This is the accepted version of the journal article: Zeng, Zhaoqi; Wu, Wenxiang; Li, Yamei; [et al.]. «Increasing meteorological drought under climate change reduces terrestrial ecosystem productivity and carbon storage». One Earth, Vol. 6, issue 10 (Oct. 2023), p. 1326-1339. DOI 10.1016/j.oneear.2023.09.007 This version is available at https://ddd.uab.cat/record/284222 under the terms of the license 1 Increasing meteorological drought under climate change reduces 2 terrestrial ecosystem productivity and carbon storage 3 4 Zhaoqi Zeng1,2, Wenxiang Wu1,3*, Yamei Li3, Chong Huang1, Xueqin Zhang1, Josep 5 Peñuelas4,5*, Yao Zhang6, Pierre Gentine7, Zhaolei Li8, Xiaoyue Wang1, Han Huang1,2, 6 Xinshuai Ren1,2, and Quansheng Ge1* 1 7 Key Laboratory of Land Surface Pattern and Simulation, Institute of Geographic Sciences 8 and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China; 2 9 Department of Environment and Resources, University of Chinese Academy of Sciences, 10 Beijing 100049, China; 3 11 State Key Laboratory of Tibetan Plateau Earth System, Environment and Resources 12 (TPESER), Institute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing 13 100101, China; 4 14 CSIC, Global ecology Unit CREAF-CSIC-UAB, Bellaterra 08193, Catalonia, Spain; 5 15 CREAF, Cerdanyola del Vallès 08193, Catalonia, Spain; 6 16 Sino-French Institute for Earth System Science, College of Urban and Environmental 17 Sciences, Peking University, Beijing, China; 7 18 Department of Earth and Environmental Engineering, Center for Learning the Earth with 19 Artificial intelligence and Physics, Columbia University, New York, NY, USA; 8 20 College of Resources and Environment, and Academy of Agricultural Sciences, Southwest 21 University, Chongqing 400715, China; 22 *Correspondence to: W.X. Wu, wuwx@igsnrr.ac.cn; J. Peñuelas, Josep.Penuelas@uab.cat; Q.S. 23 Ge, geqs@igsnrr.ac.cn 1 24 Abstract: The terrestrial biosphere absorbs about 30% of the carbon dioxide emitted by human 25 activities each year, playing an important role in regulating global carbon budgets. The 26 persistence of such a carbon sink, however, critically depends on vegetation responses to future 27 increases in atmospheric aridity, decreases in soil water availability, and greater perturbations 28 associated with meteorological droughts under global warming. While the evidence for 29 increasing frequency and intensity of meteorological drought is growing, their potential 30 systematic adverse impacts on vegetation productivity for the coming decades have not been 31 quantified. Using newly-released data from 13 Sixth Coupled Model Intercomparison Project 32 (CMIP6) models and basing on multiple meteorological drought indices, we show that the 33 global mean drought-associated reductions in gross primary productivity (GPP) and net 34 primary productivity (NPP) are projected to increase by 3.5-fold (p < 0.01) under the SSP5-8.5 35 scenario and by 2.3-fold (p < 0.01) under the SSP1-2.6 scenario during the period from 2076 36 to 2100 relative to the historical baseline period (1851–2000). Especially, the terrestrial carbon 37 costs due to meteorological drought increase faster than the mean vegetation productivity 38 enhanced by CO2 fertilization effect in tropical and temperate ecosystems and particularly for 39 cropland. Increased potential evapotranspiration in response to global warming (i.e., radiative 40 effects of rising CO2) is likely to play either a dominant and direct role in increasing drought- 41 associated reductions in GPP and NPP, by intensifying meteorological droughts, or an indirect 42 role, by increasing the sensitivity of vegetation productivity to fluctuations in precipitation, or 43 both. Our results indicate that the exacerbation of meteorological droughts under future 44 warming scenarios increase a pressure on global food security and raise the concerns about the 45 transformation of terrestrial ecosystem from a carbon sink into a carbon source. 46 Keywords: meteorological drought; vegetation productivity; CMIP6; global warming 47 Introduction 48 As a key component of the terrestrial carbon cycle and ecosystem process, terrestrial ecosystem 2 49 production uniquely involves ecological, climatic, and anthropogenic impacts on the global 50 carbon cycle.1–3 Its alternation, in either magnitude or trend, could profoundly affect CO2 51 exchange between the land and the atmosphere, with great implications for the global 52 climate.4,5 For example, terrestrial C uptake removed about 3.61 Pg of C from the atmosphere 53 annually primarily driven by the acceleration of plants’ photosynthesis and water use efficiency 54 in response to the increased concentration of CO2 (i.e., physiological effects of CO2) during 55 the period from 2007 to 2016. This accounts for 33.7% of total anthropogenic C emissions 56 from industrial activity and land-use change5 and thus provides an important negative climate- 57 C feedback. However, there is increasing evidence that the greater atmospheric water demand 58 with rising temperatures (i.e., radiative effects of CO2) may lead to an increased intensity and 59 frequency of meteorological drought,6 which could notably affect vegetation growth7–9 and 60 crop yields,10 and even drive widespread forest mortality.11,12 In particular, hotter droughts are 61 an inciting factor in C sinks reduction from insects and may also increase the frequency, size, 62 and intensity of forest fires,13,14 and thus bring about an associated release of C to the 63 atmosphere that may further accelerate the rate of climate warming through a positive climate- 64 C cycle system feedback loop. Given the increased likelihood of both negative and positive 65 feedback effects of rising levels of CO2 under future climate conditions, it has been remained 66 an internationally concerned issue of whether future increasing meteorological droughts under 67 continuous global warming will lead to systematic adverse shifts in vegetation productivity at 68 regional and global scales? An improved projection of future drought impacts on terrestrial 69 ecosystem productivity is thus essential to reduce uncertainties in predicting land C uptake and 70 to better understand atmosphere–biosphere interactions. 71 Precisely quantifying the development of meteorological drought is one of the 72 prerequisites to effective assessment of drought impacts, yet it remains methodologically 73 challenging as using different drought indices to calculate drought characteristics can introduce 3 74 different uncertainties and even produce apparently conflicting results.15,16 For example, the 75 Standardized Precipitation Index (SPI) proposed by Mckee, et al.17 has increasingly been used 76 due to its simplicity and versatility.18,19 However, the SPI bases only on precipitation data and 77 does not consider other critical variables (e.g., temperature and evapotranspiration) that can 78 markedly influence droughts, the drying trends quantified by the SPI may thus be 79 underestimated under global warming. Alternatively, the Standardized Precipitation 80 Evapotranspiration Index (SPEI_PET-RC), calculated as the difference between precipitation 81 and potential evapotranspiration (PET) that commonly estimated by using reference crop 82 Penman–Monteith equation (PET-RC),20,21 could better capture the drought dynamics than the 83 SPI especially for regions with substantially higher temperature. However, recent studies 84 suggested that reference crop Penman–Monteith equation prescribed a constant surface 85 resistance (𝑟𝑠 ) at 70 s m−1, which is appropriate for an idealized reference crop in the current 86 climate but does not account for the fact that 𝑟𝑠 increases with elevated CO2 over vegetated 87 surfaces in climate model projections.22–24 Drought projections based on SPEI_PET-RC may 88 thus be overestimated due to the overrated PET under the background of continuously 89 enhanced atmospheric CO2 concentration. Despite their individual uncertainties, both SPI- 90 based and SPEI_PET-RC-based drought projections may set a lower and upper limit, 91 respectively, for the future drying trend. Given the inherent limitations of any single drought 92 index,15,25 a multi-index evaluation could better quantify meteorological drought events and 93 provide more important information for understanding the changes in future meteorological 94 drought characteristics and its main causes and impacts. 95 Drought impacts on vegetation productivity has been examined extensively, but most 96 studies were limited to a regional scale and/or focused on past few decades.7,26 A recent study 97 produced a global map of long-term projected impacts of soil moisture deficits on vegetation 98 productivity, and suggested that the magnitude of vegetation productivity reduction associated 4 99 with extreme low soil moisture will increase dramatically.27 This study provides a preliminary 100 understanding of future drought impacts on vegetation productivity at the global scale. But soil 101 moisture only reflects the amount of water resources available in the underground part of 102 vegetation (root zone), and is thus unable to represent the full influences by other 103 meteorological factors in the aboveground part of vegetation under the drought conditions.28 104 Several recent studies indicated that vegetation growth not only suffered from water stress by 105 soil moisture deficit, but also by other meteorological factors, such as extreme high temperature 106 and vapor pressure deficit (VPD).29–31 Using only soil moisture as drought indicator may thus 107 underestimate the impacts of drought on vegetation productivity in water-limited regions. By 108 contrast, the development of meteorological drought, such as quantified by SPEI, involving 109 interactions between precipitation and temperature that directly controlling the levels of soil 110 moisture and VPD, can more synthetically and accurately capture the response of vegetation 111 growth to drought.21,32 However, a comprehensive global assessment of projected changes in 112 long-term vegetation productivity response to meteorological droughts is still missing. This 113 knowledge gap prevents a deeper understanding of vegetation response to the expected 114 intensification of drought frequency, severity, and duration under continuous global warming. 