Bayesian Spatial Design of Optimal Deep Tube Well Locations in Matlab, Bangladesh

Citation

Warren, Joshua L.; Perez-Heydrich, Carolina; & Yunus, Mohammad (2013). Bayesian Spatial Design of Optimal Deep Tube Well Locations in Matlab, Bangladesh. Environmetrics, 24(6), 377-386. PMCID: PMC4018770

Abstract

We introduce a method for statistically identifying the optimal locations of deep tubewells (dtws) to be installed in Matlab, Bangladesh. Dtw installations serve to mitigate exposure to naturally occurring arsenic found at groundwater depths less than 200 meters, a serious environmental health threat for the population of Bangladesh. We introduce an objective function, which incorporates both arsenic level and nearest town population size, to identify optimal locations for dtw placement. Assuming complete knowledge of the arsenic surface, we then demonstrate how minimizing the objective function over a domain favors dtws placed in areas with high arsenic values and close to largely populated regions. Given only a partial realization of the arsenic surface over a domain, we use a Bayesian spatial statistical model to predict the full arsenic surface and estimate the optimal dtw locations. The uncertainty associated with these estimated locations is correctly characterized as well. The new method is applied to a dataset from a village in Matlab and the estimated optimal locations are analyzed along with their respective 95% credible regions.

URL

http://dx.doi.org/10.1002/env.2218

Reference Type

Journal Article

Year Published

2013

Journal Title

Environmetrics

Author(s)

Warren, Joshua L.
Perez-Heydrich, Carolina
Yunus, Mohammad

PMCID

PMC4018770