Regional feedbacks under changing climate and land-use conditions

Introduction Conclusions References


Introduction
Terrestrial ecosystems account for large carbon stocks and explain large fractions of the variance of carbon exchange between the atmosphere and land surface. In addition, vegetation governs physical properties of the land surface (e.g. albedo, rooting depth and roughness) that affect water and energy exchange with the atmosphere. 5 Vegetation controls the surface energy partitioning, distributing net radiation over sensible and latent heat, which affects surface temperature and evapotranspiration. Humaninduced land-use/land-cover change (LULCC) has the effect of a (regional) climate forcing by modifying land properties and the carbon cycle (de Noblet-Decoudré et al., 2012). 10 The response of terrestrial ecosystems to these anthropogenically-induced climate forcings can positively or negatively feed back to the climate system, and thus magnify or reduce the initial perturbation (Field et al., 2007;Chapin et al., 2008;Arneth et al., 2010). Global modeling studies show a net positive carbon-climate feedback on surface temperature and atmospheric CO 2 -levels (Cox et al., 2000;Dufresne et al., 2002;15 Friedlingstein et al., 2006). However, these projections are still uncertain due to the complexity of the involved processes in the carbon-climate feedback loops (Meir et al., 2006;Heimann and Reichstein, 2008;Luo et al., 2009;Friedlingstein and Prentice, 2010). While biogeophysical feedbacks (albedo and water cycle) show a global negative feedback (cooling effect), they are relatively more significant at the regional scale. 20 Their sign varies among regions (Bonan, 2008;Davin and de Noblet-Ducoudré, 2010), with a usually cooling effect at high latitudes and warming in the tropics, partly related to cloud feedbacks (van der Molen et al., 2011a). However, mutual agreement among models to evaluate biogeophysical impacts of human-induced land-cover change is small (Pitman et al., 2009).
Vegetation sequesters atmospheric carbon dioxide (CO 2 ) via photosynthesis, known as Gross Primary Production (GPP) at the ecosystem level (Fig. 1). Part of the sequestered carbon is released through respiration for plant growth and maintenance (autotrophic respiration, R A ). The difference between GPP and R A is the net gain of carbon by plants, the Net Primary Production (NPP). Part of this sequestered carbon is released back to the atmosphere through organic matter decomposition by soil biota (heterotrophic respiration, R H ). The net carbon uptake by an ecosystem, being the balance between NPP and R H , is known as the Net Ecosystem Exchange (NEE). Furthermore, disturbances from both natural (e.g. fires) and anthropogenic (e.g. deforestation) sources imply a CO 2 source to the atmosphere. When also accounting for 15 these disturbances, the total carbon flux is called Net Biome Productivity (NBP). When NPP exceeds the sum of disturbances and R H , NBP is positive, implying a net carbon uptake by an ecosystem.
Current research on interactions between the carbon balance of terrestrial ecosystems and climate focuses on the capability of plants to enhance photosynthesis due 20 to enriched atmospheric CO 2 , and the sensitivity of all relevant carbon fluxes to the increasing temperature and changing moisture conditions. Enhancing photosynthesis -and thus plant growth -in response to elevated atmospheric CO 2 -levels implies a negative feedback on CO 2 -perturbations. This photosynthesis response to increasing CO 2 -levels varies across plant types, such as between C 3 and C 4 (greater for C 3 Introduction water-use efficiency (WUE; ratio of carbon fixation to water loss) is also enhanced under enhanced CO 2 -levels since it tends to reduce the stomatal opening to reduce water losses. In water-limited systems, this indirect stomatal effect can be as important as the direct effect of CO 2 fertilization (Campbell et al., 2000). However, enhancement of carbon fixation by terrestrial ecosystems might be restricted by Nitrogen (N) availability 5 (Norby et al., 2005(Norby et al., , 2010Arneth et al., 2010). The response of GPP to warming is usually positive at low temperatures and reduces at higher temperatures. The rising projected temperatures may also enhance soil respiration, the second-largest carbon flux from the surface to the atmosphere. However, this response is highly uncertain since the dynamics and processes of this 10 carbon flux are still not fully understood (Bond-Lamberty et al., 2004;Davidson and Janssen, 2006;Friedlingstein et al., 2006). Nevertheless, recent studies at the global scale showed a positive correlation between soil respiration and temperature anomalies (Bond-Lamberty and Thomson, 2010), and a global increase of soil respiration (a positive feedback) with rising temperature (Wei et al., 2010). 