Prediction of saturated hydraulic conductivity from soil physical properties under different forest vegetation using multivariate analysis techniques

Abstract

Saturated hydraulic conductivity of five types of forest soils at two depths was assessed from the physical properties using multivariate analysis techniques. All the soil variables had very strong correlations with each other. Multiple regression models linking original data set demonstrated that clay and bulk density predicted 81.7% of total variation in saturated hydraulic conductivity. The principal component analysis (PCA) involving several soil parameters explained 92% of the variance. The regressive model for saturated hydraulic conductivity using minimum data set (MDS) from PCA such as sand and bulk density accounted for 81.2% of the variability. It was almost competitive with multiple regression equations in assessing the saturated hydraulic conductivity, but less predictive than PCA because of the involvement of several contributory soil parameters. All these statistical approaches may thus provide an alternative way of measuring the saturated hydraulic conductivity of forest soils indirectly from the measured values of soil physical properties.