runs counter to the better-known prediction 
of loss of soil carbon in response to warming. In contrast to previous estimates that 
the future NHL soil will lose carbon due to warming (e.g., Davidson et al., 2006), our 
analysis of the C4MIP models indicate that the NHL soil can still continue to be a 
carbon sink through 2100. We estimated that the NPP will increase by about 90% by 
2100 in NHL. The warming in the NHL region projected from C4MIP models does 
enhance soil decomposition and drain soil carbon; however, high turnover from the 
 
 
 102 
 
high productive vegetation provides input to the soil and cancels this soil 
decomposition in the early 21st century. Davidson et al. (2006) discussed that 
terrestrial carbon models mostly use Q
10
 or Arrhenius equation for their soil 
decomposition dependence on temperature. In both parameterizations, decomposition 
rates have an exponential dependence on temperature, and this is probably why 
heterotrophic respiration is projected to accelerate at higher temperature. Nevertheless, 
after 2060, the growth rate of NHL soil carbon begins to decrease because the soil 
decomposition accelerates at higher temperatures and catches up with the biomass 
turnover input. Thus, there is a possibility that the NHL soil may lose carbon by 
intense warming after 2100, rather than in 21st century. The vegetation biomass 
turnover is of great importance in linking vegetation and soil carbon change. 
Further examination of the effects of CO
2
 fertilization and climate change in 
the carbon climate feedback context of C4MIP suggests that besides CO
2
 fertilization, 
the warming in NHL tends to stimulate vegetation productivity to an even greater 
degree. In contrast to the robust global land carbon loss due to climate-carbon 
feedback, our analysis suggests that the intense warming in the NHL may help to 
enhance the terrestrial carbon sink there. The loss of global carbon is dominated by 
the soil decomposition in the tropics, while the gain in the NHL is dominated by the 
vegetation growth. This is consistent with current ?greening? and a northward shift of 
vegetation distributions in the NHL. Such insight into biomass exchanges between the 
vegetation and soil in the NHL highlights the critical role of carbon cycle-climate 
interaction under global warming. 
 
 
 
 103 
 
 
 
Figure 6.10 The conceptual diagram to show effects CO
2
 fertilization and intense warming in 
NHL on the changes of carbon fluxes and storages by 2100 from C4MIP coupled simulation. The 
multi-model mean and one standard deviation are provided to indicate the relative magnitudes of 
these changes. All changes are relative to 1860. Unit for carbon flux is PgC/yr and PgC for carbon 
storage. 
 
A major caveat to the C4MIP simulations, and therefore to our conclusion 
predicting an increasing NHL soil carbon pool, is that permafrost and peatland carbon 
pools are not explicitly represented in these models. When the top soil layer of 
permafrost thaws, much of the large substrate pool is likely to decompose quickly, 
and large amounts of otherwise mostly unprotected carbon become available for 
decomposition (Goulden et al., 1998; Serreze et al., 2000). As the NHL warms 
significantly, the warming-induced deepening of the layer of seasonal biological 
activity may cause a net loss of deep soil carbon in the boreal forest and tundra. We 
 
 
 104 
 
are unfortunately limited to the currently available C4MIP model results, and such 
possibilities can only be fully addressed in future models/projects that better 
represent/report these processes.  
Another effect that may be of major importance is the change in the 
hydrological cycle, as precipitation is expected to increase and soil hydrology 
changes dramatically as permafrost thaws. Angert et al. (2005) studied the NDVI and 
climate data and reported that drier summers in mid and high latitudes can cancel out 
the CO
2
 uptake enhancement induced by warmer springs, demonstrating that a 
warming climate does not necessarily lead to higher CO
2
 growing-season uptake. 
Unfortunately, this study could not investigate the effect of soil moisture on NHL 
ecosystems because no information on precipitation and soil content from C4MIP is 
provided by the standard output. Soil moisture has been suggested to play an 
important role for constraining rates of decomposition (Post et al., 1982, Qian et al. 
2008), though this effect may be small compared to the intense warming in NHL. 
Also soil decomposition is known to depend on soil carbon pool size change. We 
excluded this effect in our analysis because the NHL soil pool only increases by about 
8%, compared to the 78% increase of heterotrophic respiration. What is more, the 
NHL warming has been suggested to have the potential to lengthen the fire season 
and increase the probability of fires in boreal ecosystems (Randerson et al., 2006; 
Bond-Lamberty et al., 2007). Land use change could also have a important role in the 
carbon storage in the NHL (McGuire et al., 2001; Cramer et al., 2001). All of these 
processes are still poorly understood and need to be addressed in greater depth in 
future research. 
 
 
 105 
 
 
Chapter 7: Conclusion and Future Directions 
 
7.1 Summary 
The overall focus of this dissertation was to investigate the variability of the 
terrestrial carbon cycle and its interaction with climate over a wide range of temporal 
and spatial scales. The core of my research was based on the terrestrial vegetation and 
carbon cycle model VEGAS. Using this model, we designed sensitivity simulations 
with different purposes to explore the scientific questions, and have obtained 
encouraging results. 
7.1.1 Evaluation of VEGAS  
We used observed climate forcing to drive VEGAS from 1901-2006. With 
this offline simulation, we have evaluated VEGAS against FLUXNET observations. 
In general, VEGAS was able to capture the interannual variation of the land-
atmosphere carbon flux against the observations. However, we found that VEGAS 
simulated the seasonal cycle with the correct amplitude but opposite phase at the 
Tapajos FLUXNET station site, similar to other state-of-the art models, such as TEM 
and IBIS. The cause of this behaviour in these models may be the lack of inclusion of 
some physical processes currently, such as ?hydraulic? adaptive process for tree 
growth during the dry season, as suggested by Saleska et al. (2003). However, these 
processes are still not well understood. The extent to which, soil moisture, 
temperature or both, control the heterotrophic respiration, is still an ongoing debate. 
 
 
 106 
 
Observational stations in the tropical forests are still very sparse and observational 
records are short. In the meantime, it is encouraging to note that, VEGAS is able to 
capture a large part of the interannual variation of the land-atmosphere carbon flux 
compared to the observations there. In the boreal forest of North America, VEGAS 
can reproduce phenology of the vegetation growth, respiration and land-atmosphere 
carbon flux at the station site. Despite the deficiency of the seasonal cycle simulated 
by VEGAS in the tropics, VEGAS captures the variability in both locations on the 
interannual timescale. This evaluation of VEGAS against FLUXNET observations 
encourages the application of process-based modeling in studying the response of the 
terrestrial cycle to climate variation. At the same time, it calls for the model 
development and improvement, not only for VEGAS, but for other vegetation and 
carbon cycle models. More observation sites spanning the globe is needed in the 
future. 
7.1.2 Terrestrial carbon cycle in response to ENSO 
The interannual variability of atmospheric CO
2
 shows an association with 
ENSO cycle. It can change by as much as 4-5 PgC/yr during ENSO years. Because 
CO
2
 emissions from fossil fuel burning and land use change increase slowly with 
small year-to-year variation, understanding the mechanism of variability of the 
terrestrial carbon ecosystem in response to climate variation is of significance for the 
atmospheric CO
2
 growth rate. With this goal, we used offline simulations of VEGAS 
and studied the responses of the terrestrial carbon ecosystem to the physical climate 
variations. We isolated those climatic effects of temperature and precipitation. In 
 
 
 107 
 
particular, this was done to examine the role of soil moisture in the terrestrial 
ecosystem associated with ENSO. 
Our analysis indicates that the simulated global total land-atmosphere carbon 
flux of VEGAS agrees well on interannual timescales with the results of an inversion 
model used by R?denbeck et al. (2003), especially in the tropics. Both models are in 
agreement with the atmospheric CO
2
 growth rate variation at MLO. An evaluation of 
VEGAS against observations shows that the simulated LAI by VEGAS has better 
correlation with observed NDVI in the extratropics than the tropics. Our simulation 
shows that the land-atmosphere carbon flux in the tropics has a similar magnitude as 
the global total, confirming the dominant role of the tropics on the interannual 
timescale. This underscores the importance of extending the observational network in 
the tropics in the future. 
Further study of terrestrial timing with respect to ENSO evolution indicates 
that in the tropics VEGAS simulates a robust signal of F
ta
 with a 6-month lag of 
ENSO cycle during 1980-2000. This is largely caused by the suppression of 
vegetation and enhancement of soil decomposition, as suggested by the ?conspiracy? 
theory of Zeng et al. (2005a). Though the anomalous precipitation lags ENSO cycle 
by 1-2 months, the vegetation-dependent soil moisture takes about 5 months to 
respond due to soil memory. Less precipitation, with higher temperature drains the 
soil moisture and results in drought conditions unfavorable for vegetation growth. 
This results in a 6-month lag of the NPP from El Ni?o. Simultaneously, higher 
temperature directly enhances soil decomposition. As a result, reduction of NPP and 
 
 
 108 
 
enhancement of R
h
 contribute in same direction, and provide a large amount of 
carbon release to the atmosphere.  
The precipitation and temperature have very different mechanisms in 
regulating the above variation of vegetation growth and soil decomposition during 
ENSO events. In total, precipitation variations contributes about 56% of the F
ta
 
anomaly, mostly through soil wetness in the photosynthesis process, while 
temperature accounts for 44%, which includes 25% from direct effect on soil 
decomposition, 7% from its direct effect on vegetation respiration, and 12% from its 
indirect effect on the photosynthesis through soil wetness. Such an indirect effect, 
though significant, has not been studied in the past. The unique design and analysis of 
the appropriate sensitivity simulations in this study helped delineate the indirect effect 
of temperature affecting photosynthesis through its effect on soil wetness. The 
experimental results thus indicate that in the tropics vegetation activity suppression 
(NPP) contributes 68% of variation to F
ta
, with the remaining 32% coming from soil 
decomposition under warmer and drier conditions. Within, the total land-atmosphere 
carbon flux, about 25% is due to fire burning, which is included in the vegetation and 
soil respiration. The additive responses to temperature and precipitation forcing 
observed in the tropics helped quantify the mechanisms through which each forcing 
affects the carbon flux directly and indirectly, thus highlighting the critical role of soil 
moisture in the ecosystem and carbon cycle. This is critical, as soil wetness is a 
poorly constrained factor, observationally. 
 