15 At the global scale, NPP is about 60 PgC yr −1 , and an annual increase of about 0.15-0.40 % yr −1 is estimated over the past 3 decades (Ito and Sasai, 2006;Piao et al., 2009;Pan et al., 2011). CO 2 fertilization, N deposition, forest regrowth and warming are considered to be related to the increased NPP at high latitudes, while the CO 2 fertilization component dominates the enhanced NPP in the tropics. Higher solar radi-Introduction savannas to tropical forest in Brazil. Comparing forest, transition forest and savannas, a clear difference in evaporation rates (4, 2.5 and 1 mm d −1 , respectively), annual rainfall (higher in forested areas) and the length of the dry season (longer in savanna regions) were reported. Soil moisture plays an important role. In soil moisture controlled evaporation regimes, reduction of soil moisture decreases evapotranspiration and increases sensible heat, which further increases evapotranspiration and thus reduces soil moisture via a positive soil moisture -temperature feedback loop (Seneviratne et al., 2010). The relationship between soil moisture and evapotranspiration depends on the climate regime (Koster et al., 2004;Seneviratne et al., 2006). In dry regions, evapotranspiration is highly sensitive to soil moisture, while in wetter regions it depends on the net radiative energy. In this context, da Rocha et al. (2009)  Land-use such as agriculture plays a significant role in land-atmosphere interactions, particularly in hydrological systems with extensive irrigation. Puma and Cook (2010) evaluated the global impact of observed irrigation changes in the 20th century. Evapotranspiration was strongly increased with irrigation, especially in the Indian region, while evaporation change in China's irrigation area was small due to energy limitation.
Regions with high irrigation had a cooling of about 1-2 K, with the largest cooling (∼3 K) in the northwest of the Indian subcontinent. However, a warming effect was shown in eastern India due to a weakening monsoon, which caused a decrease of soil moisture and enhanced sensible heat flux. The increasing irrigation in the Sahel region also showed a rise of cloud cover and precipitation. Although a cooling effect due to 10 irrigation has been found at the regional scale, no detectable effect on global mean temperature has been reported (Sacks et al., 2009).

Vegetation cover-albedo feedback
Surface albedo (the fraction of incident solar radiation reflected by the surface) is a function of the vegetation coverage, and thus related to ecosystem dynamics and pro- 15 cesses, which in turn are affected by global warming or enriched CO 2 levels. Changes in surface albedo alter the energy and water balances through changes in net surface radiation. For instance, the increase of vegetation cover in the boreal region decreases surface albedo and therefore increases available surface energy. This enhances transpiration, which may lead to greater precipitation in summer and lower snow cover 20 in late spring/early summer. Subtropical regions might also show a positive vegetation cover-albedo feedback as a result of climate change. With a fully coupled model, Zeng and Yoon (2009) explored the effect of this feedback to world's increasing desert in 2100. The expansion of desert increased from 10 % to 34 % when the vegetation cover-albedo feedback was considered.

25
Surface albedo change is the most significant driver of deforestation-induced climate change (Berbet and Costa, 2003). Simulation experiments of deforestation in boreal regions with global circulation models (GCM) coupled to a dynamic global vegetation 207 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | model (DGVM) have reported a cooling effect due to increased albedo, reduced sensible heating and a longer snow season (Bonan et al., 1992;Snyder et al., 2004;Cook et al., 2008;Notaro and Liu, 2008). Conversely, deforestation in tropical regions shows less cooling, or even warming. The warming effect of reduced cloudiness, resulting from decreased evapotranspiration, might nearly offset the cooling of increasing albedo 5 due to deforestation (Bala et al., 2007;van der Molen et al., 2011a).

