The Use of Cokriging Algorithm for Arsenic Mapping in Groundwater Systems

Abstract

Accurate mapping of the spatial distribution of arsenic in groundwater is an important but equally difficult task to complete due to a number of uncertainties. Classical univariate interpolation algorithms could sometimes be insufficient to capture high concentration and high gradient areas. Under these circumstances, the use of an auxiliary parameter could provide better estimates of arsenic distribution. Based on this premise, arsenic cokriging with a correlated parameter can improve the performance of interpolation and can enhance the quality of predictions. In order to test this hypothesis, a water quality dataset from an arsenic containing aquifer in Simav Plain, Turkey is used to develop arsenic distribution maps. Arsenic is cokriged with correlated parameters such as manganese, iron and dissolved oxygen; and the results are compared with univariate interpolation algorithms such as ordinary kriging and inverse distance weighing. The comparisons were performed with cross validation at sampling locations and assessed based on mean and root mean squared errors. The results revealed that maps developed using arsenic cokriging with iron have given the smallest error value and have shown closest fit to the extreme values in the dataset. Accordingly, arsenic cokriging with iron is believed to be a promising approach in mapping arsenic distributions in groundwater.