Droughts, traditionally viewed as slow-onset events, are increasingly recognized in their rapid-onset form as "flash droughts." These events, characterized by rapid intensification within weeks or even days, pose significant challenges due to their unexpected nature and the limited time available for mitigation. The 2012 US flash drought, causing over $30 billion in losses, and the 2017-2018 Australian flash drought, resulting in widespread tree mortality and livestock losses, exemplify their devastating impact. The increasing atmospheric water demand and strengthened land-atmosphere feedbacks under global warming are expected to exacerbate the frequency and intensity of flash droughts. This study addresses the urgent need for a comprehensive understanding of the characteristics and drivers of flash droughts, particularly concerning their impact on human and natural systems. The research aims to quantify global and regional changes in flash drought characteristics between 1981-2000 and 2001-2020, characterize changes in exposure of agriculture, forests, and populations, understand the role of precipitation deficits and temperature anomalies, and evaluate the ability of CMIP6 models to simulate these phenomena. This knowledge is crucial for developing effective vulnerability reduction strategies and improving mitigation and response capacities.

Previous research has highlighted the increasing frequency of flash droughts and their link to anthropogenic climate change. Several studies have employed various indices, including the Standard Evaporative Stress Ratio, Evaporative Stress Index, and soil moisture anomalies, to monitor and characterize flash droughts. A general consensus exists on the importance of root-zone soil moisture anomalies as a key indicator, reflecting the sensitivity of this variable to changes in precipitation and temperature. Existing studies have focused on regional and global trends, emphasizing the role of human-induced warming in the long-term increase of flash droughts. However, these studies often lack a comprehensive comparison of the changing characteristics and drivers before and after 2000, a period marked by a dramatic increase in atmospheric water demand. This omission creates a significant knowledge gap, hindering a deeper understanding of the underlying physical mechanisms and accurate risk assessment for interconnected systems.

The study employed a multi-faceted approach incorporating diverse data sources and analytical techniques. First, daily soil moisture, temperature, and precipitation data from 36 FLUXNET sites (473 site-years) covering 1996-2014 were used to define and characterize flash droughts at the site scale. A flash drought was defined by a rapid decrease in pentad (5-day) soil moisture from above the 40th percentile to below the 20th percentile, with a minimum duration of three pentads but less than 12 pentads to differentiate it from long-term droughts. Two global observation-based datasets, GLEAM and ERA5, provided root-zone soil moisture data (1981-2020) to analyze global spatial and temporal changes in flash drought characteristics. Daily precipitation and temperature data from ERA5 were used to analyze the contribution of precipitation deficits and temperature anomalies to flash drought occurrence. Twenty-two CMIP6 Earth system models were used to evaluate model performance in simulating flash drought characteristics and drivers, comparing them to GLEAM and ERA5 results. The study further calculated changes in exposure to flash droughts for agricultural areas, forests, and populations between 1981-2000 and 2001-2020. Statistical analyses included the Mann-Kendall test, linear detrending of soil moisture, and calculation of standardized anomalies for precipitation and temperature. The study also investigated the dependence between low precipitation and high temperatures using Pearson's correlations and bivariate copulas, examining the sensitivity of soil moisture to precipitation and temperature using partial correlation. Data sources included FLUXNET2015, GLEAM Soil Moisture V3.5a, ERA5 soil moisture, precipitation and temperature data, CMIP6 simulations from 22 Earth system models, population data from Gao (2020), cropland and pastureland fractions from NASA SEDAC (2010), urban extents data from Zhao et al. (2022), and forest cover data from Hansen et al. (2013).

The study's key findings confirm a significant increase in flash droughts since 2000, driven primarily by increased concurrent hot and dry conditions. Based on FLUXNET data, the frequency of flash droughts increased significantly (p<0.01, Mann-Kendall test) in the last two decades, with most occurring during spring and summer. Analysis of GLEAM and ERA5 data showed a substantial increase in the frequency and affected areas of flash droughts globally, particularly since 2000. The percentage of flash droughts developing within one pentad increased from 10% in the 1980s to over 20% recently. The global exposure of agricultural areas to flash droughts increased by 20.3%, forested areas by 17.1%, and populations by 30.0% during 2001-2020 compared to 1981-2000. The Amazon Basin, eastern and southern Asia experienced disproportionately high increases in integrated risks. Analysis of the physical drivers revealed a shift from precipitation-deficit flash droughts to a dominance of heat-wave and concurrent hot and dry flash droughts. While CMIP6 models correctly simulated the geographical distribution of flash droughts, they significantly underestimated the frequency, particularly those driven by precipitation deficits or heat waves. This underestimation likely stems from the models' misrepresentation of the short-timescale dependence between precipitation and temperature and their underestimation of soil moisture's sensitivity to these variables. The models showed a tendency to overestimate the frequency of droughts driven by combined low precipitation and high temperatures, especially in dry regions, but greatly underestimated this for most temperate and high-latitude regions of the Northern Hemisphere.

The findings directly address the research question by demonstrating a significant increase in flash drought frequency and intensity, largely attributable to the rising co-occurrence of extreme hot and dry conditions. The shift toward more frequent compound hot and dry flash droughts highlights the amplified impact of climate change. The results emphasize the limitations of CMIP6 models in simulating these events accurately, raising concerns about the reliability of current projections. The significant increase in exposure for agricultural and forested areas, and especially densely populated areas, underscores the urgent need for adaptation strategies and early warning systems. The observed changes in the types of flash droughts and their increased frequency, particularly in vegetated regions, warrant further investigation into the role of land-atmosphere interactions. The accelerated onset and longer durations observed could be linked to self-intensification and propagation mechanisms amplified by warming.

This study provides a comprehensive assessment of flash drought changes, highlighting the crucial role of concurrent hot and dry conditions amplified by global warming. The findings underscore the limitations of current CMIP6 models in accurately projecting future flash drought risks. Future research should focus on improving model representations of short-timescale land-atmosphere interactions and the sensitivity of soil moisture to temperature and precipitation to enhance prediction accuracy and inform effective mitigation and adaptation strategies. The disproportionate impacts on densely populated and vegetated regions necessitate targeted interventions to minimize societal and ecological vulnerability.

The study's conclusions are based on specific definitions of flash drought and data availability. While the use of multiple datasets helps to mitigate some biases, uncertainties remain in soil moisture measurements, especially in data-sparse regions. The analysis relied on CMIP6 outputs, which have limitations, as the paper notes. Further research with improved model outputs is warranted. The focus on the recent period (1981-2020) may not fully represent long-term patterns. Future studies may benefit from employing other methods of defining flash droughts and incorporating a more expansive range of datasets.