1. Introduction
Deforestation remains at the forefront of climate policy discussions, with the 2010 COP16/CMP61 negotiations in Cancun coming to an agreement on approaches and incentives for reducing emissions from deforestation and forest degradation (REDD), conservation, sustainable management of forests, and enhancement of forest carbon stocks in developing countries (collectively referred to as REDD+) (Section III.C, Decision 1/CP.16, Outcome of the Ad Hoc Working Group on Long-term Cooperative Action; the texts of all COP/CMP decisions are available online at http://unfccc.int). The following COP17 in Durban further tackled policy approaches and incentives (Section II.C, Decision 1/CP.17) and guidance for providing information on safeguards and reference levels (Decision 12/CP.17) for REDD+. Thus, research into the impacts of deforestation pathways remains timely and relevant.
While REDD+ focuses only on carbon impacts, current research has established the importance of both carbon and noncarbon or biophysical mechanisms through which tropical deforestation in particular affects climate (such as through changes in albedo and latent heat and sensible heat fluxes) (e.g., Anderson et al. 2011; Davin and de Noblet-Ducoudré 2010; Coe et al. 2009; Bonan 2008; Nepstad et al. 2008; Bala et al. 2007; Gibbard et al. 2005; Snyder et al. 2004; Claussen et al. 2001; Zhang et al. 2001, among others). Gaps in the literature still exist, however, as most modeling studies prescribe the land cover or surface parameters without allowing for interactive vegetation, biophysics, biogeochemistry, and climate (e.g., Lawrence and Chase 2010; Davin and de Noblet-Ducoudré 2010; Findell et al. 2006; Avissar and Werth 2005; Gibbard et al. 2005; Cramer et al. 2004; Snyder et al. 2004; Claussen et al. 2001) or implement deforestation as a mass replacement of trees rather than as a gradual process (e.g., Bala et al. 2007; Findell et al. 2006; Avissar and Werth 2005; Gibbard et al. 2005; Zhang et al. 2001; Costa and Foley 2000).
The simulation of interactive incremental deforestation and the characterization of both transient and equilibrium impacts are still in an exploratory stage. Preliminary work by Gotangco Castillo and Gurney (2012) simplified “deforestation pathways” to be defined by the rate of tree cover loss. Using the Community Climate System Model, version 3, (CCSM3) with the Dynamic Global Vegetation Model (DGVM), Gotangco Castillo and Gurney showed that it is indeed possible to simulate incremental deforestation using a dynamic vegetation model coupled to a climate system model, and to differentiate the biophysical and climate feedbacks of low versus high deforestation rates. The transient “sensitivity” to deforestation was defined as the change in biophysical or climate variables per million square kilometers (Mkm2) of change in tree cover [similar to the metric used by Claussen et al. (2001) for their equilibrium simulation]. However, the warm, dry bias over the Amazon in the model (Lawrence et al. 2007; Bonan and Levis 2006) precluded definitive conclusions regarding sensitivity, its relative strengths over the forests of the Amazon basin, central Africa, and Southeast Asia, and the impact of rate of change. Furthermore, the carbon impacts were not included.
This work expands on the preliminary experiment with CCSM3-DGVM (Gotangco Castillo and Gurney 2012) using instead the newer Community Climate System Model, version 4, (CCSM4) with the prognostic carbon–nitrogen model with dynamic vegetation (CNDV). CNDV has been evaluated in both the offline and fully coupled mode to show that it simulates a reasonable plant functional type (PFT) distribution compared to Moderate Resolution Imaging Spectroradiometer (MODIS)-derived cover (Gotangco Castillo et al. 2012). CCSM4 contains many notable improvements over CCSM3 that have resulted in improved soil moisture dynamics and overall hydroclimate that remedy known biases in CCSM3 [details of improvements can be found in Lawrence et al. (2011), Kluzek (2010), and Oleson et al. (2010)].
The first objective is to conduct sensitivity analyses of biophysical, carbon, and climate impacts in CCSM4-CNDV to tropical deforestation pathways. CCSM4-CNDV is used to simulate a range of deforestation pathways, as defined by the rate of tree cover loss, over the pan-tropics. This work conducts sensitivity analyses on two levels—first, changes in biophysical, carbon, and climate variables caused by a per unit area change in tree cover are quantified compared to a control scenario (Control). Second, these quantities are compared among the different tropical deforestation pathways to determine if the rate of tree cover loss affects the magnitude and direction of these changes. It is hypothesized that faster rates would result in greater sensitivities since these would entail a bigger perturbation to the CCSM4-CNDV system each year compared to a slower rate. Comparisons are made between the pan-tropical and subcontinental (in forested areas of the Amazon basin, central Africa, and Southeast Asia) responses.
In addition, because the methodology (described in the succeeding section) involves a two-step process that isolates the biophysical feedbacks before combining them with carbon flux impacts, the second objective of this study is to assess the relative magnitude of the biophysical impacts on climate compared to the combined biophysical and carbon feedbacks of pan-tropical deforestation.
2. Methods
a. Simulations
CCSM4 was utilized here. As with previous versions, CCSM4 consists of a coupler (CPL7) linking atmosphere, land, ocean, and sea ice components—these are the Community Atmosphere Model 4 (CAM4), Community Land Model 4 (CLM4; Lawrence et al. 2011), Parallel Ocean Program 2 (POP2), and Community Ice Code 4 (CICE4; Gent et al. 2011). In addition, CLM4 is extended with the option to activate the prognostic carbon–nitrogen model (Thornton et al. 2007) with dynamic vegetation (CNDV) (Gotangco Castillo et al. 2012). Simulations performed in this study used the standard fully coupled CCSM4 with CNDV.
The program code previously introduced into the DGVM in CCSM3 to implement deforestation as described in Gotangco Castillo and Gurney (2012) was adapted for CNDV with a few modifications. Deforestation was implemented uniformly over the entire pan-tropical band (30°S–30°N) by removing a percentage (based on the deforestation rate) of the population of each tree plant functional type in a grid cell annually. This results in the incremental reduction of tree foliar cover: a “deforestation curve” (DFC). In CLM4-CNDV, carbon pools of the trees removed (e.g., leaf, stem, wood, roots) were moved to appropriate litter or debris pools or to product pools [for more information, see Bonan and Levis (2010) and Oleson et al. (2010)]. Grasses, as the closest approximation for crops, were allowed to colonize the areas left by trees if climatic conditions were appropriate. Tree regrowth and establishment, however, was not allowed in the pan-tropics to represent permanent or irreversible conversion (e.g., for agriculture).
A forest preservation target (PT) of 10% of the vegetated area was imposed on each tree PFT’s cover. For example, if the tropical broadleaf evergreen originally occupied 40% of the vegetated land area of a grid cell, it would have been deforested until it occupied 10%. Tree PFTs in tropical grid cells are dominated by two types, tropical broadleaf evergreens and tropical deciduous trees, making the total preservation target typically 20% of the vegetated area. This was the same as was done in CCSM3-DGVM, but with an additional condition that if the prescribed deforestation will cause the tree PFT’s cover to fall below 10%, then only as many individuals as needed to meet the 10% requirement were removed.
