Increasing risk of simultaneous occurrence of flash drought in major global croplands

Simultaneous occurrence of flash drought in major croplands can pose challenges for global food security. However, drivers of flash drought co-occurrence in the observed and future climate remain unrecognized. Using observations and climate model simulations, we examine the simultaneous occurrence of flash droughts in 16 major global croplands that grow wheat, rice, and maize. We show that significant warming combined with decreasing precipitation led to an increased frequency of flash droughts in several major croplands during the observed climate (1981–2020). The simultaneous occurrence of flash drought in many croplands in the same year has significantly (p-value = 0.007) increased during 1981–2020 and is likely to continue. Nine out of 16 major global croplands are projected to witness an increased frequency of flash droughts under the warming climate. The observed and projected rise in flash droughts in global croplands is mainly driven by vapor pressure deficit. The positive phase of El Nino Southern Oscillation influences flash drought co-occurrence in 10 out of 16 major cropland regions and remains a dominating factor of flash droughts co-occurrence in the future. Enhanced climate warming and increased frequency of El Nino events can further enhance the occurrence of simultaneous flash droughts in several major croplands, with substantial implications for food production.


Introduction
Global croplands (1.6 billion ha) are the backbone of food security. Climate change and human interventions considerably influenced croplands and food production in the last few decades (Mbow et al 2019, FAO 2021, which are projected to increase (Fitton et al 2019). For instance, in recent decades, adverse climatic conditions have become more frequent in global croplands (wheat, maize, rice, and soybeanproducing regions) in the recent decades (Gaupp et al 2019). Similarly, with increasing drought frequency, approximately 11% of the existing croplands are vulnerable to reduced water availability, which can cause a decline in agricultural production (Fitton et al 2019). Due to management and agricultural mechanization, the cropland area has increased by 5% (0.07 billion ha) in the previous four decades (Food and Agriculture Organization 2021). However, the lack of climate adaptation has exacerbated the risk of crop failures across multiple global croplands (Singh et al 2021(Singh et al , 2022. Moreover, increased compound extremes may threaten major cereal crops such as maize, wheat, and rice (Matiu et al 2017).
Major global croplands consist of wheat, rice, and maize growing regions spread over a large area. Croplands are vital in meeting agricultural food demands (Leff et al 2004). Wheat, rice, and maize contribute around 89% of cereal worldwide. Maize contributes about 43% of total production, followed by wheat (27.7%) and rice (18.5%) (Food and Agriculture Organization 2021). Moreover, these three major crops (wheat, rice, and maize) support almost two-thirds of human calorific intake (Zhao et al 2017). Agricultural intensification has met rising food demands (Godfray et al 2010, Tilman et al 2011 by enhancing crop yields and food production in the major global croplands. Despite increasing crop yields (Ray et al 2012(Ray et al , 2013, food demand is also expected to rise considerably due to the increasing population (Tilman et al 2011, Kastner et al 2012. However, climate extremes remain a major threat to sustaining future food demand (Vogel et al 2019). In addition, climate warming and increasing population can further pose challenges to crop production and sustain the rising food demands (Mauser et al 2015).
Climate warming has caused substantial changes in the hydrological cycle (Wu et al 2013), as evidenced by the rising precipitation extremes (Mukherjee et al 2018). Extreme dry and wet events have become more frequent and intense recently (Donat et al 2016, Kumar andMishra 2020). In addition, climate warming has increased the likelihood of excessive heat exposure to major croplands (Zhu and Troy 2018). An increase in compound hot and dry extremes has been observed in croplands in the United States . Compound hot and dry extremes can cause rapid drying of soil moisture (SM) due to strong landatmospheric coupling (Miralles et al 2012, Rajeev et al 2022, Wang and Yuan 2022, which can result in flash droughts (Qing et al 2022).
Flash droughts are caused by the rapid depletion of SM (Otkin et al 2018), which directly impacts agricultural production , Mahto and Mishra 2020, Yao et al 2022, ecosystem health (Otkin et al 2016, Nguyen et  Widespread and spatially compounding droughts increased nearly three times over multiple regions, posing pressures on the global food system and economy from 1981 to 2018 (Singh et al 2021). Simultaneous occurrence of drought in several major croplands can reduce food production and pose challenges to global food security (Sarhadi et al 2018, Mehrabi andRamankutty 2019). Food shortage often causes inflated food prices, which directly impacts lower-income population (Mbow et al 2019). Moreover, food shortages are not just limited to food prices but can also lead to famine and malnutrition. For instance, south Asian countries witnessed deadly famine during the late 18th and 19th centuries (Mishra et al 2019). Similarly, a severe drought led to disease and famine in Africa's agricultural regions (Miller 2009). Enhanced warming over croplands is likely to increase the co-occurrence of warm and dry extremes (Sarhadi et al 2018) that can potentially cause simultaneous failure of crop production (Gaupp et al 2019). For instance, the risk of simultaneous drought conditions across the global wheat and maize producing regions has increased over the last four decades and is projected to rise in the future (Tigchelaar et al 2018, Gaupp et al 2019.
Simultaneous occurrence of widespread drought at multiple locations is often linked to regional and global scale circulations (Singh et al 2021, Mondal et al 2023. While the co-occurrence of traditional droughts has been widely studied (Singh et al 2022, Mukherjee et al 2023, the occurrence of flash droughts in major global croplands in the observed and projected climate remains unrecognized. In addition, it remains to be seen if climate warming poses a greater risk of flash droughts to global croplands or if flash droughts become less likely due to increased precipitation. Using observations, reanalysis datasets, and climate projections, we examine the global croplands' risk, drivers, and co-occurrence of flash droughts in the observed and future warming climate. We show that major global croplands are projected to witness an increased frequency of flash droughts under the warming climate. The projected increase in the frequency of flash droughts in the major global croplands is largely driven by elevated warming and large-scale climate variability, which enhances the risk of flash drought co-occurrence.

