1 Classification : 1 SOCIAL SCIENCES: Sustainability Science 2 PHYSICAL SCIENCES: Earth, Atmospheric, and Planetary Sciences 3 4 Consequences of rapid ice-sheet melting on the 5 Sahelian population vulnerability 6 7 Author Affiliation: 8 Dimitri Defrance (1,2), Gilles Ramstein (1), Sylvie Charbit (1), Mathieu Vrac (1), Adjoua Moïse 9 Famien (3,2), Benjamin Sultan (2), Didier Swingedouw (4), Christophe Dumas (1), François 10 Gemenne (5,6), Jorge Alvarez-Solas (7), and Jean-Paul Vanderlinden (5) 11 12 (1) LSCE/IPSL, CEA-CNRS-UVSQ, Université Paris-Saclay, Gif-sur-Yvette, France 13 (2) Sorbonne Universités, UPMC - CNRS-IRD-MNHN, LOCEAN/IPSL, Paris, France 14 (3) LAPA, Université Félix Houphouet Boigny, Abidjan, Côte-d’Ivoire 15 (4) EPOC, Université de Bordeaux, Pessac, France 16 (5) CEARC, OVSQ, Université Paris-Saclay, Guyancourt, France 17 (6) The Hugo Observatory, FNRS, University of Liège, Belgium 18 (7) PalMA Group, Universidad Complutense de Madrid, Madrid, Spain 19 20 Corresponding Author : 21 Dimitri Defrance 22 Université Pierre et Marie Curie 23 Boîte 100, 4 place Jussieu 24 75252 Paris Cedex 05 25 France 26 Tel : +33 6 49 49 89 48 27 E-mail : dimitri.defrance@ird.fr 28 29 Keywords: 30 climate change ; ice-sheet melting ; impact ; agriculture ; sahel 31 2 Abstract 32 The acceleration of ice sheet melting has been observed over the last few decades. Recent 33 observations and modeling studies have suggested that the ice sheet contribution to future 34 sea level rise could have been underestimated in the latest Intergovernmental Panel on 35 Climate Change report. The ensuing freshwater discharge coming from ice sheets could have 36 significant impacts on global climate, and especially on the vulnerable tropical areas. During 37 the last glacial/deglacial period, megadrought episodes were observed in the Sahel region at 38 the time of massive iceberg surges, leading to large freshwater discharges. In the future, such 39 episodes have the potential to induce a drastic destabilization of the Sahelian agroecosystem. 40 Using a climate modeling approach, we investigate this issue by superimposing on the 41 Representative Concentration Pathways 8.5 (RCP8.5) baseline experiment a Greenland flash 42 melting scenario corresponding to an additional sea level rise ranging from 0.5 m to 3 m. Our 43 model response to freshwater discharge coming from Greenland melting reveals a significant 44 decrease of the West African monsoon rainfall, leading to changes in agricultural practices. 45 Combined with a strong population increase, described by different demography projections, 46 important human migration flows could be potentially induced. We estimate that, without 47 any adaptation measures, tens to hundreds million people could be forced to leave the Sahel 48 by the end of this century. On top of this quantification, the sea level rise impact over coastal 49 areas has to be superimposed, implying that the Sahel population could be strongly at threat 50 in case of rapid Greenland melting. 51 Significance Statement 52 A major uncertainty concerning the 21st century climate is the ice sheet response to global 53 warming. Paleodata indicate rapid ice sheet destabilizations during the last deglaciation, 54 which could lead to an underestimation of sea level rise, as suggested in recent publications. 55 Therefore, we explore the impact of different scenarios of Greenland partial melting in the 56 very sensitive Sahel region. We first demonstrate that such a melting induces a drastic 57 decrease of West African monsoon precipitation. Moreover, we quantify the agricultural area 58 losses due to monsoon changes. Consequently, we pinpoint a large potential for migration of 59 millions of people in the coming decades. Thus, the ice sheet destabilization provokes not only 60 coastal damages but also large population migration in monsoon area. 61 3 The Sahel is particularly exposed to extreme climate variability, as evidenced by the impacts 62 of the severe droughts in the late 20th century (1). Paleoclimatic records have also shown that 63 megadrought episodes occurred in this area during past glacial/deglacial periods (2⇓⇓–5) at 64 the time of huge surges of icebergs (i.e., the so-called Heinrich events), causing outlet glacier 65 acceleration and thus sea level rise (6, 7) (SLR). Several modeling studies performing water-66 hosing experiments confirmed the close correspondence between the West African monsoon 67 weakening and the freshwater flux (FWF) released to the ocean (8⇓–10) due to ice sheet 68 melting. These studies raise the question as to whether such episodes could occur during this 69 century in response to a massive freshwater discharge triggered by a significant ice sheet 70 destabilization or surface melting and, if so, what would be the related environmental and 71 human impacts in the Sahel area. 72 73 According to the latest Intergovernmental Panel on Climate Change Fifth Assessment Report 74 (AR5) (11), the likely range of global mean SLR under the Representative Concentration 75 Pathways 8.5 (RCP8.5) scenario is 0.52 m to 0.98 m by the end of the 21st century. Although 76 considerable progress has been made in ice sheet modeling over the last decade, this range is 77 provided with only medium confidence, due to large remaining uncertainties in the ice sheet 78 dynamic response and to an improper representation of the ice–ocean interactions (12). 