Predicting effects of non-point source Pollution emission control schemes Based on VMD-BiLSTM and MIKE21

Abstract

Abstract Controlling non-point source (NPS) pollution is crucial for implementing water environment management, and simulating the water quality response to NPS pollution emission control schemes is of great importance. Variational mode decomposition (VMD) can overcome endpoint effects and modal aliasing issues, effectively separating intrinsic mode components. Bidirectional long short-term memory (BiLSTM) can fully mine the information contained in time series and has good predictive performance. MIKE21, when coupled with the Ecolab module, can well simulate the diffusion process of NPS pollution. The Weihe River water environment prediction model was constructed using VMD-BiLSTM and MIKE21, with ammonia nitrogen (NH3-N), total phosphorus (TP), and chemical oxygen demand (COD) as pollution indicators, showing the water quality response of the Weihe River within a few years after the implementation of agricultural and urban NPS pollution emission control schemes. Among them, the COD concentration decreased by up to 71.3%, the NH3-N concentration decreased by up to 31.4%, and the TP concentration decreased by up to 43.1%. The results show that the water quality of the Weihe River can be significantly improved by controlling NPS pollution emission, and reducing agricultural NPS pollution emission is key to decreasing ammonia nitrogen and total phosphorus concentrations and improving water quality.