Diagnosing drought–flood vulnerability of rice systems across growth stages using Sentinel-1/2 and CHIRPS time series

Abstract

Compound and rapidly alternating drought and flood events pose a major challenge to rice production because crop impacts depend strongly on when hydroclimatic stress occurs and how crop condition evolves afterwards. Here, rice vulnerability is characterized by stage-specific exposure, sensitivity, and recovery. We developed a Stage-resolved Drought–flood Rice Vulnerability (SDRV) framework to diagnose vulnerability formation in stable single-season rice across the middle and lower reaches of the Yangtze River (MLYR), China, during 2020–2025. SDRV integrates CHIRPS precipitation and Sentinel-1/2 time-series within four regionally harmonized growth-stage windows: establishment to early tillering (S1), tillering to jointing (S2), booting/heading to grain filling (S3), and grain filling to maturity (S4). Stage-specific percentile mapping places the underlying hydroclimatic and canopy observations in a common 0–1 diagnostic space, after which the three components are integrated into stage-wise and annual vulnerability. Results show that annual vulnerability ranked as 2022 > 2025 > 2023 > 2020 > 2021 > 2024. Within the study period, drought-dominant and drought-involved pathways were associated with greater annual vulnerability than the flood-dominant pathway represented by 2020. Vulnerability amplification was concentrated mainly in S2–S3. In 2022, persistent drought-dominant stress intensified vulnerability toward S3, producing the highest annual mean vulnerability (0.599) and the largest high-vulnerability fraction (50.4%). In 2025, alternating drought and flood conditions generated a different pathway, with vulnerability concentrated in S2 and partially reduced in S3. Spatially, high vulnerability repeatedly formed a northern and central MLYR corridor, especially across Anhui, Jiangsu and adjoining parts of the middle Yangtze region. Province-level comparison with a yield-loss proxy further showed a positive external-consistency relationship for annual vulnerability after accounting for year effects. These results demonstrate that rice vulnerability under drought–flood stress develops through stage-dependent interactions among hydroclimatic forcing, canopy response, and subsequent recovery. SDRV therefore provides a process-oriented approach for identifying when and where vulnerability intensifies within the growing season.