Reactive organic carbon (ROC, defined as the family of all atmosphere organics in any phase, other than methane 1 ) compounds are emitted into the atmosphere from anthropogenic, pyrogenic, and natural sources [2][3][4] , whereupon they are oxidized primarily by OH, O3, and NO3 to form a suite of products, including particulate matter (PM), ozone, and CO2 5 . At a global scale, models suggest that ROC is the dominant source for the production of these secondary species 1,[6][7] . The first two of these secondary products, PM and O3, are the principal global air pollutants, associated with the premature mortality of over 8 million people each year 8 . All three of these products impact climate, the last two as greenhouse gases, and organic aerosols that, through scattering and cloud formation, act to counter-balance some of the warming from greenhouse gases 9 .
Despite the central place of ROC in tropospheric chemistry, our understanding of its abundance, lifecycle, and ultimate impacts remain highly uncertain. This is largely due to the complexity of ROC, which is estimated to consist of many thousands of compounds 10 , a small subset of which are routinely measured and modeled in the atmosphere. Previous attempts to develop approaches to measure total carbon mass in the gas phase [11][12][13] have lacked the precision necessary to constrain a global budget. The alternative is to try to build a picture of the total mass of carbon by summing up speciated measurements from more routine instrumentation. Such an approach demonstrated that average observed ROC concentrations over North America spanned two orders of magnitude from 4 to 456 µgCm - 3 14 . This approach fails to identify whether carbon closure is achieved, but can provide insight into the abundance of key species of ROC and their fate. Over the last two decades, instrumentation has been developed to routinely detect a wider range of compounds 5 .
The most comprehensive effort to date to sum measurements that span the range of volatility (and thus are thought to achieve mass closure), suggested that compounds that are not typically measured during field campaigns contributed ~1/3 of the mass of ROC in a forested environment in Colorado 15 .
The atmospheric burden of ROC is dominated by long-lived species such as alkanes and acetone, whereas reactivity is dominated by short-lived species such as isoprene 1 . Together, these complementary metrics capture both those species highest in abundance and those most important for chemical processing. Techniques to measure total OH reactivity (OHR, the inverse of the OH lifetime) were developed over a decade ago, and are now routinely deployed in field campaigns (see Yang et al. 16 for a review). OHR is typically dominated by ROC, other than in some urban or remote environments where the bulk of OHR may come from inorganics or long-lived species (e.g. CH4, CO), respectively [17][18][19] . When deployed alongside speciated measurements, OHR measurements can be used to estimate the "missing OHR" not accounted for by routine instrumentation. This missing OHR has been estimated at 30-50% in urban environments, and higher (more than half missing in some cases) in forests 16 . This missing OHR may represent unmeasured emitted species or oxidation products.
While the measurements of OHR provide a powerful framework to assess chemical processing in the atmosphere, it is crucial to acknowledge that the term "missing reactivity" has been applied somewhat fluidly. The "missing reactivity" can be the shortfall between the contribution from measured species and the total OHR measurement (as described above and used in this study), or it can be the difference between modeled OHR and the total OHR measurement. The former having the virtue of a direct constraint, the latter accounting for known unmeasured compounds. When "missing reactivity" is assessed based solely on measurements, it reflects the suite of instrumentation deployed under varying operating conditions in any given campaign (which may, for example, dictate detection limits and the number of species reported). This suite can be more or less comprehensive, and is certainly not consistent across campaigns, challenging our ability to compare "missing reactivity" estimates. The summed reactivity also relies on assumed rate constants, which may differ across studies. Furthermore, the measurement of total OHR is itself operationally defined. For example, some semi-volatile or intermediate volatility compounds (S/IVOC) may not be sampled through an inlet, or chemical reactions may rapidly reform OH, resulting in the measured OH loss being lower than the OHR calculations that do not consider this recycling. These definitional and operational factors may influence the interpretation of OHR results.
