Geostatistical methods for detection of outliers in groundwater quality spatial fields

Abstract

Detection of discordant groundwater quality observations is of primary importance. Because of the large amount of data to be analysed, methods of plausibility analysis need to be simple and not too time-consuming. Two geostatistical methods were studied, which allow the identification of outlying data points based on the pattern present in the remaining data. The two methods used, point kriging and the IRF- k method, differ in severity of assumptions. Both approaches were tested on concentration of chloride and on total hardness in groundwater and performed satisfactorily. They lend themselves well also to operational applications.