A combined emission and receptor-based approach to modelling environmental noise in urban environments

The state of practice for noise assessment utilizes established standards for emission and propagation modelling of linear and point sources. Recently, land use regression (LUR) modelling has emerged as an alternative method due to relatively low data and computing resource demands. However, a limitation of LUR modelling is that is does not account for noise attenuation and reflections by features of the built environment. This study demonstrates and validates a method that combines the two modelling frameworks to exploit their respective strengths: Emission and propagation based prediction of traffic noise, the predominant source of noise at the level of streetscapes, and a LUR-based correction for noise sources that vary on spatial scales beyond the streetscape. Multi-criteria analysis, location-allocation modelling and stakeholder consultation identified 220 monitoring sites with optimal coverage for a 1-week sampling period. A subset of sites was used to validate a road traffic noise emission and propagation model and to specify a LUR model that predicted the contribution of other sources. The equivalent 24-h sound pressure level (LAeq) for all sites was 62.9 dBA (SD 6.4). This varied by time of day, weekday, types of roads and land uses. The traffic noise emission model demonstrated a high level of covariance with observed noise levels, with R-2 values of 0.58, 0.60 and 0.59 for daytime, nighttime and 24-h periods, respectively. Combined with LUR models to correct for other noise sources, the hybrid models R-2 values were 0.64, 0.71 and 0.67 for the respective time periods. The study showed that road traffic noise emissions account for most of the variability of total environmental noise in Toronto. The combined approach to predict fine resolution noise exposures with emission and receptor-based models presents an effective alternative to noise modelling approaches based on emission and propagation or LUR modelling. (C) 2018 Elsevier Ltd. All rights reserved.