115 In this study, we first create SPEI (including SPEI_PET-RC, which does not take account 116 for the effect of increased CO2 on PET, and SPEI_PET[CO2], which does) and SPI, using 117 projection data from 13 state‐of‐the‐art Earth system models (ESMs) in CMIP6, to 118 synthetically characterize and project spatiotemporal variations in meteorological drought 119 characteristics, including frequency, intensity, and duration during the period from 1851 to 120 2100. We then calculate the difference between modeled and expected GPP and NPP (Figure 121 S1) for each drought month and location, to systematically quantify the drought-associated 122 reduction in GPP and NPP under two contrasting future climate scenarios. We finally use an 123 idealized experiment, in which CO2 is increased from preindustrial levels by 1% each year only 5 124 in the atmospheric model (CO2rad) or in the vegetation model (CO2phy), or in both (FULL) 125 for four ESMs from the CMIP6 (see Materials and Methods) to specifically assess the 126 individual impacts of radiative and physiological effects of rising CO2 on projected changes in 127 drought and its associated GPP reduction. Our findings demonstrated that the dominant role of 128 radiative effects of CO2 in increasing meteorological droughts and its related carbon cost has 129 the great potential to transform terrestrial ecosystem from a carbon sink into a carbon source. 130 Results 131 Projected changes in drought characteristics 132 Under the two future climate scenarios, our analysis indicates that meteorological droughts will 133 become more frequent and longer in duration (Figure 1), particularly after 2025 under the 134 SSP5-8.5 scenario. During the period from 2076 to 2100, drought frequency is projected to 135 increase by 2.58 fold (p < 0.01) and 1.55 fold (p < 0.01) under the SSP5-8.5 and SSP1-2.6 136 scenarios, respectively (Figures 1A and 1E), and the mean drought duration is expected to 137 increase by 1.76 (p < 0.01) and 0.45 (p < 0.01) months (Figures 1C and 1G), respectively, with 138 the longest drought duration being increased by 6.47 (p < 0.01) and 2.02 (p < 0.01) months 139 (Figures 1D and 1H) per drought event (p < 0.01), respectively, compared with the historical 140 period (1851–2000). The levels of drought intensity are also projected to increase significantly 141 (p < 0.01) under the two climate scenarios, but after 2075, this increasing trend greatly slows 142 down under the SSP1-2.6 scenario (Figures 1B and 1F). Seasonally, consistent significant (p < 143 0.01) increases in droughts are projected to occur mainly during the boreal winter months of 144 December, January, and February (i.e., summer months for the Southern Hemisphere), with 145 weaker evidence for such increases occurring also in boreal summer months in the Northern 146 Hemisphere and in the latter spring months of November in the Southern Hemisphere (Figure 147 S2). In contrast, the droughts in September and October are projected to decrease for both the 6 148 Northern and Southern Hemispheres. By using SPEI_PET[CO2], we found that the increasing 149 trends in drought frequency, intensity, and duration are generally comparable to that estimated 150 by using SPEI_PET-RC, and all these trends show a significant increase (p < 0.01), despite a 151 slight decrease in magnitudes (Figure S3). The increasing trend in these drought characteristics 152 are also notably larger under the SSP5-8.5 scenario than that under the SSP1-2.6 scenario, 153 mainly as the result of the larger increase in PET under the high GHG emission scenario (Figure 154 S4). 155 Areas where drought frequency is projected to increase significantly (p < 0.01) under 156 future climate change are widely distributed across central-southern North America, 157 southwestern Eurasia, western and southern Africa, and much of South America and Australia, 158 which together account for 36.6 and 52.9% of the global land surface (vegetation-covered area) 159 under scenarios SSP1-2.6 and SSP5-8.5, respectively (Figures 2A–B). During the period from 160 2076 to 2100, drought events are projected to occur biennially on average in most regions listed 161 above under the SSP5-8.5 scenario (Figure S5); at higher latitudes (> 50N), however, drought 162 frequency is likely to decrease, especially under the low emission scenario. Spatial patterns of 163 drought duration are projected to be similar to those for drought frequency, with significant (p 164 < 0.01) increases in 21.9 and 42.3% of the global land surface under the SSP1-2.6 and SSP5- 165 8.5 scenarios, respectively (Figures 2E–F). Drought intensity is projected to exhibit a large 166 increase across nearly all the land surface under the SSP5-8.5 scenario and much smaller in the 167 SSP1-2.6 scenario, with the projected significant (p < 0.01) increase in drought intensity areas 168 being decreased from 75.9 to 32.1% of the global land surface. These spatial pattern of drought 169 characteristics are in closer agreement with that indicated by SPEI_PET[CO2], except for 170 central Africa, where a wetting trend was projected by SPEI_PET[CO2] under the two future 171 climate scenarios (Figure S6). 172 In contrast to increases in drought conditions, as indicated by our analyses of the SPEI, 7 173 water availability is projected to increase (i.e., wetting trends) under the two future climate 174 scenarios, as indicated by analysis of the SPI (Figure S7), notably across northeastern North 175 America and Europe, central-eastern Africa, and much of Asia (Figure S8). Analysis of the SPI 176 also indicates large and long-lasting decreases in water availability across the Amazon Basin, 177 Southern Africa, the southwestern United States, and southwestern Europe, particularly under 178 the SSP5-8.5 scenario. In general, current conditions are expected to intensify (i.e., a tendency 179 for the wet areas to get wetter and dry areas to get drier), as indicated by the SPI, but with no 180 overall effect at the global scale between scenarios SSP5-8.5 and SSP1-2.6 (Figure S8). 181 Projected changes in sensitivity of vegetation productivity to drought 182 The magnitudes of total reductions in GPP and NPP associated with SPEI-based drought are 183 projected to increase similarly under the two future climate scenarios (Figure 3), where trends 184 in reduced water availability identified by the SPEI become more pronounced under the SSP5- 185 8.5 scenario than under the SSP1-2.6 scenario. Between the historical (1851−2000) and future 186 (2076–2100) periods, total reductions in the global GPP (NPP) associated with SPEI-based 187 drought are predicted to increase by ~3.5-fold (from 7.49 (5.14) gC m−2 year−1 to 25.36 (18.24) 188 gC m−2 year−1; p < 0.01) under the SSP5-8.5 scenario and by ~2.3-fold (from 9.73 (6.76) gC 189 m−2 year−1 to 22.79 (15.67) gC m−2 year−1; p < 0.01) under the SSP1-2.6 scenario. In addition 190 to results based on SPEI at the 3-month timescale, we also used 2-, 4-, 5-, and 6-month 191 timescales of SPEI to evaluate the drought-related GPP reduction under the SSP5-8.5 scenario. 192 The results show great consistency with those from the 3-month SPEI (Figure S9). It should be 193 noted that the projected increase in drought-related reduction in GPP and NPP occurs in the 194 context of future CO2 fertilization, which would drive an increase in mean vegetation 195 productivity (Figure S1). Therefore, to better understand the drought impacts in relative terms, 196 it is also need to calculate the percentage reductions in GPP and NPP related to meteorological 197 droughts (i.e., ratios of total drought-related reduction in GPP to total modeled GPP in each 8 198 period). Results indicated that the percentage reductions in GPP, between the 1851−2000 and 199 2076–2100 periods, are projected to increase from ~0.75 to ~1.96% per year (p < 0.01) under 200 the SSP5-8.5 scenario and from ~0.97 to ~1.79% per year (p < 0.05) under the SSP1-2.6 201 scenario (Figure S10A). The drought assessments are not sensitive to the definition of the 202 metrics when using either SPEI_PET-RC or SPEI_PET[CO2]; the temporal variations in the 203 percentage reduction are highly consistent (Figure S10). These results further indicate a larger 204 increase in C cost resulting from meteorological droughts and the effectiveness of traditional 205 meteorological data-based drought models (i.e., SPEI_PET-RC) in evaluating the drought 206 impacts. 207 Spatial patterns of areas projected to experience greater drought-related reductions in GPP 208 and NPP during the period from 2076 to 2100 than during the historical period of 1851 to 2000 209 are correlated well with projected patterns of increased drought frequency, severity, and 210 duration, particularly in southeastern North America, central-eastern South America, 211 southwestern Eurasia, central-southern Australia, and southern Africa (Figure 4). Although 212 meteorological droughts are expected to be more widespread and severe under the SSP5-8.5 213 scenario, the predicted spatial distribution and changes in drought-associated reductions in GPP 214 and NPP are generally similar under the SSP1-2.6 and SSP5-8.5 scenarios in most temperate 215 regions. In addition, a latitudinal variation in the projected impact of drought on vegetation 216 productivity is clearly revealed, with the greatest reductions in drought-related GPP and NPP 217 for 2076–2100 likely to occur in tropical and temperate regions and the smallest reductions at 218 high latitudes (> 60°N). 