Identifying Irrigated Areas With a kNDMI – NDVI Framework and Assessing Distance‐Related Runoff Responses in the Upper Dahei River Basin, 2000–2020

ABSTRACT

In semi‐arid regions, climate variability and intensive human activities, particularly, agricultural irrigation, jointly affect long‐term runoff change. Given the challenges in monitoring irrigation withdrawals, irrigated area is frequently used as a spatial proxy for irrigation water use in hydrological attribution studies. Traditional normalized difference vegetation index (NDVI)‐based methods often fail to differentiate irrigated fields from rainfed croplands, introducing uncertainty into runoff attribution. Here, we developed a soil‐moisture‐sensitive fusion of the kernel normalized difference moisture index (kNDMI) and NDVI to identify irrigated croplands, achieving a multi‐year average agreement of 74% with local statistical yearbook data. Based on the mapped irrigation distribution, we improved the dynamic Fu‐type Budyko framework using a two‐stage parameterization of ω, a catchment‐specific parameter that integrates the effects of underlying‐surface characteristics on the long‐term partitioning of precipitation between evapotranspiration and runoff. This scheme preserves the background ω‐cultivated‐area relationship while introducing an irrigation‐informed intercept adjustment to represent irrigation‐associated variation in catchment water use, without interpreting the adjustment as an independently isolated irrigation contribution. To evaluate irrigation‐related hydrological responses, records from the Qixiaying and Zhuozishan streamflow‐gauging stations were used to characterize runoff from their respective contributing catchments. Annual streamflow declined significantly over their respective records ( p  < 0.001), with variation‐period means decreasing by 50.0%–62.5% relative to corresponding station‐specific baseline means. Buffer regressions showed scale‐related variation in irrigation–runoff associations, but no statistically supported distance‐decay gradient was established at either station. Together with contrasting irrigation water sources, these patterns suggest that hydraulic connectivity between irrigation water sources and the river network may provide a plausible process‐based explanation for spatial variation in irrigation‐related runoff responses. These findings highlight the value of considering water‐source connectivity when interpreting irrigation–runoff relationships in changing semi‐arid watersheds.