 
 109 
 
7.1.3 Impact of Midlatitude drought on the terrestrial ecosystem 
The above study illustrates the canonical tropical dominance of the terrestrial 
carbon flux in response to ENSO. It quantifies the mechanisms through which 
precipitation and temperature affect the carbon flux directly and indirectly, 
highlighting the importance of soil moisture on tropical terrestrial ecosystems. 
Changes in the carbon sources and sinks in Midlatitude regions often tend to cancel 
each other, so the overall variability is weakly correlated with ENSO and it 
contributes only to a small fraction of the interannual variability in atmospheric CO
2
. 
However, during 1998-2002 the Northern Hemisphere Midlatitudes were undergoing 
a long-term drought, with the most severity in the regions of the western US, southern 
Europe, Southwest Asia, eastern Asia and Siberia (Waple et al., 2002; Hoerling and 
Kumar, 2003). During 2002-2003, the MLO station witnessed an unprecedented 
growth rate of over 2 ppm/yr. Based on our VEGAS offline simulation, we have 
established a relationship between this drought and the sustained CO
2
 growth rate 
during 2002?2003. The 1998?2002 Midlatitude drought was characterized by reduced 
precipitation and increased temperature. The anomalous climate caused widespread 
consequences to the terrestrial ecosystem there. Our results suggest that during 1998-
2002, the Northern Hemisphere land between 20?N and 50?N was a nearly constant 
source releasing 1.3 PgC/yr more relative to the 1980-98 average, consistent with the 
similar trend found from the inversion modeling by R?denbeck et al. (2003). In 
VEGAS, an enlarged spatial extent of semi-arid regions with reduced NPP followed 
the precipitation changes in the Midlatitudes during 1998-2002. Meanwhile, the long-
term warming trend during that period led to yet more respiration. As a result, the 
 
 
 110 
 
Northern Hemisphere Midlatitude was predominantly a CO
2
 source to the atmosphere 
during 1998 ?2002 with a spatial extent larger than merely the area affected by 
reduced precipitation.  
7.1.4 Carbon cycle-climate feedback 
The magnitude of future climate change for a given anthropogenic fossil fuel 
emission depends critically on the efficiency of carbon uptake by land and ocean. 
There are large uncertainties in the predicted strength of carbon-climate feedback and 
its impact on climate prediction. Using a fully coupled carbon-climate model (UMD) 
we conducted experiments to investigate the feedback on future climate change and 
compared this with the modeling results of the Hadley and IPSL models. 
Our results indicate a positive feedback from the interactive response of the 
carbon cycle to climate change. Due to this feedback, there is about 90 ppm 
additional CO
2
 by 2100
,
 which implies a 0.6?C additional warming in UMD, 
compared to the Hadley model with 250 ppm more CO
2
 and a 1.5?C additional 
warming. The IPSL model results are more similar to ours in terms of the CO
2
 
increase and warming. However, the terrestrial carbon uptake change due to the 
carbon cycle differs significantly among the three models. UMD predicts that the land 
will be a carbon source of size 20 PgC (1860-2100 cumulative) in coupled mode, 
which is the same direction as the Hadley model. However, the Hadley model 
simulated a much more robust land carbon loss due to positive carbon cycle-climate 
feedbacks, and their amplitude is 720 PgC, 6 times larger than ours. The IPSL (200 
PgC) model released a similar amount of carbon from the land to the atmosphere as  
ours, but their results differ from ours and the Hadley Center?s in that both the soil 
 
 
 111 
 
and vegetation pools continue to increase over time, in contrast to the reduction of the 
soil carbon pool storage in the Hadley and UMD models. Our sensitivity experiments 
show that if a stronger CO
2
 fertilization is applied in UMD, there will be a land 
carbon sink of 150 PgC in 2100, comarped to a 20 PgC source in the standard run. 
This modification forces UMD in the direction of the IPSL result. In a separate 
scheme using only a single soil pool, UMD was highly sensitized to warming and had 
a cumulative land carbon source of 100 PgC through year 2100, more similar to the 
Hadley model. These sensitivity experiments have helped to understand how large 
differences among the three models can arise.   
7.1.5 Future carbon uptake in Northern High Latitudes 
These large differences among the models call for intensive investigative 
projects like C4MIP. Based on the results from C4MIP, Friedlingstein et al. (2006) 
indicate a positive carbon cycle-climate feedback from all C4MIP models in the 21
st
 
century, which causes an additional 20-200 ppm CO
2
 to be injected into the 
atmosphere by 2100, leading to 0.1-1.5?C additional warming, globally. During the 
past decades, Northern High Latitudes (NHL: poleward of 60?N) have witnessed 
dramatic changes where the annual average temperatures increased by 1-2?C in 
northern Eurasia and northwestern North America, changes much larger than the 
increase of global average surface temperature of 0.65?0.2?C over the 20
th
 century 
(IPCC, 2007). The ongoing and continued projected rapid Arctic warming has been 
changing the physical, biological, and societal conditions in the NHL via ?greening? 
and northward shifts of the vegetation distribution.  
On one hand, warming and higher CO
2
 stimulate vegetation growth wherein 
 
 
 112 
 
tundra transitions to boreal forests, thus taking up CO
2
. On the other hand, warming 
accelerates decomposition of dead organic matter, thus losing soil carbon to the 
atmosphere. The resulting net land-atmosphere CO
2
 flux is important for future 
carbon cycle-climate feedback and the degree of climate change. NHL stores a large 
amount of carbon in the permafrost and peatland under conditions of -10?C annual 
mean surface temperature. This sealed carbon is potentially vulnerable to release 
under future intense warming in the NHL. Using the C4MIP model results, we have 
investigated the future carbon uptake scenario by 2100 in Northern High Latitudes 
poleward of 60?N. C4MIP models project an intense warming in the NHL with the 
ensemble mean warming there of 5.9?C, in contrast to 3.2?C for the global mean 
increase in temperature. Our analysis based on the C4MIP simulations indicates that 
the NHL is likely to be a carbon sink with a mean of 0.4 PgC/yr by 2100, with a 
cumulative increase of about 40 PgC land carbon in the C4MIP coupled simulations. 
Of this, the vegetation carbon increases by about 18 PgC, and the soil carbon 
increases by about 22 PgC. Though the magnitude of the increment in vegetation 
carbon is slightly smaller, it is robust - a 47% increase. This is of significance to the 
NHL land carbon budget in the future, when compared to the only 8% increase in soil 
carbon. In contrast to previous estimates that future NHL soil will lose carbon due to 
intense warming, our analysis indicates that the NHL soil can still continue to be a 
carbon sink by 2100. The intense warming in the NHL region does enhance soil 
decomposition and thereby drain some soil carbon; however, the NPP will increase 
by about 88% by 2100 in the NHL, and this leads to a high vegetation turnover and 
overall carbon input to the soil. We project that this fertilization will cancel the soil 
 
 
 113 
 
decomposition effects in the early 21
st
 century. Nevertheless, after 2060, the growth 
rate of NHL soil carbon begins to decrease because the soil decomposition begins to 
accelerate under intense warming. Eventually, decomposition should catch up with 
the increased biomass fertilization and turnover, sometime after 2100.  
Another interesting point is that in contrast to the robust global land carbon 
loss due to climate-carbon feedback discussed by Friedlingtein et al. (2006), our 
analysis suggests that the intense warming in the NHL may help to enhance the 
terrestrial carbon sink there. The loss of global carbon is dominated by enhanced soil 
decomposition in the tropics, while the gain in the NHL is dominated by the 
vegetation growth. Besides CO
2
 fertilization, the intense warming in the NHL tends 
to stimulate vegetation productivity. Our results show that the future NHL carbon 
storage change by 2100 is consistent with current ?greening? and a northward shift of 
vegetation distributions in the NHL, reported from the observations. Such insight into 
biomass exchanges between the vegetation and soil in the NHL highlights the critical 
role of carbon cycle-climate interactions for carbon uptake under global warming. 
 
7.2 Future directions 
7.2.1 Multi-decadal variability of terrestrial carbon 
On decadal time scales, variations of atmospheric CO
2
 growth rate are  related 
to the variability of the carbon absorption by land and ocean, land use change, and 
change in fossil fuel emissions, etc. The simulation of terrestrial carbon of the last 
century implies a relationship between the multi-decadal variability of global total 
terrestrial carbon flux with climate decadal variation, such as the PDO (Pacific 
 
 
 114 
 
Decadal Oscillation) (Figure 7.1). The examination and quantification of decadal 
variability of global and regional average terrestrial carbon would help us to 
understand the pathway of carbon destinations, such as the difference of land carbon 
sinks between the 1980s and the 1990s (IPCC, 2001; House et al., 2003).  
 
Figure 7.1 (a) Land-atmosphere carbon flux of 1901-2004 from VEGAS offline simulation; (b) 
Pacific Decadal Oscillation (PDO) Index of 1901-2004. 6-year running mean applied to smooth 
the curves to extract the low frequency signal (red). 
 
7.2.2 Soil moisture effect on the vegetation growth and soil respiration  
                
In Chapter 3, we highlighted the importance of soil moisture on the vegetation 
growth, in particular temperature?s indirect effect on the vegetation growth via 
changes in soil moisture. The role of soil moisture on vegetation growth is, however, 
potentially very model-dependent. I will extend this work with the aid of other 
vegetation-carbon models to better quantify this effect. Moreover, the conclusions 
were entirely based on our analysis for tropical land. The extratropics, especially the 
semi-arid areas in the Midlatitudes, are important cropland and agriculture regions. 
 
 
 115 
 
Water availability there plays an important role for crop growth. The regional scale of 
soil moisture?s effect on vegetation growth in the Midlatitudes is, thus, of interest. 
Soil temperature and moisture are two of the most important environmental 
parameters controlling variations in soil CO
2
 flux; however, the relationship between 
soil respiration and soil moisture is still not well understood. In Figure 3.9, our result 
shows that VEGAS simulates a relatively small effect of soil moisture on the soil 
respiration; however, Saleska et al. (2003) suggested that the soil wetness, rather than 
temperature, plays a more important role on the soil decomposition during the wet 
season at Tapajos. Mechanisms by which water content can influence soil respiration 
are by way of the influence of soil moisture on dissolved organic matter content 
loading, on soil aeration, and on the energy allocation of microbes and plants. Thus, 
further study on relationship between soil respiration and water content will focus on 
discerning potential effects of water distribution changes on the pattern of soil 
respiration in terrestrial ecosystem along with global change. 
7.2.3 Improvement of VEGAS  
The development and improvement of VEGAS is a pathway for our better 
understanding of natural physical processes. Though we have encouraging results 
from modeling interannual variability of the terrestrial carbon cycle, we are currently 
developing a new version of VEGAS with a better seasonal cycle, and incorporating 
land use change into VEGAS to examine the carbon change due to historical land use 
change. 
 
