Abstract: Accurately obtaining reference crop evapotranspiration (ET₀) is crucial for assessing crop water requirements and optimizing irrigation water resource management.To achieve efficient and accurate estimation of ET₀ under conditions of limited meteorological observation data,a hybrid optimization model,WOA-LightGBM,was developed by applying the Whale Optimization Algorithm (WOA) to optimize the hyperparameters of the Light Gradient Boosting Machine (LightGBM).Daily meteorological data from 15 stations in Henan Province were selected to evaluate the model’s performance,which was compared with LightGBM,K-nearest neighbor (KNN),and Random Forest (RF) models.An ensemble embedded feature selection (EEFS) method,constructed using RF,Adaptive Boosting (AdaBoost),and Gradient Boosting Decision Tree (GBDT),was employed to analyze meteorological factors significantly influencing ET₀ estimation,thereby determining the optimal input combinations.The results show that models using the four important combinations derived from the EEFS method provide more accurate ET₀ estimates.In ET₀ estimation in Henan Province,all ET₀ models exhibit significant spatial variability.Compared to the LightGBM,KNN,and RF models,the WOA-LightGBM model demonstrates stable and high estimation accuracy across different regions,with ranges of the coefficient of determination (R²),nash Sutcliffe efficiency (NSE),root mean square error (RMSE),and mean absolute error (MAE) of 0.897~0.998,0.897~0.998,0.071~0.545mm·d^-1,and 0.052~0.409mm·d^-1,respectively.Seasonal variations affect the estimation accuracy,with humidity and wind speed having greater impacts in spring and autumn,especially in winter.Even when data from target stations are missing,the WOA-LightGBM model trained on data from nearby stations maintained high accuracy.This study provides a reliable reference basis for accurately estimating ET₀ under limited meteorological conditions.