Spatial Prediction of Nitrate Concentration in Drinking Water Using pH as Auxiliary Co-kriging Variable

Abstract

Lognormal ordinary cokriging (LnOCK) with auxiliary variables can sometimes improve estimates for a less densely sampled primary variable. The objective of this study was to compare lognormal ordinary cokriging (LnOCK) with lognormal ordinary kriging (LnOK) and lognormal inverse distance weighting (LnIDW) for the spatial prediction of NO3-N in drinking water using pH as an auxiliary variable in LnOCK. We collected 345 drinking water samples from all villages in the Bijar and Qorveh Plains of western Iran and analyzed them for NO3-N and pH. The NO3-N concentration exceeded the US Environmental Protection Agency maximum contaminant level (EPA-MCL) guide value of 10mg l-1 NO3-N in more than 89% of the samples. The distribution of NO3-N was highly skewed. In terms of mean error (ME) and root mean squared error (RMSE) LnIDW performed much better than LnOK for NO3-N. However, LnIDW was consistently less effective than LnOCK using pH as auxiliary variable.