Frequent occurrence of the global flash drought has posed severe threats to regional agricultural production and food security in recent decades, leading to substantial yield reductions and increasing uncertainties in food systems. Therefore, it is of great importance to quantify the impacts of the flash droughts on crop yields in order to improve the drought risk assessment and agricultural adaptation. A few previous studies have statistically analyzed the temporal and spatial evolution of the flash drought events in the major grain-producing regions in China. However, it is still lacking in the fine-scale and process-based quantitation. Particularly, it is often required to consider the crop growth characteristics, phenological stages, and multiple interacting climatic factors. This study aims to assess the impacts of the flash droughts on crop yield and agricultural productivity using a crop model. Flash drought events were also identified in China during 1950-2022 using soil moisture reanalysis data. Multiple datasets were integrated, including the long-term climatic observations, soil properties, and crop-specific parameters. An AquaCrop model was established to simulate the responses of the major grain crops to various drought conditions. A “flash drought-phenology-yield” multi-scenario simulation framework was developed after the simulation. The temporal distribution of the crop phenological stages was explicitly incorporated to evaluate the yield responses to different drought intensities and durations. A systematic and spatially explicit assessment was performed on the evolution characteristics of the flash droughts and their impacts on the crop growth and yield in the diverse climatic zones. The results revealed that: 1) The calibrated AquaCrop model demonstrated high simulation accuracy, with coefficients of determination (R²) for maize, rice, and wheat reaching 0.909, 0.821, and 0.769, respectively. 2) There was a significant increase in the frequency of the flash drought events in the major grain-producing regions from 1950 to 2022, indicating an outstanding upward trend, particularly after the 1980s. The vegetative and reproductive growth stages were identified as the two most vulnerable periods for the flash drought occurrence, accounting for approximately 38.6% and 15.0% of the total events, respectively. 3) The yield losses induced by flash droughts were consistently higher than those by conventional slow-onset droughts within the growing season, with an average increase of about 10% in the yield reduction. Furthermore, the rice exhibited the highest sensitivity to the flash droughts, followed by maize, whereas the wheat was relatively less affected among the staple crops. Moreover, the reproductive growth stage was identified as the most critical period for the yield formation, indicating the most susceptible to flash drought stress (yield losses exceeding 47.10%). Overall, there was a quantitative assessment of the flash drought impacts on the crop yields. The crop phenology and rapid soil moisture depletion were then incorporated into the drought evaluation in the future. The “flash drought-phenology-yield” framework can offer a flexible approach to couple the flash drought dynamics with the crop growth simulations, thereby improving the accuracy of yield impact assessments under complex climates. Consequently, these findings can also provide a scientific basis to develop adaptive strategies, thus enhancing early warning on the agricultural risks with the increasing frequency and intensity of flash droughts.

Soil moisture is one of the most important indicators for the exchange of energy and water in the landing gas. The resulting surface evapotranspiration, water distribution, and soil heat capacity can be changed under the local, regional, and even global climate. Therefore, soil moisture can be measured to control the dry and wet state of the soil, as well as the climate. This study aims to explore the spatiotemporal distribution characteristics and evolution trends of soil moisture memory at multiple time scales. The soil moisture data was collected from the European central reanalysis products (ERA5-Land) in China from 1979-2018. The lag autocorrelation was then combined to calculate the soil moisture memory. Finally, a systematic investigation was implemented to determine the spatiotemporal distribution and evolution trend of soil moisture memory under different time scales (season, year, and inter-year) and dry/wet conditions. The results show that: 1) There was no autocorrelation coefficient in the 0.01 significance test after the lag time over 300 d from the perspective of the annual average state, indicating the average regional memory time of 300 d. Spatially, the memory time presented a spatial pattern of high in the north and low in the south. Among them, the memory time was less than 20 d in the southwest, less than 40d in the central, the East, and South China, whereas, there was over 80 d in the northwest, and North China, with a maximum of 240 d. 2) In season, the memory time of soil humidity in most areas was characterized by short summer, long winter, as well as the second autumn and spring. There were small seasonal differences in the Qinghai-Tibet Plateau, whereas, the large seasonal differences were found in North China, with the winter exceeding 70, 20 d in summer, and 30-50 d in spring and autumn. There was also the spatial distribution pattern of high in the north and low in the south under different dry and wet conditions. But the memory time was much longer in the dry period than the wet one. For example, the Qinghai-Tibet Plateau, North China, and northeast China were recorded as more than 90 d during the dry period, whereas, the wet period was less than 60 d. 3) The evolution trends of soil moisture memory and duration were relatively consistent in the northeast, central, South China, and the northern part of the Qinghai-Tibet Plateau. The memory time of soil moisture decreased in the northern, especially in the north, and Qinghai-Tibet Plateau. There was an increase in the southern regions. In addition, the spatial distribution of the memory evolution trend was changed with the increase in lag time, indicating a lag time of 30 d in most parts of Chinese mainland.

Based on ground meteorological observations and land surface reanalysis data, flash drought and high temperature events in the Yangtze River Basin from 1950 to 2021 were identified, and the spatiotemporal distribution characteristics of these two extreme events were analyzed. Combined with the Copula function, the joint probability distribution and conditional probability distribution functions of high temperature and flash drought were constructed, and the onset time periods when the soil moisture content percentile reached moderate, severe, extreme, and excessively extreme droughts under different temperature conditions were estimated. The results show that the times of occurrence of flash droughts and days of high temperature have significantly increased since 1990. The onset time period of droughts is significantly negatively correlated with daily maximum temperature, and shortens as the temperature increases, indicating the higher likelihood of flash drought occurrence at higher temperatures. The joint probability of high temperature and flash drought in the upper and middle reaches of the Yangtze River Basin, including Yunnan, Sichuan, Chongqing, Guizhou, and Hubei, is higher than that in the downstream areas in typical events in 2006, 2011, and 2013. Under low temperature conditions, the onset time period of droughts in the source region of the Yangtze River Basin is twofold of that in the middle and lower reaches. With the rising temperatures, the onset time period of droughts in the source region and the upper reaches shortens. When the daily maximum temperature exceeds 30℃, the development of droughts shows a spatial pattern of rapid onset in the upstream and delayed onset in the downstream.