Individual Parameter Perturbation and Error Analysis of Fish Bioenergetics Models

Abstract

We quantified the sensitivity of predicted rates of growth and consumption to parameter variation for models of yellow perch (Perca flavescens), largemouth bass (Micropterus salmoides), and alewife (Alosa pseudoharengus). We used statistical analyses of the results of Monte Carlo simulations to rank parameter importance. The order of parameter importance was model specific, although the results emphasized the need for accurate estimation of the realized fraction of maximum consumption rate (P) and allometric parameters for consumption (a₁, b₁) and respiration (a₂, b₂). Excretion and egestion parameters contributed little to prediction errors. The Monte Carlo methods were used to examine the relative importance of parameter variation and diet composition, an external forcing function, on forecasts of alewife growth. If standard deviations of model parameters were known within 2% of their expected values, uncertainty in diet composition could contribute as much as 47% of the variance in predicted alewife weight. When standard deviations of model parameters are realistically defined, diet uncertainty contributed less than 10% of the variance in predicted weight.