A novel groundwater pollution risk assessment method for producing-enterprises sites: Integrating scenario-specific pollution evaluation with Gaussian mixture model clustering

Groundwater pollution risk assessment is crucial for effective environmental management, particularly in industrial settings where producing-enterprises are major sources of pollution. Traditional risk assessment methods often lack precision and adaptability for these sites. This study developed a novel evaluation index system integrating anti-pollution performance, pollution status, and pollution loading, focusing on three specific scenarios: pollution from the three wastes, unorganized emissions during production processes, and storage tank leaks. To address limitations in traditional weight determination methods, we applied the Gaussian mixture model (GMM) for data clustering, completing the risk assessment. A chemical producing-enterprise in Southern China was selected as the study site, revealing that 81 % of the total enterprise area was at low or relatively low risk, while 3 % remained at high risk, requiring immediate remediation. To validate the effectiveness of the method, we compared GMM with two commonly used clustering algorithms Hierarchical Clustering and DensityBased Spatial Clustering of Applications with Noise, evaluating clustering performance using indices like the Silhouette Coefficient, Calinski-Harabasz Index, and Davies-Bouldin Index. The results confirmed that GMM with logarithmic transformation outperforms other algorithms, offering a valuable reference for enhancing the management efficiency of enterprises in controlling pollution at source.