Effects of parameter uncertainty on predictions of unsaturated flow

Abstract

This study uses first-order first and second moment analysis to evaluate the error in predictions of unsaturated flow models caused by parameter uncertainty, expressed in terms of a mean value for each parameter and an error covariance matrix. These are related to the mean and variance in the model predictions by a first-order Taylor expansion. Two applications of the methodology are presented. The first involves the estimation of parameters and their error covariance matrix from particle size distribution data for a layered field soil. These are used to predict water contents and associated errors for a static retention test and a dynamic drainage experiment. The second case involves parameter estimation using a numerical inversion method from a transient flow experiment for a hypothetical homogeneous soil. Predictions of water contents and their errors for a rainfall-redistribution event using the estimated parameters compare well with predictions corresponding to actual parameters. Uncertainty predictions with the first-order error analysis procedure also agree well with results of a Monte Carlo simulation study.