Feasibility of decision tree application (M5 model) for determining soil moisture characteristic curve from easily available soil parameters

One of the major issues making difficult to model soil water and crop relationship is difficulty of determining soil hydraulic characteristics as soil moisture and unsaturated hydraulic conductivity curves. Because these characteristics are affected by chemical and physical agents. In this study، some readily available soil parameters such as sand، clay and silt percentage، moisture versus matric suction، bulk density، organic matter، saturated electrical conductivity (EC) and sodium adsorption ratio (SAR) were measured. Also soil hydraulic parameters are determined by optimization method. Finally، soil unsaturated moisture parameter is formulated by M5 decision tree method as functions of matric suction، sand، clay and silt percentage، bulk density، organic material، sodium adsorption ratio and saturated electrical conductivity. Also computed moisture data is compared with measured soil moisture curve data. Next، the R2 Determination coefficient، root mean square error (RMSE) and mean of bios error (MBE) is determined between measured and decision tree model (M5) fitted moisture data for all soil textures and salinities. Result showed that besides matric potential data. The sand، silt and clay data have also special effect on precision of moisture determining with M5. But matric potential data has greater effect on valid moisture prediction comparing to observed data. Generally، results of modeling with decision tree model (M5) showed that although some physical and chemical soil data could be available and measurable but matric potential data is required as key input parameter for accuratly simulation moisture retention curve.