Field observations provide a snapshot of the local atmospheric environment and can be used to probe specific sources and atmospheric processes. While field campaigns are often constructed with specific scientific questions in mind, the observational capital may be relevant to additional future lines of inquiry. In this study, we revisit measurements made during two field campaigns over the United States: the Southern Aerosol and Oxidant Study (SOAS) of 2013 and the California Research at the Nexus of Air Quality and Climate (CalNex) campaign of 2010. While numerous studies have explored the processes controlling atmospheric composition for each of these two campaigns, this study applies a new lens to contrast the abundance and role of ROC in these two very different environments.
The CalNex campaign was based in Southern California, a region which has undergone a major air quality clean-up since the 1970s but remains under the influence of over 10 million people who live in Los Angeles County 20 . Southern California represents a modern urban environment, characterized by reductions in vehicle emissions over recent decades 21 , with increasingly important emissions from other source categories including volatile consumer products (VCPs) 22 .
Here we investigate measurements made at the Pasadena, California ground site from May 15-June 16, 2010, from towers located 10m above an empty parking lot (note that the AMS had a separate inlet sampling above a trailer at a height of ~6m). See Table 1 for details and Ryerson et al. 20 for more information. In particular, we include here an expanded set of ROC measurements made by gas chromatography mass spectrometry (GC-MS) based on new peak identification methods [22][23] . A 2D-TAG was deployed during the last 4 days of the campaign, but given the lack of overlap with other measurements, that data is not included here [24][25] . Meteorological conditions varied during the CalNex campaign with both sunny, clear days and cool, overcast days 20 .
SOAS was a part of the larger Southeast Atmospheric Study (SAS) of 2013 26 . The Southeast United States (SEUS) represents a mix of biogenic and anthropogenic sources, with urban centers embedded in large swaths of forests. Air quality regulations have led to declines in anthropogenic emissions of NOx, SOx, and organic aerosol in the region for more than a decade preceding SOAS [27][28] . Here we explore the measurements made at the ground site in Centreville, Alabama, from June 1 to July 15, 2013. The majority of the gas phase measurements analyzed here were made from a tower located 16.5 m above the ground in a clearing near the edge of a forest and well above the forest canopy (Table 2). The aerosol measurements and the gas phase measurements operated by the SEARCH network (see Table 2) were made nearby at ground-level near the center of the clearing. This rural site is located within the Talladega National Forest, a dense mixed forest of pine and broadleaf trees. The summer of 2013 was consistently cooler and rainier than average conditions 26 .
In this study, we revisit the online measurements of ROC, OH reactivity and related trace species measured at Centreville and Pasadena. For both campaigns, all observations are first averaged to an hourly time base and individual ROC species concentrations are converted into units of µgCm - 3 at standard temperature and pressure (STP, 298 K, 1 atm). Given that not all instrumentation operated continuously over the entirety of the campaigns, in particular for SOAS (see Table 2), we only present diurnal averages or 24-hour averages for each campaign. We further note that while much of the same key instrumentation was deployed at both sites, this suite and its operation are not identical, and not all species are observed at concentrations above detection limits at both sites, thus the list of ROC species measured/reported for the SOAS and CalNex campaigns differ.
Figure 1 shows the average ROC mass concentrations measured for the CalNex and SOAS campaigns, highlighting the contributions from different chemical families. Table S1 provides average concentrations for each individual species.
Average ROC concentrations during SOAS total 36.5 µgCsm -3 , with sizeable contributions from primary biogenic species, in particular isoprene (8.6 µgCsm -3 ) and monoterpenes (4.8 µgCsm -3 ).
