A novel nonlinear direct-mapping approach for multiple time scale driving force analysis of surface water quality variations under intense human interference

A multi-objective optimization model integrating machine learning and time-frequency analysis for supporting nitrogen and phosphorus pollution reduction in Guangzhou city, China

The unbridled discharge of nitrogen and phosphorus (NP) pollutants is believed to have surpassed ecosystem resilience limits for many regions, which is of great concern to research and governmental communities. In this research, a multi-objective optimization model was developed based on integrating advanced optimization, time-frequency analysis, and machine learning approaches into a general modeling framework. Nonlinear relationships among a variety of driving forces and variations of NP pollution can be effectively reflected and handled under multiple time scales, directly capturing the intricacy and uncertainty of water surface system within certain regions. At the same time, impacts of climate change and industry structure adjustment were addressed for deeply analyzing complexities of NP pollution. Results of the model can be used for harmonizing economic development with multi-dimensional ecological requirements, which can then be employed for supporting the mitigation of NP pollution and the reduction of extreme pollution frequency. The developed model was demonstrated through a real-world case study in Guangzhou of south China, a city grappling with the daunting task of reducing NP pollution while addressing economic needs. The results showed that reasonable adjustments to the industrial production structure would effectively reduce NP pollution while maintaining stable economic growth. Guangzhou would reduce mean NP concentrations by 7.10 % and decrease extreme pollution frequencies by 52.57 % in 2025. This approach provided substantial value for quarterly production structure adjustment in a transitional environment.