79 80 In Greenland, recent observations of fjords standing well below sea level suggest important 81 processes of glacier front destabilization (13) that are not included in the current dynamic ice 82 sheet models (14). Moreover, although there are only a few ice shelves surrounding 83 Greenland compared with West Antarctica, post-AR5 remote sensing observations reveal that 84 ice shelves have experienced a continuous thinning for several years, resulting in a buttressing 85 weakening (15, 16), not only in the Antarctic ice sheet but also in Greenland. This leads to a 86 significant ice stream acceleration and possibly to a massive discharge of grounded ice, similar 87 to what occurred during Heinrich events or, more recently, after the collapse of the Larsen B 88 ice shelf (17). Moreover, past episodes of rapid SLR acceleration, such as the Meltwater Pulse 89 1A (18), are still raising questions about our ability to evaluate the future SLR under current 90 understanding of physical mechanisms. 91 92 4 Results from these past climate studies combined with present-day observations suggest that 93 the ice sheet contribution to SLR could have been underestimated. Here, we consider different 94 freshwater discharge scenarios equivalent to an additional SLR ranging from 0.5 m to 3 m 95 coming from ice sheet melting and/or destabilization, which is not accounted for in the 96 baseline RCP8.5 climate simulation. We explore the related climatic effects on the West 97 African monsoon over the 21st century and their ensuing impacts on the Sahelian cultivable 98 areas and thus on the local population. 99 100 Using the Institut Pierre Simon Laplace low resolution coupled ocean–atmosphere model 101 (IPSL-CM5A-LR) [same version as in the Coupled Model Intercomparison Project, Phase 5 102 (CMIP5) exercise (19); Methods] run under the RCP8.5 radiative forcing from 2006 to 2100, 103 we performed four different water-hosing experiments superimposed to the RCP8.5 scenario 104 in which we added, respectively, a 0.11-, 0.22-, 0.34-, and 0.68-Sv FWF (1 Sv = 106 m3⋅s−1) 105 released in the North Atlantic from 2020 to 2070 and corresponding, respectively, to 0.5-, 1-, 106 1.5-, and 3-m SLR. Our goal is, first, to investigate the climatic impacts of the FWF coming from 107 Greenland (GrIS scenarios) in the West African region and, second, to show the impacts on 108 the cereal cultivation in the Sahel area, and the consequences for the local population, which 109 is already facing chronic malnutrition problems. 110 Changes in tropical precipitation systems 111 It has been shown that the freshening of the North Atlantic has global climatic impacts (9, 112 20⇓⇓⇓⇓–25), including a strong cooling of the Northern Hemisphere down to the Sahara 113 (26⇓⇓–29) related to a very strong slowdown of the North Atlantic Deep Water (NADW) 114 leading to the slowdown of the Atlantic meridional overturning circulation (AMOC) (9, 20⇓⇓⇓–115 24). The maximum decrease of the mean annual NADW outflow at 30°N occurs around 2060 116 and corresponds to a reduction of 90% (60%) of the initial NADW value associated with a sea 117 level rise of 3 m (0.5 m) (Fig. 1A). This feature is associated with a large decrease of Sahel 118 rainfall (10% to 60%) between 2030 and 2060 with respect to the RCP8.5 scenario (Fig. 2). This 119 spatial pattern of precipitation changes is similar to the one inferred from the large surge of 120 iceberg discharges that occurred in the past (2, 30). The tropical rainfall changes are linked to 121 the Northern Hemisphere cooling through atmospheric teleconnections. A north–south 122 thermal gradient between the Sahara (cooler) and the Guinea Gulf (warmer) appears (Fig. 1B). 123 This gradient leads to a rise of sea level pressure gradient, inducing low-level southward winds, 124 5 which block the monsoon system farther south (Fig. 1C). The Sahel becomes drier, and the 125 surface temperature increases; this causes an additional local temperature gradient that 126 strengthens the African Easterly Jet, causing a moisture export from this area (2, 31) (Fig. 1D). 127 These mechanisms underlying the drastic reduction of Sahelian precipitation are robust in 128 different climatic contexts with several models (9, 22). 129 Here we focus on the Western African Sahel region (8°N to 18°N; 17°W to 15°E). Because the 130 Sahelian population is strongly dependent on pastoralism and rainfed agriculture for 131 subsistence (32), our analysis is made in terms of summer precipitation changes [June to 132 September (JJAS)] during which most of the rainfall occurs (between 80% and 90%). To 133 circumvent the acknowledged difficulties of CMIP5 models (33) to properly capture the 134 mesoscale processes and therefore the monsoon rainfall, we applied a statistical method to 135 improve the IPSL simulated precipitation in the West African region with respect to the Water 136 and Global Change project (WATCH) Forcing Data methodology applied to the latest global 137 atmospheric reanalysis data produced by the European Centre for Medium-Range Weather 138 Forecasts (ERA-Interim) (WFDEI) reanalysis (34). This method, called “Cumulative Distribution 139 Function transform” (CDF-t), has been successfully applied in many climate-related studies 140 (e.g., refs. 35–38; Methods). 