Source identification of trace elements in groundwater combining APCS-MLR with geographical detector

Abstract

Accurately identifying the sources of trace elements in groundwater is difficult, especially in the areas with intensive human activities. This study proposes a framework to identify and quantify the contributions of potential sources of trace elements in groundwater. Health risk assessment is first conducted to identify the trace elements with the highest threats to groundwater quality, and then the geographical detector is used to qualitatively investigate the potential sources and processes dominating the pollution from identified trace elements, and the absolute principal component scores-multivariate linear regression (APCS-MLR) is used to quantify the contributions of different sources to the accumulation of trace elements. The proposed approach was applied to China’s Loess Plateau, a typical water-limited area with rapid urbanization and intensive agricultural activities, after conducting two groundwater sampling campaigns in a flood and a dry season. The trace element concentrations in groundwater exhibited significant spatial and seasonal differences. Health risk assessment revealed that Cr and As through oral ingestion were responsible for non-carcinogenic and carcinogenic risks in different seasons. Land use was the primary factor influencing the spatial heterogeneity of Cr in the flood season, while elevation and distance from city were dominant factors in the dry season. Normalized difference vegetation index (NDVI) and distance from city contributed the most to As change in the flood season, but NDVI was the major contributor in the dry season. The average contributions of potential pollution sources to trace elements in groundwater were in the order of natural source (35%) > geogenic process (17%) > industrial source (13%) in the flood season, but mixed sources (47%) > agricultural activities (18%) in the dry season. Distinguished from the traditional source apportionment methods, the newly proposed approach is not limited to identification of pollutant sources, but also can obtain the exact contributions of different sources. The findings provide useful information for groundwater management in the Loess Plateau.