Atmospheric tracer experiment uncertainties related to model evaluation

The verification and validation of air quality models is based on comparisons with observational data that can be collected in tracer experiments. The goal of this work is to assess the typical errors that can affect the model evaluation and validation when using real field measurements. The KATREX dataset was chosen for this purpose, since two different teams sampled and analysed concentrations at co-located samplers, therefore providing independent estimates in the same meteorological conditions. Tracers were emitted at two different heights, therefore four datasets are available for the analysis. Comparing the observations of the two teams, also through a statistical analysis, a mean error of 22.5% and a median error of 14% were found. The effect of this uncertainty in the validation of models was then investigated considering the predictions of a Gaussian and a Lagrangian particle models. It followed that the performances of the models could be considered 'good' or not depending on which dataset was used for the evaluation.