The top 10 species (isoprene, methanol, acetone, organic aerosol, ethanol, α-pinene, MVK, ethane, other monoterpenes, and ß-pinene) contribute 70% of the mean ROC mass concentrations observed during SOAS. In contrast, CalNex ROC (average 73.8 µgCsm -3 ) is dominated by anthropogenic species, including alkanes (24.9 µgCsm -3 ) and ethanol (9.0 µgCsm -3 ), the latter of which has been highlighted as a product of VCP emissions as well as from mobile sources in the Los Angeles region 22,29 . The top 10 species measured during CalNex (ethanol, acetone, ethane, propane, i-pentane, methanol, organic aerosol, propanol, propionic acid, and n-butane) make up 58% of the mean ROC. Overall, measured ROC is twice as abundant in the atmosphere during CalNex compared to SOAS, consistent with expectations of higher ROC in a "polluted environment". When compared to a previous compilation of ROC concentrations over North America 14 , the ROC reported at CalNex is second only to Mexico City, and exceeds ROC reported in Pittsburgh in 2002. Similarly, ROC levels during SOAS are higher than all non-urban sites reported in Heald et al. 14 , as well as the Manitou Forest site where Hunter et al. 15 report average ROC concentrations of 26.7 µgCm -3 . Despite the substantial differences in ROC abundance and speciation during CalNex and SOAS, we find that organic aerosol contributes ~5% of ROC in both environments.
CalNex. This maximum is driven by isoprene for SOAS (this daytime peak is dampened by the night-time peak in monoterpenes), and OA, acetone, and propane for CalNex. There are modest differences between weekday and weekend ROC, with slightly higher values (<10%) on weekends in both Pasadena and Centreville. The slightly higher weekend concentrations of biogenics such as isoprene during SOAS may be the result of lower NOx and lower OH over the weekend extending the lifetime of these primary species.
We note that our analysis includes only the identified and calibrated S/IVOC (semi-volatile to intermediate volatility, with saturation vapor pressures C * =1-10 5 µg m -3 ) species measured by SV-TAG during SOAS 30 ; these compounds contribute 0.25 µgCsm -3 . A preliminary calibration of all the additional SV-TAG measured compounds with the formulas C15H24 and C15H22 31 , which should provide an upper-bound on sesquiterpene concentrations, indicates that sesquiterpenes contribute less than 100 ngCm -3 throughout the SOAS campaign. It is not clear how much additional mass may be contributed by unmeasured/unidentified S/IVOC species. We also note that a nitrate CIMS was deployed during SOAS which detected an additional 87 compounds which make a negligible contribution to ROC mass concentrations (0.03 µgCsm -3 ) and are not included in further analysis. No online speciated S/IVOC measurements are available during CalNex. Unspeciated gas-phase SVOC concentrations (C * =1 µg m -3 ) estimated by Ma et al. 32 average 2.77 µg m -3 (assuming an OM:OC=1.3, equivalent to ~2.1 µgC m -3 ), or less than 3% of the measured ROC. Zhao et al. 33 report 3-hour integrated IVOC measurements (both speciated and unspeciated, equivalent to C12-C22 n-alkanes) made offline from Tenax glass tubes during CalNex, with a campaign average concentration of primary IVOC of 6.3 µgm -3 (assuming an OM:OC of 1.1 for a C15 n-alkane this gives ~5.7 µgCsm -3 ). This represents an additional 8% in ROC mass concentrations. Measurement and quantification challenges for these samples are discussed in Zhao et al. 33 .
