Comparison of Multiple Linear Regression and Regression Tree for Prediction of Saturated Hydraulic Conductivity and Macroscopic Capillary Length (α*)

Ruhollah Taghizadeh-Mehrjardi, Sirous Dehghani, Ehsan Sahebjalal


The saturated hydraulic conductivity and macroscopic capillary length of the upper part of the soil are two important soil properties which control water infiltration and help us to create different models for predicting soil water status. Measurement of these parameters is tedious, time consuming and expensive so to overcome this problem, indirect methods such as Pedotransfer functions (PTFs) have been used for determining such parameters by using easily measurable input data. The objective of this study was, comparing the performance of multiple linear regression and regression tree for estimating saturated hydraulic conductivity and inverse of macroscopic capillary length parameter (a*). Therefore, saturated hydraulic conductivity) Kfs (and macroscopic capillary length (α*) in 60 study points of Azadegan plain in Shahrekord were measured using single ring and multiple constant head method. Also, some of the readily available soil data of two first pedogenic horizons of the soils were obtained. Then the multiple linear regression and regression tree were used to derive PTFs. The accuracy and reliability of both methods were evaluated using root mean square error (RMSE), mean error (ME), relative error (RE) and Pearson correlation coefficient (r). The results indicated that the performance of regression tree was better than multiple linear regression. Furthermore, by comparing results it is obvious that bulk density, geometric mean and weight mean of peds diameter had the most effects on saturated hydraulic conductivity and macroscopic capillary length


infiltration, inverse of macroscopic capillary length, pedotransfer functions, regression

Full Text: