In all of our modeling work, we need to understand the uncertainties in our calculations. There are different sources of uncertainty in evaluating the modelling results for the spatial range of a chemical. We classify the various types of uncertainty based on a framework presented by Adam Finkle, which recognises four different types of uncertainty (reference). Parameter uncertainty is the uncertainty in associated with model inputs are used in a calculation (i.e. chemical properties). Model uncertainty results from the effect of simplifying nature for use in models and the associated uncertainty. This is the uncertainty confronting the modeller when selecting among a number of alternate assumptions and algorithms when constructing a model. Decision rule uncertainty addresses disagreements or poor specification of social objectives (i.e. are the models designed to address the decision makers' questions and are these questions clear). Natural variability is the variation in value of a parameter at different locations.
We often carry out uncertainty analyses using Monte Carlo techniques. The first step to complete a Monte Carlo simulation is to determine an appropriate distribution representing the likely values for each parameter. Second, a value is randomly selected from each distribution and the characteristic travel distance is calculated from these values. This is referred to as a single Monte Carlo simulation. The process of selecting input values and calculating the output is repeated over and over, generally hundreds or thousands of times. After all of the simulations are completed, we have a distribution of the possible values for the model output.
Current Projects:
The Office of Emergency and Remedial Response (OERR) plays a lead role in developing national guidance and planning future activities that support the EPA Superfund Program. The purpose of this project is to develop for OERR methods for scoring the quality, relevance and reliability of probability density functions.