Datasets
Cropland is defined as an area used for growing annual and perennial crops for human consumption, fodder, and biofuel (Potapov et al 2022). We obtained an updated (Phalke et al 2020, Thenkabail et al 2021 major global cropland (maize, rice and wheat) information from Leff et al (2004). Major global croplands consist of a wide area and high-density wheat, rice, and maize growing regions (Leff et al 2004). An archive of the global cropland dataset is available at the United States Geological Survey (USGS) Land Processes Distributed Active Archive Centre (LP DAAC) server (https://lpdaac.usgs.gov/news/releaseof-gfsad-30-meter-cropland-extent-products/). The USGS prepared the highest spatial resolution (30 m) global agricultural dataset using long-term (1972-2020) Landsat satellite imagery, which was a part of the Global Food Security-Support Analysis Data @ 30-m (GFSAD30) Project (Thenkabail et al 2021). The dataset has an overall accuracy of 92%. The cropland dataset has been widely used for global drought and agriculture applications (See et al 2015, Ghazaryan et al 2020, Lesk and Anderson 2021, Potapov et al 2022. We obtained daily air temperature (T), dewpoint temperature (Tdew), sea surface temperature (SST), and SM at 0.25 • spatial resolution from the European Centre for Medium-Range Weather Forecasts (ECMWF) atmospheric reanalysis version 5 (ERA5 reanalysis) for the observed period   (Hersbach et al 2020). ERA5 reanalysis provides a reliable dataset for croplands regions across the globe (Mahto and Mishra 2019, Rolle et al 2021, Du et al 2022. In addition, ERA5 is an updated reanalysis widely used for flash drought identification ( (see table S1 for GCMs related details). We used four emission scenarios [SSP1-2.6 (lowest emission), SSP2-4.5 (moderate emission), SSP3-7.0 (high emission), and SSP5-8.5 (highest emission)] to incorporate more scenarios and corresponding projected changes in the future climate. Temperature, SST, relative humidity (RH), surface SM, and root zone SM were obtained from the nine CMIP6-GCMs. We regridded climate model output from their native resolution to 1 • (∼110 km) to make them consistent across all the models.

Estimation of vapor pressure deficit (VPD)
We obtained surface air temperature (T, in • C) and dew-point temperature (Tdew, in • C) from ERA5, which were used to estimate the observed climate's VPD . We used temperature and RH from CMIP6-GCMs to estimate VPD for the historical (1950-2014) and projected future climate (2015-2100). VPD (hPa) was estimated by taking the difference between saturated and actual vapor pressure. Saturation vapor pressure (es, in hPa) was calculated using surface air temperature (Alduchov and Eskridge 1996) for the observed, historical, and projected future climate. Dew-point temperature and RH (Zotarelli et al 2010) were used to calculate the actual vapor pressure (ea, in hPa). VPD was estimated as: Saturation vapour pressure (es) = 6.11 * e ( 17.27 * T 237.3+T ) Actual vapour pressure (ea) = 6.11 * e ( 17.27 * Tdew 237.3+Tdew ) Relative humidity (%) = ea es * 100 (3) Vapor pressure deficit (VPD) = es − ea. (4) We used different variables to estimate VPD in ERA5 (using T and Tdew) and CMIP6-GCMs (using T and RH). Therefore, we compared VPD from ERA5 and CMIP6 during the common period 1981-2014 to ensure a reliable agreement (r > 0.85) between figure S1.

