부산대학교 도서관

The spatial distribution of arsenic contamination of groundwater resources is important in estimating the health risks of millions of people worldwide, particularly in the densely populated river deltas of Vietnam. The predictive Arsenic (As) probability at the village - level in shallow ground waters in the Red River Delta was identified using 512 rural groundwater samples collected from a large – scale survey during the period May 2005 – Jan 2007 from residential small – scale tube wells between 5 - 135 m depth. Nine principal hydro-chemical parameters, pH, NO3-, NH4+, Fe, PO43-, DOC, Eh, SO42- and Mn contributing significantly to the model were applied to predict the probability of occurrence of As concentrations in different threshold limits in groundwater over the study area. The As prediction model was finally validated upon an internal set of 102 samples. In this study, the regression analysis confirmed that there is a good correlation of some environmental explanatory variables with binary – coded As concentrations data in ground waters. This modeling result is broadly consistent with the already known findings of previous studies. We estimate that Hanoi and Hanam are the highest potential areas of As hazard exceeding 5 µg/L, indicating the requirement for development and implementation of policy control measures. The statistical model based in stepwise logistic regression for 5µg/L correctly explained 71.6% of the variation in the validation set as contaminated – and uncontaminated regions (predicted probability cut – off threshold of 0.24) and shows good results for the prediction of elevated As concentrations exceeding 5µg/L. The developed logistic regression model indicates that three parameters (pH, NH4, Fe) can accurately predict probability of arsenic concentration ≥ 5 µg/L in groundwater. These parameters provide an explanation for release of arsenic by reductive dissolution of As – rich FeOOH in NH4+ containing groundwater. The results can provide a more realistic description of the distribution of As if it captures the large - scale variation of As in the study area. The predicted probability of As hazard in this region is associated with the known spatial distribution of arsenic contamination, and further indicated elevated risks at sampled sites where groundwater studies are not practical and economical or no arsenic concentration data exists. Thus, the statistical risk model can be used as a tool to assist sampling in adequate villages showing high risk. However, arsenic sampling is still needed and required in arsenic – assessment stages in other areas and the need for long – term monitoring and maintenance is not precluded.