Glyoxal (CHOCHO) and formaldehyde (HCHO) are short-lived products of the atmospheric oxidation of volatile organic compounds (VOCs). Both are detectable from space by solar backscatter . Isoprene emitted by terrestrial vegetation accounts for about a third of the global source of non-methane VOCs (NMVOCs ) and drives large enhancements of CHOCHO and HCHO in the continental boundary layer . Satellite observations of HCHO have been widely used as a proxy to estimate isoprene emissions , but there are uncertainties related to the HCHO yield from isoprene oxidation and the role of other NMVOCs as HCHO precursors . CHOCHO observations from space could provide a complementary constraint . Here we use CHOCHO and HCHO aircraft observations over the southeast US from the summer 2013 Southeast Nexus (SENEX) campaign , interpreted with the Goddard Earth Observing System chemical transport model (GEOS-Chem), to test understanding of the CHOCHO yield from isoprene oxidation, its dependence on nitrogen oxide radicals (NOx ≡ NO + NO2), and the combined value of the CHOCHO–HCHO pair measured from space to constrain isoprene emissions and chemistry.

Isoprene impacts air quality and climate as a precursor to ozone and secondary organic aerosol (SOA ), and also affects concentrations of hydrogen oxide radicals (HOx ≡ OH + HO2; ) and NOx . Atmospheric oxidation of isoprene by OH takes place on a timescale of less than an hour to produce organic peroxy radicals (ISOPO2). Reaction of ISOPO2 with NO drives production of ozone and of organic nitrates that serve as a reservoir for NOx . At lower NOx levels, ISOPO2 reacts dominantly with HO2 to produce isoprene epoxydiols (IEPOX) via isoprene peroxides (ISOPOOH; ), and from there isoprene SOA . ISOPO2 can also isomerize to generate HOx radicals .

The fate of ISOPO2 determines the production rates and overall yields of CHOCHO and HCHO. Several studies have provided insight on the time- and NOx-dependent yield of HCHO . Under high-NOx conditions, HCHO production is sufficiently prompt that observed HCHO columns can be locally related to isoprene emission rates . This assumption is the basis of many studies that have used satellite HCHO observations to constrain isoprene emissions . HCHO production is much slower under low-NOx conditions, spatially “smearing” the local relationship between isoprene emissions and HCHO columns. This has been addressed by using concurrent satellite data for NO2 columns to correct the isoprene–HCHO relationship or by using adjoint-based inverse modeling to relate HCHO columns to isoprene emissions including the effect of transport .

Isoprene is estimated to account for about ∼50 % of global CHOCHO production , but there is large uncertainty regarding the yield of CHOCHO from isoprene oxidation. Open fires and aromatic VOCs can also be major sources of CHOCHO . Several studies have used the measured CHOCHO–HCHO concentration ratio RGF=[CHOCHO]/[HCHO] as an indicator of the dominant VOC precursors. found higher RGF values (>0.04) from GOME-2 satellite observations in regions where biogenic VOCs are dominant, and lower values where anthropogenic VOCs are dominant. However, the opposite behavior is observed in ground-based studies . Our recent CHOCHO retrieval from the OMI satellite instrument is in better agreement with surface observations of CHOCHO and RGF compared to those from GOME-2 and SCIAMACHY as a result of improved background corrections and removal of NO2 interferences. There remains the question of how observed CHOCHO–HCHO relationships are to be interpreted.

The Southeast Nexus (SENEX) aircraft campaign was conducted over the southeast US in June–July 2013. The aircraft had a detailed chemical payload including in situ CHOCHO and HCHO . Thirteen daytime flights were conducted over the campaign with extensive boundary layer coverage. recently used the SENEX observations to evaluate CHOCHO formation from isoprene in the AM3 chemical transport model (CTM). They found that the AM3 mechanism had closer agreement with observations than the explicit Master Chemical Mechanism v3.3.1 (MCMv3.3.1; ), and suggested that CHOCHO yields from isoprene epoxydiols are underestimated in MCMv3.3.1. Here we take a more rigorous look at potential missing pathways in MCMv3.3.1. In doing so, we present an improved chemical mechanism for CHOCHO formation from isoprene for the GEOS-Chem CTM, and evaluate it against the SENEX observations, including the time and NOx dependence of the CHOCHO yield from isoprene. We discuss the implications of the new mechanism for the interpretation of satellite observations, and present a first validation of the CHOCHO retrieval from the OMI satellite instrument .

