Managing uncertainty in modelling of wicked problems: theory and application to Sustainable Aquifer Yield

Abstract

This thesis presents two approaches to help manage uncertainty in modelling for the resolution of wicked problems , which have no clear problem definition, solution or measure of success. It focuses on Sustainable Aquifer Yield (SAY) as an example. SAY is defined as the pumping volume obtained by a management plan that is expected to satisfy objectives under future conditions within a groundwater system. Integrated modelling can help express, systematise and use knowledge of relevant behaviour of the system, while engaging diverse stakeholders and addressing their interests. Uncertainty is however a key and multifaceted issue when dealing with wicked problems. While many modelling methods exist to help address this uncertainty, there is a need for modellers to be able to integrate these methods purposefully for an applied problem. The research presented involved iteratively proposing two approaches to manage uncertainties in integrated modelling that supports decision making, and exploring the value of each approach by applying it to case studies. For each approach, the applications specifically a) address a technical problem, b) push boundaries on how the problem is viewed, specifically identifying hitherto neglected aspects, and c) address a context where accounting for contested views and surprise is imperative. This research process is described in terms of Critical Systems Practice and resulted in a compilation of linked publications. The first approach proposed is an Uncertainty Management Framework that can be used to help audit the treatment of uncertainty in a step-wise description of an analysis (e.g. evaluating a management plan). The framework provides a formal structure for managing uncertainty by incorporating an uncertainty typology and a set of fundamental uncertainty management actions, but may be too restrictive and demanding for some contexts. To address these limitations, a complementary second approach, designated Iterative Closed Question Modelling, addresses uncertainty by constructing models to test whether each possible answer to a closed question is plausible. The question, assumptions about plausibility and the process of constructing models are all considered uncertain and therefore themselves iteratively critiqued. This approach is formalised in terms of Boundary Critique such that it provides a philosophical foundation justifying the use of a broad range of methods to manage uncertainty in predictive modelling. The thesis concludes that uncertainty needs to be embraced as a natural part of researchers, policy makers and community coming to grips with an evolving situation, rather than being an obstacle to be eliminated. Training of modellers to manage uncertainty needs to specifically address: identification of model scenarios that contradict dominant conclusions; critique of model assumptions and questions from multiple stakeholders’ points of view; and negotiation of the modeller’s role in anticipating surprise (e.g. through understanding consequences of error, design of monitoring, contingency planning and adaptive management). The resulting emphasis on critical thinking about alternative models helps to remind the user that modelling is not a magic trick for seeing the future, but a structured way to reason about both what we do and do not know. Show more