Linear understanding of a huge aquatic ecosystem model using a group-collecting sensitivity analysis

Abstract

Huge complex ecosystem models with several hundred parameters and large input data sets escape standard attempts at integral assessment. We introduce the concept of group-collecting sensitivity analysis in which related model parameters or forcing coefficients are combined into subsets. Since means and standard deviations of subsets are varied instead of individual coefficients, the method is numerically efficient and produces a condensed amount of results. Application to the Aquatic Ecosystem Model (AQEM) is presented. AQEM is a descendant of the European Regional Seas Ecosystem Model (ERSEM) with a finer process and spatial resolution with respect to the Wadden Sea. A two-dimensional sub-structured sensitivity table, which is the major result of this approach, enables an immediate perception of sensitive functional relationships and dependencies between individual parameters and the relevant characteristics of a near-shore aquatic ecosystem. Special emphasis is placed on differences in average and seasonal behaviour. The response of selected result variables to the variations of the majority of group parameters is correlated, i.e. result variables show a similar sensitivity to variations in a specific parameter group. We show that exceptions to this rule lead to a deeper insight into the model system.