Biogeochemical and Biophysical Responses of the Land Surface to a Sustained Thermohaline Circulation Weakening

Home , Biosphere model, Climate change scenario, HadCM3

1NOVEMBER 2004 HIGGINS 4135

PAUL A. T. HIGGINS University of California, Berkeley, Berkeley, California

(Manuscript received 20 November 2003, in ®nal form 14 May 2004)

ABSTRACT Biotic responses to climate change may constitute signi®cant feedbacks to the climate system by altering biogeochemistry (e.g., carbon storage) or biophysics (i.e., albedo, evapotranspiration, and roughness length) at the land surface. Accurate projection of future climate change depends on proper accounting of these biological feedbacks. Similarly, projections of future climate change must include the potential for nonlinear responses such as thermohaline circulation (THC) weakening, which is increasingly evident in paleoclimate reconstructions and model experiments. This article uses of¯ine simulations with the Integrated Biosphere Simulator (IBIS) to determine long-term biophysical and biogeochemical responses to climate patterns generated by the third Hadley Centre Coupled Ocean±Atmosphere General Circulation Model (HadCM3) under forced THC weakening. Total land surface carbon storage decreases by 0.5% in response to THC weakening, suggesting that the biogeochemical response would not constitute a signi®cant climate feedback under this climate change scenario. In contrast, large regional and local changes in leaf area index (LAI) suggest that biophysical responses may constitute signi®cant feedbacks to at least local and regional climate. Indeed, the LAI responses do lead to changes in midday direct and diffuse beam albedo over large regions of the land surface. As a result, there are large local and regional changes in the land surface's capacity to absorb solar radiation. Changes in energy partitioning between sensible and latent heat ¯uxes also occur. However, the change in latent heat ¯ux from the land surface is primarily attributable to changes in precipitation that occur under forced THC weakening and not a result of the subsequent changes in vegetation.

1. Introduction system (Betts et al. 1997; Cox et al. 2000; Levis et al. 2000). In some region-speci®c cases these biotic respons- Structural and functional ecosystem responses to cli- es magnify climate changes by as much as a factor of 4 mate change may constitute signi®cant feedbacks to the (Ganopolski et al. 1998) and can greatly amplify even climate system through changes in biophysical and biogeochemical characteristics of the land surface (Field small changes in external forcing (Claussen et al. 1999). and Avissar 1998; Pielke et al. 1998; Sellers et al. 1997). These atmosphere±biosphere interactions may be partic- Important biophysical characteristics of the land surface ularly important in arid regions (Foley et al. 2003; Hig- include 1) albedo, or re¯ectivity of incoming solar ra- gins et al. 2002; Nicholson 2000) as a consequence of diation, which depends on vegetation type and density; greater climate sensitivity to changes in transpiration, 2) evapotranspiration, which in¯uences atmospheric roughness length, and albedo (Nicholson 2000; Wang and moisture and partitioning of surface energy between Eltahir 2000a,c). As a result, projection of future climate sensible and latent heat (leaf area, rooting depth, and change depends on proper accounting of biophysical and photosynthetic rates all in¯uence rates of transpiration); biogeochemical responses at the land surface. and 3) surface roughness, which in¯uences friction be- Paleoclimate and model studies also increasingly sug- tween the atmosphere and the land surface and atmo- gest that changes in ocean circulation can trigger non- spheric mixing in the boundary layer. Biogeochemical linear climate responses to external forcing (Broecker changes (i.e., the carbon and nutrient cycles) can alter 2003; Clark et al. 2002; Hall and Stouffer 2001; Knutti carbon storage at the land surface and therefore in¯u- and Stocker 2002; Rahmstorf and Ganopolski 1999). ence atmospheric concentrations of greenhouse gases. For example, the paleoclimate record suggests that at Therefore, biophysical and biogeochemical responses least three modes of the ocean's thermohaline circulation can amplify or dampen external changes to the climate (THC) are stable: 1) the modern mode, 2) a weaker mode that was common during the previous glacial period, and 3) an ``off mode'' (Alley and Clark 1999). In Corresponding author address: Dr. Paul Higgins, University of California, Berkeley, 151 Hilgard Hall, Berkeley, CA 94720-3110. some models, multiple stable THC equilibria are also E-mail: [email protected] possible with the equilibria attained depending on tem-

