The superior coupled model with innovative strategy for accurate imputation of missing hydrological monitoring data in water research-A case of groundwater level data

• Propose a missing hydrological monitoring data imputation theory based on the "Decomposition-Classification-Feature Extraction-Imputation". • Extreme-point Symmetric Mode Decomposition (ESMD) can decompose time series into multiple real feature IMFs and R. • Permutation Entropy (PE) and Singular Value Decomposition (SVD) can reasonably reduce the number of components and further extract features. • Whale Optimization Algorithm-Bidirectional Long Short-Term Memory (WOA-BiLSTM) algorithm significantly enhances the data imputation accuracy of the BiLSTM model. • The ESMD-PE-SVD-WOA-BiLSTM coupled model possesses strong imputing capability with the MAPE of 0.32 %. The missing hydrological monitoring data pose challenges to regional water resource analysis and research. This paper proposes a strategy of “Decomposition-Classification-Feature Extraction-Imputation” and establishes a coupled model integrating Extreme-point Symmetric Mode Decomposition (ESMD), Permutation Entropy (PE), Singular Value Decomposition (SVD), and Whale Optimization Algorithm-Bidirectional Long Short-Term Memory (WOA-BiLSTM) for the hydrological monitoring data imputation. The practicality of this model in imputing discrete and continuous missing data is verified using daily groundwater level data from monitoring stations around Qinghai Lake, spanning from April 1, 2019, to June 30, 2020. The results show that ESMD can decompose the groundwater level sequences into IMFs and R; the PE-SVD method for complexity-based reclassification and feature extraction can reasonably reduce the number of components while guaranteeing imputing accuracy. Compared with other models, the ESMD-PE-SVD-WOA-BiLSTM coupled model has the best imputing accuracy (MAE=17.43, RMSE=21.08, MAPE=0.32 %, IoA=90 %), providing an effective and reliable tool for precise hydrological monitoring data imputation and regional water resource research.