141 142 To illustrate the internal model variability, we considered a four-member dataset of the 143 RCP8.5 scenario, each member differing in initial conditions. The evolution of the corrected 144 precipitation in the West African Sahel region, obtained under the RCP8.5 dataset (baseline 145 experiments) and the four GrIS scenarios, is displayed in Fig. 3. However, the precipitation 146 signal simulated in response to the 0.5-m SLR perturbation is not statistically significant 147 compared with the four members of the RCP8.5 baseline experiment, as indicated by the t 148 test (P value <0.05; Methods), and the corresponding results will not be further discussed in 149 the following. 150 The effect of the FWF perturbation radically changes the evolution of precipitation averaged 151 over the Sahel region. The first key feature is a significant decrease of Sahel rainfall for the 152 three larger GrIS scenarios (i.e., 1-, 1.5-, and 3-m equivalent SLR) compared with the four-153 member RCP8.5 dataset. This decrease occurs almost concomitantly with the FWF release and 154 can be up to 30% over the period 2030–2060, reaching 3 mm⋅d−1, where the greatest 155 differences with the baseline experiment scenario are simulated (P value <0.05). When the 156 6 freshwater perturbation stops, Pav increases slightly, and values close to those of the baseline 157 experiment are recovered. 158 Increasing vulnerability 159 The Sahelian agroecosystem is likely to be strongly disturbed by these large precipitation 160 changes; this could have significant impacts on populations extremely dependent upon 161 rainfed agriculture for subsistence. It is documented that the rainfall decrease and the 162 temperature elevation in the Sahel will negatively impact yields of staple food cereals, such as 163 sorghum and millet (39). The water demand for these crops is calculated by Food and 164 Agriculture Organization (FAO) formulations (Methods) and depends on temperature. The 165 north–south gradient of water demand has a similar amplitude for sorghum and millet, 166 directly related to the temperature gradient. In the Sahel area, the sorghum needs, currently, 167 between 520 mm and 660 mm per growing period. The millet growth period is shorter than 168 that of sorghum and needs therefore less water (460 mm to 600 mm per growing period). The 169 water demand increases over the 21st century, due to the temperature increase. In average 170 on the Sahel area, the water demand values rise from 580 mm (515 mm) to 650 mm (580 mm) 171 per growing period for the sorghum (millet). 172 173 To quantify the impacts of rainfall decrease on the population, we analyze the gain or loss of 174 available area for agriculture relative to the adequacy between the sorghum water 175 requirement and the JJAS precipitation. Fig. 4A displays the variations of available area for 176 sorghum cultivation. Under the GrIS scenarios, a strong decrease of the cultivable area with 177 respect to 1976–2005 is observed between 2025 and 2100, up to ∼1,100,000 km2 for the 1-178 m GrIS melting scenario and even more for the 1.5- and 3-m GrIS melting scenarios. After 179 2070, the cultivable area slightly increases, and the RCP8.5 values are progressively recovered, 180 except for the 3-m scenario. 181 182 The large impact of the GrIS scenarios on the local population may be enhanced by a strong 183 demography dynamics in the Sahel. All of the projections of the demography evolution 184 suggest an increase of the population over Africa (40). However, these projections remain 185 uncertain and are strongly dependent on socioeconomic changes that will occur throughout 186 the 21st century (40, 41). To estimate the range of people affected by monsoon variations, we 187 analyze the human impacts related to a loss of cultivable areas for a demography fixed to that 188 7 of 2011 (lower bound) and for an evolving demography deduced from a shared socioeconomic 189 pathway (41) (SSP3 hereafter), which is consistent with the RCP8.5 scenario (upper bound). 190 191 Considering the Sahelian population fixed to its 2011 level (i.e., 135 million people, Fig. 4B), 192 the GrIS scenarios lead to a rapid growth (in less than 20 y) of people impacted by the loss of 193 cultivable area, up to ∼60 million people in the 1.5- and 3-m GrIS melting scenarios between 194 2040 and 2065, due to change in precipitation regimes. This number slightly decreases at the 195 end of the FWF perturbation. However, the most dramatic consequences are observed when 196 the population dynamics are accounted for (Fig. 4C). According to the SSP3 scenario, the 197 number of people living below the water threshold (Methods) for sorghum cultivation 198 undergoes a rapid and continuous increase, up to ∼360 million by the end of the 21st century. 199 This number represents one third of the population living in the Sahel area, showing that the 200 climatic impact is widely amplified by the demography explosion. This situation will put a 201 considerable strain on millet and sorghum subsistence agriculture. For local farmers, 202 migration might thus appear as a necessary option, especially if one considers the rapid 203 development of African metropolises. Options are, indeed, likely to be limited for local 204 farmers, and staying on their land would require substantial changes in agricultural techniques 205 and the abandonment of subsistence agriculture (42). 