 
 116 
 
 
Appendices 
 
A.1 Vegetation carbon dynamics in VEGAS 
 
In VEGAS, the gross photosynthetic rate Ag is a function of, temperature, soil 
moisture, light and atmospheric CO
2
 concentration: 
02
)}(),,,(),(),({
gg
ACOHLAIPARWTGA ????=                   (A.1) 
  
where ?()T is a temperature dependent (Figure A.1), ?()W  is soil moisture 
dependent, ? (,,)PAR LAI H  is light dependent (Figure A.2) and ?()CO
2
 is CO2 
dependent growth factor (Figure A.3), and function G is a co-limiting function. ?()W  
uses a linear function of soil moisture. Vegetation parameters used in VEGAS can be 
found in Table A.1. A
g0
 is the maximum carbon assimilation rate, taken as 5.5 
KgC/m
2
/yr. ? is a function of Photosynthetically Active Radiation (PAR), Leaf Area 
Index (LAI), and vegetation Height (H). The PFT-dependent height and LAI 
determine the relative competitive advantage of trees vs. grasses, as well as young 
and old trees. The gross carbon assimilation is partly used for vegetation growth, so 
that the net assimilation is: 
ggrowg
ArAA ?=                                                                        (A.2) 
 
Where rgrow is the growth respiration fraction, taken as 0.3. 
The change of vegetation leaf, wood and root carbon pools are: 
                       )(
lllll
l
FSTRAf
dt
dC
+++?=                                 (A.3) 
 
 
 
 117 
 
                   
dC
dt
ffA R T S
r
lr r r r
=? ? ++() ( )1                                   (A.4) 
 
           )()1)(1(
wwwwrl
FSTRAff
dt
dCw
+++???=                            (A.5) 
 
where C
l
, C
r
, C
w
are the leaf, root and wood carbon pools, with subscript l, r, w 
representing leaf, root and wood, respectively. The loss terms are proportional to the 
respective carbon pool, such as RrfTC
llTlairl
= (), TtC
lll
=  and so on, where r
l
 is the 
rate of carbon loss; and this applies to R , T , S  for l, r, w. R is respiration, T is 
turnover, S is drought and cold stress turnover. F is carbon loss due to fire, only for 
vegetation leaf and wood. The leaf carbon pool gets a fraction of the total assimilation 
carbon at an allocation factor ffLAIH
ll
= (,) which is a function of LAI and 
vegetation height H, and f
r
 is a root allocation factor, and the rest goes into the wood 
carbon pool.  
Tree height is a function of wood mass using the bolometric function: 
 )]/exp(1[
max
k
wsw
kk
CCHH ??=                                                 (A.6) 
 
Where H
max
 is maximum height and C
ws
  is a wood mass scaling factor and k refers to 
the different 4 PFTs in VEGAS. 
The light extinction coefficient is a function of vegetation height: 
               ?? ? ?=? ? ?
max max min
( ) exp( / )HH
k
s
                          (A.7) 
 
Where lambda is light extinction coefficient and Hs is the tree height scalar. The 
fraction allocated to leaf is initially large, but decreases as the plant grows 
fL
l
=?exp[ . ]
max
015
?
?
                                                  (A.8) 
 
 
 
 118 
 
Where L is Leaf Area Index (LAI). The fraction of the rest which is allocated to roots 
f
r
 is 0.3 for trees and 0.8 for grasses. The remainder goes to the wood carbon pool. 
The light dependent growth rate is 
 ??
?
?
=?? +
0
1015{ exp[ . [ ]}
max
min
LL                               (A.9) 
 
Where?
0
 is the normalized PAR at the top of the vegetation canopy and the term in 
the braces is a shading factor related to canopy structure. This is modified from 
Sellers (1991) such that given the same LAI, taller plants are better at trapping light 
because of their three-dimensional structure. 
The competition between the PFTs depends on their fractional cover of each 
PFT. It is simply proportional to tree height: 
? =
?
H
H
k
k
k
                                                                          (A.10) 
Since tree height is the result of the balance between growth and loss, this deceptively 
simple competition rule actually includes a major factor. For instance, grasses would 
grow faster than tree initially as they allocate more resources to grow leaves. But 
since grasses have smaller maximum height they would be eventually out competed 
by trees (Eq. A.10). 
The temperature dependence of respiration is: 
 
             fkT
T
= ? ?exp[ ( )]25                                                        (A.11) 
 
For leaf and wood: k = 0.08 (Q
10 
= 2.2), root and fast soil pool: k = 0.045 (Q
10 
=1.5) 
intermediate soil pool: k = 0.03 (Q
10 
=1.35) and slow soil pool: k = 0.01(Q
10 
=1.105), 
where Q
10
 is the ratio of temperature dependence of respiration function f
T
with 10?C 
temperature change to the original one at 25?C. Leaves are subject to additional cold 
 
 
 119 
 
and drought stress under which leaves are removed at a rate more rapid than merely 
represented by the imbalance of reduced photosynthesis and normal respiration and 
turnover loss: 
SS TT w
ll ccod s
= ? + ?{exp[ . ( ) exp[ ( . )]}01 02 ?                (A.12) 
 
Where Tcod = min (Tair , 5) such that there is no stress over 5?C ; Tc is a critical 
temperature and Tc = 5, -45, -5, 5 for the 4 PFTs; ws is a drought resistant factor. 
 
Table A.1 Vegetation parameters used in VEGAS for 4 different Plant Functional Types (PFTs), 
namely, broadleaf tree, needle leaf tree, cold grass and warm grass. In the following chart, PFT1 
means broadleaf tree, PFT2 for needle leaf tree, PFT3 for cold grass, and PFT4 for warm grass. 
Parameters Physical name Value Unit 
A
g0
 Max carbon assimilation rate 5.5 KgC m
-2 
yr
-1
 
r
grow
 Growth respiration fraction 0.3 
 
H
max
 Max tree/grass height   PFT1 50 
PFT2 50 
PFT3 2 
PFT4 2 
 
m 
C
ws
  Wood mass scalar PFT1 20 
PFT2 20 
PFT3 0.1 
PFT4 0.1 
 
Kg 
H
s
 Tree height scalar 10.0 
m 
?
max
 Max light extinction coefficient 0.8 
 
?
min
 Min light extinction coefficient 0.3 
 
R
l
 Leaf respiration rate 1 
yr
-1
 
R
w
 Sapwood respiration rate 1/10 
yr
-1
 
R
r
 Root respiration rate 1/2 
yr
-1
 
T
l
 Leaf turnover rate 1 
yr
-1
 
T
w
 Sapwood turnover rate 1/30 
yr
-1
 
T
r
 Root turnover rate 1/2 
yr
-1
 
S
l
 Leaf drought and cold stress rate 1/2 
yr
-1
 
S
w
 Sapwood drought and cold stress rate 1/100 
yr
-1
 
S
r
 Root drought and cold stress rate 1/100 
yr
-1
 
T
cod
 Critical temperature for the cold 
stress 
PFT1 5.0 
PFT2 -45.0 
PFT3 -5.0 
PFT4 5.0 
 
?C 
W
s
 Water stress coefficient  10.0 
 
Q
10
 Temperature dependent for 
respiration 
Leaf and wood: 2.2 
root: 1.5 
 
 
 
 120 
 
 
 
 
Figure A.1 The temperature dependent function )(T? for 4 PFTs in VEGAS. (a) Warm tree and 
cold tree; (b) Warm grass (C4) and cold grass (C3). 
 
 
  
 
Figure A.2 The light dependent ),,( HLAIPAR?  on the radiation in VEGAS. 
 
 
 
 
 121 
 
 
Figure A.3 The CO
2
 dependent growth factor )(
2
CO?  for C3 (warm/cold tree and cold grass) 
and C4 (warm grass) in VEGAS. 
 
A.2 Soil carbon dynamics in VEGAS 
 
The total turnover from vegetation biomass Tv = Tl + Tr + Tw + Sl + Sr + Sw +  
Fvt is the input for the first soil pool Csf  (litter-fall): 
)(
fffffv
sf
FbETRT
dt
dC
+++?=                           (A.13)   
 
The carbon is lost through respiration RrfTC
f f Tf soil sf
= (), and turnover to the 
intermediate soil pool TtC
ffsf
= , and erosion EeC
ffsf
= . F
vt
 Is the turnover from 
the dead vegetation due to fire. b
f
F
f
 is the flux of fast soil pool due to fire (see the 
following section of fire module). 
 
The intermediate (Csm) and slow (Css) soil pools 
are modeled similarly as the fast soil except that there is no turnover for the slow soil 
pool (end of cascade). The soil parameters used in VEGAS can be found in Table 
A.2. 
dC
dt
TRTE
sm
fmmm
=? ++()                                                    (A.14) 
   
 
 
 122 
 
dC
dt
TRE
ss
mss
=? +()                                             (A.15) 
 
 
Table A.2 Soil parameters used in VEGAS for three soil carbon pools: fast soil, intermediate soil 
and slow soil carbon pool. 
Parameters Physical name Value Unit 
R
f
 Fast soil pool respiration rate 1/2 
 yr
-1
 
R
m
 Intermediate soil pool respiration rate 1/100 
yr
-1
 
R
s
 Slow soil pool respiration rate 1/1000 
yr
-1
 
T
f
 Fast soil pool turnover rate 1/80 
yr
-1
 
T
m
 intermediate soil pool turnover rate 1/1000 
yr
-1
 
E
l
 Fast soil pool erosion rate 1/400 
yr
-1
 
E
w
 Intermediate ast soil pool erosion rate 1/5000 
yr
-1
 
E
r
 Slow soil pool erosion rate 1/50000 
yr
-1
 
Q
10
 Temperature dependence of 
respiration 
Fast soil: 1.5 
Intermediate: 1.35 
Slow soil: 1.105 
 
 
 
 
 
Figure A.4 Soil wetness effect on the soil respiration in VEGAS with different topography. The 
topography information is scaled to the topography gradient in VEGAS and doesn?t change in 
modeling with time. The plot indicates that at a higher soil wetness condition, the respiration 
tends to be reduced somewhat (e.g., wetland) by the topographic effects.  
 
