Modelling uncertainty and errors

by Barry
(Calgary)

Here is what the EPA has to say about modelling uncertainty and errors. It incorporates some theory put forth by Carl Gauss:

Modeling software is more trustworthy for coming up with longer time-period concentrations than for calculating one-hour concentrations at precise locations.

The programs are okay for estimating highest concentrations that may take place anywhere in the domain. For example,

"...errors in highest estimated concentrations of ± 10 to 40 % are found to be typical (assuming appropriate inputs)."

Concentration calculations for an exact time and place correlate poorly with monitored concentrations (corresponding to the same time and space) and reliability is not as good.

Uncertainties fall into two categories:

1) Reducible - where plume locating errors arise from errors in wind direction in the meteorological data.

2) Unspecified uncertainties intrinsic to the atmosphere not incorporated into the model. These errors can be quite large, up to 50%, and are no indication that specific concentrations do not actually happen. These are examples of natural modelling uncertainty and errors. We just can't tell precisely where and when.



Search this site for more information now.