Assessing the Internal Variability of Large-Eddy Simulations for Microscale Pollutant Dispersion Prediction in an Idealized Urban Environment

Abstract

This study aims at estimating the inherent variability of microscale boundary-layer flows and its impact on air pollutant dispersion in urban environments. For this purpose, we present a methodology combining high-fidelity large-eddy simulation (LES) and a stationary bootstrap algorithm, to estimate the internal variability of time-averaged quantities over a given analysis period thanks to sub-average samples. A detailed validation of an LES microscale air pollutant dispersion model in the framework of the Mock Urban Setting Test (MUST) field-scale experiment is performed. We show that the LES results are in overall good agreement with the experimental measurements of wind velocity and tracer concentration, especially in terms of fluctuations and peaks of concentrations. We also show that both LES estimates and the MUST experimental measurements are subject to significant internal variability, which is therefore essential to take into account in the model validation. Moreover, we demonstrate that the LES model can accurately reproduce the observed internal variability.