Spatial Variation of Mineral Particles of the Soil using Remote Sensing Data and Geostatistics to the Soil Texture Interpolation

Soil texture is the average size of soil particles which depends on the relative proportion of sand, silt and clay contents. Soil texture is one of the most important features used by soil and environmental scientists to describe soils. Soil texture directly affects the soil porosity, which in turn, determines water-retention and flow characteristics, nutrient-holding capacity, internal drainage, sorption characteristics and long-term soil fertility. High-resolution soil maps are essential for land-use planning and other activities related to forestry, agriculture and environment protection. Given the soil texture roles in controlling the soil functions, it is necessary to understand the spatial distribution of this feature in regional scale. As soil texture is a staticproperty, regional scale soil texture maps can thus help environmental scientists to predict different soil-related processes. The objective of this study was to develop a soil textural class map using Terra satellite MODIS sensor images. To achieve this goal, the digital elevation model SRTM radar of the studied area for soil samples from different altitudes and slopes was prepared in foursen consecutive 30 meters time frame. The nearest neighbor method with an error of less than 0.5 pixels was used and the elevation layers were mosaicked and transmitted to the UTM ZON-38 coordinate system and GIS Ready Became. The normalized difference vegetation index of bands 1 and 2 of the matrix was obtained to isolate the reflection of the electromagnetic spectrum of vegetation and soil. This final mosaicked digital elevation model was then divided into different altitudes to accurately evaluate the surface texture. The 60 spatial points were selected to estimate the texture of surface soil in thestudied area with systematic randomized sampling. In the current study, soil texture was determined forthe air-dried samplesusing hydrometer. The SWIR bands of MODIS with resolution of 500 meters were selected for sampling dates. After corrections, DN values of the bands for sampling points were extracted. The Pearson correlation coefficient and step wise regression techniques were used to establish proper relationships between the DN values of the SWIR bands and the soil particles. Kriging and cokriging methods were also employed to create a spatially distributed map of the soil textural classes. The results showed that there is a close correlation between the SWIR bands of the terra satellite and the MODIS sensor with band 3, and using this auxiliary variable significantly reduces the estimation error. The best model for fitting semivariogram for clay, silt and sand contents were spherical, spherical and exponential models and the best fitting Cross-semivariogram for clay, silt and sand contents were spherical, exponential and exponential models, respectively. The highest and lowest error estimation was, respectively, related to sand and clay content. The maximum and minimum decrease of estimation error by the auxiliary variables was found for sand and clay content, respectively. The nugget/sill ratio of the kriging semivariograms was greater than 25%for sand and claycontentand lower than 25%for sand and silt content. This indicates that sand and silt contents had a strong spatial dependency, and clay content hada moderate spatial dependency. These ratios for cokriging cross-semivariograms of sand, silt and clay contentsware less than 25%. The interpolation of estimated soil texture was also determined using the cokriging method with 70% of the soil texture measured in the laboratory. Our results indicated thatcokriging method estimated the soil particles more accurately as compared with linear multi-variable stepwise regression and kriging methods. Application of cokriging method also reduces the number of sampling points and the estimation error of soil texture zoning. Therefore, cokriging method seems to be better suited in impact assessments for data-scarceregions such as Iran.