UAV‐Based Canopy Equivalent Water Thickness for Variable‐Rate Drip Irrigation in Winter Wheat

ABSTRACT Canopy equivalent water thickness (EWT) is a quantitative indicator of canopy water status. However, its application in variable‐rate drip irrigation remains insufficiently investigated. A two‐year lysimeter experiment was conducted in the North China Plain (NCP) using two winter wheat cultivars and four irrigation treatments: FI0, FI60, FI90 and variable‐rate irrigation (VRI). Unmanned aerial vehicle (UAV) multispectral imagery was used to estimate plot‐averaged canopy EWT at jointing, booting, flowering and grain filling. Twelve vegetation indices and four machine‐learning models were evaluated for EWT inversion. LightGBM performed best in 2025 (coefficient of determination [ R 2 ] = 0.716, root mean square error [RMSE] = 0.043 g cm −2 , relative prediction deviation [RPD] = 1.91), whereas the decision tree performed best in 2026 ( R 2 = 0.791, RMSE = 0.033 g cm −2 , RPD = 2.32). Plot‐averaged EWT increased from jointing to flowering and declined during grain filling, while red‐edge indices maintained relatively stable relationships with EWT. Compared with FI90, VRI reduced crop evapotranspiration by approximately 40% while maintaining grain yield close to FI60. The EWT‐based VRI achieved a favourable balance between grain yield and water input, demonstrating its potential for precision drip irrigation.