• Achard, F., , H. Eva, , A. Glinni, , P. Mayaux, , T. Richards, , and H. J. Stibig, 1998: Identification of deforestation hot spots areas in the humid tropics. TREES Publications Series B Research Rep. 4, 84 pp.

    • Search Google Scholar
    • Export Citation
  • Arakawa, A., , and R. V. Lamb, 1977: Computational design of the basic dynamical processes of the UCLA general circulation model. Methods in Computational Physics, Vol. 17, Academic Press, 174–265.

    • Search Google Scholar
    • Export Citation
  • Avissar, R., , and D. Werth, 2005: Global hydroclimatological teleconnections resulting from tropical deforestation. J. Hydrometeor., 6 , 134145.

    • Search Google Scholar
    • Export Citation
  • Bonan, G. B., , S. Levis, , L. Kergoat, , and K. W. Oleson, 2002: Landscapes as patches of plant functional types: An integrating concept for climate and ecosystem models. Global Biogeochem. Cycles, 16 , 1021. doi:10.1029/2000GB001360.

    • Search Google Scholar
    • Export Citation
  • Briegleb, B. P., , C. M. Bitz, , E. C. Hunke, , W. H. Lipscomb, , M. M. Holland, , J. L. Schramm, , and R. E. Moritz, 2004: Scientific description of the sea ice component in the Community Climate System Model, version 3. Tech. Rep. NCAR/TN-463+STR, 78 pp.

  • Carpenter, S., , T. Frost, , D. Heisey, , and T. K. Kratz, 1989: Randomized intervention analysis and the interpretation of whole-ecosystem experiments. Ecology, 70 , 11421152.

    • Search Google Scholar
    • Export Citation
  • Charney, J. G., 1975: Dynamics of deserts and drought in the Sahel. Quart. J. Roy. Meteor. Soc., 101 , 193202.

  • Collins, W. D., and Coauthors, 2006: The Community Climate System Model, version 3 (CCSM3). J. Climate, 19 , 21222143.

  • 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 , 1834.

    • Search Google Scholar
    • Export Citation
  • da Rocha, H. R., , C. A. Nobre, , J. P. Bonatti, , I. R. Wright, , and P. J. Sellers, 1996: A vegetation-atmosphere interaction study for Amazonia deforestation using field data and a ‘single column’ model. Quart. J. Roy. Meteor. Soc., 122 , 567594.

    • Search Google Scholar
    • Export Citation
  • Delire, C., , P. Behling, , M. T. Coe, , J. A. Foley, , R. Jacob, , J. Kutzbach, , Z. Liu, , and S. Vavrus, 2001: Simulated response of the atmosphere-ocean system to deforestation in the Indonesian Archipelago. Geophys. Res. Lett., 28 , 20812084.

    • Search Google Scholar
    • Export Citation
  • Dickinson, R. E., , and A. Henderson-Sellers, 1988: Modeling tropical deforestation—A study of GCM land surface parametrizations. Quart. J. Roy. Meteor. Soc., 114 , 439462.

    • Search Google Scholar
    • Export Citation
  • Eltahir, E. A. B., , and R. L. Bras, 1993: On the response of the tropical atmosphere to large-scale deforestation. Quart. J. Roy. Meteor. Soc., 119 , 779793.

    • Search Google Scholar
    • Export Citation
  • Findell, K. L., , and T. R. Knutson, 2006: Weak simulated extratropical responses to complete tropical deforestation. J. Climate, 19 , 28352850.

    • Search Google Scholar
    • Export Citation
  • Gedney, N., , and P. J. Valdes, 2000: The effect of Amazonian deforestation on the northern hemisphere circulation and climate. Geophys. Res. Lett., 27 , 30533056.

    • Search Google Scholar
    • Export Citation
  • Hack, J. J., , J. M. Caron, , S. G. Yeager, , K. W. Oleson, , M. M. Holland, , J. E. Truesdale, , and P. J. Rasch, 2006: Simulation of the global hydrological cycle in the CCSM Community Atmosphere Model version 3 (CAM3): Mean features. J. Climate, 19 , 21992221.

    • Search Google Scholar
    • Export Citation
  • Hansen, J., , G. Russell, , D. Rind, , P. Stone, , A. Lacis, , S. Lebedeff, , R. Ruedy, , and L. Travis, 1983: Efficient three-dimensional global models for climate studies: Models I and II. Mon. Wea. Rev., 111 , 609662.

    • Search Google Scholar
    • Export Citation
  • Henderson-Sellers, A., , and V. Gornitz, 1984: Possible climatic impacts of land cover transformations, with particular emphasis on tropical deforestation. Climatic Change, 6 , 231257.

    • Search Google Scholar
    • Export Citation
  • Henderson-Sellers, A., , R. E. Dickinson, , T. B. Durbidge, , P. J. Kennedy, , K. McGuffie, , and A. J. Pitman, 1993: Tropical deforestation—Modeling local-scale to regional-scale climate change. J. Geophys. Res., 98 , 72897315.

    • Search Google Scholar
    • Export Citation
  • Holtslag, A. A. M., , and C-H. Moeng, 1991: Eddy diffusivity and countergradient transport in the convective atmospheric boundary layer. J. Atmos. Sci., 48 , 16901698.

    • Search Google Scholar
    • Export Citation
  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437471.

  • Lean, J., , and P. R. Rowntree, 1993: A GCM simulation of the impact of Amazonian deforestation on climate using an improved canopy representation. Quart. J. Roy. Meteor. Soc., 119 , 509530.

    • Search Google Scholar
    • Export Citation
  • Lin, S. J., , and R. B. Rood, 1996: Multidimensional flux-form semi-Lagrangian transport schemes. Mon. Wea. Rev., 124 , 20462070.

  • Lin, S. J., , and R. B. Rood, 1997: An explicit flux-form semi-Lagrangian shallow-water model on the sphere. Quart. J. Roy. Meteor. Soc., 123 , 24772498.

    • Search Google Scholar
    • Export Citation
  • Mabuchi, K., , Y. Sato, , and H. Kida, 2005a: Climatic impact of vegetation change in the Asian tropical region. Part I: Case of the Northern Hemisphere summer. J. Climate, 18 , 410428.

    • Search Google Scholar
    • Export Citation
  • Mabuchi, K., , Y. Sato, , and H. Kida, 2005b: Climatic impact of vegetation change in the Asian tropical region. Part II: Case of the Northern Hemisphere winter and impact on the extratropical circulation. J. Climate, 18 , 429446.

    • Search Google Scholar
    • Export Citation
  • Matthews, E., 1983: Global vegetation and land use: New high-resolution data bases for climate studies. J. Climate Appl. Meteor., 22 , 474487.

    • Search Google Scholar
    • Export Citation
  • McGuffie, K., , A. Henderson-Sellers, , H. Zhang, , T. B. Durbridge, , and A. J. Pitman, 1995: Global climate sensitivity to tropical deforestation. Global Planet. Change, 10 , 97128.

    • Search Google Scholar
    • Export Citation
  • Myers, N., 1991: Tropical forests: Present status and future outlook. Climatic Change, 19 , 332.

  • Oleson, K. W., and Coauthors, 2004: Technical description of the Community Land Model (CLM). Tech. Rep. NCAR/TN-461+STR, 174 pp. [Available online at http://www.cgd.ucar.edu/tss/clm/distribution/clm3.0/TechNote/CLM_Tech_Note.pdf.].

    • Search Google Scholar
    • Export Citation
  • Peixoto, J. P., , and A. H. Oort, 1992: Physics of Climate. American Institute of Physics, 520 pp.

  • Rosenzweig, C., , and F. Abramopoulos, 1997: Land-surface model development for the GISS GCM. J. Climate, 10 , 20402054.

  • Russell, G. L., , and J. A. Lerner, 1981: A new finite-differencing scheme for the tracer transport equation. J. Appl. Meteor., 20 , 14831498.