206 207 We demonstrated that Greenland melting during the 21st century could drastically affect the 208 climate, not only in high-latitude locations but also over the tropical areas, through 209 atmospheric and oceanic teleconnections. Although most studies focus on the coastal impacts 210 of SLR (43), we pointed out that Greenland melting could produce drastic droughts in the 211 Sahel, with many consequences for agricultural practices and for population migrations. In the 212 past, monsoon-dependent farmers have used the cities (44) and the coastal zones as a refuge 213 or a final migration destination following rainfall deficit years. Under the 1-m SLR scenario or 214 one involving higher SLR, coastal zones will be extremely destabilized, and migration to these 215 regions will be difficult, with a possible “coastal squeeze” (45), making the urban areas the 216 primary destination for migrants. Today, most migrant flows related to environmental 217 disruptions occur within their national or regional boundaries (46). A rapid melting of ice 218 sheets, however, is likely to lead to dramatic population shifts that would develop beyond 219 borders and would entail irreversible demographic impacts. 220 8 Methods 221 222 Model and experimental details 223 224 Model and Experimental Details 225 All of the experiments presented in this study have been carried out with the coupled 226 atmosphere–ocean IPSL-CM5A-LR model (19), which has been used for the CMIP5 exercise. 227 The atmospheric component has a spatial resolution of 3.75° × 1.875° in longitude and 228 latitude, respectively, with 39 vertical levels; the oceanic component uses an irregular grid 229 with a nominal resolution of 2°, and a higher latitudinal resolution of 0.5° in the equatorial 230 ocean, and 31 vertical levels. The locations of freshwater inputs have been designed to 231 produce a rapid response of the model. We therefore chose to release the freshwater in 232 locations of deep water formations, in the North Atlantic (45°N to 65°N, 45°W to 5°E), which 233 also coincides with regions of input of Greenland meltwater (47). Recent papers pointed out 234 relationships between Greenland melting and AMOC variations (48). The spread of FWF values 235 (0.11 Sv to 0.68 Sv) has been chosen to explore the impact of a large and rapid freshwater 236 input due to partial melting of the Greenland ice sheet. The highest FWF (0.68 Sv) accounts 237 for the fact that current climate models are possibly too stable in response to freshwater 238 release (49). A growing number of modeling results support this assumption by invoking (i) 239 intrinsic model biases in advection (50⇓⇓⇓–54) and/or in the stratification of the subpolar gyre 240 (55), (ii) an incorrect representation of freshwater pathways in the absence of an iceberg drift 241 module (10), or (iii) a too coarse resolution of current ocean models that are not eddy-242 resolving (56, 57). All these factors could potentially lead to a limited sensitivity of projected 243 AMOC to freshwater input. Thus, we analyze here the progressive reduction of the AMOC 244 corresponding to increased FWF and its potential impacts on the Sahelian region. More 245 importantly, moderate scenarios (corresponding to 0.11 Sv to 0.34 Sv) have to be considered 246 regarding the most recent ice sheet observations (e.g., refs. 13, 16, and 58). 247 Statistical method to adjust the IPSL simulated precipitation 248 The simulated precipitation have been corrected with respect to the WFDEI reanalyses 249 interpolated to a 0.5°x0.5° spatial resolution (34), used as a reference. Here, the “calibration” 250 9 period covers the 34-year time period 1979-2013, while the “projection” period covers the 251 94-year time period 2006-2099. 252 The bias correction method used in this study is a variant of the “Quantile-Mapping” approach 253 (e.g., (59, 60)) and allows to account for the climate change signal into the correction (37). 254 This method called “Cumulative Distribution Function – transform” (CDF-t) was initially 255 developed by (61) and has then been applied in many climate-related studies (e.g., (35–38), 256 among others). If X denotes the random variable representing the modelled variable to be 257 corrected, and Y the random variable representing the reference variable, CDF-t first 258 estimates the cumulative distributions 𝐹𝑌𝑝 and 𝐹𝑋𝑝 of the random variables 𝑌𝑝 and 𝑋𝑝 over 259 the projection (future) time period before applying a distribution-derived quantile-mapping, 260 i.e. trying to map a modelled value 𝑥𝑝 to a value 𝑦𝑝 such that their distributions are equivalent 261 (62) : 262 264 𝐹𝑌𝑝(𝑦𝑝) = 𝐹𝑋𝑝(𝑋𝑝) ⇔ 𝑦𝑝 = 𝐹𝑌𝑝−1[𝐹𝑋𝑝(𝑋𝑝)] (1) 263 265 If 𝐹𝑋𝑝 can be directly modelled – parametrically or not – from the data to be corrected in the 266 projection period, the modelling of 𝐹𝑌𝑝 is based on the assumption that a mathematical 267 transformation T allows going from 𝐹𝑋𝑐 to 𝐹𝑌𝑐 – the distribution of the random variables 𝑌𝑐 268 and 𝑋𝑐 in the calibration period, 269 𝑇[𝐹𝑋𝑐(𝑧)] = 𝐹𝑌𝑐(𝑧) (2) 270 271 for any z, and that T is still valid in the projection period: that is, 272 𝑇[𝐹𝑋𝑝(𝑧)] = 𝐹𝑌𝑝(𝑧) (3) 273 274 Replacing z by 𝐹𝑋𝑐 −1(𝑢) in eq. (2), where u is any probability in [0, 1], we obtain 275 𝑇(𝑢) = 𝐹𝑌𝑐[𝐹𝑋𝑐 −1(𝑢)] (4) 276 277 corresponding to a simple definition for T. Inserting eq. (4) in eq. (3) leads to a modelling of 278 𝐹𝑌𝑝: 279 𝐹𝑌𝑝(𝑧) = 𝐹𝑌𝑐 [𝐹𝑋𝑐 −1[𝐹𝑋𝑝(𝑧)]] (5) 280 281 10 Once 𝐹𝑋𝑝 and then 𝐹𝑌𝑝 are modelled, a distribution-based quantile-mapping is applied as in 282 eq. (1). Hence, this CDF-t approach includes the information about the distributions over the 283 projection time period in the quantile-mapping technique. More details can be found in (37). 