Flash drought identification
We used root zone (60 cm) SM to identify flash droughts in observed (ERA5), historical (CMIP6historical), and projected future (CMIP6-ssp126, CMIP6-ssp245, CMIP6-ssp370, and CMIP6-ssp585) periods. The ERA5 SM is reasonably correlated (r > 0.7) against satellite-based SMAP-enhanced SM at a global scale (Lal et al 2022) and has been widely used to analyze flash droughts at regional and global scale (Christian et al 2021, Parker et al 2021, Shah et al 2022. We compared SM from ERA5 reanalysis against the GLEAM dataset and found a good agreement (r > 0.8) between ERA5 and GLEAM SM in most regions during 1981-2019 (figure S2). Similarly, we compared precipitation from ERA5 with GPCP, which also compares very well (correlation coefficient, r > 0.95) in all the cropland regions during 1981-2019 (figure S3).
We constructed five day average (or accumulated, defined as pentads) time series of precipitation, air temperature, VPD, and SM to analyze the flash droughts. We used pentads to avoid the high-frequency variability that may present daily (Christian et al 2019, Mahto and Mishra 2020). Effective root depth for most crops ranges between 40 and 100 cm (Lilley and Kirkegaard 2016). Therefore, we used root-zone SM (up to 60 cm depth) to identify flash drought (Mahto and Mishra 2020) across the major global cropland regions. We used the flash drought identification methodology Mahto and Mishra (2020) developed to identify flash drought events during observed and future warming climate. The method incorporates the necessary criteria (rapid development, significant impact, and termination) for flash drought identification (Pendergrass et al 2020). The onset of flash drought occurs when SM declines from above the 40th percentile to less than the 20th percentile in a maximum of three pentads (criteria for rapid development). After the flash drought onset, SM must persist below the 20th percentile for at least the next two consecutive pentads (criteria for significant stress (Pendergrass et al 2020)). Flash drought terminates when SM bounces back to the 30th percentile or more, which signifies a normal condition (Svoboda et al 2002, Farahmand andAghaKouchak 2015). A minimum-duration flash drought event should have three pentads below the 20th percentile between the onset and termination pentads (Yuan et al 2019).
We defined the frequency of flash drought as a ratio of flash drought pentads and the total number of pentads in a given period. The frequency of flash drought is given as follows: Similarly, the intensity of a flash drought is defined as the average SM percentile (up to 60 cm) during a flash drought, except for the pentads of onset and termination. Percentile and anomaly were estimated relative to the climatological mean period of 1981-2010 for observed (ERA5), historical (CMIP6), and projected future climate (CMIP6). We estimated the standardized anomaly by subtracting the corresponding long-term mean and dividing it by its standard deviation (σ) for each pentad.
We defined the onset-development phase as the duration between above the 40th percentile and the pentad below the 20th percentile with a nonincreasing slope between two consecutive pentads during a flash drought event. Pentads during the flash drought onset-development phase were considered for estimating the flash drought development rate. Changes in precipitation and VPD were assessed corresponding to SM desiccation during the onset-development phase to examine the dominant driver of flash drought occurrence. A schematic illustration of the definition of a flash drought event (onset, propagation, and termination) is provided in figure S4. Moreover, the area affected by flash drought was estimated using the number of grids below the 20th percentile SM during a flash drought event. Moreover, we considered the simultaneous occurrence of flash droughts at an annual scale. Here we expect that the onset of flash drought among different cropland regions may vary but should occur within the same year. The maximum crop damage is expected when flash drought occurs during the peak crop growing period. However, flash droughts can affect multiple crops during a calendar year based on their occurrence.

