Letter to the Editor on Wormhoudt, J.; Wood, E.; Knighton, W.; Kolb, C.; Herndon, S.; Olaguer, E. 2015. Vehicle emissions of radical precursors and related species observed in the 2009 SHARP campaign; J. Air Waste Manage. Assoc. 65: 699–706

Dear Editor,

We appreciate the opportunity to provide comments on the publication by Wormhoudt and colleagues (Wormhoudt et al., 2015). Their paper makes numerous references to our paper (Rappenglück et al., 2013), which reports traffic related radical precursors and related species as observed and modeled at an urban highway junction in Houston. According to our understanding the publication by Wormhoudt et al. is missing descriptions and explanations at various points and may draw incorrect conclusions.

General Remarks

The motivation of the Rappenglück et al. 2013 paper was to obtain measurements of mobile emissions in order to compare it with the mobile on-road emissions calculated by models MOBILE6 and MOVES2010 available at that time. This included all the various sources for mobile emissions represented in MOBILE6 and MOVES2010, the use of link-based information of the VMT and VMT mix [distribution of vehicle miles traveled (VMT) according to vehicle type and fuel], speed activity estimates and “off-network” emissions such as idling and starting trucks. The research approach was not intended to determine emission factors for individual vehicles as it is never mentioned in the paper.

The location of the ambient measurements was selected in that way that the sampling site (surrounded by freeways, ramps, feeder roads) was exposed to mobile emissions anytime regardless of the wind direction and could be considered as a good representation of the regional traffic on freeways. The location was close enough to traffic related emissions as reflected in the strong relationships of CO vs. NO_x (r² = 0.91) and NO vs. NO_x (r² = 0.96), for instance, but sufficiently well suited to capture more than only plumes related to on-road emissions (e.g. idling and starting trucks as discussed for the HCHO vs. CO relationship due to the existence of parking lots in the area).

In addition, in another study, the Rappenglück group performed the HONO measurements at the University of Houston (UH) Moody Tower which the Wormhoudt et al paper is referring to as “UH-LOPAP” (please note: the caption for Table 1 suggests that all measurements of chemical species listed in the table for the Moody Tower were obtained with UH-LOPAP, which is not correct since only HONO was measured with UH-LOPAP, but not the other trace gases). The Moody Tower site located at about 70 m agl is an excellent site to investigate complex air chemistry processes as shown in numerous papers (e.g. overview papers by Olaguer et al. 2009, Olaguer et al. 2014, and references therein), but it is also a challenging site to study specific emission sources, as they are complex and at times overlapping, and critically depend on wind flow conditions (e.g. Leuchner and Rappenglück, 2010). Very careful data screening is essential, if representative fresh emissions are to be studied. We therefore believe that it is more suitable to perform observations of traffic related emissions in the immediate vicinity of traffic related activities instead, like it was done in our study (Rappenglück et al., 2013).

In fact, the results presented in Wormhoudt et al. for the Moody Tower site reflect more aged air masses as reflected in the overall high NO₂ fraction of NO_x. The authors present Figure 2 as an example for their data analysis. The SHARP data depository for the Moody Tower, which the authors of the Wormhoudt et al paper are referring to in their acknowledgements, includes more data than it is mentioned in their paper. For instance, this data depository shows about 10-37 ppb O₃ and 200-280 ppt PAN during most of the time frame indicated by the two vertical lines in Figure 2 in the Wormhoudt et al. paper, which is an indication of a more aged air mass. Also, the NO₂ fraction of NO_x is almost 100%. It is not before 5:25 am CST that O₃ and PAN decrease to 5 ppb and 125 ppt, respectively, and that the NO₂ fraction of NO_x decreases to about 70%. We thus believe that this data selection may have missed most of the rush hour impacts, which start around 5- 6 am CST and peak around 8 am CST (see Figure 4 about the diurnal variation of VMT in Rappenglück et al., 2013). While it may be true that Wormhoudt et al. considered the maximum morning peak in NO_x, CO, and HONO in their data analysis represented in Figure 2, this may not have necessarily been the traffic rush hour peak, as the location of the monitor is not close to the freeway and it may have been masked by sources closer to the Moody Tower and the increasing boundary layer height after 6 am CST.

It is mentioned that only for the UH-LOPAP results Wormhoudt et al. used the same method of data selection as in Rappenglück et al. 2013, i.e. a correlation of data streams over the entire morning rush hour period. In case they did, then the data of May 24 should have been discarded, simply alone due to the fact that PAN > 50 ppt. Looking into the time frame which Wormhoudt et al. mention in the table caption for Table 1 (i.e. May 17–31, 2009), only May 26 and 27 would have passed this test, but apparently these days were not included among those 11 days selected by Wormhoudt et al. as at some point in the main text they define the days selected as “…. May 17 through 23, 25, and 29 through 31….”.