Four deforestation pathways were defined using four annual rates [based on FAO (2010) data] paired with the 10% target: 0.5% (DFC0.5-PT10), 1% (DFC1-PT10), 2% (DFC2-PT10), and 5% (DFC5-PT10). All the simulations were performed at a 1.9° × 2.5° resolution.
A 100-yr CCSM4-CNDV equilibrium modern-day (year 2000) spinup continuing from the end of a multicentury spinup [described in Gotangco Castillo et al. (2012)] was generated. From there, two sets of five simulations (a control and the four deforestation pathways) were performed. In the first set, only the biophysical feedbacks were interactive (Biophysics Only). Although the land model calculated carbon pools and fluxes, the atmospheric CO2 concentration remained fixed at year 2000 levels (368.9 ppmv). The other greenhouse gases were held at default year 2000 values, which are 316.0 ppbv for N2O, 1760 ppbv for CH4, 0.65345 ppbv for chlorofluorocarbon 11 (CFC11), and 0.5350 ppbv for CFC12.
The net ecosystem exchange (NEE) derived from the five simulations in this first set was then used as a basis for creating time series of varying atmospheric CO2. The global NEE time series were converted into a time series of annual CO2 fluxes to the atmosphere. Assuming the saturation of ocean sinks and any remaining land sinks not subject to deforestation (i.e., zero sinks), the accumulating CO2 was then converted directly into equivalent atmospheric concentration (in ppmv). The land–atmosphere flux of each year contributed to the calculated CO2 concentration at the start of the following year in the time series. No other sources of anthropogenic greenhouse gas emissions are included other than the pan-tropical deforestation.
The second set of simulations then incorporated both the biophysics and carbon impacts (Biophysics+Carbon) of pan-tropical deforestation. The simulations (including the control) were repeated but with CAM reading in the prepared atmospheric CO2 dataset and passing these values to CLM. CAM interpolated the monthly CO2 concentrations between the time series of annual values. Currently, CCSM4 does not have the capability to interactively adjust atmospheric CO2 concentrations based on the carbon flux from the land; hence, this method approximated the combined biophysical and carbon feedbacks of time-dependent deforestation. As in the Biophysics Only set, the other greenhouse gases were kept constant at their year 2000 values.
All in all, a total of 11 simulations were performed: one fully coupled spinup, a control and four deforestation pathways in the Biophysics Only set, and another control and four deforestation pathways in the Biophysics+Carbon set.
b. Analysis
The analyses focused on changes in selected biophysical, carbon, and climate variables during both active and postdeforestation periods. These variables are averaged over the pan-tropics as a whole, but to supplement the analysis they are averaged over major forested lands in the tropics as well, namely the Amazon basin (10°S–0°, 70°–50°W), central African (5°S–5°N, 10°–30°E) and Southeast Asian (10°S–10°N, 90°–150°E) subcontinents (refer to the map in Fig. 1, which is based on the standard map of regions of the diagnostic analysis package of the CCSM4 Land Model Working Group). This was to see how local variations in the response to the same applied forcing compare.
Subcontinental areas of analysis, from left to right: the Amazon basin, central Africa, and Southeast Asia.
Citation: Journal of Climate 26, 3; 10.1175/JCLI-D-11-00382.1
Subcontinental areas of analysis, from left to right: the Amazon basin, central Africa, and Southeast Asia.
Citation: Journal of Climate 26, 3; 10.1175/JCLI-D-11-00382.1
Subcontinental areas of analysis, from left to right: the Amazon basin, central Africa, and Southeast Asia.
Citation: Journal of Climate 26, 3; 10.1175/JCLI-D-11-00382.1
The total tropical tree PFT cover (tropical broadleaf evergreens and tropical deciduous trees) was plotted in time and the “transition point” (20% of the original cover) when the majority of grid cells were reaching the preservation target and ceasing forced deforestation was identified. This was done for the four spatial domains under consideration for all simulations. Two sets of analyses were then performed.
Active deforestation period: The data from the years prior to the transition were used to analyze sensitivity in the period of active deforestation. Because the perturbation in this study was the tree-to-grassland conversion, selected variables (albedo, net radiation, surface fluxes, total ecosystem carbon, ground temperature, and precipitation) were correlated against total tropical tree PFT cover, rather than time, for each deforestation rate and each spatial domain. Correlating against tree cover area also normalized or compensated for the slight differences in initial tree cover among the subcontinents. Sensitivity to tree cover was calculated by applying a linear regression and expressed in terms of the change in the variables per million square kilometers of tree cover removed. The Pearson correlation coefficients were evaluated using a 95% confidence level for statistical significance. As a supplement, for those variables that had significant trends in all cases, a test for homogeneity of regressions (based on the analysis of variance) was performed.
Postdeforestation period: The 30 years following the 20% mark in the Biophysics+Carbon set were used to characterize climatologies for each deforestation rate during the period immediately following the forced tree-to-grass conversion. For comparability, the same years were used to define the 30-yr postdeforestation climatologies of the Biophysics Only set. While the primary objective was to analyze the long-term climates resulting purely from different tropical deforestation pathways, a secondary objective was to isolate the Biophysics Only impact for a given time period and compare it against the combined Biophysics+Carbon impact. Within each set, the ground temperature and precipitation means of the deforestation cases were compared against the means of the same period of their respective Control simulations using a t test with a significance level of 95%. As a supplement, standard deviations from the means were also determined, and for those variables which were significantly different from the Control case across all deforestation cases, a test of equality of means was also performed [using the analysis of variance (ANOVA)]. Note that ground temperature here refers to surface soil temperature. (Supplemental Material S2 contains analyses of vegetation or leaf temperature and 2-m air temperature as well.)
The magnitudes of biophysical versus combined biophysical and carbon impacts on climate during a specific time period and in each pathway were then compared by looking at their respective temperature and precipitation differences from their Control cases. This differencing performed in both the analyses of active and postdeforestation periods served to remove background trends.
3. Results
a. Land cover change
Figure 2 plots the distribution of the PFTs averaged over the last decade of the fully coupled 100-yr spinup. The results are in accordance with the PFT distributions in the initial evaluation of the performance of CNDV (Gotangco Castillo et al. 2012). The Amazon basin, central Africa, and Southeast Asia contain predominantly tropical broadleaf evergreen and deciduous trees with an approximately equal mix. As deforestation progress, tree foliar cover averaged over the pan-tropics, as well as over the three major subcontinental forested areas, follows a decay curve until the preservation target is reached, as seen in Fig. 3, a sample from the Biophysics Only case. The period until the 20% mark is the active deforestation period (summarized in Table 1), and the biophysical, carbon, and climate results prior to this point are used to calculate the sensitivity to changes in tree foliar cover.
Mean foliar projective cover of plant functional types, expressed in percentage of grid cell vegetated area, averaged over the last 10 years of the 100-yr CCSM4-CNDV year 2000 equilibrium spinup.
Citation: Journal of Climate 26, 3; 10.1175/JCLI-D-11-00382.1
Mean foliar projective cover of plant functional types, expressed in percentage of grid cell vegetated area, averaged over the last 10 years of the 100-yr CCSM4-CNDV year 2000 equilibrium spinup.
