Water Quality

Integration of Conventional Statistical and Fuzzy Theories for Improved Characterization of Contaminated Hydro-systems

[Note: This area of research was closed in 2009; However, as an opportunity arises, this area is pursued to build synergy with other SASWE driven research foci]
Since the discovery of large-scale arsenic contamination of groundwater in Bangladesh more than a decade ago, studies related to its spatial characterization at the regional scale have relied on geo-statistical approaches and the classical notion of linear stochastic dynamics (Figure 1). Such approaches (e.g., kriging methods), however, solve only the pattern completion problem, and not the pattern recognition problem. As a result, the ‘field’ of arsenic estimated in this fashion from a limited amount of information is subject to greater uncertainty at non-sampled locations due to measurement and sampling errors of in-situ arsenic tests. In this research we argue that it is no longer defensible to continue to use purely geo-statistical approaches of pattern filling while evidence accumulates on the deterministic geochemical factors that dictate the spatial variability of arsenic contamination in groundwater. Therefore, in order to obtain a more accurate estimate of the arsenic field, the geo-statistical approach should be complemented with pattern recognition approach (e.g., non-linear deterministic and chaotic dynamic theory). We discuss the potential opportunities that this combined paradigmatic approach (pattern filling + pattern recognition) could offer towards greater cost-effectiveness in locating clusters of safe shallow wells than that would otherwise be possible through a purely geo-statistical (pattern filling) or a purely non-linear dynamic (pattern recognition) framework. Our research however also emphasizes that integration of these complementary paradigms should not be construed as a replacement of the conventional geo-statistical pattern-filling techniques but only to minimize the limitations of such techniques.


Figure 1. A field of arsenic contamination produced by the geo-statistical technique of kriging (after BGS-DPHE, 2001).

A particular focus of this research is to develop low-cost and non-structural simulation methods for prevention of arsenic contaminated water supply in a rural setting. We had earlier investigated an approach that aims at identification of safe/unsafe wells. The proposal is founded on the premise that the near impossibility of testing every single shallow well in a rural setting requires a simulation methodology that can accurately characterize a well as being safe/unsafe without the need for extensive and expensive in-situ sampling tests. Such a method can then act as a fast-running and inexpensive proxy to the time-consuming field campaigns and save considerable testing resources by judiciously directing them to those wells pre-determined by the simulation approach to have a high likelihood of being unsafe. Furthermore, by flagging a safer cluster of wells functional (and an unsafe cluster of wells as non-operational), villagers are expected to find this approach socially more convenient than the expensive house-hold treatment options currently available in rural areas of developing countries. We are currently investigating the effectiveness of our scheme in Northwestern Bangladesh (Figure 2) on the basis of Field kits and laboratory measurements of arsenic concentrations.


Figure 2. Study-region for assessment of proposed low-cost simulation-based scheme for prevention of arsenic contaminated water supply.