Improving the relevance and impact of water quality modelling for decision-making

Catchment managers often turn to computer-based water quality models to support catchment and natural resource management (CNRM). However, model use by managers is inherently problematic. Often-reported problems include inadequate or poor quality input data, miscommunication between scientists and managers, inappropriate treatment of model uncertainty and excessive model complexity or simplicity. This paper reports on the methodological lessons learned from several CNRM projects in New South Wales, Australia. Six modelling project management problems that significantly impact on the utility of models in decision-making are discussed: 1. Relevance and impact. Model evaluation usually focuses on the technical quality of models. Relevance to the decision-making problem and impact on the decision that is made are more useful indicators of model effectiveness in decision-support. 2. Methodological tension. Disparate thinking amongst practitioners within knowledge communities can inhibit effective model use as much or more than disagreement or misunderstanding between scientists and managers. 3. Model uncertainty. Most treatments of uncertainty focus on quantifiable uncertainties and their assessment using sensitivity and uncertainty analyses. Conceptual uncertainties, which are difficult or impossible to quantify, often predominate. 4. Excessive information gathering. Timeliness is critical in decision-support. Excessive information gathering can contribute to 'information overload' and lead to 'analysis paralysis'. 5. Inadequate information transformation. The identification or creation of useful information in useful quantities and useful form may be more critical to CNRM than facets of information and knowledge management that receive greater attention, such as elimination of perceived barriers to knowledge transfer between scientists and managers. 6. Vested interests. The behaviour of individuals in a modelling project is often affected by incentives, biases and value-judgements that can contribute to poor modelling outcomes. Guidelines on good practice in modelling often present a relatively narrow view of how scientific knowledge and modelling is used. Modelling challenges are usually framed as technical or technological problems that can be overcome by selecting appropriate models, using rigorous scientific procedures, and by open communication between managers and scientists. An implied expectation is that if a transparent and rigorous modelling process is used, science and modelling will be usable in CNRM. Much research and development effort in the area of computer-based decision support is being expended on tool creation, data collection and management and on the resolution of technical issues that impact on the quality and reliability of the solutions generated by models and other decision support software. This paper argues that model effectiveness (i.e. the relevance and impact of the model on decision-making) can be improved by focusing greater attention on implementation methodology rather than technology. For example, effort should be focused on evaluating the appropriateness of modelling during program design and planning, and acknowledging and dealing with incongruity between the motivations, methodological preferences and epistemologies of individuals involved in the modelling project, especially where this occurs within rather than between knowledge communities.