Impact and Prediction of Pollutant on Mangrove and Carbon Stocks: A Machine Learning Study Based on Urban Remote Sensing Data

Mangrove ecosystems have important ecological and economic values, especially their ability to store carbon. However, in recent years, human disturbance has accelerated mangrove degradation. Among them, the emission of pollutants cannot be ignored. It is of great significance for carbon emission reduction and ecological protection to study the impacts of different pollutants on mangroves and their carbon stocks. Based on the remote sensing data of coastal areas south of the Yangtze River in mainland China, this paper builds the ensemble learning model Random Forest (RF) and Gradient Boosting Regression (GBR) to empirically analyse the relationship between industrial wastewater, industrial sulfur dioxide (SO2), PM2.5 and mangrove forests. The results show that the pollutant concentration of meteorological normalisation is more stable. The importance of pollutants presents regional heterogeneity. The area of mangroves in different cities and the corresponding total carbon stocks show different trends with the increase or decrease of pollutants, and there is a dynamic balance between urban pollutant discharge and mangrove growth in some cities. The research in this paper provides an analysis and explanation from the perspective of machine learning to explore the relationship between mangroves and pollutants and at the same time, provides scientific suggestions for the formulation of future pollutant emission policies in different cities.