Sensitivity and Uncertainity

Abstract

Ecological models are subject to considerable uncertainty, which needs to be addressed explicitly. This can be achieved by using sensitivity (SA) and uncertainty (UA) analyses. Both attempt to propagate uncertainties in model inputs and parameters to model output(s). In SA, uncertainties in model inputs and parameters are represented by likely ranges. In contrast, in UA, the uncertainty in each of the inputs and parameters is characterized by a probability density function. SA can be used to assess model validity, as well as the relative sensitivities of model outputs to model inputs/parameters. Due to the highly nonlinear nature of ecological models, and the strong possibility of interactions between model inputs and parameters, sampling-based SA methods are most appropriate in an ecological modeling context. UA can be used to determine distributions of model outputs, enabling confidence limits on predictions and a range of risk-based performance measures to be obtained. If no calibration data are available, input distributions can be estimated independently of model outputs and propagated through the model using Monte Carlo methods or first- or second-order approximations. When calibration data are available, more sophisticated methods, such as generalized likelihood uncertainty estimation (GLUE) or Markov Chain Monte Carlo (MCMC), have to be used, and output distributions generated by sampling from the parameter/input vectors obtained.