Estimation of global warming levels
We estimated the global mean annual temperature from the nine CMIP6-GCMs during 2020-2100. We considered 1861-1890 as the reference period to calculate the global warming levels (i.e. temperature changes in • C concerning the 1861-1890 period) to estimate the changes in Earth's average air temperature since the pre-industrial era under the different emission scenarios. We created discrete bins on an interval of ten years (2020-2029, 2030-2039…2090-2099) for all four emission scenarios to incorporate the interannual temperature variability and to increase the sample size of warming periods. We estimated the decades with a range (0.3-0.7; 0.8-1.2; 1.3-1.7, 1.8-2.2, 2.3-2.7 … and so on) to estimate the global warming levels (0.5, 1, 1.5, 2.0, and 2.5 … so on). For instance, if a decade's mean annual temperature change varies between 1.3 • C-1.7 • C, we consider that decade of 1.5 • global warming level.
We used SST anomaly of the Niño 3.4 region (120 W-170 W, 5 N-5 S) to estimate the Oceanic Niño Index (ONI) for observed (ERA5) and the future warming climate (CMIP6-GCMs). ONI (usually ranges between −3 • C-3 • C) was used to identify El Niño (ONI more than 0.5 • C) and La Nina (ONI less than −0.5 • C) phases of El Niño-Southern Oscillation (ENSO) in the eastern equatorial Pacific Ocean (Trenberth 1997, Trenberth andStepaniak 2001). We used a three month average SST to estimate ONI for the observed, historical, and future periods as recommended by the National Oceanic and Atmospheric Administration (NOAA) (https://origin.cpc.ncep. noaa.gov/products/analysis_monitoring/ensostuff/ ONI_v5.php). For a year to be classified as El Niño or La Niña, the ONI must exceed 0.5 • C or −0.5 • C for at least five consecutive months. We estimated the number of major global croplands that experienced flash drought co-occurrence in any given year. Then we identified years for which the number of major global croplands experiencing flash drought co-occurrence exceeded one standard deviation from the mean (number). For these selected years, we constructed SST anomalies composites to examine the role of ENSO variability on flash drought cooccurrence in the major global croplands.

Analysis approach
Considering the highest emission scenario (SSP5-8.5), we show the effect of maximum warming on global flash drought co-occurrence. We performed trend analysis using the nonparametric Mann-Kendall trend test (Mann 1945) with Sens's slope estimator (Sen 1968) to examine the changes in flash drought frequency, precipitation, temperature and SM in the observed and future climate. We used Pearson's Correlation Coefficient (r) to estimate the linear relationship between two variables. For instance, the linear relationship between flash drought frequency change and climatic variables (precipitation and temperature) was established. Similarly, the relationship between increasing global warming levels and changes in flash drought frequency (and SM) was also evaluated using a correlation coefficient. In addition, we performed dominance analysis (Budescu 1993, R and DV 2003) to evaluate the relative contribution (%) of precipitation and VPD in flash drought occurrence (i.e. development) during observed and future warming climate. Relative contribution uses a linear regression approach, where a fraction of total variability (R 2 ) in flash drought occurrence (represented by SM anomalies in the drought development phase) can be explained by precipitation or VPD. The conditional dominance of a variable for each predictor is evaluated, and the predictor with the highest average conditional dominance was identified as the most significant contributor (Budescu 1993, Nimon and Oswald 2013).

