Keywords
sensitivity analysis, integrated assessment modelling, decision-making
Start Date
1-7-2006 12:00 AM
Abstract
Integrated Assessment Modelling (IAM) incorporates knowledge from different disciplines to provide an overarching assessment of the impact of different management decisions. Integrated models generally require numerous parameters from varying sources, many not known with certainty. Rapid increases in model size and complexity, particularly in the case of integrated models for decision-making, pose new challenges for effective sensitivity analysis. Some of the identified shortcomings of existing sensitivity analysis methods in the context of IAM include: computational inefficiency, failure to assess parameter interactions, excessive data requirements (e.g. requiring parameter probability distributions), assumptions of model linearity and monotonicity and, in particular, difficulty of use in decision-making. To overcome these shortcomings, a new, rank-equivalence method of sensitivity analysis is proposed. The method operates on the assumption that model outputs will be used for ranking of management options. Where models are used for decision-making it is important to ensure that the solution is robust and that rankings will not alter with small changes in model parameters or inputs. The Rank-Equivalence method incorporates parameter bounding as well as numerical optimisation methods in order to find the minimum combined change in parameters or inputs that will result in the ranking of two management options becoming equal. This allows a translation of the set of acceptable model outcomes into a corresponding range of model inputs, thus allowing decision-makers to directly assess whether the current uncertainties of model parameters and inputs are adequate for differentiating between management options. The Rank-Equivalence method is tested using a case study of an integrated catchment model of the Namoi River. The SA results from the case study indicate that while there are several solutions of similar fitness, the solutions may be comprised of different changes in several parameters
Rank-Equivalence Method for Sensitivity Analysis of an Integrated Model of a River Catchment
Integrated Assessment Modelling (IAM) incorporates knowledge from different disciplines to provide an overarching assessment of the impact of different management decisions. Integrated models generally require numerous parameters from varying sources, many not known with certainty. Rapid increases in model size and complexity, particularly in the case of integrated models for decision-making, pose new challenges for effective sensitivity analysis. Some of the identified shortcomings of existing sensitivity analysis methods in the context of IAM include: computational inefficiency, failure to assess parameter interactions, excessive data requirements (e.g. requiring parameter probability distributions), assumptions of model linearity and monotonicity and, in particular, difficulty of use in decision-making. To overcome these shortcomings, a new, rank-equivalence method of sensitivity analysis is proposed. The method operates on the assumption that model outputs will be used for ranking of management options. Where models are used for decision-making it is important to ensure that the solution is robust and that rankings will not alter with small changes in model parameters or inputs. The Rank-Equivalence method incorporates parameter bounding as well as numerical optimisation methods in order to find the minimum combined change in parameters or inputs that will result in the ranking of two management options becoming equal. This allows a translation of the set of acceptable model outcomes into a corresponding range of model inputs, thus allowing decision-makers to directly assess whether the current uncertainties of model parameters and inputs are adequate for differentiating between management options. The Rank-Equivalence method is tested using a case study of an integrated catchment model of the Namoi River. The SA results from the case study indicate that while there are several solutions of similar fitness, the solutions may be comprised of different changes in several parameters
