• View raw image

• View raw image

• View raw image

• View raw image

• View raw image

• View raw image

• View raw image

• View raw image

• View raw image

• View raw image

• View raw image

1. Introduction

The earth’s climate responds to a variety of internal and external forcing terms (Baede et al. 2001). There is very high confidence that the net effect of human activities since 1750 has caused warming and that the amount of warming is large relative to changes in solar irradiance (Forster et al. 2007). The 2001 and 2007 reports by the Intergovernmental Panel on Climate Change (IPCC) focus principally on the impact of increasing greenhouse gases, but both reports also identify other forcing terms, including land cover change.

The significance of land cover change on large-scale climate is disputed. Land cover change (LCC) alters the surface energy and moisture balances as a result of changes in rooting depth and canopy height and structure of vegetation (Findell et al. 2007). The partitioning of available energy between sensible and latent heat fluxes is sensitive to land cover conditions, as is the partitioning of precipitation between runoff and evaporation (Pitman 2003). Removal of natural vegetation tends to decrease rooting depths, which reduces moisture availability, leading to increased sensible heat flux and warmer and deeper boundary layers. In contrast, reforestation typically increases the latent heat flux via transpiration and interception loss, which cools and moistens the boundary layer (Betts et al. 1996). These changes in the root and canopy structure of land cover are accompanied by changes in albedo, which impacts radiative forcing (Forster et al. 2007;Findell et al. 2007). The change associated with deforestation is normally a decrease in radiative forcing via an increase in albedo (hence this acts to cool the surface through the radiation balance), and the albedo increase can be amplified via a positive feedback with snow (Betts 2000). Models that emphasize the albedo change rather than the changes in available moisture predict a regional cooling associated with deforestation in the midlatitudes (e.g.,Bonan 1999;Govindasamy et al. 2001;Matthews et al. 2003). However, studies that also include the changes to the partitioning of available energy tend to predict a regional warming in the midlatitudes due to deforestation (e.g.,Findell et al. 2007).

The exploration of the impact of land cover change on the atmosphere is one of the oldest experiments conducted with climate models (Henderson-Sellers and Gornitz 1984). A vast literature has developed, focused on the impact of tropical deforestation (Dickinson and Henderson-Sellers 1988;McGuffie et al. 1995;Sud et al. 1996;Chase et al. 2000;Avissar and Werth 2005;Findell et al. 2006) and the impact of global LCC, including the impact of temperate deforestation (Bonan 1999;Govindasamy et al. 2001;Bounoua et al. 2002;Zhao and Pitman 2002;Findell et al. 2007). The impact of perturbations to the land cover under enhanced greenhouse gas levels has also been examined (Costa and Foley 2000;Zhao et al. 2001;Bala et al. 2007).

There is no doubt that large-scale land cover change strongly affects the regional climate over the areas subject to land clearance. Model experiments conducted using global climate models and regional climate models confirm this (Chase et al. 1999;Wang et al. 2003;Hahmann and Dickinson 1997;Heck et al. 2001;Narisma and Pitman 2003). However, there is no agreement on the impact of LCC on climate in areas far from the perturbation, nor has an assessment of the impact of LCC relative to other more well-studied forcing terms been systematic. The examination of the regional significance of land cover change in comparison to increased greenhouse gases has been undertaken (Pitman and Zhao 2000;Bala et al. 2007;Betts 2001;Betts et al. 2004) but more work on this is needed. The issue of remote teleconnections—where LCC is in, for example, South America—is used to explain changes over another continent has been addressed many times. Some authors find clear teleconnections (e.g.,Gedney and Valdes 2000;Werth and Avissar 2002,2005), while others do not (Findell et al. 2007).

While large-scale land cover change can affect regional climate, it is useful to place the scale of this impact into some context. There are two important questions that remain outstanding. First, do SST anomalies resulting from natural variability [e.g., El Niño–Southern Oscillation (ENSO), North Atlantic Oscillation (NAO)] or anthropogenic factors (i.e., the twentieth-century warming trend) impact regional climates more or less than land cover change? Clearly, SST anomalies are generally remote from the continents where people live; hence the effect of SST anomalies on regional climate compared to LCC can provide a relative measure that prioritizes where effort might best be expended to improve regional simulations. Second, can LCC anomalies cause remote, nonlocal changes in climate? ENSO, for example, manifests as a well-known global forcing on climate. If land cover change can be shown to affect the global climate on a scale equivalent to ENSO or other major SSTs anomalies, then the omission of land cover change in global projections (e.g., those used in the fourth assessment report of the IPCC) would be very limiting.

