Premium
Bayesian uncertainty estimation methodology applied to air pollution modelling
Author(s) -
Romanowicz Renata,
Higson Helen,
Teasdale Ian
Publication year - 2000
Publication title -
environmetrics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.68
H-Index - 58
eISSN - 1099-095X
pISSN - 1180-4009
DOI - 10.1002/(sici)1099-095x(200005/06)11:3<351::aid-env424>3.0.co;2-z
Subject(s) - uncertainty analysis , range (aeronautics) , stability (learning theory) , sensitivity (control systems) , wind speed , reliability (semiconductor) , uncertainty quantification , covariance , atmospheric dispersion modeling , bayesian probability , atmospheric instability , statistics , mathematics , econometrics , computer science , meteorology , air pollution , physics , power (physics) , materials science , quantum mechanics , machine learning , electronic engineering , engineering , composite material , chemistry , organic chemistry
The aim of the study is an uncertainty analysis of an air dispersion model. The model used is described in NRPB‐R91 (Clarke, 1979), a model for short and medium range dispersion of radionuclides released into the atmosphere. Uncertainties in the model predictions arise both from the uncertainty of the input variables and the model simplifications, resulting in parameter uncertainty. The uncertainty of the predictions is well described by the credibility intervals of the predictions (prediction limits), which in turn are derived from the distribution of the predictions. The methodology for estimating this distribution consists of running multiple simulations of the model for discrete values of input parameters following some assumed random distributions. The value of the prediction limits lies in their objectivity. However, they depend on the assumed input distributions and their ranges (as do the model results). Hence the choice of distributions is very important for the reliability of the uncertainty analysis. In this work, the choice of input distributions is analysed from the point of view of the reliability of the predictive uncertainty of the model. An analysis of the influence of different assumptions regarding model input parameters is performed. Of the parameters investigated (i.e. roughness length, release height, wind fluctuation coefficient and wind speed), the model showed the greatest sensitivity to wind speed values. A major influence on the results of the stability condition specification is also demonstrated. Copyright © 2000 John Wiley & Sons, Ltd.