Sustainable management strategy for phosphorus in large-scale watersheds based on the coupling model of substance flow analysis and machine learning

Abstract

Clarifying the quantitative response relationship between socio-economic factors and water quality is key to developing a sustainable phosphorus (P) management strategy. we established a regulation framework for the feedback loop between socio-economic factors and water quality (i.e., P fluxes) from a spatial perspective based on a substance flow analysis (SFA) process model and machine learning (ML) model using a Bayesian optimization. The study demonstrated that utilizing long-term and intensive monitoring records, along with a ML algorithm to model the P response of the water body, resulted in good robustness and accuracy. Watershed P flows have a significant impact on P flux, and the response of P flux exhibits non-linear and non-lagged characteristics. The SFA–ML coupled model advances the current understanding of how P flows contribute to guiding P cycling in a watershed. P-SFA can serve as reliable feedback medium on the interaction between socio-economic activities and water quality in watersheds.