Comparison and application of random forest, chaid and geostatistics models in predicting soil saturated hydraulic conductivity

Soil saturated hydraulic conductivity (Ks) is one of the important factors involved in water, soil, and agricultural sciences. Ks measurement is important for solute and water movement modeling and, in turn, is costly and time consuming. It is also impractical to spatially and temporarily measure the Ks in large scale studies. This parameter can be estimated using early soil parameters. The present research was conducted in order to predict the hydraulic conductivity of soil saturation using random forest, chaid and geostatistical methods in Hirmand city. For this purpose, 130 soil samples were collected from the surface (0-30 cm) and transferred to the laboratory for testing and analysis. In the laboratory, the parameters of hydraulic conductivity of soil saturation, soil texture, organic carbon, acidity, electrical conductivity and Percentage of lime were measured. It was then estimated using measurement parameters and using CHAID, Random Forest and geostatistics models. 20 different patterns of the combination of soil moisture parameters were considered as input to the Chaid model and random forest. Among the different combinations, the best combination was selected based on lower MAE and higher R. The results showed that the random forest model with the highest R2 (0.98) and the lowest MAE (0.0019) is the best model for predicting the saturated hydraulic conductivity of soil in Hirmand region. The results of zoning showed that the amount of saturated hydraulic conductivity of the soil was higher in the west, center and northeast than in other places.