Figure shows the observed and simulated median vertical profiles of CHOCHO, HCHO, and NOx concentrations along the SENEX flight tracks. Figure  shows maps of concentrations below 1 km altitude (above ground level) taken as the mixed layer. Here and elsewhere we only include daytime observations (10:00–17:00 LT) and exclude targeted sampling of biomass burning plumes (diagnosed by acetonitrile concentrations above 200 pptv). CHOCHO, HCHO, and NOx were measured by the Airborne Cavity Enhanced Spectrometer (ACES; ), In Situ Airborne Formaldehyde (ISAF) instrument , and the NOAA chemiluminescence instrument , with stated accuracies of 6, 10, and 5 %, respectively.

Simulated median NOx concentrations in the mixed layer are within 10 % of observations, supporting the 50 % reduction in EPA NEI NOx emissions previously inferred from the analysis of SEAC4RS observations by , also included here (Sect. ). Half of isoprene oxidation in the model under the SENEX conditions takes place by the low-NOx pathways (Fig. ). Simulated median CHOCHO and HCHO concentrations in the mixed layer are within 20 % of observations, but the model is too low at higher altitudes. During SENEX the mixed layer was typically capped by a neutrally stable transition layer of shallow cumulus convection which extended up to 3 km , which could suggest that the model underestimates transport via this mechanism. However, the model does not underestimate other isoprene oxidation products in the transition layer, such as MVK + methacrolein (Fig. S8 in the Supplement). Another possible source of CHOCHO in the transition layer is via heterogeneous aerosol oxidation . However, specific aerosol precursors that produce CHOCHO at yields required to match the SENEX observations are currently unknown .

The CHOCHO observations in the free troposphere (> 3 km) have to be treated with caution since they are below the reported instrument precision (32 pptv, ). It is therefore difficult to determine whether the bias is due to a missing CHOCHO source in the model or instrument artifact. Elevated CHOCHO concentrations above the boundary layer have also been observed in previous campaigns over the southeast US , California , and the remote Pacific . There could be a free-tropospheric source missing in the model, but it is unclear what this source could be, and correlative analysis of observed free-tropospheric CHOCHO with other species measured in SENEX offer no insight (r<0.3 for all observed VOCs).

The mixed layer concentrations maps in Fig.  show that the model captures some of the horizontal variability in the observations. The spatial correlation for HCHO is high (r=0.75) as in SEAC4RS (r=0.64, ), and reflects isoprene emission patterns. Correlation for CHOCHO is also relatively strong (r=0.51). Temporally averaged CHOCHO and HCHO concentrations simulated by the model for the SENEX period (background in Fig. ) are much more uniform than those sampled along the SENEX flight tracks because of day-to-day variability in isoprene emissions, mostly driven by temperature .

Figure compares simulated and observed CHOCHO vs. HCHO relationships in the mixed layer, color coded by NOx concentrations. Correlation between the two species is strong. The model better captures the observed slope (0.028 modeled vs. 0.024 observed) compared to the AM3 CTM (0.045 and 0.035 with and without CHOCHO production from δ-ISOPO2+NO, respectively; ). Inclusion of aerosol uptake further reduces the bias to the observed slope (0.026, Fig. S10). On average, CHOCHO is produced more promptly in AM3 compared to GEOS-Chem, which may lead to the higher slope. In the first few hours of oxidation this is due to a higher CHOCHO yield from ISOPO2 isomerization. Beyond the initial stages of isoprene oxidation, CHOCHO is produced faster in AM3 because of the increased fraction of CHOCHO produced from IEPOX over GLYC oxidation (Fig. ).

The strong correlation between CHOCHO and HCHO might suggest that they provide redundant information for constraining isoprene emissions. However, examination of Fig.  indicates higher observed CHOCHO-to-HCHO ratios (RGF) at low-NOx concentrations, not captured by GEOS-Chem. Figure  shows the RGF ratio as a function of NOx below 1 km in the SENEX observations and as simulated by GEOS-Chem. Points are color coded by OH exposure time tOH (Eq. ), derived from PTR-MS observations of the methylvinylketone + methacrolein-to-isoprene ratio following . The median and interquartile RGF values binned in 250 pptv NOx increments are also shown. The observed median RGF values (0.02 to 0.024 mol mol-1) show no significant dependence on NOx, while GEOS-Chem shows a weak dependence.

The observations contain a subset of low-NOx points with higher RGF values (0.03–0.06). The model also produces a subset of enhanced RGF values under low-NOx conditions, although peak RGF values are lower than the observations. In both cases, the enhanced RGF values coincide with short OH exposure times, which are caused by OH titration by isoprene. The high RGF reflects the relatively faster production of CHOCHO than HCHO in the early stage of isoprene oxidation under low-NOx conditions as shown by Fig. . The presence of that population in the observations provides support for fast glyoxal production from the isomerization pathway of isoprene oxidation (Fig. ) that is present in GEOS-Chem but not in MCMv3.3.1. The model may not capture the highest observed RGF values due to uncertainties in the yield of DHDC from isoprene and its photolysis rate, both of which have been estimated based on literature proxies (Sect. S3).