The conceptual framework of responses and feedbacks
Forcings, interactions and processes that control terrestrial ecosystems-climate feedbacks are summarized in a conceptual framework (Fig. 2). Arrows connect system components, and closed loops between components represent feedbacks. Red arrows with a (+) sign imply that an increase of A will increase B or vice versa; green arrows with a (−) sign imply the reverse response; and double colored arrows with a (±) sign imply a response that can be either positive or negative. The product of all processes in a closed loop determines the overall sign of the feedback. Feedback loops are assumed to represent processes at a comparable spatial and temporal 15 scale. Some feedbacks reflect regional processes during specific episodes (such as droughts), while others reflect the entire global carbon-climate system. In the diagram shown in Fig. 2, changes in atmospheric CO 2 are induced by fossil fuel combustion, deforestation and net carbon uptake from terrestrial ecosystems. Higher atmospheric CO 2 increases carbon sequestration via photosynthesis (GPP), 20 which in turn reduces atmospheric CO 2 . In addition, CO 2 enrichment reduces stomatal conductance, which reduces water losses via evapotranspiration (ET) while maintaining carbon sequestration. This combination implies an increasing WUE (= GPP/ET).
Temperature affects both GPP and ecosystem respiration (R A and R H ). The response of GPP to temperature can be either positive or negative, since it increases at 25 low temperatures and it usually decreases at high temperatures. Ecosystem respiration is positively stimulated by higher temperature and is also highly related to GPP. The net effect of the feedback depends on which temperature response will be dominant. If 208 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | GPP is accelerated by higher temperature and water is not limiting, and the response of changes in ecosystem respiration is limited, the overall feedback loop is negative (Fig. 3a). Conversely, if respiration is strongly affected by changes in temperature, the feedback loop can be positive and therefore lead to increasing CO 2 emissions to the atmosphere (Fig. 3b). However, if primary productivity is reduced by elevated tem-5 perature, and thus also the carbon supplied to the soil is decomposed, this positive feedback is reduced or might disappear (Fig. 3c). In addition, N availability affects both GPP and respiration. In this conceptual framework the N cycle is, however, not considered explicitly. Soil moisture can play an important role in the carbon cycle, by influencing photosynthesis activity and ecosystem respiration. Finally, the sign of the net biogeochemical feedbacks will depend on the net balance between processes, which is depicted by a black circle around GPP and respiration in Fig. 3. Changes in albedo, due to natural vegetation cover changes, modify net radiation and ultimately surface temperature. Vegetation cover is influenced by natural ecosystem production (GPP) and deforestation, which in turn depends on food demand and 15 crop productivity. The level of adoption of agro-technologies plays an important role in agricultural expansion. Proper agricultural practices might improve crop production, and thus reduce the required land area, and consequently deforestation. In this conceptual framework, irrigation is added and it depends on available water resources. Other factors such as economy, food repartition and consumption patterns should be 20 added but are not considered in this study.

