Characterizing ∆CO:∆NO
x values from observational data can be useful for identifying important sources of CO and NO
x and constraining emissions estimates from these sources. In this study the methods are intended to isolate impacts of the roadway, so it would be expected that the predominant sources would be gasoline and diesel onroad vehicles. The ∆CO:∆NO
x values from this analysis generally fall between 3 and 7 with some outliers. CO:NO
x MOVES emissions ratios from onroad diesel vehicles are around 1 and ratios from onroad gasoline vehicles are between 13 and 14 [
35]. Therefore, values calculated here suggest a mix of diesel and gasoline vehicles. Somewhat higher values at the 300 m site might indicate influence from nonroad gasoline equipment which can have emitted CO:NO
x ratios in the range of 19–22 [
35] and may have been operating either at the nearby golf course or airport. While the study design did not include direct accounting of diesel and gasoline vehicles, we estimated the number of light-duty and heavy-duty vehicles based on collected vehicle length information resulting in fraction of light-duty vehicles between 0.80 and 0.92 at all times during the study period with a mean value of 0.89. Based on these fractions, MOVES estimated that heavy duty vehicles emitted 2.3 times more NO
x than light-duty vehicles emitted, which is consistent with the relatively low observed ∆CO:∆NO
x values. Despite this general consistency, we found little temporal correlation between the estimated fraction of light-duty vehicles with observationally-derived ∆CO:∆NO
x values (r ≤ 0.33 for all method/monitor combinations).
Figure S9 further depicts the lack of correlation between ∆CO:∆NO
x estimates and the fraction of light-duty vehicles. It is possible that the low correlation reported here and shown in
Figure S9 are due to fairly low variability in the light-duty fraction which was near 0.9 for the majority of hours included in this analysis.
To further explore whether these methods provide information that is useful for qualitatively identifying emissions sources we compare ratios from this work to those derived in the literature from observational studies (
Figure 3;
Table S5). ∆CO:∆NO
x values in
Figure 3 are color-coded by the method of data screening that they used to isolate a signal from onroad vehicles. Most of the literature values show a decreasing trend over time since the mid 1980s. This decrease over time is likely due to several rounds of vehicle regulations over the past 20 years which have reduced CO and NO
x emissions from cars and trucks in the United States at different rates as has also been noted by previous studies [
26,
56]. Specifically, results from Parish [
29] suggest that the ∆CO:∆NO
x values decreased from 18.9 in 1989 to 8.9 in 1998 in Boulder Colorado and similarly decreased from 10.2 in 1994 to 6.3 in 1999 in Nashville. More recent measurements in Houston [
34] and Boise [
32] since 2006 have ∆CO:∆NO
x estimates between 5 and 7 and are consistent with the ∆CO:∆NO
x values inferred from cross-road gradients and OLS in this study but are lower than values inferred from orthogonal regressions reported here. The most recent measurements described by Anderson et al. [
23] in Baltimore are substantially higher than the ∆CO:∆NO
x values from this study and are similar to values reported in the literature for measurements made in the late 1990s. The higher ∆CO:∆NO
x values from Anderson et al. [
23] may be caused by chemical aging during the travel of these pollutants due to the fact that measurements were taken in an urban area but not in a near-road environment [
35]. Comparisons with international studies show much higher ∆CO:∆NO
x values at sites in Mexico City [
24] and Sao Paulo [
31] (
Table S5). The higher ratios in Mexico City and Sao Paulo may reflect gasoline vehicles that have less stringent emissions controls, similar to what was seen in the earlier US studies. These comparisons suggest that regression-based ∆CO:∆NO
x values can provide qualitative insight into the types of sources affecting near-road locations and the change in those sources over the past several decades. The data in this study, as well as data used in the studies shown in
Figure 3 are all at least nine years old and do not represent emissions from the current U.S. vehicle fleet. To better understand how this older data might compare to CO:NO
x ratios from onroad vehicles today, we looked at National US emissions estimates available for download from the US EPA [
57] for 2011, 2014, and 2016, as well as data projected to 2023 and 2028 (we looked at EPA’s “ff” case for all years except 2014 for which the “fd” case was the only case available). These data are derived from MOVES simulations. Taking only onroad running vehicle emissions, we found that the 2011 values of 5.8 is predicted to have increased to 6.2 and 6.9 in 2014 and 2016, respectively, and is expected to increase further by 2023 and 2028 to 11.0 and 12.3. It is important to note that while the CO:NO
x ratio is predicted to have increased since 2011, the total emissions of CO and NO
x are both predicted to have declined from this source category. This predicted change in CO:NO
x for the US fleet could be the results of several different factors. First, it is possible that the relative mix of diesel and gasoline vehicles is predicted to change. A larger fraction of gasoline vehicles on the road would result in higher MOVES estimates of CO:NO
x. Another possibility is that newer vehicle emissions technologies taking effect in this time period have resulted in larger reductions in NO
x than in CO relative to emissions control technologies that were dominant in vehicles operating during 2011.