We translate the observed mass concentrations during CalNex and SOAS into calculated OH reactivity (cOHR), which is defined here as Σ(kOH+Xi [Xi]), the sum of OH reactivity contributed by each species (Xi). We note that this includes not only ROC species, but also other key gas phase species that react with OH (CO, O3, NO, NO2, SO2, OH, HO2). We use here reaction rate constants (kOH+Xi) at 298K, taken from JPL kinetics compilation 34 , the IUPAC evaluated kinetic data 35 the GEOS-Chem chemical mechanism v12.7.0 (www.geos-chem.org), or from EPI Suite 36 (see Tables S1 andS2). As in Hunter et al. 15 , we make no effort to adjust reaction rates for ambient temperatures during SOAS and CalNex; this is reasonable for SOAS where average temperatures are similar to 298K, but likely leads to a slight overestimate of cOHR at night during CalNex when temperatures drop to ~288K. Aerosols are not expected to contribute significantly to OHR. For example, following equation 3.38 of Seinfeld and Pandis 37 , we calculate that under moderately polluted conditions (50 μgm -3 , assuming a mean diameter of 200nm) the upper-limit first order rate coefficient for OH loss to particles (given an average molecular velocity of OH of 609 ms -1 and assuming an upper limit for the uptake coefficient of 1) is 0.23 s -1 , equivalent to ~1% of the measured OHR during CalNex and SOAS (see below). It is therefore reasonable to neglect the aerosol contribution to OHR during these campaigns.
Figure 3 shows the average cOHR for CalNex and SOAS. The total cOHR for SOAS is 16.6 s -1 , with a large contribution of 9.4 s -1 from a single species, isoprene, and a further 0.7 s -1 from its immediate oxidation products MVK and methacrolein. Monoterpenes contribute an additional 2.9 s -1 . Over 85% of cOHR at the SOAS site is from ROC species, largely biogenics. By contrast, the average cOHR during CalNex (13.9 s -1 ) is dominated by NOx (4.7 s -1 ), with ROC species contributing 51% of the reactivity. Despite the urban nature of this site, isoprene remains the single largest ROC contributor to reactivity during CalNex (0.9 s -1 ), though the amount is an order of magnitude lower than for SOAS. Washenfelder et al. 38 noted that oak and eucalyptus trees upwind of Pasadena likely represent a local biogenic source during CalNex. The second most important individual ROC contributor to reactivity during CalNex is acetaldehyde, a product of ethanol and alkene oxidation, concentrations of which have increased in ambient air, including over California, in recent decades due to ethanol fuel additives and VCP emissions 22,29 . In general, we find that many of the measured ROC species that contribute to mass concentrations during CalNex (Figure 1) do not substantially impact calculated reactivity (Figure 3), suggesting that OH loss in southern California is not primarily driven by anthropogenic (or biogenic) ROC emissions. The reactivity per mass concentration of ROC is more than 4 times higher for SOAS than for CalNex (0.10 for CalNex vs 0.42 for SOAS). However, we note that despite all of these differences, the total cOHR estimated for CalNex and SOAS are quite similar.
Figure 2b shows that the diurnal cycle for cOHR is even more muted during CalNex than ROC concentrations shown in Figure 2a, given that cOHR from highly-reactive alkenes peaks at night 39 , compensating for the daytime peak driven by isoprene and acetaldehyde. The large contribution of isoprene to reactivity during SOAS enhances the diurnal cycle observed in ROC concentrations at that site (Figures 2a and2b). Weekday and weekend differences in cOHR are negligible (not shown).
