Improving ecological simulation through sensitivity analysis

Ecosystem models require extensive data for initial ization and calibration. Since data collection is expensive and time-consuming, it should be guided by knowledge of the model's sensitivity to initial con ditions, coefficients, and update intervals. These factors were studied using an ecological model of a reservoir. For this model, it was found that (a) biological and chemical variables need to be considered in addition to temperature during calibra tion, (b) changes in the effective diffusion coeffi cients produce the greatest changes in the model's responses, (c) nutrient update intervals of less than one week yield results similar to those for one week but dissimilar to those for two weeks or one month, and (d) nutrient concentrations strongly affect the model's sensitivity to small changes in a number of coefficients.