Measurements of atmospheric ethene by solar absorption FTIR spectrometry

. Atmospheric ethene (C 2 H 4 ; ethylene) amounts have been retrieved from high-resolution solar absorption spectra measured by the JPL MkIV interferometer. Data recorded from 1985 to 2016 from a dozen ground-based sites have been analyzed. At clean-air sites such as Alaska, Sweden, New Mexico, or the mountains of California, the ethene column was always less than 10 15 molec.cm -2 and therefore undetectable. In urban sites such as Pasadena, California, 10 ethane was measurable with column amounts of 20x10 15 molec.cm -2 observed in the 1990's. Despite the increasing population and traffic in Southern California, a factor 3 decrease in ethene column density is observed over Pasadena in the past 25 years, accompanied by a decrease in CO. This is likely due to Southern California's increasingly stringent vehicle exhaust regulations and tighter enforcement over this period. here long-term remote sensing measurements of C 2 H 4 in the lower troposphere, where the vast majority of C 2 H 4 resides, by ground-based MkIV observations. We also present MkIV balloon measurements of C 2 H 4 in the upper troposphere. Distribution, magnitudes, reactivities, ratios and diurnal patterns of volatile organic compounds in the Valley of Mexico the MCMA 2002 2003


Ground-based MkIV Results
is blacked out at 948.25 cm -1 . There are also eight CO 2 lines (orange) in this window with depths of 40-60%, one of which sits directly atop the C 2 H 4 Q-branch at 949.35 cm -1 . These CO 2 lines are temperature sensitive, having ground-state energies in the range 1400 < E" <1600 cm -1 . It is not possible to clearly see the C 2 H 4 absorption in Fig. 1, and so Fig. 2 zooms into the Q-branch region. The lower panel reveals that the peak C 2 H 4 absorption is less than 1% deep and strongly 100 overlapped by CO 2 . It is also overlapped by absorption from H 2 O, SF 6 , NH 3 , N 2 O, and solar OH Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2017-403 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 8 June 2017 c Author(s) 2017. CC BY 3.0 License. lines. NH 3 absorption lines exceed 1% in this window on this particular day but do not overlap the strongest part of the C 2 H 4 Q-branch. The SF 6 ν 3 Q-branch at 947.9 cm -1 also exceeds 1% but fortunately does not overlap the C 2 H 4 Q-branch. The SF 6 R-branch, however, underlies the C 2 H 4 Q-branch with about 0.3% absorption depth. The lower panel shows the same spectrum fitted 105 without any C 2 H 4 absorption. This causes a ~0.5% dip in the residuals around 949.35 cm -1 and an increase in the overall rms from 0.245 to 0.260%. The 0.5% dip in the residuals is weaker than the 0.9% depth of the C 2 H 4 feature in the upper panel because the other fitted gases have adjusted to try to compensate for the missing C 2 H 4 . Their inability to completely do so supports the attribution to C 2 H 4 .
Given the severity of the interference, especially the directly-overlying 60%-deep CO 2 line, we were at first skeptical that C 2 H 4 could be retrieved to a worthwhile accuracy from this window, or any other. But given the good quality of the spectral fits, we nevertheless went ahead and analyzed the entire MkIV ground-based spectral dataset, consisting of 4200 spectra acquired on 1090 different days over the past 30 years. Figure 3. Averaging kernels (upper panels) and a priori profiles (lower panels) pertaining to the ground-based C 2 H 4 retrieval illustrated in Figs. 1 and 2 The solid lines in Figure 3 shows the averaging kernel (upper panels) and a priori profile 125 (lower panels) pertaining to the C 2 H 4 retrieval illustrated in Figures 1 and 2. The kernel represents the change in the total retrieved column due to the addition of one C 2 H 4 molecule.cm -2 at a particular altitude. In a perfect column retrieval, the kernel would be 1.0 at all altitudes, but in reality the retrieval is more sensitive to C 2 H 4 at high altitudes than near the surface, as is typical for a profile-scaling retrieval of a weakly absorbing gas. The a priori vmr profile has a 130 value of 500 ppt at the surface, dropping rapidly to 10 ppt by 5 km altitude. An even larger fractional drop, from 10 to 0.5 ppt occurs in the lower stratosphere between 15 and 21 km. The slight kink in the averaging kernel (solid line) over this same altitude range is due to this large drop in vmr. Since 99% of the C 2 H 4 lies in the troposphere, the stratospheric portion of the averaging kernel is of academic interest only for total column retrievals.

