Uncertainty management during conceptual modelling: A cognitive mapping case study

‘Scoping’ and ‘problem framing’ begin the integrated assessment and modelling (IAM) process and result in a conceptual model that frames the system of interest according to a defined problem. A conceptual model is mainly qualitative and provides a basis for the later development of a numerical model. Given the significant complexities of interaction among the cross-disciplinary system components that IAM explores, this process of abstracting reality to a more manageable model requires subjective assumptions and decisions tailored to the specific modelling question and purpose. This abstraction and subjectivity may cause a loss of confidence by policy-makers planning to base decisions on model results. To avoid this, best practice guidance from the literature focuses on process and makes recommendations to engage stakeholders, avoid ambiguity in problem framing, maintain transparency on model decisions, and avoid overly-complex model representations that waste resources. The focus on stakeholder engagement and transparency reflects a belief that, ultimately, model utility is largely defined by the acceptance of its outputs by stakeholders, which will be facilitated by an understanding of the modelling process. What is often not clear is at what point each of these recommendations has been addressed sufficiently for the modelling process to progress. In addition, non-modellers may find it difficult to understand the implications of ambiguous problem framing and a lack of model parsimony. This reduces transparency as well as their ability to contribute on multi-disciplinary, multi-sectoral modelling projects. This paper suggests that the concept of uncertainty may be a central driver of best practice in the early phases of scoping, problem framing and conceptual modelling. To illustrate, an existing uncertainty management framework (UMF) is used to guide decisions leading towards a conceptual model for a Cambodian groundwater use case study. Two iterations of the UMF application are performed, alternately using the ‘scoping’ and ‘problem framing’ phases as sources of uncertainty to be managed. The first iteration triggered the decision to undertake formal stakeholder engagement to collect additional knowledge about the system to be modelled, and the selection of Eden’s cognitive mapping approach to structure and analyse this data. This new information helped to reduce or ignore the iteration 1 uncertainties, as well as facilitate the identification of second-iteration uncertainties (only a selection is illustrated). Strengths of the UMF approach as applied to the case study included the iterative identification and treatment of uncertainties, its structured, action-oriented, step-by-step nature, and the guidance and flexibility on the choice of methods. Implementation challenges were mainly peripheral, such as choosing how uncertainties should be prioritized, choosing how to prompt the identify task, and sourcing appropriate methods for a given level and nature of uncertainty. Another challenge was communicating the concept of uncertainty to non-modellers. In terms of handling ambiguity, driving transparency and pursuing parsimony, the uncertainty management approach a) encouraged clarity on what options were available to handle ambiguity and how it might affect problem framing; b) provided transparency through an iterative, structured process that communicated uncertainties about the modelling choices to be made; and c) encouraged questioning of assumptions about model structure, multiple pathways, and which concepts should be made explicit in the conceptual model. The deliberation, transparency, and awareness of resource limits encouraged by the UMF all generated confidence that the resulting system abstraction was defensible and ‘enough’ to progress through the model-building process. It is hoped that the utility and relative simplicity of this approach as demonstrated with this case study will encourage a stronger and more explicit focus on uncertainty during the scoping, problem framing and conceptualization phases of IAM.