Impacts of economy-society-environment on surface water resource in China are complex and unclear. Revealing these connections is vital to understand responses of surface water quality to anthropogenic activities. This study made an attempt to explore potential indications on surface water quality from economic, social and environmental factors in eight separate regions of China during the period of 2000–2019. Five machine learning models were employed including Greedy Thick Thinning Bayesian Belief Network, Naive Bayes, Augmented Naive Bayes (ANB), Logistic Regression and Random Forest. A total of 8 economic variables, 5 social variables and 8 environmental variables were introduced into the models. Results showed that ANB presented the best performance in estimating the surface water quality class with the highest accuracies of 81%, 75% and 87% for three surface water quality groups (Class I–III, Class IV–V and worse than Class V), respectively. The higher the surface water environmental carrying capacity in a region, the better the estimation performance of ANB on the surface water quality class. Surface water quality with Class I–III was more closely related to economic and social development, while environmental variables largely interpreted the quality of surface water with Class IV–V in most regions. The critical factors filtered by the importance analysis were indicative on surface water quality. This study provided a feasible framework in revealing the economy-society-environment nexus in the context of comprehensive management on regional surface water quality.
Read full abstract