Agricultural expansion in sub-Saharan Africa (SSA)
Agriculture accounts for 65 % of the economical sector in Africa and 35 % of the gross national product in sub-Saharan Africa (SSA). About 90 % of the agricultural area is 25 small-scale rainfed farming, which can be negatively affected by a warming climate and the increasing frequency of extreme events, particularly droughts. The assessment of climate change impacts on agriculture is difficult due to the considerable uncertainties in climate projections, the response of crop yields and the beneficial effects of adaptation practices. From a meta-analysis, Roudier et al. (2011) evaluated the impacts of future changes 5 in crop yields due to climate change in West Africa. A median yield loss of 11 % was projected due to higher temperature and enhanced water stress, with a larger impact in northern West Africa, where warmer and drier projections prevailed. Lobell et al. (2011) used a statistical model to explore the potential effect of 1 K warming in SSA on two types of maize management: (a) "optimal" management with no water 10 deficit; and (b) "water stress" management representing rainfed agriculture, in which irrigation was interrupted after maize plants emerged. Under optimal management, the imposed warming reduced maize yields up to 30 % in 65 % of the current maize area in SSA. Conversely, all maize areas showed yield declines under the water stress management regime, and the reduction was at least 20 % in more than 75 % of the 15 areas. Liu et al. (2008) projected the effect of climate change and increased CO 2 concentration on SSA rainfed agriculture for 2030 with six main crops (cassava, maize, wheat, sorghum, rice and millet). A small increase of crop yields in response to climate change was reported, and wheat was the only crop with a negative effect. Great spatial variability was found in SSA; 7 countries showed a reduction in crop yield and 20 23 countries produced higher crop yields for all climate scenarios. Nevertheless, no changes in crop area were considered, and the discernible effect of crop area variability on future variations in temperature and precipitation require consideration in future climate assessments (Paeth et al., 2009). A strong expansion of agricultural land is anticipated in SSA to increase crop pro-25 duction. Large investments in agricultural technology and management are required to adapt to future climate change. Figure 4a shows the positive carbon feedback. If warming reduces crop yields (via an adverse effect on GPP), and land clearing increases (assuming no adoption of agro-technologies or irrigation), CO 2 emissions in the short Introduction  . 4a) and the long run (Fig. 4b) are enhanced and a positive feedback to the initial greenhouse forcing may take place. However, the sign of this feedback will depend on the adoption of agro-technologies and crop selection, the responses of crops to increasing temperature, the food demand and the relative role of the local carbon balance in the global carbon cycle. 5 Soil moisture is a major limiting factor for crop production in SSA. Crop failures are in this region highly correlated to drought events (Rojas et al., 2011), which are projected to become more frequent. Figure 4c shows the positive soil moisture-temperature feedback during a drought event. Soil water deficit reduces evapotranspiration, which positively feeds back to the increasing surface temperature. Crop productivity will depend on agricultural management, and irrigation is an important factor for increasing crop yields. If policies promote agricultural management such as irrigation, these will feed back negatively by reducing surface temperature due to increased evapotranspiration. However, limited availability of water resources can provide a serious constraint. Crop selection is an important means for adapting to a changing climate, which may 15 be regulated by policy measures for specific regions and adequate crop and soil management.

Heat waves and vegetation feedbacks in European ecosystems
Increasing interannual variability of summer climate in Europe (Seneviratne et al., 2006) is expected to increase the frequency of heat waves, usually linked to prolonged 20 anticyclonic circulation anomalies. Land -atmosphere interactions played an important role in the 2003 heat wave (Fischer et al., 2007b). A low precipitation anomaly in winter and spring is shown to be related to warm summers during various years in Europe (Vautard et al., 2007). Fischer et al. (2007a illustrated the importance of landatmosphere interactions by comparing coupled and uncoupled (prescribed soil mois-25 ture) land surface representations in a regional climate model. For a selection of heat waves (1976, 2003 and 2005), land-atmosphere interactions amplified precipitation anomalies four months before the particular event and increased evaporation during 211 Introduction spring, which increased daily temperature extremes for the hottest days in summer and augmented heat wave duration. The surface response to heat waves, is sensitive to the type of land-cover (Teuling et al., 2010). Observations from flux towers in forest and grassland areas showed that during the 2003 heat wave forests reduced evapotranspiration, most probably due to stomatal closure. In contrast, grasslands showed 5 a higher evapotranspiration early in the event, and thus a cooler environment than forests. For prolonged dry conditions, the resulting depletion of soil moisture eventually led to a reduction of latent heat flux and enhanced surface heating compared to forests. Forest areas tend to reduce water losses during a heat wave and prevent heat wave intensification in the long term. Figure 5a depicts the positive soil moisture -10 temperature feedback in ecosystems during a heat wave. Future hot spots of biophysical vegetation feedbacks in European forest areas were analyzed with the DGVM LPJ-GUESS by Wramneby et al. (2010). These hot spots were diagnosed by mapping the different greenhouse scenario response between regional climate model simulations with and without interactive vegetation responses. A 15 negative cooling feedback was found in central Europe, where CO 2 fertilization promoted plant growth and thus increased evapotranspiration. Scandinavian Mountains exposed a positive warming feedback due to the increased forest area extent and the reduced snow cover leading to a decrease of surface albedo (Fig. 5b). Drier summers in southern Europe showed a positive feedback by reduction of evapotranspiration due 20 to decreased plant growth.