    • Search Google Scholar
    • Export Citation
  • Russell, G. L., , J. R. Miller, , and D. Rind, 1995: A coupled atmosphere-ocean model for transient climate change studies. Atmos.–Ocean, 33 , 683730.

    • Search Google Scholar
    • Export Citation
  • Semazzi, F. H. M., , and Y. Song, 2001: A GCM study of climate change induced by deforestation in Africa. Climate Res., 17 , 169182.

  • Sud, Y. C., , G. K. Walker, , J. H. Kim, , G. E. Liston, , P. J. Sellers, , and W. K. M. Lau, 1996: Biogeophysical consequences of a tropical deforestation scenario: A GCM simulation study. J. Climate, 9 , 32253247.

    • Search Google Scholar
    • Export Citation
  • Voldoire, A., , and J. F. Royer, 2005: Climate sensitivity to tropical land surface changes with coupled versus prescribed SSTs. Climate Dyn., 24 , 843862.

    • Search Google Scholar
    • Export Citation
  • von Storch, H., , and F. W. Zwiers, 2001: Statistical Analysis in Climate Research. Cambridge University Press, 484 pp.

  • Werth, D., , and R. Avissar, 2002: The local and global effects of Amazon deforestation. J. Geophys. Res., 107 , 8087. doi:10.1029/2001JD000717.

    • Search Google Scholar
    • Export Citation
  • Werth, D., , and R. Avissar, 2005a: The local and global effects of Southeast Asian deforestation. Geophys. Res. Lett., 32 , L20702. doi:10.1029/2005GL022970.

    • Search Google Scholar
    • Export Citation
  • Werth, D., , and R. Avissar, 2005b: The local and global effects of African deforestation. Geophys. Res. Lett., 32 , L12704. doi:10.1029/2005GL022969.

    • Search Google Scholar
    • Export Citation
  • Wigley, T. M. L., , and B. D. Santer, 1990: Statistical comparison of spatial fields in model validation, perturbation, and predictability experiments. J. Geophys. Res., 95 , 851865.

    • Search Google Scholar
    • Export Citation
  • Xue, Y. K., , K. N. Liou, , and A. Kasahara, 1990: Investigation of biogeophysical feedback on the African climate using a two-dimensional model. J. Climate, 3 , 337352.

    • Search Google Scholar
    • Export Citation
  • Xue, Y. K., , H. M. H. Juang, , W. P. Li, , S. Prince, , R. DeFries, , Y. Jiao, , and R. Vasic, 2004: Role of land surface processes in monsoon development: East Asia and West Africa. J. Geophys. Res., 109 , D03105. doi:10.1029/2003JD003556.

    • Search Google Scholar
    • Export Citation
  • Zhang, G. J., , and N. A. McFarlane, 1995: Sensitivity of climate simulations to the parameterization of cumulus convection in the Canadian Climate Centre general circulation model. Atmos.–Ocean, 33 , 407446.

    • Search Google Scholar
    • Export Citation
  • Zhang, H., , A. Henderson-Sellers, , and K. McGuffie, 1996a: Impacts of tropical deforestation. Part I: Process analysis of local climatic change. J. Climate, 9 , 14971517.

    • Search Google Scholar
    • Export Citation
  • Zhang, H., , K. McGuffie, , and A. Henderson-Sellers, 1996b: Impacts of tropical deforestation. Part II: The role of large-scale dynamics. J. Climate, 9 , 24982521.

    • Search Google Scholar
    • Export Citation
  • Zheng, X. Y., , and E. A. B. Eltahir, 1997: The response to deforestation and desertification in a model of West African monsoons. Geophys. Res. Lett., 24 , 155158.

    • Search Google Scholar
    • Export Citation
  • Zheng, X. Y., , and E. A. B. Eltahir, 1998: The role of vegetation in the dynamics of West African monsoons. J. Climate, 11 , 20782096.

  • View in gallery

    Global land-cover map from Matthews (1983), emphasizing with rectangles the three regions in which tropical forests (black) are replaced with a mixture of shrubs and grassland in our deforestation experiments. Note that all areas within the boxes are converted (including places like Panama), but the changes are dominated by the Amazon, Africa, and Southeast Asia. Original (color) land-cover map and legend can be found in Matthews (1983).

  • View in gallery

    Seasonal and annual mean precipitation (mm day−1) for (left) the GPCP dataset and (right) differences between GM II and the GPCP. Seasonal averages are December–Februrary (DJF), March–May (MAM), June–August (JJA), and September–November (SON). Note that the GPCP data are interpolated to the model grid.

  • View in gallery

    (Continued) Seasonal and annual mean precipitation differences (mm day−1) between (left) AM and (right) CCSM and the GPCP dataset.

  • View in gallery

    Temporal standard deviation (mm day−1) of annual (ANN) and seasonal precipitation for the (left) NCEP reanalysis dataset and (right) GM II.

  • View in gallery

    (Continued) Temporal standard deviation (mm day−1) of ANN and seasonal precipitation for (left) AM and (right) CCSM.

  • View in gallery

    Worldwide locations where precipitation has either significantly (top) decreased or (bottom) increased during a period of at least three months of the year, as a result of tropical deforestation as simulated by an MME of three GCMs: GM II, AM, and CCSM. The MME mean annual cycle of precipitation (mm day−1) for the control (blue) and deforested (red) cases at continental locations most severly affected by the deforestation is also represented. The color scale indicates the number of months registering a statistically significant change (Student’s t test 95% significance level) during the annual cycle.

  • View in gallery

    Worldwide locations where precipitation has significantly (left) increased and (right) decreased during a period of at least three months of the year, as a result of tropical deforestation as simulated by (top) GM II, (middle) AM, and (bottom) CCSM. The color scale indicates the number of months registering a statistically significant change (Student’s t test 95% significance level) during the annual cycle.

  • View in gallery

    Ranking of the “true” ensemble among the “false” ensembles for the SITES statistics calculated for the MME and all three models individually. Both precipitation (black) and 247-mb geopotential (white) are shown (top) for the tropics and (bottom) for northern midlatitudes only.

  • View in gallery

    March geopotential change in the MME: Contour interval is 10 m, zero contour not shown. The black curve indicates 95% significance. Contour indicates geopotential changes at the 95% significance level, and the contour interval is 10 m (zero contour not drawn).

  • View in gallery

    Vertically averaged meridional potential energy flux (thick line) and precipitation (thin line) for the solstice and equinox months for (a) the control runs MME, and for the differences between deforested and control runs in (b) GM II, (c) AM, and (d) CCSM.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 32 32 15
PDF Downloads 19 19 10

Effects of Tropical Deforestation on Global Hydroclimate: A Multimodel Ensemble Analysis

View More View Less
  • 1 National Center for Atmospheric Research, Boulder, Colorado
  • 2 Savannah River National Laboratory, Aiken, South Carolina
  • 3 Department of Civil and Environmental Engineering, Duke University, Durham, North Carolina
© Get Permissions
Full access

Abstract

Two multimodel ensembles (MME) were produced with the GISS Model II (GM II), the GISS Atmosphere Model (AM), and the NCAR Community Climate System Model (CCSM) to evaluate the effects of tropical deforestation on the global hydroclimate. Each MME used the same 48-yr period but the two were differentiated by their land-cover types. In the “control” case, current vegetation was used, and in the “deforested” case, all tropical rain forests were converted to a mixture of shrubs and grassland. Globally, the control simulations produced with the three GCMs compared well to observations, both in the time mean and in the temporal variability, although various biases exist in the different tropical rain forests.

The local precipitation response to deforestation is very strong. The remote effect in the tropics (away from the deforested tropical areas) is strong as well, but the effects at midlatitudes are weaker. In the MME, the impacts tend to be attenuated relative to the individual models.

The significance of the geopotential and precipitation responses was evaluated with a bootstrap method, and results varied during the year. Tropical deforestation also produced anomalous fluxes in potential energy that were a direct response to the deforestation. These different analyses confirmed the existence of a teleconnection mechanism due to deforestation.

