COVID-19 stay-at-home orders, with reduction in automotive traffic, appear to reduce NO2 concentrations in near-road urban settings, as would be expected from the known role of traffic in primary emissions. The COVID crisis is not a practical or meaningful example of anti-pollution measures, but inadvertently presents a stark test of how human behavior impacts environmental pollutants. The environmental consequence of changing behavior may be observed in near-real-time with analysis of public, local datasets.
Nitrogen Dioxide (NO2) is a pervasive environmental air pollutant linked to short-term respiratory irritation and development of respiratory disease with long-term exposure (Heroux et al., 2015; Brunekreef and Holgate, 2002; see also information collected at https://www.epa.gov/no2-pollution). Fossil fuel burning directly emits trace NO2 to the atmosphere, but this is only one component of a complex atmospheric cycle in which NO2 is created or destroyed by reaction with other gases and solar radiation (Anttila et al., 2011; Crutzen, 1971). Meteorological conditions may change the rate at which localized concentrations of NO2 mix on a regional scale, and directly impact nitrogen oxide concentrations through sunlight and rainfall. One must take care attributing rising or falling NO2 to any single factor. Over a long period of observation, those factors that may cause short-term changes average out.
Here in hourly measurements dating back to January, 2016, we see that mean NO2 concentrations measured at 2 near-freeway locations (Seattle 10th and Weller, Tacoma S 26th St.) respond to the increased traffic of the workweek, and show a two-peak diurnal pattern characteristic of morning and evening rush-hours. Past research in the peer-reviewed literature has shown similar traffic-related patterns (e.g., Anttila et al., 2011; Kendrick et al., 2015). With abrupt disruption to work-related traffic, local stay-at-home orders mitigating the spread of COVID-19 inadvertently test the significance of primary automotive NO2 emissions.
In showing that the time-window of the stay-at-home order exhibits suppressed NO2 concentrations, we may compare that window to all comparable timespans (beginning with the same day of the week, and covering the same number of days). For the Seattle example, a 29-day window has 274 comparisons in the dataset. The comparison is flawed to the extent that NO2 concentrations are impacted by long-term secular and seasonal trends unrelated to the COVID mitigation measures. We may use the full dataset to construct a model of “expected” NO2 concentrations by successively subtracting 3 components from the observed data: a linear regression line fit to the full dataset, a 31-day “monthly” median across the 365-day calendar year, and a weekday median. The model, subtracted from the observed measurements, leaves a residual “detrended” dataset for which values above and below 0 respectively represent increased and decreased NO2 relative to “normal.”
The target stay-at-home window may then be compared to the equivalent time ranges. We have done this for the 33-day span from 3/22 to 4/23 in New York, at a Bronx air monitoring station, and for the 32-day span from 3/23 to 4/23 in Seattle and Tacoma. With all windows, we calculate the proportion of days with detrended values less than 0, the median difference in concentration (vs. the model) and the mean difference. We then calculate how many of the comparison windows have more negative days, lower or equal median, and lower or equal means relative to the target. If the stay-at-home order has no impact on NO2 concentrations, we would expect these measurements of the target window to fall within the typical range of the comparison windows. The proportion of comparisons (P-value) with lower values therefore provides an estimated probability of such low NO2 concentrations arising through random selection of an equivalent date range.
|neg. days||median||mean||neg days||median||mean||avg.|
For each station, the stay-at-home window exhibits lower NO2, relative to values expected by the model, than 99% or more of comparable timespans in the dataset. Although many factors may drive NO2 concentrations so low on short time scales, this is strong evidence that decreasing traffic due to COVID mitigation measures has reduced atmospheric pollution.
Anttila, P., Tuovinen, J., Niemi, J.V., 2011, Primary NO2 emissions and their role in the development of NO2 concentrations in a traffic environment. Atmospheric Environment 45: 986-992.
Brunekreef, B., Holgate, S.T., 2002. Air pollution and health. Lancet 360: 1233-42.
Crutzen, P.J., 1979, The Role of NO and NO2 in the chemistry of the troposphere and stratosphere. Annual Review of Earth and Planetary Sciences 7: 443-472.
Héroux, M., Anderson, H.R., Atkinson, R., Brunekreef, B., Cohen, A., Forastiere, F., Hurley, F., Katsouyanni, K., Krewski, D., Krzyzanowski, M., Kunzli, N., Mills, I., Querol, X., Ostro, B., Walton, H., 2015, Quantifying the health impacts of ambient air pollutants: recommendations of a WHO/Europe project. International Journal of Public Health 60: 619-627.
Kendrick, C.M., Koonce, P., George, L.A., 2015, Diurnal and seasonal variations of NO, NO2 and PM2.5 mass as a function of traffic volumes alongside an urban arterial. Atmospheric Environment 122: 133-141.