On the occurrence of the worst drought in South Asia in the observed and future climate

We obtained precipitation, maximum and minimum temperatures, and wind speed at 0.5° spatial resolution from the four global datasets [Climatic Research Unit (CRU-v4.02) (Harris et al 2020), Princeton Data (Sheffield et al 2006), University of Delaware (UDEL-v5.01) (Legates and Willmott 1990), and WATCH-Forcing-Data-ERA-Interim (WFDEI) (Weedon et al 2014)] for South Asia for 1951–2016. These datasets are developed using different methods and observations to prepare the gridded data of precipitation, temperature, and wind speed. For instance, CRU data was developed using the angular-distance weighting interpolation method on the monthly climate anomalies from a large network of observed stations (Harris et al 2020). Princeton forcing data was developed by combining the number of global observation-based datasets (station and satellite-based) and National Centers for Environmental Prediction–National Center for Atmospheric Research reanalysis (Sheffield et al 2006). The UDEL data was developed using the global gauge measurements from Global Historical Climatology Network, Global Synoptic Climatology Network archive, Global Surface Summary of Day, and station observations from different countries (Legates and Willmott 1990). The WFDEI meteorological data was developed using the WATCH forcing data (WFD) based on the bias-corrected ERA-40 reanalysis (Uppala et al 2005, Weedon et al 2014). These datasets have been widely used in the previous drought assessments (Sheffield et al 2009, Wang et al 2011, Hoerling et al 2012, Stagge et al 2017) and other hydroclimatic extremes (Ali et al 2018, Aadhar and Mishra 2019, Philip et al 2020). Using daily precipitation, maximum and minimum temperatures, and wind speed from the four observational datasets, we developed the four sets of meteorological forcing for variable infiltration capacity (VIC; version: 4.1) model to simulate the soil moisture for the period 1951–2016 (please see supplemental text S1 for more details (available online at stacks.iop.org/ERL/16/024050/mmedia)). We also used rice yield and vegetation index data to evaluate the impact of worst drought on agriculture productivity. Rice yield data was obtained from the Food and Agriculture Organization (FAO; www.fao.org/faostat/en/#data/QC) for India, Pakistan, Sri Lanka, Nepal, Bhutan, and Bangladesh for the period 1961–2016. We also used the Normalized Difference Vegetation Index (NDVI) data based on Advanced Very High-Resolution Radiometer (AVHRR) sensor from National Oceanic and Atmospheric Administration (NOAA) for the period 1982–2016.

The VIC model (Liang et al 1994, 1996b, Cherkauer et al 2003) is a physically-based land surface hydrological model that is used to simulate the sub-daily or daily soil moisture for each grid using full water and energy budget modes. We set up the VIC model at 0.5° spatial resolution and simulated daily soil-moisture using the full water budget mode for the period 1951–2016. Soil moisture in the VIC model is simulated in the three soil layers using Richard's equation, where the top two soil layers contribute to surface runoff. The VIC model does not consider any lateral flow from one grid to another. The bottom (third) soil layer contributes to the baseflow and considers the gravity-based drainage. Soil moisture simulated using the VIC model considers the sub-grid variability of vegetation, topography, and storage capacity (Liang et al 1996a). We obtained the vegetation parameter from NOAA-AVHRR (Hansen et al 2000) and derived soil parameters using the Harmonic World Soil Database. The calibrated soil parameters for the Indian sub-continent basins were used from Mishra et al (2019). Moreover, we evaluated the VIC simulated soil moisture using the satellite-based soil-moisture (Global Land Evaporation Amsterdam Model; GLEAM). The VIC simulated soil-moisture performs well compared to the GLEAM (Martens et al 2017) soil moisture during the period 1980–2016 (figure S1). The GLEAM soil moisture is based on the assimilation of satellite-based soil moisture. We find that all the products capture the year-to-year variability in soil moisture, however, there are large differences in the soil moisture trends due to different observations (figure S1). Moreover, GLEAM uses satellite-based soil moisture, dynamic vegetation, and considers the role of human influence (e.g., irrigation) in soil moisture estimation. As soil moisture simulations conducted using the VIC model did not consider the role of human influence, the differences between the VIC and GLEAM soil moisture are expected. The VIC model has been widely used to reconstruct the historical droughts and estimation of soil moisture in the previous studies (Sheffield 2004, Mishra and Cherkauer 2010, Nijssen et al 2014, Shah and Mishra 2015).