Observed changes in precipitation and temperature in the global croplands
Nearly one-third (16.7 million km 2 ) of agricultural land is currently used as cropland, which meets global food demands (Food and Agriculture Organization 2021). We selected the major global croplands comprising 16 regions dominated by wheat, rice, and maize crops. These three staple crops grow in different seasons in the major global croplands ( figure 1(a)). Wheat is mainly grown in the cropland regions of the United States, Europe, Russia, India, China, and Australia. Rice is produced in monsoon-dominating areas, including India and China. At the same time, maize is cultivated in almost all the major cropland regions ( figure 1(a)). Rice and maize have a relatively higher crop water requirement (500-800 mm) as compared to wheat (450-650 mm) during the crop growing period (Brouwer and Heibloem 1986). Therefore, spatial variability in crop water availability is a primary driver of the heterogeneity in crop distribution across the major global croplands. Climate warming has altered precipitation patterns in the last few decades (Shiu et  The major global cropland experienced significant warming (∼1 • C-2 • C) in the observed period of 1981-2020 (figures 1(b) and S5). The most notable (p-value < 0.05) warming occurred in the croplands of Europe, where the temperature increased by 2.5 • C from the 1981-2020 period. The increased warming in the major croplands in Europe caused severe heat waves that have become more frequent, widespread, and devastating in recent decades (Schär et al 2004, Rousi et al 2022. In contrast, croplands in Canada and south Asia experienced relatively lesser warming (∼0.5 • C) between 1981 and 2020. Observed changes in precipitation reveal a higher spatial variability compared to temperature across the major global croplands (figure 1(c)). For instance, precipitation changes vary between 15% (south Asia) and −28% (east Australia). In contrast, almost all the cropland regions experienced a positive temperature change (between 0.5 • C and 2 • C) during the observed climate ( figure S5). Regionally, croplands in Northern North America, South Asia, and tropical island groupings witnessed an increase in precipitation during the observed climate, while precipitation slightly increased in the Western Europe croplands ( figure 1(c)). On the other hand, mean annual precipitation decreased by more than 15% (p-value < 0.01) in South America, Eastern Europe, East Africa, China, and Australia ( figure 1(c)). Overall, mean annual precipitation has decreased significantly (p-value < 0.05) in 9 out of 16 cropland regions ( figure S5(a)). In contrast, precipitation has increased significantly in only two cropland regions during the observed climate. The other five major croplands witnessed moderate changes (but insignificant) in precipitation from 1981-2020 ( figure S5(a)). Nevertheless, globally, precipitation has decreased by nearly 8% (p-value < 0.01), whereas temperature has increased by 1.2 • C (p-value < 0.01) in the major croplands during the observed climate ( figure  S5). Therefore, robust and significant warming combined with the decline in precipitation can adversely affect crop yields and food production (Otkin et al 2018) with an increased likelihood of flash droughts  in the major croplands during the observed climate.

Observed flash drought occurrence in major global croplands
The frequency of flash drought (percentage of time-step) varies from 10%-20% among the major global croplands during 1981-2020 (figure 2). Flash droughts occur more often in croplands in eastern North America, South America, Western Europe, South East Asia, and the tropical group of islands (figures 2(a) and (b)). On a global scale, flash droughts have increased significantly in most cropland regions. For instance, croplands in South America, Europe, South East Asia, east Africa, and Australia experienced an increased (more than 10%) frequency of flash droughts during the observed climate (figures 2(c) and (d)). On the other hand, croplands in Canada (∼5%), South Asia (∼8%), and the group of tropical islands (∼4%) witnessed a decline in the flash drought frequency during the observed period (figures 2(c) and (d)). The increase in flash drought occurrence in the major croplands can be attributed to the decline in mean annual precipitation. However, there may be other drivers of the rise in flash drought frequency. For instance, despite a moderate increase in precipitation in Western Europe, flash drought frequency has increased (figures 2(c) and (d)), which can be attributed to significant observed warming in Western Europe. Globally, the observed decline in precipitation and warming contributed to the increased frequency of The increased frequency and area affected (figures 2 and S7) can enhance the likelihood of simultaneous flash droughts in several major global croplands. The number of major croplands that experienced flash drought in any given year has significantly increased (p-value = 0.001) at a rate of one region per decade during the observed climate ( figure S8). Ten or more major croplands simultaneously experienced flash droughts in 14 years during the last two decades (figure S8). Moreover, 11 of 16 major croplands experienced flash droughts in 2012 (figure S9). Croplands in North America, South America, Europe, and northwest Asia witnessed a substantial yield loss due to multiple flash droughts (figure S9). For instance, the United States seen about a ∼30% reduction in corn yield (Rippey 2015). Nearly 80% of the agricultural land in the central United States was affected during the 2012 flash drought (USDA 2015), which resulted in $34.5 billion in economic losses (NOAA 2012). Similarly, croplands in Europe experienced a yield loss of 25%-37.5% (1-1.5 tons ha −1 ) for wheat and 40% (3 tons ha −1 ) for maize (Crocetti et al 2020). Maize yield was significantly reduced in Kazakhstan (55%) and Argentina (40%) during the 2012 flash drought (Rojas et al 2019). Agricultural loss due to the 2012 flash droughts was relatively lesser (∼$3 billion) in Asia as flash droughts did not affect all the crop-growing seasons (FAO 2021). Similar to 2012, extreme heat exposure compounded with flash droughts affected the major croplands that grow wheat, maize, and rice (Zhu and Troy 2018).

Occurrence and drivers of flash drought in the projected future climate
The frequency of flash droughts is projected to rise in 9 out of the 16 major global croplands under the warming climate ( figure 3). The most notable increase in flash drought frequency is projected in major croplands in Europe (10%-12%), eastern South America (∼10%), North America, South Africa, and the tropical group of islands (∼8%-10%). In contrast, the frequency of flash droughts is projected to decline in 7 of 16 major croplands in the future climate with increasing global warming levels (see methods for details). Flash droughts are projected to decrease in major croplands in northern North America, southern South America, central Africa, South Asia, and areas of high-latitude Asia in the future ( figure 3(a)), primarily due to the increase in SM under the warming climate. On the other hand, the decline in SM is projected to enhance the flash drought frequency in the rest of the major croplands (9 out of 16) in Europe, central North America, eastern South America, South Africa, eastern Asia, and Australia under increasing global warming levels (figures 3(a) and (b)). Several croplands that experienced increased flash drought frequency in the observed climate (also see figure 2) can also see a rise in the future climate (figure 3(a)) with severe implications due to an increase in the flash drought frequency in the future. Northern North America and South Asia are the only regions where croplands are projected to experience a decline in flash drought occurrence in the future, consistent with the observed changes. We examined flash drought sensitivity to changes in SM under the warming climate (figure S10). Croplands with a projected increase in the flash drought frequency are more sensitive to SM change in response to the warming in the future. For instance, flash droughts in the croplands of eastern North America, East Asia, and a tropical group of islands are projected to rise by 8%-10% in response to a 1% decline in SM (figure S10). On the other hand, flash droughts are projected to decrease by ∼2% in response to a 1% increase in SM in the major croplands in southern South America, central Africa, and South Asia (figure S10).
We estimated the relative contribution of precipitation and VPD in response to declining SM during the flash drought onset-development phase (see methods for details). VPD is the dominant driver of flash droughts in the major global croplands in both the observed and the future warming climate. VPD and precipitation contribute 60% and 40%, respectively, to flash drought occurrence in the major global croplands ( figure 4(a)). Precipitation dominates VPD in driving flash droughts in 7 out of the 16 major croplands (central North America, southern South America, central East Africa, South Asia, and northern Asia) during the observed climate ( figure 4(a)). On the other hand, VPD dominates precipitation in the remaining nine major croplands ( figure 4(a)). There can be uncertainty in the individual climate models, which is accounted by the taking ensemble mean of CMIP6-GCMs. We also find that CMIP6 models reasonably agreed in 13 out of 16 major global croplands against ERA5 in explaining the dominant drivers of flash drought occurrence during the 1981-2014 period (figure S11). VPD is projected to remain the dominant driver over precipitation for flash drought occurrence in all the major croplands except in central West Africa under the future warming climate ( figure 4(b)).

ENSO variability and flash drought occurrence
We examined the role of ENSO variability on flash drought co-occurrence in the global croplands during the observed and future climate. The distribution of ENSO's positive, negative, and neutral phases was examined for the years of SST anomaly composite (see methods for details). We find that more than 60% of croplands (10 out of 16) experienced flash drought during the positive phase of ENSO (El Niño). In contrast, the remaining six croplands experience flash drought mainly during the negative phase of ENSO (La Nina). Croplands in Asia, Europe, and Africa mainly experienced flash droughts during the El Nino phase, while croplands in North and South America experienced flash droughts during the La-Nina phase (figure S12). El Nino contributed about 36% (50%) of the years that experienced above average (mean plus one standard deviation) flash drought co-occurrence in the major croplands during the observed (historical) period of 1981-2014 (figures 5(a) and (b)). The average flash drought co-occurrence is higher (∼9 regions per year) during the El Nino phase compared to the La Nina phase (∼7 regions per year) (figure S13). The flash drought co-occurrences during El Niño are projected to increase further in the near future (2021-2054) ( figure 5(c)). The El Niño events are projected to rise by more than 30% under the warming climate (Singh et al 2022), a major driver of the flash drought co-occurrence in the global croplands. The El Niño contribution to flash drought co-occurrence is projected to increase further under the SSP5-8.5 scenario (figure S14). The increased frequency of El Niño under the warming climate (Cai et al 2014, Ying et al 2022 can enhance the simultaneous occurrence of flash droughts in the global croplands. We find an elevated risk of simultaneous occurrence of flash drought contributed by the warming climate and ENSO variability in the global croplands.

