Nonlinear threshold effects of waterlogging stress on crop productivity across China: A remote sensing-based macro-assessment

CONTEXT Against the backdrop of intensifying climate change and frequent extreme precipitation, waterlogging increasingly threatens agricultural production. Ensuring food security requires clarifying how waterlogging stress affects crop productivity at a macro scale. OBJECTIVE This study aims to systematically analyze the waterlogging response characteristics of maize and rice across China's major agricultural regions, and to identify empirical thresholds that can inform region-specific adaptation strategies. METHODS We developed a novel framework integrating remote sensing retrieval and constraint line analysis. Waterlogging was identified using relative water content (RWC) based on daily soil moisture and field capacity data from 2003 to 2023. Three waterlogging metrics were calculated: waterlogging frequency (WF), waterlogging duration (WD), and average relative water content (ARWC). Crop biomass was represented by net primary productivity (NPP) retrieved from the CASA model. Constraint line analysis (XGBoost-based fitting) was applied to quantify the nonlinear threshold effects of waterlogging on maize, single-season rice, double-season early rice, and double-season late rice across nine agricultural regions in China. RESULTS AND CONCLUSIONS From 2003 to 2023, waterlogging intensity significantly decreased in most agricultural regions of China ( P < 0.01), but significantly increased in Northeast China (NE) and North China (NC) ( P < 0.01). Maize in NE was severely affected by waterlogging ( ρ = −0.86; k = −3.37), with NPP declining as WD increased. Maize in Northwest China (NW) (ARWC threshold = 49.22%) and the Loess Plateau (LP) (ARWC threshold = 51.40%) showed higher sensitivity to waterlogging. Single-season rice in the Middle-Lower Yangtze River (MLYR) exhibited strong waterlogging tolerance, with an ARWC threshold as high as 70.97% during its reproductive growth stage. The empirical thresholds for both double-season early and late rice in the MLYR were higher than those in South China (SC). In both MLYR and SC, rice biomass showed a significant monotonic decrease with increasing WF or WD ( ρ = −0.91 to −0.46). These findings demonstrate that crop responses to waterlogging are highly region-specific, with clear threshold effects. SIGNIFICANCE Unlike previous studies that have examined plant responses at the field or genetic level, or used remote sensing to identify areas prone to waterlogging without quantifying nonlinear thresholds for individual crops. Our framework overcomes spatial limitations by explicitly deriving empirical response boundaries for maize and rice across major agricultural regions, bridging the gap between coarse satellite observations and crop-specific physiological tolerance. This enables a direct linkage between macro-scale remote sensing signals and physiological tolerance limits, capturing the ecological context dependence of waterlogging stress. These regional thresholds provide actionable benchmarks for emergency dewatering in NE and NC maize systems, for irrigation–salinity co-management in NW and LP, and for drainage scheduling by growth stage in MLYR and SC rice systems, offering a transferable framework for remote sensing-based agricultural adaptation planning.