284 To refine the bias correction method, a multivariate mapping could be performed, notably to 285 better account for effect of the meso-scale processes (e.g. AEJ instabilities giving rise to squall 286 lines) that could counteract those of the large-scale circulation. Such a multivariate approach 287 would require wind data in altitude that are not currently available. Moreover, multivariate 288 statistical bias correction methods are only emerging in the literature and are not yet ready 289 to be used. However, paleoclimatic data reveal that ice-sheet melting produced in the past a 290 strong decrease of the West African monsoon (3–5) with underlying mechanisms fully similar 291 to those highlighted in the present study (2), suggesting that the effect of jet instabilities is 292 insufficient to counterbalance the effect of large-scale circulation (i.e decrease of the 293 monsoon rainfall). 294 295 t-test for each simulation 296 To investigate the significance of the monsoon variations due to the freshwater input, we 297 use the t-test. We average the total monsoon precipitation on the Sahel area (8°N-18°N; 298 17°W-15°E) and compare each scenario with the RCP8.5 baseline experiment. The t-test (eq. 299 6) must be done with stationary series: 300 𝑡 = 𝑋𝑠𝑐𝑒𝑛− 𝑋𝑟𝑐𝑝85 √𝑠𝑠𝑐𝑒𝑛 2 𝑛𝑠𝑐𝑒𝑛 + 𝑠𝑟𝑐𝑝85 2 𝑛𝑟𝑐𝑝85 (6) 301 302 where: 303 t is the t-test result 304 X is the sample mean for the scenario under study and the RCP8.5 baseline scenarios 305 S² is the unbiased estimator of the variance of the two samples 306 n is the number simulated precipitation value in each scenario (i.e 10 for the RCP8.5 307 baseline experiment and 10 for each GrIS scenario) 308 However our scenarios are used in transient experiments. To circumvent this problem, we 309 calculate the t-test values 10 years by 10 years with a time lag of 1 year (i.e. 2006-2015, 2007-310 11 2016,...) to obtain 84 pseudo-stationarity periods by subsampling. We obtain a t-value for each 311 year between 2011 and 2094. For each t-test we have, 10 values for one GrIS scenario and 10 312 for the RCP8.5 one, leading to 18 degrees of freedom allowing to have a robust test. A longer 313 period would lead to non-stationarity of our time series and a shorter period to a test with a 314 too large variability, and therefore not usable. Using a probability threshold of 97.5% 315 combined with these 18 degrees of freedom, the critical value is 2.101. 316 317 Water demand of crops 318 The threshold of crop water demand evolves with time as a function of temperature: the crops 319 need more water when the temperature increases. The water demand of sorghum cultivation 320 can be obtained for each model grid point in the Sahel area (8°N-18°N; 17°W-15°E). It is 321 estimated with the evapotranspiration (ETo) given by the Blaney-Criddle technique (63) (eq. 322 7) with a correction factor kc, as suggested by the FAO eq. 8 (64), that accounts for specific 323 characteristics of a given crop specie: 324 𝐸𝑇𝑐𝑟𝑜𝑝 = 𝑘𝑐 × 𝐴 (7) 325 𝐸𝑇0 = 𝑝 (0.46 𝑇𝑚𝑒𝑎𝑛 + 8) (8) 326 327 where: 328 ETo is the potential evapotranspiration (mm/day) 329 ETcrop is the water demand for crop (mm/growing period) 330 Tmean = mean temperature over the monsoon period (° C) 331 A is the crop growing period duration (i.e 120 days for sorghum, 105 for millet) 332 p = percentage of daytime duration. 333 kc = crop factor: 0.78 for sorghum, 0.79 for millet 334 335 336 Surface area and population impacted by rainfall changes 337 To estimate the variations of the agricultural area due to rainfall changes and the number of 338 inhabitants impacted by the weakening of precipitation, we computed the land surface area 339 receiving an amount of precipitation below the required precipitation threshold for sorghum 340 cultivations. Since the number of inhabitants is given by a 0.5°x0.5° spatial resolution dataset, 341 12 provided by the Potsdam Institute for Climate Impact Research from a preliminary version of 342 the SSP population data (the 2012-05-11 data in the IIASA database), the rainfall has been bi-343 linearly interpolated on a 0.5°x0.5° grid. For each scenario (RCP 8.5 and GrIS), the area 344 impacted by rainfall change (R(t)) in the Sahel area (8°N-18°N; 17°W-15°E) is obtained year by 345 year with the following equation: 346 𝑅(𝑡) = ∑ 𝑅𝑠𝑐𝑒𝑛(𝑡) − ∑ 𝑅𝑟𝑒𝑓 (9) 347 348 where: 349 Rscen(t) represents the area covered by the grid points where the precipitation volume 350 is above the water demand of crops 351 Rref represents the area covered by the grid points where the precipitation averaged 352 over the last thirty-year climatic period (1976-2005) is above the water demand of crops. 353 354 To estimate the evolution of the cultivable area affected by a precipitation deficit we express 355 the number of corresponding pixels in km². When the number of pixels is negative (positive), 356 the area available for agriculture is smaller (larger) than that of the 1976-2005 climatic period. 