219 Across climate gradients, we found that areas with greater projected changes in drought- 220 related reductions in vegetation productivity (DRP) tend to occur in water-limited (arid) 221 regions, while mean changes in DRP decrease from water- to energy-limited regions and along 222 an aridity gradient, where areas with greater reductions in DRP tend to be concentrated in semi- 9 223 arid ecosystems (0.05 < aridity < 0.5; Figures 5A–B and S11); projected increases in PET are 224 likely to be greater in these arid regions than in humid, energy-limited areas under the SSP1- 225 2.6 and SSP5-8.5 change scenarios (Figure S12). Across plant functional types, the greatest 226 increases in DRP during the period from 2076 to 2100 are projected to occur in cropland, 227 followed by forest and grassland (Figure 5D), whereas within plant functional type, the greatest 228 changes in DRP are likely to occur in the tropics and subtropics (TSGSS, TSMBF, and TSDBF), 229 followed by the temperate regions (TGSS, TBMF, and TCF), and the smallest changes in DRP 230 are likely to occur in the montane and cold regions (MGS, BF, and TUN; Figures 5C and S11). 231 Despite projected reductions in drought frequency, intensity, and duration, as indicated by 232 the SPI (Figure S7), with greater levels of water availability across the majority of the global 233 land surface, we projected an increased likelihood of drought-related (SPI-based) reductions in 234 GPP and NPP during the period from 2076 to 2100 (Figure S13), predominantly in 235 southwestern North America, southern Africa, northern-central South America, and 236 southwestern Europe (Figure S14). Our analysis demonstrates that SPI indicators of vegetation 237 productivity responses to drought will be more sensitive than those based on the SPEI, 238 indicating that the vegetation productivity will be more sensitive to decreases in precipitation 239 as a result of substantially elevated atmospheric water demand under future climate scenarios, 240 especially for northern China, southwestern Asian, and central Africa (Figure S15). Vegetation 241 drought sensitivity, indicated by the SPEI, will decrease during the period from 2076 to 2100 242 over more than half of the total vegetated land surfaces, predominantly in the southern 243 hemisphere, and this decrease will be larger under the SSP5-8.5 scenario, suggesting an 244 improvement in drought resistance under the high GHG emission scenario (Figure S15). In 245 addition to SPI and SPEI, we also used soil moisture (SM) as another drought indicator to 246 quantify drought-related GPP (NPP) reduction. These analysis methods are generally 247 consistent with Xu, et al.27 but used newly released CMIP6 data. Results indicated that SM 10 248 drought-related GPP (NPP) reduction are predicted to increase from 14.23 (9.52) gC m−2 year−1 249 to 40.71 (28.85) gC m−2 year−1 (p < 0.01) under the SSP5-8.5 scenario and from 14.66 (9.85) 250 gC m−2 year−1 to 29.37 (20.24) gC m−2 year−1 (p < 0.01) under the SSP1-2.6 scenario (Figure 251 S16), which are larger than those quantified by SPEI, especially in the Amazon Basin and the 252 high latitudes (Figure S17). 253 Mechanisms of projected changes in drought-related reduction in vegetation productivity 254 With rising atmospheric CO2 concentrations, plants tend to increase stomatal closure and 255 enhance water use efficiency to minimize water lose, which can induce reduction in 256 transpiration at the leaf level. On the other hand, increased vegetation productivity and leaf 257 biomass due to CO2 fertilization effects can generate a larger evaporative surface and thus 258 increase transpiration and thus the actual evapotranspiration (ET) at the ecosystem level. By 259 analyzing the projected changes in transpiration and ET in the CO2phy simulation, we found 260 that both of them show similar spatial patterns and are reduced in about 87% and 74% of the 261 vegetated land surface, respectively, in response to a quadruple increase in CO2 (Figures 6A 262 and S18). This suggested that extensive leaf area increases are not enough to offset the 263 influence of decreasing stomatal conductance on transpiration and ET. The reduction in ET 264 implies less water vapor available to drive rainfall and more sensible flux to rise surface 265 temperature, which contributes to widespread decrease in precipitation and enhancement in 266 VPD and PET (Figures 6, S18, and S19). These findings collectively indicated that even 267 without the CO2rad effects, physiological responses alone can also cause a slight 268 meteorological drying (global annual mean SPEI decreased by 0.1), especially for the wet 269 regions (Figure S19). However, the great improvement in water use efficiency (WUE, i.e., 270 GPP/transpiration) can slow down the loss of soil moisture (SM) and protect vegetation from 271 this mild meteorological drying, resulting in little change or even a widespread decrease in 11 272 drought-related reduction in vegetation productivity, except in the Amazon Basin region 273 (Figure 6C). 274 In the CO2rad simulation, despite the fact that 78% of the global land surface experienced 275 an increase in precipitation with a global average increase by 3.87%, this increase is small 276 relative to the enhancement in atmospheric water demand (i.e., VPD and PET), causing a 277 widespread drying trend that indicated by both SM and SPEI (Figure S19). Especially, our 278 results indicated that the spatial patterns of projected changes in precipitation, ET, VPD, SM, 279 and SPEI, as well as the drought-related reduction in vegetation productivity in the full 280 simulation (FULL) are highly consistent with those in the CO2rad simulation (Figures 6B, 6D, 281 S18, and S19). This further verified that the radiative effects will dominate over the water 282 saving effects of plants’ physiology in response to increasing CO2, resulting in similar surface 283 drying and reduction patterns in climate model simulations with or without the physiologic 284 response. But for some tropical and temperate regions (e.g., Amazon Basin), drought-related 285 GPP reductions are projected to increase in both CO2phy and CO2rad simulations, suggesting 286 that the reduction in stomatal conductance due to rising CO2 and VPD and decrease in SM 287 induced by increasing ET jointly contribute to the larger carbon cost in these regions. 288 Discussion 289 Projected changes in drought characteristics 290 This study provides a comprehensive evaluation of projected changes in meteorological 291 drought characteristics (frequency, intensity, and duration, as indicated by the SPI and SPEI) 292 under low (SSP1-2.6) and high (SSP5-8.5) GHG emission scenarios. The contrasting trends in 293 global mean drought characteristics projected by the analysis of the SPEI and SPI indicate that 294 PET increases more rapidly than precipitation under future climate scenarios, causing an 295 imbalance between water vapor availability and atmospheric demand and thus driving an 12 296 increase in meteorological drought. The greatest increases in drought conditions will occur 297 predominantly in tropical and subtropical latitudes, with a marked poleward expansion (Figure 298 2). These findings support the view anticipated by Sherwood, et al.33 that PET would increase 299 substantially in most tropical and mid-latitude areas in response to global warming, resulting 300 in shifts in local climates to more arid conditions. Chiang, et al.34 specifically investigated the 301 individual impacts of GHG and aerosol forcing on levels of PET and reported that the negative 302 effects of aerosol forcing may be masked by the positive impacts of GHG forcing in tropical 303 and subtropical regions, which eventually contribute to the increase in PET in these regions. 304 Thus, under anthropogenically induced global warming resulting from an increased 305 concentration of atmospheric CO2 and other heat-trapping gases, there is a strong expectation 306 of a general increase in PET that is directly related to a greater incidence of drought events 307 with greater severity and longer duration, especially for tropical and subtropical areas. 308 However, recent studies have indicated that increasing PET driven by global warming may 309 be severely overpredicted in the traditional offline calculations (e.g., the reference crop 310 Penman–Monteith equation), as they neglect the impacts of elevated CO2 on 𝑟𝑠 . When we take 311 the impacts of elevated CO2 concentration on 𝑟𝑠 into account in calculating PET, the projected 312 trends in global mean PET and drought frequency, intensity, and duration still increase 313 significantly (p < 0.01), although the magnitudes of these trends show a slight decrease (Figure 314 S3). These findings indicate that the trends in PET can be mostly explained by increased 315 temperature (i.e., radiative effects of rising CO2) rather than by elevated 𝑟𝑠 (i.e., physiological 316 effects of rising CO2) under rising atmospheric CO2 concentrations, which is further verified 317 by our mechanism analysis (Figure 6) and is consistent with recent observations.35 All these 318 results support the premise that the water cycle will intensify in a warming climate because of 319 greater atmospheric water demand. 320 Future levels of precipitation deficit under anthropogenic climate change, as indicated by 13 321 the global mean frequency, intensity, and duration of drought events, based on analysis of the 322 SPI, were shown to decrease, inferring that anthropogenic-mediated climate change may not 323 lead to increased droughts as a result of reduced levels of precipitation in most regions. 