A novel hybrid model combined with ensemble embedded feature selection method for estimating reference evapotranspiration in the North China Plain

• DOI: 10.1016/j.agwat.2024.108807
• Publication Date: 2024-04-02T17:14:13Z
• Authors: Hanmi Zhou; Linshuang Ma; Xiaoli Niu; Youzhen Xiang; Jiageng Chen; Yumin Su; Jichen Li; Sibo Lu; Cheng Chen; Qi Wu
• Keywords: Feature (linguistics)

The reference evapotranspiration (ETo) is a key parameter in achieving sustainable use of agricultural water resources. To accurately acquire ETo under limited conditions, this study combined the northern goshawk optimization algorithm (NGO) with extreme gradient boosting (XGBoost) model to propose novel NGO-XGBoost model. performance was evaluated using meteorological data from 30 stations North China Plain and compared XGBoost, random forest (RF), k nearest neighbor (KNN) models. An ensemble embedded feature selection (EEFS) method results RF, adaptive (AdaBoost), categorical (CatBoost) models used obtain importance factors estimating ETo, thereby determine optimal combination inputs indicated that by top 3, 4, 5 important as input combinations, all achieved high estimation accuracy. It worth noting there were significant spatial differences precisions four models, but exhibited consistently precisions, global indicator (GPI) rankings 1st, range coefficient determination (R2), nash efficiency (NSE), root mean square error (RMSE), absolute (MAE) bias (MBE) 0.920–0.998, 0.902–0.998, 0.078–0.623 mm d−1, 0.058–0.430 −0.254–0.062 respectively. Furthermore, accuracy varied across different seasons, which more significantly affected humidity wind speed winter. When target station insufficient, trained historical neighboring still maintained precision. Overall, recommends reliable for provides calculating absence data.