Using the SWAP model for the stochastic analysis of hydraulic parameter uncertainty propagation in soil water balance components

The SWAP model employs a finite difference numerical solution of the Richards equation, including root water uptake, to simulate the movement and predict the state of soil water and associated quantities in the vadose zone. The relation between hydraulic conductivity K, pressure head h, and water content θ can be described by parameters of the Van Genuchten-Mualem (VGM) relations, where the quality of these parameters determines the quality of the model output. We developed a stochastic procedure to evaluate the outputs of the SWAP model according to the uncertainty and correlations in the VGM parameters. Specific software was developed to (1) fit VGM parameters to observed retention and conductivity data to obtain values, standard errors, and correlations of transformed parameters (software HPFit); (2) generate p stochastic realizations of the VGM parameters using Cholesky decomposition (software StochHyProp), and (3) run the SWAP model with each of the p parameter realizations for specific scenarios, extracting simulation results (software RunSWAP), e.g., the simulated water balance components evaporation, transpiration, bottom flux, and runoff. The results from the last step yield the respective frequency distributions of the output values. Examples will show that the most commonly performed prediction using the average VGM parameter values does not always agree with the median of the stochastic realizations. The developed procedure allows the quantitative analysis of the uncertainty of SWAP model output, adding to the interpretation of the predictive power of hydrological models like SWAP.