Citation: Journal of Climate 26, 3; 10.1175/JCLI-D-11-00382.1
Mean foliar projective cover of plant functional types, expressed in percentage of grid cell vegetated area, averaged over the last 10 years of the 100-yr CCSM4-CNDV year 2000 equilibrium spinup.
Citation: Journal of Climate 26, 3; 10.1175/JCLI-D-11-00382.1
Decline in tropical tree PFT (tropical broadleaf evergreens and tropical deciduous trees) foliar projective cover (Mkm2) over tropical land in the Biophysics Only set after the year 2000 spinup. The horizontal dashed bar marks the approximated transition point (20% of original cover) from grid cells being dominantly in active deforestation to dominantly at or below the preservation target.
Citation: Journal of Climate 26, 3; 10.1175/JCLI-D-11-00382.1
Decline in tropical tree PFT (tropical broadleaf evergreens and tropical deciduous trees) foliar projective cover (Mkm2) over tropical land in the Biophysics Only set after the year 2000 spinup. The horizontal dashed bar marks the approximated transition point (20% of original cover) from grid cells being dominantly in active deforestation to dominantly at or below the preservation target.
Citation: Journal of Climate 26, 3; 10.1175/JCLI-D-11-00382.1
Decline in tropical tree PFT (tropical broadleaf evergreens and tropical deciduous trees) foliar projective cover (Mkm2) over tropical land in the Biophysics Only set after the year 2000 spinup. The horizontal dashed bar marks the approximated transition point (20% of original cover) from grid cells being dominantly in active deforestation to dominantly at or below the preservation target.
Citation: Journal of Climate 26, 3; 10.1175/JCLI-D-11-00382.1
Number of years of active deforestation for each spatial domain (as defined in Fig. 1) under the Biophysics Only set and the Biophysics+Carbon set, as counted from the start of each deforestation case that branches from the end of the 100-yr fully coupled spinup.
Note that in addition to the prescribed rate of tree removal, tree cover is also affected by natural causes of mortality and feedbacks from the forced deforestation. Thus, the decrease from year to year does not correspond to what the deforestation rate alone may predict, and varies among subcontinents. The initial values of tree cover (the sum of the broadleaf evergreen and deciduous tropical trees) of the three areas are comparable (2.46, 2.64, and 2.44 Mkm2 in the Amazon basin, central Africa, and Southeast Asia, respectively). However, the Amazon loses tree cover the fastest and reaches the transition mark the soonest, despite having the same perturbation applied as the other two areas. Southeast Asia is the slowest, with more differentiation across rates. This is true for all deforestation cases in both the Biophysics Only and the Biophysics+Carbon simulations.
Note also that in all cases the tree cover could not be sustained at the preservation target, although the rate of decrease is slower than during the active deforestation period. This is due to the control of tropical tree regrowth and establishment combined with other natural causes of mortality, consistent with what was seen previously in Gotangco Castillo and Gurney (2012).
b. NEE and the atmospheric CO2 time series
Global NEE in terms of 1 petagram of carbon (PgC) yr−1 (positive for source) is extracted from the output of the Control and deforestation simulations in the Biophysics Only set. Figure 4 shows (top) the time series of global NEE and (bottom) the rising atmospheric CO2 concentrations for each case. NEE increases initially as the bulk of trees are removed but declines as the fraction of trees removed also declines and the preservation target is reached. This results in annual atmospheric CO2 concentrations that rapidly climb initially then gradually level off as the NEE decreases. This trend is due to a combination of changes in gross primary production and heterotrophic respiration. (See Figs. S1.1–S1.3 in Supplemental Material S1 for additional plots of carbon fluxes and pools.) Gross and net primary productivity initially decreases rapidly as trees are removed but recovers partially as grasses spread. Conversely, heterotrophic respiration increases sharply but eventually decreases to below Control levels, with carbon being lost mainly from the wood and stem carbon pools (see Figs. S1.4 and S1.5 in Supplemental Material S1). This results in an overall decrease in total ecosystem carbon in the pan-tropics, in accordance with the loss of tree cover, and a net flux of carbon to the atmosphere.
(top) Global net ecosystem exchange (NEE) from the Biophysics Only set (positive for source) after year 2000. (bottom) Corresponding rise in atmospheric CO2 concentrations. Dashed horizontal bar indicates when 2 × CO2 concentrations are surpassed.
Citation: Journal of Climate 26, 3; 10.1175/JCLI-D-11-00382.1
(top) Global net ecosystem exchange (NEE) from the Biophysics Only set (positive for source) after year 2000. (bottom) Corresponding rise in atmospheric CO2 concentrations. Dashed horizontal bar indicates when 2 × CO2 concentrations are surpassed.
Citation: Journal of Climate 26, 3; 10.1175/JCLI-D-11-00382.1
(top) Global net ecosystem exchange (NEE) from the Biophysics Only set (positive for source) after year 2000. (bottom) Corresponding rise in atmospheric CO2 concentrations. Dashed horizontal bar indicates when 2 × CO2 concentrations are surpassed.
Citation: Journal of Climate 26, 3; 10.1175/JCLI-D-11-00382.1
Because the land and atmosphere remain coupled in the Biophysics+Carbon simulations, NEE is actually recalculated each year as both the deforestation proceeds and the atmospheric CO2 climbs. In this set, the increased productivity resulting from a carbon fertilization effect lowers the NEE slightly (not shown). However, the NEE is still a net source to the atmosphere, which indicates that the carbon fertilization effect cannot compensate for deforestation even at low deforestation rates. As a result, the atmospheric CO2 concentrations still rise in Biophysics+Carbon in the same manner as in Biophysics Only.
c. Active deforestation period: Sensitivity to tropical tree foliar cover
1) Biophysics Only
In the Biophysics Only simulations, the general result is that sensitivities do not vary widely or consistently across the deforestation rates within a subcontinent. For most of the variables, where the trends are significant across the different rates, these are too close to each other to make any practical difference. The results from the supplementary statistical test of homogeneity of regressions also do not reveal any coherent patterns among the trends across the four deforestation cases.
However, even with the normalization into per unit area quantities, sensitivities do vary more across the three specific subcontinental areas—the Amazon basin, central Africa, and Southeast Asia—than across rates. Not surprisingly, trends are stronger when averaged over these areas rather than over the pan-tropics as a whole. (Complete tables of trends and correlations for the different variables are shown in Supplemental Material S2, while selected variables are presented in Table 2 to illustrate these results.)
Sensitivities of selected variables in the Biophysics Only set, averaged over the respective deforestation periods of each spatial domain (as defined in Fig. 1). Variables are regressed against tropical tree cover in Mkm2 to obtain the change in each variable per Mkm2 change in tree cover (note that a negative sign denotes an inverse relationship). Numbers in parentheses are the Pearson coefficients. Numbers in italics indicate trends not statistically significant at a 95% confidence level.
Albedo, for example, exhibits a first-order response and is, as expected, negatively correlated with forest cover (i.e., albedo increases when tree cover decreases). Albedo over the Amazon basin shows the strongest sensitivity with a change of −1% of reflected radiation per million square kilometers of tree cover compared to approximately −0.5% in central Africa, −0.3% in Southeast Asia, and −0.07% over the pan-tropics. Across the four different rates, however, the range of the sensitivity of albedo in the Amazon basin is only −0.06% Mkm−2 of tree cover. This range is even smaller in central Africa and Southeast Asia (both approximately −0.03% Mkm−2 of tree cover).