135
A major uncertainty in the retrieved column amounts is likely to be the smoothing error, which represents the effect of error in the a priori vmr profile. If the averaging kernel were perfect (i.e., 1.0 at all altitudes) this wouldn't matter, but in fact the C 2 H 4 kernels vary from 0.96 at the ground to 1.4 at 40 km altitude. To investigate the sensitivity of the retrieved column to the assumed a priori profile, we also performed retrievals with a different a priori vmr profile in 140 which the C 2 H 4 vmr profile had been halved in the 0-4 km altitude range and increased in the stratosphere, as depicted by the dashed line in Figure 3. The resulting change in the retrieved C 2 H 4 column was less than 2% with no discernable change to the rms fitting residuals, which are dominated by the interfering gases. This smallness of the C 2 H 4 column perturbation is a result of the averaging kernel being close to 1.0 at the altitudes with the largest a priori vmr errors (0 to 3 145 km). Note that only errors in the shape of the a priori vmr profile affect the retrieved columns in a profile scaling retrieval. Figure 4 shows the resulting MkIV ground-based C 2 H 4 columns from a dozen different observation sites. The plot is color-coded by the pressure altitude of the site. This was preferred over geometric altitude to prevent all the points from a given site piling up at exactly the same x-150 value. The pressure altitude varies by up to ±1.5% at the high altitude sites, which is equivalent to ±0.2 km. Only points with C 2 H 4 uncertainties <1x10 15 were included in the plot, representing 95.7% of the total data volume. One day (out of 255) at Barcroft (3.8 km altitude) was omitted from the plotted data because it had abnormally high C 2 H 4 , as well as other short-lived gases -clearly a local pollution event.  At JPL the C 2 H 4 column is highly variable. JPL is located at the Northern edge of the Los Angeles conurbation, and so when winds are from the Northern sector, or strong from the ocean, pollution levels are much smaller than during stagnant conditions. This is seen in the large

180
The CO also shows a substantial decline since the 1990's at JPL, although not as dramatic as that of C 2 H 4 since CO never falls below 1.5x10 18 molec.cm -2 at JPL, even under the cleanest conditions. Figure 5a shows a tight correlation between C 2 H 4 and CO at JPL (green) suggesting a common local source for both. We believe this common source to be mainly vehicle exhaust and the declines in CO and C 2 H 4 to be a result of increasingly stringent requirements on vehicle 185 emissions imposed by the US Environmental Protection Agency (EPA; e.g., the 1990 Clean Air Act) and the California Air Resources Board (CARB, LEV2) over the past decades and stronger enforcement thereof (e.g., smog checks).  /c/d also shows correlations between C 2 H 4 and other gases: C 2 H 2 and C 2 H 6 , and H 2 CO for all the MkIV measurements. These correlations are not as tight as that with CO, due to C 2 H 2 and C 2 H 6 have other sources. For example, C 2 H 6 also comes from natural gas leaks. The fact that these trace gases are much less abundant than CO means that their measurements are noisier, which also degrades the correlations.  Figure S1 plots the gas column relationship for the JPL ground-based data only, each panel containing ~1700 observations. The decreases in the CO, C 2 H 2 , C 2 H 4 and H 2 CO since the 1990's are evident by the lack of red points in the upper right of the panels plotting these gases. C 2 H 6 seems not to have decreased significantly as is evident from the large values of the red 205 points in the third row. In fact, on November 10, 2015, we observed a factor 2-3 enhancement of the C 2 H 6 column as a result of JPL being directly downwind of the Aliso Canyon natural gas leak on that day [Conley et al., 2916]. Although this event was associated with a 2.5% enhancements of column CH 4 (not shown here), there were no enhancements of CO, C 2 H 2 , C 2 H 4 , so these particular C 2 H 6 points (red) in the third row of Fig. S1 protrude upwards from the main clusters. The highest correlations are between CO and C 2 H 2 (0.91) and CO and C 2 H 4 (0.92). The 225 correlation coefficient between C 2 H 2 and C 2 H 4 is only 0.84, probably reflecting the fact that C 2 H 2 and C 2 H 4 are much more difficult (i.e. noisier) measurements than CO. The worst correlation is between C 2 H 6 and H 2 CO (-0.04).
The overall gradient of the C 2 H 4 /CO relationship using all JPL data is 3.8±0.3 ppt/ppb, as in

