Abstract
Cold-region spring dryland grain crops are highly sensitive to water stress during the seedling stage, making spring agricultural drought a major constraint on crop productivity in semi-arid regions. Conventional drought monitoring often relies on the Standardized Soil Moisture Index (SSMI), which does not account for vertical root distribution and can produce ''pseudo-wet'' signals in deep soil layers due to seasonal freeze–thaw. To address these limitations, we developed a Profile-Attenuation-based Standardized Soil Moisture Index (PASSMI).By incorporating a profile-attenuation weighting function, PASSMI amplifies shallow moisture signals while suppressing deep-layer interference, thereby better aligning soil moisture profiles with crop physiological demands.Compared to SSMI, PASSMI is more sensitive in identifying drought events and shows stronger consistency with actual agricultural drought impacts. Using the semi-arid region of western Heilongjiang as a study area, we integrated in-situ observations with CMA-RA reanalysis data to reconstruct vertical soil moisture profiles (1980–2020) through a hybrid framework combining Light Gradient Boosting Machine with Gaussian Process Regression (LightGBM-GPR). Comparative analysis shows that PASSMI improves sensitivity in identifying spring drought events by 58% over SSMI, and exhibits approximately 2.3 times stronger correlation with crop yield anomalies. Spatially, spring drought in the study area exhibits a ''wet east, dry west'' pattern. Temporally, the drought cycle followed four stages: quiescence, intensification, peak, and recovery, with a 70.3% occurrence frequency over the 41-year period, dominated by mild to moderate events. This study provides a more effective tool for spring drought monitoring in cold-region dryland agriculture and offers scientific support for water resource management and irrigation optimization in similar semi-arid environments.