An Inverse Approach to the Analysis of Uncertainty in Models of Environmental Systems

Abstract

An inverse methodology that integrates qualitative and quantitative aspects of uncertainty in environmental modeling is presented. The methodology, RIMME (Random-search Inverse Methodology for Model Evaluation), comprises three Monte Carlo procedures: (i) Regionalized Sensitivity Analysis (RSA); (ii) Tree-Structured Density Estimation (TSDE); and (iii) Uniform Covering by Probabilistic Rejection (UCPR). Unlike conventional direct predictive approaches, inverse methods work backwards to identify attributes of the model, and the corresponding real system, that are critical to attaining a prescribed endpoint. RIMME is capable of integrating scientific uncertainties, in the model and empirical data, with non-standard, qualitative forms of uncertainty, such as the value-laden policy and stakeholder issues that feature prominently in contemporary environmental assessments. RIMME is applied to a case study of Lake Lanier, Georgia (USA), a key resource whose water quality and ecological integrity is perceived by society to now be threatened by rapid urbanization within and around its watershed. Results indicate that RIMME provides an effective bridge across the gap between traditional science and the now emerging post-normal science era.

Keywords:

• uncertainty
• environmental models
• integrated methodology
• Lake Lanier
• Monte Carlo
• RIMME