We used daily meteorological forcing (precipitation, minimum temperature, maximum temperature, and wind speed) from the Community Earth System Model Version 1 (CESM) for the historic (1920–2005) and future (2006–2100) periods (Kay et al 2015). CESM Large Ensemble (hereafter: CESM-LENS) provides simulations for 40 ensemble members for the period 1920–2100. Each ensemble member from CESM-LENS uses slightly different atmospheric initial conditions with the same radiative forcing (unique climate trajectory and same external forcing in each run; Kay et al 2015). The CESM-LENS simulations were performed for the representative concentration pathway 8.5 (RCP-8.5) that assumes an increase of 8.5 W m⁻² in the radiative forcing at the end of 21st century. The CESM-LENS simulations provide access to large datasets (40 ensemble members) to examine the role of internal climate variability. Moreover, to examine the role of anthropogenic activities (greenhouse gas emission and industrial aerosol) on droughts in South Asia, we used daily meteorological forcing from the 'ALL-but-one-forcing-fixed' experiments of CESM-LENS (Deser et al 2020b; www.cesm.ucar.edu/working_groups/CVC/simulations/cesm1-single_forcing_le.html). The 'ALL-but-one-forcing-fixed' experiments follow the same simulation protocol as CESM-LENS, except that one forcing agent (XGHG or XAER) is held fixed at 1920 conditions (Deser et al 2020a). The XGHG (fixed greenhouse gas) and XAER (fixed industrial aerosol) provide daily meteorological forcing (precipitation, minimum temperature, maximum temperature, and wind) for 20 different initial conditions for the period 1920–2080 to examine the role of anthropogenic activities.

We estimated characteristics (intensity, duration, and severity) of soil moisture droughts during the observed and future climate over South Asia using the VIC model simulated soil moisture for the historical (1951–2016) and future climate (2035–2100). We selected periods of 66 years in both historic and future climate to compare the occurrence of droughts in South Asia. The droughts were identified using the Standardized Soil Moisture Index (SSI, Hao and Aghakouchak 2013). SSI was estimated at the monthly timescale using the monthly root zone (60 cm, Mishra et al 2018, 2019) soil moisture. We estimated the intensity, area, and duration of drought spell using the area-averaged time-series of 1 month SSI over South Asia. A drought spell was identified using the onset and termination of a drought. The onset of a drought spell was identified if 1 month SSI is below zero for a continuous four months. Similarly, the termination of the drought was identified if 1 month SSI is above zero for a continuous four months. The area under drought (in percentage area of South Asia) was estimated for each month using the 1 month SSI below −0.5 considering the abnormal drought condition (Svoboda et al 2002). For each drought spell, we estimated the duration, mean intensity, and mean area under the drought. Then, the overall drought severity score (mean intensity × duration × area) was estimated for each drought spell to identify the worst drought during the observed climate of 1951–2016 (Mishra 2020a).

First, we estimated the drought characteristics (intensity, duration, and area) for the drought spells using monthly SSI based on the VIC model-simulated soil moisture using meteorological forcing from CRU, Princeton, UDEL, and WFDEI data (table 1, figure S3). We estimated the overall severity score (intensity × area × duration; in months × %) for all the identified drought spells in all four datasets (table 1). We find that the 1999–2006 drought spell was the worst drought in South Asia during the 1951–2016 period in all the four datasets (table 1, figure S3). However, as expected, the worst drought had different overall severity score due to differences in the forcing from the four observational products. The worst drought started in 1999 and continued till 2006 (figure 1) in all the datasets (CRU, Princeton, UDEL, and WFDEI). The overall severity score of the drought varied between 746 and 1019 among the four datasets (table 1). The drought achieved its peak intensity in July 2002 and affected more than 65% of South Asia (figures 1 and S4; hereafter, the 2002 drought for 1999–2006 drought spell). The 2002 drought covered entire South Asia except the northeastern parts of India (figure 1).

