Assessing the capability of satellite data for soil mapping

Abstract

The purpose of this research is to evaluate the potential of using landsat ETM+ data for soil mapping. The study area is located in center of Iran and covers about 1300 ha. The database of studied area was created by introducing topographic map (1:25,000), soil map (1:20,000) and reports and satellite data. After pre-possessing stage, selection of the best informative bands was carried out using optimum index factor (OIF) calculation and principle component analysis (PCA). Results showed that bands TM1, TM4 and TM6-2 contain the highest information and the lowest redundancy. Besides the mentioned bands, TM5 and TM7 were considered for digital image classification. The images were classified using maximum likelihood classifier into seven mapping units. Separability of mapping units examined at 95% confidence level. Comparison of the prepared soil map from satellite data and ground truth showed a relatively high accuracy of 80%. Also, comparison of prepared soil map from satellite data and detailed soil maps prepared using conventional methods showed imagery data could increase the classification and interpretative purity percentage up to 50% and 85%, respectively. The results indicated high potential of imagery data for inventory and increasing the precision of existing soil maps. Therefore, incorporation of high-resolution satellite data for soil survey especially in arid and semi-arid regions is highly recommended.