 
 123 
 
 
Figure A.5 Temperature effect Q
10
 on the respiration in VEGAS. Q
10
 =2.2 for leaf and wood, 1.5 
for the root and fast soil, 1.35 for the intermediate soil and 1.105 for the slow soil carbon pool.  
 
A.3 Fire module in VEGAS 
 
VEGAS has a fire module which calculates the land-atmosphere carbon flux 
due to biomass burning based on the effects of  moisture availability, fuel load and 
PFT dependent resistance (Eq. A.16).  When fire occurs in VEGAS, leaves suitable as 
fuel load burns completely, while wood and fast soil carbon get only partly 
combusted. The rest of dead vegetation is a turnover going into the fast soil pool that 
is decomposed later according to local soil and climate condition (indirect fire carbon 
flux) (Eq. A.17) 
ffwwllire
FbFbFbCF ++=                                (A.16) 
wwllvt
FbFbF )1()1( ?+?=                             (A.17) 
Where b
l 
, b
w
 and b
f  
are the fractions of leaf, wood and surface soil pool suitable to 
burn completely into the atmosphere. b
l
 = 1 (all leaves are burned), b
w 
= 0.2,0.5,0.9 
for broadleaf tree, needle leaf tree and grasses, respectively, b
f
 =0.2 is for the surface 
soil pool (our fast soil pool includes litter-fall). Needle leaf tree has a higher burnt 
 
 
 124 
 
fraction because of the more flammable material (sap etc.) such as the boreal forests 
in Canada and Siberia. 
lfirefirel
CPrF =                                           (A.18) 
wfirefirew
CPrF =                                          (A.19) 
ffirefiref
CPrF =                                         (A.20) 
Where the r
fire
 is a unit loss rate and P
fire
 is the probability of fire occurrence. 
fire
P  is 
a PFT weighted average of P
fire
.  
The probability of fire occurrence is predicted as: 
fueltempswetfire
PPPP =                                      (A.21) 
Where the soil dependence P
swet
 is as: 
 )1.0/exp( swetP
swet
?=                                (A.22) 
And the temperature dependence is not used, i. e. P
temp
=1. The fuel load dependence 
requires a minimum fuel load of 0.2 kg/m
2 
in order for fire to become self-sustainable 
and increases linearly afterwards: 
0.1/)2.0( ?=
fuelfuel
CP                                (A.23) 
Where C
fuel
 is the fuel load for PFT type, k dependent: 
sffwwllfuel
CbCbCbC ++=                           (A.24) 
  In VEGAS, CF
ire
 is included in the vegetation and soil respiration (R
a
 and R
h
). 
The carbon flux associated with biomass burning of vegetation leaf/wood is included 
in the vegetation respiration of vegetation (R
a
) and the part of fast soil pool due to fire 
is included in soil decomposition (R
h
). The dead vegetation due to fire turnover into 
fast soil pool, is an indirect part of fire, which is much smaller with respect to the 
 
 
 125 
 
total vegetation turnover. We estimate that 83% of the carbon fluxes due to biomass 
burning is included in the R
a
 and the remaining 17% is included in R
h. 
 
Often observationally available is fire recurrence interval (FRI) which can be 
diagnosed as: 
firefire
Pr
FRI
1
=  
                                          (A.25) 
And 
firefire
Pr  is the probability of fire per unit time. 
 
Table A.3 Parameters used in the fire module in VEGAS. 
Parameters Physical name Value Unit 
r
fire
 
A unit loss rate 
1/2.0  
yr
-1
 
b
l
 Fraction of leaf suitable to burn 
completely                                  
1.0 
 
b
w
 Fraction of wood suitable to burn 
completely 
PFT1 0.2 
PFT2 0.5 
PFT3 0.9 
PFT4 0.9 
 
 
b
f
 
Fraction of fast soil pool suitable 
to burn completely 
0.2 
 
 
 
A.4 Carbon fluxes at steady stage in VEGAS 
 
 At steady state in the VEGAS offline simulation (see the Chapter 2 for the 
description of offline simulation),  the global total GPP is 124 PgC/yr with NPP of 61 
PgC/yr, and vegetation and soil carbon pools contain 641 PgC, 1848 PgC, 
respectively. In vegetation, leaf contains 15 PgC, wood has 615 PgC and the root has 
21 PgC. As for the three soil carbon pools, the fast soil carbon pool contains 307 PgC, 
610 PgC in intermediate pool and 910 PgC in the slow soil carbon pool. The 
following figure (Figure A.7b) indicates the magnitude of the individual carbon 
fluxes in VEGAS.  
 
 
 126 
 
 
 
 
 
 
 127 
 
 
 
Figure A.6 (a) Annual mean of carbon fluxes (PgC/yr) in VEGAS. The figure indicates the 
magnitude of the individual carbon fluxes in VEGAS. The values in red in parenthesis are 
calculated at steady state of offline VEGAS simulation. The arrows in colors are used to indicate 
different biological processes. (b) Annual mean of GPP, Ra and Rh (PgC/yr) and individual 
carbon pool (PgC) of vegetation and soil in VEGAS. The physical meaning of these variables in 
this figure with respect to their names in Appendices can be found in the following table: 
 
 
 128 
 
  
Variable in the  
Appendices 
Variables in 
Figure 6 
Physical meaning of variables 
A GPP Gross Primary Productivity 
 Growth Fraction of GPP into vegetation growth 
f
l 
A leaf Fraction of GPP into vegetation leaf 
(1-f
l
)(1-f
r
)A wood Fraction of GPP into vegetation wood 
(1-f
l
)f
r 
A root Fraction of GPP into vegetation root 
r
grow 
A
g
 rspgrow Growth respiration  
R
l
 rspleaf Vegetation leaf respiration 
R
w
 rspwood Vegetation wood respiration 
R
r
 rsproot Vegetation root respiraiton 
T
l
 tovleaf Vegetation leaf turnover into soil 
T
w
 tovwood Vegetation wood turnover into soil 
T
r
 tovroot Vegetation root turnover into soil 
S
l
 Stsleaf Vegetation leaf stress into soil  
S
w
 Stswood Vegetation wood stress into soil  
S
r
 Stsroot Vegetation root stress into soil  
   
b
l
 F
l
 burnleaf Vegetation leaf burnt into atmosphere 
b
w 
F
w
 burnwood Vegetation wood burnt into atmosphere 
F
l
+F
w
 Biomass burning Vegetation wood/leaf suitable for burning 
(1-b
l
) F
l
 tovfireleaf Vegetation dead leaf after biomass burning into soil  
(1-b
w
)F
w
 tovfirewood Vegetation dead wood after biomass burning into soil 
b
f
 F
f
 firesfast Fast soil pool burnt into atmosphere 
   
R
f
 rspsfast Fast soil respiration 
R
m
 rspsmed Intermediate soil respiration 
R
s
 rspsslow Slow soil respiration 
T
f
 tovfast Fast soil turnover  
T
m
 tovmed Intermediate soil turnover 
E
f
 erosfast Fast soil erosion  
E
m
 erosmed Intermediate soil erosion 
E
s
 erosslow Slow soil erosion 
   
NPP  Net Primary Productivity, GPP-R
a
 
R
a
 R
a
 Autotrophic respiration or plant respiration 
R
h
 R
h
 Heterotrophic respiration or soil decomposition 
NEE  Net Ecosystem Exchane 
   
 
C
l
 Cleaf Vegetation leaf carbon pool 
C
w
 Cwood Vegetation wood carbon pool 
C
r
 Croot Vegetation root carbon pool 
C
sf
 Csfast Fast soil carbon pool 
C
sm
 Csmed Intermediate soil carbon pool 
C
ss
 Csslow Slow soil carbon pool 
 
 
 
 129 
 
A.5 Variables used in the carbon research community 
 
The most important common variables used by the carbon research 
community are as follows: 
GPP A
g
=                                                                         (A.26)  
 
NPP GPP R
a
= ?                                                              (A.27)  
 
NEP NPP R
h
= ?                                                            (A.28)  
 
      NEE NEP= ?                                                                 (A.29) 
 
Where GPP (Gross Primary Production) is the difference of total photosynthetic 
carbon and growth respiration. NPP denotes net primary production, R
a
 is autotrophic 
respiration, NEP is net ecosystem product, and R
h
 is heterotrophic respiration. NEE is 
the land-atmosphere carbon flux.  
 
 
 130 
 
 
Glossary 
AVHRR Advanced Very High Resolution Radiometer 
C4MIP Coupled Carbon Cycle Climate Model Inter-comparison Project 
CASA Carnegie-Ames-Stanford Approach 
CCMLP Carbon Cycle Model Linkage Project 
CMDL Climate Monitoring & Diagnostics Laboratory 
CRU Climate Research Unit 
CO
2
 Carbon Dioxide  
DGVM Dynamics Global Vegetation Model 
ENSO El Ni?o-Southern Oscillation 
ESM Earth System Model 
gCO
2
 CO
2 
growth rate, time derivative of the CO
2
 concentration of Mauna 
Loa Observatory in this thesis  
EVI Enhanced Vegetation Index  
GIMMS Global Inventory Monitoring and Modeling Studies 
GPP Gross Primary Production 
GRACE 
 
Gravity Recovery and Climate Experiment 
FLUXNET 
 
A global network of micro-meteorological tower sites with CO
2
 flux 
F
ta
 Land-atmosphere carbon flux, same as NEE 
 
GISS Goddard Institute for Space Studies 
IBIS Integrated Biosphere Simulator 
 
 
 131 
 
IGBP International Geosphere-Biosphere Programme 
IPCC Intergovernmental Panel on Climate Change 
IPSL Institut Pierre Simon Laplace, France 
LAI Leaf Area Index 
MEI Multivariate ENSO Index 
MLO Mauna Loa Observatory 
MODIS Moderate Resolution Imaging Spectroradiometer 
NASA National Aeronautics and Space Administration 
NDVI Normalized Difference Vegetation Index 
NEE Land-atmosphere carbon flux; -NEP 
NEP Net Ecosystem Production 
NHL Northern High Latitudes poleward of 60?N in this thesis 
NPP Net Primary Production 
NOAA National Oceanic and Atmospheric Administration 
OAGCMs Coupled atmosphere-ocean general circulation models   
PDO Pacific Decadal Oscillation 
PER Precipitation-Evaporation-and-Runoff 
PFT Plant Functional Types 
PgC petagrams of carbon, 1 PgC = 10
15
 gram carbon 
ppm parts per million by volume 
Q
10
 Function of the temperature dependence of soil respiration, this factor 
represents the difference in respiration rates over a 10?C interval 
QTCM Quasi-equilibrium Tropical Circulation Model 
 