Zoom In Zoom Out Reset image size

We find that the 2002 drought affected the central parts of South Asia more severely in all the four datasets (figure 1), which is the major agricultural region in South Asia and may be affected the food and agricultural productivity in the region. Therefore, to evaluate the impact of drought on crop yield, we estimated the first difference of rice yield (Lobell and Field 2007, Mishra et al 2020) for India, Pakistan, Bhutan, Bangladesh, Nepal, and Sri Lanka (figure S5). The 2002 drought resulted in a considerable reduction in rice yield in India, Nepal, and Sri-Lanka (figure S5). For instance, rice yield was decreased by 470 kg ha⁻¹ in India in 2002 due to the severe drought condition, which was the highest reduction in yield during the 1961–2016 period. Moreover, we find an increase in rice yield in Bhutan and Bangladesh in 2002, which can be attributed to wet soil moisture conditions in the Northeast region (figure 1). We also estimated the NDVI anomaly for the 2002 drought during the monsoon season (June–September) over South Asia (figure S5), which indicates a considerable decline in NDVI in the northwestern parts of South Asia. However, the positive NDVI anomalies in Indo-Gangetic plains are associated with groundwater irrigation and human activities. The results of 2002 drought impact on agricultural productivity are consistent with the previous studies (Mishra et al 2014, 2020, Aadhar and Mishra 2017). The decline in the agricultural yield in 2002 was also associated with the concurrent hot and dry conditions (Mishra et al 2020). The droughts during the monsoon season in South Asia often overlap with positive air temperature anomalies (Mishra et al 2020). Therefore, we estimated temperature anomaly for the year 2002 over South Asia using the mean air temperature from CRU, WFDEI, UDEL, and Princeton (figure 1). We find that the temperature anomaly in July 2002 was significantly high (∼2 °C) in the parts of South Asia, which can be partly attributed by depletion of soil-moisture due to land–atmospheric interaction (Fischer et al 2007, Miralles et al 2014, Horton et al 2016) (figure 1). Central India experienced a temperature anomaly of more than 2 °C during July 2002 in all the four datasets (figure 1), which might have exacerbated soil moisture deficit and water availability in South Asia.

All the four datasets captured the soil moisture drought and associated temperature anomalies during 2002 over South Asia (figure 1). However, soil moisture droughts based on the four observed datasets show considerable uncertainty in their characteristics (intensity, area, duration, and severity) (figures 1, S3 and S4), which highlights the need of considering different observational products for the drought assessment. For instance, the 2002 drought was the worst drought in CRU, WFDEI, UDEL, and Princeton, however, with the overall severity scores of 1019, 1000, 818, and 746, respectively (figure 1). Moreover, the intensity, duration, and areal coverage of the 2002 drought exhibited a large uncertainty (figures 1 and S4, table 1). For instance, the duration of the 2002 drought varies between 43 and 78 months in all the four datasets (table 1). Similar to the 2002 drought, the four datasets showed considerable uncertainty in the other drought spells and their characteristics during 1951–2016 (table 1). We identified 17 drought spells using 1 month SSI based on WFDEI data during 1951–2016. However, only 14 drought spells were identified using SSI based on the CRU and UDEL datasets during the same period. The difference in the drought spells was mainly due to uncertainty in the area average drought intensity in all the four datasets.

We find a large uncertainty in the area under drought based on 1 month SSI obtained using the VIC model simulations using the four datasets (figure S4). The uncertainty in drought characteristics and spells can be attributed to the difference in the precipitation and temperature in the four different global datasets (figure S6). For instance, all the four datasets showed a considerable variation in mean annual precipitation and temperature, which translates into soil moisture simulations from the VIC model (figure S6). Despite the large variation among the four datasets, we find a total of 11 drought spells (1952–1955, 1957–1958, 1960–1961, 1965–1967, 1968–1970, 1972–1973, 1974–1975, 1984–1988, 1991–1994, 1999–2006, and 2008–2010), which were consistent in all the four datasets (CRU, Princeton, UDEL, and WFDEI) (table 1, figure S3). These 11 drought spells based on monthly SSI also exhibit uncertainty in the associated drought characteristics (figures S3 and S4, table 1). Our results show that more attention is required in the identification of droughts. For instance, there is considerable uncertainty in soil moisture droughts and associated characteristics based on the four observational datasets in South Asia. Nonetheless, the 2002 drought was the worst drought that affected more than 60% of South Asia and had severe implications for food production and water availability in the region.

Since the 2002 drought was the worst drought with considerable implications in South Asia, it is imperative to examine the risk of such drought under the warming climate in the future. We conducted simulations using the VIC model and meteorological forcing from 40 members of CESM-LENS data for the historical (1951–2016) and future (2035–2100) period. The CESM-LENS simulations have an advantage as these are available for a large-ensemble and simulate the summer monsoon variability reasonably well (Mishra et al 2020). In addition, simulations based on the CESM-LENS forcing can be useful to estimate the occurrence of rare events such as the worst drought of 2002 during the observed climate. A large number of CESM-LENS data (2640 years for each period) provides an opportunity to perform a robust analysis on the occurrence of worst droughts in the historical and future periods over South Asia. We considered the overall severity score from all four datasets (as a threshold; TH_CRU, TH_WFDEI, TH_UDEL, and TH_Princeton) to estimate the occurrence of the 2002 drought under the warming climate. We estimated the major drought spells in each member of CESM-LENS data using 1 month SSI based on the VIC model-simulated soil moisture.

