Aquifer vulnerability assessment to heavy metals using ordinal logistic regression

A methodology using ordinal logistic regression is proposed to predict the probability of occurrence of heavy metals in ground water. The predicted probabilities are defined with reference to the background concentration and the maximum contaminant level. The model is able to predict the occurrence due to different influencing variables such as the land use, soil hydrologic group (SHG), and surface elevation. The methodology was applied to the Sumas-Blaine Aquifer located in Washington State to predict the occurrence of five heavy metals. The influencing variables considered were (1) SHG; (2) land use; (3) elevation; (4) clay content; (5) hydraulic conductivity; and (6) well depth. The predicted probabilities were in agreement with the observed probabilities under existing conditions. The results showed that aquifer vulnerability to each heavy metal was related to different sets of influencing variables. However, all heavy metals had a strong influence from land use and SHG. The model results also provided good insight into the influence of various hydrogeochemical factors and land uses on the presence of each heavy metal. A simple economic analysis was proposed and demonstrated to evaluate the cost effects of changing the land use on heavy metal occurrence.