Improving the Prediction Accuracy of Groundwater Salinity Mapping Using Indicator Kriging Method

The saline groundwater irrigation is an important problem in the arid and semiarid region because it can cause soil salinization and reduce crop productivity. The accurate spatial distribution of groundwater salinity can be helpful to managers and decision makers. In this study, the mapping of salinity risk through groundwater electrical conductivity (EC) irrigation was performed based on data collected from 88 wells in the Lower-Cheliff plain (Algeria). The EC data showed a normal distribution based on elementary statistics. The EC classified using Riverside method point out a high risk of groundwater salinity (Class C3) or very high risk (Class C4) for soil salinization. The EC estimated by ordinary kriging method (OK) revealed on one hand, an underprediction of a high value, on the other hand, an overprediction of low value. The methodology of nonparametric and nonlinear of indicator kriging (IK) was performed by three thresholds: EC>2.25, EC>3, and EC>5dS/m. The map has been obtained from the combination of the local conditional cumulative distribution function (CCDF). The interpolated map by IK indicates the same overall spatial distribution of salinity with the one obtained by OK, enlightening differences in the shape and size of the area. The comparison between the groundwater EC estimated by OK and the one using IK demonstrates that IK has a better spatial prediction of salinity in terms of area and uncertainty. The groundwater salinity map was improved and accurately predicted by IK interpolation method.