Whitehill Andrew R, Lunden Melissa, Kaushik Surender, Solomon Paul
Center for Environmental Measurement and Modeling, Office of Research and Development, United States Environmental Protection Agency, Research Triangle Park, North Carolina, 27711, USA.
Aclima, Inc, San Francisco, CA, 94111, USA.
Atmos Environ X. 2020 Oct 1;7. doi: 10.1016/j.aeaoa.2020.100080.
Mobile mapping of air pollution has the potential to provide pollutant concentration data at unprecedented spatial scales. Characterizing instrument performance in the mobile context is challenging, but necessary to analyze and interpret the resulting data. We used robust statistical methods to assess mobile platform performance using data collected with the Aclima Inc. mobile air pollution measurement and data acquisition platform installed on three Google Street View cars. They were driven throughout the greater Denver metropolitan area between July 25, 2014 and August 14, 2014, measuring ozone (O), nitrogen dioxide (NO), nitric oxide (NO), black carbon (BC), and size-resolve particle number counts (PN) between 0.3 μm and 5.0 μm diameter. August 6, 2014 was dedicated to parked and moving collocations among the three cars, allowing an assessment of measurement precision and bias. We used the median absolute deviation (MAD) to estimate instrument precision from outdoor, parked collocations. Bias was assessed by measurements obtained from parked cars using the standard deviation of median values over a collocated measurement period, as well as by Passing-Bablok regression statistics while the cars were moving and collocated. For the moving collocation periods, we compared the distribution of 1-σ standard deviations among the 3 cars to the estimated distribution assuming only measurement uncertainty (precision and bias). The distribution of mobile measurements agreed well with the theoretical uncertainty distribution at the lower end of the distribution for O, NO, and PN. We assert that the difference between the actual and theoretical distributions is due to real spatial variability between pollutants. The agreement between the parked car estimates of uncertainty and that measured during the mobile collocations (at the lower quantiles) provides evidence that on-road collocation while parked could be sufficient for estimating measurement uncertainties of a mobile platform, even when extended to the moving environment.
空气污染的移动测绘有潜力在前所未有的空间尺度上提供污染物浓度数据。在移动环境中表征仪器性能具有挑战性，但对于分析和解释所得数据是必要的。我们使用稳健的统计方法，利用安装在三辆谷歌街景车上的Aclima公司移动空气污染测量和数据采集平台收集的数据，评估移动平台的性能。在2014年7月25日至2014年8月14日期间，这些车辆在丹佛大都市区行驶，测量臭氧（O）、二氧化氮（NO₂）、一氧化氮（NO）、黑碳（BC）以及直径在0.3μm至5.0μm之间的粒径分辨颗粒物数量计数（PN）。2014年8月6日专门用于三辆车之间的停放和移动配置，以评估测量精度和偏差。我们使用中位数绝对偏差（MAD）从户外停放配置中估计仪器精度。通过在并置测量期间使用中位数的标准偏差从停放车辆获得的测量值来评估偏差，以及在车辆移动并置时通过帕森斯 - 巴布洛赫回归统计来评估偏差。对于移动配置期间，我们将三辆车之间1 - σ标准偏差的分布与仅假设测量不确定度（精度和偏差）的估计分布进行比较。对于O、NO和PN，移动测量的分布在分布的下端与理论不确定度分布吻合良好。我们断言实际分布与理论分布之间的差异是由于污染物之间真实的空间变异性。停放车辆的不确定度估计与移动配置期间测量的不确定度（在较低分位数处）之间的一致性提供了证据，表明即使扩展到移动环境，停放时的道路配置对于估计移动平台的测量不确定度可能就足够了。
