A distributed hybrid model for agricultural water-land resource matching

Under climate and land use changes, the spatio-temporal matching patterns between agricultural water and land resources are crucial for making informed irrigation decisions that help reduce the risk of agricultural water scarcity. Traditional methods lack predictive capacity due to data limitations. In this study, a distributed hybrid model (VIC-RCM) based on the Variable Infiltration Capacity-3Layers (VIC-3L), Random Forest Cellular Automata Markov model (RF-CA-Markov), and Global Climate Models (GCMs) was developed. This model integrates VIC-3L for water resource simulation, RF-CA-Markov for land use simulation, and GCMs for regional climate projection, employing a water-land matching coefficient to quantify their spatial matching status. This model offers two advantages: its simulated grid area is adjustable without constraint by administrative boundaries, and it exhibits predictive capability. Application to the Nanpan River Basin demonstrated that the VIC-RCM model achieved high accuracy in simulating agricultural water-land resource matching. The average gridded matching coefficients for the entire basin from the 1970s to the 2010s displayed a decreasing trend, with water-scarce areas mainly concentrated in the central and northwestern regions. Under four shared socioeconomic pathway (SSP) scenarios(SSP1–2.6, SSP2–4.5, SSP3–7.0, and SSP5–8.5) of Coupled Model Intercomparison Project Phase 6 (CMIP 6), the mean gridded matching coefficients from the 2010s to the 2090 s continued to decline. Spatial distribution remained consistent across periods, with lower matching coefficients in the central and south-central parts of the basin. This study provides more precise risk classification zones and corresponding combination measures based on the model, which will support regional agricultural planning and irrigation decision-making.