Regardless what time frame was used, it should be noted that sunrise occurred early in the morning, e.g. on May 24 sunrise was at 5:24 am CST. Rappenglück et al. 2013 did not consider any data sets where global radiation was higher than 10 W m⁻². At this point it remains unclear whether the authors applied the same strict data screening as in Rappenglück et al. 2013, or a different one since no detailed information related to data screening is given in the Wormhoudt et al. paper.

Wormhoudt et al. provide some speculation about whether major freeways located at a considerable larger distance to the Moody Tower or local micro-scale variations in the vehicle mix might have had an impact on the observed CO/NO_x ratio at the Moody Tower. As an example they refer to potential diesel truck emissions at the Hilton Hotel adjacent to the Moody Tower. While vertical mixing at this early morning time is at a minimum, the authors could also include the UH power plant, which is located about 500 m north of Moody Tower in the upwind sector the authors selected as shown in Figure 2. The UH power plant is closer than any freeway in the vicinity of UH campus under these meteorological flow conditions and its plumes can easily reach the Moody Tower platform.

Again, this shows the challenge of retrieving representative information about emission sources at the Moody Tower, more specifically how difficult it is to measure mobile emissions in vicinity of all these other emission sources.

HONO/NO_x

In Figures 3 and 4 Wormhoudt et al. present correlation plots between ΔHONO/ΔNO_x vs. wind speed and ΔHONO/ΔCO vs. wind speed. No values of r² are given, but the relationship is obviously very poor. It is hard to retrieve even qualitative statements. The authors themselves state that “There are many other variables involved which influence the ratios from day to day, including wind direction and distance to the nearest downwind highway.” This statement also indicates that the authors did not perform an adequate wind direction screening to define potential upwind highways. Regardless, the figures suggest decreasing ΔHONO/ΔNO_x and ΔHONO/ΔCO ratios with increasing wind speed. At certain wind speed thresholds these ratios would approach zero indicating no HONO emission at all, which is unreasonable. We reiterate our statement that the Moody Tower is not the best site to characterize specific emission sources, which are not located in its immediate vicinity, like freeways. Apart from that, it remains unclear whether Wormhoudt et al used exactly the same criteria for calculating the HONO vs NO_x and HONO vs CO relationships using the UH-LOPAP as we did in our paper for the Galleria site (Rappenglück et al., 2013).

Wormhoudt et al refer to a recent measurement of diesel-powered generator exhaust by Lee et al. 2011 using the cw-TILDAS instrument, yielding a HONO/NO_x value of 0.0082 ± 0.0005 kg/kg. Unfortunately, Lee et al. 2011 do not specify the type of generator. However, it is known that non-road diesel engines are not regulated by the US EPA and are exempt from fuel requirement and exhaust gas aftertreatment and as a consequence there can be large differences in emissions between different engines and operating modes (Watson et al., 2008). In terms of traffic related HONO emissions the recent publication by Xu et al. 2015 appears to be more relevant. They report HONO/NO_x ratios in fresh plumes (NO/NO_x > 0.80), with the range of 0.5-1.6% for conditions in Hong Kong. This supports our suggestion that it may be unreasonable to assume a constant HONO/NO_x ratio of 0.8% in traffic emissions, regardless of the type of vehicle and its driving mode as currently adopted by the US EPA in the MOVES model, and we would like to recommend the US EPA to review this approach. At this point we also would like to emphasize a side aspect that may also be of relevance in air quality modeling that was documented in Rappenglück et al. 2013, but not considered in Wormhoudt et al., which is the increased NO₂ fraction of NO_x in traffic emissions.

HCHO/CO

The caption for Table 1 from the Wormhoudt et al. paper indicates that for the Moody Tower a data set within the time frame May 17-31, 2009 was taken. Earlier in the paper specific 11 days within this time frame are mentioned explicitly, which were used for Table 1: May 17 through 23, 25, and 29 through 31.

From the discussion in the text related to HCHO/CO it can be deduced that Wormhoudt et al. used two criteria to discriminate against industrial emissions: (1) SO₂ > 0.6 ppb and (2) no east wind. It is unclear what it is actually meant exactly by east wind. According to previous work at the Moody Tower (Rappenglück et al., 2010; Leuchner and Rappenglück, 2010; Czader et al., 2013) it appears reasonable to assume a wind sector rather than one specific local wind direction.

In particular the authors refer to the days April 21, 22 and May 12, 18, 24, 27, and 29. The question arises why the authors refer to these dates as only May 18 and 29 fall into the 11 days used for Table 1. Another question is why not the same days were used for the other ratios. If industrial emissions might have an impact on HCHO/CO, it appears reasonable that the same emissions might also have an impact on the other ratios listed in Table 1. In addition, looking into the SHARP data set it seems that about 5 of those 11 selected days show SO₂ > 0.6 and winds from ESE to NE in the 4–8 am CST time frame.