Discussion and conclusions
Flash droughts in major croplands significantly affect agriculture activities and food production (Zhao et al 2017). The majority of the global croplands experienced an increase in flash drought frequency. However, a decreasing trend in the frequency of flash drought in northern North America, South Asia, and tropical islands can be associated with increased precipitation and lesser warming than the other global croplands. Strengthening moisture transport during the summer monsoon season (June-September) is a crucial reason for increased precipitation over the South Asian region (Roxy et al 2017). Similarly, increased annual precipitation (rainfall and snowfall) contributed by the polar jet stream (Masson-Delmotte et al 2021) reduced flash drought frequency in the northern North American croplands.
Increased flash drought frequency across multiple croplands enhanced the agricultural exposure to flash drought co-occurrence. For instance, nearly 70% (11 of 16) of croplands experienced flash drought in 2012, leading to a global rise in food prices (Food and Agriculture Organization 2021). Similarly, a considerable loss in crop yields was reported due to recordbreaking hot and dry conditions in 2022 in South Asia (USDA 2022). Over 70% of the global croplands are projected to experience increased flash drought under the warming climate. As a result, the risk of simultaneous flash drought in multiple croplands is also projected to rise in the future warming climate. Flash drought severity can affect exposed croplands differently. For instance, less severe flash droughts can be terminated by an intense precipitation spell or can be managed by irrigation. On the other hand, extreme evaporative demand can restrict flash drought termination, leading to crop failure (Mahto and Mishra 2019).
The global large-scale precipitation systems are highly influenced by the ENSO phenomenon (Cai et al 2021, Sorí et al 2021, which governs the longterm changes in precipitation and VPD. ENSO is a prominent factor that controls the drivers of flash drought occurrence (i.e. precipitation and VPD) in the observed and future climates. In particular, flash droughts during El Nino affect nearly 65% (10 of 16 regions) of the global croplands. The increasing frequency of El Nino directly impacts the growing risk of flash drought co-occurrence in the major croplands.
Simultaneous occurrence of flash drought in the major agricultural areas poses a risk of food insecurity risks and increased food prices. Notwithstanding technological advancements in agriculture (Peng et al 2022), the food crisis often threatened the global population driven by flash droughts and heatwaves (Alizadeh et al 2020). Compound heatwaves and flash droughts cause higher crop damage and pose unbearable losses. Flash droughts are projected to develop at a faster rate. They can be induced by the compounding effect of SM depletion and increased VPD (Qing et al 2022), further increasing the challenges in flash drought preparedness and mitigation. Therefore, the simultaneous occurrence of flash droughts during El Niño under the warming climate can harm agriculture and food security.
We used ERA5 reanalysis data which is among the better-performing reanalysis products (Mahto and Mishra 2020) to analyze flash drought occurrence in the major croplands. SM from ERA5 reanalysis captures the spatial and temporal variability of insitu and satellite-based datasets Mishra 2020, Lal et al 2022) and has been widely used for drought assessments (Christian et al 2021, Parker et al 2021). However, there can be uncertainty in different reanalysis products in different croplands (Singh et al 2022). Therefore, analyzing multiple reanalysis datasets can help us capture some dataset-related uncertainty. Like reanalysis products, there can be considerable intermodal uncertainty in the CMIP6-GCMs . Notwithstanding the uncertainty in CMIP6-GCMs, the projected changes in drought are robust (Ukkola et al 2020). Our analysis did not consider the role of human interventions (e.g. irrigation) on flash drought occurrence. Irrigation can considerably alter the flash drought occurrence and temperature (Ambika and Mishra 2020). However, most CMIP6-GCMs need to consider the role of irrigation in future projections. Considering both climate change and human interventions can provide more insights into the flash drought occurrence in the observed and projected future climate.

Data availability statement
The dataset used in this study is publically available and can be obtained from the European Centre for Medium-Range Weather Forecasts (ECMWF) 5th generation atmospheric reanalysis (ERA5), which is distributed by the C3S CDS (Hersbach et al 2020). The Coupled Model Intercomparison Project Phase 6 (CMIP6) data are obtained from the World Climate Research Programme (WCRP) server (https://esgfnode.llnl.gov/projects/cmip6/). We have processed and analyzed the data in MATLAB, Version R2020a, under the institute license agreement. The figures are created using the Generic Mapping Tools (GMT), Version 6.1.1, which is distributed under the GNU Lesser General Public License (LGPL), available at www.generic-mapping-tools.org/download/.