Uncertainty Assessment of Phosphorus Risk to Surface Waters
Cost: £140k
Duration: 2006-2007
Details
The focus of the first year of this project has been the elicitation of experts' opinions on the scale-dependent functioning of terrestrial and aquatic systems with respect to the risk to ecology from elevated levels of phosphorus (P) loss. This was achieved by identification of the functionally significant components essential for inclusion in a risk analysis and at WFD catchment scales. This has been driven by the need to meet requirements for the Water Framework Directive (WFD) in the form of ecological risk analyses and the uncertainties highlighted in the WFD River Basin Characterisation (RBC1) P risk analysis. The RBC1 P risk analysis generated estimates of relative risk to surface waters (a P pressure risk) given a number key catchment characteristics, but was shown to have a poor correlation with EA monitoring data. In light of this, experts' opinions were also sought on the magnitudes and sources of uncertainty associated with any such risk analysis. The results from the elicitation are to be used to build a framework that builds uncertainty and conceptual understanding into the development of revised risk analyses and highlights areas of future research which may reduce the uncertainties associated with our estimates.
For such a problem, where large scale assessments are to be made of complex and heterogeneous systems, a systematic approach that could assist in the translation of research findings at various scales to the scale of interest was needed. A methodology for the elicitation of expert opinion was developed specifically for this project (Figure 1). The methodology involved using structured one-to-one interviews with experts identified for particular research disciplines within the broad area of P transfers and cycling in terrestrial and aquatic catchment science. The approach developed draws from diverse methods such as: NUSAP (Funtowicz and Ravetz, 1990), Bayesian Belief Networks (Barton 2006), fuzzy logic and subjective probabilities (Tversky and Kahneman, 1974), which include mathematical, psychological and social science elements. Experts were asked to provide information in a variety of quantitative (i.e. as fuzzy distributions; Figure 2) and qualitative ways describing their beliefs regarding chosen components of conceptual models. Scale-dependent conceptual models were developed with the experts (eg. Lake Model); these models were central in the stimulation of conversation and as reference for identification of dominant model components. For these components, factors such as scientific understanding, scale-dependent transferability of scientific principles and data quantity and quality were considered to try and identify the origins of the primary uncertainties. For the quantitative part of the study, experts' opinions were combined to gain an overview of area of agreement and disagreement.
