Using Deterministic and Geostatistical Techniques to Estimate Soil Salinity at the Sub-Basin Scale and the Field Scale

Abstract

Two different techniques are evaluated in this study to estimate soil salinity in the lower Arkansas River Valley area in Colorado for both the sub-basin and field scales. Inverse Distance Weight (IDW) and Ordinary Kriging (OK) are evaluated as deterministic and geostatistical techniques respectively. Deterministic techniques depend on the assumption that the interpolating surface should be influenced mostly by the nearby points and less by the more distant points. Kriging techniques rely on the notion of autocorrelation and assume the data comes from a stationary stochastic process. The objectives of this study are: 1) compare the performance of the deterministic versus the geostatistical kriging techniques on the sub-basin and the field scales; 2) evaluate the effect of sampling density on the accuracy of both the deterministic and the geostatistical kriging techniques; 3) evaluate the effect of sampling distribution on both techniques; and 4) evaluate the effect of the existence of autocorrelation among data for both techniques. Different data sets for both field scale and sub-basin scale were collected in the study area where soil salinity impacts the crop productivity. Several data sets collected at the field scale and the sub- basin scale in the downstream area were evaluated. These data sets represent different sampling densities and spacing, different data distributions, and different amount of autocorrelation among the data. The results of this study indicate that there is no significant difference in the performance of both deterministic and geostatistical techniques at the field scale. However, the performance of the deterministic technique was significantly better than the geostatistical technique for the sub- basin scale (due to the lack of autocorrelation at this scale). The data distribution has no significant role on the performance of both techniques.