The interactive natural drivers of global geogenic arsenic contamination of groundwater

Abstract

The effects of interactions between natural factors on the genesis of high arsenic groundwater on a global scale remain poorly understood. This paper develops a neural network model of global groundwater arsenic contamination using 26 indicators including climate, soil physical and chemical properties, and more than 70,612 data points worldwide to identify and quantify typical interaction effect. Importance analysis shows that the pivotal main effects are produced by climate and soil texture indicators. Among these main effects, seven true and six second-order interactions, with five related to precipitation in the latter, were identified to be responsible for about 10.5% of the variation in the occurrence probability of high arsenic groundwater. For the second-order interactions, the interaction strength in the torrid zone is stronger than that in the temperate zone. On a global scale, high precipitation is more likely to generate a dramatic interaction, with such interactions less likely to occur at the low end of the interactive indicators. High probabilities of elevated arsenic concentration are usually accompanied by synergy and vice versa.