Corresponding author address: Roni Avissar, Department of Civil and Environmental Engineering, Edmund T. Pratt Jr. School of Engineering, 123 Hudson Hall, Duke University, Durham, NC 27708-0287. Email: avissar@duke.edu

Abstract

Two multimodel ensembles (MME) were produced with the GISS Model II (GM II), the GISS Atmosphere Model (AM), and the NCAR Community Climate System Model (CCSM) to evaluate the effects of tropical deforestation on the global hydroclimate. Each MME used the same 48-yr period but the two were differentiated by their land-cover types. In the “control” case, current vegetation was used, and in the “deforested” case, all tropical rain forests were converted to a mixture of shrubs and grassland. Globally, the control simulations produced with the three GCMs compared well to observations, both in the time mean and in the temporal variability, although various biases exist in the different tropical rain forests.

The local precipitation response to deforestation is very strong. The remote effect in the tropics (away from the deforested tropical areas) is strong as well, but the effects at midlatitudes are weaker. In the MME, the impacts tend to be attenuated relative to the individual models.

The significance of the geopotential and precipitation responses was evaluated with a bootstrap method, and results varied during the year. Tropical deforestation also produced anomalous fluxes in potential energy that were a direct response to the deforestation. These different analyses confirmed the existence of a teleconnection mechanism due to deforestation.

Corresponding author address: Roni Avissar, Department of Civil and Environmental Engineering, Edmund T. Pratt Jr. School of Engineering, 123 Hudson Hall, Duke University, Durham, NC 27708-0287. Email: avissar@duke.edu

1. Introduction

Numerous simulations of tropical deforestation suggest that replacing rain forests with grassland would decrease precipitation and evaporation and increase surface temperature over the deforested area (see Henderson-Sellers et al. 1993 for a review). Typically conducted over Amazonia (Henderson-Sellers and Gornitz 1984; Dickinson and Henderson-Sellers 1988; Eltahir and Bras 1993; Lean and Rowntree 1993; da Rocha et al. 1996; Costa and Foley 2000; Gedney and Valdes 2000; Werth and Avissar 2002), similar results were obtained for the deforestation of tropical Africa (Xue et al. 1990; Zheng and Eltahir 1997, 1998; Semazzi and Song 2001; Xue et al. 2004; Werth and Avissar 2005b), Southeast Asia (Delire et al. 2001; Xue et al. 2004; Mabuchi et al. 2005a,b; Werth and Avissar 2005a), and the entire tropical region (Sud et al. 1996; Zhang et al. 1996a,b; Avissar and Werth 2005; Voldoire and Royer 2005; Findell and Knutson 2006). The physical processes involved in deforestation include an increase in surface albedo (Charney 1975) and the modification of other land surface properties (e.g., roughness length and Bowen ratio). The overall effect is the reduction of the net energy transferred from the surface into the atmospheric boundary layer, thereby reducing moist convection. In addition to the impact on the deforested region itself, these processes can alter both the large-scale circulation (Henderson-Sellers and Gornitz 1984; Henderson-Sellers et al. 1993) and the Rossby waves, which propagate from the tropics into the midlatitudes (Gedney and Valdes 2000). These changes in circulation can ultimately modify the remote climate, a mechanism known as “teleconnection.”

While studies of the local impact of large-scale deforestation have produced fairly consistent results (Henderson-Sellers et al. 1993), possible teleconnections are still controversial (Findell and Knutson 2006). Earlier studies indicate that tropical deforestation (the Amazon in particular) has no detectable impact on the global hydroclimate (Henderson-Sellers and Gornitz 1984), but this has since been challenged. McGuffie et al. (1995) already identified disturbances of the Asian monsoon and changes in climate at mid and high latitudes due to tropical deforestation, and Zhang et al. (1996b) identified a possible Rossby wave propagation induced by tropical deforestation. In their numerical simulation of Amazon deforestation, Gedney and Valdes (2000) noted a close correlation between precipitation anomalies over the North Atlantic and Europe and upper-level circulation changes due to Amazon deforestation, suggesting that deforestation in the Amazon affects precipitation over Europe through Rossby wave propagation. Similarly, Werth and Avissar (2002) identified a reduction of precipitation in North America when the Amazon is deforested. African (Semazzi and Song 2001; Werth and Avissar 2005b) and Southeast Asian (Mabuchi et al. 2005b; Werth and Avissar 2005a) deforestation also modifies Rossby wave trains, with their attendant precipitation changes. Avissar and Werth (2005) used the NASA Goddard Institute for Space Studies (GISS) general circulation model (GCM) version II (Hansen et al. 1983) (hereafter referred to as “GM II”) to identify teleconnections due to individual rain forest removal as well as teleconnections resulting from the deforestation of all tropical rain forests simultaneously. They found that deforestation of Amazonia and central Africa severely reduces rainfall in the U.S. Midwest, deforestation of Southeast Asia affects China and the Balkan Peninsula most significantly, and the elimination of any of these tropical forests considerably enhances summer rainfall in the southern tip of the Arabian Peninsula. They also emphasize that the combined effect of deforestation of these three tropical regions does not correspond to the sum of their separate regional effects and also causes, for example, a significant decrease in winter precipitation in California not seen when any rain forest is removed individually.

Inevitably missing from single-GCM studies is the possible effect of the specific model dependence on the experimental results. Models differ in their dynamical core, numerical schemes, as well as their parameterizations (e.g., land surface, turbulence, radiation, and cloud microphysics), which could affect the response of precipitation to deforestation. Model intercomparison projects, in which several models perform the same highly controlled numerical experiment, can be used to assess the various model dependencies and measure the uncertainty in our understanding of both the studied processes and our ability to simulate them. Given the alarming estimated and projected rates of deforestation (Myers 1991; Achard et al. 1998) and their significant potential impact on the global climate, it is important to increase our confidence in the pattern and strength of the teleconnections as found by Avissar and Werth (2005) by reproducing their experiment with different models. The main objective of the study described here, which is a continuation of the studies done by Avissar and Werth (2005) and Werth and Avissar (2002; 2005a,b), was to perform both an intercomparison study and a multimodel ensemble (MME) of tropical deforestation experiments to assess the importance of teleconnections resulting from such land-cover change in all tropical regions. Three different GCMs were used for this purpose.

2. Numerical experiments

To create an MME, three different GCMs are used: GM II, the atmosphere model also developed at GISS (extensive references on the model can be found at http://aom.giss.nasa.gov/publica.html; Russell et al. 1995) hereafter refer to as “AM,” and the Community Climate System Model, version 3.0 [CCSM3; an extensive description of CCSM3 can be found at http://www.ccsm.ucar.edu/ and in a special issue of the Journal of Climate (2006, Vol. 19, No. 11) devoted to CCSM; Collins et al. 2006]. Each model is configured with 4° × 5° latitude–longitude horizontal elements and with either 12 (GM II and AM) or 26 (CCSM3) vertical layers. Each model is run with both a “control” and a “deforested” scenario for a single, 52-yr simulation each. The first 4 years of each simulation are discarded as spinup, thus resulting in each month of the year being simulated 48 times for each simulation. Observed monthly-mean sea surface temperatures (SSTs) and ice cover derived from a multiyear climatological record (1978–2002) from the National Center for Environmental Predictions (NCEP) reanalysis dataset [NCEP reanalysis data provided by the National Oceanic and Atmospheric Administration (NOAA) Cooperative Institute for Research in Environmental Sciences (CIRES) Climate Diagnostics Center, Boulder, Colorado, from their Web site at http://www.cdc.noaa.gov/; Kalnay et al. 1996] are used for all simulations (i.e., each ocean point cycles through the same temperature and ice cover values each year). As explained in Werth and Avissar (2005a) this use of prescribed climatological SSTs reduces the climate system “memory” to a few months and eliminate any year-to-year autocorrelation that would preclude use of the t test (which presumes complete statistical independence). A calculation of autocorrelation for precipitation for all three models reveals that, for any month, the annual autocorrelation exceeds 0.5 (or falls below −0.5) only at very few points (6 or fewer, out of a field of 3312). Thus, under such conditions, each simulation produces 48 independent realizations for each month of the year.