357 The number of inhabitants impacted by rainfall changes is estimated by summing the number 358 of people living in the corresponding surface area. To count only the rural population with 359 only rainfed agriculture practices, the surface area where the current population density is 360 above 200 inhabitants/km² is excluded. A positive (negative) value means that a greater 361 (smaller) number of people is affected by rainfall changes compared to the reference period 362 (1976-2005). 363 Code and data availability 364 All data generated in this study by the IPSLCM5A-LR model for the Greenland scenarios as well 365 as and the Ferret and Python scripts produced for their analysis are available from the 366 corresponding author. Other results supporting this study are based on CMIP5 model, WFDEI 367 Re-analysis data and populations projection, which are available 368 respectively from http://cmip-pcmdi.llnl.gov/cmip5/data_portal.html, 369 http://www.eu-watch.org/data_availability and http://clima-dods.ictp.it/Users/fcolon_g/ISI-370 MIP/. 371 http://cmip-pcmdi.llnl.gov/cmip5/data_portal.html http://www.eu-watch.org/data_availability http://clima-dods.ictp.it/Users/fcolon_g/ISI-MIP/ http://clima-dods.ictp.it/Users/fcolon_g/ISI-MIP/ 13 Acknowledgements 372 We thank Michel Crucifix and an anonymous reviewer for their constructive comments 373 and suggestions that helped improve the manuscript, as well as Serge Janicot and Juliette 374 Mignot for fruitful discussions. We are also very grateful to Sarah Amram, Jean-Yves 375 Peterschmitt, and Aurélien Quiquet for technical support, and to Sandra Bouneau and 376 Sylvain David for numerous exchanges. This work was supported by the French Atomic 377 Commission (CEA) within the framework of the Variations Abruptes du Climat: 378 Conséquences et Impacts éNergétiques project funded by the Département des sciences 379 de la matières (DSM) with the DSM-Energie Program. It benefited from the high 380 performance computing (HPC) resources made available by Grand Equipement National 381 de Calcul Intensif, CEA, and Centre National de la Recherche Scientifique. The authors 382 thank the Potsdam Institute for Climate Impact Research for providing the gridded data 383 population (SSP3) based on a preliminary version of the SSP population data (the 2012-384 05-11 data in the IIASA database). This database has been elaborated within the Inter-385 Sectoral Impact Model Intercomparison Project. 386 387 References 388 1. Hulme M, Doherty R, Ngara T, New M, Lister D (2001) African climate change: 1900-2100. Clim 389 Res 17(2):145–168. 390 2. Mulitza S, et al. (2008) Sahel megadroughts triggered by glacial slowdowns of Atlantic 391 meridional overturning. Paleoceanography 23(4):PA4206. 392 3. Stager JC, Ryves DB, Chase BM, Pausata FSR (2011) Catastrophic drought in the Afro-Asian 393 monsoon region during Heinrich event 1. Science 331(6022):1299–302. 394 4. Itambi AC, Dobeneck T Von, Mulitza S, Bickert T, Heslop D (2009) Millennial-scale northwest 395 African droughts related to Heinrich events and Dansgaard-Oeschger cycles : Evidence in 396 marine sediments from offshore Senegal. 24:1–16. 397 5. Tjallingii R, et al. (2008) Coherent high- and low-latitude control of the northwest African 398 hydrological balance. Nat Geosci 1(10):670–675. 399 6. Broecker W, Bond G, Klas M, Clark E, McManus J (1992) Origin of the northern Atlantic’s 400 Heinrich events. Clim Dyn 6(3-4):265–273. 401 7. Alvarez-Solas J, Robinson A, Montoya M, Ritz C (2013) Iceberg discharges of the last glacial 402 period driven by oceanic circulation changes. Proc Natl Acad Sci 110 (41 ):16350–16354. 403 8. Kageyama M, Mignot J (2009) Glacial climate sensitivity to different states of the Atlantic 404 Meridional Overturning Circulation: results from the IPSL model. Clim Past 5(3):551–570. 405 9. Swingedouw D, et al. (2009) Impact of Freshwater Release in the North Atlantic under 406 Different Climate Conditions in an OAGCM. J Clim 22(23):6377–6403. 407 10. Swingedouw D, et al. (2013) Decadal fingerprints of freshwater discharge around Greenland in 408 a multi-model ensemble. Clim Dyn 41(3-4):695–720. 409 11. Church JA, et al. (2013) Sea Level Change. Climate Change 2013: The Physical Science Basis. 410 Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental 411 Panel on Climate Change [Stocker, T.F., D. Qin, G.-K. Plattner, M. Tignor, S.K. Allen, J. 412 Boschung, A. Nauels, Y. Xia, pp 1137–1216. Cambridge . 413 12. Straneo F, et al. (2013) Challenges to Understanding the Dynamic Response of Greenland’s 414 Marine Terminating Glaciers to Oceanic and Atmospheric Forcing. Bull Am Meteorol Soc 415 14 94(8):1131–1144. 416 13. Rignot E, Fenty I, Xu Y, Cai C, Kemp C (2015) Undercutting of marine-terminating glaciers in 417 West Greenland. Geophys Res Lett 42(14):5909–5917. 418 14. Gillet-Chaulet F, et al. (2012) Greenland ice sheet contribution to sea-level rise from a new-419 generation ice-sheet model. Cryosph 6(6):1561–1576. 420 15. Paolo FS, Fricker HA, Padman L (2015) Volume loss from Antarctic ice shelves is accelerating. 421 Science 348(6232):327–332. 422 16. Mouginot J, et al. (2015) Fast retreat of Zachariæ Isstrøm , northeast Greenland. Science 423 350(6266):1357–1361. 424 17. Alvarez-Solas J, Ramstein G (2011) On the triggering mechanism of Heinrich events. Proc Natl 425 Acad Sci 108(50):E1359–E1360. 426 18. Gomez N, Gregoire LJ, Mitrovica JX, Payne AJ (2015) Laurentide-Cordilleran Ice Sheet saddle 427 collapse as a contribution to meltwater pulse 1A. Geophys Res Lett 42(10):3954–3962. 