 
 132 
 
R
a
 Autotrophic Respiration 
R
e
 Total Ecosystem Respiration 
 
R
h
 Heterotrophic Respiration 
SLand Simple Land surface Model 
SOI Southern Oscillation Index 
SST Sea Surface Temperature 
Swet Soil wetness 
TEMs Terrestrial Ecosystem Model 
UMD University of Maryland 
VEGAS VEgetation-Global-Atmosphere-Soil 
WCRP World Climate Research Programme 
 
 
 
 
 
 
 
 133 
 
Bibliography 
 
Alexeev, V.A., P. L. Langen, and J. R. Bates, 2005: Polar amplification of surface 
warming on an aquaplanet in ?ghost forcing? experiments without sea ice 
feedbacks. Climate Dynamics, 24, 655?666. 
Arctic climate impact assessment. 2005: Lead authors, Jim Berner and others. 
Carolyn Symon, Lelani Arris, and Bill Heal, editors. pp. 101-144. Cambridge 
University Press. 
Angert, A., S. Biraud, C. Bonfils, W. Buermann, and I. Fung, 2004: CO
2
 seasonality 
indicates origins of post-Pinatubo sink. Geophys. Res. Lett., 31, L11103, 
doi:10.1029/2004GL019760. 
Angert, A., S. Biraud, C. Bonfils, C. C. Henning, W. Buermann, J. Pinzon, C. J. 
Tucker, and I. Fung, 2005: Drier summers cancel out the CO
2
 uptake 
enhancement induced by warmer springs. Proc. Natl. Acad. Sci. U.S.A., 
102(31), 10,823?10,827. 
Bacastow, B. R., 1976: Modulation of atmospheric carbon dioxide by the Southern 
Oscillation. Nature, 261, 116?118. 
Bala, G., K.Caldeira, A. Mirin, M. Wickett, C. Delire, and T. J. Phillips, 2006: 
Biogeophysical effects of CO
2
-fertilization on global climate. Tellus B, 58, 620?
27. 
Baker, D. F., R. Law, K. Gurney, P. Rayner, P. Peylin, A. Denning, P. Bousquet, L. 
Bruhwiler, Y-H. Chen, P. Ciais, I. Fung, M. Heimann, J. John, T. Maki, S. 
Maksyutov, K. Masarie, M. Prather, B. Pak, S. Taguchi, and Z. Zhu, 2006: 
Transcom 3 inversion intercomparison: Impact of transport model errors on the 
interannual variability of regional CO
2
 fluxes, 1988-2003. Global Biogeochem. 
Cycles, 20, GB1002, doi:10.1029/2004GB002439. 
Berthelot, M., P. Friedlingstein, P. Ciais, P. Monfray, J. L. Dufresne, H. Le Treut, and 
L. Fairhead, 2002: Global response of the terrestrial biosphere to CO
2
 and 
climate change using a coupled climate-carbon cycle model. Global 
Biogeochem. Cycles, 16, 1084, doi:10.1029/2001GB001827.  
 
 
 134 
 
Betts, R., P. Cox, S. Lee, and F. Woodward, 1997: Contrasting physiological and 
structural vegetation feedbacks in climate change simulations. Nature, 387, 
796?799. 
Betts, R. A, 2004: Global vegetation and climate: self-beneficial effects, climate 
forcings and climate feedbacks. Journal de Physique, IV 121, 37?60. 
Bond-Lamberty, B., C. K. Wang, and S. T. Gower, 2004: Contribution of root 
respiration to soil surface CO
2
 flux in a boreal black spruce chronosequence. 
Tree Physiol., 24(12), 1387?1395. 
Bond-Lamberty, B. S. D. Peckham, D. E. Ahl S. T. Gower, 2007: Fire as the 
dominant driver of central Canadian boreal forest carbon balance. Nature, 450, 
89-92. 
Bousquet, P., P. Peylin, P. Ciais, C. Le, Quere, P. Friedlingstein, and P. Tans, 2000: 
Regional changes in carbon dioxide fluxes of land and oceans since 1980. 
Science, 290, 1342?1346. 
Braswell, B., D. Schimel, E. Linder, and B. Moore, 1997: The response of global 
terrestrial ecosystems to interannual temperature variability. Science, 278, 870?
872.  
Bronson, D.R., S. Gower, M. Tanner, S. Linder, and Van Herk, 2008: Response of 
soil surface CO
2
 flux in a boreal forest to ecosystem warming. Global Change 
Biol., 14, 856-867. 
Budyko, M. I., 1974: Climate and life, Academic, New York. 
Buermann, W., B. Anderson, and C. Tucker, R. Dickinson, W. Lucht, C. Potter, and 
R. Myneni, 2003: Interannual covariability in northern hemisphere air 
temperatures and greenness associated with El Ni?o-Southern Oscillation and 
the Arctic Oscillation. J. Geophys. Res., 108(D13), 4396, doi:10.1029/2002JD 
002630.  
Cao, M., and F. Woodward, 1998: Dynamic responses of terrestrial ecosystem carbon 
cycling to global climate change. Nature, 393, 249?252. 
Chapin, F. S., and E. Matthews, 1993: Boreal carbon pools: approaches and 
constraints in global extrapolations. In: Vinson, T. S. and Kolchugina, T. P. 
 
 
 135 
 
(Eds) Carbon Cycling in Boreal Forest and Sub-Arctic Ecosystems, pp. 9-20. 
Proceedings International Workshop, EPA 600/R-93/084. 
Cook, B., N. Zeng, and J-H. Yoon, 2008: Future drying of the southern Amazon and 
central Brazil. Submitted to Proc. Natl. Acad. Sci. 
Cox, P. M., R. A. Betts, C. D. Jones, S. A. Spall, and I. J. Totterdell, 2000: 
Acceleration of global warming due to carbon-cycle feedbacks in a coupled 
climate model. Nature, 408, 184?187. 
Cox, P. M., R. A. Betts, M. Collins, P. P. Harris, C. Huntingford, and C. D. Jones, 
2004: Amazonian forest diaback under climate-carbon projections for the 21st 
century. Theor. Appl. Climatol., 78, 137?156. 
Cramer, W., D. W. Kicklighter, A. Bondeau, B. Moore, C. Churkina, B. Nemry, A. 
Ruimy, and A. L. Schloss, 1999. Comparing global models of terrestrial net 
primary productivity (NPP): overview and key results. Global Change Biol. 5, 
1?15, Suppl. 1. 
Cramer, W., A. Bondeau, F. Woodward, I. Prentice, R. Betts, V. Brovkin, P. Cox, V. 
Fisher, J. Foley, A. Friend, C. Kucharik, M. Lomas, N. Ramankutty, S. Sitch, B 
Smith, A. White, and C. Young-Molling, 2001: Global response of terrestrial 
ecosystem structure and function to CO
2
 and climate change: results from six 
dynamic global vegetation models. Global Change Biol. 7(4), 357?373. 
Dargaville, R., M. Heimann, A. McGuire, I. Prentice, D. Kicklighter, F. Joos, J. 
Clein, G. Esser, J. Foley, J. Kaplan, R. Meier, J. Melillo, B. Moore, N. 
Ramankuty, T. Reichenau, A. Schloss, S. Sitch, H. Tian, L. Williams, and U. 
Wittenberg, 2002: Evaluation of terrestrial carbon cycle models with 
atmospheric CO
2
 measurements: Results from transient simulations considering 
increasing CO
2
, climate, and land-use effects. Global Biogeochem. Cycles, 
16(4), 1092, doi:10.1029/2001GB001426.  
Davidson, E. A., I. A. Janssens, and Y. Luo, 2006: On the variability of respiration in 
terrestrial ecosystems: Moving beyond Q
10
. Global Change Biol., 12, 154?164. 
DeFries, R. S., R. A. Houghton, M. C. Hansen, C. B. Field, D. Skole, and J. 
Townshend, 2002: Carbon emissions from tropical deforestation and regrowth 
 
 
 136 
 
based on satellite observations for the 1980s and 1990s. Proc. Natl. Acad. Sci., 
99(22), 14,256?14,261. 
Dufresne, J., P. Friedlingstein, M. Berthelot, L. Bopp, P. Ciais, L. Fairhead, H. Le, 
Treut, and P. Monfray, 2002: On the magnitude of positive feedback between 
future climate change and the carbon cycle. Geophys. Res. Lett., 29(10), 1405, 
doi:10.1029/2001GL013777. 
Eliasson, P. E., R. E. McMurtrie, D. A. Pepper, M. St?mgren, S. Linder, and G. I. 
?gren, 2005: The response of heterotrophic CO
2
 flux to soil warming. Global 
Change Biol., 11, 167?181. 
Feely, R., J. Boutin, C. Cosca, Y. Dandonneau, J. Etcheto, H. Inoue, M. Ishii, C. Le 
Quere, D. Mackey, M. McPhaden, N. Metzl, A. Poisson, and R. Wanninkhof, 
2002: Seasonal and interannual variability of CO
2
 in the equatorial pacific. 
Deep Sea Res. Part II, 49, 2443?2469. 
Field, C. B., 2001: Plant physiology of the ?missing? carbon sink. Plant Physiol., 
125, 25? 28. 
Friedlingstein, P., L. Bopp, P. Ciais, J. Dufresne, L. Fairhead, H. LeTreut, P. 
Monfray, and J. Orr, 2001: Positive feedback between future climate change 
and the carbon cycle. Geophys. Res. Lett., 28 (8), 1543?1546. 
Friedlingstein, P., J. Dufresne, P. Cox, and P. Rayner, 2003: How positive is the 
feedback between climate change and the carbon cycle? Tellus B, 55, 692?700. 
Friedlingstein, P., P. Cox, R. Betts, L. Bopp, W. von Bloh, V. Brovkin, S. Doney, M. 
Eby, I. Fung, B. Govindasamy, J. John, C. Jones, F. Joos, T. Kato, M. 
Kawamiya, W. Knorr, K. Lindsay, H. D. Matthews, T. Raddatz, P. Rayner, C. 
Reick, E. Roeckner, K.-G. Schnitzler, R. Schnur, K. Strassmann, S. Thompson, 
A. J. Weaver, C. Yoshikawa, and N. Zeng, 2006: Climate-carbon cycle 
feedback analysis: Results from the C4MIP model intercomparison, J. Clim., 
19(14), 3337?3353. 
Fung, I., 2000: IGBP NEWS, 41. 
Fung, I., S. Doney, K. Lindsay, and J. John, 2005: Evolution of carbon sinks in a 
changing climate. Proc. Natl. Acad. Sci. 102(32), 11,201?11,206. 
 