We identified all the drought spells in the CESM-LENS that had the overall severity score higher than the thresholds for the 2002 drought in the observed climate. For instance, we identified all the drought spells in CESM-LENS that exceeded the overall severity score of 1019 to understand the risk of the occurrence of the worst drought based on the threshold of the CRU data (table S1). Since the risk of occurrence of worst drought in CESM-LENS varies with the threshold of an overall severity score based on the four observational products, we considered the uncertainty of observations to identify the worst droughts in CESM-LENS (TH_CRU, TH_WFDEI, TH_UDEL, and TH_Princeton; table S1). We estimated the frequency of the worst drought in the historical (1951–2016) and future (2035–2100) period in 40-members of CESM-LENS using overall severity score from CRU (TH_CRU), WFDEI (TH_WFDEI), UDEL (TH_UDEL), and Princeton (TH_Princeton) (table S1). We estimated the occurrence of worst droughts per 100 years in the historical and future periods using the frequency droughts spells that exceeded the thresholds of the worst observed drought based on the four datasets (CRU, UDEL, WFDEI, and Princeton) (figure 2). The occurrence of the worst drought per 100 years was estimated using the number of worst droughts in the CESM-LENS multiplied by 100 and divided by the total number of years (40 × 66 = 2640 years) in the historical and future climate (table S1).

Zoom In Zoom Out Reset image size

The frequency of occurrence of the worst droughts per 100 years varied from 1.1 to 2.6 during the historical period of CESM-LENS (1951–2016). The occurrence of the worst drought in CESM-LENS exceeding TH_CRU had the lowest (1.1 events in 100 years on average) frequency of occurrence in 100 years due to the highest overall severity score of the 2002 drought. We find a lesser number of droughts exceeding TH_CRU in comparison to the WFDEI, UDEL, and Princeton datasets. Since the overall severity score of Princeton data was lowest (TH_Princeton; table S1), the highest frequency of occurrence (2.6) per 100 years was identified in the historical period. Therefore, the risk of the worst drought in the historical period of CESM-LENS is more than two-fold in the Princeton data in comparison to the risk based on the CRU dataset. Our results show the need of the identification of the occurrence of the worst drought based on the different observational datasets.

The uncertainty in the occurrence of the worst drought can have implications for the planning of water resources structures. Moreover, as the frequency of the worst drought is higher based on the Princeton data (table S1), lower risk in the historical and future periods were found in comparison to the other observational datasets. Similar to the historical period (1951–2016), we estimated the frequency of occurrence of the worst drought (like 2002) in the future (2035–2100) period in CESM-LENS simulations (table S1, figure 2). We find that the frequency of the worst drought is projected to increase in the period 2035–2100 in CESM-LENS based on the thresholds of all the four datasets. Therefore, droughts like 2002 may occur more frequently in the projected future climate in South Asia. We find a similar increase in the occurrence of droughts per 100 years in the future climate for all the four thresholds based on the four observational products. The frequency of the worst drought during the future climate varies between 1.7 and 4 droughts in 100 years based on the four observational datasets. Therefore, the occurrence of the worst drought per 100 years is projected to increase by 1.5-times (1.17–1.7 in TH_CRU; 1.29–1.93 in TH_WFDEI; 2–3 in TH_UDEL; and 2.6–4 in TH_Princeton) in the future (2035–2100) in comparison to the historical period. Other than the identification of the worst drought based on the observed overall severity score, we estimated the frequency of the longest and the most widespread drought (considering the 1999–2006 drought spell as a benchmark) in the historical (1951–2016) and future (2035–2100) periods. We find that the frequency of widespread droughts is projected to increase in the future (table S2), while the frequency of prolonged droughts is projected to decrease in the future climate (table S3).

