Soil Texture Classes Mapping as a Way in Soil Sustainable Management

Abstract

The spatial distribution map of soil classes is highly useful in planning sustainable agriculture. One of the most important soil characteristics is soil texture, and providing a map of it can help identify lands suitable for various agricultural products. For growers, the soil texture class map is more important than the map of the percentage of soil texture particles. The aim of this study was to compare the efficiency of two models—Decision Tree (DT) and Artificial Neural Network (ANN)—in forecasting soil texture class in the Bardeh region of Chaharmahal and Bakhtiari province, Iran. In this study, 96 soil profiles were excavated, described, and sampled from surface horizons across an area of 6,875 hectares, with the locations recorded using GPS. The percentages of clay, sand, and silt were measured, and the soil texture classes were determined. Results showed that the RMSE, Kappa Index, R square, and overall accuracy for estimating soil texture class based on test data were 0.6, 0.09, 0.24, and 0.59 for the Neural Network model, and 0.76, 0.75, 0.6, and 0.41 for the Decision Tree model, respectively. According to the results, the Decision Tree model was the better predictor for providing soil texture class maps. The results suggest that digital mapping of soil texture class could be an effective method for soil resource assessment.