The PTF models stored on this home page represent both parametric PTF models and point PTF models. The models are based on a Danish soil data set representing 1618 horizon observations. The point PTF models predict the total porosity, and the water content at 10hPa, 33hPa, 100hPa, 1000 hPa and at 15000hPa. The parametric PTF models predict the van Genuchten Mualem retention parameters. The PTF models are developed by using neural network analysis and the results of this optimisation isare weight matrices that can be used by the Matlab programs also stored on this page. To run the models, you need to have the Matlab program package including the Neural Network toolbox installed on your computer.

Nine weight matrices areis stored on the page, each having 50 sets of parameters predicting the van Genuchten parameters. The weight matrices are named and ending on P1.bin, P2.bin….P9.bin representing different models using different input parameters. The files are stored as binary files that can be opened by the programmes. The programs predict the average results based on the 50 different sets of parameters. If you use the weight matrices in other programmes, remember to rescale the alpha and n parameters (see on line 86-88 , and line 112-114 in the program Wrcpar_ptf.M).

Eigth weight matrices isare stored on the page, each having 50 sets of parameters predicting the porosity and the water contents at 10hPa, 33hPa, 100hPa, 1000 hPa and at 15000hPa. The weight matrices are named and ending on M1.bin, M2.bin…M8.bin). The models are using different input parameters.

To run the programmes, I have given a test data set which can be used for testing the programmes. On the page, I have included the output files using the different models. The format of both the input- and file and the output file is text files.

The format of the input data to the matrices is described in the two programmes Parptf_221104.M (L58 –L69 for P1, L82-90 for P2 and etc.) tofor predicting the van Genuchten parameters. A and in Pointptf_221104.M L60-69 for M1, M2… to predict the water contents at the different tensions.

The texture- and organic matter data hasve to be given as percentage [%], horizon as either 1 for top horizons and 2 for sub horizons,bulk density as [g/cm3], water contents as decimal (exampelle theta at 10hPa= 0.35). The data in the testdata.txt file follows the request of the PTF models.

We are working on setting up a user friendly soft ware to make it easy to run the models. We are also working on expanding the models including predictions of the Mualem Ko and the L parameters and on makeing the models using not only Danish soil texture classes but also USDA texture classification.

The results of the calibration and the validation of the PTF models isare printed in the paper “Point and Parameter Pedotransfer Functions for Water Retention Predictions for Danish Soils”Christen D. Børgesen, Marcel G. Schaap accepted Geoderma, October 2004.