In GM II, heat and humidity are advected with a quadratic upstream scheme, and momentum is advected with a second-order scheme. The model has both shallow and deep convection, and a second-order closure planetary boundary layer scheme for moisture and heat transfer is applied at the surface. The model uses six soil layers and a hydrology scheme that accounts for soil moisture transfer and root extraction (Rosenzweig and Abramopoulos 1997), the latter of which depends on the vegetation specified within a grid element.

AM is similar to GM II except that the atmospheric mass and momentum equations are solved on a modified version of the “C grid” scheme of Arakawa and Lamb (1977) instead of the “B grid” adopted in GM II. The model also uses an added linear-upstream scheme (Russell and Lerner 1981) for heat and moisture advection, with subgrid-scale linear gradients in three dimensions, allowing for subgrid-scale effects to be included in the physics. The land surface scheme is the same as in GM II.

The CCSM3 configuration used in this study couples an atmosphere, land, and sea ice model. The Community Atmosphere Model (CAM) horizontal discretization is based on a conservative “flux-form semi-Lagrangian” scheme described by Lin and Rood (1996; 1997). The vertical discretization can be best described as Lagrangian with a conservative remapping, which essentially makes it quasi-Lagrangian. The model has both shallow and deep convection (Zhang and McFarlane 1995) and uses an explicit, nonlocal atmospheric boundary layer parameterization (Holtslag and Moeng 1991). The Community Land Model, version 3 (CLM3) is based on a nested subgrid hierarchy of scales representing land units, soil or snow columns, and plant functional types (Bonan et al. 2002; Oleson et al. 2004). The model uses 10 soil layers with vertically varying soil textures and a hydrology scheme that accounts for soil moisture transfer and root extraction, depending on the soil texture and plant functional type. It also includes the effects of competition for water among plant functional types. The hydrology component of the model includes river routing for the runoff. The Community Sea Ice Model, version 5.0 (CSIM5; Briegleb et al. 2004) accounts for ice dynamics, varying sea ice albedos, and exchanges of salt between sea ice and the surrounding ocean, while keeping the sea ice fraction as prescribed in the input dataset.

In the “control” MME, a vegetation map developed by Matthews (1983) for the period starting in 1960 and ending in 1979 (before heavy deforestation started) is adopted for the simulations. The dataset is translated into the nine biome classes of the GM II and AM (Matthews 1983) and into the 17 plant functional types of CLM3 using the method developed by Bonan et al. (2002). In the “deforested” MME, all tropical forests within the boxes in Fig. 1 are replaced with a mixture of shrubs and grassland. While corresponding to a single biome class in both GM II and AM, the proportion of shrubs and grasses in the CLM3 was set so that it matched the GM II shrubs and grassland biome’s albedo, or set to 60% shrub when the GM II albedo could not be matched.

There are differences in the representation of land-cover types (in both the forested and deforested simulations) between CLM3 and GM II/AM, resulting in different albedo, leaf area index, roughness length, and displacement height, among others. As some of these parameters vary interactively in the land surface scheme as the simulations evolve, a straightforward comparison between them is not possible, and they are part of the reasons (together with the other parameterizations in each model) for the differences in the simulation results.

3. Model performance

a. Mean precipitation fields

The three GCMs selected for this study have been used extensively to produce climate scenarios under various forcing conditions, and their ability to reproduce the current climate has been described in various publications (Russell et al. 1995; Avissar and Werth 2005; Collins et al. 2006). Nonetheless, it is important to understand their ability to reproduce regional and global precipitation, a key variable in the present study, and particularly to emphasize their performance differences with respect to observations.

The performance evaluation was done by comparing the control run produced with each one of the three GCMs to four precipitation datasets: the Legates and Microwave Sounding Unit (MSU) Precipitation Climatology (Legates/MSU) (data and documentation of an updated version of the original dataset can be found at http://jisao.washington.edu/legates_msu/), the Global Air Temperature and Precipitation Climatologies, version 2.1 (Willmott–Matsuura–Legates; data and documentation can be found at http://climate.geog.udel.edu/~climate/), the NASA Global Precipitation Climatology Project (GPCP) (data and documentation can be found at http://precip.gsfc.nasa.gov), and the Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) (data and documentation can be found at http://www.cdc.noaa.gov/cdc/data.cmap.html). All datasets exist on a 2.5° × 2.5° latitude–longitude grid except for the Willmott–Matsuura–Legates dataset, which is on a 0.5° × 0.5° grid. It is important to emphasize that these datasets are merely derived from observations. All datasets were interpolated to the model grids for comparison purposes. Figure 2 shows the seasonal-mean precipitation (mm day−1) for the GPCP dataset compared to the model simulations. All three GCMs show a similar general pattern of biases: positive biases over the tropical oceans and a dry bias over the land.

Table 1 summarizes the global and regional differences between models and observations. The models tend to have a wet bias for the global, annual-mean precipitation when compared to observations, with AM being the wettest. Most of this wet bias is located over the oceans, particularly over the Indian Ocean, where a wet bias occurs in all three models as compared to GPCP (Fig. 2). In CCSM, the Indian Ocean wet bias also extends into the Indian subcontinent and Arabian Peninsula as well as most of sub-Saharan Africa. From July to November, this is associated with a dry bias over coastal and southern China and the Indochina Peninsula, which indicates a shift in the monsoonal system to the west, a known problem of the model. This feature is also present (yet weaker) in AM and GM II. To a lesser extent, the highest mountainous regions of the Andes, the Himalayas, and even the Rockies also contribute to the wet bias, particularly for AM.

Table 1 indicates that, globally, all models have annual-mean dry biases compared to the Legates/MSU dataset, mainly over the Northern Atlantic and Pacific Oceans, as well as the southern high-latitude Pacific (not shown). This worldwide dry bias is only seen with the Legates/MSU dataset and not with the other datasets. All models position the ITCZ as observed (not shown). Precipitation mismatches between simulations and observations over the oceans could be partly due to the model forcing by multiyear-averaged monthly SSTs. Both prescribed climatological SSTs and fixed 1960s land-cover boundary conditions eliminate some of the interannual variability, which might affect the monthly, annual, and global mean precipitation through nonlinear interactions in the climate system. Nevertheless, most of these systematic biases are known problems in GCMs even when forced with observed, transient SSTs (Hack et al. 2006). Over mountainous areas, the models’ coarse horizontal resolution and their schematic representation of the topography tended to create erroneous motion in the mountain lee. Nonetheless, global patterns are generally well represented with all models (Fig. 2), and the mismatch in the precipitation amounts partly depends on the observation dataset to which they are compared. Globally, the models have similar seasonal behavior, and the resulting differences with observations have typically the same sign and comparable magnitude.

Of particular importance for this study are the regional precipitation patterns over the tropical forests. Over Amazonia, all models experience dry biases annually (Table 1) and these are strongest for GM II. CCSM is the only model to show some seasonality in the biases, being somewhat wet during the wet season (September through March) and dry for the remaining months. Over central Africa, CCSM shows a general wet bias except from June to August, while AM and GM II have seasonally and latitudinally contrasting results (Fig. 2). Over Southeast Asia, all models show a general dry bias, with differences in magnitude: a strong dry bias exists in AM and a weak one is seen in CCSM. CCSM is also the only model to show a wet bias from December through February. Note that the numbers given in Table 1 only refer to the land areas included in the boxes of Fig. 1 and that it was further restricted in the Southeast Asian case to the tropical areas of Indonesia and the Indochina Peninsula. The differences observed between models are probably inherent to their ability to represent the ITCZ structure, the meridional migration of tropical precipitation, and the patterns of the monsoonal systems. Predominantly over the Amazon basin, the land portion of the water cycle, namely the evapotranspiration recycling, may also be poorly resolved.

b. Temporal variability in precipitation

To illustrate the capability of the models to simulate interannual variability, Fig. 3 shows their standard deviation of seasonally averaged precipitation compared to that from the NCEP reanalysis (Kalnay et al. 1996). The latter was calculated for 48 years (from 1956 to 2003). As expected, the variability of the reanalysis exceeds that of the models in all seasons, with values greater than 0.5 mm day−1 over most of the model domain. Of the models, GM II seems to have the highest variance and the AM the lowest with the CCSM in between. All models have too little variability over Africa and the Pacific Ocean, and CCSM and AM have too little over Southeast Asia.

