Using Geostatistics to Predict the Spatial Variability of Soil Salinity from Electromagnetic Induction Measurements

Soil salinity is a major threat to crop and soil productivity in arid and semiarid regions of the world. This is often due to the nature of the underlying parent materials as well as anthropogenic factors such as the quality of the applied irrigation waters; resulting in a spatially variable crop development and production. On a 21,000-ha study site, apparent electrical conductivity (ECa) was measured using the EM38DD sensor in 37 fields to quantify the spatial variation in soil salinity. Based on the ECa values, ground truth samples were also collected to calibrate the easily measured ECa with the expensive to measure electrical conductivity of the saturation paste extract (ECe). The main objective of this study was to describe the extent of spatial distribution of soil salinity on the landscape. The geostatistical analyst tool of ArcGIS software was used to model semivariograms and to predict soil salinity at unknown locations by ordinary kriging. Probability maps of ECe exceeding the saline threshold of 4dSm-1 were produced using indicator kriging. Cross validation statistics was performed to evaluate the accuracy of the ECa prediction maps. The prediction standard error maps were also compared