 
 137 
 
Goulden, M. L., S. C. Wofsy, J. W. Harden, S. E. Trumbore, P. M. Crill, S. T. Gower, 
T. Fries, B. C. Daube, S.-M. Fan, D. J. Sutton, A. Bazzaz, and J. W. Munger, 
1998: Sensitivity of boreal forest carbon balance to soil thaw. Science, 279, 
214?221. 
Gu, L. H., D. D. Baldocchi, S. C. Wofsy, J. W. Munger, J. J. Michalsky, S. P. 
Urbanski, and T. A. Boden, 2003: Response of a deciduous forest to the Mount 
Pinatubo eruption: Enhanced photosynthesis. Science, 299, 2035?2038. 
Gurney, K., R. Law, A. Denning, P. Rayner, D. Baker, P. Bousquet, L. Bruhwiler, Y. 
Chen, P. Ciais, S. Fan, I. Fung, M. Gloor, M. Heimann, K. Higuchi, J. John, T. 
Maki, S. Maksyutov, K. Masarie, P. Peylin, M. Prather, B. Pak, J. Randerson, J. 
Sarmiento, S. Taguchi, T. Takahashi, and C. Yuen, 2002: Towards robust 
regional estimates of CO
2
 sources and sinks using atmospheric transport 
models. Nature, 415, 626?630. 
Hansen, J., A. Lacis, D. Rind, G. Russell, P. Stone, I. Fung, R. Ruedy, and J. Lerner, 
1983: Climate sensitivity: Analysis of feedback mechanisms in climate 
processes and climate sensitivity, in Climate Processes and Climate Sensitivity. 
Geophys. Monogr. Ser., vol. 29, pp. 130? 163, AGU, Washington, D. C. 
Hansen, J., E. R. Ruedy, J. Glascoe, and M. Sato, 1999: GISS analysis of surface 
temperature change. J. Geophys. Res., 104, 30997?31022. 
Hashimoto, H., R. Nemani, M. White, W. Jolly, S. Piper, C. Keeling, R. Myneni, and 
S. Running, 2004: El Ni?o-Southern Oscillation-induced variability in terrestrial 
carbon cycling. J. Geophys. Res., 109, D23110, doi:10.1029/2004JD04959. 
Heimann, M., and M. Reichstein, 2008: Terrestrial ecosystem carbon dynamics and 
climate feedbacks. Nature, 451, 289-292. 
Heimann, M., G. Esser, and A. Haxeltine, 1998: Evaluation of terrestrial carbon cycle 
models through simulations of the seasonal cycle of atmospheric CO
2
: First 
results of a model intercomparison study. Global Biogeochem. Cycles, 12(1), 1?
24. 
Hoerling, M., and A. Kumar, 2003: The perfect ocean for drought. Science, 299, 691? 
694. 
 
 
 138 
 
Holland, M. M., and C. M. Bitz, 2003: Polar amplification of climate change in 
coupled models. Climate Dynamics, 21, 221?232. 
Houghton, R. A., 1999: The annual net flux of carbon to the atmosphere from 
changes in land use 1850?1990. Tellus, 51B, 298?313. 
Houghton, R. A., J. L. Hackler, and K. T. Lawrence, 1999: The US carbon budget: 
Contributions from land-use change. Science, 285, 574?578. 
Houghton, R. A., 2000: Interannual variability in the global carbon cycle. J. Geophys. 
Res., 105, 20,121?20,130. 
Houghton, R. A., 2003a: Revised estimates of the annual net flux of carbon to the 
atmosphere from changes in land use and land management 1850?2000, Tellus, 
Ser. B, 55, 378?390. 
Houghton, R. A., 2003b: Why are estimates of the terrestrial carbon balance so 
different?, Global Change Biol., 9, 500?509. 
House, J. I., I. C. Prentice, N. Ramankutty, R. A. Houghton, and M. Heimann, 2003: 
Reconciling apparent inconsistencies in estimates of terrestrial CO
2
 sources and 
sinks. Tellus, Ser. B, 55(2), 345?363. 
Huete, A., K. Didan, Y. Shimabukuro, P. Ratana, S. Saleska, L. Hutyra, W. Yang, R. 
Nemani, and R. Myneni, 2006: Amazon rain forests green-up with sunlight in 
dry season. Geophys. Res. Lett., 33, L06405, doi:10.1029/2005GL025583  
Ichii, K., H. Hashimoto, R. Nemani, and M. White, 2005: Modeling the interannual 
variability and trends in gross and net primary productivity of tropical forests 
from 1982 to 1999. Global Planet. Change, 48, 274?286. 
IPCC (2007) Climate change 2007: the physical science basis. Contribution of 
Working Group I. In: Fourth Assessment Report of the Intergovernmental Panel 
on Climate Change (eds Solomon S, Qin D, Manning M, Chen Z, Marquis M, 
Averyt KB, Tignor M, Miller HL), Cambridge University Press, Cambridge, 
996pp. 
Jones, C. D. and P. M. Cox, 2001: Modeling the volcanic signal in the atmospheric 
CO
2
 record. Global Biogeochem. Cycles, 15(2), 453?465. 
 
 
 139 
 
Jones, C., M. Collins, P. Cox, and S. Spall, 2001: The carbon cycle response to 
ENSO: A coupled climate-carbon cycle model study. J. Climate, 14, 4113?
4129. 
Jones, C. D., P. Cox, and C. Huntingford, 2003a: Uncertainty in climate carbon-cycle 
projections associated with the sensitivity of soil respiration to temperature, 
Tellus, Ser. B, 55(2), 642?648.  
Jones, C. D., P. M. Cox, R. L. H. Essery, D. L. Roberts, and M. J. Woodage, 2003b: 
Strong carbon cycle feedbacks in a climate model with interactive CO
2
 and 
sulphate aerosols. Geophys. Res. Lett., 30(9), 1479, doi:10.1029/2003GL016 
867. 
Jones, C., and P. Cox, 2005: On the significance of atmospheric CO
2
 growth rate 
anomalies in 2002?2003. Geophys. Res. Lett., 32, L14816, doi:10.1029/2005 
GL023027. 
Joos, F., I. C. Prentice, S. Sitch, R. Meyer, G. Hooss, G. Plattner, S. Gerber, and K. 
Hasselmann, 2001: Global warming feedbacks on terrestrial carbon uptake 
under the Intergovernmental Panel on Climate Change (IPCC) emission 
scenarios. Global Biogeochem. Cycles, 15, 891?908. 
Kaplan, J. O., and M. New, 2006: Arctic climate change with a 2?C global warming: 
Timing, climate patterns and vegetation change. Clim. Change, 79, 213?241, 
doi: 10.1007/s10584-006-9113-7. 
Keeling, D. C., and R. Revelle, 1985: Effects of ENSO on the atmospheric content of 
CO
2
. Meteroritics, 20, 437?450. 
Keeling, C., J. Chin, and T. Whorf, 1996: Increased activity of northern vegetation 
inferred from atmospheric CO
2
 measurements. Nature, 384, 146?149. 
Kindermann, J., G. Wurth, G. Kohlmaier, and B. FW, 1996: Interannual variation of 
carbon exchange fluxes in terrestrial ecosystems. Global Biogeochem. Cycles, 
10, 737?755. 
Knorr, W., 2000: Annual and interannual CO
2
 exchanges of the terrestrial biosphere: 
process-based simulations and uncertainties. Global Ecol. Biogeogr., 9(3), 225?
252. 
 
 
 140 
 
Knyazikhin, Y., J. V. Martonchik, D.J. Diner, R. B. Myneni, M. Verstraete, B. Pinty, 
and N. Gobron, 1998: Estimation of vegetation canopy leaf area index and 
fraction of absorbed photosynthetically activeradiation from atmosphere-
corrected MISR data. J. Geophys. Res., 103, 32239?32256. 
Langenfelds, R. L., R. J. Francey, B. C. Pak, L. P. Steele, J. Lloyd, C. M. Trudinger, 
and C. E. Allison, 2002: Interannual growth rate variations of atmospheric CO
2
 
and its ?13C, H2, CH4, and CO between 1992 and 1999 linked to biomass 
burning. Global Biogeochem. Cycles, 16(3), 1048, doi:10.1029/2001GB001466. 
Lawrence, D. M., and A. G. Slater, 2005: A projection of severe near surface 
permafrost degradation during the 21
st 
century. Geophys. Res. Lett., 32, L24401, 
doi:10.1029/2005GL025080. 
Le Quere, C., O. Aumont, L. Bopp, P. Bousquet, P. Ciais, R. Francey, M. Heimann, 
C. Keeling, R. Keeling, H. Kheshgi, P. Peylin, S. Piper, I. Prentice, and P. 
Rayner, 2003: Two decades of ocean CO
2
 sink and variability. Tellus B, 55, 
649?656. 
Levitus, S., J. I. Antonov, T. P. Boyer, and C. Stephens, 2000: Warming of the world 
ocean. Science, 287, 2225?2229. 
Linderholm, H.W., 2006: Growing season changes in the last century. Agricultural 
and Forest Meteorology, 137, 1?14. 
Liu, J., J. A. Curry, Y. Dai, and R. Horton, 2007: Causes of the northern high-latitude 
land surface winter climate change. Geophys. Res. Lett., 34, L14702, 
doi:10.1029/2007GL030196. 
Lotsch, A., M. A. Friedl, B. T. Anderson, and C. J. Tucker, 2005: Response of 
terrestrial ecosystems to recent Northern Hemispheric drought. Geophys. Res. 
Lett., 32, L06705, doi:10.1029/2004GL022043. 
Lucht, W., I. Prentice, R. Myneni, S. Sitch, P. Friedlingstein, W. Cramer, P. 
Bousquet, W. Buermann, and B. Smith, 2002: Climatic control of the high-
latitude vegetation greening trend and pinatubo effect. Science, 296, 1687?1689. 
Lucht, W., S. Schaphoff, T. Erbrecht, U. Heyder, and W. Cramer, 2006: Terrestrial 
vegetation redistribution and carbon balance under climate change. Carbon 
Balance Manage., 1(6), doi:10.1186/1750-0680-1-6. 
 