324 However, our analysis of SPEI data indicates that when they do occur, the extra heat from 325 global warming will increase the rates of evapotranspiration and surface water evaporation, 326 leading to more rapidly occurring drought conditions that are likely to be more intense. In this 327 regard, despite projected increases in precipitation, the anomalous fluctuations in high levels 328 of anthropogenic GHG emissions may also lead to an increased frequency and intensity of 329 meteorological drought conditions. However, because a warmer atmosphere can hold more 330 moisture according to the Clausius–Clapeyron scaling, it is also possible that increasingly 331 intense precipitation under warmer atmospheric conditions may lead to an intensification of 332 moisture levels in currently wet areas, causing a strong tendency for the wetter areas to become 333 wetter under improved levels of anthropogenic GHG emissions (Figure S8).36–38 This may 334 further exacerbate the problem of uneven distribution of water resources. 335 Projected changes in drought-reduced productivity 336 From a C perspective, although the analysis of SPEI data indicated greater increases in the 337 frequency, intensity, and duration of drought under the SSP5-8.5 climate scenario than under 338 the SSP1-2.6 scenario, global mean drought-associated reductions in GPP and NPP tended to 339 be similar (Figures 1 and 3). This paradoxical phenomenon may be explained in two ways. 340 First, the drought resistance of vegetation is improved under the elevated atmospheric CO2 341 concentrations (Figure S15). Numerous studies have indicated that rising CO2 concentrations 342 could stimulate photosynthetic activity39 and increase intrinsic vegetation water use efficiency 343 with lower stomata conductance,40 which was also identified by our mechanism analysis 344 (Figure S19). These physiological responses are of particular importance in plant communities 345 subjected to seasonal water shortage or drought conditions, as plants could maintain similar 14 346 rates of C assimilation with a reduced demand for water.41 Second, the reduction in GPP (NPP) 347 in drought-sensitive regions may be partially offset by increased vegetation productivity in 348 energy-limited regions during meteorological drought periods. This was verified by our results 349 projecting a substantial increase in vegetation productivity at high latitudes and across some 350 humid regions, such as southeastern China, even during periods of meteorological drought; in 351 particular, these regions were predicted to further expand under the high GHG emission 352 scenario (Figure 4). The increased vegetation productivity during meteorological drought 353 periods indicates that the water limitation caused by drought may not offset the positive effects 354 of higher temperatures and CO2 concentrations on vegetation growth in these colder and wetter 355 regions, supporting previous studies showing that temperature and photoperiod play more 356 important roles than water availability in vegetation growth at high latitudes.42,43 357 Additionally, the fact that vegetation productivity can increase during drought conditions 358 may also explain why our projected magnitudes of globally averaged reductions in GPP 359 associated with meteorological droughts (25.36 gC m−2 year−1; i.e., ~2.85 PgC year−1) were 360 lower than that estimated by Xu, et al.27 (~4.7 PgC year−1, which is consistent with our 361 estimations by using CMIP6 SM data, i.e., 4.57 PgC year−1; Figure S16) during the last quarter 27 362 of this century under the high GHG emission scenario. As Xu, et al. focused only on the 363 months during which GPP were reduced owing to extreme SM deficit, ignoring the fact that 364 drought is a long-term and gradual developing phenomenon and the responses of vegetation 365 growth to water stress can vary during different drought developing stages. For example, in 366 most energy-limited regions, favorable climate conditions (e.g., enhanced temperature and 367 abundant sunshine) may provide a more important role in promoting vegetation growth during 368 the early drought developing stage, which may partially or even entirely compensate the 369 reduction in GPP caused by extreme SM deficit in the middle or later drought stages.32,44 In 370 particular, our results indicated that most projected drought events will be initialized by 15 371 enhanced atmospheric water demand (i.e., PET) due to radiative effects of rising CO2, which 372 is responsible for the subsequent SM deficits during drought conditions. This further highlight 373 that the impacts of abundant sunshine and increased temperature as well as its tightly related 374 climate factors (e.g., VPD) preceding extreme SM deficit (i.e., during the early drought 375 developing stage) are nonnegligible to comprehensively capture drought-associated reduction 376 in vegetation productivity under climate change. 377 Overall, continued climate change threatens terrestrial ecosystems far more than it benefits 378 high latitudes. As Penuelas, et al.45 hypothesized that future increases in global temperature, 379 drought frequency, and key nutrient limitations, such as phosphorous, may drive a shift from 380 one period dominated by the positive effects of atmospheric fertilization on C sink to another 381 characterized by the saturation of these positive effects and a rise in negative impacts on climate 382 change. In terms of drought impacts, our analyses indicate that this transformation has a very 383 high possibility to occur during the period from 2076 to 2100, as the percentage reduction in 384 GPP related to meteorological droughts is projected to increase by a factor of 2 under both the 385 SSP5-8.5 and SSP1-2.6 climate scenarios (Figure S10), suggesting a faster increase in 386 terrestrial C cost resulting from droughts than the mean GPP (NPP) because of CO2 fertilization 387 effects in the future. But spatially, we highlight that this transformation may be more 388 pronounced across tropical and temperate ecosystems, such as the Amazon, Mediterranean 389 Basin, Southern Africa, and the southwestern United States. This is because larger increases in 390 both SPEI- and SPI-based droughts were projected to occur during the period from 2076 to 391 2100 in these regions (Figures 2 and S8), indicating a continued rise in atmospheric water 392 demand (PET) under global warming and a decrease in precipitation and associated increase 393 in water limitation that are expected to reduce vegetation productivity in these regions. It is 394 important to note that the optimal air temperature for ecosystem-level GPP, particularly in 395 tropical forests, is close to current growing-season air temperatures but is projected to fall 16 396 below the actual air temperatures under all future climate scenarios, indicating that 397 temperatures above the optimum may also occur for vegetation productivity in these 398 ecosystems.29 Thus, the integrated effects of water and temperature limitations associated with 399 both radiative and physiological effects of rising CO2 may mask the positive impacts of CO2 400 fertilization (Figure 6), causing catastrophic impacts on vegetation productivity and thus the 401 transformation of terrestrial ecosystems into C sources in these regions. These predictions of 402 large increases in drought-induced GPP and NPP reductions in tropical and temperate regions 403 support observations of changes in drought-induced vegetation productivity in the Amazon 404 Basin,46 the western United States,47 and across the globe (GPP) when using remote sensing- 405 based estimates,48 implying a further rising threat to the stability of the land C sink. In addition, 406 the largest increases in drought-induced vegetation productivity were projected to occur over 407 croplands (Figure 5C), which may be due to their lower coping capacity in times of water 408 scarcity as compared with woody vegetation with shallower roots and thus more limited access 409 to deeper soil water.26 Such findings highlight the urge for societies to take actions to reduce 410 increasing pressures of climate change on crop yields and guarantee a global food security. 411 Uncertainties of predictions and implications for the global C cycle and food security 412 Several limitations of our study reflect important challenges and open questions. First, many 413 models suffer from substantial tropical sea surface temperature biases that affect the accuracy 414 of the El Niño/Southern Oscillation (ENSO) simulations,49,50 which would thus affect the 415 reliability of drought event simulations in regions strongly connected to ENSO events; 416 therefore, further research is needed to identify the origins and impacts of these biases. Second, 417 rising temperatures and levels of CO2 tend to be correlated with regional hydroclimatic 418 conditions,34,51 and models may not fully capture the vegetation responses to changes in these 419 climate conditions, especially the response of 𝑟𝑠 to elevated atmospheric CO2 in water-limited 420 regions. This merits further attention because the largest increases in meteorological droughts 17 421 were projected to be concentrated mainly in these water-limited regions. Third, studies of plant 422 physiology have shown that plants have “memory” that allows them to store and recall 423 information from previous events and adjust their responses to future stress conditions 424 accordingly.52 Future increases in drought frequency may repeatedly trigger a memory of water 425 scarcity in plants and improve their tolerance to extreme drought events, through a reduction 426 in sensitivity.53 The lack of plant memory effects in CMIP6 models may therefore have led to 427 the overestimation of drought-related reduction in vegetation productivity. In addition to 428 memory effects, droughts elicit legacy effects—multiyear recovery of trees from drought—in 429 plants, which results from the physiological impairment caused by drought-induced water 430 stress.54,55 Even when climate conditions return to long-term average conditions, surviving 431 trees do not recover their expected growth rates for an average of 2 to 4 years.56 In CMIP6 432 models, however, plants’ physiological recovery from drought is often assumed to be complete 433 and relatively fast, leading to an underestimation of drought-related GPP (NPP) reduction. 