᭧ 2004 American Meteorological Society

Unauthenticated | Downloaded 09/27/21 01:26 PM UTC 4136 JOURNAL OF CLIMATE VOLUME 17 perature and salt conditions, details of surface forcing, sition, land-use patterns, and species invasions further initial THC strength, and additional model parameters complicate projections of ecosystem responses to global (Dijkstra and Neelin 2000; Rahmstorf 1996; Schneider change disturbance. This article presents simulated bio- and Thompson 2000; Stocker and Schmittner 1997). geochemical and biophysical responses of the land sur- Projection of future climate change requires assessment face to a sustained THC weakening scenario under con- of and accounting for the biophysical and biogeochem- stant GHG concentrations. This is a necessary ®rst step ical responses of the land surface that could result from for projecting biological responses and feedbacks to fu- such nonlinear climate change. ture climate changes. Climate simulations performed with the third Hadley Centre Coupled Ocean±Atmosphere General Circula- tion Model (HadCM3) under preindustrial greenhouse 2. Methods gas (GHG) forcing suggest that THC weakening could a. Climate scenarios alter temperature and precipitation throughout the world (Vellinga and Wood 2002; Vellinga et al. 2002). Vellinga I use four previously described climate scenarios un- and Wood (2002) imposed a large salinity disruption in der constant CO2 (280 ppm) that vary in monthly tem- the North Atlantic, which caused an initial decrease in perature and precipitation (Higgins and Vellinga 2004). HadCM3's THC from 20 Sverdrups (1 Sv ϵ 106 m3 sϪ1) Temperature and precipitation scenarios are based on to 0 Sv in the North Atlantic at 48ЊN and 666-m depth. monthly means from the third decade of two simulations Over the next 120 yr of the simulation, THC recovers from HadCM3 (described above), a control simulation to its full strength as the off mode is not stable in in which THC remains stable and active, and a THC HadCM3 under preindustrial climate conditions and weakening simulation in which a prescribed salinity de- model parameters. During the ®rst decade annually av- cline greatly reduces THC (Vellinga and Wood 2002; eraged temperature decreases by 6Њ±8ЊC in the atmo- Vellinga et al. 2002). The four climate scenarios are 1) sphere above the northwest Atlantic while THC strength a control scenario in which temperature and precipita- remains below roughly 6 Sv. Over subsequent decades, tion are taken from HadCM3's control simulation, 2) a climate system teleconnections slowly spread through- weakened THC scenario in which both temperature and out the world while the strength of THC continues to precipitation are taken from HadCM3's weakened THC recover. During the third decade, portions of western simulation, 3) weakened THC temperature combined Europe cool by over 3Њ with 1Њ±2ЊC cooling occurring with HadCM3's control precipitation, and 4) weakened throughout much of Europe and in parts of Asia, north- THC precipitation combined with HadCM3's control ern Africa, and North America. Large shifts in precip- temperature. Since temperature and precipitation in¯u- itation in the Tropics also occur during the third decade ence one another (temperature affects precipitation by as a consequence of a shift in the intertropical conver- altering atmospheric circulation, relative humidity, and gence zone (ITCZ; Vellinga and Wood 2002; Vellinga evapotranspiration, while precipitation affects temper- et al. 2002). ature by altering the net energy available to the surface Attribution of these climate responses to THC weak- and the partitioning of that energy between sensible and ening requires statistical distinction of the forced re- latent heat ¯uxes), these last two scenarios are not phys- sponse (signal) from the internal variability of the cli- ically consistent. By isolating temperature and precip- mate system (noise) (Chervin and Schneider 1976). The itation changes, however, this approach provides a mea- HadCM3 climate experiments identify statistically sig- sure of ecosystem sensitivity to each. ni®cant climate responses to THC weakening as anom- Climate in HadCM3 does not reach equilibrium for alies that exceed 2 times the decadal mean standard THC weakening as recovery occurs over subsequent deviation from the control, sampled over 300 yr (Vel- decades in this model implementation. As previously linga and Wood 2002; Vellinga et al. 2002). By this test, described (Higgins and Vellinga 2004), this study uses the Northern Hemisphere temperature changes and the the third decade of the transient HadCM3 scenarios in ITCZ shift are both signi®cant. order to balance the ongoing THC recovery with the Previous examination of ecosystem responses to this slow spread of climate teleconnections throughout the scenario demonstrates that large structural and func- world, even though anomalies around the North Atlantic tional changes may occur throughout the world in re- are strongest in the ®rst decade of the experiment. This sponse to thermohaline circulation weakening (Higgins approach assumes the third decade is generally consis- and Vellinga 2004). These broadly distributed ecosys- tent with a plausible long-term shift to a stable weak tem responses are potentially suf®cient to alter bio- THC equilibrium and holds climate constant for the eco- physical and biogeochemical characteristics of the land system simulations, described below. Alternatively, eco- surface and feedback to the climate system. Of course, system responses to the transient climate changes as- projection of future climate and biological system re- sociated with the HadCM3 simulated THC weakening sponses represents a complex integrated assessment of and recovery could have been examined. Doing so the global system (Higgins and Vellinga 2004). The di- would provide a more realistic examination of ecosys- rect effects of elevated carbon dioxide, nitrogen depo- tem responses over time to this particular climate sce-