Next, we estimated the ensemble mean of SSI based on the VIC model simulations using CESM-LENS forcing for all the droughts that exceeded the overall severity score of the worst drought based on the four observational datasets (table S1) for the historical (1951–2016) and the future (2035–2100) periods (figure 2). To construct the ensemble mean of all the worst drought identified during the historical and future periods, we considered peak intensity of drought spells (figures 2 and S6). We find that the ensemble mean of SSI for the worst droughts occurred in the historical climate of CESM-LENS showed exceptional drought (SSI < −2) condition over the central and western parts of South Asia in the ensembles based on all four thresholds (TH_CRU, TH_WFDEI, TH_UDEL, and TH_Princeton) (figures 2(c)–(f)). Moreover, the majority of South Asia was affected by the droughts that exceeded the overall severity of the worst drought (figures 2(c)–(f)). As expected, the ensemble mean of the worst droughts based on TH_CRU showed higher drought intensity in comparison to the other three (TH_WFDEI, TH_UDEL, and TH_Princeton) as TH_CRU was the highest.

Similar to the historical period of CESM-LENS, we constructed an ensemble mean of all the droughts that exceeded the overall severity score of the four observational datasets in the future climate (figures 2(g)–(j)). We find that the ensemble mean of droughts that exceeded the threshold of the worst drought (like 2002) show an exceptional drought condition over the central and western parts of South Asia (figures 2(g)–(j)). The area under the exceptional drought condition (SSI < −2) is higher in the ensemble mean of the worst droughts in the future in comparison to the historical period (figure 2). Moreover, our results show that future drought spells are projected to be more severe, intense, and widespread in South Asia (figures 3(a) and S7, table S2). To examine the temperature anomalies associated with the worst droughts in South Asia, we constructed the ensemble mean of temperature anomalies during the peak drought intensity in the historical and future periods (figure S8). We find that the temperature anomalies exceed 2 °C–3 °C during the worst droughts that were identified in the historical climate based on the CESM-LENS simulations. Moreover, air temperature anomalies associated with the worst droughts in the future can reach to 6 °C–7 °C (figure S8). This substantial rise in temperature during the worst droughts can be detrimental to the agricultural production and water resources in the region. In addition, the concurrent soil moisture drought and high temperature can be lead to heat waves and heat stress in the region (Mishra et al 2017, Nanditha et al 2020). Our results support the previous findings showing the increase in the monsoon and annual drought frequency in the future climate over South Asia (Aadhar and Mishra 2018, 2020, Mishra et al 2020). Overall, we find that the frequency of the worst droughts (like 2002) is projected to rise in the future climate. In addition, the worst droughts in the future are projected to be more severe, intense, and widespread in South Asia.

Zoom In Zoom Out Reset image size

Next, we estimated the linkage between the sea surface temperature (SST) and the worst droughts (overall severity score more than TH_Princeton; in figure 4(a)) that occurred in the historical and future climate in South Asia using CESM-LENS simulations. We selected the overall severity score threshold of the Princeton data as it was the lowest among all the four datasets. Therefore, the droughts that exceed TH_Princeton will also exceed the thresholds of the other three datasets (CRU, UDEL, and WFDEI). All the droughts that exceeded the threshold of the overall severity score (TH_Princeton) were identified during the historical and future periods in all the 40 ensemble members of CESM-LENS (figure 3(a)). It is evident that there are several droughts that exceeded the overall severity score (1019) of the worst drought identified using the soil moisture simulations based on the CRU data.

Zoom In Zoom Out Reset image size

Since SST from CEMS-LENS shows a strong trend in the warming climate, we used the ensemble empirical mode decomposition method (Wu and Huang 2009) to remove the time-varying secular trend (Wu et al 2011) from the SST data. After removing the trend from SST, we constructed the monsoon (June to September) season SST anomaly composite for the all droughts that exceeded the threshold (TH_Princeton) during the historical (1951–2016) and future (2035–2100) periods. We find that the worst droughts occurred in South Asia are associated with the warm SST anomalies over the central Pacific Ocean (5° N–5° S, 170° W–120° W; Nino 3.4 region) in the historical and future periods (figures 3(b) and (c)), which is consistent with the observations (Mishra et al 2012, 2020, Roxy et al 2015). Moreover, we estimated the contribution of Nino 3.4 in the occurrence of the worst droughts during the historical and future periods (table S4). We find that more than 90% of the worst droughts are associated with the warm SST over Nino 3.4 region in the historical period (1951–2016) based on the CESM-LENS simulations. In the future period (2035–2100), about 80% of the worst droughts associated with warm SST anomalies showing that SST anomalies over Nino 3.4 during the monsoon season will continue to play an essential role in the future droughts. Overall, our results show that warm SST anomaly over Nino 3.4 is the primary driver of the worst droughts over South Asia.

