Uncertainty in Environmental Decision Making: Issues, Challenges and Future directions

Abstract

Environmental decision making is complicated by the complexity of natural systems and the generally competing needs of multiple stakeholders. Modelling tools are often used to assist at various stages of the environmental decision-making process. If such models are to provide effective decision support, the uncertainties associated with all aspects of the decision-making process need to be taken into account explicitly, including those associated with data, models and human factors. However, as models become more complex to better represent integrated environmental, social and economic systems, achieving this goal becomes more difficult. Some of the important issues that need to be addressed in relation to the incorporation of uncertainty in environmental decision-making processes include:. •The development of appropriate risk-based performance criteria that are understood and accepted by a range of scientific disciplines. Risk-based criteria generally relate to the concept of likelihood, the likely magnitude and the likely duration of failure, where failure is defined as the inability of an environmental system to perform its desired function. However, the terminology used in various disciplines differs. Given the increase in the use of integrated models, and the resulting collaboration between people from different disciplines, there is a need to develop a common lexicon, or at least a shared understanding of the meaning of the terminology used.•The development of methods for quantifying the uncertainty associated with human input (see Chapter 6). This includes the development of uncertainty analysis methods that are able to cater for subjective and non-quantitative factors, human decision-making processes (which may be influenced by political and other external factors), and uncertainties associated with the model development process itself.•The development of approaches and strategies for increasing the computational efficiency of integrated models, optimisation methods, and methods for estimating risk-based performance measures. Examples include the use of efficient Monte Carlo sampling techniques (e.g. Latin hypercube sampling) or first- and second-order approximations (e.g. first- and second-order reliability methods), the use of innovative sensitivity analysis methods to skeletonise complex integrated models and the replacement of computationally expensive process models with data-driven metamodels (e.g. artificial neural networks).•The development of integrated software frameworks that enable all sources of uncertainty to be incorporated in the environmental decision-making process (see Chapter 7). © 2008 Elsevier B.V. All rights reserved.