Unauthenticated | Downloaded 09/27/21 01:26 PM UTC 1NOVEMBER 2004 HIGGINS 4137 nario, but would provide no opportunity to examine compared over an additional 120-month period added ecosystem responses to the long-term changes in THC to the end of the 200-yr simulation with differences demonstrated throughout the paleoclimate record and in determined by a t test. The t test is appropriate assuming other climate models. that vegetation responses do not provide longer-term memory of the interannual variability introduced by the weather generator. b. Ecosystem responses To test the potential implications of the changes in Structural and functional ecosystem responses to land surface albedo, rough estimates of the change in THC weakening are tested using the Integrated Bio- absorbed solar radiation (⌬R) are calculated globally sphere Simulator (IBIS) 2.1, a process-based dynamic and locally according to global ecosystem model (Foley et al. 1996; Kucharik et ⌬R ϭ S ϫ (Ϫ⌬␣ ϫ T ), (1) al. 2000). IBIS takes monthly climate input for average grnd T frac temperature, precipitation, relative humidity, cloudi- where Sgrnd is the amount of solar radiation reaching the ness, temperature range, wind speed, and number of land surface, ⌬␣T is the change in midday direct beam rainy days and uses a weather generator to produce daily land surface albedo (from above), and Tfrac is the land variability. Monthly average temperature and precipi- fraction. For local calculations Tfrac ϭ 1.0 while Sgrnd tation come from the four climate scenarios created from varies by latitude and season. Here Sgrnd is calculated HadCM3 data, all others are from historical climatology locally according to (New et al. 1999). IBIS also requires input data for S ϭ S cosZ ϫ (1 Ϫ ␣ ), (2) topography, soil texture (Global Soils Data Task Group grnd 0 atm Ϫ2 2000), and minimum temperature. where S 0 is the solar constant (1366 W m ), cosZ is The results presented here come from ®ve IBIS sim- the cosine of the zenith angle, calculated as described ulations: 1) dynamic vegetation forced by HadCM3's in (Washington and Parkinson 1986), and ␣atm is the control climate (control), 2) dynamic vegetation forced atmospheric albedo. Although ␣atm does vary due to an by HadCM3's weak THC temperature and precipitation uneven distribution of clouds, this calculation holds ␣atm (TP), 3) dynamic vegetation forced by HadCM3's weak constant at 0.23, the global average, since model data THC temperature and control precipitation (T only), 4) on cloud distributions are not available. As a result, this dynamic vegetation forced by HadCM3's weak THC approach overestimates the solar radiation absorbed by precipitation and control temperature (P only), and 5) the surface in areas with heavy cloud cover and under- static vegetation forced by HadCM3's weak THC tem- estimates surface radiation in areas with light cloud cov- perature and precipitation (static TP). In static vegeta- er. Furthermore, this approach ignores potential changes tion mode, IBIS holds the carbon stored in biomass (e.g., in cloud cover that could alter the amount of solar ra- leaf and structural tissues) constant. Thus, the leaf area diation that reaches the earth surface and thereby damp- for a particular location remains constant when leaves en these changes. Nevertheless, this rough estimate of are displayed. Drought and winter deciduous leaf dis- the change in absorbed radiation at the land surface play patterns may vary due to climate changes, however. provides a useful approximation of the importance of Each IBIS scenario begins with a 100-yr spin up un- the vegetation's responses. For the global calculation der HadCM3's control temperature and precipitation. At Tfrac ϭ 0.3, re¯ecting the amount of the earth surface Ϫ2 the end of the 100-yr spin up, the weak THC scenarios covered by land, Sgrnd ϭ 197 W m , and this rough experience a step change in climate while the control calculation assumes that land surface albedo changes is held constant. IBIS then runs for an additional 100- are evenly distributed with respect to latitude, season, yr to account for transient vegetation changes. and cloud cover. c. Analysis 3. Results Globally aggregated biomass, litter, soil, and total car- Total terrestrial carbon storage decreases by 0.5% in bon storage are compared using a 20-yr mean from the the TP scenario (Table 1). This small change results end of the 200-yr IBIS experiment for each of the four from the combination of a 4.5% decrease in live biomass dynamic vegetation scenarios. This approach shows and nearly compensating increases in carbon stored in long-term ecosystem responses to these climate scenar- soil and litter. The 1.9% reduction in total carbon storage ios but does not consider the pathway of those ecosys- in the P-only and the 1.1% increase in the T-only sce- tem changes. Changes in mean annual leaf area index narios show that the temperature and precipitation (LAI) and mean annual latent heat ¯ux are compared changes have small and opposing effects on total carbon between the control and TP simulations and between storage when viewed in isolation. the TP and static TP simulations over the last 10 yr of In contrast, large local and regional changes in LAI the 200-yr IBIS experiment with differences determined occur in the TP scenario relative to the control (Fig. by a t test. Mean annual midday direct and diffuse land 1a). LAI increases by roughly 1±3 m2 mϪ2 in eastern surface albedo for the control and TP simulations is South America and by 1±2 m2 mϪ2 throughout much of