The reduced variability is partly caused by the use of prescribed climatological SSTs. The main consequence of a weaker variability for the subsequent analysis is that the same changes in the mean could appear more significant when a Student’s t test is used. However, the modeled values seen here are reasonably close to the reanalysis values, justifying a comparison of the control to the deforested runs.

4. Impacts of deforestation on precipitation

To assess the impact of deforestation on precipitation, the null statistical hypothesis H0, stating that the control and deforested populations had the same mean precipitation, can be tested with a Student’s t value for each month at each grid point. For any given month, the year-to-year autocorrelation is negligible owing to the use of prescribed climatological SST, which eliminates the model source of long-term persistence. Each grid point that could reject H0 at the 95% significance level for three months or more can be considered to have experienced a significant impact from the deforestation process. The subjective restriction to three or more months is done largely to reduce the number of cells that show a statistical difference in precipitation solely due to the internal variability of the model or to statistical artifacts, and to better emphasize those regions that are affected by the deforestation in a meaningful way from a hydroclimatological point of view, not merely statistically. One-tailed t values are used to separate the increased/decreased precipitation impact of the deforestation. The analysis is twofold: one test is performed on the control and deforested MMEs (i.e., combining the three model results into one large dataset), and a second test is done for each model separately to identify regions where impacts could be identified with a specific model.

A map of t-test significance values must consider the natural spatial correlation of meteorological variables—points on the map cannot be considered independent, so if one point has a large t-test value, nearby points are more likely to have similar values, which could create large areas of significant change that occur by chance. Such maps do, however, provide a good first look at the effects of deforestation relative to the location of the deforested areas. In section 5, a bootstrap analysis is performed to sort out further how significant the changes are.

a. MME analysis

Figure 4 shows those locations worldwide where precipitation has either significantly decreased or increased for at least three months of the year as a result of tropical deforestation simulated by the MME. Depending on the month, precipitation significantly decreases over 30%–75% of the deforested area (Fig. 4, top). Over the areas that experience a decrease, annual precipitation is reduced by an average of 0.4–0.7 mm day−1, representing a 7%–18% decrease, while the most strongly affected months average a 0.7–1.1 mm day−1 reduction (14%–25%), peaking as high as −38 mm day−1 or 80%, consistent with previously published results. Unlike in Avissar and Werth (2005) who used GM II for their study, only a weak seasonality is observed in the precipitation decreases in any of the tropical regions. Some 6%–13% of the deforested grid cells also experience an annual increased precipitation averaging 0.06–0.18 mm day−1, or <5% of the annual mean (in those cells). Generally, this increase occurs during the wet season.

Outside the deforested regions, statistically significant changes in precipitation can be found at a few locations in North America, Africa, and in the western tropical Atlantic and Pacific (Fig. 4). When averaged over all colored grid cells shown in Fig. 4b, the annual precipitation increases by 0.03 mm day−1, or 3% on average. When averaged over the colored cells of Fig. 4a, it decreases by 0.06 mm day−1, representing about 4%. These amounts average a 0.14 mm day−1 increase and a 0.19 mm day−1 decrease when focusing on the affected months only, corresponding to a 14% and 12% change, respectively. The variability in change is large and standard deviations of 0.9–1.1 mm day−1 could be found for both the increased and decreased precipitation. Regions with the largest changes were usually located close to the deforested areas, as seen for the precipitation decreases over the west Atlantic. Nonetheless, several remote regions experience significant changes as well. For example, a few grid cells in the eastern United States show an average 7%–11% decrease in precipitation (0.14–0.17 mm day−1) for the months of November through April.

b. Single model analysis

Figure 5 shows those locations worldwide where precipitation either significantly decreases or increases for at least three months of the year as a result of tropical deforestation simulated by the three models individually. Generally, in all models, more sites experience a statistically significant precipitation change owing to deforestation compared to the MME, particularly over the oceans and for remote locations where the spatial discrepancies between models are larger. Significantly affected locations could be found in Europe and North America, and about 50% of deforested tropical Africa experiences a decrease in precipitation in the individual simulations, compared to only 28% of the area for the MME.

GM II shows greatly decreased precipitation over the deforested and downstream areas, as well as the U.S. East Coast and the Atlantic summertime ITCZ area. Marginal increases in precipitation appear in the Caribbean and along the tropical African coast. AM simulates areas of decreased precipitation that do not spread far beyond the deforested regions, but does present a larger number of locations where precipitation significantly increased for several months, such as the African Sahel, the Indian Ocean, and the east Pacific. CCSM exhibits the most months and locations with a precipitation increase, with large areas of the tropics experiencing precipitation changes over both land and sea regions. Places where changes in precipitation are found in the MME are generally those where at least two out of the three models show precipitation changes, which suggests that no one of the models has a dominant impact on the MME.

Tropical deforestation locations that experienced a precipitation increase are mainly coastal areas. Areas of precipitation increase can also be found over southeastern Amazonia, mainly in the CCSM. The precipitation in this area increases during the wet season (December through February) but decreases for the rest of the year (not shown). Similarly, the Sahel region, which is at the margin of the tropical area, experiences a significant precipitation increase in both CCSM and AM. The area to the northwest of the Amazon experiences a decrease in precipitation in all models.

While less consistent than over tropical regions, some similarities can be found in the different model extratropical regions. All three models experience increases in precipitation in the tropical west Pacific, and all three show decreases in precipitation in the tropical east Pacific (near South America). Increased precipitation can be found over the Arabian Peninsula, the Sahel region, as well as in the region encompassing Iran to western China in the AM and CCSM simulations. Furthermore, the extratropical locations significantly affected in the MME do not necessarily show a significant precipitation change in any one of the models separately. Interestingly, no increase is seen in the Saudi Arabian region in GM II, despite such a change existing in another simulation with this model (Avissar and Werth 2005). This is due to the different SSTs used in this experiment, which are 0.4°–1.0°C warmer in the northern Indian Ocean in the current experiment. An additional simulation with GM II using the SSTs from Avissar and Werth (2005) confirms their previous findings and suggests a strong sensitivity of GM II to SSTs.

5. Bootstrap analysis of precipitation and upper-level geopotential

For each model, differences will exist between realizations of the same ensemble, even if they are run with the same vegetation conditions. Therefore, it can be assumed that the differences between the control and deforested ensembles are at least partly due to the natural variability in the simulated climate system. The spatial autocorrelation in precipitation and geopotential can create widespread changes in those variables that are the result of random variability alone (von Storch and Zwiers 2001). While averaging and the three-month criteria applied to the t test are meant to eliminate this, the question remains as to what observed differences are actually induced by the change in land surface. This can be determined by creating “false” ensembles in which the control and deforested members are combined randomly (maintaining a 1:1 ratio) and analyzing the difference between the false “control” and “deforested” runs. This can serve as a “placebo” to which we can compare the differences between the actual control and deforested ensemble (the “true” ensemble).

For the MME, as well as for each model separately, a bootstrap process is used to create 1000 false control and 1000 false deforested datasets (Carpenter et al. 1989) in which each false control and false deforested dataset comprises 72 (24 when applied to the models separately) control and 72 (24) deforested runs selected randomly. The “true” ensemble comprises the actual 144 (48) member control and deforested datasets. The difference between the false control and false deforested ensembles can be taken as a null statistic since no net vegetation differences exist between them. To compare them, we examine the 247-mb geopotential and precipitation differences by calculating the SITES statistic (Wigley and Santer 1990), a measure of mean square differences between the time-mean control and deforested fields normalized by the spatial means of the time variance at each point:
i1520-0442-22-5-1124-eq1
where dx and mx are the values of the two spatial fields at location x, nt is the number of points in time, and σD and σM are the spatial means of the time variances of D and M at each point.