434 Finally, we suggested that the increasing drought-related GPP (NPP) reduction may also be 435 partially compensated for by the positive anomalies of GPP and NPP related to favorable 436 wetness, temperature and radiation, and enhanced water-use efficiency, especially at high 437 latitudes, resulting in increased GPP and NPP, including in drought periods. However, this 438 projection may be too optimistic because the current CMIP6 models do not explicitly consider 439 insect dynamics,27 which are driven by temperature and drought and that contribute to tree 440 mortality and C cycles at a range of scales.57,58 Thus, further attention to including these 441 physiological and wider ecosystem processes in future models will lead to an improved 442 understanding of the effects of global change on GPP (NPP) and C cycles. 443 Conclusions 444 Compared with historical (1851–2000) levels, CMIP6 models project that the global mean 445 SPEI-based drought frequency, intensity, and duration will significantly increase during the 18 446 period from 2076 to 2100 under high (SSP5-8.5) and low (SSP1-2.6) future GHG emission 447 scenarios, whereas SPI-based drought hazards are likely to decrease. These contrasting trends 448 highlight the dominant role of PET (i.e., radiative effects of rising CO2) in the occurrence of 449 drought events for most regions under global warming, which are further verified by 450 mechanism experiments. The projected drought-associated reduction in GPP and NPP (DRP) 451 increased under the two climate scenarios, regardless of the drought indicator (SPI, SPEI), 452 indicating that vegetation productivity will become more sensitive to precipitation fluctuations 453 under enhanced atmospheric water demand. Areas with the greatest increases in DRP are likely 454 to occur in cropland, highlighting the potential threat of meteorological drought to global food 455 security. Spatially, larger DRP areas are projected to concentrate in the tropical and temperate 456 regions, including the Amazon Basin, Southern Africa, the southwestern United States, and 457 Europe, with smaller DRP occurring at high latitudes, where GPP (NPP) even increased during 458 meteorological drought periods. Such spatial patterns of drought and DRP are driven by 459 tradeoffs among the effects of water, temperature, and rising CO2 concentrations. Improved 460 quantification of the individual and combined impacts of these climate factors on vegetation 461 growth will lead to more reliable projections of ecosystem productivity and thus a better 462 understanding of atmosphere–biosphere feedbacks under future global climate change. 463 Materials and Methods 464 CMIP6 model data 465 We used monthly historical (1850–2014) and future (2015–2100) precipitation, maximum 466 temperature, minimum temperature, relative humidity, wind speed, shortwave radiation, GPP, 467 and NPP data derived from the CMIP6 simulations. At the time of writing this paper, there 468 were 13 earth system models that produced these data (Table S1; https://esgf- 469 node.llnl.gov/projects/cmip6/). Data were bilinearly interpolated to a spatial resolution of 0.5° 19 470 × 0.5°. In contrast to CMIP5, CMIP6 data employ a shared socioeconomic pathways (SSPs) 471 framework that describes five alternative evolutions of future society in the absence of climate 472 change or climate policy (SSP1 to SSP5),59 among which SSP1 and SSP5 envision contrasting 473 trends for human development, as they assume an increasing shift toward sustainable practices 474 (SSP1) and an energy-intensive, fossil-based economy (SSP5). On the basis of assumptions for 475 the SSPs, combined with four representative concentration pathways (RCPs), CMIP6 generates 476 four radiative forcing pathways (SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5) with associated 477 warming to the end of the 21st century in updated versions of integrated assessment models, 478 which has improved performance in many drought-related aspects than CMIP5, from 479 projecting ecosystem productivity (e.g., GPP and NPP) to hydrological process.60 In this study, 480 we considered SSP1-2.6 and SSP5-8.5 as two future scenarios, where SSP1-2.6 updates the 481 RCP2.6 pathways and represents a low-ending range of future scenarios, as measured by its 482 radiative forcing pathway (2.6 Wm−2 in 2100; low forcing sustainability pathway), whereas 483 SSP5-8.5 stabilizes radiative forcing at 8.5 Wm−2 in 2100 and is considered a high radiative 484 forcing scenario.59 Thus, the SSP1-2.6 and SSP5-8.5 scenarios capture the potential influence 485 of future ranges of GHG emissions on drought impacts under a relatively realistic range of 486 socioeconomic development pathways. Additionally, to include the physiological response of 487 vegetation to rising CO2 in calculating PET (PET[CO2]) and the SPEI (SPEI_PET[CO2]), 488 monthly and latitudinally (0.5°) resolved CO2 concentration data during the period from 1850 489 to 2100 under the two climate scenarios were used in our analysis.61 490 To specifically investigate the relative influence of radiative and physiological effects of 491 rising CO2 on projected changes in drought and its related GPP (NPP) reduction, monthly 492 precipitation, maximum temperature, minimum temperature, relative humidity, wind speed, 493 shortwave radiation, transpiration, actual evapotranspiration (ET), soil moisture (SM), and 494 GPP that outputted in three different simulations with 1% per year CO2 increases from four 20 495 CMIP6 models (BCC-CSM2-MR, CanESM5, CMCC-ESM2, and IPSL-CM6A-LR) were used. 496 The three simulations include: (1) 1% per year increase in CO2 from pre-industrial to quadruple 497 pre-industrial levels for both radiative and physiological processes (FULL; 1pctCO2 in CMIP6 498 terminology); (2) same as (1) but the CO2 increase is only in radiative processes, and is fixed 499 to the pre-industrial level for the physiological processes (CO2rad; 1pctCO2-rad in CMIP6 500 terminology); and (3) same as (1) but the CO2 increase is only in physiological processes, and 501 is fixed to the pre-industrial level for the radiative processes (CO2phy; 1pctCO2-bgc in CMIP6 502 terminology). These three simulations thus allow for the partitioning of changes in each water 503 and C cycle flux into two components of CO2rad and CO2phy. In each simulation, the change 504 in a field (e.g., ET, VPD, PET, and SPEI) due to increasing CO2 was calculated as the difference 505 between the average of the last 25 years with that of the first 25 years (Figures S18 and S19). 506 VPD was calculated from temperature and relative humidity; PET was calculated by Penman– 507 Monteith equation (see below); annual SPEI was calculated at a 12-month timescale, and the 508 baseline period for SPEI calculation was set to the first 30 years of the FULL model run for all 509 experiments (including CO2rad and CO2phy) in a given model. Before calculation, all monthly 510 original data in each simulation and model were first bilinearly interpolated to a common 0.5° 511 × 0.5° grid. 512 Observed and reanalyzed data 513 Monthly precipitation and potential evapotranspiration data for the 1980–2018 period were 514 obtained from the Climatic Research Unit Times Series (CRU-TS) data set to calculate an 515 aridity index, defined as the ratio of mean annual precipitation to PET; the latest version of this 516 database (v. 4.03) covers the period from 1901 to 2018 at a spatial resolution of 0.5° × 0.5° 517 over land surfaces (https://data.ceda.ac.uk/badc/cru/data/cru_ts). This data set, which has been 518 widely used in previous studies of climate change,62,63 was generated from interpolated 519 monthly climatic anomalies derived from more than 4,000 globally distributed meteorological 21 520 stations. 521 To characterize the terrestrial ecosystems that were water-limited or energy-limited (see 522 below for specific methods), monthly root-zone soil moisture (SM) and transpiration (trans) 523 data derived from the Global Land Evaporation Amsterdam Model (Gleam v3.3a; 524 https://www.gleam.eu/) were used. Data were generated based on a reanalysis of radiation and 525 air temperature data, a combination of gauge-based, satellite-based, and reanalyzed 526 precipitation data, and satellite-based vegetation optical depth data.64 Gleam data comprised 527 daily and monthly temporal resolution at 0.25° spatial resolution for the period from 1980 to 528 2018 and had undergone rigorous correction, preprocessing, and validation.65 Soil moisture 529 and transpiration data were aggregated to a spatial resolution of 0.5°. 530 Vegetation distribution data 531 Vegetation land cover data, based on the International Geosphere–Biosphere Program (IGBP) 532 classification, were extracted from the MCD12Q1 Land Cover Science Data product at a spatial 533 resolution of 0.05° (https://modis.gsfc.nasa.gov/data/dataprod/mod12.php), which were then 534 aggregated to 0.5°. 535 Changes in drought-related vegetation productivity were quantified by biome type 536 (https://www.worldwildlife.org/publications/terrestrial-ecoregions-of-the-world),66 which comprise 537 temperate tropical and subtropical moist broadleaf forests (TSMBF), tropical and subtropical 538 dry broadleaf forests (TSDBF), tropical and subtropical coniferous forests (TSCF), temperate 539 broadleaf and mixed forests (TBMF), temperate coniferous forests (TCF), boreal forests or 540 taiga (BF), tropical and subtropical grasslands, savannas, and shrublands (TSGSS), temperate 541 grasslands, savannas, and shrublands (TGSS), flooded grassland and savannas (FGS), montane 542 grasslands and shrublands (MGS), tundra (TUN), Mediterranean forests, woodlands, and scrub 543 (MFWS), deserts and xeric shrublands (DXS), and mangroves (MG). Samples of TSCF (3 544 pixels) and MG (10 pixels) were limited at a spatial resolution of 0.5°, so they were removed 22 545 from our analysis (Figure S11). 