Indicators of performance of dispersion models and their reference values
A methodology of validation of dispersion models is analysed in this paper for situations when measurements, corresponding to a given set of governing meteorological parameters, are contaminated with stochastic noise resulting from the turbulent nature of atmospheric processes. Deterministic dispersion models in these situations predict certain non-stochastic values of concentrations and, therefore, cannot reproduce measured patterns of stochasticity. Resulting discrepancies are reflected in all indicators of performance of the models. The values of these indicators are estimated in this paper for an idealised case of a 'perfect model' that can predict exact values of characteristics of the probability distribution function of measured concentrations in assumption that this PDF can be fitted with a lognormal distribution. The estimates obtained can be used as reference values of the indicators of performance of dispersion models.
Keywords: dispersion, modelling, validation, performance, indicator