Milando Chad W, Batterman Stuart A
Environmental Health Sciences, University of Michigan, 1415 Washington Heights, Ann Arbor, MI USA 48109.
Atmos Environ (1994). 2018 May;181:135-144. doi: 10.1016/j.atmosenv.2018.03.009. Epub 2018 Mar 21.
The development of accurate and appropriate exposure metrics for health effect studies of traffic-related air pollutants (TRAPs) remains challenging and important given that traffic has become the dominant urban exposure source and that exposure estimates can affect estimates of associated health risk. Exposure estimates obtained using dispersion models can overcome many of the limitations of monitoring data, and such estimates have been used in several recent health studies. This study examines the sensitivity of exposure estimates produced by dispersion models to meteorological, emission and traffic allocation inputs, focusing on applications to health studies examining near-road exposures to TRAP. Daily average concentrations of CO and NO predicted using the Research Line source model (RLINE) and a spatially and temporally resolved mobile source emissions inventory are compared to ambient measurements at near-road monitoring sites in Detroit, MI, and are used to assess the potential for exposure measurement error in cohort and population-based studies. Sensitivity of exposure estimates is assessed by comparing nominal and alternative model inputs using statistical performance evaluation metrics and three sets of receptors. The analysis shows considerable sensitivity to meteorological inputs; generally the best performance was obtained using data specific to each monitoring site. An updated emission factor database provided some improvement, particularly at near-road sites, while the use of site-specific diurnal traffic allocations did not improve performance compared to simpler default profiles. Overall, this study highlights the need for appropriate inputs, especially meteorological inputs, to dispersion models aimed at estimating near-road concentrations of TRAPs. It also highlights the potential for systematic biases that might affect analyses that use concentration predictions as exposure measures in health studies, e.g., to estimate health impacts.
鉴于交通已成为城市主要的暴露源，且暴露估计会影响相关健康风险的估计，因此为与交通相关的空气污染物（TRAPs）的健康效应研究开发准确且合适的暴露指标仍然具有挑战性且十分重要。使用扩散模型获得的暴露估计可以克服监测数据的许多局限性，并且此类估计已在最近的几项健康研究中得到应用。本研究考察了扩散模型产生的暴露估计对气象、排放和交通分配输入的敏感性，重点关注其在研究TRAP近路暴露的健康研究中的应用。将使用研究线源模型（RLINE）和时空分辨移动源排放清单预测的一氧化碳（CO）和一氧化氮（NO）的日平均浓度与密歇根州底特律近路监测点的环境测量值进行比较，并用于评估队列研究和基于人群的研究中暴露测量误差的可能性。通过使用统计性能评估指标和三组受体比较名义模型输入和替代模型输入来评估暴露估计的敏感性。分析表明，暴露估计对气象输入相当敏感；一般来说，使用每个监测点的特定数据可获得最佳性能。更新后的排放因子数据库带来了一些改进，尤其是在近路站点，而与更简单的默认概况相比，使用特定站点的日交通分配并未提高性能。总体而言，本研究强调了针对旨在估计TRAPs近路浓度的扩散模型，需要合适的输入，尤其是气象输入。它还强调了可能影响健康研究中使用浓度预测作为暴露测量指标的分析（例如估计健康影响）的系统偏差的可能性。
