|
|||
|
Application of indirect methods for prediction of soil hydraulic properties are widely pursued to provide for their high demand as input to numerical simulation models. Specifically, pedotransfer functions to predict soil water retention have been widely developed. However, few datasets that include unsaturated hydraulic conductivity data are available. Moreover, all current soil hydraulic datasets employ a wide variety of measurement techniques. It is expected that prediction of soil water retention and unsaturated hydraulic conductivity data from basic soil properties can be improved if hydraulic data are determined using a single measurement method that is consistently applied to all soil samples. Such a unique dataset consisting of about 310 soil water retention and unsaturated hydraulic conductivity functions is presented, with hydraulic data obtained using the multi-step outflow method. Using neural network analysis, soil water retention and hydraulic conductivity characteristics are predicted from basic soil properties, i.e., sand, silt, and clay content, bulk density, saturated water content, and saturated hydraulic conductivity. The results show that the prediction accuracy increased if the training data sets consists of soil hydraulic data that are obtained using the same measurement technique. Specifically, the prediction errors of water content were about 3 to 4 %. Unsaturated hydraulic conductivity predictions improved significantly when the performance-based algorithm was used, minimizing residuals of soil hydraulic data rather than hydraulic parameters. The error for predicted values of water content and unsaturated hydraulic conductivity were reduced by about 50 %, when compared to predicted hydraulic functions using program Rosetta. Reference:
Minasny, B., Hopmans, J.W., Harter,
|
|||
|
|
|||