During both SOAS and CalNex measurements of total OHR were deployed using laser-induced fluorescence techniques and both datasets were corrected for chemical interferences [40][41] . Thus, these measurements and the cOHR compiled here provide a critical opportunity to contrast the "missing OH reactivity" at two very different sites in the United States. However, as discussed in Section 1, operational differences in measured OHR may remain. Measured OHR is similar at Pasadena and Centreville, with average values of 20.1 and 19.3 s -1 , respectively. Figure 3 shows that 31 and 14% of the measured reactivity is not accounted for by the cOHR (6.3 s -1 for for CalNex and 2.6 s -1 for SOAS). The "missing reactivity" for SOAS is well within the uncertainty of the OHR measurement and rate constants used to estimate cOHR, and may indicate that all the reactivity is accounted for at this site. Kaiser et al. 42 suggest that in their analysis missing OHR during SOAS may be attributed to unmeasured primary emissions (e.g. sesquiterpenes). Some studies [43][44] (though not all 45 ) have shown that sesquiterpene concentrations peak at night this may offer an explanation for Figure 2b which shows a nighttime peak in missing OHR, with little missing OHR in daytime. However, while sesquiterpene measurements made by SV-TAG during SOAS (see Section 3.2) show a pronounced nighttime peak, concentrations were substantially lower than monoterpene concentrations, consistent with previous studies that suggest that sesquiterpene concentrations are relatively low outside of the forest canopy 44 , and thus unlikely to fully explain the missing OHR at night. In Pasadena, however, there may be a larger reservoir of unmeasured ROC species (or inorganics). As shown in McDonald et al. 22 , the ROC measurements during CalNex (including the expanded suite of species measured by GC-MS, see Table 1) capture much of the reactivity from VCP sources, however they also suggest that unmeasured ROC species from VCP sources could contribute additional OHR (see Table S8 of that study). An analysis of comprehensive ROC measurements at Manitou Forest in Colorado during 2011 suggested that previously unmeasured S/IVOCs compounds, contributed ~25% of the OHR at that site. S/IVOC compounds may also make an important contribution to OHR in Southern California and in the SEUS. However, the subset of SVOCs measured by SV-TAG during SOAS contribute only 3% (0.56 s -1 ) of the cOHR. Applying OH reaction rates to the average primary IVOCs concentrations measured offline during CalNex (see Tables S5 and S7 of 33 ), yields an additional 0.33 s -1 of cOHR, suggesting that these IVOC contribute only ~5% of the missing OHR during CalNex. The missing OHR during CalNex appears to be consistent throughout the day (Figure 2b), and therefore is not suggestive of a clear local primary emission signature.
We can use the estimated cOHR to explore ozone formation associated with ROC during CalNex and SOAS. Ozone formation is initiated by the reaction of peroxy radicals with NO, as described by Jacob 46 and applied in Sinha et al. 47 :
with the resulting NO2 photolyzing to produce ozone. If we assume the rapid cycling of radicals in a polluted atmosphere:
therefore,
Figure 2c shows the estimated diurnal pattern in PO3 based on equation 3 for both CalNex and SOAS. As described in Section 3.2, CalNex is characterized by much lower ROC reactivity (7.1 s -1 ) than SOAS (15.5 s -1 ), however OH concentrations during CalNex are about a factor of four higher than during SOAS, with mid-day peaks of 4x10 6 molec cm -3 and 1x10 6 molec cm -3 , respectively. Thus the resulting ozone production rate is about 50% higher during CalNex, with peak early afternoon values of ~9 ppb hr -1 , compared to ~6 ppb hr -1 for SOAS. We note that there is little weekday-weekend difference in measured OH during CalNex in contrast with estimated OH concentrations based on simple-steady state models 48 ; constraining these estimates to measured OHR yields better agreement with measured OH 41 .
Figure 2d shows the mid-day increase in ozone concentrations observed during CalNex and SOAS.
CalNex is characterized by ~40 ppb increase over 7 hours, with a more modest increase of ~15 ppb over the same time interval for SOAS. These daytime increases likely reflect local production integrated over the immediate upwind region, as well as the mixing down of the residual layer from the previous day. The diurnal profile of typical ozone deposition velocities for urban and forested land classes 49 produce a small sink of 0.3-0.6 ppb hr -1 peaking mid-day; this is negligible compared to the daytime ozone production rates calculated at Pasadena and Centreville. The observed ozone increase in Southern California is roughly consistent with the ozone production rate calculated using equation 3, however this relationship appears to substantially overestimate the observed daytime increase in ozone in the SEUS. This is true even when the calculated ozone production rate is tempered by the estimated ozone dry deposition flux over forested land.