546 Definition of drought events and characteristics 547 Drought events become apparent after substantial periods without precipitation, but 548 quantification of their onset and end times and spatial extent is problematic; as a result, a range 549 of drought indices have been developed.21 Against the background of rapid global warming, 550 robust drought indices have to include the dynamics of temperature and its closely correlated 551 factors, such as PET. Therefore, we defined drought events in our study based on the SPEI, 552 calculated as the difference between precipitation and PET, to describe drought conditions with 553 respect to normal conditions for a given period.21 This approach accounts for both the effect of 554 climate warming on drought and the role of land–atmosphere feedback effects in drought 555 development and persistence. It can be calculated for contrasting timescales, where a short- 556 term (e.g., 3-month) SPEI reflects a high frequency of variability in soil moisture that is 557 important for vegetation production, whereas a long-term (e.g., 12-month) SPEI indicates 558 medium-term trends in precipitation and provides annual estimates of water availability that 559 are relevant to hydrological drought. Given that our investigation focused on drought impacts 560 on vegetation productivity, we mainly focus on the 3-month SPEI,67,68 in which PET (PET-RC) 561 was calculated by the reference crop Penman–Monteith equation20 (SPEI_PET-RC): 900 0.408∆(𝑅𝑛 −𝐺)+𝛾( )𝑈 (𝑒 −𝑒𝑎 ) 𝑇mean +273 2 𝑠 562 PET-RC = ， (1) ∆+𝛾(1+0.34𝑈2 ) 563 where 𝑅𝑛 is the net radiation at the vegetation surface [MJ·m−2·day−1]; G is the soil heat flux 564 density [MJ·m−2·day−1]; 𝑇mean is the mean air temperature at 2 m above ground level [℃]; 𝑈2 565 is the wind speed 2 m above ground level [m·s−1]; 𝑒𝑠 is the saturation pressure of water vapor 566 [kPa]; 𝑒𝑎 is the actual water vapor pressure [kPa]; ∆ is the slope of the vapor pressure curve 567 [kPa·℃−1]; and, 𝛾 is the psychrometric constant [kPa·℃−1]. In the reference crop Penman– 568 Monteith model, the surface resistance (𝑟𝑠 ) is prescribed as 70 s m−1, and this parameter value 23 569 is embedded in the equation. 570 However, recent studies have suggested that PET and drought conditions may be 571 overestimated by the SPEI_PET-RC, as it does not consider the impact of rising CO2 on 𝑟𝑠 .22,23 572 To reduce such uncertainties, we also used a modified reference crop Penman–Monteith PET 573 model that takes the biological effect of elevated [CO2] into account, as derived by Yuan, et al. 30 574 , to calculate the PET ( PET[CO2 ] ) and SPEI (SPEI_ PET[CO2 ] ). The PET[CO2 ] was 575 calculated as follows: 900 0.408∆(𝑅𝑛 −𝐺)+𝛾( )𝑈 (𝑒 −𝑒𝑎 ) 𝑇mean +273 2 𝑠 576 PET[CO2 ] = ∆+𝛾{1+𝑈 −4 ， (2) 2 [0.34+2.4×10 ([𝐶𝑂2 ]−300)]} 577 The monthly SPEI-3 (SPEI at a 3-month timescale) series for each pixel from 1851 to 2100 578 was calculated to construct the drought duration, intensity, and frequency. The onset and end 579 times of a drought event were defined as the month when the SPEI fell below and returned to 580 −1, respectively. Given that soil moisture stored prior to a drought buffers the impacts on 581 vegetation growth of short-term moisture deficiency, we defined a drought event as occurring 582 just when the SPEI value was less than −1 for at least three consecutive months. Thus, the 583 drought duration was calculated as the number of months between the onset and end of a 584 drought event; the drought intensity was calculated as the mean of monthly SPEI values during 585 the drought period; the drought frequency was defined as the number of drought events that 586 occurred over a specific period. Because the pronounced seasonal shifts in drought can also 587 alter vegetation productivity patterns, monthly and seasonal changes in drought frequency 588 during 2076 to 2100 were also assessed (Figure S2). Additionally, given that the timescales at 589 which different vegetation types respond to drought may differ,69 in addition to SPEI at a 3- 590 month timescale, we also calculated 2-, 4-, 5-, and 6-month timescales of SPEI to further 591 estimate the projected changes in drought-related GPP reduction under the SSP5-8.5 scenario. 592 To explore the individual contributions of changes in precipitation and atmospheric water 24 593 demand (i.e., PET) on long-term trends in drought frequency, intensity, and duration, we also 594 employed the standardized precipitation index (SPI) to define drought events. In contrast to the 595 SPEI, which accounts for precipitation and PET, the SPI accounts for only precipitation. By 596 comparing the results generated by the SPEI and SPI, the individual contributions of 597 precipitation and PET on drought as well as on vegetation productivity can thus be separated. 598 Note that the results generated based SPEI_PET-RC were shown in main text, while results 599 generated based SPEI_PET[CO2 ] and SPI were shown in supplementary. In addition to SPEI 600 and SPI, soil moisture (SM) data was used as an additional drought indicator to assess drought 601 impacts. SM drought was defined as the month when SM was below tenth percentiles for the 602 same month during 1851–2000 and caused a decrease in GPP, which is referenced to Xu, et al. 27 603 . 604 Definition of drought-related reduction in vegetation productivity 605 We defined drought-related reductions in GPP and NPP (DRP) as a departure in the modeled 606 productivity (CMIP6-exported GPP and NPP data) from the expected productivity (i.e., the 607 theoretical value of GPP and NPP in the absence of drought) during a drought episode: 608 𝐷𝑅𝑃𝑚𝑜𝑛 (𝑖, 𝑗) = 𝐸𝑃𝑚𝑜𝑛 (𝑖, 𝑗) − 𝑀𝑃𝑚𝑜𝑛 (𝑖, 𝑗), (3) 609 𝐷𝑅𝑃𝑒𝑣𝑒𝑛𝑡 (𝑖, 𝑗) = ∑𝐷𝐷 𝑚=1 𝐷𝑅𝑃𝑚𝑜𝑛 (𝑖, 𝑗), (4) 610 where 𝐷𝑅𝑃𝑚𝑜𝑛 (𝑖, 𝑗) is the drought-related reduction in vegetation productivity (i.e., GPP or 611 NPP), 𝐸𝑃𝑚𝑜𝑛 (𝑖, 𝑗) is the expected productivity, and 𝑀𝑃𝑚𝑜𝑛 (𝑖, 𝑗) is the modeled productivity 612 for a specific drought month 𝑚𝑜𝑛 in a pixel at longitude 𝑖 and latitude 𝑗; 𝐷𝑅𝑃𝑒𝑣𝑒𝑛𝑡 (𝑖, 𝑗) is the 613 total reduction during a drought episode, and 𝐷𝐷 is the duration of a drought event. 614 To quantify the expected productivity for a given drought month m, a smooth spline 615 (“smooth.spline” function in the R package) was used to fit a smooth curve over noisy 616 simulations of GPP and NPP for 𝑚 during the period from 1851 to 2100. This method 617 minimizes an objective function that considers the goodness of fit and smoothness of the 25 618 curve.27 The difference between the expected and observed productivity for month 𝑚 during 619 the drought year was defined as 𝐷𝑅𝑃𝑚𝑜𝑛 , with positive values indicating a drought reduction 620 in productivity and negative values indicating the reverse (Equation (2)). Estimation of 621 𝐷𝑅𝑃𝑚𝑜𝑛 using the smooth spline is described in detail in Figure S1. 622 For the analysis of temporal global shifts in DRP, we summed 𝐷𝑅𝑃𝑒𝑣𝑒𝑛𝑡 (𝑖, 𝑗) across 623 drought events and spatial locations for period 𝑝: 𝑁𝑗 𝑁𝑓 624 𝐷𝑅𝑃𝑔𝑙𝑜𝑏𝑎𝑙 (𝑝) = ∑𝑗=1 ∑𝑁𝑖 𝑖=1 ∑𝑓=1 𝐷𝑅𝑃𝑒𝑣𝑒𝑛𝑡 (𝑖, 𝑗, 𝑓) 𝐴(𝑖, 𝑗), (5) 625 where N𝑓 is the drought frequency and 𝐴(𝑖, 𝑗) is the area (m2) for the pixel at longitude 𝑖 and 626 latitude 𝑗 during period 𝑝. 627 Definition of sensitivity of vegetation productivity to drought 628 Biological sensitivity is the degree to which a system responds to (or is affected by) climate 629 change70 and is used as a key parameter to quantify the vulnerability of the ecosystem to a 630 climate disturbance.71 In our study, the sensitivity of vegetation productivity to drought in a 631 pixel for a specific period 𝑆𝑒𝑛𝑠𝑝 was defined as the mean reduction in productivity per drought 632 unit, defined as the product of the frequency, mean duration, and mean intensity: 𝑁𝑓 ∑𝑓=1 𝐷𝑅𝑃𝑒𝑣𝑒𝑛𝑡 (𝑖,𝑗,𝑓) 633 𝑆𝑒𝑛𝑠𝑝 = , (6) 𝑁𝑓×𝐷×𝐼 634 where 𝑁𝑓, 𝐷, and 𝐼 represent the drought frequency, mean drought duration, and mean drought 635 intensity, respectively, in a pixel at longitude 𝑖 and latitude 𝑗 during period 𝑝. 636 Statistical Analysis 637 We tested for differences at a specific pixel at longitude 𝑖 and latitude 𝑗 in the mean drought 638 frequency, duration, and intensity and drought-related reduction in GPP (NPP) between future 639 periods (2076−2100) and the historical baseline period (1851–2000) by using the Welch two- 640 sample t-test at p < 0.01. Temporal trends in mean drought sensitivity were calculated by using 26 641 the nonparametric Theil–Sen estimator, and pixels in which there were shifts in trend (p < 0.01) 642 were identified by using the Mann–Kendall trend test. The dependence of drought-related 643 reduction in GPP (NPP) on climate gradients was tested by a Pearson correlation analysis of 644 annual mean root-zone soil moisture and transpiration (r(SM, trans)), where positive values 645 indicate water-limited conditions and negative values represent energy-limited conditions.24,72 646 Acknowledgments 647 This research was funded by the Strategic Priority Research Program of the Chinese Academy 648 of Sciences (Grant No. XDA20090000) and the National Natural Science Foundation of China 649 (Grant No. 42271317). J.P.’s research was funded by the Spanish Government (Grant No. 650 PID2019-110521GB-I00), the Catalan Government (Grant No. SGR 2017-1005), the 651 Fundación Ramon Areces (project ELEMENTAL-CLIMATE), and the European Research 652 Council (Synergy Grant No. ERC-SyG-2013-610028, IMBALANCE-P). 653 Conflict of interest 654 The authors declare no competing financial interests 655 References 656 1. Gatti, L.V., Basso, L.S., Miller, J.B., Gloor, M., Domingues, L.G., Cassol, H.L.G., Tejada, 657 G., Aragao, L.E.O.C., Nobre, C., Peters, W. et al. (2021) Amazonia as a carbon source linked 658 to deforestation and climate change. Nature 595 (7867), 388–393. 659 2. Qin, Y., Xiao, X., Wigneron, J.-P., Ciais, P., Brandt, M., Fan, L., Li, X., Crowell, S., Wu, X., 660 Doughty, R. et al. (2021) Carbon loss from forest degradation exceeds that from deforestation 661 in the Brazilian Amazon. Nature Climate Change 11 (5), 442–448. 662 3. Beer, C., Reichstein, M., Tomelleri, E., Ciais, P., Jung, M., Carvalhais, N., Roedenbeck, C., 663 Arain, M.A., Baldocchi, D., Bonan, G.B. et al. (2010) Terrestrial Gross Carbon Dioxide Uptake: 27 664 Global Distribution and Covariation with Climate. Science 329 (5993), 834–838. 