Direct Data Manipulation for Local Decision Analysis as Applied to the Problem of Arsenic in Drinking Water from Tube Wells in Bangladesh

Abstract

A wide variety of tools are available, both parametric and nonparametric, for analyzing spatial data. However, it is not always clear how to translate statistical inferences into decision recommendations. This article explores the possibilities of estimating the effects of decision options using very direct manipulation of data, bypassing formal statistical analysis. We illustrate with the application that motivated this research, a study of arsenic in drinking water in nearly 5000 wells in a small area in rural Bangladesh. We estimate the potential benefits of two possible remedial actions: (1) recommendations that people switch to nearby wells with lower arsenic levels; and (2) drilling new community wells. We use simple nonparametric clustering methods and estimate uncertainties using cross-validation.

Effectiveness of Different Approaches to Arsenic Mitigation over 18 Years in Araihazar, Bangladesh: Implications for National Policy

“…14 Target depths that are likely to be low in As based on village-level BAMWSP data can be recombined to identify larger areas where households are likely to be able to lower exposure by installing an intermediate well. The subset of 1558 villages identified by the search algorithm 27 covers as many as 691 unions, almost half of the total of 1633 unions encompassing the 11,174 BAMWSP villages with a minimum of 20 wells and at least >20% high As wells. For villages within each of these 691 unions and possibly others, an intermediate aquifer low in As would probably be identified by a new blanket testing campaign because of the installation of new and somewhat deeper wells.…”

“…9,22−26 A search algorithm developed for Araihazar was applied to the countrywide BAMWSP data to determine where these measurements indicate an intermediate aquifer that is systematically low in As, along with an estimate of the reliability of this assessment using an approximate Bayesian approach. 27 When applied to 11,173 villages in the BAMWSP data set (see the Supporting Information) with at least 20% of wells containing >50 μg/L As and a minimum of 20 wells, the algorithm indicates a target depth in the 45–90 m depth range with an estimated probability of at least 0.8 that it is correct in a subset of 1558 villages (Figure 2b). Many of these villages are located within the most affected regions of the country (Figure 2a).…”

Evidence for Elevated Levels of Arsenic in Public Wells of Bangladesh Due To Improper Installation

“…The 90‐m threshold distinguishes private and public wells but has no particular geological significance. The vast majority of wells >90 m in Araihazar are low in As but in some parts of the study area, Pleistocene low‐As aquifers can be reached at considerably shallower depths (Gelman et al ). In other parts of the country, groundwater that is consistently low in As is reached only at depths that exceed 150 m (Ravenscroft et al ).…”

Design of a rural water provision system to decrease arsenic exposure in Bangladesh

“…Many believe that the most effective way to reduce arsenic exposure in Bangladesh is to encourage villagers to switch to arsenic-free shallow tubewells, or deep tubewells (31,80). Due to Bangladesh's complex geochemistry, tubewells 10 meters apart can deliver water of vastly different arsenic concentrations because they tap different aquifers.…”