Droughts and deforestation in the Amazon basin
The Amazon has been warming at a rate of 0.25 • C per decade in the past decades and a mean increase of 3.3 • C is projected for the 21st century (Malhi et al., 2008). SST variability explains about half of the rainfall variability in the Amazon (Zeng et al., 2008), while about 50 % of the rainfall in the Amazon is provided by water vapor originating from local sources (Li et al., 2006). Surface evaporation is the main source of moisture in the dry and early transition period (between dry and wet seasons) in the Amazon (Li et al., 2006; Li and 5 Fu, 2004) . Land surface characteristics can contribute significantly to the interannual variability of the wet season onset (Fu and Li, 2004). Deforestation may reduce soil moisture, delay the transition period and therefore prolong the dry season. Costa et al. (2007) highlighted the effect of deforestation on increasing surface albedo and consequent 10 reduction of evapotranspiration, which was well correlated to declined precipitation. This feedback loop was also confirmed by Sampaio et al. (2007), who projected an increase of surface temperature, and a decrease of evapotranspiration and precipitation in eastern Amazonia, especially during the dry season. Moreover, Amazon deforestation might enhance ENSO over the Pacific and thus induce greater reductions of rainfall 15 (positive feedback) as reported by Nobre et al. (2009).
More frequent, extended and severe droughts are projected in the Amazon Basin for this century (Li et al., 2006). The ENSO-related droughts (e.g. 1982/1983 and 1997/1998) particularly concern the northern and northeastern regions. However, the intense drought in 2005, associated to anomalously warm tropical North Atlantic SSTs, 20 highly affected the southern and southwestern regions, especially in the dry season. The 2005 drought was reported as a one-in-a hundred year event (Marengo et al., 2008). However, a more severe drought during the dry season in 2010 occurred (Lewis et al., 2011), which could be related again to the high Atlantic SST. Biomass and NPP might be reduced by increasing drought event frequency in the Amazon (Samanta et the Amazon forests to droughts. Two intense droughts in a decade might offset the net gains of an undisturbed Amazon forest (Lewis et al., 2011) and an increased frequency of these events could lead to forest dieback (negative NPP) or savannization (Nobre and Borma, 2009;Sampaio et al., 2007), especially over eastern and southeastern Amazonia (Salazar et al., 2007). 5 Several studies have evaluated the dynamics of the Amazon forests and biome shifts under a changing climate, which might feed back to the climate system. Oyama and Nobre (2003) reported a second climate-vegetation equilibrium in the Brazilian Amazon with tropical savannas replacing forests in a large area of the Amazon Basin. Malhi et al. (2009) suggested that seasonal and deciduous forests in the eastern Amazon could replace current vegetation. Nobre and Borma (2009) discussed thresholds that could cause the Amazon forest to switch to a stable savanna equilibrium, and concluded two main tipping points for the Amazon forest: 40 % of cleared Amazon forest and a temperature increase greater than 3-4 K.
The feedbacks related to deforestation in the Amazon are depicted in Fig. 6a. The to elevated CO 2 -concentrations when no CO 2 fertilization effect was considered. Conversely, a wetter climate and no changes in forest dynamics were simulated when including the CO 2 fertilization. Nevertheless, a biome shift towards tropical savanna was projected if the dry season would systematically be longer than 4 months. Betts et al. (2004)  Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | evaporation (positive biogeophysical feedback) and carbon sequestration (positive biogeochemical feedback), which caused 20 % and 5 % reduction of precipitation, respectively. Both feedback mechanisms are represented in Fig. 6b: (a) the effect of stomatal conductance to the soil moisture feedback, and (b) the reduction of evaporative water recycling due to forest cover decline.

