Measuring unsaturated hydraulic characteristic parameters, such as capillary soil water, saturated soil moisture, field moisture, saturated conductivity etc., is not only expensive but also accuracy uncontrolled. Soil hydraulic characteristic parameters can be estimated accurately using the Pedo-transfer Functions (PTFs) depending on soil fundamental parameters which are easy to get such as grading of soil, organic content and bulk density. Pedo-transfer functions (PTFs) have been developed by two methods in this paper with data measured at keerqin sandy land. (1) PTFs have been developed by using multivariate step wise regression analysis, nonlinear analysis and principle factor analysis. (2) Back-propagation neural network (BP) was used in simulation and prediction of PTFs. We use theory of difference to evaluate PTFs developed whether regression method or BP model. Accuracy of PTFs is evaluated by the root of the mean squared difference (RMSD). Last, we compared measuring values with estimation values of regression method and BP model. The evaluating results indicate that PTFs developed by regression method or BP model satisfy to use to the education, research and production practice for Keerqin sandy land.
|