Abstract Specific yield (SY) is a key parameter in estimating groundwater storage change. Accurately predicting SY in unconfined aquifers remains challenging, as SY varies with groundwater‐level fluctuations and unsaturated‐zone moisture fluxes. This challenge is particularly evident in the North China Plain (NCP), which has experienced substantial fluctuations in groundwater resources. This study constructed and selected data‐driven models to predict SY using hydroclimatic and hydrogeologic predictors. Additionally, input features were refined by assessing the contributions of driving factors. An ensemble prediction model for SY was developed using random forest and light gradient boosting machine as base learners, with seven input features including precipitation. The model achieved a NSE of 0.794 and a root mean square error of 0.0191 during the validation period. Based on the prediction model, the spatiotemporal distribution of SY across the NCP from 2004 to 2023 was updated and analyzed. The results show a mean SY of 0.070 for the NCP, with a declining trend of −2.62 × 10−3 decade−1 from 2003 to 2023. Compared to the shallow groundwater storage anomalies (GWSA) calculated using a fixed SY, the shallow GWSA derived from the updated SY exhibits a more pronounced declining trend of 15.48 mm/yr and a stronger correlation (CC = 0.76) with the GWSA based on gravity recovery and climate experiment satellite data. The systematic framework proposed in this paper is effective for SY prediction and accurate GWSA assessment, providing critical technical support for sustainable groundwater resource management.
Han et al. (Sun,) studied this question.