Unauthenticated | Downloaded 09/27/21 01:26 PM UTC 4138 JOURNAL OF CLIMATE VOLUME 17 1507.6 1499.9 1478.6 1524.1 0.1 0.4 0.2 0.4 1091.3 1099.0 1075.4 1113.0 0.04 0.06 0.06 0.03

FIG. 1. Change in the global distribution of LAI (m2 mϪ2) under the TP simulation, relative to (a) LAI from the control simulation and (b) LAI from the static TP simulation. Only locations where 18.3 19.1 17.8 19.5 litter Std dev Soil carbon Std devsigni®cant Total changes occur are shown (P Ͻ 0.01 as measured by t test). Below-ground North America and in parts of sub-Saharan Africa. LAI decreases by roughly 1±4 m2 mϪ2 in northern South America, by 1±2 m2 mϪ2 in portions of Australia and 2 Ϫ2

0.05 0.04 0.03 0.06 western Africa, and by 0±2 m m throughout the Northern Hemisphere higher latitudes. Similar differences in LAI appear when the TP and static TP simu-

1. Terrestrial carbon storage. All values in Pg Carbon. lations are compared (Fig. 1b). Therefore, LAI responses to temperature and precipitation under unchanging

ABLE vegetation distribution do not confound comparisons be- T tween the TP and static TP scenarios designed to dis- litter Std dev 47.0 47.5 45.3 48.9 entangle vegetation- and climate-induced changes at the Aboveground land surface. The global distribution of annually averaged land surface albedo from the control simulation for direct (Fig. 2a) and diffuse (Fig. 2b) beam radiation broadly tracks the vegetation distribution in the Tropics and the com- 0.18 0.13 0.27 0.12 bination of vegetation patterns and snow cover in the higher latitudes. Land surface albedo values range from a low of roughly 0.1 in heavily vegetated regions to over 0.6 in some high-latitude regions. Comparison between direct beam midday albedo

351.0 334.3 340.1 342.7 from the TP and control simulations demonstrates that changes occur in a pattern broadly similar to that of the LAI responses (Fig. 3). For example, land surface albedo in the TP simulation decreases by up to roughly 0.09, and 0.15 in parts of eastern South America and in the center of North America, respectively (Fig. 3a). Control TP P only T only

