Abstract
Accurate estimation of reference evapotranspiration (ET 0 ) and crop coefficients (K c ) is critical for irrigation planning, particularly in data‐limited regions where agriculture dominates freshwater consumption. Although machine learning (ML) methods have been widely applied to ET 0 and K c estimation, most studies address these parameters separately or focus on a single seasonal K c value rather than stage‐specific coefficients. This study presents an integrated machine learning framework that simultaneously estimates monthly ET 0 and wheat stage‐specific crop coefficients (K c_ini , K c_mid , K c_end ) using long‐term district‐level meteorological data from Türkiye. A total of 19 algorithms, including linear models, tree‐based ensembles, boosting methods, support vector machines (SVR) and deep learning models, were evaluated using 10‐fold cross‐validation and standard performance metrics. The results showed that linear models provided the most consistent performance for monthly and annual ET 0 estimation ( R 2 ≈0.85–0.87), while nonlinear ensemble models achieved superior accuracy for K c prediction ( R 2 > 0.95). Feature importance analysis identified solar radiation and temperature as dominant ET 0 drivers, whereas elevation and relative humidity were key controls for stage‐specific K c . The proposed framework provides a scalable approach for regional agricultural water management and supports model selection under data‐limited conditions.