Sensitivity Analysis of a Network-Based, Catchment- ScaleWater-Quality Model

Careful consideration of the uncertainties and sensitivities associated with model outputs is essential when critical decisions are made on the basis of such results. Consideration of uncertainty is particularly important in the context of natural resource management, where models are often used to tackle complex and conflicting issues across multiple scales, as is the case in evaluating management options to reduce surface water pollution. This paper describes an analysis of uncertainty in the catchment-scale integrated hydrologic, economic, stream sediment and nutrient export model known as CatchMODS. The paper briefly describes the linked components of CatchMODS and its application in the Ben Chifley Dam catchment, Australia. An initial investigation to investigate some of the most important sources of output uncertainty is described. First-order sensitivities to selected model parameters are found analytically by linearising parts of the model and used, together with knowledge of where non-linearity has most effect, to point to conditions to be investigated further. The extent of non-linear effects is also checked by comparing the analytical results with the results of parameter perturbation tests. Results from the analysis are used to prioritise continuing model development and data-collection activities. The results are also to be incorporated into a decision-analysis framework to evaluate management options to reduce surface water pollution. The decision-analysis framework and incorporation of uncertainty analysis into it are outlined.