Uncertainty in soil hydraulic properties parameterization and its propagation in vadose zone hydrological model output

Vadose zone hydrological models employing finite difference numerical solutions of the Richards equation allow simulating the movement and predicting the state of soil water and associated quantities in the vadose zone. Nowadays, robust algorithms like Hydrus and SWAP are available to perform such simulations. Since most numerical issues with these algorithms have been solved, hydraulic parameters describing the relation between conductivity K, pressure head h and water content θ determine the quality of model output. Whichever method is used to obtain soil hydraulic properties, resulting parameters include an uncertainty, which may be expressed as a confidence interval. Existing correlations between parameters may be expressed in a correlation matrix. Using a stochastic approach, uncertainty and correlation may be considered in simulations and their propagation in results can be assessed. We developed a software which generates n stochastic realizations of the hydraulic parameter set considering uncertainty (standard error) and correlation matrix, runs the SWAP model for each realization and extracts the model output of interest. To apply the software and assess the propagation of uncertainties in the model output, hydraulic properties were measured in soils from south-east Brazil using inverse modeling of laboratory evaporation experiments, resulting in Van Genuchten parameter values, respective errors and correlations. These data were used to obtain n (n=10⁴) stochastic realizations of deep drainage, runoff and evaporation in a bare-soil scenario. Similarly, for a pasture cropped scenario the water balance components (including transpiration), and relative yield were evaluated. The effect of uncertainty in these parameters on the mentioned output variables prediction is presented and discussed.