Multivariateanalysisoftheheterogeneousgeochemicalprocessescontrollingarsenicenrichmentinashallowgroundwatersystem

黄爽兵*,刘昌蓉

Abstract

The effects of various geochemical processes on arsenic enrichment in a high-arsenic aquifer at Jianghan Plain in Central China were investigated using multivariate models developed from combined adaptive neuro-fuzzy inference system (ANFIS) and multiple linear regression (MLR). The results indicated that the optimum variable group for the AFNISmodel consisted of bicarbonate, ammonium, phosphorus, iron, manganese, fluorescence index, pH, and siderite saturation. These data suggest that reductive dissolution of iron/manganese oxides, phosphate-competitive adsorption, pH-dependent desorption, and siderite precipitation could integrally affect arsenic concentration. Analysis of the MLR models indicated that reductive dissolution of iron(III) was primarily responsible for arsenicmobilization in groundwaters with lowarsenic concentration. By contrast, for groundwaters with high arsenic concentration (i.e., > 170 μg/L), reductive dissolution of iron oxides approached a dynamic equilibrium. The desorption effects from phosphatecompetitive adsorption and the increase in pH exhibited arsenic enrichment superior to that caused by iron(III) reductive dissolution as the groundwater chemistry evolved. The inhibition effect of siderite precipitation on arsenic mobilization was expected to exist in groundwater that was highly saturated with siderite. The results suggest an evolutionary dominance of specific geochemical process over other factors controlling arsenic concentration, which presented a heterogeneous distribution in aquifers.