Figure also shows that there is a small subset of points in GEOS-Chem with RGF values less than 0.01, reflecting low CHOCHO values in the model that are not found in the observations where the concentration floor is 0.05 ppbv (Fig. ). There may be a CHOCHO background missing from the model, possibly contributed by monoterpenes; MCMv3.3.1 predicts that the total CHOCHO yield from common monoterpenes is high , and that they produce CHOCHO over a timescale of days (Fig. S11).

Knowledge gained from SENEX enables an improved interpretation of CHOCHO and HCHO column observations from space in isoprene dominated environments. We use for this purpose June–August in 2006 and 2007 observations of CHOCHO, HCHO, and tropospheric NO2 columns from the Ozone Monitoring Instrument (OMI). OMI was launched onboard the NASA Aura satellite in July 2004, and provides daily global coverage in sun-synchronous orbit with an equatorial crossing time of 13:40 LT. The CHOCHO data are from the Smithsonian Astrophysical Observatory (SAO) retrieval described in and hereby referred to as OMI SAO. The HCHO and NO2 data are from the OMI Version 3 product release . Retrievals are in the 435–461 nm spectral range for CHOCHO, 328.5–356.5 nm for HCHO, and 405–465 nm for NO2. We use 2006–2007 data because 2013 data for CHOCHO are very noisy (Fig. S12), possibly because of sensor degradation. The OMI observations are compared to a GEOS-Chem simulation covering the same period, at 2∘ × 2.5∘ horizontal resolution.

Slant columns along the optical path of the backscattered solar radiation are fitted to the observed spectra and converted to vertical columns by division with an air mass factor (AMF) that accounts for the viewing geometry, atmospheric scattering, and the vertical profile of the gas :

AMF=∫0∞w(z)s(z)dz.

Here w(z) is the scattering weight measuring the sensitivity of the retrieval to the gas concentration at altitude z, and s(z) is a normalized vertical profile of gas number density. Here we recomputed the AMFs for the three retrievals using vertical profiles from GEOS-Chem, as it is necessary for comparing simulated and observed vertical columns .

We remove observations impacted by the row anomaly (http://www.knmi.nl/omi/research/product/rowanomaly-background.php), and those with cloud fractions over 20 %. Previous validation of the OMI HCHO retrievals with SEAC4RS aircraft observations revealed a 43 % uniform low bias , corrected in the data shown here.

Figure compares CHOCHO and HCHO vertical columns from GEOS-Chem and OMI, and Fig.  shows spatial correlations over the eastern US. Excellent agreement is found for HCHO, providing an independent test of the correction to the OMI HCHO retrieval inferred from the SEAC4RS data . Since GEOS-Chem can also replicate the CHOCHO–HCHO correlation in the SENEX data, the simulated CHOCHO columns can be used to indirectly validate the OMI CHOCHO observations. CHOCHO from OMI is highly correlated with GEOS-Chem (r=0.81), indicative of the isoprene source. However OMI CHOCHO shows a higher continental background and a factor of 2 weaker enhancement over the southeast US.

suggested that errors in the assumed surface reflectivities affecting the AMFs were an important source of the bias in the OMI HCHO retrievals. CHOCHO retrievals are even more sensitive to surface reflectivity because of the longer wavelengths. previously pointed out that the OMI surface reflectivities used in the standard NO2 retrievals were too high and replaced them with high-resolution (0.05∘ × 0.05∘) reflectivity observations from MODIS to produce the Berkeley High-Resolution (BEHR) OMI NO2 retrieval. CHOCHO and NO2 are retrieved at similar wavelengths so the sensitivity to surface reflectivity should be similar. Figure  (bottom right) shows the mean CHOCHO scattering weights computed from the OMI-SAO and BEHR. The lower BEHR surface reflectivity values result in a lower AMF and hence a higher vertical column (Fig. , bottom left panel). The slope of the regression between GEOS-Chem and OMI CHOCHO columns increases from 0.48 to 0.62, improving but not reconciling the differences.

As pointed out above, SENEX and other observations suggest that GEOS-Chem may be missing a background source of CHOCHO. Integration of the median CHOCHO profile above 2 km in Fig.  shows a negative model bias of 1.3×1014 molecules cm-2, comparable to the continental background intercept in Fig.  (1.9×1014 molecules cm-2). The nonzero intercept may in part reflect an underestimate of CHOCHO concentrations caused by a missing CHOCHO source over the southeast US, such as monoterpenes (Sect. ). The presence of free-tropospheric CHOCHO would further impact the AMF calculation under continental background conditions since the retrieval sensitivity as measured by the scattering weights increases with altitude. Thus the retrieved continental background would be overestimated.