The purpose of this study is, therefore, to compare the relative climatic impacts of these different forcing fields—SST anomalies and historical LCC—with experiments using the same general circulation model and the same experimental framework. This will help resolve open questions regarding the strength of the climatic response to LCC, although the inherently different time scales associated with the different forcing fields must be considered. The LCC scenario used in these experiments represents many centuries of anthropogenic alterations to the earth’s surface, while an SST anomaly such as ENSO is a transient phenomenon. These issues will be discussed in more detail insection 5. First, the model and the experiments are briefly described in the next section. Globally averaged impacts are then considered, followed by a detailed consideration of the regional response to LCC and to imposed SST anomaly patterns. A discussion follows insection 5, and conclusions are given insection 6.

2. Model description and experimental design

The climate model is composed of an atmospheric general circulation model coupled to an ocean with prescribed SST. The model grid cells are 2° latitude × 2.5° longitude, with 24 vertical levels in the atmosphere. The model has a seasonal and a diurnal cycle of insolation. The model’s treatment of the atmosphere and land is the same as that in the Climate Model version 2.1 (CM2.1;Delworth et al. 2006). General characteristics of the land model are described below, since it is such a crucial factor in this series of experiments. Additional model documentation is available from the Geophysical Fluid Dynamics Laboratory (GFDL) Web site (online athttp://nomads.gfdl.noaa.gov).

3. Global-scale impacts

Table 3 shows, for each of the four experiments, the percent of global area and land area with significantly different annual mean 2-m air temperature from the control simulation. Precipitation, evaporation, sensible heat flux, root-zone soil moisture, and runoff are also included. Additionally,Fig. 3 shows the magnitude of the globally averaged annual mean differences from the control experiment for nine variables for each of the four perturbation experiments, whileFigs. 4 and5 show global maps of the annual mean differences for 2-m air temperature and precipitation, respectively.

The near-global SST forcing in the WTr experiment produces a global surface air temperature response that passes the 95%-level modifiedt test in 72% of the globe and over 43% of land area (Table 3;Fig. 4a). The average surface air temperature is 0.18°C warmer in the WTr experiment than in the control run (Fig. 3a). Globally averaged precipitation and evaporation both increase by about 0.0012 cm day⁻¹ (Figs. 3d,e), while the sensible heat flux decreases by a statistically insignificant amount (about 0.04 W m⁻²;Fig. 3f). Though the global average precipitation change is statistically significant, only 14% of the total area and 9% of the land area have significantly different annual precipitation, and the magnitude of the change is nowhere greater than 0.1 cm day⁻¹ (Fig. 5a). Evaporation and sensible heat flux are altered over about a quarter of the globe (27% and 21%, respectively), but the impact only reaches about 10% of land area. The strong temperature response to the WTr perturbation (affecting 43% of land area) compared to the weaker response in precipitation (9%) is an indication of why it is more difficult to detect changes in rainfall and attribute them to global warming (e.g.,Zhang et al. 2007).

The WENSO SST forcing also has broad, global impacts, with 60% of global area and 55% of land area experiencing significantly different temperatures than the control simulation. Unlike the WTr experiment, the surface temperature impact is not just a warming signal: much of the United States, Mexico, eastern Russia, and northern China are significantly cooler (by more than 1°C;Fig. 4b). Larger areas, however, do show warming in excess of 1°C (e.g., northern North America and northern South America, Australia, India, and eastern Africa;Fig. 4b). The global average annual mean temperature is 0.16°C warmer than the control simulation (Fig. 3a). This value gives an indication of the impact that an El Niño event has on the often-cited metric of global mean air temperature, for example, the high temperatures noted in the long-term record in 1998 (see, e.g.,Thompson et al. 2008). The WENSO forcing produces an impact on other variables that is significant over more than half the globe (Table 3) and leads to significant globally averaged differences in eight of the nine variables shown inFig. 3 (all but net radiation at the surface). Even over land, where the SST forcing is obviously not local, 39%–45% of the area has significantly different precipitation, evaporation, sensible heat flux, and soil moisture (Table 3). In general, the nonsignificant land areas for these four variables are north of 50°N in North America and north of 40°N in Asia (precipitation shown inFig. 4b). A strong precipitation response is obvious in the Indian and Pacific Oceans.