Figure 6 (top row) shows the ranking of the tropical (26°S–26°N) SITES value for the true ensemble compared to the false ensembles for the MME and for each model separately. To eliminate the local effect of deforestation, we have not used the deforested grid cell in calculating the SITES statistic—the tropical changes are entirely due to the remote effects of deforestation.

The MME “true” precipitation beats over 60% of the “false” ensembles for seven months, while the geopotential ranks very low throughout the year. In only one month of the year does the precipitation ranking fall below 0.5. Of course, with 12 precipitation bars and 12 geopotential bars, we might expect half of them (12) to lie above 0.5 just by chance. However, the fact that almost all the precipitation bars are above 0.5 is more suggestive that the precipitation effect is due to deforestation. This behavior is not reproduced by the individual models, however, which all show high precipitation and geopotential ranking in the tropics throughout most of the year. The “true” geopotential changes in GM II rank over 98% for five months of the year, and precipitation ranks at over 95% for seven months. The “true” geopotential changes in AM rank among the top 10% of the “false” changes for six months of the year, and precipitation ranks over 98% for six months. CCSM shows the best ranking with both precipitation and geopotential ranking 100% throughout the year, in agreement with Fig. 5 which shows that the CCSM experiences the largest precipitation changes due to deforestation.

The large discrepancy between the MME and the individual models for the upper-level geopotential is consistent with the t-test results that show the CCSM geopotential to rise in the tropics overall, while that of the other two models falls (not shown). This contrasting behavior is less apparent in the precipitation fields in which both increases and decreases can be observed over the tropics, particularly in CCSM and AM, leading to a better agreement between models and hence a more consistent MME behavior. In other words, models differ in the process responses to deforestation (exemplified here by the upper-level geopotential field) but not necessarily in the resulting precipitation field.

Figure 6 shows that the tropics are most sensitive to the effects of deforestation. To determine the strength of the more remote teleconnection associated with tropical deforestation, the same bootstrap analysis is performed again but with the SITES calculation limited to the region north of 30°N (Fig. 6, bottom row). The MME results are poor for both geopotential and precipitation. We see values lower than we might expect to see if the true ensemble members came from a normal distribution. The individual models show effects that are significant (Fig. 6) but tend to cancel each other out when the model results are combined into the MME. For example, the GISS and AM simulations show strong anomalies of opposite sign in July (not shown), which then sees little change when the model results are combined. In March, however, the geopotential anomalies tend to reinforce each other, yielding a strong wave pattern over Asia and North America (Fig. 7). While one month of 12 scoring high could also be due to chance, the fact that the March geopotential towers over the others is at least suggestive of a deforestation effect.

6. Changes in potential energy flux

If the large-scale circulation is affected by a change in land surface, this can possibly be detected by a change in the mean meridional circulation (MMC). A good measure of the strength of the tropical MMC is the geopotential height flux (Peixoto and Oort 1992), defined as
i1520-0442-22-5-1124-eq2
where square brackets denote a zonal mean and a prime denotes the deviation from the zonal mean. Normally, the equation includes terms for the time mean and deviation, but these are small for the monthly-mean values of this study, so they have been ignored. Also, eddies tend to be quasigeostrophic and are largely unable to transport potential energy (PE) in the zonal mean (Peixoto and Oort 1992). Therefore, the term for the eddy transport of geopotential height [υz′] is small, and we concentrate on the transport due to the mean meridional circulation [υ][z].

The transport is calculated for each model and vertically averaged at each latitude. Figure 8a shows the January transport of height flux (or PE) and the movement of potential energy out of the tropics (where it is created diabatically) into the Northern Hemisphere can be seen. The January changes induced by deforestation (Figs. 8b–d) show many differences, but are consistent in that they predict an increase in northward transport at ∼10°S, near the southern edge of the peak in Fig. 8a. This results from a redistribution of precipitation seen in Figs. 4 and 5. The zonal-mean precipitation changes are in quadrature with the fluxes: in all the models, precipitation rises slightly at around 20°S and falls sharply at about 5°S. This imposes a gradient in diabatic heating and potential energy production, leading to an induced northward flux.

Figure 8 also shows the same analysis for July, during which geopotential is moved south from a generation point at 5°N. The changes induced by deforestation are much less dramatic during this solstice month, but both GM II and AM do show an increase in (southward) flux at the lagging edge of the maximum southward transport, while CCSM shows the opposite.

Figure 8 shows the results for the equinox months of April and October, when potential energy is generated at the equator and passes into both hemispheres. After deforestation, the flux out of the tropics seems to be reduced in April, while it is (weakly) increased in October, particularly in CCSM. Again, this is indicative of changes of potential energy production at the equator. Since they have identical irradiance profiles, it is surprising that April and October are different. In April, however, the precipitation increases from 20°S to 15°S, while it falls at about 5°S, similar to January. The October precipitation changes are weaker, and this suggests a deforestation response with a thermal dependence.

7. Summary and conclusions

Three different GCMs were used in this study to produce a MME to detect possible hydrometeorological teleconnections due to total tropical deforestation. The Student’s t-test analysis of the MME showed a strong decrease in mean annual precipitation over 30%–75% of the deforested areas with a few monthly precipitation increases in some particular grid cells, especially in coastal regions or close to mountain ranges. Precipitation changes could also be found outside the deforested areas, mainly in the tropics and at a few locations in the northern midlatitudes, reaching up to a 34% change in monthly precipitation. The bootstrap analysis revealed a strong tropical precipitation effect but a contradictory geopotential effect—two models showed geopotential decreases, while the third showed increases. The effect in the northern midlatitudes is weaker, but we do see some evidence of a wave train forced by the tropical changes. Our interpretation is that the effect exists in the individual models, but it is buried in the natural variability of the climate system. When the results are combined into a multimodel ensemble, however, this has the effect of making the signal stronger relative to the noise. With more months simulated, a signal must be stronger to rise above the noise, so only the strongest signals will come through. The multimodel ensemble is not creating a signal where none exists. Rather, it acts as a filter to help eliminate spurious signals.

The zonal- and vertical-mean potential energy fluxes further emphasized the change in energy transport due to deforestation from the tropics into the northern midlatitudes. While these fluxes and geopotential height modifications confirmed the existence of a teleconnection mechanism, there was no straightforward answer on where and when precipitation changes might occur due to tropical deforestation.

By using three different GCMs in an MME analysis, model-specific sensitivities were somewhat eliminated and showed that some tropical deforestation-induced precipitation changes are robust to model particularities and confirm previous findings (Avissar and Werth 2005). Nonetheless, each model predicted different patterns of midlatitude geopotential and precipitation change due to tropical deforestation, and the reasons for these discrepancies as well as the details of the mechanisms involved in the teleconnection still need further investigation. Studies using GCMs to predict the large-scale effects of tropical deforestation have produced a range of responses from weak (Findell and Knutson 2006) to strong (Gedney and Valdes 2000), with the current research being somewhere in between. Based on our findings here, the results from global simulations are inconsistent, but do suggest the existence of a remote effect. Furthermore, the use of prescribed climatological SSTs can act to dampen the model interannual variability, which could increase or decrease the magnitude of changes due to deforestation. Reproducing these results with transient observed SSTs, or with coupled ocean–atmosphere models is an essential next step to this study to provide additional insights on the spatial and temporal variability of the teleconnections suggested here.

Acknowledgments

This research was funded by the National Science Foundation (NSF) under Grants ATM-0346554 and ATM-0634745. The views expressed herein are those of the authors and do not necessarily reflect the views of NSF. We are very grateful to the Terrestrial Science Section at the National Center for Atmospheric Research (NCAR) in Boulder Colorado for their hosting, help, advice and support in using CCSM. We would especially like to thank Dave Schimel, Gordon Bonan, Samuel Levis, and Mariana Vertenstein.