428 19. Dufresne J-L, et al. (2013) Climate change projections using the IPSL-CM5 Earth System Model: 429 from CMIP3 to CMIP5. Clim Dyn 40(9-10):2123–2165. 430 20. Manabe S, Stouffer RJ (1988) Two Stable Equilibria of a Coupled Ocean-Atmosphere Model. J 431 Clim 1(9):841–866. 432 21. Vellinga M, Wood RA, Gregory JM (2001) Processes Governing the Recovery of a Perturbed 433 Thermohaline Circulation in HadCM3. J Clim 15:764–780. 434 22. Stouffer R, Yin J, Gregory J (2006) Investigating the causes of the response of the 435 thermohaline circulation to past and future climate changes. J Clim 19(8):1365–1387. 436 23. Kageyama M, et al. (2013) Climatic impacts of fresh water hosing under last glacial Maximum 437 conditions: A multi-model study. Clim Past 9(2):935–953. 438 24. Jackson R, et al. (2015) Global and European climate impacts of a slowdown of the AMOC in a 439 high resolution GCM. Clim Dyn:3299–3316. 440 25. Swingedouw D, Braconnot P, Marti O (2006) Sensitivity of the Atlantic Meridional Overturning 441 Circulation to the melting from northern glaciers in climate change experiments. 33(April):1–442 4. 443 26. Peterson LC, Haug GH, Hughen KA, Röhl U (2007) Tropical Atlantic During the Last Glacial 444 Rapid Changes in the Hydrologic Cycle of the Tropical Atlantic During the Last Glacial. Science 445 (80- ). doi:10.1126/science.290.5498.1947. 446 27. Leduc G, et al. (2007) Moisture transport across Central America as a positive feedback on 447 abrupt climatic changes. Nature 445(7130):908–11. 448 28. Chiang JCH, Bitz CM (2005) Influence of high latitude ice cover on the marine Intertropical 449 Convergence Zone. Clim Dyn. doi:10.1007/s00382-005-0040-5. 450 29. Arbuszewski J a., deMenocal PB, Cléroux C, Bradtmiller L, Mix A (2013) Meridional shifts of the 451 Atlantic intertropical convergence zone since the Last Glacial Maximum. Nat Geosci. 452 doi:10.1038/ngeo1961. 453 30. Niedermeyer EM, et al. (2009) Extratropical forcing of Sahel aridity during Heinrich stadials. 454 Geophys Res Lett 36:L20707. 455 31. Liu Y, Chiang JCH, Chou C, Patricola CM (2014) Atmospheric teleconnection mechanisms of 456 extratropical North Atlantic SST influence on Sahel rainfall. Clim Dyn 43(9-10):2797–2811. 457 15 32. Ickowicz A, et al. (2012) Crop-Livestock Production Systems in the Sahel - Increasing Resilience 458 for Adaptation to Climate Change and Preserving Food Security. Building Resilience for 459 Adaptation to Climate Change in the Agriculture Sector, pp 261–294. 460 33. Christensen JH, et al. (2013) Climate Phenomena and their Relevance for Future Regional 461 Climate Change. Climate Change 2013: The Physical Sci - Ence Basis. Contribution of Working 462 Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change 463 [Stocker, T.F., D. Qin, G.-K. Plattner, M. Tignor, S.K. Allen, J. Boschung, A. Nauels, Y. X. 464 34. Weedon GP, et al. (2014) The WFDEI meteorological forcing data set: WATCH Forcing Data 465 methodology applied to ERA-Interim reanalysis data. Water Resour Res 50(9):7505–7514. 466 35. Oettli P, Sultan B, Baron C, Vrac M (2011) Are regional climate models relevant for crop yield 467 prediction in West Africa? Environ Res Lett 6(1):14008. 468 36. Colette A, Vautard R, Vrac M (2012) Regional climate downscaling with prior statistical 469 correction of the global climate forcing. Geophys Res Lett 39(13):L13707. 470 37. Vrac M, et al. (2012) Dynamical and statistical downscaling of the French Mediterranean 471 climate: uncertainty assessment. Nat Hazards Earth Syst Sci 12(9):2769–2784. 472 38. Vigaud N, Vrac M, Caballero Y (2013) Probabilistic downscaling of GCM scenarios over 473 southern India. Int J Climatol 33(5):1248–1263. 474 39. Sultan B, et al. (2013) Assessing climate change impacts on sorghum and millet yields in the 475 Sudanian and Sahelian savannas of West Africa. Environ Res Lett 8(1):014040. 476 40. KC S, Lutz W (2014) The human core of the shared socioeconomic pathways: Population 477 scenarios by age, sex and level of education for all countries to 2100. Glob Environ Chang. 478 doi:10.1016/j.gloenvcha.2014.06.004. 479 41. O’Neill BC, et al. (2014) A new scenario framework for climate change research: the concept 480 of shared socioeconomic pathways. Clim Change 122(3):387–400. 481 42. Juana JS, Kahaka Z, Okurut FN (2013) Farmers’ Perceptions and Adaptations to Climate 482 Change in Sub-Sahara Africa: A Synthesis of Empirical Studies and Implications for Public 483 Policy in African Agriculture. J Agric Sci 5(4):121–135. 484 43. Neumann B, Vafeidis AT, Zimmermann J, Nicholls RJ (2015) Future Coastal Population Growth 485 and Exposure to Sea-Level Rise and Coastal Flooding - A Global Assessment. PLoS One 486 10(3):e0118571. 487 44. Afolayan AA, Adelekan IO (1999) The role of climatic variations on migration and human 488 health in Africa. Environmentalist 18(4):213–218. 489 45. Pontee N (2013) Defining coastal squeeze: A discussion. Ocean Coast Manag 84:204–207. 490 46. McLeman R a., Hunter LM (2010) Migration in the context of vulnerability and adaptation to 491 climate change: Insights from analogues. Wiley Interdiscip Rev Clim Chang 1(3):450–461. 492 47. Bamber J, van den Broeke M, Ettema J, Lenaerts J, Rignot E (2012) Recent large increases in 493 freshwater fluxes from Greenland into the North Atlantic. Geophys Res Lett 39(19):L19501. 494 48. Rahmstorf S, et al. (2015) Exceptional twentieth-century slowdown in Atlantic Ocean 495 overturning circulation. Nat Clim Chang 5(5):475–480. 