Treatment Live biomass Std dev These absolute albedo changes constitute percentage changes of up to 40% and 50% for eastern South Amer-

Unauthenticated | Downloaded 09/27/21 01:26 PM UTC 1NOVEMBER 2004 HIGGINS 4139

FIG. 2. Midday albedo for the land surface in the control FIG. 4. Change in diffuse beam midday albedo for the TP simu- simulation for (a) direct and (b) diffuse beam radiation. lation, relative to the control in (a) absolute and (b) percentage terms. Only locations where signi®cant changes occur are shown (P Ͻ 0.01 as measured by t test).

ica and Northern America, respectively (Fig. 3b). In contrast, direct beam albedo increases by up to 0.09 in northern South America, 0.06 in parts of West Africa and in western Autsralia, 0.09 in high-latitude North America, and 0.18 in parts of Europe and Asia (Fig. 3a). These absolute albedo changes constitute percentage changes that range from a 50% reduction to an 80% increase (Fig. 3b). A similar pattern of albedo change occurs for diffuse beam radiation in both absolute (Fig. 4a) and percentage terms (Fig. 4b), except that diffuse beam albedo decreases in parts of northern South Amer- ica. Globally averaged direct beam land surface albedo equals 0.217 and 0.2208 for the control and TP simulations, respectively. This small increase in land surface albedo under THC weakening (0.0038) leads to a small decrease in globally averaged net surface radiation (0.23 WmϪ2). Large local and regional changes in absorbed surface radiation occur, however (Fig. 5). For example, absorbed solar radiation increases by more than 15 W mϪ2 in parts of eastern South America and North America. Large decreases also occur in northern South America (up to 20 W mϪ2), parts of the Sahel (5±15 W mϪ2), northern Australia (5±10 W mϪ2), Europe (5±20 W mϪ2), FIG. 3. Change in direct beam midday albedo for the TP simulation, Ϫ2 relative to the control in (a) absolute and (b) percentage terms. Only Asia (5±20 W m ), and high-latitude North America Ϫ2 locations where signi®cant changes occur are shown (P Ͻ 0.01 as (5±10 W m ). For comparison, the radiative forcing measured by t test). associated with anthropogenically greenhouse gas emis-

Unauthenticated | Downloaded 09/27/21 01:26 PM UTC 4140 JOURNAL OF CLIMATE VOLUME 17

FIG. 5. Change in absorbed solar radiation at the land surface (W mϪ2) for the TP simulation, relative to the control. Only locations where signi®cant changes in terrestrial albedo (P Ͻ 0.01 as measured by t test) occur are shown. sions to date is approximately 2.5 W mϪ2 (Houghton et al. 2001). Comparison of latent heat ¯ux at the surface between the TP and control simulations (Fig. 6a) shows a response generally similar to that of LAI and consistent with changes in precipitation. In particular, increases of roughly 10±40 W mϪ2 occur in eastern South America Ϫ2 with reductions of 10±50 W mϪ2 in northern South FIG. 6. Change in the latent heat ¯ux difference (W m ) for the Ϫ2 TP simulation, relative to (a) the latent heat ¯ux from the control America. Reductions of 10±30 W m occur in West simulation and (b) the latent heat ¯ux from the static TP simulation. Africa and Australia as do smaller reductions throughout Only locations where signi®cant changes occur are shown (P Ͻ 0.01 Europe and in parts of Asia and North America. How- as measured by t test). ever, few clear patterns of change in latent heat ¯ux are evident when the TP scenario is compared to the static TP scenario except for a small increase in eastern South and diffuse land surface albedo in some areas. Fully America (Fig. 6b). quantifying the climate implications of these changes would require studies using fully coupled climate and ecosystem models in order to close feedback loops. Nev- 4. Discussion ertheless, numerous studies show that such changes in The globally aggregated biogeochemical and bio- albedo can cause local and regional temperature and physical responses to this THC weakening scenario are precipitation changes (Copeland et al. 1996; Dickinson small. The 0.5% change in global carbon storage dem- and Kennedy 1992; Dirmeyer and Shukla 1996; Hah- onstrated here after 100 yr is roughly equivalent to the mann and Dickinson 1997; Lean and Rowntree 1997; annual anthropogenic emissions of carbon dioxide from Xue 1997; Zeng et al. 1996), suggesting that the re- 1990 to 2000, while the 0.23 W mϪ2 change in absorbed sponses shown here may also constitute important feed- solar radiation by the land surface is roughly an order backs to local and regional climate in those portions of of magnitude smaller than the ϳ2.5 W mϪ2 forcing due Australia, South America, and North America where to anthropogenic greenhouse gas emissions to date. changes occur or even more distantly through telecon- These small global-scale biogeochemical and biophys- nections. ical responses are somewhat surprising given the large Regional climates throughout the world may be more local and regional ecosystem responses demonstrated or less sensitive to changes in the land surface depending here and previously (Higgins and Vellinga 2004) but on patterns of atmospheric circulation, soil properties, are consistent with the ®nding that global aggregation topography, continent size, and latitude (Betts et al. of biomass and net primary productivity (NPP) masks 1997; Claussen 1998; Dirmeyer and Shukla 1996; Klei- large regional and local changes (Higgins and Vellinga don et al. 2000; Zhang et al. 1996). Climate in South 2004). America and in arid regions such as the West African The absence of important globally aggregated re- Sahel (10Њ±17.5ЊN, 15ЊW±15ЊE), where biotic responses sponses does not preclude signi®cant responses over appear strongest in these simulations, appears to be par- smaller spatial scales. Indeed, the large local and re- ticularly sensitive to land surface changes (Kleidon and gional changes in LAI lead to large changes in direct Heimann 1999, 2000; Wang and Eltahir 2000a,b,c).

