Modeled Groundwater Decline is Associated with Lower Arsenic Concentrations in Hanoi

Abstract Elevated arsenic concentrations in groundwater pose health risks. Geogenic arsenic contamination is driven by biogeochemical mobilization and groundwater dynamics. Intensive pumping can initially exacerbate contamination by drawing arsenic or mobilizing organic matter into aquifers, but over longer time scales may also promote dilution and flushing in some settings. To better understand the mechanisms controlling arsenic concentrations, we examined changes over two decades in Hanoi (Vietnam), a hotspot of geogenic arsenic contamination and groundwater pumping. Arsenic concentrations from 470 wells sampled in the year 2000 were compared with 434 wells sampled in the same area in 2013 and 2023. Median concentrations declined from 14 to 8 μg/L between 2000 and 2023, while mean concentrations and the proportion of high-arsenic wells (>100 μg/L) remained similar. Using hydrochemical and groundwater-level data, CatBoost models combined with SHAP analysis assessed changes in predictor importance. In 2000, arsenic was more strongly associated with redox predictors, whereas groundwater-level predictors became more important in 2023. Within wells used to evaluate groundwater-level change (n = 251), larger decadal declines were associated with lower arsenic concentrations (median = 31 μg/L) than wells without declines (median = 120 μg/L). We hypothesize a decadal mechanism shift, from redox to hydrological controls, particularly driving arsenic declines.