496 49. Valdes P (2011) Built for stability. Nat Geosci 4(7):414–416. 497 50. Liu W, Liu Z, Brady EC (2014) Why is the AMOC monostable in coupled general circulation 498 models? J Clim 27(6):2427–2443. 499 16 51. Liu W, Xie S, Liu Z, Zhu J (2017) Overlooked possibility of a collapsed Atlantic Meridional 500 Overturning Circulation in warming climate. Sci Adv 3(January):e1601666. 501 52. Deshayes J, et al. (2013) Oceanic hindcast simulations at high resolution suggest that the 502 Atlantic MOC is bistable. Geophys Res Lett 40(12):3069–3073. 503 53. Hofmann M, Rahmstorf S (2009) On the stability of the Atlantic meridional overturning 504 circulation. Proc Natl Acad Sci U S A 106(49):20584–20589. 505 54. Weber SL, et al. (2006) The modern and glacial overturning circulation in the Atlantic ocean in 506 PMIP coupled model simulations. Clim Past Discuss 2(5):923–949. 507 55. Sgubin G, Swingedouw D, Drijfhout SS, Mary Y, Bennabi A (2017) Abrupt cooling over the 508 North Atlantic in modern climate models. Nat Commun 8:14375. 509 56. Spence P, Saenko OA, Sijp W, England MH (2013) North Atlantic Climate Response to Lake 510 Agassiz Drainage at Coarse and Ocean Eddy-Permitting Resolutions. 2651–2667. 511 57. Weijer W, Maltrud ME, Hecht MW, Dijkstra HA, Kliphuis MA (2012) Response of the Atlantic 512 Ocean circulation to Greenland Ice Sheet melting in a strongly-eddying ocean model. 513 39(May):1–6. 514 58. Noël B, et al. (2017) A tipping point in refreezing accelerates mass loss of Greenland’s glaciers 515 and ice caps. 8:14730. 516 59. Déqué M (2007) Frequency of precipitation and temperature extremes over France in an 517 anthropogenic scenario: Model results and statistical correction according to observed values. 518 Glob Planet Change 57(1):16–26. 519 60. Gudmundsson L, Bremnes JB, Haugen JE, Engen-Skaugen T (2012) Technical Note: 520 Downscaling RCM precipitation to the station scale using statistical transformations–a 521 comparison of methods. Hydrol Earth Syst Sci 16(9):3383–3390. 522 61. Michelangeli P, Vrac M, Loukos H (2009) Probabilistic downscaling approaches: Application to 523 wind cumulative distribution functions. Geophys Res Lett 36(11):L11708. 524 62. Panofsky HA, Brier GW (1958) Some applications of statistics to meteorology (Mineral 525 Industries Extension Services, College of Mineral Industries, Pennsylvania State University). 526 63. Blaney HF, Criddle WD (1965) Determining Water Requirements For Settling Water Disputes. 527 Nat Resour J 4(1938):29–41. 528 64. FAO (2016) 2. Water and soil requirements. Available at: 529 http://www.fao.org/docrep/u3160e/u3160e04.htm. 530 531 532 17 Figure 1: Teleconnection mechanisms linking the Greenland ice sheet melting and the decrease of the Sahelian rainfall. A) AMOC evolution (in % with respect to the present-day values),; B) Mean annual temperature anomaly between the 1.5 m GrIS and the RCP8.5 scenarios averaged over the period 2030-2060; C) Same as B) for the sea-level pressure anomaly; and the 10-m winds (black arrows); D) Same as B) for the mean summer (JJAS) temperature anomaly and the 850 hPa winds (black arrows). Figure 2: JJAS precipitation anomaly between the 1.5 m GrIS scenario and the RCP 8.5 baseline experiment normalized to the RCP 8.5 values and averaged over 2030-2060. A value of 100 corresponds to a doubling of precipitation and -100 corresponds to zero precipitation. The precipitation values are obtained after applying the statistical method (see Methods). The blue box (8°N-18°N , 17°W-15°E) represents the region under study. 18 Figure 3: Evolution of JJAS precipitation during the 21st century averaged over the Sahel area (8°N- 18°N, 17°W-15°E) for the RCP8.5 and the GrIS scenarios. The orange star indicates the simulated JJAS precipitation over the climatic reference period (1976-2005) deduced from the IPSL-CM5A simulated precipitation (4.96 mm). To illustrate the internal model variability, we considered a 4-member dataset of the RCP8.5 scenario, each member differing in initial conditions. The area delimited by the two grey curves represents the range of model variability deduced from the 4-member dataset. 533 19 534 Figure 4: Impacts of rainfall change on sorghum cultivation and on population. A) Evolution of the surface area available for sorghum cultivation (i.e. when the average JJAS precipitation is above the sorghum water demand) for each GrIS scenario and for the baseline experiment. The evolution of the available cultivable area is given with respect to the available area averaged over the 1976-2005 reference period deduced from the historical IPSL-CM5A simulated precipitation Negative (positive) values indicate a loss (gain) of cultivable area; B) Evolution of the number of inhabitants living under the sorghum water demand with respect to the 1976-2005 historical reference period. This evolution is estimated with the assumption that the number of inhabitants is fixed to its 2011 level; C) Evolution of the number of inhabitants living under the sorghum water demand with respect to the 1976-2005 historical reference period.Here, this evolution accounts for the evolution of demography provided by the SSP3 scenario. Both in (B) and (C), positive values indicate that the number of inhabitants living under the sorghum water demand increases with respect to the reference period.