Spatial patterns of probability distributions for concentration fluctuations at the street network scale

Klippel, K., Goulart, E. V. ORCID: https://orcid.org/0000-0002-0051-0778, Auerswald, T. ORCID: https://orcid.org/0009-0007-5827-5487, Galvão, E., Marques Ciarelli, P. ORCID: https://orcid.org/0000-0003-3177-4028, Costa Reis, N. ORCID: https://orcid.org/0000-0002-6159-4063 and Coceal, O. ORCID: https://orcid.org/0000-0003-0705-6755 (2025) Spatial patterns of probability distributions for concentration fluctuations at the street network scale. Building and Environment, 285 (Part A). 113491. ISSN 1873-684X doi: 10.1016/j.buildenv.2025.113491

Abstract/Summary

This study investigates concentration fluctuations in street networks, focusing on the spatial dependence of their probability distribution functions (PDFs) and their representation through statistical models. Direct numerical simulations (DNS) of passive scalar dispersion were performed to characterize these fluctuations in a regular street network composed of rectangular buildings. The scalar was continuously released from a ground level point source at an intersection, with a wind forcing angle of 45 degrees to the street orientation. Clustering analysis identified three regions within the plume: (1) the plume edge, with exponential-like distributions, high intermittency, and extreme skewness and kurtosis; (2) the transition region, with asymmetric distributions and reduced intermittency; and (3) the plume centre, exhibiting Gaussian-like distributions with negligible intermittency and near-zero skewness. Four theoretical distributions—Gamma, Beta, Lognormal, and Weibull — were evaluated for their ability to model these fluctuations. Gamma and Weibull provided the best fits overall, capturing variations in distribution shapes and accurately modelling moderate to extreme concentrations across the different regions. The Gamma distribution was the most consistent, demonstrating strong performance for both the 50th and 98th percentiles of the inverse cumulative density function (ICDF). Analyses of model performance for separate moments revealed that the Gamma model performed best in predicting the variance in high-fluctuation regions (cluster 1) whereas the Beta model showed superior results for skewness and kurtosis in low-fluctuation areas (cluster 3). All models failed to accurately capture higher-order moments (skewness and kurtosis) in cluster 1, with none demonstrating significant superiority, highlighting challenges in accurately modelling fat-tailed distributions.

Item Type	Article
URI	https://centaur.reading.ac.uk/id/eprint/131337
Identification Number/DOI	10.1016/j.buildenv.2025.113491
Refereed	Yes
Divisions	Science > School of Mathematical, Physical and Computational Sciences > Department of Meteorology
Publisher	Elsevier
Download/View statistics	View download statistics for this item