Trends in moisture fluxes are also consistent with theory: where trends are significant, the correlations of tree cover with latent heat and canopy evapotranspiration are positive, and the correlations with ground evaporation and sensible heat flux are negative. Taking latent heat flux, for example, the Amazon again exhibits the highest sensitivity. Latent heat decreases by approximately 5.0 W m−2 Mkm−2 of tree cover removed, except in DFC5-PT10 where the sensitivity is higher at 7.1 W m−2 Mkm−2. Whether this increase in sensitivity is due to the rate cannot be ascertained here because the other three deforestation pathways are consistent with each other and no higher rate was tested.
In contrast, the sensitivity of latent heat flux in central Africa ranges only from 2.10 to 2.56 W m−2 Mkm−2. In Southeast Asia, the range across rates is wider, from 2.43 to 3.95 W m−2 Mkm−2. Still, this is a smaller value compared to the differences in latent heat fluxes between these two areas and the Amazon basin within a given deforestation case.
In terms of total ecosystem carbon, Southeast Asia loses the most, more than double that of the Amazon basin at 80 PgC Mkm−2 in the former compared to 36 PgC Mkm−2 in the latter. The sensitivity in Africa is close to that of Southeast Asia at about 74–76 PgC Mkm−2. The changes in total ecosystem carbon are directly connected with NEE. NEE and heterotrophic respiration are also lowest in the Amazon compared to the other areas (see Supplemental Material S1). This may be related to reduced soil water content in upper soil levels in the Amazon basin (not shown; likely due to decrease in precipitation), which inhibits soil respiration and CO2 production (Wu et al. 2011; Cook and Orchard 2008; Davidson et al. 2000). While CNDV does not explicitly model microbial communities, decomposition proceeds at its potential (water- and temperature-limited) rates (Oleson et al. 2010) using a log relationship with water potential (Andrén and Paustian 1987; Orchard and Cook 1983).
With only biophysical feedbacks, sensitivities of ground temperature to the loss of tree cover are inconsistent except in the Amazon basin where ground temperatures respond with a 0.87- to 1.16-K increase per million square kilometers of tree cover removed. Precipitation is also weakly correlated to land cover across all the deforestation rates and spatial domains considered. The only significant trends are the positive correlations of DFC0.5-PT10 and DFC1-PT10 in the Amazon, which indicate a drying over land.
2) Biophysics+Carbon
Compared to the results in the Biophysics Only set, more significant sensitivities are seen when both biophysical and carbon feedbacks are present (again, complete tables are shown in Supplemental Material S2, while selected variables are presented in Table 3 to illustrate results). The ranges of sensitivities across rates are still narrower than the range of sensitivities across the three subcontinental areas.
Albedo sensitivity, which is still highest over the Amazon basin, is slightly affected by the increasing CO2. This is due to the CO2 fertilization effect on vegetation that darkens the surface and decreases the average surface albedo (as seen in Table 3 compared to Table 2). The differential impact of the increasing atmospheric CO2 is stronger in the areas that take longer to reach the preservation target. (Note from Table 1 that Southeast Asia and central Africa have comparable durations of active deforestation periods, especially for the faster rates, versus the Amazon, which transitions into postdeforestation a decade sooner for the faster rates and up to three decades sooner for the slower rates).
Latent heat flux is most sensitive in Southeast Asia (at 9–10 W m−2 Mkm−2) instead of the Amazon basin because of greater decreases in canopy transpiration compared to canopy and ground evaporation in response to the loss of tree cover (see Supplemental Material S2). However, the variability in moisture flux sensitivity seen over the Amazon basin in the Biophysics Only case and is also seen in the Biophysics+Carbon set (the range is 5–8 W m−2 Mkm−2). Again, it is unclear whether the heat fluxes over the Amazon are exhibiting some nonlinearity in response to deforestation rates. Still, these differences within subcontinents across rates are still smaller than differences across subcontinents given the same deforestation rate.
Again, total ecosystem carbon is still most sensitive in Southeast Asia, followed by central Africa and the Amazon basin. These trends are only slightly lower compared to the Biophysics Only set because the carbon fertilization effect is minimal. This carbon sink, though slightly increased, cannot offset the carbon source from the heterotrophic respiration.
The sensitivity of ground temperatures to tropical tree cover loss is significant in all deforestation pathways and across all subcontinents, even over pan-tropical land. The pan-tropics experiences an increase of 0.17 K in ground temperature per million square kilometers of tree cover lost. The Amazon basin is still the area with the highest ground temperature sensitivity at an increase of close to 2 K compared to approximately 1 K Mkm−2 of tree cover lost in central Africa and Asia. In addition, the Amazon is the only area with a positive correlation between annual precipitation and tree cover (i.e., precipitation decreases as tree cover decrease) although the magnitude of the drying varies across the rates (as low as 6.06 cm yr−1 lost per million square kilometers of tree cover lost in DFC2-PT10, and as much as 16.63 cm yr−1 lost per million square kilometers of tree cover lost in DFC1-PT10). In central Africa and Southeast Asia, annual precipitation is negatively correlated, which indicates that these areas are getting progressively wetter in response to the combined carbon feedback (i.e., the increased warming).
Of the three subcontinents, only over Southeast Asian land do the rates result in statistically different sensitivities in temperature and precipitation in the Biophysics+Carbon set, according to the test for homogeneity of regressions. This may be traced only to the nonhomogenous trends in the loss of ecosystem carbon because biophysical sensitivities still remain homogenous across rates.
d. Postdeforestation period: Climate contributions
First, results show that the 30-yr climatologies are similar (Tables 4 and 5) despite the differences in deforestation pathways. This is consistent with the lack of significant differences among the feedback sensitivities across the rates. The ground temperatures averaged over each subcontinent again fall within a narrow range that is of comparable magnitude to interannual variabilities, as denoted by the standard deviations (in parentheses in Tables 4 and 5). For example, averaged over the pan-tropics, ground temperatures for all of the deforestation cases in the Biophysics Only set are significantly different from their respective controls, but fall within 298.53–298.57 K.
Mean annual ground temperatures (K) during the respective postdeforestation climatologies of each deforestation case, and the differences (denoted by “Δ”) from respective Controls. Subcontinental spatial domains are as defined in Fig. 1. Numbers in parentheses represent standard deviations. Numbers in italics represent means not significantly different from Control at the 95% level.
Subcontinentally, the ranges are slightly wider across rates but are still comparable in magnitude to interannual variability. The same holds for the ground temperature and precipitation in the Biophysics+Carbon cases. There is no consistent or robust pattern in the climate differences caused by the rates across the different areas. Tests for equality of means likewise show that only the means over Southeast Asian land and global land are unequal, statistically, across the rates. As was also seen in the active deforestation period, greater differences exist across the Amazon basin, central Africa, and Southeast Asia within the same deforestation case. The warming over the Amazon basin is consistently the highest in both the Biophysics Only and Biophysics+Carbon cases. In terms of precipitation, the impacts move in opposite directions: the Amazon is the only area that experiences a drying in response to deforestation. Southeast Asia is the wettest of the three, followed by central Africa.
