ROSETTA: a computer program for estimating soil hydraulic parameters with hierarchical pedotransfer functions

Soil hydraulic properties are necessary for many studies of water and solut e transport but often cannot be measured because of practical and/or financ ial constraints. We describe a computer program, ROSETTA, which implements five hierarchical pedotransfer functions (PTFs) for the estimation of water retention, and the saturated and unsaturated hydraulic conductivity. The h ierarchy in PTFs allows the estimation of van Genuchten water retention par ameters and the saturated hydraulic conductivity using limited (textural cl asses only) to more extended (texture, bulk density, and one or two water r etention points) input data. ROSETTA is based on neural network analyses co mbined with the bootstrap method, thus allowing the program to provide unce rtainty estimates of the predicted hydraulic parameters. The general perfor mance Of ROSETTA was characterized with coefficients of determination, and root mean square errors (RMSEs). The RMSE values decreased from 0.078 to 0. 044 cm(3) cm(-3) for water retention when more predictors were used. The RM SE for the saturated conductivity similarly decreased from 0.739 to 0.647 ( dimensionless log(10) units). The RMSE values for unsaturated conductivity ranged between 0.79 and 1.06, depending on whether measured or estimated re tention parameters were used as predictors. Calculated mean errors showed t hat the PTFs underestimated water retention and the unsaturated hydraulic c onductivity at relatively high suctions. ROSETTA's uncertainty estimates ca n be used as an indication of model reliability when no hydraulic data are available. The ROSETTA program comes with a graphical user interface that a llows user-friendly access to the PTFs, and can be downloaded from the US S alinity Laboratory website: http://www.ussl.ars.usda.gov/. (C) 2001 Elsevie r Science B.V. All rights reserved.