Spatial estimation of surface soil texture using remote sensing data.

Abstract

Soil texture is a principal soil property, affecting physical characteristics and behavior of the soil, such as soil water retention, nutrient holding capacity and susceptibility to erosion. Areas with high intensity agriculture need information about the soil types, thus there is a need for an effective method of soil mapping. This study aimed at exploring the potential of remote sensing data in estimating surface soil texture. Fifteen surface soil samples were collected from the study area. Soil samples were analyzed using hydrometer method and the soil percentage contents were plotted on USDA textural triangle to determine soil texture types. Areas with vegetation and built-up were excluded from analysis of reflectance values by assigning their pixel values with no data. A composite bare soil image was obtained Multiple linear regression (MLR) analysis technique was used to relate soil variables and spectral data from Sentinel 2 images. Then MLR equations were used to estimate soil texture for the study area by using reflectance values from band2, band3, band4, band8 and band11. Prediction of clay content showed the highest correlation with an R value of 0.777. silt content prediction showed an R of 0.608, while sand particles showed the least value of R as 0.593. In order to produce surface soil texture map, limiting texture classes for each soil type were set using a raster calculator, a map algebra tool in ArcGIS software. Kriging was used in interpolation for the areas that were assigned no data during the construction of a bare soil image. It was observed that that clay had the highest pixel count between 40% to 50% with a mean value of 47%. Sand had the highest pixel count between 30% and 40% with a mean value of 38.3% While for silt, the highest pixel count was between 10% to 15% with a mean of 13.7% The results indicated that spectral reflectance data can be used in estimating soil texture if a well bare soil composite image is produced by eliminating all other features like built up areas and vegetation.