Research progress on decomposing ensemble models for streamflow forecasting

Climate change has significantly impacted hydrological conditions across diverse river basins. Traditional hydrological models fall short in accurately representing the complex dynamics of daily streamflow sequences. The decomposition ensemble model, designed to manage nonlinear and non-stationary streamflow forecasting effectively, has not been comprehensively studied or summarized. This gap limits its development and practical application in streamflow forecasting. This paper provides a detailed overview of the theoretical foundations, application developments, and practical challenges of the decomposition ensemble model. It systematically categorizes and summarizes mainstream algorithmic applications and outlines prospective research directions. The analysis reveals that: ①The decomposition ensemble model excels at managing nonlinear and non-stationary streamflow forecasting. ②The applicability of various decomposition ensemble models to streamflow forecasting varies, necessitating the selection of suitable model structures and parameter configurations for specific scenarios. ③Decomposition ensemble models must address mode mixing and also tackle boundary effects and data leakage, challenges that could be mitigated through more sophisticated coupled models and advanced decomposition strategies. Despite the promising performance and broad potential applications of the decomposition ensemble model in streamflow forecasting, it is crucial to refine the algorithm and overcome practical challenges to improve the model's accuracy and reliability. Such enhancements will make it a more effective tool for managing complex hydrological variations.