Numerical models make an important contribution to the assessment of potential environmental impacts, such as the potential cumulative impacts of coal resource development on water resources and water-dependent assets, including rivers, wetlands and groundwater systems. Uncertainty analysis, where potential sources of uncertainty in model predictions are analysed using quantitative or qualitative techniques, are used to build confidence in model predictions. While techniques to quantify the uncertainty in model predictions are well documented, effective communication of predictive uncertainty that instils confidence in the model and its predictions remains a challenge.
This paper describes the challenges and solutions used to communicate the uncertainty analysis developed for the Bioregional Assessment Programme in Australia. To do this, numerical models are run hundreds or thousands of times to quantify the sensitivity of model predictions to input parameters using a credible range of possible input parameters. The resulting range of possible model predictions is a high-dimensional dataset, which is challenging to visualise and summarise.
Bioregional Assessments have simplified the communication of these often very skewed distributions of model predictions by reporting the probability of exceeding a specified threshold or percentile estimates of a model output. Model predictions are reported in three ways: 1. A conservative zone to rule-in and rule-out impacts; 2. Maps, tables and plots that show the range of estimates; and 3. Median, 5th and 95th percentile estimates of model outputs described in the text of the report.
The open and transparent communication of the uncertainty associated with the numerical models used in the assessment of potential environmental impacts ultimately improves confidence in the model and its predictions.