Emission Factors

As explained above, the motivation of the paper by Rappenglück et al. 2013 was to obtain a very realistic observation of mobile emissions in order to compare it with the mobile emissions calculated by the mobile models MOBILE6 and MOVES2010 available at that time. As the emission models calculate emissions in units of mass, we decided to compare the observations with model results using emission ratios. In addition, we want to clarify that the emission factors used by the mobile models are expressed in units of grams/mile if the vehicle is moving and grams/hr or grams/vehicle if the vehicle is not moving, not in terms of fuel based emissions. Nevertheless, the research approach for the Rappenglück et al. 2013 paper was not intended to determine vehicle emission factors.

However, we do agree with Wormhoudt et al. that there are broad ranges of CO emission factors found in literature. They critically depend on the year the observations were done and the location. They do also depend on the age of the vehicles. Wormhoudt et al. cite McDonald et al. 2013 who tabulate 53 measurements of CO on-road emission factors for gasoline vehicles. Their table shows strikingly how significantly CO emission factors decreased over the last years. For instance, while in July/August 2006 the CO emission factor was 24.0 g/kg(fuel) determined in the Caldecott Tunnel, it was 14.3 g/kg(fuel) in July 2010 for the same location. The reported age of the vehicles was 6.3 years (2006) and 6.4 years (2010). On the other hand in August 2010 other measurements made in Van Nuys reported a CO emission factor of 20.9 g/kg(fuel) (Bishop et al., 2012). For tunnel measurements in Van Nuys at the same time a CO emission factor of 21.3 g/kg(fuel) (Fuijita et al., 2012) was determined as listed by McDonald et al. 2013. In the latter two cases the reported age of the vehicles was 9.1 years. McDonald et al. 2013 also show that the mean light-duty vehicle fleet was on the order of 2 years older in California than in Houston. We suspect the average vehicle age at the Galleria site might be less than for the measurements reported by Wormhoudt et al. for the Washburn tunnel, for instance, as this tunnel is located in the industrial area of Houston. For this reason we think that the measurements presented for the Washburn tunnel and “on-road Houston plumes” suggest an either different vehicle age mix or VMT mix or a combination of both than for the Galleria site.

Wormhoudt et al. also cite Dallmann et al. (2012, 2013), who performed measurements of vehicle emissions in the Caldecott tunnel in 2010. Dallmann et al. 2012 reported fleet-average emission factors for heavy-duty diesel trucks of 8.0 ± 1.2 g/kg(fuel) for CO and 28.0 ± 1.5 g/kg(fuel) for NO_x. Dallmann et al. 2013 reported fleet-average emission factors for light-duty gasoline vehicles of 14.3 ± 0.7 g/kg(fuel) for CO and 1.90 ± 0.08 g/kg(fuel) for NO_x. In the Rappenglück et al. 2013 paper we did not calculate emission factors. However, here we provide some estimate. Considering these latest emission factors by Dallmann et al. (2012, 2013) and assuming a diesel-gasoline split for the freeway during 6–9 am CST as shown in Table 4 (5% diesel and 95% gasoline driven vehicles) in Rappenglück et al. 2013, this would yield a CO emission factor of about [(0.05*8.0 g/kg) + (0.95*14.3 g/kg)] ≈ 14.0 g/kg(fuel) and a NO_x emission factor of [(0.05*28.3 g/kg) + (0.95*1.90 g/kg)] ≈ 3.22 g/kg(fuel). These emission factors are in reasonable agreement with our observations from the whole fleet at the Galleria site, which would yield an emission factor of 10.4 ± 0.6 g (CO)/kg (fuel) and 2.82 ± 0.07 g (NO_x)/kg (fuel) considering the observed emission ratios for that area. It might even be still closer, assuming that the vehicle fleet in Houston is 2 years younger than in California according to McDonald et al. 2013. The corresponding values reported by Wormhoudt et al. for the Washburn tunnel and the “on-road Houston plumes,” for which no VMT mix information is given, deviates significantly from the Dallmann et al. (2012, 2013) and our values, which may be the most representative average values of today’s traffic fleet.

We would like to add that Wormhoudt et al. make the following statement: “Indeed, Rappenglück et al. 2013 tabulate the maximum values of excess CO and CO₂ measured at the Galleria site, and converting those values yields an emission factor for the highest signals observed of 18 g(CO)/kg(fuel).” It was unclear to us where Wormhoudt et al. obtained this information from, but from a private communication through JAWMA we found out that Wormhoudt et al. got this information from the Table 3 in our paper. This is clearly a misinterpretation since the data in Table 3 is unscreened data. Table 5 in Rappenglück et al. 2013 tabulates the values for the ratios of the screened data, from where emission factors should be calculated, only. Nevertheless, we would like to emphasize that the point of the Rappenglück et al. 2013 paper is to indicate that although the MOVES model shows improvements over MOBILE6, still the emission ratios of very important species are off and would need to be adjusted in the model.

Sincerely,