665 4. Reichstein, M., Bahn, M., Ciais, P., Frank, D., Mahecha, M.D., Seneviratne, S.I., 666 Zscheischler, J., Beer, C., Buchmann, N., Frank, D.C. et al. (2013) Climate extremes and the 667 carbon cycle. Nature 500 (7462), 287–295. 668 5. Keenan, T.F. and Williams, C.A. (2018) The Terrestrial Carbon Sink. In Annual Review of 669 Environment and Resources, Vol 43 (Gadgil, A. and Tomich, T.P. eds), pp. 219–243. 670 6. Dai, A. (2013) Increasing drought under global warming in observations and models. Nature 671 Climate Change 3 (2), 52–58. 672 7. Jiao, W., Wang, L., Smith, W.K., Chang, Q., Wang, H. and D'Odorico, P. (2021) Observed 673 increasing water constraint on vegetation growth over the last three decades. Nature 674 Communications 12 (1), 3777. 675 8. Restaino, C.M., Peterson, D.L. and Littell, J. (2016) Increased water deficit decreases 676 Douglas fir growth throughout western US forests. Proceedings of the National Academy of 677 Sciences of the United States of America 113 (34), 9557–9562. 678 9. Wang, S., Zhang, Y., Ju, W., Chen, J.M., Ciais, P., Cescatti, A., Sardans, J., Janssens, I.A., 679 Wu, M., Berry, J.A. et al. (2020) Recent global decline of CO2 fertilization effects on 680 vegetation photosynthesis. Science 370 (6522), 1295–1300. 681 10. Lobell, D.B., Roberts, M.J., Schlenker, W., Braun, N., Little, B.B., Rejesus, R.M. and 682 Hammer, G.L. (2014) Greater Sensitivity to Drought Accompanies Maize Yield Increase in the 683 US Midwest. Science 344 (6183), 516–519. 684 11. Adams, H.D., Zeppel, M.J.B., Anderegg, W.R.L., Hartmann, H., Landhausser, S.M., Tissue, 685 D.T., Huxman, T.E., Hudson, P.J., Franz, T.E., Allen, C.D. et al. (2017) A multi-species 686 synthesis of physiological mechanisms in drought-induced tree mortality. Nature Ecology & 687 Evolution 1 (9), 1285–1291. 688 12. Williams, A.P., Allen, C.D., Macalady, A.K., Griffin, D., Woodhouse, C.A., Meko, D.M., 28 689 Swetnam, T.W., Rauscher, S.A., Seager, R., Grissino-Mayer, H.D. et al. (2013) Temperature as 690 a potent driver of regional forest drought stress and tree mortality. Nature Climate Change 3 691 (3), 292–297. 692 13. Williams, A.P., Allen, C.D., Millar, C.I., Swetnam, T.W., Michaelsen, J., Still, C.J. and 693 Leavitt, S.W. (2010) Forest responses to increasing aridity and warmth in the southwestern 694 United States. Proceedings of the National Academy of Sciences of the United States of 695 America 107 (50), 21289–21294. 696 14. Flannigan, M.D., Krawchuk, M.A., de Groot, W.J., Wotton, B.M. and Gowman, L.M. (2009) 697 Implications of changing climate for global wildland fire. International Journal of Wildland 698 Fire 18 (5), 483–507. 699 15. Trenberth, K.E., Dai, A., van der Schrier, G., Jones, P.D., Barichivich, J., Briffa, K.R. and 700 Sheffield, J. (2014) Global warming and changes in drought. Nature Climate Change 4 (1), 17- 701 22. 702 16. Sheffield, J., Wood, E.F. and Roderick, M.L. (2012) Little change in global drought over 703 the past 60 years. Nature 491 (7424), 435-+. 704 17. Mckee, T.B., Doesken, N.J. and Kleist, J. (1993) The relationship of drought frequency and 705 duration to time scales. In Proceedings of the 8th Conference on Applied Climatology, 706 Anaheim, CA, USA, 179–183. 707 18. Gocic, M. and Trajkovic, S. (2014) Spatiotemporal characteristics of drought in Serbia. 708 Journal of Hydrology 510, 110-123. 709 19. Spinoni, J., Naumann, G., Carrao, H., Barbosa, P. and Vogt, J. (2014) World drought 710 frequency, duration, and severity for 1951-2010. International Journal of Climatology 34 (8), 711 2792-2804. 712 20. Allen, R.G., Pereira, L.S., Raes, D. and Smith, M. (1998) Crop evapotranspiration- 713 Guidelines for computing crop water requirements-FAO Irrigation and drainage paper 56. Fao 29 714 Roma 300, D05109. 715 21. Vicente-Serrano, S.M., Begueria, S. and Lopez-Moreno, J.I. (2010) A Multiscalar Drought 716 Index Sensitive to Global Warming: The Standardized Precipitation Evapotranspiration Index. 717 Journal of Climate 23 (7), 1696–1718. 718 22. Yang, Y., Roderick, M.L., Zhang, S., McVicar, T.R. and Donohue, R.J. (2019) Hydrologic 719 implications of vegetation response to elevated CO2 in climate projections. Nature Climate 720 Change 9 (1), 44–48. 721 23. Milly, P.C.D. and Dunne, K.A. (2016) Potential evapotranspiration and continental drying. 722 Nature Climate Change 6 (10), 946–949. 723 24. Berg, A. and McColl, K.A. (2021) No projected global drylands expansion under 724 greenhouse warming. Nature Climate Change 11 (4), 331–371. 725 25. Ukkola, A.M., De Kauwe, M.G., Roderick, M.L., Abramowitz, G. and Pitman, A.J. (2020) 726 Robust Future Changes in Meteorological Drought inCMIP6Projections Despite Uncertainty 727 in Precipitation. Geophysical Research Letters 47 (11). 728 26. Gampe, D., Zscheischler, J., Reichstein, M., O'Sullivan, M., Smith, W.K., Sitch, S. and 729 Buermann, W. (2021) Increasing impact of warm droughts on northern ecosystem productivity 730 over recent decades. Nature Climate Change 11 (9), 772–779. 731 27. Xu, C., McDowell, N.G., Fisher, R.A., Wei, L., Sevanto, S., Christoffersen, B.O., Weng, E. 732 and Middleton, R.S. (2019) Increasing impacts of extreme droughts on vegetation productivity 733 under climate change. Nature Climate Change 9 (12), 948–953. 734 28. Zheng, F., Ciais, P., Prentice, I.C., Gentine, P. and Hajima, T. (2022) Atmospheric dryness 735 reduces photosynthesis along a large range of soil water deficits. Nature Communications 13 736 (1), 1234567890. 737 29. Huang, M., Piao, S., Ciais, P., Penuelas, J., Wang, X., Keenan, T.F., Peng, S., Berry, J.A., 738 Wang, K., Mao, J. et al. (2019) Air temperature optima of vegetation productivity across global 30 739 biomes. Nature Ecology & Evolution 3 (5), 772–779. 740 30. Yuan, W., Zheng, Y., Piao, S., Ciais, P., Lombardozzi, D., Wang, Y., Ryu, Y., Chen, G., 741 Dong, W., Hu, Z. et al. (2019) Increased atmospheric vapor pressure deficit reduces global 742 vegetation growth. Science Advances 5 (8). 743 31. Grossiord, C., Buckley, T.N., Cernusak, L.A., Novick, K.A., Poulter, B., Siegwolf, R.T.W., 744 Sperry, J.S. and McDowell, N.G. (2020) Plant responses to rising vapor pressure deficit. New 745 Phytologist 226 (6), 1550-1566. 746 32. Zeng, Z., Wu, W., Ge, Q., Li, Z., Wang, X., Zhou, Y., Zhang, Z., Li, Y., Huang, H., Liu, G. 747 et al. (2021) Legacy effects of spring phenology on vegetation growth under preseason 748 meteorological drought in the Northern Hemisphere. Agricultural and Forest Meteorology 310. 749 33. Sherwood, S. and Fu, Q. (2014) A Drier Future? Science 343 (6172), 737–739. 750 34. Chiang, F., Mazdiyasni, O. and AghaKouchak, A. (2021) Evidence of anthropogenic 751 impacts on global drought frequency, duration, and intensity. Nature Communications 12 (1), 752 2754. 753 35. Wang, R., Li, L., Gentine, P., Zhang, Y., Chen, J., Chen, X., Chen, L., Ning, L., Yuan, L. 754 and Lu, G. (2022) Recent increase in the observation-derived land evapotranspiration due to 755 global warming. Environmental Research Letters 17 (2). 756 36. Zhang, X., Zwiers, F.W., Hegerl, G.C., Lambert, F.H., Gillett, N.P., Solomon, S., Stott, P.A. 757 and Nozawa, T. (2007) Detection of human influence on twentieth-century precipitation trends. 758 Nature 448 (7152), 461–465. 759 37. Min, S.-K., Zhang, X. and Zwiers, F. (2008) Human-induced arctic moistening. Science 760 320 (5875), 518–520. 761 38. Seager, R., Naik, N. and Vecchi, G.A. (2010) Thermodynamic and Dynamic Mechanisms 762 for Large-Scale Changes in the Hydrological Cycle in Response to Global Warming. Journal 763 of Climate 23 (17), 4651–4668. 31 764 39. Gunderson, C.A. and Wullschleger, S.D. (1994) Photosynthetic acclimation in trees to 765 rising atmospheric CO2: A broader perspective. Photosynthesis Research 39 (3), 369–388. 766 40. Keenan, T.F., Hollinger, D.Y., Bohrer, G., Dragoni, D., Munger, J.W., Schmid, H.P. and 767 Richardson, A.D. (2013) Increase in forest water-use efficiency as atmospheric carbon dioxide 768 concentrations rise. Nature 499 (7458), 324–327. 769 41. Driscoll, A.W., Bitter, N.Q., Sandquist, D.R. and Ehleringer, J.R. (2020) Multidecadal 770 records of intrinsic water-use efficiency in the desert shrub Encelia farinosa reveal strong 771 responses to climate change. Proceedings of the National Academy of Sciences of the United 772 States of America 117 (31), 18161–18168. 773 42. Piao, S., Tan, J., Chen, A., Fu, Y.H., Ciais, P., Liu, Q., Janssens, I.A., Vicca, S., Zeng, Z., 774 Jeong, S.-J. et al. (2015) Leaf onset in the northern hemisphere triggered by daytime 775 temperature. Nature Communications 6, 6911. 776 43. Berner, L.T., Massey, R., Jantz, P., Forbes, B.C., Macias-Fauria, M., Myers-Smith, I., 777 Kumpula, T., Gauthier, G., Andreu-Hayles, L., Gaglioti, B.V. et al. (2020) Summer warming 778 explains widespread but not uniform greening in the Arctic tundra biome. Nature 779 Communications 11 (1), 4621. 780 44. Tang, H. and Dubayah, R. (2017) Light-driven growth in Amazon evergreen forests 781 explained by seasonal variations of vertical canopy structure. Proceedings of the National 782 Academy of Sciences of the United States of America 114 (10), 2640-2644. 783 45. Penuelas, J., Ciais, P., Canadell, J.G., Janssens, I.A., Fernandez-Martinez, M., Carnicer, J., 784 Obersteiner, M., Piao, S., Vautard, R. and Sardans, J. (2017) Shifting from a fertilization- 785 dominated to a warming-dominated period. Nature Ecology & Evolution 1 (10), 1438–1445. 786 46. Doughty, C.E., Metcalfe, D.B., Girardin, C.A.J., Farfan Amezquita, F., Galiano Cabrera, 787 D., Huaraca Huasco, W., Silva-Espejo, J.E., Araujo-Murakami, A., da Costa, M.C., Rocha, W. 788 et al. (2015) Drought impact on forest carbon dynamics and fluxes in Amazonia. Nature 519 32 789 (7541), 78–82. 790 47. Breshears, D.D., Cobb, N.S., Rich, P.M., Price, K.P., Allen, C.D., Balice, R.G., Romme, 791 W.H., Kastens, J.H., Floyd, M.L., Belnap, J. et al. (2005) Regional vegetation die-off in 792 response to global-change-type drought. Proceedings of the National Academy of Sciences of 793 the United States of America 102 (42), 15144–15148. 