Geostatistically based management of arsenic contaminated ground water in shallow wells of Bangladesh

This paper investigates the effectiveness of geostatistical approaches, specifically ordinary kriging, for regional management of arsenic contaminated shallow ground water in Bangladesh. The arsenic database for reference comprised the nation-wide survey (of 3534 drinking wells) completed in 1999 by the British Geological Survey (BGS) in collaboration with the Department of Public Health Engineering (DPHE) of Bangladesh. A Monte Carlo (MC) framework was devised for selection of randomly-sampled networks of wells from this reference database. Each randomly sampled network was assumed an equi-probable exploratory field campaign designed commensurably with the requirements of rapidity and cost-effectiveness in a rural setting. In general, the kriging method was found to underestimate the arsenic concentration at non-sampled locations. This underestimation exceeded the safe limits at the Holocene region of Southcentral Bangladesh. The probability of successful prediction of safe wells for this region was found to be 72% (WHO safe limit – 10 ppb) and 78% (Bangladesh safe limit – 50 ppb). For the Pleistocene Northwestern region of Bangladesh, the safe well prediction probability was in the ranges of 90%–97%. The relatively more contaminated Holocene region in Southcentral Bangladesh, on other hand, was found more amenable to accurate geostatistical prediction of unsafe wells. Findings from this study exemplify that, while mainstream geostatistical approaches (e.g., ordinary kriging) may not provide the most accurate prediction of mean arsenic concentration at non-sampled locations, they can delineate an approximate strategy for management of arsenic contaminated shallow ground water if applied carefully. The kriging methodology is applied to a test case in Bangladesh; the approach, however, is general and is expected to have application in rural settings for other developing countries where arsenic contamination of ground water is also widespread (e.g., parts of India, Vietnam, Taiwan and Mexico). Copyright Springer Science + Business Media B.V. 2007