While model parameter uncertainties are easily obtained from parameter estimators such as PEST, they are often of limited use from a management perspective. With regard to the decision making process, we tend to be more interested in estimates of fluxes and water levels instead. This research examines the extent to which model output estimates such as upward seepage and drainage intensities are influenced by model parameter estimates and uncertainties obtained from PEST.
Model Output Uncertainties in Groundwater Modelling
While model parameter uncertainties (e.g. hydraulic conductivity, entrance resistance) are easily obtained from parameter estimators such as PEST, they are often of limited use from a management perspective. With regard to the decision making process, we tend to be more interested in computed fluxes and water levels instead, and the uncertainty surrounding these outcomes. This research examines the assessment of model output uncertainties of quantities such as upward seepage and drainage fluxes as a function of parameter uncertainties.
Two methods are compared: linearized approximations and Monte Carlo analysis. Compared to Monte Carlo analyses, linearized approximations are less computationally demanding (and thus faster), but provide no information on the distribution of uncertainties.
Both methods were applied to a 1-dimensional Hooghoudt model, as well as a case study of the Hooge Raam area (approx. 150 km2, in Noord-Brabant, the Netherlands) in MODFLOW.
Results of the case study show that, while model parameter estimates are highly influenced by the choice of adjustable parameters, the influence on model output values is limited.