Presenter/Author Information

Wendy S. Merritt
B. G. W. Croke
A. J. Jakeman

Keywords

water resources, catchcrop, ihacres, northern thailand, sensitivity analysis

Start Date

1-7-2002 12:00 AM

Description

This paper investigates the sensitivities of model outputs to model parameter values within a Biophysical Toolbox developed as part of a Decision Support System (DSS) for integrated catchment assessment and management of land and water resources in the highland regions of northern Thailand. The toolbox contains a hydrological module based upon the IHACRES rainfall-runoff model, a crop model (CATCHCROP), and an erosion model (USLE) modified to suit conditions in northern Thailand. Emphasis in the development of the individual models within the Biophysical Toolbox was placed upon limiting model complexity. Limited data availability commonly restricts the complexity of the model structure that can justifiably be used to model natural systems. The challenge under conditions with limited data is then to strike a balance in the model(s) between statistical rigour and model complexity. Once encompassed within the Biophysical Toolbox, linkages between the models increase the complexity of the system, despite the relative simplicity of the individual models. Consequently, the impacts of outputs from individual models on the outputs of other models deserve considerable attention. Understanding model sensitivity is of particular importance where there is a lack of data with which to support or adequately verify model behaviour. Sensitivity analysis potentially allows the identification of model components that require attention in terms of improved parameter estimation or improvement in model structure. Preliminary testing of the individual models within the Biophysical Toolbox has been reported previously within the literature and the Biophysical Toolbox as a whole has been described. This paper explores sensitivities within the Biophysical Toolbox, targeting in particular the identification of components of the toolbox in which sensitivities are propagated throughout the model.

Share

COinS
 
Jul 1st, 12:00 AM

Sensitivity Testing of a Biophysical Toolbox for Exploring Water Resources Utilisation and Management Options

This paper investigates the sensitivities of model outputs to model parameter values within a Biophysical Toolbox developed as part of a Decision Support System (DSS) for integrated catchment assessment and management of land and water resources in the highland regions of northern Thailand. The toolbox contains a hydrological module based upon the IHACRES rainfall-runoff model, a crop model (CATCHCROP), and an erosion model (USLE) modified to suit conditions in northern Thailand. Emphasis in the development of the individual models within the Biophysical Toolbox was placed upon limiting model complexity. Limited data availability commonly restricts the complexity of the model structure that can justifiably be used to model natural systems. The challenge under conditions with limited data is then to strike a balance in the model(s) between statistical rigour and model complexity. Once encompassed within the Biophysical Toolbox, linkages between the models increase the complexity of the system, despite the relative simplicity of the individual models. Consequently, the impacts of outputs from individual models on the outputs of other models deserve considerable attention. Understanding model sensitivity is of particular importance where there is a lack of data with which to support or adequately verify model behaviour. Sensitivity analysis potentially allows the identification of model components that require attention in terms of improved parameter estimation or improvement in model structure. Preliminary testing of the individual models within the Biophysical Toolbox has been reported previously within the literature and the Biophysical Toolbox as a whole has been described. This paper explores sensitivities within the Biophysical Toolbox, targeting in particular the identification of components of the toolbox in which sensitivities are propagated throughout the model.