Runoff potential index for upland-lowland drought assessment in rainfed rice using earth observation and mechanistic crop modelling

Drought vulnerability assessment in agricultural systems remains increasingly critical under climate change, yet current approaches are constrained by limitations of existing topographic indices, particularly in low-gradient terrains where the widely-used Topographic Wetness Index (TWI) exhibits numerical instability and fails to detect critical microtopographic variations that control water retention at field scales. This study introduces the Runoff Potential Index (RPI), a curvature-based terrain metric that addresses specific limitations of slope-dependent indices for climate-resilient agricultural drought assessment: RPI(x, y) = ∇2z/(|∇z|+ε), integrating local terrain curvature (via Laplacian of elevation) with slope magnitude. The analysis presents complementary approaches combining: (1) RPI terrain analysis using satellite-derived elevation data for upland-lowland differentiation based on terrain-controlled water redistribution, identifying runoff-prone uplands versus water-retaining lowlands, and (2) CERES-Rice mechanistic crop modeling driven entirely by Earth observation data to evaluate drought stress patterns across varying sowing dates, supporting climate adaptation strategies in data-scarce regions. The RPI maintained analytical sensitivity across subtle elevation gradients (0.7-1.8 m variations) where TWI becomes numerically unstable, successfully detecting centimeter-scale microtopographic variations critical for water retention. Terrain analysis revealed distinct upland-lowland differentiation patterns, with lowland areas achieving 200 kg/ha higher yields compared to upland areas. CERES-Rice simulations across 20 years (2000-2019) identified optimal sowing windows that minimize drought stress, with delayed sowing causing yield reductions exceeding 1,500 kg/ha. Critically, terrain-based yield advantages (200-300 kg/ha) are substantially smaller than temporal optimization benefits, exposing limitations in current mechanistic models that fail to adequately represent topographic water redistribution effects captured by RPI analysis. The Earth observation-based framework enables drought vulnerability mapping without ground-based data requirements, supporting climate adaptation in agricultural systems globally. The findings reveal conceptual limitations in bucket-based crop models and demonstrate scalable approaches for drought-resilient agriculture under changing climate conditions. This framework enables practical climate adaptation through: (1) field-specific sowing recommendations that prevent 45-73\% yield losses from suboptimal timing, (2) identification of drought-vulnerable zones requiring targeted water management, and (3) satellite-based drought risk assessment accessible to smallholder farmers in data-scarce regions, directly supporting SDG 13.1 (strengthen resilience and adaptive capacity to climate-related hazards) and SDG 13.3 (improve education and capacity-building on climate change adaptation).