 
 141 
 
Marland, G., T. A. Boden, and R. J. Andres, 2005: Global, regional, and national CO
2
 
emissions. Trends: A compendium of data on global change, Carbon Dioxide 
Information Analysis Center, Oak Ridge National Laboratory, U.S. Department 
of Energy, Oak Ridge, TN. 
Markgraf, V., and H. F. Diaz, 2000: The past ENSO record: A synthesis. In El Ni?o 
and the Southern Oscillation: Multiscale Variability and Global and Regional 
Impacts, Diaz, H. F., and Markgraf, V. (Eds.), Cambridge University Press, 
465-488. 
Matthews, H. D., M. Eby, A. J. Weaver, and B. J. Hawkins, 2005: Primary 
productivity control of simulated carbon cycle-climate feedbacks. Geophys. Res. 
Lett., 32, L14708, doi:10.1029/2005GL022941. 
McGuire, A. D., and J. E. Hobbie, 1997: Global climate change and the equilibrium 
responses of carbon storage in arctic and subarctic regions. pp. 47-48. IN: Arctic 
System Science Modeling Workshop Report. The Arctic Research Consortium 
of the United States. Fairbanks, Alaska. 
McGuire, A. D., J. S. Clein, J. Melillo, D. Kicklighter, R. Meier, C. Vorosmarty, and 
M. Serreze, 2000: Modeling carbon responses of tundra ecosystems to historical 
and projected climate: sensitivity of pan-Arctic carbon storage to temporal and 
spatial variation in climate. Global Change Biol., 6, 141? 159. 
McGuire, A., S. Sitch, J. Clein, R. Dargaville, G. Esser, J. Foley, M. Heimann. F. 
Joos, J. Kaplan, D. Kicklighter, R. Meier, J. Melillo, B. Moore, I. Prentice, N. 
Ramankutty, T. Reichenau, A. Schloss, H. Tian, L. Williams, and U. 
Wittenberg, 2001: Carbon balance of the terrestrial biosphere in the twentieth 
century: Analyses of CO
2
, climate and land use effects with four process-based 
ecosystem models. Global Biogeochem. Cycles, 15, 183?206. 
McGuire, A.D., F. Chapin, J. Walsh, and C. Wirth, 2006: Integrated regional changes 
in arctic climate feedbacks: implications for the global climate system. Annual 
Review of Environment and Resources, 31, 61?91. 
Melillo, J. M., P. A. Steudler, J. D. Aber, K. Newkirk, H. Lux, F. P. Bowles, C. 
Catricala, A. Magill, T. Ahrens, and S. Morrisseau1, 2002: Soil warming and 
carbon-cycle feedbacks to the climate system. Science, 298, 2173?2176. 
 
 
 142 
 
Melillo J. M., D. W. Kicklighter, A. D. McGuire, W. T. Peterjohn, K. M. Newkirk, 
1995: Global change and its effects on soil organic carbon stocks. In: Role of 
Nonliving Organic Matter in the Earth's Carbon Cycle (eds Zepp RG, Sonntag 
Ch), pp.175-189. John Wiley & Sons Ltd. 
Mitchell, T. D., and P. D. Jones, 2005: An improved method of constructing a 
database of monthly climate observations and associated high-resolution grids. 
Int. J. Climatol., 25, 693?712. 
Montaigne, F., 2002: The Great Northern Forest: Boreal. National Geographic, 201, 
42?65. 
Myneni, R., C. Keeling, C. Tucker, G. Asrar, and R. Nemani, 1997: Increased plant 
growth in the northern high latitudes from 1981 to 1991. Nature, 386, 698?702. 
Myneni, R. B., C. J. Tucker, G. Asrar, and C. D. Keeling, 1998: Interannual 
variations in satellite-sensed vegetation index data from 1981 to 1991. J. 
Geophys. Res., 103(D6), 6145?6160. 
Neelin, J., and N. Zeng, 2000: The first quasi-equilibrium tropical circulation model 
formulation. J. Atmos. Sci., 57, 1741?1766.  
Nemani, R., C. D. Keeling, H. Hashimoto, W. M. Jolly, S. C. Piper, C. J. Tucker, R. 
B. Myneni, and S. W. Running, 2003: Climate-driven increases in global 
terrestrial net primary production from 1982 to 1999. Science, 300, 1560?1563. 
Nemani, R. R., and S. W. Running, 1989: Testing a theoretical climate-soil-leaf area 
hydrologic equilibrium of forests using satellite data and ecosystem simulation. 
Agric. For. Meteorol., 44, 245-260. 
Nemry, B., L. Francois, J. Gerard, A. Bondeau, and M. Heiman, 1999: Comparing 
global models of terrestrial net primary productivity (NPP): analysis of the 
seasonal atmospheric CO
2
 signal. Global Change Biol., 5, 65?76 Suppl. 1. 
New, M. G., M. Hulme, and P. D. Jones, 1999: Representing 20
th
 century space-time 
climate variability. I. Development of a 1961?1999 mean monthly terrestrial 
climatology. J. Climate, 12, 829?865. 
Oliveira, R., T. Dawson, S. Burgess, and D. Nepstad, 2005: Hydraulic redistribution 
in three Amazonian trees. Oecologia, 145, 354?363. 
 
 
 143 
 
Overland, J. E., M. C. Spillane and N. N. Soreide, 2004: Integrated analysis of 
physical and biological Pan-Arctic change. Clim. Change, 63, 291?322. 
Page, S. E., F. Siegert, J. O. Rieley, H. Boehm, A. Jaya, and S. Limin, 2002: The 
amount of carbon released from peat and forest fires in Indonesia during 1997. 
Nature, 420, 61?65.  
Piao, S. L., P. Ciais, P. Friedlingstein, P. Peylin, M. Reichstein, S. Luyssaert, H. 
Margolis, J. Y. Fang, A. Barr, A. P. Chen AP, A. Grelle, D. Y. Hollinger, T. 
Laurila, A. Lindroth, A. D. Richardson, T. Vesala,  2008: Net carbon dioxide 
losses of northern ecosystems in response to autumn warming. Nature 451, 49?
53. 
Piao, S. L., P. Friedlingstein, O. Ciais, L. M. Zhou, and A. P. Chen, 2006: The effect 
of climate and CO
2
 changes on the greening of the Northern Hemisphere over 
the past two decades. Geophys. Res. Lett., 33, L23402, doi:10.1029/2006GL 
028205. 
Piao, S. L., P. Friedlingstein, P. Ciais, N. Viovy, and J. Demarty, 2007: Growing 
season extension and its impact on terrestrial carbon cycle in the Northern 
Hemisphere over the past 2 decades. Global Biogeochem. Cycles, VOL. 21, 
GB3018, doi:10.1029/2006GB002888. 
Plattner, G., F. Joos, and T. Stocker, 2002: Revision of the global carbon budget due 
to changing air-sea oxygen fluxes. Global Biogeochem. Cycles, 16(4), 1096, 
doi:10.1029/2001GB001746. 
Post, W. M., W. R. Emanuel, P. J. Zinke, and A. G. Strangenberger, 1982: Soil 
carbon pools and world life zones. Nature, 298, 156?159. 
Potter, C., S. Klooster, M. Steinbach, P. Tan, V. Kumar, S. Shekhar, R. Nemani, and 
R. Myneni, 2003: Global teleconnections of climate to terrestrial carbon flux. J. 
Geophys. Res., 108(D17), 4556, doi:10.1029/2002JD002979  
Prentice, I. C. et al., 2001: The carbon cycle and atmospheric CO
2
, in The 
Intergovernmental Panel on Climate Change (IPCC) Third Assessment Report, 
edited by J. T. Houghton and D. Yihui, chap. 3, pp. 185?237, Cambridge Univ. 
Press, New York. 
 
 
 144 
 
Qian, H., R. Joseph, and N. Zeng, 2008: Carbon uptake by the Northern High 
Latitude Terrestrial Biosphere under Climate Change in the 21
st
 Century in 
C4MIP models. Submitted to Global change boil. 
Qian, H., R. Joseph, and N. Zeng, 2008: Response of the Terrestrial Carbon Cycle to 
the El Ni?o-Southern Oscillation, Tellus B, doi: 10.1111/j.1600-0889.2008.00 
360.x. 
Randerson, J., C. Field, I. Fung, and P. Tans, 1999: Increases in early season 
ecosystem uptake explain recent changes in the seasonal cycle of atmospheric 
CO
2
 at high northern latitudes. Geophys. Res. Lett., 26, 2765?2768. 
Rasmusson, E. M. and T. H. Carpenter, 1982: Variations in tropical sea surface 
temperature and surface wind fields associated with the Southern Oscillation / 
El. Ni?o. Mon. Wea. Rev, 110, 354-384. 
Raupach, M., G. Marland, P. Ciais, C. L. Qu?r?, J. G. Canadell, G. Klepper, and C. 
Field, 2007: Global and regional drivers of accelerating CO
2
 emissions. Proc. 
Natl. Acad. Sci. U.S.A., 104, 10,288?10,293. 
Rayner, P., R. Law, and R. Dargaville, 1999: The relationship between tropical CO
2
 
fluxes and the El Ni?o-Southern Oscillation. Geophys. Res. Lett., 26, 493?496. 
Reichstein, M., D. Papale, R. Valentini, M. Aubinet, C. Bernhofer, A. Knohl, T. 
Laurila, A. Lindroth, E. Moors, K. Pilegaard, and G. Seufert, 2007: 
Determinants of terrestrial ecosystem carbon balance inferred from European 
eddy covariance flux sites. Geophys. Res. Lett., 34, L01402; doi:10.1029/2006 
GL027880. 
R?denbeck, C., S. Houweling, M. Gloor, and M. Heimann, 2003: CO
2
 flux history 
1982-2001 inferred from atmospheric data using a global inversion of 
atmospheric transport. Atmos. Chem. Phys., 3, 1919?1964. 
Ropelewski, C. F., and M. S. Halpert, 1987: Global and regional scale precipitation 
patterns associated with the El-Nino Southern Oscillation. Mon. Weather Rev., 
115, 1606?1626. 
Russell, J. L., and J. M. Wallace, 2004: Annual carbon dioxide drawdown and the 
Northern Annular Mode. Global Biogeochem. Cycles, 18, GB1012, doi:10.1029 
/2003GB002044. 
 