235
We also looked for ethene in MkIV balloon spectra using exactly the same window, spectroscopy and fitting software (GFIT) as used for MkIV ground-based measurements. The advantage of the balloon spectra is that the airmass is much larger and the solar and instrumental features are removed from the occultation spectra by ratioing them against a high-sun spectrum taken at noon from float altitude.
240 Figure 6 shows a spectral fit to the MkIV balloon spectrum at 6 km tangent altitude measured above Esrange Sweden in Dec 1999.  The balloon results are consistent with the ground-based measurements, confirming that C 2 H 4 exists in measurable quantities only in the polluted PBL, which is generally inaccessible by solar occultation due to the high aerosol content making the limb path opaque. The typical 1-3 280 day lifetime of C 2 H 6 at mid-and low-latitudes implies that it will only be measureable in the free troposphere soon after rapid uplift.   From spectra acquired during one of the most intense of these fires (Jan 1, 2002), Rinsland et al. [2005] retrieved a total C 2 H 4 column of 380±20x10 15 through a dense smoke 300 plume and inferred a huge mole fraction of 37 ppb peaking at about 1 km above ground level.
This retrieval used information from the shape of the Q-branch feature, which was nearly as deep as the overlapping CO 2 line. These C 2 H 4 amounts are 20 times larger than anything seen by MkIV, even from polluted JPL.
Coheur et al. [2007] reported a C 2 H 4 vmr of 70±20 ppt at 11.5 km altitude (their Table 2) 305 in a biomass-burning plume, observed by the Atmospheric Chemistry Experiment (ACE) [Bernath et al., 2005] off the East coast of Africa. Their Fig. 2 shows measured C 2 H 4 exceeding 100 ppt below 8 km. Simultaneous measurement of elevated C 2 H 2 , CO, C 2 H 6 , HCN and HNO 3 confirm their biomass-burning hypothesis.
Herbin et al. [2009] reported zonal-average ethene profiles above 6 km altitude based on 310 global measurements by ACE. Figure 2

315
Herbin also report increasing C 2 H 4 with latitude. Although the ACE zonal means agree with the in situ measurements made during the PEM-West and TRACE-P, these campaigns were designed to measure the outflow of Asian pollution and therefore sampled some of the worst pollution on the planet. So one would expect lower values in a zonal average. Based on the total absence of negative values in any of their retrieved vmr profiles, we speculate that Herbin et al. [2009] 320 performed a log(vmr) retrievial, imposing an implicit positivity constraint. This would have led to a noise-dependent, high bias in their retrieved profiles in places where C 2 H 4 was undetectable.  Figure 9 of Blake et al. [2003]).
Ethene was measured during the HAIPER Pole-to-Pole (HIPPO; Wofsy et al. 2011Wofsy et al. , 2012) mission by the Advanced Whole Air Sampler. Figure 8 plots Washenfelder et al., [2011] also report a factor 6 drop in C 2 H 4 amounts since the September 1993 CalNEX campaign, but note that the 1993 readings occurred during a smog 395 episode, implying higher than normal levels of pollution. This drop is larger than the factor 3 decrease seen in the MkIV column data, but not inconsistent given the sparse statistics together with the large day-to-day variability seen in the MkIV data.
Measurements of ethene in Mexico City ranged between 10-60 ppb, with higher levels in the commercial sectors and lower values in residential areas (Altuzar et al., 2001(Altuzar et al., , 2005Velasco 400 et al., 2007). These are 5-30 times larger than the 2.16 ppb measured by Washenfelder et al. [2011] in Pasadena in 2010.

Discussion
To see whether the ground-based MkIV C 2 H 4 measured in Pasadena was correlated with the airmass origin, we performed HYSPLIT back-trajectories, and computed the amount of time 405 that airmasses arriving 500 m above JPL had spent over the highly populated areas of Los Angeles conurbation. When column C 2 H 4 was plotted versus this time-over-conurbation, the correlation was very poor. Column CO also had a poor correlation. The fact that the C 2 H 4 correlates well with CO tends to discount the possibility that the C 2 H 4 measurements are wrong, since the CO measurements are very reliable. So this implies that either the trajectories are 410 wrong, or that the urban pollution is not a major source of the C 2 H 4 or CO observed by MkIV.
We point out that JPL is located at the foot of the San Gabriel mountains, which rise over 1 km above JPL over a horizontal distance of less than 5 km. This extreme topography might give rise