Agricultural drought occurring in southwest China’s major grain-producing region: Spatiotemporal evolution, crop growth risks and yield estimation

• Ten-day VTCI accurately identified multi-year and spatial variation in drought conditions in the Sichuan Basin. • Hodrick-Prescott filtering method performed well for yield-detrending in three crops examined (rice, wheat, maize). • The response to drought of main crops presented spatial heterogeneity. • Random forest provided qualified yield estimates of main crops in the Sichuan Basin. Drought has been the key environmental stress limiting agricultural development and crop productivity, especially in major grain production areas. This study aims to monitor drought conditions in time and assess accurately the impacts of drought on crop yield of major grain production areas (the Sichuan Basin, PR China), and to further establish a technical framework for drought monitoring and crop yield estimation considering drought stress and crop growth status. We firstly employed vegetation temperature condition index (VTCI) at ten-day intervals, to capture the drought characteristics, especially in a typical historical drought event. Then, several detrending methods were introduced in the study to accurately separate meteorological yields of main crops including wheat, maize and rice. Finally, the correlation between VTCI at ten-day intervals and detrended crop yield was analyzed to assess drought-crop risk and determine the key growth period. VTCI during the key growth period and crop yields were input into random forest (RF) model for yield estimation of main crops. Spatiotemporal evolution of drought condition showed that drought characteristics exhibited obvious spatial heterogeneity, and Chengdu Plain suffered from longer duration and stronger intensity of drought events. Sichuan Basin suffered from two severe droughts in 2022. Hodrick Prescott filtering method effectively separated meteorological yields of main crops, and 66.67 % counties of wheat, 70.87 % counties of maize and 70.45 % counties of rice passed the ADF stationary test. In addition, VTCI from early January to late May showed stronger correlation with wheat yield and VTCI from early March to late August was closely related to yields of rice and maize. RF model provided a qualified estimated yield of main crops in the Sichuan Basin with R 2 more than 0.5 in most cases. This study revealed spatiotemporal evolution of drought events and drought-crop risk in the Sichuan Basin, PR China, which provided a reference for improving the productivity of main crops in the drought-prone region and major grain-producing region.