ABSTRACT
Groundwater arsenic contamination is a public health disaster in many areas of Bangladesh. Elevated concentrations of arsenic (largely as inorganic arsenic) have also been reported in rice with concentrations variably above guideline values. Consumption of rice is therefore one of the major exposure routes of arsenic to humans. We have used random forest modelling to identify areas of high arsenic-in-rice hazard. Datasets of spatially continuous environmental parameters, including climate, soil and modelled groundwater arsenic along with six crop-specific in-situ variables, were used to map the probability of total arsenic in rice exceeding 120 µg/kg. Results shows a 4 ± 3% greater probability of high arsenic rice with shallow groundwater irrigation. Greater probabilities of high rice arsenic are also evident in High Yielding Varieties in areas with high groundwater arsenic. The models may be used to inform land and resource management strategies to reduce human exposure to arsenic.
