South Asia suffers severe water scarcity even under the Paris Agreement

Mehnaz Rashid National Taiwan University https://orcid.org/0000-0002-0602-5298 Ren-Jie Wu National Taiwan University Yoshihide Wada The International Institute for Applied Systems Analysis https://orcid.org/0000-0003-4770-2539 Hannes Müller Schmied Goethe University Frankfurt https://orcid.org/0000-0001-5330-9923 Min-Hui Lo (  minhuilo@ntu.edu.tw ) Department of Atmospheric Sciences, National Taiwan University, Taipei, Taiwan https://orcid.org/0000-0002-8653-143X

water availability changes seasonally in sub-tropical and tropical regions due to monsoonal rainfall 79 and river discharge changes 15,20,21 . Therefore, WS and the changes in the hydrological components 80 should be evaluated on a seasonal scale.

127
Results from the ensemble of two hydrological and three climate models show that WS in SA will 128 increase significantly in a 1.5 º C warming scenario with a robust spatiotemporal variability (Fig.  129 1a), and the absolute changes compared to the pre-industrial period are considerable (Fig. 1b). 130 Spatial WS hotspots in the region are located in North-central, Northwestern and Southern India, 131 Eastern Pakistan, Northern, and Northwestern Bangladesh; given the socio-economic situation, 132 these areas can be most vulnerable to water crises in the future. On the other hand, the Himalayan 133 region (Nepal, Bhutan, North and Northeastern India, and Northern Pakistan), Central-eastern 134 India, Western Pakistan, and most of Sri Lanka are relatively unaffected. A robust seasonality in 135 WS is observed, which is lowest in JJAS and peaks in DJF and MAM (Fig.1). To explore how 136 strongly seasonality affects water security in the future, Table 1 shows the exposure of the human 137 population to WS in each season. For example, in JJAS, 157 million (7.6% of the total population 138 in SA) people are projected to face severe WS, and the number increases to 341 million (16.5%), 139 610 million (29.4%), and 558 million (28.4%) in ON, DJF, and MAM, respectively. Of the 140 numbers mentioned above, India alone is projected to have 494 million (31.3% of the total 141 population in India) and 503 million (31.8%) people under severe WS in DJF and MAM, 142 respectively (Table 1 and Supplementary Table 1).  143   144 We observe a considerable increase in the number of people living under severe WS in a 1.5 ˚C 145 warmer world compared to the pre-industrial period. For example, population living under severe 146 WS will increase from 1.5% to 7.6 %, 3.7% to 16.5%, 5% to 29.4%, and 6% to 28.4% (% of total 147 population) in JJAS, ON, DJF, and MAM, respectively (Table 1 and Supplementary Table 1 Altogether, about 42% of the South Asian population will suffer from moderate to severe WS for 157 half of the year (Table 1). This will result in water crises in this already vulnerable region with 158 severe over-exploitation of freshwater resources. WS in both 1.5 and 2 º C warming scenarios is 159 analogous compared to the pre-industrial period and additional warming of 0.5 º C does not reveal 160 a significant change in WS pattern (Fig. 1b, Supplementary Fig. 1b). This is primarily due to stable  Table 2). 165 166 Two major factors affecting the WS in a warmer world are water availability and water 167 consumption (Fig. 2). Water availability shows enormous heterogeneity in its seasonal and spatial 168 patterns, and the absolute changes compared to the pre-industrial period are extensive (Fig. 2a). 169 JJAS is the most plentiful water season, with the average availability going up to 1200 mm in some 170 of the prominent river basins ( Supplementary Fig. 4a), signifying the importance of both summer 171 monsoon (which supplies approximately 80% of the rainfall to the region) 47  In this study, we evaluate WS in SA (India, Pakistan, Bangladesh, Nepal Bhutan, Sri Lanka, and 291 the Maldives) under the Paris Agreement temperature targets of 1.5 and 2 º C temperature increase 292 compared to the pre-industrial period. We also examine the spatial as well as temporal WS hotspots. 293 Further, we evaluate the two important dimensions of WS, availability and consumption, and the 294 hydrological variables and compute their absolute changes compared to the preindustrial period. 295 To do so, we use the outputs from the two global hydrological models (GHMs where the water 296 consumption is available) PCR-GLOBWB 54,55 and WaterGAP2 56 , forced by three global climate 297 models (GCMs), MIROC5, HadGEM2-ES, and GFDL-ESM2M from Inter-Sectoral Model 298 Intercomparison Project Phase 2b (ISIMIP2b) under the RCP 6.0 middle emission scenario 46 . Thus 299 we generate data for three worlds: one representing natural conditions in the pre-industrial 300 reference period (1661-1860; as mentioned in the ISIMIP protocol with pre-industrial socio-301 economic conditions fixed at 1860) 46 and two future scenarios: 1) a 1.5 º C warmer world and 2) a 302 2 º C warmer world; using the RCP 6.0 middle emission scenario 46 . For the RCP6.0 scenario, future 303 climate and CO2 concentration vary as per RCP 6.0, while human interferences including land use 304 represent 2005 societal conditions. 305