Irrigation in South and Southeast Asia
South and Southeast Asia, together with East Asia, are known as the monsoon Asia, where half of the world's population lives. Agricultural production depends on monsoon rainfall during the summer season (June-August). However, technological innovation and irrigation expansion in the 1960s permitted to both reduce potential water deficit 10 in the monsoon season, and to allow a second crop in agricultural systems during the dry season. India and Southeast Asia together use more than half of the global irrigation (Sacks et al., 2009). Pakistan and Bangladesh have the highest percentage of irrigation in arable land (83 % and 56 %, respectively), and 35.4 % of the arable land in India (about 1.6 million km 2 , the greatest in South Asia) is under irrigation (Alauddin 15 and Quiggin, 2008). Furthermore, Bangladesh has the highest level of irrigation intensity due to irrigation in the dry season. However, Southeast Asia already faces water supply shortage and Sri Lanka is the only country in South Asia with little or no scarcity (Chuan, 2003) . Water shortage might induce a positive feedback as shown in Fig. 7a. Less water availability reduces soil moisture, which affects crop productivity. This will 20 reduce evapotranspiration and thus increase surface temperature, which would intensify evapotranspiration and further reduce soil moisture. More irrigation is needed to keep crop production at a sufficient level. Human-induced deforestation for agricultural expansion might affect monsoon patterns, as shown by Fu (2003) who reported an increase of winter monsoon and a weaker summer monsoon in East Asia due to LULCC. By using an atmospheric GCM, Takata et al. (2009)  weakening of Asian summer monsoon circulation between 1700 and 1850, especially over South and Southeast Asia. However, the impact of irrigation on rain patterns in the Indian monsoon region might be greater than LULCC, as stated by Douglas et al. (2009). Irrigation increases soil moisture and thus latent heat flux, which in turn reduces surface temperature and changes regional circulation patterns and precip-

Summary and conclusions
In this review study, the strength of terrestrial ecosystem-climate feedbacks at the re-15 gional and global level was explored. We identified four hot spots with ongoing or projected changes in land-use or management and in climate: sub-Saharan Africa (SSA), Europe, the Amazon Basin and South and Southeast Asia. We applied a conceptual framework (Fig. 2) as a tool to explore potential regional feedbacks. The diverse climatic, ecologic and socio-economic conditions led to different feedbacks among the 20 hot spots. In Europe, the high surface temperatures during a heat wave causes a soil water deficit, which reduces evapotranspiration and consequently further raises surface temperatures (positive soil moisture-temperature feedback). Conversely, during drought events in the Amazon, soil moisture-precipitation feedback dominates; soil moisture Introduction in forest dieback and carbon emissions, which also have an impact on the global climate. Human-induced land-use changes (deforestation) in the Amazon also influence ecosystem dynamics and initiate a positive vegetation albedo feedback through further declining vegetation cover due to its impact on the water balance. Irrigation is essential for agriculture to adapt to changes in rainfall patterns and 5 drought events, and it plays an important role in different feedbacks. In SSA, where agriculture is almost exclusively rainfed, investment in irrigation has high potential (de Fraiture and Wichelns, 2010). More frequent use of irrigation can have a local cooling effect due to enhanced evapotranspiration, leading to a negative soil moisturetemperature feedback. In contrast, South and Southeast Asian agriculture is largely 10 irrigation based, and more frequent water shortage, due to further water use, will lead to a positive soil moisture-temperature feedback. Deforestation and irrigation can also enhance changes in the Asian monsoon patterns. Irrigation increases latent heat flux in the pre-monsoon season. The subsequent surface temperature reduction, and thus the decrease in land-sea temperature gradient, changes regional circulation patterns 15 and consequently the early monsoon becomes weaker (Lee et al., 2009). Furthermore, Gordon et al. (2005) reported that projected extended deforestation in SSA changed the hydrological cycle, including enhanced changes in the Asian monsoon patterns. Thus enhanced food production in one region potentially leads to changes in the production in other areas. 20 Also global feedbacks, such as the carbon cycle-climate feedback, can be explored with the conceptual framework. Furthermore, if the local or regional cooling effect of irrigation has a signal on the global mean temperature, the feedback attains a global scope. For each hotspot, the ultimate sign of a feedback relies on the individual processes in the feedback chain. Further research is needed to gain confidence in the 25 way the balance between processes may affect the ultimate feedback sign.