Another major result is the relatively large impact of biophysics compared to the combined biophysics and carbon impacts on long-term subcontinental climate, although the magnitude of the impact varies across the subcontinents. Tables 4 and 5 also tabulate the differences in temperature and precipitation between the deforestation cases and their respective control cases during their respective postdeforestation periods for the Biophysics Only and Biophysics+Carbon simulations. In Southeast Asia, the changes in ground temperature are significant across all deforestation pathways for both the Biophysics Only and Biophysics+Carbon sets. The biophysical impact is variable, ranging from 11% to as much as 21% of the combined biophysical and carbon impacts. [This is just for comparison, however, and not to suggest that biophysical and carbon feedbacks combine linearly; Zhang et al. (2001), for example, have already demonstrated the nonlinearity of results with simulations looking at the impacts of deforestation and doubled CO2 separately and jointly.] Southeast Asia is the only area in which we see this wide range in biophysical contribution to ground temperatures. The strength of the biophysical impact on precipitation is variable as well; it is insignificant and in the opposite direction (compared to the Biophysics+Carbon case) in the DFC0.5-PT10 simulation, but significant and strong in the DFC5-PT10 simulation at 37% of the Biophysics+Carbon change in precipitation over land.
In central Africa, again the biophysical impact is strongest on ground temperatures at 26%–29% of the combined biophysical and carbon impact. The range of the biophysical contribution to the increase in annual precipitation is wide, ranging from 34% to 53% (where significant).
In the Amazon basin, the biophysical impact on both temperature and precipitation is the strongest. With ground temperatures, it is as high as 55% of the increase in the combined Biophysics+Carbon scenario. The precipitation change is in the opposite direction in this area compared to the others (drying instead of getting wetter), but again the biophysical contribution is up to 61% of the drying in Biophysics+Carbon simulations. In addition, the fact that this area undergoes such significant changes despite experiencing lower atmospheric CO2 concentrations (as seen in Table 1, the Amazon basin enters into postdeforestation period sooner than the other areas) speaks to the importance of non-CO2 feedbacks of land cover change in determining subcontinental climate.
For reference, the transient climate response (TCR) to CO2 doubling of CCSM4 at a 1° resolution is estimated at 1.72°C, while the equilibrium climate sensitivity is 3.20°C (Bitz et al. 2012). The changes in temperature attained here in the Biophysics+Carbon set are well beyond the TCR. Note, however, that the atmospheric CO2 concentrations are not strictly capped here at 2×CO2 (which is 737.8 ppmv here) because they are calculated from the NEE. They increase slightly further to as high as 776.5 ppmv (in the last year included in the analysis of 30-yr means).
Lastly we note that the impact of tropical deforestation is detectable at a global scale when both the biophysical and carbon feedbacks are present. Tables 4 and 5 summarize the Biophysics Only and Biophysics+Carbon impacts over global land for the 30-yr periods during which the pan-tropics on average have moved past the transition period (since there is no forcing outside of the pan-tropics). The latter causes more than a 2-K increase in ground temperature and a 2.47–3.97-cm increase in precipitation. However, the Biophysics Only simulations also result in significant warming of ground and vegetation temperatures with the higher deforestation rates. The increases are 0.11 and 0.22 K for ground temperatures, in the DFC2-PT10 and the DFC5-PT10 cases, respectively. These represent very small fractions of the combined Biophysics+Carbon impacts (which are dominated by the CO2 forcing), but whether the higher deforestation rates lead to changes that can be detected on a larger scale is a hypothesis that can be further explored.
Figure 5, which is based on the DFC5-PT10 results, is presented here as representative of the Biophysics Only and Biophysics+Carbon sets. This maps the changes in ground temperatures and precipitation compared to respective Control scenarios. It is clear that the greatest changes are concentrated in the forests of the Amazon basin, central Africa, and Southeast Asia. Other deforestation cases reflect similar results in the pan-tropics across the different rates. Potential teleconnections in the extratropics will require ensemble modeling and further analysis.
Differences in annual mean (top) ground temperatures and (bottom) precipitation during the postdeforestation period (simulation years 136–165) for DFC5-PT10 vs Control in the (left) Biophysics Only and (right) Biophysics+Carbon sets. Areas with significant change (at a 95% confidence level) are stippled.
Citation: Journal of Climate 26, 3; 10.1175/JCLI-D-11-00382.1
Differences in annual mean (top) ground temperatures and (bottom) precipitation during the postdeforestation period (simulation years 136–165) for DFC5-PT10 vs Control in the (left) Biophysics Only and (right) Biophysics+Carbon sets. Areas with significant change (at a 95% confidence level) are stippled.
Citation: Journal of Climate 26, 3; 10.1175/JCLI-D-11-00382.1
Differences in annual mean (top) ground temperatures and (bottom) precipitation during the postdeforestation period (simulation years 136–165) for DFC5-PT10 vs Control in the (left) Biophysics Only and (right) Biophysics+Carbon sets. Areas with significant change (at a 95% confidence level) are stippled.
Citation: Journal of Climate 26, 3; 10.1175/JCLI-D-11-00382.1
4. Discussion
The sensitivity analysis of CCSM4-CNDV shows similar trends and means in biophysical, carbon, and climate variables regardless of the rate of conversion. The spatially averaged changes in these variables per million square kilometers of tree cover lost fall within a narrow range across the different deforestation pathways. Note that there are certain caveats to the results here. First, CCSM4-CNDV was not truly completely interactive in that atmospheric CO2 did not vary in response to NEE. This means that the carbon fertilization effect seen in the Biophysics+Carbon case did nothing to bring down the NEE and atmospheric CO2 compared to what was calculated offline. In calculating the atmospheric CO2 time series offline, it was also assumed that there were no other carbon sinks. The increase in atmospheric CO2 concentrations, and hence the carbon impacts, will be lower if an ocean sink and forest sinks in the temperate latitudes are considered. How the strength of these sinks, particularly the land sink, will evolve as deforestation progresses represents another uncertainty.
There are also other processes that may affect sensitivities to rates of deforestation but are not yet represented in the model. One example is the acclimation of plants to elevated CO2 levels. Slower rates may give vegetation more opportunity to acclimate. Another example is the change in speciation of forests and the plasticity of these species in adapting to a range of environmental conditions. In this experiment, for example, the loss of Amazonian forest is rapid compared to central Africa or Southeast Asia despite experiencing the same rate of tree cover loss. However, the forest dieback (Cox et al. 2004) may not actually occur as fast as is being predicted by here and by models because of this type of succession or transition to tree species with higher tolerance or plasticity (K. Sakaguchi 2011, personal communication). However, in CLM, the plant functional types represent very broad categories that do not capture these variances in species’ plasticity.
In addition, forest growth was controlled in the pan-tropics to represent permanent conversion, as in agricultural systems. However, mainly grasses colonized the areas formerly occupied by trees, and grasses differ from crops in at least two important ways: 1) they are unmanaged (agricultural areas would be more regulated in terms of fertilizer and water inputs) and 2) crops have more of a seasonal phenology. The conversion from forest to urban areas or roads would also result in different sensitivities. The albedo of urban surfaces can potentially be even higher than that of grasses or bare ground but the latent heat flux may be much lower (urban heat island effect).