The WNAO forcing is associated with warming in excess of 0.5°C over the region of the SST anomaly, much of western Asia through the Middle East, and throughout central North America (Fig. 4c). Modest warming is also seen over much of South America and Africa. In total, 37% of global area and 41% of land area have a significantly different 2-m air temperature from the control simulation, with a statistically significant globally averaged annual mean increase of 0.09°C (Fig. 3a). The areas of significant differences are smaller for other variables: 19%–27% for global precipitation, evaporation, and sensible heat flux, and 13%–17% for land-only values of those three variables, soil moisture, and runoff (Table 3). Surprisingly, the magnitude of the globally averaged annual mean increase in both precipitation and evaporation is equivalent to that of the WENSO experiment (about 0.0007 cm day⁻¹;Fig. 3), and the mean global temperature increase of 0.09°C is more than half the 0.16°C of the WENSO experiment.

In contrast to the broad impact of each of the SST anomalies, the LCC experiment produces statistically significant surface air temperature changes over only 8% of global area and 13% of land area. The globally averaged temperature difference of 0.023°C is not statistically significant at the 95% level (Fig. 3a). For the other variables listed inTable 3, only between 5% and 10% of global area shows a significant difference from the control simulation. Given the discussion of field significance insection 2c, it is unlikely that the global impact of the LCC experiment would pass field significance requirements. The land-only values inTable 3 show that 8%–14% of land area is significantly different from the control run. Two-meter air temperature, evaporation, and root-zone soil moisture all exceed the 12.5% threshold needed for a field with 40 dof to be statistically significant at the 95% level. Nevertheless, this 8%–14% of land area passing the significance test should be considered alongside the fact that 11.6% of the land surface has altered vegetation in the LCC experiment. Given the bulk of previous research indicating that this sort of LCC should induce a local response (Betts et al. 1996;Pitman 2003), the 8%–14% range might simply indicate the direct response to LCC and does not necessarily imply the existence of teleconnections beyond the altered regions. However, the globally averaged values of soil moisture, surface runoff, precipitation, evaporation, longwave heat flux, and net radiation at the surface are each significantly different from the control run values at the 95% confidence level, suggesting that, in the areas where changes occur, their magnitudes are relatively large and consistent in sign (Fig. 3).

Thus, in broad terms,Table 3 shows that the WTr, the WENSO, and the WNAO SST forcings drive a large-scale response in temperature. The scales of the changes in temperature globally and over land are large, exceeding 40% of land area at a 95% significance level. In contrast, the land cover change anomaly causes a much smaller impact at the scale of the continents. The WENSO forcing dominates the precipitation impact over land in terms of spatial extent, though all forcings lead to statistically significant changes in the global average annual mean precipitation rate (Fig. 3). When we focus on terrestrial quantities such as evaporation, sensible heat, soil moisture, and runoff, while the WENSO pattern remains the forcing with the broadest impact, LCC is at least equivalent to the WTr and WNAO forcings over land.

While thecontinental-scale impacts of LCC are relatively small, though statistically significant in some cases, the impacts of LCC are much broader when analyzed regionally. The LCC perturbation is likely to be most important over the regions where the perturbation is imposed. We therefore assess the impact of LCC over those regions shown inFig. 2 relative to the regional impact of the SST anomalies. Given that the imposed SST anomalies are commonly understood to have important impacts on regional climate, identifying the impact of LCC relative to these SST anomalies over the regions of LCC provides a clear demonstration of the significance of LCC to regional climate. However, regions of LCC may not necessarily coincide with regions of extensive natural variability associated with the SST patterns considered here, so this selection of regions may bias the results toward LCC significance. In particular,Figs. 4 and5 show that precipitation and temperature in northern South America and the islands of Oceania are sensitive to the imposed SST anomalies. These regions are not considered here because of a lack of large-scale LCC in the depiction shown inFig. 2. Furthermore,Figs. 4 and5 also show that LCC does lead to some significant differences in areas without LCC, most notably the broad area of warming north of the Black and Caspian Seas (Fig. 4d). Given that there is no statistically significant accompanying change in precipitation (Fig. 5d), evaporation, sensible heat flux, soil moisture, or runoff (not shown), it is likely that the surface temperature signal results from being in the predominately downwind direction from European regions with substantial LCC.

