Evaluation of probability distributions for concentration fluctuations in a building array

The wide range of values observed in a measured concentration time series after the release of a dispersing airborne pollutant from a point source in the atmospheric boundary layer, and the hazard level associated with the peak values, demonstrate the necessity of predicting the concentration probability distribution. For this, statistical models describing the probability of occurrence are preferably employed. In this paper a concentration database pertaining to a field experiment of dispersion in an urban-like area (MUST experiment) from a continuously emitting source is used for the selection of the best performing statistical model between the Gamma and the Beta distributions. The skewness, the kurtosis as well as the inverses of the cumulative distribution function were compared between the two statistical models and the experiment. The evaluation is performed in the form of validation metrics such as the Fractional Bias (FB), the Normalized Mean Square Error and the factor-of-2 percentage. The Beta probability distribution agreed with the experimental results better than the Gamma probability distribution except for the 25th percentile. Also according to the significant tests using the BOOT software the Beta model presented FB and NMSE values that are statistical different than the ones of the Gamma model except the 75th percentiles and the FB of the 99th percentiles. The effect of the stability conditions and source heights on the performance of the statistical models is also examined. For both cases the performance of the Beta distribution was slightly better than that of the Gamma.