|Van Genuchten, Martinus|
Submitted to: Vadose Zone Journal
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/9/2008
Publication Date: 1/5/2009
Publication URL: hdl.handle.net/10113/2885
Citation: Guber, A.K., Pachepsky, Y.A., Van Genuchten, M.T., Simunek, J., Jacques, D., Nemes, A., Nicholson, T., Cady, R.E. 2009. Multimodel Simulation of Water Flow in a Field Soil using Pedotransfer Functions . Vadose Zone Journal. 8:1-10. Interpretive Summary: Pedotransfer functions (PTFs) are routinely utilized to relate soil hydraulic parameters to readily available data on soil and sediment properties. The accuracy of PTFs outside of the development region is generally unknown, and this causes uncertainty in the accuracy of soil water flow simulations with hydraulic parameters generated with a specific PTF. The existence of several models with the unpredictable accuracy for a particular site is common in various environmental science fields. One solution, called multimodeling, is to assign some weights to the simulation results from different models, and to use these weights to combine the simulation results from different models. Our objective was to compare different methods of computing the model weights for simulation of the field soil water regime using pedotransfer functions (PTF). Experimental data on soil water contents and basic soil properties at 5 depths along a 6-m transect in a layered loamy soil were used to develop the input for PTFs, and to evaluate the multimodel prediction. We solved the Richards flow equation using the computer code HYDRUS-1D with parameter sets derived from PTFs from different regions, and then compared the six different methods of combining the simulation results. The similarity in simulations obtained with different models appeared to be the biggest difficulty in obtaining robust weights. The difficulty could be overcome by using the technique called regression with the singular value decomposition. The accuracy of multimodel simulations using this method was comparable to the accuracy of simulations with site-specific soil hydraulic parameters. Using monitoring data to derive site-specific model weights instead of site-specific hydraulic parameters may be a promising approach to improve the efficiency of soil water monitoring.
Technical Abstract: Calibration of variably saturated flow models with field monitoring data is complicated by the strongly nonlinear dependency of the unsaturated flow parameters on the water content. Combining predictions using various independent models, often called multimodel prediction, is becoming a popular modeling technique. The objectives of this study were to (a) compare different methods of multimodel simulation of the field soil water regime using pedotransfer functions (PTF), and (b) investigate whether the calibration of a flow model with field data can be replaced by multimodel simulations. We solved the Richards flow equation using HYDRUS-1D with parameter sets derived from 19 published PTFs and compared different methods of combining the simulation results from the 19 individual models by using (1) only the best model, (2) equal weights, (3) regressing measured values to results of individual models, (4) singular value decomposition (SVD) in the regression, (5) Bayesian model averaging, and (6) weights derived from Kullback-Leibler information. Data on soil water contents and basic soil properties at 5 depths along a 6-m transect in a layered loamy soil were used to calibrate the Richards equation and to develop the input for PTFs. The SVD multimodel was the best method with the accuracy of about 0.01 cm3cm-3 at the 35-cm depth, and of about 0.005 cm3cm-3 at larger depths for one month monitoring and 13 months of testing. This indicates that monitoring of the soil water regime in combination with multimodel simulations can be a viable approach to simulating water flow in the vadose zone.