Unauthenticated | Downloaded 09/27/21 01:26 PM UTC 1NOVEMBER 2004 HIGGINS 4141

Changes in energy partitioning also accompany the Claussen, M., 1998: On multiple solutions of the atmosphere-vege- change in LAI associated with ecosystem shifts, as dem- tation system in present-day climate. Global Change Biol., 4, 549±559. onstrated by the changes that occur in latent heat ¯ux. ÐÐ, C. Kubatzki, V. Brovkin, A. Ganopolski, P. Hoelzmann, and These ecosystem shifts occur due to changes in precip- H. J. Pachur, 1999: Simulation of an abrupt change in Saharan itation, however, which also affects evaporation and en- vegetation in the mid-Holocene. Geophys. Res. Lett., 26, 2037± ergy partitioning between sensible and latent heat. Com- 2040. parison between the dynamic and static vegetation sce- Copeland, J., R. Pielke, and T. Kittel, 1996: Potential climatic impacts of vegetation change: A regional modeling study. J. Geophys. narios demonstrates that the changes in precipitation Res., 101D, 7409±7418. constitute the primary driver for the changes in latent Cox, P., R. Betts, C. Jones, S. Spall, and I. Totterdell, 2000: Accel- heat ¯ux in the TP scenario and that vegetation re- eration of global warming due to carbon-cycle feedbacks in a sponses are not primarily responsible for the changes coupled climate model. Nature, 408, 184±187. in energy partitioning. Dickinson, R., and P. Kennedy, 1992: Impacts on regional climate of Amazon deforestation. Geophys. Res. Lett., 19, 1947±1950. Some caution is necessary in extrapolating these re- Dijkstra, H. A., and J. D. Neelin, 2000: Imperfections of the ther- sults to other climate change scenarios in general and mohaline circulation: Latitudinal asymmetry and preferred to THC weakening experiments in particular. Globally northern sinking. J. Climate, 13, 366±382. aggregated responses to the climate system may be more Dirmeyer, P., and J. Shukla, 1996: The effect on regional and global substantial to less modest changes in temperature or climate of expansion of the world's deserts. Quart. J. Roy. Me- teor. Soc., 122, 451±482. precipitation, which would likely accompany greater Field, C., and R. Avissar, 1998: Bidirectional interactions between weakening of THC or a complete THC collapse. Fur- the biosphere and the atmosphereÐIntroduction. Global Change thermore, climate changes due to anthropogenic emis- Biol., 4, 459±460. sions of greenhouse gases may amplify or dampen the Foley, J. A., I. C. Prentice, N. Ramankutty, S. Levis, D. Pollard, S. effects of THC weakening in region-speci®c patterns, Sitch, and A. Haxeltine, 1996: An integrated biosphere model of land surface processes, terrestrial carbon balance, and veg- while other global changes (e.g., the direct effects of etation dynamics. Global Biogeochem. Cycles, 10, 603±628. elevated carbon dioxide, nitrogen deposition, land use ÐÐ, M. T. Coe, M. Scheffer, and G. L. Wang, 2003: Regime shifts patterns, and species invasions) further complicate the in the Sahara and Sahel: Interactions between ecological and projection of ecosystem responses. Nevertheless, the re- climatic systems in northern Africa. Ecosystems, 6, 524±539. sults presented here suggest that important biogeochem- Ganopolski, A., C. Kubatzki, M. Claussen, V. Brovkin, and V. Pe- ical and biophysical responses may accompany THC toukhov, 1998: The in¯uence of vegetation±atmosphere±ocean interaction