794 48. Zhao, M. and Running, S.W. (2010) Drought-Induced reduction in global terrestrial net 795 primary production from 2000 through 2009. Science 329 (5994), 940–943. 796 49. McKenna, S., Santoso, A., Sen Gupta, A., Taschetto, A.S. and Cai, W. (2020) Indian Ocean 797 Dipole in CMIP5 and CMIP6: characteristics, biases, and links to ENSO. Scientific Reports 10 798 (1), 11500. 799 50. Zhu, Y., Zhang, R.-H. and Sun, J. (2020) North Pacific Upper-Ocean Cold Temperature 800 Biases in CMIP6 Simulations and the Role of Regional Vertical Mixing. Journal of Climate 33 801 (17), 7523–7538. 802 51. DeAngelis, A.M., Qu, X. and Hall, A. (2016) Importance of vegetation processes for model 803 spread in the fast precipitation response to CO2 forcing. Geophysical Research Letters 43 (24), 804 12550–12559. 805 52. Galviz, Y.C.F., Ribeiro, R.V. and Souza, G.M. (2020) Yes, plants do have memory. 806 Theoretical and Experimental Plant Physiology 32 (3), 195–202. 807 53. Fang, O., Zhang, Q.-B., Vitasse, Y., Zweifel, R. and Cherubini, P. (2021) The frequency 808 and severity of past droughts shape the drought sensitivity of juniper trees on the Tibetan 809 plateau. Forest Ecology and Management 486, 118968. 810 54. Bastos, A., Ciais, P., Friedlingstein, P., Sitch, S., Pongratz, J., Fan, L., Wigneron, J.P., Weber, 811 U., Reichstein, M., Fu, Z. et al. (2020) Direct and seasonal legacy effects of the 2018 heat wave 812 and drought on European ecosystem productivity. Science Advances 6 (24). 813 55. Kannenberg, S.A., Novick, K.A., Alexander, M.R., Maxwell, J.T., Moore, D.J.P., Phillips, 33 814 R.P. and Anderegg, W.R.L. (2019) Linking drought legacy effects across scales: From leaves 815 to tree rings to ecosystems. Global Change Biology 25 (9), 2978–2992. 816 56. Anderegg, W.R.L., Schwalm, C., Biondi, F., Camarero, J.J., Koch, G., Litvak, M., Ogle, K., 817 Shaw, J.D., Shevliakova, E., Williams, A.P. et al. (2015) Pervasive drought legacies in forest 818 ecosystems and their implications for carbon cycle models. Science 349 (6247), 528–532. 819 57. Lehmann, P., Ammunet, T., Barton, M., Battisti, A., Eigenbrode, S.D., Jepsen, J.U., 820 Kalinkat, G., Neuvonen, S., Niemela, P., Terblanche, J.S. et al. (2020) Complex responses of 821 global insect pests to climate warming. Frontiers in Ecology and the Environment 18 (3), 141– 822 149. 823 58. Anderegg, W.R.L., Hicke, J.A., Fisher, R.A., Allen, C.D., Aukema, J., Bentz, B., Hood, S., 824 Lichstein, J.W., Macalady, A.K., McDowell, N. et al. (2015) Tree mortality from drought, 825 insects, and their interactions in a changing climate. New Phytologist 208 (3), 674–683. 826 59. O'Neill, B.C., Tebaldi, C., van Vuuren, D.P., Eyring, V., Friedlingstein, P., Hurtt, G., Knutti, 827 R., Kriegler, E., Lamarque, J.-F., Lowe, J. et al. (2016) The Scenario Model Intercomparison 828 Project (ScenarioMIP) for CMIP6. Geoscientific Model Development 9 (9), 3461–3482. 829 60. Cook, B.I., Mankin, J.S., Marvel, K., Williams, A.P., Smerdon, J.E. and Anchukaitis, K.J. 830 (2020) Twenty-First Century Drought Projections in the CMIP6 Forcing Scenarios. Earths 831 Future 8 (6), e2019EF001461. 832 61. Meinshausen, M., Vogel, E., Nauels, A., Lorbacher, K., Meinshausen, N., Etheridge, D.M., 833 Fraser, P.J., Montzka, S.A., Rayner, P.J., Trudinger, C.M. et al. (2017) Historical greenhouse 834 gas concentrations for climate modelling (CMIP6). Geoscientific Model Development 10 (5), 835 2057–2116. 836 62. Humphrey, V., Berg, A., Ciais, P., Gentine, P., Jung, M., Reichstein, M., Seneviratne, S.I. 837 and Frankenberg, C. (2021) Soil moisture–atmosphere feedback dominates land carbon uptake 838 variability. Nature 592 (7852), 65–69. 34 839 63. Li, X., Piao, S., Wang, K., Wang, X., Wang, T., Ciais, P., Chen, A., Lian, X., Peng, S. and 840 Penuelas, J. (2020) Temporal trade-off between gymnosperm resistance and resilience 841 increases forest sensitivity to extreme drought. Nature Ecology & Evolution 4 (8), 1075–1083. 842 64. Martens, B., Miralles, D.G., Lievens, H., van der Schalie, R., de Jeu, R.A.M., Fernandez- 843 Prieto, D., Beck, H.E., Dorigo, W.A. and Verhoest, N.E.C. (2017) GLEAM v3: satellite-based 844 land evaporation and root-zone soil moisture. Geoscientific Model Development 10 (5), 1903– 845 1925. 846 65. Miralles, D.G., De Jeu, R.A.M., Gash, J.H., Holmes, T.R.H. and Dolman, A.J. (2011) 847 Magnitude and variability of land evaporation and its components at the global scale. 848 Hydrology and Earth System Sciences 15 (3), 967–981. 849 66. Olson, D.M., Dinerstein, E., Wikramanayake, E.D., Burgess, N.D., Powell, G.V.N., 850 Underwood, E.C., D'Amico, J.A., Itoua, I., Strand, H.E., Morrison, J.C. et al. (2001) Terrestrial 851 ecoregions of the worlds: A new map of life on Earth. Bioscience 51 (11), 933–938. 852 67. Jiang, P., Liu, H., Piao, S., Ciais, P., Wu, X., Yin, Y. and Wang, H. (2019) Enhanced growth 853 after extreme wetness compensates for post-drought carbon loss in dry forests. Nature 854 Communications 10, 195. 855 68. Wu, C., Wang, J., Ciais, P., Penuelas, J., Zhang, X., Sonnentag, O., Tian, F., Wang, X., 856 Wang, H., Liu, R. et al. (2021) Widespread decline in winds delayed autumn foliar senescence 857 over high latitudes. Proceedings of the National Academy of Sciences of the United States of 858 America 118 (16), e2015821118. 859 69. Vicente-Serrano, S.M., Gouveia, C., Camarero, J.J., Beguería, S., Trigo, R., López-Moreno, 860 J.I., Azorín-Molina, C., Pasho, E., Lorenzo-Lacruz, J. and Revuelto, J. (2013) Response of 861 vegetation to drought time-scales across global land biomes. Proceedings of the National 862 Academy of Sciences 110 (1), 52-57. 863 70. Beaugrand, G. (2014) Marine Biodiversity, Climatic Variability and Global Change, 35 864 Routledge. 865 71. Choat, B., Jansen, S., Brodribb, T.J., Cochard, H., Delzon, S., Bhaskar, R., Bucci, S.J., Feild, 866 T.S., Gleason, S.M., Hacke, U.G. et al. (2012) Global convergence in the vulnerability of 867 forests to drought. Nature 491 (7426), 752–755. 868 72. Berg, A. and Sheffield, J. (2018) Soil Moisture-Evapotranspiration Coupling in CMIP5 869 Models: Relationship with Simulated Climate and Projections. Journal of Climate 31 (12), 870 4865–4878. 36 871 Figure legends 872 Figure 1. Temporal changes in SPEI-based (SPEI_PET-RC) drought characteristics from 1851 873 to 2100 under the SSP5-8.5 (A–D) and SSP1-2.6 (E–H) climate scenarios comprising changes 874 in drought frequency (A, E), mean drought intensity (B, F), mean drought duration (C, G), and 875 mean longest drought duration (D, H). The solid red line represents the overall multi model 876 mean, and the shading represents the mean ± SD. 877 Figure 2. Spatial patterns of mean changes in drought characteristics, based on SPEI_PET-RC, 878 between future (2076–2100) and historical (1851–2000) periods under the SSP5-8.5 (A, C, E) 879 and SSP1-2.6 (B, D, F) climate scenarios. The stippling denotes regions with changes in mean 880 drought at p < 0.05. The insets in A, C, and E represent frequency distributions of mean changes 881 in drought frequency, intensity, and duration, respectively, under the SSP5-8.5 (red line) and 882 SSP1-2.6 (blue line) climate scenarios; vertical lines represent average values. The insets in B, 883 D, and F show proportions (%) of areas with increases (*+) or decreases (*−) in drought 884 frequency, intensity, and duration, respectively, under the SSP5-8.5 (red bar) and SSP1-2.6 885 (blue bar) climate scenarios at p < 0.05. 886 Figure 3. Temporal changes in the drought-related reduction in vegetation productivity, based 887 on the SPEI_PET-RC, during period from 1851 to 2100 under the SSP5-8.5 (A, B) and SSP1- 888 2.6 (C, D) climate scenarios: drought-related reduction in GPP (A, C) and drought-related 889 reduction in NPP (B, D). The solid red line represents overall multi-model means, and the 890 shading represents the mean ± SD. 891 Figure 4. Spatial patterns of mean changes in the drought-related reduction in GPP (A, D) and 892 NPP (B, E) between future (2076–2100) and historical (1851–2000) periods under the SSP5- 893 8.5 (A, B) and SSP1-2.6 (D, E) climate scenarios. The stippling indicates regions with mean 894 changes in drought-related GPP and NPP reduction at p < 0.05. (C, F) Latitudinal comparison 895 of mean changes in drought-related reduction in GPP (red) and NPP (blue); the solid line 37 896 indicates means, and the shading represents the SD. 897 Figure 5. Dependence of changes in the drought-related reduction in vegetation productivity 898 on climate and vegetation gradients. Boxplots of mean changes in drought-related reduction in 899 GPP (tomato boxes) and NPP (orange boxes) across cor(SM, trans) gradients (A); *+, +, −, and 900 *− for significant (p < 0.05) positive, positive, negative, and significant (p < 0.05) negative 901 values, respectively, where positive values indicate water-limited transpiration and negative 902 values reflect energy limitation. Boxplots of mean changes in the drought-related reduction in 903 GPP (tomato boxes) and NPP (orange boxes) across aridity gradients (B), biome types (C), and 904 plant functional types (D). 905 Figure 6. Global mean relative changes in precipitation (Pre), transpiration (Tran), actual 906 evapotranspiration (ET), vapor pressure deficit (VPD), potential evapotranspiration (PET), and 907 water use efficiency (WUE), and annual mean values changes in the Standardized Precipitation 908 Evapotranspiration Index (SPEI) between last 25 years with the first 25 years under the FULL, 909 CO2phy, and CO2rad simulations (A). Spatial patterns of mean changes in the drought-related 910 reduction in GPP between last 25 years with the first 25 years under the FULL (B), CO2phy 911 (C), and CO2rad (D) simulations. 38 912 Figure 1 913 39 914 Figure 2 915 40 916 Figure 3 917 41 918 Figure 4 919 42 920 Figure 5 921 43 922 Figure 6 923 44