4. Regional-scale impacts

Figures 6 –10 show regionally averaged seasonal cycles of seven variables (2-m air temperature, root-zone soil moisture, precipitation, evaporation, sensible heat flux, net radiation at the surface, and surface runoff) for five areas with substantial land cover conversion (see boxes inFig. 2b). In addition to the seasonal cycles for each of the five model simulations, monthly differences from the control run are displayed, along with the number of months that each experiment is significantly different from the control, according to the 95%-level modifiedt test. Conservatively, we might expect one month per year to pass this significance test by chance for each experiment. We now discuss each of these five regions in detail.

5. Discussion

At the global scale, our results show significant changes in air temperature, rainfall, evaporation, and surface runoff induced by each of the SST anomalies. Over land only, the WENSO and WNAO experiments dominate the perturbation to the precipitation and temperature fields with at least 2 and 3 times the land area, respectively, affected by each SST anomaly in comparison to LCC (Table 3). Thus, the land-only impact from LCC on temperature and precipitation covers a small area relative to the SST anomalies and is at the margin of statistical field significance. However, if changes in surface quantities (evaporation, sensible heat flux, soil moisture, runoff) are assessed, LCC impacts about as much area as the WNAO and WTr perturbations (Table 3), and the magnitude of these impacts is large enough to lead to statistically significant differences in the annual global mean values of soil moisture, surface runoff, precipitation, and evaporation (Fig. 3).

Despite these relatively large magnitude changes in regions of LCC, the results presented here show that changes to terrestrial quantities such as soil moisture and sensible and latent heat fluxes that result from LCC rarely propagate into the atmosphere in a significant way.Figure 11 shows a clear example of how dramatically the atmosphere over the Indian Ocean and Southeast Asia is altered by the WENSO experiment (Fig. 11b). The other forcing scenarios do not significantly alter the 850-mb wind or specific humidity fields (Figs. 11a,c,d), despite the substantial impact of LCC on evaporation, sensible heat flux, soil moisture, runoff, and even surface temperature in regions A1, A2, and T1 (Figs. 7 –9).Figures 7 –9 show that LCC and WENSO are essentially equally important at the terrestrial surface in these three regions, butFig. 11 shows that while a broad atmospheric response is present in the ENSO-like experiment, the LCC impacts do not propagate beyond the surface. Nevertheless, the impact of LCC on regional surface hydrology is highly relevant to society at large, even when these changes do not impact local or remote atmospheric fields.

Our experiments were designed to place the four forcing scenarios in a common framework. However, there are important differences in the time scales associated with the four scenarios. The SST trend and the LCC pattern both represent longer time-scale processes than the transient signals of an ENSO or an NAO anomaly. The trend anomaly is representative of the century of data (1901–2004) from which the SST forcing patterns were derived. The land cover change scenario is representative of cumulative land cover conversions spread out over several centuries, and our focus of the regional analyses on areas with known LCC may overemphasize the impact of LCC. Despite these differences, the results presented here show that LCC must be accounted for to reliably simulate terrestrial quantities, and they reinforce the need for proper representation of the land surface when considering regional climate.

Although the scale of the SST anomalies in the ENSO and NAO patterns is realistic, it is a simplification to apply these anomalies in the positive phase consistently for many decades. Given the perpetual nature of these anomalies over the 60-yr experiment, we have compared the first 15 yr to the last 15 yr of the WENSO experiment to determine if the experiment can be approximately considered an independent 60-member ensemble of 1-yr ENSO-forced experiments, or whether a cumulative atmospheric response develops wherein precipitation and/or temperature anomalies build up during the course of the experiment. Differenced fields for precipitation and surface air temperature between the start and end of the 60-yr run (not shown) confirm the former, with no significant global or regional-scale coherent differences developing over the course of the experiment. Thus, we can take this experiment to be equivalent to a 60-member ensemble set, consistent with the short intraseasonal time scale for the atmosphere to respond to global SST forcing.

