Application of Digital Soil Mapping in Soil Particle Size Zonation and Estimation of Saturated Soil Hydraulic Conductivity for Optimal Management of Watersheds (Case Study: Damghanrood Watershed)

The soil particle size distribution is one of the most important of soil properties that effect on the soil hydraulic properties, including saturated hydraulic conductivity. Therefore, accurate knowledge of spatial distributon of soil particle size in the watershed is very effective on the optimal management of the watershed. In this study, the spatial distribution of sand, silt and clay particles were predicted in the Damghanrood watershed with a spatial resolution of 30 m at the depths of 0-30, 30-60 cm. For this purpose, 110 soil sampling points were determined using conditional Latin hypercube sampling (cLHS) method. Environmental variables were extracted from Landsat 8 Operational Land Imager (OLI) satellite and digital elevation model (DEM). The random forest (RF) model was used for determined the relationship between soil particles and environmental variables. The results showed that the coefficient of determination (R2) of the RF model at a depth of 0-30 cm for clay, sand and silt particles with a range of 0.6, 0.52 and 0.71, respectively, and at a depth of 30-60 cm, respectively. It was obtained with 0.69, 0.67 and 0.49. In the surface layer, the auxiliary variables extracted from the remote sensing data and in the deep layer, the variables extracted from the most part were related to the soil particle data. The results showed that the coefficient of determination (R2) of the RF model for prediction clay, sand and silt fractions at depth of 0-30 cm was of 0.6, 0.52 and 0.71, respectively, and at a depth of 30-60 cm, for prediction of these fraction the R2 value was 0.69, 0.67 and 0.49, respectively. In the surface layer, the auxiliary variables extracted from the remote sensing data were more important variables for prediction of particle fraction but in deep layer, the terrain attributes were the most important variables in prediction of particle size fractions. The values of saturated hydraulic conductivity (Ks) estimated using pedotransfer functions varied between 0.08 to 1 m / day. The lowest amount of Ks was observed in lands with rock outcrops and marl soils. The results showed that the spatial distribution of Ks derived from sand and clay data was well overly with the reality of the region. So that the lowest values of Ks were observed in areas with rock outcrops and in marly soils.