A MATHEMATICAL ANALYSIS OF ARSENIC IN GROUNDWATER - A CASE STUDY OF NEPAL

Arsenic contamination in groundwater is a serious health hazard in the Ganges floodplains, yet the root cause of high groundwater arsenic in this region remains uncertain. By far the most abundant data on this phenomenon are in the form of regional "blanket" testing results for arsenic in shallow tube wells. Relatively little information is available in subsurface hydrology in these areas.

We postulate that there exist some measurable physical quantities that are correlated with arsenic which elucidate the mechanism of arsenic pollution. Mathematical modeling and analysis of the regional data can extract constraints on these mechanisms. Arsenic values for some 400,000 wells in the Terai of Nepal, Ganges floodplain, are available. This study focuses on 1850 samples from Nawalparasi, the most affected district in Nepal.

We model arsenic concentration as a function of other measured physical quantities including well 'age', 'depth', 'number of users', and 'water pH'. We seek a suitable statistical model that discards the insignificant quantities and include the significant ones in terms of variation of groundwater arsenic. These models can then serve as the basis for quantitative risk assessments and cost/benefit analysis to optimize remedial measures.

The models consider the risk of exceeding international and national threshold for arsenic in drinking water in 3D. Classical Multiple Linear Regression and the recently developed Quantile Regression are used for the quantitative analysis of arsenic contamination in the groundwater. Initial results show a positive spatial correlation between the bottom of the high arsenic zone and apparent (anecdotal) thickness of superficial clays. The models indicate a dependence of high arsenic on a positive linear function of depth (downward oxygen transport?) and negative quadratic function of depth (proximity to deeper course sediments?).