Figure shows CHOCHO vs. HCHO relationships for OMI (using the BEHR scattering weights) and GEOS-Chem, color coded by tropospheric NO2 columns. Individual points are seasonal averages (data points from Fig. ) in order to limit noise. The slope is steeper in GEOS-Chem because the CHOCHO columns are higher. Since GEOS-Chem reproduces the aircraft CHOCHO–HCHO relationship without bias (Fig. ), this is further evidence of bias in the OMI CHOCHO observations. The CHOCHO–HCHO relationship is tight in both OMI (r=0.86) and GEOS-Chem (r=0.99), with no indication of a separate population of low-NOx points with high RGF as there was in the SENEX data. It thus appears from the OMI data that satellite observations of CHOCHO and HCHO in isoprene-dominated environments are redundant. This may reflect the higher NOx levels in 2006–2007 compared to 2013 . However since median RGF shows no significant variation with NOx in the SENEX data (Fig. ), the required temporal averaging of satellite observations is a more likely explanation for the tight correlation. Finer-scale and more temporally resolved data, as will be available from the TEMPO geostationary instrument to be launched in the 2018–2020 time frame , may provide new perspectives of the utility of the CHOCHO retrieval.

We have used aircraft observations of glyoxal (CHOCHO), formaldehyde (HCHO), and related species from the SENEX aircraft campaign over the southeast US together with OMI satellite data to better understand the CHOCHO yield from isoprene and the complementarity of CHOCHO and HCHO observations from space for constraining isoprene emissions. This work includes a first validation of the CHOCHO retrieval from the OMI satellite instrument.

We began with an analysis of the time- and NOx-dependent CHOCHO and HCHO yields from isoprene oxidation in the GEOS-Chem chemical transport model and in the Master Chemical Mechanism (MCMv3.3.1). The GEOS-Chem mechanism features several updates relevant to CHOCHO formation. These include a decrease in the δ-ISOPO2+NO branching ratio leading to prompt CHOCHO production under high-NOx conditions, and a proposed low-NOx pathway for prompt CHOCHO formation by photolysis of a di-hydroperoxide dicarbonyl compound (DHDC) product from (1,5)H-shift isomerization of dihydroperoxy α-formyl peroxy radicals in the ISOPO2 isomerization pathway (Fig. ). GEOS-Chem and MCMv3.3.1 show similar HCHO yields from isoprene, increasing with increasing NOx. CHOCHO yields from isoprene in MCMv3.3.1 show behavior similar to HCHO but GEOS-Chem has a higher yield at low NOx from the ISOPO2 isomerization pathway.

Comparison of GEOS-Chem to the SENEX observations of CHOCHO and HCHO shows good agreement in the boundary layer but a negative CHOCHO model bias in the free troposphere. This could reflect an instrument artifact but may also imply a missing background source in the model. Mixed layer (<1 km) observations show a strong CHOCHO–HCHO relationship that is reproduced in GEOS-Chem and is remarkably consistent across all conditions except at very low NOx where the [CHOCHO] / [HCHO] ratio (RGF) can be unusually high. This reflects prompt formation of CHOCHO under low-NOx conditions, which was missing from MCMv3.3.1 and is now simulated in our updated GEOS-Chem mechanism by DHDC photolysis. A previous model comparison to SENEX showed that MCMv3.3.1 underestimates the CHOCHO yield from isoprene . Our work shows the missing DHDC production pathway can explain approximately 60 % of this underestimate, with the remainder caused by an underestimate of the δ-ISOPO2 branching ratio (3.4 % in MCMv3.3.1 vs. 10 % in GEOS-Chem).

The SENEX observations enable indirect validation of the OMI CHOCHO satellite data using GEOS-Chem as an intercomparison platform. The OMI data show a continental background that is consistent with the SENEX free-tropospheric observations, and an enhancement over the southeast US that is consistent with the isoprene source. However this enhancement is a factor of 2 too low in the OMI data. A partial explanation is that surface reflectivities assumed in the standard OMI retrieval are too high. The satellite data show strong CHOCHO–HCHO correlation consistent with the model and imply that the two gases provide redundant information for constraining isoprene emissions in regions where isoprene is their dominant precursor. This redundancy may reflect the seasonal averaging in the OMI data required to reduce noise. Recent validation of the HCHO satellite data revealed negative retrieval biases , which can be corrected using spatially uniform scaling factors (as done in this study). Since similar biases may exist for the CHOCHO retrieval, the scaled HCHO data should at present be preferentially used as proxy for isoprene emissions. Future geostationary observations from TEMPO will require less temporal averaging and this may reveal the utility of CHOCHO observations for estimating isoprene emissions under low-NOx conditions when isoprene oxidation is titrated.