Investigation of relationship between polycyclic aromatic hydrocarbons and human activities in urban soils of China using machine learning methods

Abstract

Clarifying the impact of human activities on polycyclic aromatic hydrocarbons (PAHs) in soil can provide a scientific basis to reduce PAHs pollution and formulate control measures to manage soil PAHs pollution, which are important to reduce global warming and support China's implementation of carbon neutrality goals. Here, we characterize the PAHs in 1055 samples of urban soil collected in China from 2000 to 2020. The total concentration of PAHs ranged from 2.75 to 38,865 ng g−1. PAHs concentrations in the Yangtze River Delta (YangtzeD) and Pearl River Delta exhibited an inverted U-shaped curve over time, and PAHs concentrations in Beijing have decreased. Air pollution prevention policies and vehicle emission standards in these regions reduced PAHs in the soil. We measured PAHs in soils of 18 major cities in YangtzeD, representing a typical area. Total concentration of PAHs was 3.88–2153 ng g−1 in YangtzeD, with 40.5%, 36.3%, and 23.2% of PAHs coming from industry and transportation, coal combustion, and biomass combustion, respectively by Positive matrix factorization (PMF) model. To explore how human activities affect the PAH concentrations, we screened three machine learning models and selected SHapley Additive explanation-extreme gradient boosting (XGB-SHAP) as the best model. The results of XGB-SHAP (R2 = 0.64) show that the main human activities affecting PAHs were carbon emissions, population size, and economic development (MAS = 709.1) and industrial waste gas emissions (MAS = 577.1), and these were positively correlated with PAHs. Therefore, effective ways to reduce PAHs may largely rely on industrial co-control. With the prevention and control of air pollution, clean energy, and green and low-carbon development policies constantly being implemented, PAH concentrations will continue to be reduced in the future.