REFERENCES

  • Achard, F., , H. Eva, , A. Glinni, , P. Mayaux, , T. Richards, , and H. J. Stibig, 1998: Identification of deforestation hot spots areas in the humid tropics. TREES Publications Series B Research Rep. 4, 84 pp.

    • Search Google Scholar
    • Export Citation
  • Arakawa, A., , and R. V. Lamb, 1977: Computational design of the basic dynamical processes of the UCLA general circulation model. Methods in Computational Physics, Vol. 17, Academic Press, 174–265.

    • Search Google Scholar
    • Export Citation
  • Avissar, R., , and D. Werth, 2005: Global hydroclimatological teleconnections resulting from tropical deforestation. J. Hydrometeor., 6 , 134145.

    • Search Google Scholar
    • Export Citation
  • Bonan, G. B., , S. Levis, , L. Kergoat, , and K. W. Oleson, 2002: Landscapes as patches of plant functional types: An integrating concept for climate and ecosystem models. Global Biogeochem. Cycles, 16 , 1021. doi:10.1029/2000GB001360.

    • Search Google Scholar
    • Export Citation
  • Briegleb, B. P., , C. M. Bitz, , E. C. Hunke, , W. H. Lipscomb, , M. M. Holland, , J. L. Schramm, , and R. E. Moritz, 2004: Scientific description of the sea ice component in the Community Climate System Model, version 3. Tech. Rep. NCAR/TN-463+STR, 78 pp.

  • Carpenter, S., , T. Frost, , D. Heisey, , and T. K. Kratz, 1989: Randomized intervention analysis and the interpretation of whole-ecosystem experiments. Ecology, 70 , 11421152.

    • Search Google Scholar
    • Export Citation
  • Charney, J. G., 1975: Dynamics of deserts and drought in the Sahel. Quart. J. Roy. Meteor. Soc., 101 , 193202.

  • Collins, W. D., and Coauthors, 2006: The Community Climate System Model, version 3 (CCSM3). J. Climate, 19 , 21222143.

  • 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 , 1834.

    • Search Google Scholar
    • Export Citation
  • da Rocha, H. R., , C. A. Nobre, , J. P. Bonatti, , I. R. Wright, , and P. J. Sellers, 1996: A vegetation-atmosphere interaction study for Amazonia deforestation using field data and a ‘single column’ model. Quart. J. Roy. Meteor. Soc., 122 , 567594.

    • Search Google Scholar
    • Export Citation
  • Delire, C., , P. Behling, , M. T. Coe, , J. A. Foley, , R. Jacob, , J. Kutzbach, , Z. Liu, , and S. Vavrus, 2001: Simulated response of the atmosphere-ocean system to deforestation in the Indonesian Archipelago. Geophys. Res. Lett., 28 , 20812084.

    • Search Google Scholar
    • Export Citation
  • Dickinson, R. E., , and A. Henderson-Sellers, 1988: Modeling tropical deforestation—A study of GCM land surface parametrizations. Quart. J. Roy. Meteor. Soc., 114 , 439462.

    • Search Google Scholar
    • Export Citation
  • Eltahir, E. A. B., , and R. L. Bras, 1993: On the response of the tropical atmosphere to large-scale deforestation. Quart. J. Roy. Meteor. Soc., 119 , 779793.

    • Search Google Scholar
    • Export Citation
  • Findell, K. L., , and T. R. Knutson, 2006: Weak simulated extratropical responses to complete tropical deforestation. J. Climate, 19 , 28352850.

    • Search Google Scholar
    • Export Citation
  • Gedney, N., , and P. J. Valdes, 2000: The effect of Amazonian deforestation on the northern hemisphere circulation and climate. Geophys. Res. Lett., 27 , 30533056.

    • Search Google Scholar
    • Export Citation
  • Hack, J. J., , J. M. Caron, , S. G. Yeager, , K. W. Oleson, , M. M. Holland, , J. E. Truesdale, , and P. J. Rasch, 2006: Simulation of the global hydrological cycle in the CCSM Community Atmosphere Model version 3 (CAM3): Mean features. J. Climate, 19 , 21992221.

    • Search Google Scholar
    • Export Citation
  • Hansen, J., , G. Russell, , D. Rind, , P. Stone, , A. Lacis, , S. Lebedeff, , R. Ruedy, , and L. Travis, 1983: Efficient three-dimensional global models for climate studies: Models I and II. Mon. Wea. Rev., 111 , 609662.

    • Search Google Scholar
    • Export Citation
  • Henderson-Sellers, A., , and V. Gornitz, 1984: Possible climatic impacts of land cover transformations, with particular emphasis on tropical deforestation. Climatic Change, 6 , 231257.

    • Search Google Scholar
    • Export Citation
  • Henderson-Sellers, A., , R. E. Dickinson, , T. B. Durbidge, , P. J. Kennedy, , K. McGuffie, , and A. J. Pitman, 1993: Tropical deforestation—Modeling local-scale to regional-scale climate change. J. Geophys. Res., 98 , 72897315.

    • Search Google Scholar
    • Export Citation
  • Holtslag, A. A. M., , and C-H. Moeng, 1991: Eddy diffusivity and countergradient transport in the convective atmospheric boundary layer. J. Atmos. Sci., 48 , 16901698.

    • Search Google Scholar
    • Export Citation
  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437471.

  • Lean, J., , and P. R. Rowntree, 1993: A GCM simulation of the impact of Amazonian deforestation on climate using an improved canopy representation. Quart. J. Roy. Meteor. Soc., 119 , 509530.

    • Search Google Scholar
    • Export Citation
  • Lin, S. J., , and R. B. Rood, 1996: Multidimensional flux-form semi-Lagrangian transport schemes. Mon. Wea. Rev., 124 , 20462070.

  • Lin, S. J., , and R. B. Rood, 1997: An explicit flux-form semi-Lagrangian shallow-water model on the sphere. Quart. J. Roy. Meteor. Soc., 123 , 24772498.

    • Search Google Scholar
    • Export Citation
  • Mabuchi, K., , Y. Sato, , and H. Kida, 2005a: Climatic impact of vegetation change in the Asian tropical region. Part I: Case of the Northern Hemisphere summer. J. Climate, 18 , 410428.

    • Search Google Scholar
    • Export Citation
  • Mabuchi, K., , Y. Sato, , and H. Kida, 2005b: Climatic impact of vegetation change in the Asian tropical region. Part II: Case of the Northern Hemisphere winter and impact on the extratropical circulation. J. Climate, 18 , 429446.

    • Search Google Scholar
    • Export Citation
  • Matthews, E., 1983: Global vegetation and land use: New high-resolution data bases for climate studies. J. Climate Appl. Meteor., 22 , 474487.

    • Search Google Scholar
    • Export Citation
  • McGuffie, K., , A. Henderson-Sellers, , H. Zhang, , T. B. Durbridge, , and A. J. Pitman, 1995: Global climate sensitivity to tropical deforestation. Global Planet. Change, 10 , 97128.

    • Search Google Scholar
    • Export Citation
  • Myers, N., 1991: Tropical forests: Present status and future outlook. Climatic Change, 19 , 332.

  • Oleson, K. W., and Coauthors, 2004: Technical description of the Community Land Model (CLM). Tech. Rep. NCAR/TN-461+STR, 174 pp. [Available online at http://www.cgd.ucar.edu/tss/clm/distribution/clm3.0/TechNote/CLM_Tech_Note.pdf.].

    • Search Google Scholar
    • Export Citation
  • Peixoto, J. P., , and A. H. Oort, 1992: Physics of Climate. American Institute of Physics, 520 pp.

  • Rosenzweig, C., , and F. Abramopoulos, 1997: Land-surface model development for the GISS GCM. J. Climate, 10 , 20402054.

  • Russell, G. L., , and J. A. Lerner, 1981: A new finite-differencing scheme for the tracer transport equation. J. Appl. Meteor., 20 , 14831498.