Furthermore, it is also possible that the “threshold” in CCSM4-CNDV beyond which any nonlinear response becomes more apparent is not within the current “realistic” rates of deforestation. The linear behaviors seen here may be peculiar to this model, raising the question of whether CCSM4-CNDV is able to represent or capture abrupt climatic shifts (G. Narisma 2010, personal communication). The use of model ensembles will help verify whether the linear behavior seen here is a general trend or is particular to CCSM4-CNDV. To date, the time-dependent aspect of progressive, interactive tropical deforestation with dynamic vegetation has not been well investigated using other models, so we lack benchmarks for the sensitivities calculated during the active deforestation period.
In general, however, although this work employed transient simulations, the impacts seen here can be compared against results from other equilibrium modeling studies for the purpose of contextualization. Previous sensitivity studies (Bala et al. 2007; Claussen et al. 2001) have concluded that the carbon-induced warming due to tropical deforestation dominates over the biophysical impacts in contrast to high-latitude deforestation in which the cooling effect of albedo prevails (Bala et al. 2007; Claussen et al. 2001; Betts 2000; Brovkin et al. 1999; Bonan et al. 1992). The results of this study similarly show that the atmospheric CO2 forcing is stronger over larger spatial scales, causing statistically significant warming globally, with biophysical impacts still significantly strong locally. Finally, the net biophysical feedback is also that of warming, which indicates that the changes in latent and sensible heat fluxes at the surface outweigh the increase in albedo. This is consistent with studies that isolate the biophysical impacts of tropical deforestation (e.g., Davin and de Noublet-Ducoudré 2010; Jackson et al. 2008; Claussen et al. 2001).
This work also highlights the subcontinental responses. Comparing these results with existing literature, we find that over the Amazon basin ground temperature and precipitation changes agree with other modeling experiments. Deforestation in the Amazon basin positively reinforces the impacts of increased greenhouse gas emissions causing a rapid warming and exacerbated seasonal drying (Malhi et al. 2008; Nepstad et al. 2008). Zhang et al. (2001) and Costa and Foley (2000) both simulate the complete replacement of tropical forests by grasslands combined with the doubling of CO2 in the Amazon (note, however, that the spatial coverage in these studies is larger than what was used here as shown in Fig. 1). In the former, annual precipitation decreased by 31.2 cm yr−1 and surface temperature increased by 3 K; in the latter, the decrease in annual precipitation was estimated at only 15.3 cm yr−1, and the increase in temperature at roughly 3.5 K. Atmospheric CO2 was prescribed at the onset at 660 ppmv for Zhang et al. and 690 ppmv for Costa and Foley. In contrast, conversion of trees is not complete in this study, but higher concentrations of CO2 are attained. Taking these into account, the impacts seen here (an increase in temperature of approximately 4 K and a decrease in annual precipitation of roughly 30 cm in the Biophysics+Carbon set) are still reasonable.
Other experiments that also simulated complete deforestation but without the increase in CO2 show more modest increases in temperature but a wide range of estimates for changes in precipitation. These estimates agree on the direction of change (drying). Coe et al. (2009), for example, find precipitation reduced in all watersheds studied within the Amazon basin (which encompasses a greater study area than shown in Fig. 1 here). This reduction is as much as 20% in the Xungu watershed in a business-as-usual deforestation scenario compared to a control scenario without anthropogenic vegetation change. However, estimates of precipitation change in different studies vary on the magnitude of the reduction. Lean and Rowntree (1997) found seven studies in the 1990s to be in the range of a 0.1–3.8-K warming and a range of 1.5–64.0 cm yr−1 drying over the Amazon. The work of Zhang et al. (1996), Werth and Avissar (2002), and Findell et al. (2006) also estimates the decrease in precipitation to be within this range. Again, the results for the Amazon basin from the Biophysics Only set fall within these estimates.
Over the central African region defined in the Zhang et al. (2001) study, surface warming was estimated to be 2.5 K and the increase in annual precipitation to be only 2.5 cm in response to combined deforestation and doubled CO2. This agrees with the warming seen here in the Biophysics+Carbon case, but is well below the estimated 20–26 cm yr−1 increases in rainfall. Snyder et al. (2004) simulated complete deforestation in central Africa but without the increased CO2, finding the surface warming to range from 0.3 K in the western coastal region to 1.4 K slightly inland. This agrees with temperatures from the Biophysics Only set, although the tree removal here is not complete. The impacts on precipitation, however, differ from the Snyder et al. study, which saw a decrease rather than an increase.
As with central Africa, the temperature estimates over Southeast Asia agree with other work but the precipitation impacts differ in direction. Zhang et al.’s (2001) estimated surface warming of 2.1 K, due to both deforestation and increased CO2, agrees well with Southeast Asia results from the Biophysics+Carbon set (though the spatial domain of the former extends farther north, including Cambodia, Vietnam, Thailand, and Laos, than what is used here). However, precipitation decreases by 17.2 cm yr−1 in the former, in contrast to the increase on the order of 30 cm yr−1 seen here. In Zhang et al. (1996), deforestation with present-day CO2 also resulted in a decrease in precipitation, while no significant change is found here in the Biophysics Only set, except for the DFC5-PT10 case, which shows an increase. Werth and Avissar (2005) also modeled monthly decreases in precipitation in Southeast Asia but found it was weaker than those that occurred in Africa or the Amazon basin.
Southeast Asia is archipelagic, and hence the ocean response to deforestation would greatly affect climate. In addition, precipitation comes from the monsoon rather than by convection (Kanae et al. 2001; Meehl and Washington 1993) and thus would be affected more by greenhouse gas–induced warming rather by local changes in evapotranspiration. The West African monsoon likewise affects precipitation over equatorial forests, and warming is projected to increase precipitation around the Gulf of Guinea (Christensen et al. 2007; Meehl et al. 2007). Thus, the simple slab ocean models used by studies such as Zhang et al. (2001) do not adequately represent monsoon dynamics and may not be fully capturing the impacts of deforestation on precipitation. Further analyses of the atmospheric and ocean dynamics of the simulations here is required to ascertain the mechanisms linking deforestation to these changes in climate over central Africa and Southeast Asia, and to account for the contrasting impacts seen on precipitation.
Taking the larger body of existing literature into context, the biophysical, carbon, and climate impacts seen in CCSM4-CNDV are still largely consistent with theory and other model simulations. This lends confidence that CCSM4-CNDV is an adequate representation of the earth system, and will thus respond to perturbations in a reasonable manner, despite its possible limitations in capturing transient responses to abrupt changes.
