Self-Organizing Map provides new insights into the MixSIAR model for calculating the source contributions of sulfate contamination in groundwater

Abstract

The concentration of sulfate in global groundwater has been observed a significant upward trend in recent years. Excessive sulfate levels contribute to increased groundwater salinity and acidification, thereby posing a threat to human health and ecological balance. For effective groundwater pollution management and control, accurately quantifying the sources of sulfate pollution remains a challenge. This research integrates the Self-Organizing Map (SOM) clustering method to enhance the accuracy of the Bayesian isotope mixing model (MixSIAR) in quantifying the contribution rate of groundwater sulfate. During the dry season, sulfate (SO₄^2-) primarily originates from the oxidation of pyrite, whereas SO₄^2- sources include both pyrite oxidation and the co-dissolution of carbonate rocks and gypsum during the normal and wet seasons. Incorporating SOM, the MixSIAR model demonstrates reduced values of Leave-One-Out Information Criterion (LOOIC), and Widely Applicable Information Criterion (WAIC) (LOOIC = 82.5, and WAIC = 82.3). Overall, in the study area, coal mines (accounting for 34.3% - 48.4%) are identified as the primary pollution sources, particularly in Clusters 3, 4 and 5. Clusters 1, 2, and 5 are more significantly affected by other pollution sources, with fertilizers contributing 32.7%, evaporite dissolution contributing 24.1% and 24.2%, respectively. This study supports the development of regional pollution control strategies.

Keywords: Bayesian isotope mixing model; Groundwater; Isotopic characteristic; Self-Organizing Map; Sulfate contamination source.