    • Search Google Scholar
    • Export Citation
  • Russell, G. L., , J. R. Miller, , and D. Rind, 1995: A coupled atmosphere-ocean model for transient climate change studies. Atmos.–Ocean, 33 , 683730.

    • Search Google Scholar
    • Export Citation
  • Semazzi, F. H. M., , and Y. Song, 2001: A GCM study of climate change induced by deforestation in Africa. Climate Res., 17 , 169182.

  • Sud, Y. C., , G. K. Walker, , J. H. Kim, , G. E. Liston, , P. J. Sellers, , and W. K. M. Lau, 1996: Biogeophysical consequences of a tropical deforestation scenario: A GCM simulation study. J. Climate, 9 , 32253247.

    • Search Google Scholar
    • Export Citation
  • Voldoire, A., , and J. F. Royer, 2005: Climate sensitivity to tropical land surface changes with coupled versus prescribed SSTs. Climate Dyn., 24 , 843862.

    • Search Google Scholar
    • Export Citation
  • von Storch, H., , and F. W. Zwiers, 2001: Statistical Analysis in Climate Research. Cambridge University Press, 484 pp.

  • Werth, D., , and R. Avissar, 2002: The local and global effects of Amazon deforestation. J. Geophys. Res., 107 , 8087. doi:10.1029/2001JD000717.

    • Search Google Scholar
    • Export Citation
  • Werth, D., , and R. Avissar, 2005a: The local and global effects of Southeast Asian deforestation. Geophys. Res. Lett., 32 , L20702. doi:10.1029/2005GL022970.

    • Search Google Scholar
    • Export Citation
  • Werth, D., , and R. Avissar, 2005b: The local and global effects of African deforestation. Geophys. Res. Lett., 32 , L12704. doi:10.1029/2005GL022969.

    • Search Google Scholar
    • Export Citation
  • Wigley, T. M. L., , and B. D. Santer, 1990: Statistical comparison of spatial fields in model validation, perturbation, and predictability experiments. J. Geophys. Res., 95 , 851865.

    • Search Google Scholar
    • Export Citation
  • Xue, Y. K., , K. N. Liou, , and A. Kasahara, 1990: Investigation of biogeophysical feedback on the African climate using a two-dimensional model. J. Climate, 3 , 337352.

    • Search Google Scholar
    • Export Citation
  • Xue, Y. K., , H. M. H. Juang, , W. P. Li, , S. Prince, , R. DeFries, , Y. Jiao, , and R. Vasic, 2004: Role of land surface processes in monsoon development: East Asia and West Africa. J. Geophys. Res., 109 , D03105. doi:10.1029/2003JD003556.

    • Search Google Scholar
    • Export Citation
  • Zhang, G. J., , and N. A. McFarlane, 1995: Sensitivity of climate simulations to the parameterization of cumulus convection in the Canadian Climate Centre general circulation model. Atmos.–Ocean, 33 , 407446.

    • Search Google Scholar
    • Export Citation
  • Zhang, H., , A. Henderson-Sellers, , and K. McGuffie, 1996a: Impacts of tropical deforestation. Part I: Process analysis of local climatic change. J. Climate, 9 , 14971517.

    • Search Google Scholar
    • Export Citation
  • Zhang, H., , K. McGuffie, , and A. Henderson-Sellers, 1996b: Impacts of tropical deforestation. Part II: The role of large-scale dynamics. J. Climate, 9 , 24982521.

    • Search Google Scholar
    • Export Citation
  • Zheng, X. Y., , and E. A. B. Eltahir, 1997: The response to deforestation and desertification in a model of West African monsoons. Geophys. Res. Lett., 24 , 155158.

    • Search Google Scholar
    • Export Citation
  • Zheng, X. Y., , and E. A. B. Eltahir, 1998: The role of vegetation in the dynamics of West African monsoons. J. Climate, 11 , 20782096.

Fig. 1.
Fig. 1.

Global land-cover map from Matthews (1983), emphasizing with rectangles the three regions in which tropical forests (black) are replaced with a mixture of shrubs and grassland in our deforestation experiments. Note that all areas within the boxes are converted (including places like Panama), but the changes are dominated by the Amazon, Africa, and Southeast Asia. Original (color) land-cover map and legend can be found in Matthews (1983).

Citation: Journal of Climate 22, 5; 10.1175/2008JCLI2157.1

Fig. 2.
Fig. 2.

Seasonal and annual mean precipitation (mm day−1) for (left) the GPCP dataset and (right) differences between GM II and the GPCP. Seasonal averages are December–Februrary (DJF), March–May (MAM), June–August (JJA), and September–November (SON). Note that the GPCP data are interpolated to the model grid.

Citation: Journal of Climate 22, 5; 10.1175/2008JCLI2157.1

Fig. 2.
Fig. 2.

(Continued) Seasonal and annual mean precipitation differences (mm day−1) between (left) AM and (right) CCSM and the GPCP dataset.

Citation: Journal of Climate 22, 5; 10.1175/2008JCLI2157.1

Fig. 3.
Fig. 3.

Temporal standard deviation (mm day−1) of annual (ANN) and seasonal precipitation for the (left) NCEP reanalysis dataset and (right) GM II.

Citation: Journal of Climate 22, 5; 10.1175/2008JCLI2157.1

Fig. 3.
Fig. 3.

(Continued) Temporal standard deviation (mm day−1) of ANN and seasonal precipitation for (left) AM and (right) CCSM.

Citation: Journal of Climate 22, 5; 10.1175/2008JCLI2157.1

Fig. 4.
Fig. 4.

Worldwide locations where precipitation has either significantly (top) decreased or (bottom) increased during a period of at least three months of the year, as a result of tropical deforestation as simulated by an MME of three GCMs: GM II, AM, and CCSM. The MME mean annual cycle of precipitation (mm day−1) for the control (blue) and deforested (red) cases at continental locations most severly affected by the deforestation is also represented. The color scale indicates the number of months registering a statistically significant change (Student’s t test 95% significance level) during the annual cycle.

Citation: Journal of Climate 22, 5; 10.1175/2008JCLI2157.1

Fig. 5.
Fig. 5.

Worldwide locations where precipitation has significantly (left) increased and (right) decreased during a period of at least three months of the year, as a result of tropical deforestation as simulated by (top) GM II, (middle) AM, and (bottom) CCSM. The color scale indicates the number of months registering a statistically significant change (Student’s t test 95% significance level) during the annual cycle.

Citation: Journal of Climate 22, 5; 10.1175/2008JCLI2157.1

Fig. 6.
Fig. 6.

Ranking of the “true” ensemble among the “false” ensembles for the SITES statistics calculated for the MME and all three models individually. Both precipitation (black) and 247-mb geopotential (white) are shown (top) for the tropics and (bottom) for northern midlatitudes only.

Citation: Journal of Climate 22, 5; 10.1175/2008JCLI2157.1

Fig. 7.
Fig. 7.

March geopotential change in the MME: Contour interval is 10 m, zero contour not shown. The black curve indicates 95% significance. Contour indicates geopotential changes at the 95% significance level, and the contour interval is 10 m (zero contour not drawn).

Citation: Journal of Climate 22, 5; 10.1175/2008JCLI2157.1

Fig. 8.
Fig. 8.

Vertically averaged meridional potential energy flux (thick line) and precipitation (thin line) for the solstice and equinox months for (a) the control runs MME, and for the differences between deforested and control runs in (b) GM II, (c) AM, and (d) CCSM.

Citation: Journal of Climate 22, 5; 10.1175/2008JCLI2157.1

Table 1.

Mean seasonal and annual precipitation biases (mm day−1) between GISS Model II (GM II), AM model (AM), CCSM, and the four datasets: Legates/MSU (LM), Willmott–Matsuura–Legates (WML), GPCP, and CMAP, globally and for the three tropical regions of the Amazon, central Africa, and Southeast Asia, defined as the land portion of boxes in Fig. 1. Negative values are in boldface.

Table 1.
Save