5. Conclusions and recommendations
This work is primarily a sensitivity analyses of deforestation rates simulated within CCSM4-CNDV, a fully coupled climate model with interactive carbon–nitrogen cycles and dynamic vegetation. As such it is not the objective to project future changes with accuracy, but to study CCSM4-CNDV responses to different yearly amounts of tree removal, quantified in terms of the change in selected biophysical, carbon and climate variables per million square kilometers of change in tree cover. The following major insights can be drawn from the results:
The CCSM4-CNDV system response to the deforestation does not seem to be clearly dependent on the deforestation rate, at least for the range of rates tested here. Note, however, that only a “first approximation” of the combined impacts of biophysics and the carbon cycle is employed. Thus it is still possible that nonlinearities in the sensitivities to the rates may arise if the system was completely interactive and included other important but absent processes such as species acclimation and ecological succession. A clearer effect of the rates, as seen in Table 1, is to hasten or delay climate change. This aspect is what may be relevant for planning. Over the pan-tropics, the fastest rate, 5%, results in an active deforestation time 40–45 years shorter than the slowest rate, 0.5%, which also means accelerating the warming that much sooner. Even the 1% case already results in a 12-yr difference compared to the 0.5% case, which is within the time horizon of medium-term development planning. These timelines may be particular to this sensitivity study of CCSM4-CNDV, however, so further research is required to more accurately model current and projected deforestation trends and estimate planning horizons represented by different deforestation rates.
The biophysical, total ecosystem carbon, and climate trends and means during the active and postdeforestation periods vary more across subcontinents with the same deforestation rate rather than across rates within the same subcontinent. For example, the greatest overall temperature sensitivities to deforestation are detected over the Amazon basin in both the Biophysics Only and Biophysics+Carbon sets, followed by central Africa and then Southeast Asia. Over the postdeforestation period, the increases in ground temperatures over the Amazon basin, which are on the order of almost 3–4 K in most cases, are approximately twice those calculated over Southeast Asia when both biophysics and carbon impacts are considered. In terms of time dependency, subcontinental variations exist as well. The changes over the Amazon occur the most rapidly, followed by central Africa and then Southeast Asia.
Lastly, on a regional to subcontinental scale, the biophysical impacts of deforestation on climate are significant and comparable to the combined biophysics and carbon impacts. The maximum ratio for the increase in temperature is for ground temperature in the Amazon basin: the Biophysics Only impact is as much as 55% of the Biophysics+Carbon impact. For changes in precipitation, the maximum ratio is for the decrease also over the Amazon basin. The drying seen in the Biophysics Only set is as much as 61% of the Biophysics+Carbon set. However, biophysical and carbon impacts do not combine linearly. Carbon Only simulations using the input atmospheric CO2 time series but without any deforestation are required to ascertain how the impacts combine.
Because this type of interactive, progressive land cover change modeling is still in its initial stages, this work contributes to methodology development and to the evaluation of CCSM4-CNDV while exploring scientific questions. These scientific questions seek to address our understanding of biosphere–atmosphere sensitivities to deforestation, including transient impacts. The methods here should be replicated using other climate models to test if the lack of sensitivity to deforestation rates is robust, or if this is a feature of this particular model. Simulations that isolate specific areas of interest (the Amazon basin, the Congo or central Africa, and Southeast Asia) and using more accurate values for deforestation will also yield insights on the actual contribution of deforestation in each area to current and project climate. High-resolution downscaling can also offer new knowledge regarding more localized feedbacks. In addition, there are still many improvements to be implemented in the way incremental deforestation and the preservation targets were implemented in the code, particularly with regard to sustaining the preservation target.
Future modeling work may also explore a more expanded characterization of deforestation pathways that considers not just the temporal aspect but also the spatial structure of deforestation (e.g., continuous versus patchy versus herringbone) and the particular type of land cover change (e.g., tree to unmanaged grass versus tree to croplands versus tree to urban area). On-the-ground verification and observational data analyses would also complement modeling efforts. What is needed is a systematic and statistical analyses of historical deforestation trends vis-à-vis climate trends to determine if actual historical sensitivities are within range of those estimated by models.
Deforestation pathways can be thought of as analogous CO2 concentration pathways—in the same way that CO2 emissions can be “ramped up” in models until a prescribed limit is reached, deforestation can also be sped up or slowed down until preservation targets are reached. In addition, as seen in this study, deforestation also induces a “transient climate response.” The climate sensitivity to deforestation, as we have defined it in this study, is the change per unit loss in tree cover during the active deforestation period. If climate sensitivity to deforestation can be quantified, then this suggests the possibility of defining a “climate effectiveness” metric of reducing deforestation or reforestation that will encompass both biophysical and carbon impacts of the land cover change.
Acknowledgments
CCSM4 was provided by the National Center for Atmospheric Research (NCAR). NCAR is sponsored by the National Science Foundation (NSF). The NCAR Land Model Working Group (LMWG) also developed the CLM diagnostics package on which the regional definitions and map was based. This research used resources of the National Center for Computational Sciences (NCCS) at Oak Ridge National Laboratory (ORNL), which is supported by the Office of Science of the U.S. Department of Energy under Contract DE-AC05-00OR22725. Computing time was provided through the Climate-Science Computational End Station Development–University Collaborations project of the INCITE program. Funding support for the author was provided by the Purdue Climate Change Research Center (PCCRC), the Schlumberger Faculty for the Future Fellowship and the Fulbright program. Many thanks also to Matt Huber (Purdue University) for his helpful suggestions, and to Sam Levis (NCAR) who collaborated on the code modifications and on generating the CLM4-CNDV spinup used by the simulations here.
REFERENCES
Anderson, R. G., and Coauthors, 2011: Biophysical considerations in forestry for climate protection. Front. Ecol. Environ., 9, 174–182, doi:10.1890/090179.
Andrén, O., and K. Paustian, 1987: Barley straw decomposition in the field: A comparison of models. Ecology, 68, 1190–1200.
Avissar, R., and D. Werth, 2005: Global hydroclimatological teleconnections resulting from tropical deforestation. J. Hydrometeor., 6, 134–145.
Bala, G., K. Caldeira, M. Wickett, T. J. Phillips, D. B. Lobell, C. Delire, and A. Mirin, 2007: Combined climate and carbon-cycle effects of large-scale deforestation. Proc. Natl. Acad. Sci. USA, 104, 6550–6555.
Betts, R. A., 2000: Offset of the potential carbon sink from boreal forestation by decreases in surface albedo. Nature, 408, 187–190.
Bitz, C. M., K. M. Shell, P. R. Gent, D. Bailey, G. Danabasoglu, K. C. Armour, M. M. Holland, and J. T. Kiehl, 2012: Climate sensitivity of the Community Climate System model version 4. J. Climate, 25, 3053–3070.
Bonan, G. B., 2008: Forests and climate change: Forcings, feedbacks and the climate benefit of forests. Science, 320, 1444–1449.
Bonan, G. B., and S. Levis, 2006: Evaluating aspects of the Community Land and Atmosphere Models (CLM3 and CAM3) using a dynamic global vegetation model. J. Climate, 19, 2290–2301.
Bonan, G. B., and S. Levis, 2010: Quantifying carbon–nitrogen feedbacks in the Community Land Model (CLM4). Geophys. Res. Lett., 37, L07401, doi:10.1029/2010GL042430.
Bonan, G. B., D. Pollard, and S. L. Thompson, 1992: Effects of boreal forest vegetation on global climate. Nature, 359, 716–718.
Brovkin, V., A. Ganopolski, M. Claussen, C. Kubatzki, and V. Petoukhov, 1999: Modelling climate response to historical land cover change. Global Ecol. Biogeogr., 8, 509–517.
Christensen, J. H., and Coauthors, 2007: Regional climate projections. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 847–940.
Claussen, M., V. Brovkin, and A. Ganopolski, 2001: Biogeophysical versus biogeochemical feedbacks of large-scale land cover change. Geophys. Res. Lett., 28, 1011–1014.
