New perspective on density-based spatial clustering of applications with noise for groundwater assessment

This study introduces an integrated approach combining Kernel Principal Component Analysis (Kernel PCA) and Density-Based Spatial Clustering of Applications with Noise (DBSCAN) for the assessment of groundwater quality in arid environments. Kernel PCA was employed to reduce the dimensionality of high-dimensional datasets, outlier handling, and enhanced cluster separation. Five kernel types viz: linear, polynomial, radial basis function (RBF), sigmoid, and cosine; were compared, with the polynomial kernel demonstrating superior performance in preserving variance and achieving effective dimensionality reduction. DBSCAN identified spatial clusters and anomalies (outliers) in groundwater quality, with optimal eps = 0.05 and minPts = 3, determined using the Silhouette Score (SS) and Davies-Bouldin Index (DBI). The analysis revealed higher salinity levels influenced by seawater intrusion and over-extraction due to heavily urbanized and agricultural areas. The spatial clustering analysis provides a comprehensive view of distinct physicochemical zones and contamination hotspots. This novel Kernel PCA-DBSCAN framework enhances the detailing of groundwater quality assessment of physicochemical patterns and supports sustainable resource management.