on climate during the mid-Holocene. Science, 280, weakening at least locally and regionally even when 1916±1919. globally aggregated responses remain small. Global Soils Data Task Group, 2000: Global Soil Data Products CD- ROM (IGBP-DIS). Oak Ridge National Laboratory Distributed Acknowledgments. PATH thanks the National Science Active Archive Center. Foundation and the Department of Energy's Global Hahmann, A., and R. Dickinson, 1997: RCCM2±BATS model over tropical South America: Applications to tropical deforestation. Change Education Program for funding. Steve Schnei- J. Climate, 10, 1944±1964. der, Chris Field, Hal Mooney, Liz Hadly, and two anon- Hall, A., and R. J. Stouffer, 2001: An abrupt climate event in a coupled ymous reviewers provided valuable comments and sug- ocean±atmosphere simulation without external forcing. Nature, gestions. Crown Copyright provided HadCM3 boundary 409, 171±174. climate scenarios with the help of Michael Vellinga. Higgins, P. A. T., and M. Vellinga, 2004: Ecosystem responses to abrupt climate change: Teleconnections, scale and the hydro- Ferret, a product of NOAA's Paci®c Marine Environ- logical cycle. Climatic Change, 64, 127±142. mental Laboratory, facilitated data visualization and ÐÐ, M. D. Mastrandrea, and S. H. Schneider, 2002: Dynamics of graphics (www.ferret.noaa.gov). John Foley, Dierk Pol- climate and ecosystem coupling: Abrupt changes and multiple zin, and the Center for Sustainability and the Global equilibria. Philos. Trans. Roy. Soc. London, 357, 647±655. Environment provided IBIS. Houghton, J. T., Y. Ding, D. J. Griggs, M. Noguer, P. J. van der Linden, X. Dai, K. Maskell, and C. A. Johnson, Eds., 2001: Climate Change 2001: The Scienti®c Basis. Cambridge Univer- REFERENCES sity Press, 881 pp. Kleidon, A., and M. Heimann, 1999: Deep-rooted vegetation, Am- Alley, R. B., and P. U. Clark, 1999: The deglaciation of the northern azonian deforestation, and climate: Results from a modelling hemisphere: A global perspective. Ann. Rev. Earth Planet. Sci., study. Global Ecol. Biogeogr., 8, 397±405. 27, 149±182. ÐÐ, and ÐÐ, 2000: Assessing the role of deep rooted vegetation Betts, R., P. Cox, S. Lee, and F. Woodward, 1997: Contrasting phys- in the climate system with model simulations: Mechanism, com- iological and structural vegetation feedbacks in climate change parison to observations and implications for Amazonian defor- simulations. Nature, 387, 796±799. estation. Climate Dyn., 16, 183±199. Broecker, W., 2003: Does the trigger for abrupt climate change reside ÐÐ, K. Fraedrich, and M. Heimann, 2000: A green planet versus a in the ocean or in the atmosphere? Science, 300, 1519±1522. desert world: Estimating the maximum effect of vegetation on Chervin, R. M., and S. H. Schneider, 1976: Determining statistical the land surface climate. Climatic Change, 44, 471±493. signi®cance of climate experiments with general circulation Knutti, R., and T. F. Stocker, 2002: Limited predictability of the future models. J. Atmos. Sci., 33, 405±412. thermohaline circulation close to an instability threshold. J. Cli- Clark, P. U., N. G. Pisias, T. F. Stocker, and A. J. Weaver, 2002: The mate, 15, 179±186. role of the thermohaline circulation in abrupt climate change. Kucharik, C. J., and Coauthors, 2000: Testing the performance of a Nature, 415, 863±869. Dynamic Global Ecosystem Model: Water balance, carbon bal-