6. Conclusions

Experiments with the atmosphere and land components of the GFDL CM2.1 have been used to compare the relative impacts of anthropogenic land cover conversion (primarily a conversion of forests to grasslands or agriculture) to three realistic SST perturbations: a signal of the global warming SST trend, and two experiments forced by patterns of interannual SST variability (signals similar to warm-phase ENSO and NAO events). The LCC experiment includes biophysical changes resulting from conversion of the vegetation type (changes to the surface albedo, root-zone soil water capacity, stomatal resistance, and roughness length) but does not include impacts on the carbon cycle. Human water-management practices are also not considered in these experiments. Additionally, experimental results are often model dependent, and this study considers output from only one climate model.

Model results show that LCC has a small but significant impact on annual-average global precipitation, evaporation, net radiation, longwave radiation, soil moisture, and runoff. These impacts are largely collocated with regions of land cover disturbance: they do not propagate into the atmosphere to change remote regions, though temperature effects are significant downwind of the altered European region. The SST anomalies, on the other hand, tend to impact a much larger percentage of the globe, with between 30% and 55% of the land surface significantly altered by the warm-phase ENSO perturbation.

In the regions where the land surface is altered, the impact of LCC can be equally or more important than the SST forcing patterns in determining the seasonal cycle of the surface water and energy balance. Indeed, in many regions, the land cover change experiment had the greatest impact on surface fields such as runoff and soil moisture. In some regions, these fields were significantly affected even when the regional climate was not. Thus, we provide a context for the impacts of LCC on climate: namely, strong regional-scale impacts that can significantly change globally averaged fields but that rarely propagate beyond the disturbed regions. This suggests that LCC must be accounted for to reliably simulate terrestrial quantities over regions of intensive LCC. Furthermore, these results demonstrate that proper representation of the land cover condition is critical in the design of climate model experiments, particularly if those model results are to be used as input to hydrological models and/or higher-resolution climate models, or for regional-scale assessments of climate change or variability.

This points to the need to include significant forcing terms in future climate projections and to the need for a methodology for determining which forcing terms should be included in such climate projections. We suggest that placing any experiment into the context of better-known perturbations may allow for the separation and identification of those processes or parameters that matter in a relative sense. We therefore propose that a set of benchmark simulations be designed to provide the context against which new perturbation experiments can be assessed. This will enable modelers to show that an experiment with any given model produces a regional or global anomaly that is important relative to a known signal. It will also take a small step in dealing with the uncertainty in land surface modeling made apparent byKoster et al. (2006) in their identification of challenges associated with the unknown coupling strength of climate models. We propose experiments to evaluate the relative impacts of three specific perturbations:

1. The impact of a doubling of atmospheric carbon dioxide. This is commonly undertaken and provides a global signal that is societally important and of broad interest.

2. The impact of an ENSO-like SST anomaly. This is a large-scale and recurring SST anomaly that is known to have a globally significant fingerprint.

3. The impact of a LCC anomaly. LCC is highly regionalized and is largely coincident with the distribution of the earth’s human population. Such experiments provide a regional pattern of forcing that we have demonstrated to be important for the climate over areas of dense human population. Furthermore, the impact of LCC is highly model dependent, as discussed in the introduction and inFindell et al. (2007).

For a proper assessment of impacts, experiment 1 requires an interactive ocean model (see, e.g.,Findell and Delworth 2005), while experiments 2 and 3 can be performed—to first order—with an atmosphere-only model, using fixed SSTs, as in these experiments. Any user-defined anomaly that can be shown to be as globally significant as 1 or 2 is clearly important. However, any user-defined anomaly that impacts those regions of LCC as significantly as LCC itself is also extremely important and worthy of inclusion in models. This would not, of course, be the singular criterion for the relevance of any given forcing. It would, however provide the basis for determining those forcings worthy of immediate attention in coupled climate models.