Effect of sampling duration on the estimate of pollutant concentration behind a heavy-duty vehicle: A large-eddy simulation

Abstract

Plume chasing is cost-effective, measuring individual, on-road vehicular emissions. Whereas, wake-flow-generated turbulence results in intermittent, rapid pollutant dilution and substantial fluctuating concentrations right behind the vehicle being chased. The sampling duration is therefore one of the important factors for acquiring representative (average) concentrations, which, however, has been seldom addressed. This paper, which is based on the detailed spatio-temporal dispersion data after a heavy-duty truck calculated by large-eddy simulation (LES), examines how sampling duration affects the uncertainty of the measured concentrations in plume chasing. The tailpipe dispersion is largely driven by the jet-like flows through the vehicle underbody with approximate Gaussian concentration distribution for x ≤ 0.6h, where x is the distance after the vehicle and h the characteristic vehicle size. Thereafter for x ≥ 0.6h, the major recirculation plays an important role in near-wake pollutant transport whose concentrations are highly fluctuating and positively shewed. Plume chasing for a longer sampling duration is more favourable but is logistically impractical in busy traffic. Sampling duration, also known as averaging time in the statistical analysis, thus has a crucial role in sampling accuracy. With a longer sampling (averaging) duration, the sample mean concentration converges to the population mean, improving the sample reliability. However, this effect is less pronounced in long sampling duration. The sampling accuracy is also influenced by the locations of sampling points. For the region x > 0.6h, the sampling accuracy is degraded to a large extent. As a result, acceptable sample mean is hardly achievable. Finally, frequency analysis unveils the mechanism leading to the variance in concentration measurements which is attributed to sampling duration. Those data with frequency higher than the sampling frequency are filtered out by moving average in the statistical analyses.