Unauthenticated | Downloaded 09/27/21 01:26 PM UTC 4142 JOURNAL OF CLIMATE VOLUME 17

ance, and vegetation structure. Global Biogeochem. Cycles, 14, water, and carbon between continents and the atmosphere. Sci- 795±825. ence, 275, 502±509. Lean, J., and P. Rowntree, 1997: Understanding the sensitivity of a Stocker, T. F., and A. Schmittner, 1997: In¯uence of CO2 emission GCM simulation of Amazonian deforestation to the speci®cation rates on the stability of the thermohaline circulation. Nature, of vegetation and soil characteristics. J. Climate, 10, 1216±1235. 388, 862±865. Levis, S., J. A. Foley, and D. Pollard, 2000: Large-scale vegetation Vellinga, M., and R. A. Wood, 2002: Global climatic impacts of a feedbacks on a doubled CO climate. J. Climate, 13, 1313±1325. collapse of the Atlantic thermohaline circulation. Climatic 2 Change, 54, 251±267. New, M., M. Hulme, and P. Jones, 1999: Representing twentieth- ÐÐ, ÐÐ, and J. M. Gregory, 2002: Processes governing the re- century space±time climate variability. Part I: Development of covery of a perturbed thermohaline circulation in HadCM3. J. a 1961±90 mean monthly terrestrial climatology. J. Climate, 12, Climate, 15, 764±780. 829±856. Wang, G. L., and E. A. B. Eltahir, 2000a: Ecosystem dynamics and Nicholson, S. E., 2000: Land surface processes and Sahel climate. the Sahel drought. Geophys. Res. Lett., 27, 795±798. Rev. Geophys., 38, 117±139. ÐÐ, and ÐÐ, 2000b: Biosphere±atmosphere interactions over West Pielke, R., R. Avissar, M. Raupach, A. Dolman, X. Zeng, and A. Africa. II: Multiple climate equilibria. Quart. J. Roy. Meteor. Denning, 1998: Interactions between the atmosphere and ter- Soc., 126, 1261±1280. restrial ecosystems: In¯uence on weather and climate. Global ÐÐ, and ÐÐ, 2000c: Role of vegetation dynamics in enhancing Change Biol., 4, 461±475. the low-frequency variability of the Sahel rainfall. Water Resour. Rahmstorf, S., 1996: On the freshwater forcing and transport of the Res., 36, 1013±1021. Atlantic thermohaline circulation. Climate Dyn., 12, 799±811. Washington, W. M., and C. L. Parkinson, 1986: An Introduction to ÐÐ, and A. Ganopolski, 1999: Long-term global warming scenarios Three-Dimensional Climate Modeling. University Science computed with an ef®cient coupled climate model. Climatic Books, 422 pp. Xue, Y., 1997: Biosphere feedback on regional climate in tropical Change, 43, 353±367. north Africa. Quart. J. Roy. Meteor. Soc., 123, 1483±1515. Schneider, S. H., and S. L. Thompson, 2000: A simple climate model Zeng, N., R. Dickinson, and X. Zeng, 1996: Climatic impact of Am- used in economic studies of global change. New Directions in azon deforestationÐA mechanistic model study. J. Climate, 9, the Economics and Integrated Assessment of Global Climate 859±883. Change, S. J. DeCanio et al., Eds., The Pew Center on Global Zhang, H., A. Henderson Sellers, and K. McGuf®e, 1996: Impacts Climate Change, 59±80. of tropical deforestation. Part I: Process analysis of local climatic Sellers, P., and Coauthors, 1997: Modeling the exchanges of energy, change. J. Climate, 9, 1497±1517.

Unauthenticated | Downloaded 09/27/21 01:26 PM UTC