 
 145 
 
Saleska, S., S. Miller, D. Matross, M. Goulden, S. Wofsy, H. da Rocha, P. de 
Camargo, P. Crill, B. Daube, H. de Freitas, L. Hutyra, M. Keller, V. Kirchhoff, 
M. Menton, J. Munger, E. Pyle, A. Rice, and H. Silva, 2003: Carbon in Amazon 
forests: Unexpected seasonal fluxes and disturbance-induced losses. Science, 
302, 1554?1557.  
Saleska, S., K. Didan, A. Huete, H. da Rocha, 2007: Amazon Forests green-up during 
2005 drought. Science, doi: 10.1126/science.1146663. 
Schaphoff, S., W. Lucht, D. Gerten, S. Sitch, W. Cramer, and I. C. Prentice, 2006: 
Terrestrial biosphere carbon storage under alternative climate projections. Clim. 
Change, 74, doi:10.1007/s10584-005-9002-5. 
Schimel, D. S., 1995: Terrestrial ecosystems and the carbon cycle. Global Change 
Biol., 1, 77?91. 
Schlesinger, W., 1991: Biogeochemistry: An Analysis of Global Change. Academic 
Press, San Diego. 
Schwartz, M. D., R. Ahas, and A. Aasa, 2006: Onset of spring starting earlier across 
the Northern Hemisphere. Global Change Biol., 12, 343?351. 
Sellers, P. J., 1991: Modeling and observing land?surface?atmosphere interactions on 
large scales. In: Wood, E.F. (Ed.), Land?Surface?Atmosphere Interactions for 
Climate Modeling: Observations, Models and Analysis. Kluwer Academic 
Publishers, Dordrecht. 
Serreze, M. C., J. E. Walsh, F. S. Chapin III, T. Osterkamp, M. Dyurgerov, V. 
Romanovsky, W. C. Oechel, J. Morison, T. Zhang and R. G. Barry, 2000: 
Observational evidence of recent change in the northern high-latitude 
environment. Clim. Change, 46, 159-207. 
Siegenthaler, U., T. F. Stocker, E. Monnin, D. Luthi, J. Schwander, B. Stauffer, D, 
Raynaud, J.M. Barnola, H. Fischer, V. Masson-Delmotte, and J. Jouzel, 2005: 
Stable carbon cycle-climate relationship during the late Pleistocene. Science, 
310, 1313?1317. 
Sitch, S., A. D. McGuire, J. Kimball, N. Gedney, J. Gamon, R. Engstrom, A. Wolf, 
Q. Zhuang, J. Clein, and K. McDonald, 2007: Assessing the carbon balance of 
 
 
 146 
 
circumpolar arctic tundra using remote sensing and process modeling. 
Ecological Application, 17, 213?234. 
Spahni, R., J. Chappellaz, T. F. Stocker, L. Loulergue, G. Hausammann, K. 
Kawamura, J. Fl?ckiger, J. Schwander, D. Raynaud, V. Masson-Delmotte, and J. 
Jouzel, 2005: Atmospheric methane and nitrous oxide of the late Pleistocene 
from Antarctic ice cores. Science, 310, 1317-1321. 
Stott, P. A., D. A. Stone, and M. R. Allen, 2004: Human contribution to the European 
heatwave of 2003. Nature, 432, 610? 614. 
Thompson, M. V., J. T. Randerson, C. M. Malmstr?m, and C. B. Field, 1996: Change 
in net primary production and heterotrophic respiration: How much is necessary 
to sustain the terrestrial carbon sink? Global Biogeochem. Cycles, 10(4), 711?
726. 
Thompson, S. L., B. Govindasamy, A. Mirin, K. Caldeira, C. Delire, J. Milovich, M. 
Wickett, and D. Erickson, 2004: Quantifying the effects of CO
2
-fertilized 
vegetation on future global climate and carbon dynamics. Geophys. Res. Lett., 
31, L23211, doi:10.1029/2004GL021239. 
Tian, H. Q., J. M. Melillo, D. W. Kicklighter, A. D. McGuire, J. V. K. Helfrich, B. 
Moore, and C. J. Vorosmarty, 1998: Effect of interannual climate variability on 
carbon storage in Amazonian ecosystems. Nature, 396, 664?667. 
Trenberth, K. E., 1997: The definition of El Ni?o. Bull. Am. Meteorol. Soc., 78, 
2771?2777. 
Trenberth, K. E., Branstator, G. W., Karoly, D., Kumar, A., Lau, N. C. and C. 
Ropelewski, 1998: Progress during TOGA in understanding and modeling 
global teleconnections associated with tropical sea surface temperatures. J. 
Geophys. Res., 103, 14291?14324. 
Tucker, C. J., D. A. Slayback, J. E. Pinzon, S. O. Los, R. B. Myneni, and M. G. 
Taylor, 2001: Higher northern latitude normalized difference vegetation index 
and growing season trends from 1982 to 1999. Int. J. Biometeorol., 45, 184?
190. 
Tucker, C. J., J. E. Pinzon, M. E. Brown, D. Slayback, E. W. Pak, R. V. E. Mahoney, 
and N. El Saleous, 2005: An extended AVHRR 8-km NDVI data set compatible 
 
 
 147 
 
with MODIS and SPOT vegetation NDVI data. Int. J. Remote Sens., 26(20), 
4485?4498. 
Van der Werf, G. R., J. T. Randerson, G. J. Collatz, L. Giglio, P. S. Kasibhatla, A. F. 
Arellano, S. C. Olsen, and E. S. Kasischke, 2004: Continental-scale partitioning 
of fire emissions during the 1997 to 2001 El Ni?o/La Ni?a period. Science, 303, 
73?76. 
Waple, A. M., J. H. Lawrimoreb, M. S. Halpertc, G. D. Bellc, W. Higginsc, B. Lyond, 
M. J. Mennee, K. L. Gleasone, R. C. Schnellf, J. R. Christyg, W. Thiawh, W. J. 
Wrighti, M. J. Salingerj, L. Alexanderk, R. S. Stonel, and S. J. Camargom, 2002: 
Climate assessment for 2001. Bull. Am.Meteorol. Soc., 83, S1? S62. 
Wardley, N. W., and P. J.Curran, 1984: The estimation of green leaf area index from 
remotely sensed airborne multiespectral scanner data. Int. J. Remote Sens.,, 5(4), 
671-679. 
Wetherald, R. T., and S. Manabe, 1995: The mechanisms of summer dryness induced 
by greenhouse warming. J. Clim., 12, 3096? 3108. 
Winguth, A. M. E., M. Heimann, K. D. Kurz, E. Maier-Reimer, U. Mikolajewicz, and 
J. Segschneider, 1994: El Ni?o?Southern Oscillation related fluctuations of the 
marine carbon cycle. Global Biogeochem. Cycles, 8, 39?64. 
Wittenberg, U., M. Heimann, G. Esser, A. McGuire, and W. Sauf, 1998: On the 
influence of biomass burning on the seasonal CO
2
 signal as observed at 
monitoring stations. Global Biogeochem. Cycles, 12(3), 531?544.  
Wolter, K., and M. S. Timlin, 1998: Measuring the strength of ENSO events-how 
does 1997/98 rank? Weather, 53, 315-324. 
Xiao, X., Q. Zhang, S. Saleska, L. Hutyra, P. de Camargo, S. Wofsy, S. Frolking, S. 
Boles, M. Keller, and B. Moore, 2005: Satellite-based modeling of gross 
primary production in a seasonally moist tropical evergreen forest. Remote Sens. 
Environ., 94, 105?122 
Xie, P., and P. Arkin, 1996: Analyses of global monthly precipitation using gauge 
observations, satellite estimates, and numerical model. J. Climate, 9, 840?858. 
 
 
 148 
 
Zeng, N., J. Neelin, K. Lau, and C. Tucker, 1999: Enhancement of interdecadal 
climate variability in the Sahel by vegetation interaction. Science, 286, 1537?
1540. 
Zeng, N., J. D. Neelin, and C. Chou, 2000: A quasi-equilibrium tropical circulation 
model: Implementation and simulation. J. Atmos. Sci., 57, 1767? 1796. 
Zeng, N., 2003: Glacial-interglacial atmospheric CO
2
 change-The glacial burial 
hypothesis. Adv. Atmos. Sci., 20, 677?693. 
Zeng, N., H. Qian, E. Munoz, and R. Iacono, 2004: How strong is carbon cycle-
climate feedback under global warming?. Geophys. Res. Lett., 31, L20203, doi: 
10.1029/2004GL020904. 
Zeng, N., A. Mariotti, and P. Wetzel, 2005a: Terrestrial mechanisms of interannual 
CO
2
 variability. Global Biogeochem. Cycles, 19, GB1016, doi:10.1029/2004GB 
002273. 
Zeng, N., H. Qian, C. R?denbeck, and M. Heimann, 2005b: Impact of 1998-2002 
Midlatitude drought and warming on terrestrial ecosystem and the global carbon 
cycle. Geophys. Res. Lett., 32(22),L22709, doi:10.1029/2005GL024607. 
Zeng, N., J-H Yoon, A. Mariotti, and S. Swenson, 2008: Long-term soil moisture 
variability from a new P-E water budget method. J. Climate, 15, 248-265. 
Zhou, L., C. J. Tucker, R. B. Myneni, R. K. Kaufmann, D. Slayback, N. V. Shabanov, 
and R. B. Myneni, 2001: Variations in northern vegetation activity inferred 
from satellite data of vegetation index during 1981 to 1999. J. Geophys. Res., 
106, 20069?20083. 
Zhuang, Q., V. E. Romanovsky, and A. D. McGuire, 2001: Incorporation of a 
permafrost model into a large-scale ecosystem model: Evaluation of temporal 
and spatial scaling issues in simulating soil thermal dynamics. J. Geophys. Res., 
106, 33,649?33,670.