Coe, M. T., M. H. Costa, and B. S. Soares-Filho, 2009: The influence of historical and potential future deforestation on the stream flow of the Amazon River—Land surface processes and atmospheric feedbacks. J. Hydrol., 369, 165–174.
Cook, F. J., and V. A. Orchard, 2008: Relationships between soil respiration and soil moisture. Soil Biol. Biochem., 40, 1013–1018.
Costa, M. H., and J. A. Foley, 2000: Combined effects of deforestation and doubled atmospheric CO2 concentrations on the climate of Amazonia. J. Climate, 13, 18–34.
Cox, P. M., R. A. Betts, M. Collins, P. P. Harries, C. Huntingford, and C. D. Jones, 2004: Amazonian forest dieback under climate–carbon cycle projections for the 21st century. Theor. Appl. Climatol., 78, 137–156.
Cramer, W., A. Bondeau, S. Schaphoff, W. Lucht, B. Smith, and S. Sitch, 2004: Tropical forests and the global carbon cycle: Impacts of atmospheric carbon dioxide, climate change and rate of deforestation. Philos. Trans. Roy. Soc. London, 359B, 331–343.
Davidson, E. A., L. V. Verchot, J. H. Cattânio, I. L. Ackerman, and J. E. M. Carvalho, 2000: Effects of soil water content on soil respiration in forests and cattle pastures of eastern Amazonia. Biogeochemistry, 48, 53–69.
Davin, E., and N. de Noblet-Ducoudré, 2010: Climatic impact of global-scale deforestation: Radiative versus nonradiative version. J. Climate, 23, 97–112.
FAO, 2010: Global forest resources assessment 2010. Food and Agriculture Organization Forestry Paper 163, 340 pp. [Available online at http://www.fao.org/forestry/fra/fra2010/en/.]
Findell, K. L., T. R. Knutson, and P. C. D. Milly, 2006: Weak simulated extratropical responses to complete tropical deforestation. J. Climate, 19, 2835–2850.
Gent, P. R., and Coauthors, 2011: The Community Climate System Model version 4. J. Climate, 24, 4973–4991.
Gibbard, S., K. Caldeira, G. Bala, T. J. Phillips, and M. Wickett, 2005: Climate effects of global land cover change. Geophys. Res. Lett., 32, L23705, doi:10.1029/2005GL024550.
Gotangco Castillo, C. K., and K. R. Gurney, 2012: Exploring surface biophysical-climate sensitivity to tropical deforestation rates using a GCM. Earth Interact., 16. [Available online at http://EarthInteractions.org.]
Gotangco Castillo, C. K., S. Levis, and P. Thornton, 2012: Evaluation of the new CNDV option of the Community Land Model: Effects of dynamic vegetation and interactive nitrogen on CLM4 means and variability. J. Climate, 25, 3702–3714.
Jackson, R. B., and Coauthors, 2008: Protecting climate with forests. Environ. Res. Lett., 3, 044006, doi:10.1088/1748-9326/3/4/044006.
Kanae, S., T. Oki, and K. Musiake, 2001: Impact of deforestation on regional precipitation over the Indochina Peninsula. J. Hydrometeor., 2, 51–70.
Kluzek, E., cited 2010: CCSM Research Tools: CLM4.0 Users Guide Documentation. [Available online at http://www.cesm.ucar.edu/models/ccsm4.0/clm/models/lnd/clm/doc/UsersGuide/book1.html.]
Lawrence, D. M., P. E. Thornton, K. W. Oleson, and G. B. Bonan, 2007: The partitioning of evapotranspiration into transpiration, soil evaporation, and canopy evaporation in a GCM: Impacts on land–atmosphere interaction. J. Hydrometeor., 8, 862–880.
Lawrence, D. M., and Coauthors, 2011: Parameterization improvements and functional and structural advances in version 4 of the Community Land Model. J. Adv. Model. Earth Syst.,3, M03001, doi:10.1029/2011MS000045.
Lawrence, P. J., and T. N. Chase, 2010: Investigating the climate impacts of global land cover change in the Community Climate System Model. Int. J. Climatol., 30, 2066–2087.
Lean, J., and P. R. Rowntree, 1997: Understanding the sensitivity of a GCM simulation of Amazonian deforestation to the specification of vegetation and soil characteristics. J. Climate, 10, 1216–1235.
Malhi, Y., J. T. Roberts, R. A. Betts, T. J. Killeen, W. Li, and C. A. Nobre, 2008: Climate change, deforestation and the fate of the Amazon. Science, 319, 169–172.
Meehl, G. A., and W. M. Washington, 1993: South Asian summer monsoon variability in a model with doubled atmospheric carbon dioxide concentration. Science, 260, 1101–1104.
Meehl, G. A., and Coauthors, 2007: Global climate projections. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 747–845.
Nepstad, D. C., C. M. Stickler, B. Soares-Filho, and F. Merry, 2008: Interactions among Amazon land use, forests and climate: Prospects for a near-term forest tipping point. Philos. Trans. Roy. Soc. London, 363B, 1737–1746.
Oleson, K. W., and Coauthors, 2010: Technical description of version 4.0 of the Community Land Model (CLM). NCAR Tech. Note NCAR/TN-478+STR, 257 pp.
Orchard, V. A., and F. J. Cook, 1983: Relationship between soil respiration and soil moisture. Soil Biol. Biochem., 15, 447–453.
Snyder, P. K., J. A. Foley, M. H. Hitchman, and C. Delire, 2004: Analyzing the effects of complete tropical forest removal on the regional climate using a detailed three-dimensional energy budget: An application to Africa. J. Geophys. Res., 109, D21102, doi:10.1029/2003JD004462.
Thornton, P. E., J. F. Lamarque, N. A. Rosenbloom, and N. M. Mahowald, 2007: Influence of carbon–nitrogen cycle coupling on land model response to CO2 fertilization and climate variability. Global Biogeochem. Cycles, 21, GB4018, doi:10.1029/2006GB002868.
Werth, D., and R. Avissar, 2002: The local and global effects of Amazon deforestation. J. Geophys. Res., 107, 8087, doi:10.1029/2001JD000717.
Werth, D., and R. Avissar, 2005: The local and global effects of Southeast Asian deforestation. Geophys. Res. Lett., 32, L12702, doi:10.1029/2005GL022970.
Wu, Z., P. Dijkstra, G. W. Koch, J. Peñuelas, and B. A. Hungate, 2011: Responses of terrestrial ecosystems to temperature and precipitation change: A meta-analysis of experimental manipulation. Global Change Biol., 17, 927–942.
Zhang, H., A. Henderson-Sellers, and K. McGuffie, 1996: Impacts of tropical deforestation. Part I: Process analysis of local climatic change. J. Climate, 9, 1497–1517.
Zhang, H., A. Henderson-Sellers, and K. McGuffie, 2001: The compounding effects of tropical deforestation and greenhouse warming on climate. Climatic Change, 49, 309–338.
The 16th session of the Conference of the Parties of the United Nations Framework Convention on Climate Change (COP) and the sixth session of the Conference of the Parties serving as the meeting of the Parties to the Kyoto Protocol (CMP).
