Sensitivity and structure assessment of a software tool to gauge the ecological impact of flow scenarios

Several generic problems in sensitivity assessment (SA) of large simulation models employed for decision support are examined by reference to the Murray flow assessment tool (MFAT), software used by the Murray-Darling Basin Commission to help to gauge the ecological effects of alternative flow-management strategies in the River Murray in SE Australia. Sensitivity assessment of the MFAT imposes demands common to SA of many large simulation models, namely avoidance of excessive computing demand, examination of aggregation rules for results which are dispersed spatially, temporally and over types of items, and presentation of results in sufficiently reduced form for easy assimilation without losing important detail. Computing demands are minimised by exploiting the relatively simple structure of the software to analyse much of its sensitivity behaviour algebraically, including the propagation of sensitivities through successive stages of aggregation. Some alternative aggregation formulae are reviewed in terms of their sensitivity properties. A simple scheme for presenting sensitivity results in reduced form, initially as histograms or sets of quantiles, then as classified frequencies of occurrence of significant sensitivities, are illustrated with SA results from the component of the MFAT concerning habitat condition for fish.