Sensitivity analysis of an environmental model an application of different analysis methods

A parametric sensitivity analysis (SA) was conducted on a well known model for the production of a key sulphur bearing compound from algal biota. The model is of interest because of the climatic relevance of the gas (dimethylsulphide, DMS), an initiator of cloud particles. A screening test at low sample size is applied first (Morris method) followed by a computationally intensive variance based measure. Standardised regression coefficients are also computed. The various SA measures are compared with each other, and the use of bootstrap is suggested to extract empirical confidence bounds on the SA estimators. For some of the input factors, investigators guess about the parameters relevance was confirmed; for some others, the results shed new light on the system mechanism and on the data parametrisation. (C) 1997 Elsevier Science Limited.