Soil hydraulic characteristics, especially the soil water retention curve and hydraulic conductivity, are essential for many agricultural, environmental, and engineering applications. Their measurement is time-consuming and thus costly. Hence, many researchers focused on methods enabling their indirect estimation. In this paper, Wösten’s continuous pedotransfer functions were applied to the data from a selected locality in the Czech Republic, Tišice. The available data set related to this locality consists of 140 measured soil water retention curves, and the information about the soil texture, bulk density ρ<sub>d</sub>, and organic matter content determined at the same time. Own continuous pedotransfer functions were derived, following the methodology used in continuous pedotransfer functions. Two types of fitting, 4-parameters and 3-parameters, were tested. In 4-parameter fitting, all parameters of the van Genuchten’s equation, θ<sub>s</sub>, θ<sub>r</sub>, α, n, were optimized; in 3-parameter fitting, only three parameters, θ<sub>r</sub>, α, n, were optimised while the measured value of θ<sub>s</sub> was set as constant. Based on the results, it can be concluded that the general equations of Wösten’s pedotransfer functions are not very suitable to estimate the soil water retention curves for the locality Tišice in the Czech Republic. However, the parameters of the same Wösten’s equations, which were calculated only from the data for each particular locality, performed much better. The estimates can be improved if the value for the saturated soil water content θ<sub>s</sub> is known, applied and not optimised (the case of 3-parameter fitting). It can be advantageous to estimate SWRC for a locality with no data available, using PTFs and the available basic soil properties. In addition, to measure some retention curves and/or some their parameters, like θ<sub>s</sub>, can improve the accuracy of the SWRC estimation.
Application of neural network pedotransfer functions to calculate soil water retention curve