306
We choose the two GHMs specifically because they provide actual water consumption data. 307 Furthermore, we select the three GCMs for the reason that they cross the 1.5 and 2 ℃ temperature 308 thresholds under RCP 6.0, a prerequisite for this study. The individual GCMs cross the 1.5 and 2 309 ˚C temperature thresholds compared to the preindustrial period at different times. For example, a 310 1  (Table S3; https://www.isimip.org/protocol/isimip2b-temperature-thresholds-and-time-315 slices/). 316

317
The variables used in this analysis include: water consumption, water availability, precipitation, 318 evapotranspiration (ET), and discharge, all the variables are available from ISIMIP2b at a spatial 319 resolution of 0.50×0.50, except water availability, which is calculated separately. We calculate 320 seasonal water consumption, discharge, and ET for each GCM-GHM combination independently. Here we unequivocally use water consumption, so that the water consumed within a grid cell is no 337 longer available to other users 16,57 . In SA, due to immense irrigation, a large proportion of water 338 is abstracted 58-60 ; therefore, taking withdrawal might result in overestimation of water scarcity as 339 a part of withdrawal is either available to downstream users or as return flow to the groundwater 340 system 57,61 . 341 343 Where WSI is the water scarcity index in cell "i" and for the season "s"; AWC (mm/month) is the 344 actual water consumption in cell "i" for the season "s"; and AWA (mm/month) is the actual water 345 available in cell "i" for the season "s". For example, if more than 40% of the available water is 346 consumed, WS is graded as severe; and similarly, if the consumption ranges between 20-40%, WS 347 is said to be moderate; and when the consumption is less than 20%, there is no WS, the 20% 348 threshold is set based on the amount of water that is required to maintain environmental flows, as 349  (Table S3)  for each warming scenario. Thus we show population moving "into" and "out of" the different WS 364 categories seasonally, for example, in JJAS, 157 million (7.6% of the total population in SA) 365 people are projected to face severe WS, and the number increases to 341 million (16.5%), 610 366 million (29.4%), and 558 million (28.4%) in ON, DJF, and MAM, respectively in a 1.5 º C warming 367 scenario (Table 1). Results for the 2 º C scenario are shown in supplementary table 2.  368   369 We also calculate absolute population and the percentage of the total population under different 370 WS categories in each season of the year for the pre-industrial reference period (supplementary 371 Table 1). Thus we provide the assessment of the increase in the absolute and the percentage of the 372 total population under different WS categories in each warming scenario compared to the pre-373 industrial period. For example, the total population living under severe WS will increase from 4 374 (1.5%) to 157 (7.6%), 10.5 (3.7%) to 341 (16.5%), 16 (5%) to 610 (29.4%), and 18 (6%) to 558 375 (28.4%) million in JJAS, ON, DJF, and MAM, respectively in 1.5 º C warming scenario compared 376 to the preindustrial period. 377 378

Hydrological Variables 379
Here we first perform the individual GCM-GHM simulations for discharge and ET, and only GCM 380 simulations for precipitation for 1.5 and 2 ˚C warming scenarios, producing 120 datasets 381 ( Supplementary Figs 9-14). The analysis is conducted at a seasonal scale and the ensemble 382 projections are then used to show the spatio-temporal alterations in the hydrological variables for 383 both warming scenarios ( Supplementary Figs. 2-3). The scenario analysis reveals a significant 384 alteration in the hydrological variables that contributes to the strong seasonality in water 385 availability. We further notice intensification in precipitation and ET in particular under a 2 ˚C 386 warming scenario. Hence it is important to have a detailed future study focusing on the assessment 387 of hydrological fluxes vis-à-vis hydrological extremes under the Paris Agreement. 388 389

Regression Analysis 390
For linear regression, we consider two climate-driven hydrological fluxes that is precipitation and 391 ET as explanatory variables for WS, given that discharge is one of the parameters in WS 392 assessment (see Equ.1; Methods). Following the methodology of Wu et al 2020 63 , we use the 393 following equation to perform regression analysis of seasonal precipitation and ET on WS for both 394 1.5 and 2 ºC warming scenarios. 395 396 * = 1 * 1 * + 2 * 2 * (2) 397 398 Where "y*" is the standardized (subtracting the mean and dividing by its standard deviation) 399 seasonal WS, and " * " is the corresponding standardized seasonal precipitation and ET at each 400 grid cell. a 1 * and a 2 * are the partial regression coefficient for precipitation and ET respectively. 401 Only the grid cells with the statistical significance F-test at a 95% confidence level are considered 402 for this analysis. The contribution of different variables ( ) are calculated as follows: 403 The contribution of precipitation and ET on the WS is shown in Fig. 3 and Supplementary Fig. 7