The framework also gives a conceptual representation of the effect of human activities (e.g. irrigation and land-use change) in the ecosystems-climate system. Changes in land-use and crop production are based on human decisions, which depend on demographic, socio-economic and environmental factors. Although several attempts to integrate this socio-economic dimension in Earth System Models, the representation of these feedbacks remains a big challenge (Chapin et al., 2008). Given the importance of moving towards a better representation of the interaction between natural and human systems, improving levels of integration between Earth 5 System (ES) and Integrated Assessment (IA) tools is needed (Hibbard et al., 2010). In this respect, van Vuuren et al. (2012) discern four levels of interaction, ranging from a simple force-response to a complex coupled system. These interactions can be examined in collaboration studies involving coupled ES and IA models, where the degree of required coupling depends on the type of interaction considered. In between 10 the straightforward one-way information exchange and the most complex fully coupled modeling approaches, one can improve ES representation in IA models or vice versa. A priori, an assessment of the expected feedback strength is essential to select the most feasible model coupling strategy.
Based on these four levels of interactions, different approaches are suggested for 15 the hotspots selected in this paper. For SSA, an intermediate approach is sufficient by improving the representation of the climate system in an IA model. The system dynamics are dominated by socio-economical developments (such as food demand and agro-technology) and a fairly crude representation of the interaction with the climate system is appropriate. However, if due to for instance unexpected non-linear 20 responses the local feedback turns out to be strong, a fully coupled system must be considered.    Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | R. R., and Myneni, R. B.: Amazon forests did not green-up during the 2005 drought, Geophys. Res. Lett., 37, 1-5, doi:10.1029/2009GL042154, 2010. Sampaio, G., Nobre, C., Costa, M. H., Satyamurty, P., Soares-Filho, B. S., and Cardoso, M.: Regional climate change over eastern Amazonia caused by pasture and soybean cropland expansion, Geophys. Res. Lett., 34, L17709, doi:10.1029/2007GL030612, 2007. aboveground; BG to belowground; SOC to Soil Organic Carbon; Ra to Autotrophic respiration; Rh 6 to Heterotrophic respiration; NPP to Net Primary Production; NEE to Net Ecosystem Exchange; 7 NBP to Net Biome Production. 8 9 10 Fig. 1. Carbon stocks (circles) and fluxes (squares) in a terrestrial ecosystem. AG refers to aboveground; BG to belowground; SOC to Soil Organic Carbon; R A to Autotrophic respiration; R H to Heterotrophic respiration; NPP to Net Primary Production; NEE to Net Ecosystem Exchange; NBP to Net Biome Production.

Fig. 2.
Conceptual framework of interactions and processes between terrestrial ecosystems and climate. Red arrows mean that an increase (decrease) of a process/state will increase (decrease) another process/state; green arrows imply the reverse response. Circles refer to states, squares to fluxes/processes, triangles to responses and diamonds to external drivers. T refers to temperature; Nat Cov: Natural vegetation cover; GPP to Gross Primary Production; NEE to Net Ecosystem Exchange; R A to autotrophic respiration; R H to heterotrophic respiration; ET to evapotranspiration; Agro-tech to agro-technologies. Introduction  Fig. 4 (a,b) Positive feedback between crop yields, LULCC and temperature in SSA (c)Positive soil 2 moisture--temperature feedback during a drought event. Purple arrow represents the climate 3 forcing. 4 5 36 1 Fig. 4 (a,b) Positive feedback between crop yields, LULCC and temperature in SSA (c)Positive soil 2 moisture--temperature feedback during a drought event. Purple arrow represents the climate 3 forcing. 4 5 Fig. 4. (a, b) Positive carbon cycle -temperature feedbacks, when considering a negative impact of increasing temperature on crop yields, greater deforestation and low levels of adoption of agrotechnologies in SSA (c) Positive soil moisture-temperature feedback during a drought event. Purple arrow represents the climate forcing. Introduction