Meteorological detrending of primary and secondary pollutant concentrations: Method application and evaluation using long-term (2000–2012) data in Atlanta

Abstract

The effectiveness of air pollution regulations and controls are evaluated based on measured air pollutant concentrations. Air pollution levels, however, are highly sensitive to both emissions and meteorological fluctuations. Therefore, an assessment of the change in air pollutant levels due to emissions controls must account for these meteorological fluctuations. Two empirical methods to quantify the impact of meteorology on pollutant levels are discussed and applied to the 13-year time period between 2000 and 2012 in Atlanta, GA. The methods employ Kolmogorov-Zurbenko filters and linear regressions to detrended pollutant signals into longterm, seasonal, weekly, short-term, and white-noise components. The methods differ in how changes in weekly and holiday emissions are accounted for. Both can provide meteorological adjustments on a daily basis for future use in acute health analyses. The meteorological impact on daily signals of ozone, NO x , CO, SO 2 , PM 2.5 , and PM species are quantified. Analyses show that the substantial decreases in seasonal averages of NO x and SO 2 correspond with controls implemented in the metropolitan Atlanta area. Detrending allows for the impacts of some controls to be observed with averaging times of as little as 3 months. Annual average concentrations of NO x , SO 2 , and CO have all fallen by at least 50% since 2000. Reductions in NO x levels, however, do not lead to uniform reductions in ozone. While average detrended summer average maximum daily average 8 hour ozone (MDA8h O 3) levels fell by 4% (2.2 ± 2 ppb) between 2000 and 2012, winter averages have increased by 12% (3.8 ± 1.4 ppb), providing further evidence that high ozone levels are NO x-limited and lower ozone concentrations are NO x-inhibited. High ozone days (with MDA8h O 3 greater than 60ppb) decreased both in number and in magnitude over the study period.

Estimate Contribution of Prescribed Rangeland Burning in Kansas to Ambient PM2.5 through Source Apportionment with Unmix Receptor Model

“…From 2005 to 2014, S4 showed a decreasing trend, which may be attributed to long-term reductions in regional SO 2 emissions due to rules such as the Clean Air Nonroad Diesel Rule and the Acid Rain Program for utilities. Similar long-term decreasing trends of atmospheric sulfate in the U.S. have been reported in other studies and have been linked to the implementation of a number of regulations and associated controls on sulfur emissions in the U.S. (Hubbell et al, 2009;Henneman et al, 2015).…”

Ozone Production Efficiencies at Rural New York State Locations: Relationship to Oxides of Nitrogen Concentrations

“…Meanwhile, part of Henneman, Shen, et al (2017) examined observed OPEs (ΔO 3 /ΔNO z ) in the southeast United States from 1996 to 2015 using data from the same SEARCH sites as Blanchard and Hidy (2018). At each SEARCH site, the authors applied a long-term meteorological detrending of observed [O 3 ] concentrations to approximate the upper 20th percentile for photochemical state or PS*, a parameter that the authors claim (1) describes daily photochemical activity and (2) is independent of emissions (Henneman et al, 2015;Henneman, Chang, et al, 2017;Henneman, Shen, et al, 2017). Then, Henneman, Shen, et al (2017) estimated observed OPEs over the 20-year period using daily 14:00-15:00 local time averages of NO x , NO y , and O 3 for days corresponding to the upper 20th percentile for PS*.…”

Assessing the impact of Clean Air Action Plan on Air Quality Trends in Beijing Megacity using a machine learning technique

“…Statistical analysis of ambient air quality data is another commonly used method to decouple the meteorological effects on air quality (Henneman et al, 2017;Liang et al, 2015), including Kolmogorov-Zurbenko (KZ) filter model and deep neural network (Wise and Comrie, 2005;Comrie, 1997;Eskridge et al, 1997;Hogrefe et al, 2003;Gardner and Dorling, 2001). But they usually gave a poor fitting, suggesting a poor performance of the KZ filter model, or did not allow us to investigate the effect of input variables in neural network models (therefore it is referred as a "black-box" model) (Gardner and Dorling, 2001;Henneman et al, 2015). More recently, new approaches based on classification trees are being developed, which are suitable for air quality weather detrending, including the boosted regression trees (BRT) and random forest (RF) algorithms (Carslaw and Taylor, 2009;Grange et al, 2018).…”