Bayesian analysis of groundwater quality in a semi-arid coastal county of south Texas

Abstract

Bayesian frameworks for comparing water quality information to a pre-specified standard or goal and comparing water quality characteristics among two different entities are presented and illustrated using chloride and total dissolved solids (TDS) measurements obtained in the shallower Chicot and the deeper Evangeline formations of the Gulf coast aquifer underlying Refugio County, TX. The Bayesian approach seeks to present evidence in favor of the competing hypotheses which are weighed equally and unlike classical statistics do not make a decision in favor of one hypothesis. When comparing water quality information to a specified goal, the Bayesian approach addresses the more practical question—given all the information, what is the probability of meeting the goal? Similarly, when comparing the water quality between two entities, the approach simply emphasizes the nature and extent of differences and as such is better suited for evaluative studies. Bayesian analysis indicated that average chloride concentrations in the Evangeline formation was 1.65 times the concentrations in the Chicot formation while the corresponding TDS concentration ratio was close to unity. The probability of identifying water with TDS ≤1,000 g/m³ was extremely low, especially in the more prolific Evangeline formation. The probability of groundwater supplies with mean chloride concentrations ≤500 g/m³ was relatively high in the Chicot formation but very low in the Evangeline formation indicating the possible need for blending groundwater with other sources to meet municipal water quality goals.