Our results show that the worst droughts (like the year 2002) are projected to become more frequent (1.5 times) in the future climate with an increased drought intensity and areal coverage. However, it remains unclear that to what extent do the anthropogenic activities (greenhouse-gas emission and industrial aerosol) affect the occurrence of these major soil moisture droughts in South Asia. To examine the role of anthropogenic activities (greenhouse gas emission and industrial aerosol) in the occurrence of the droughts that exceeded the overall severity score of the worst drought of the observed climate, we used 'All-but-one-forcing-fixed' CESM-LENS experiments. The 'ALL-but-one-forcing-fixed' experiments used the same simulation protocol as CESM-LENS, except that one forcing agent (XGHG or XAER) is held fixed at 1920 conditions (Deser et al 2020b). We estimated the drought spells in the fixed industrial aerosol (XAER) and XGHG conditions and compared them with the ALL (natural + anthropogenic) scenario of the CESM-LENS simulations. We conducted soil moisture simulations using the VIC model for XGHG and XAER scenarios for the historical and future periods for the 20 ensemble members of CESM-LENS. Therefore, the role of GHG/AER emissions on droughts can be identified by comparing the ALL and XGHG/XAER simulations.

The frequency of droughts that exceeded the overall severity score of the worst drought (like 2002) was estimated in the ALL and XAER scenarios using 20-members of CESM-LENS for the historical (1951–1996) and future (2035–2080) (figure 4, table S5). We find that the frequency (per 100 years) of worst droughts in the ALL and XAER are similar during the historical climate for all four thresholds (TH_CRU, TH_WFDEI, TH_UDEL, and TH_Princeton), indicating that there is no effect of aerosols on the occurrence of worst drought during the historical period of CESM-LENS. However, the frequency of the worst droughts in the future (2035–2080) is higher in the XAER in comparison to the ALL (natural + anthropogenic) scenario, indicating that the region could have faced more droughts in the future (2035–2080) in the absence of industrial aerosols. We find that the ensemble mean of SSI for the droughts that exceeded the overall severity score of the worst drought (2002) showed an exceptional drought condition (SSI < −2) and a large areal coverage during the historical and future climate (figure 4). However, the ensemble mean SSI of worst droughts occurred in future is less severe than the ensemble mean SSI of worst droughts occurred in the historical period over South India (figure 4). The decrease in the ensemble mean severity of worst droughts in South India in the future under the XAER scenario can be attributed to an increase in precipitation in the region (figure S9). Temperature anomalies of 2 °C–3 °C and 4 °C–6 °C were associated with the worst droughts in the historical and future periods, respectively (figure S10). Overall, we find that without industrial aerosol emissions, the frequency of the worst droughts in the future can be higher in comparison to the ALL scenario in South Asia. The XAER induced high temperature and results increase in heat extremes (Horton et al 2016, Xu et al 2018). Therefore, the increased atmospheric aridity and frequent heat extremes in XAER scenario result higher frequency of worst droughts in the XAER scenario.

Finally, to estimate the role of greenhouse gas (GHG) emissions on the occurrence of the worst droughts in South Asia, we estimated the frequency of droughts that exceeded the overall severity score of the four datasets (TH_CRU, TH_WFDEI, TH_UDEL, and TH_Princeton) (figure 5, table S5). For the XGHG condition, we find that the smaller number of worst droughts occurred during the historical climate in comparison to the ALL scenario. Moreover, the frequency of the worst droughts in the XGHG and ALL are similar in the future (2035–2080) climate. Since GHG emission is the primary source of atmospheric warming, we find a relatively lesser increase in the temperature over South Asia in the future climate in XGHG condition (figures S9 and S11). The lesser projected rise in temperature under the future climate in XGHG does not accelerate the soil-moisture depletion during the post-monsoon season despite a decline in the summer monsoon precipitation. Moreover, the soil-moisture during the post-monsoon season recovers after moderate rainfall and avoids the long-lasting worst droughts in the future climate under XGHG scenario. Therefore, the lesser rise in temperature does not cause an increase in the frequency of the worst droughts despite a projected decrease in summer monsoon precipitation in the future (figure S9). In addition, we find that the ensemble mean of SSI of the worst droughts is higher in the future compared to the historical period (figure 7). The higher ensemble mean SSI in the future may be associated with the consecutive failure of monsoon seasons or large deficit in precipitation over South Asia (figure S9). Overall, our results show that the frequency of worst droughts remains the same in the future climate in without greenhouse gas emission scenario compared to ALL scenario.

Zoom In Zoom Out Reset image size