Morris Method As a Global Sensitivity Analysis Tool Used to Validate a Model for Water Infiltration into Unsaturated Porous Media.

The governing equation of water flow, i.e. Richard equation, was coded to simulate water infiltration into unsaturated porous media. The equation was approximated by finite-difference solution. The code was used to simulate water infiltration into Yolo light clay. Simulated results were compared to published data on Philip's semi-analytical solution. Discrepancy between simulated results and Philip's semi-analytical solution was first investigated by local sensitivity analysis, and it was subsequently compared to one of the global sensitivity analysis tools, i.e. elementary effect method which is generally known for its ability as a screening tool. Local sensitivity analysis was obtained through sensitivity coefficient by estimating the ratio of simulation output increment or decrement, as in this case the output was water content, to the amount of changes in input parameter. This technique is better known as one-at-a-time (OAT), where each input is varied while fixing other input parameters. Morris method improved the OAT by averaging a number of elementary effects, which is comparable to sensitivity coefficient, at random input values of other parameters. This procedure detached from the dependence of specific local input space value at which the elementary effects are computed. As part of the study, we set up a hypothetical case study on the input space of parameters to identify the parameter which responsible for the discrepancy from the Philip's semi-analytical solution. Our results have shown that the input parameters would be categorized into: (a) negligible; (b) linear and additive, or (c) non-linear or interactions with other parameters, which could be retrieved from Morris method. The parameter, i.e. spatial discretization size, fell into the third category was responsible for the model discrepancy.