The lower ozone formation during SOAS is the result of NOx limitations at this forested site. This can be illustrated by a metric (ϑ=OHRNOx/OHRROC) suggested by Kirchner et al. 50 to assess ozone production regimes. The observations at CalNex produce an average ϑ of 0.52 which is well above the 0.2 boundary to indicate VOC-limitation. Average early afternoon ϑ is lower on weekends due to lower NO, but stays above 0.35 consistent with VOC-limitation, whereas using an alternate metric Griffith et al. 41 suggest that observations during afternoons on weekends are NOx-limited during CalNex. We note that if we assume that all of the missing OHR comes from unmeasured ROC species, the re-calculated ϑ briefly dips below the 0.2 boundary on weekend afternoons suggesting a transition from VOC to NOx limited. Conversely, the average ϑ at SOAS exhibits a strong diurnal cycle where ϑ drops below the NOx limitation boundary (0.01) from 9am-4pm on average (on both weekday and weekends), and increases over night (campaign average ϑ of 0.012).
Thus, despite very high reactivity from ROC in the SEUS, the lower daytime OH and NOx depress the ozone production potential of these organics during SOAS.
We note that the OHR from IVOC during CalNex discussed in Section 3.2 could contribute an additional ~5% to the ozone production rate calculated using equation 3; well within the uncertainty associated with estimating the ozone production rate from its observed daytime increase in Figure 2c. However, if the remaining missing OHR during CalNex were attributed to ROC, the resulting calculated ozone production would substantially exceed the observed rise in ozone concentrations. This suggests that either (1) some of the missing OHR may be the result of non-ROC species, or (2) conditions during CalNex may not be uniformly NOx-saturated (as assumed in equation 3), reducing ozone production efficiency, or (3) that there are larger, unaccounted for, sinks of ozone during CalNex. Deployment of instrumentation at an upwind site, as well as an O3 reactivity instrument 51 , alongside the suite of instrumentation deployed in Pasadena during CalNex, could provide further insight into the role of ROC in controlling ozone concentrations in the LA basin.
ROC oxidation leads to the formation of secondary organic aerosol (SOA), a substantial and sometimes dominant contributor to fine particulate matter 52 . The vast potential number of precursors which produce SOA at varying timescales, coupled with gaps in our understanding of this chemistry, combined with the addition of primary organic aerosol, complicate any assessment of regional SOA formation. Here we explore the instantaneous SOA formation potential of ROC at Centreville and Pasadena.
To estimate the SOA formation potential (the potential amount of SOA formed from the measured precursors), we multiply mass concentrations of ROC with carbon yields (YC) based on the approach of Hunter et al. 15 . A subset of yields for more well-studied SOA precursors (aromatics, isoprene, monoterpenes) are taken from Hunter et al. 15 based on chamber studies. The remaining yields are estimated based on the vapour pressure of each species taken from EPI Suite 36 , converted to the saturation vapor pressure (C*), and then to yield using the functional form of Donahue et al. 53 (as given in caption of Table S1).The resulting yields vary from <1% to 19% (see Table S1). We employ this approach as it allows us to systematically estimate the SOA yields for all considered species, however, we note that the uncertainty on these estimated yields is high;
Hunter et al. 15 suggest a factor of three. While a detailed review of all estimated yields is beyond the scope of this study, we note that the systematic application of the relationship between vapor pressure and carbon yield 53 leads to large estimated yields for alcohols (YC=3-5%); however, there is no laboratory evidence to date to suggest that the oxidation of alcohols yields appreciable SOA, in fact some previous CalNex studies assume that the SOA yield from alcohols is zero 22,54 . Thus, in what follows, we assume zero SOA potential for alcohols (SOA potentials for these species as calculated using the approach of Donahue et al. 53 are given in brackets in Table S1). In addition,
we note that the SOA potential metric as used here only captures SOA formed in the gas-phase. S5 andS7) to the reported average mass concentrations and assuming an OM:OC of 2 for the SOA formed would give an additional 0.80 µgCsm -3 from IVOC. This is equivalent to more than half of the SOA formation potential of all the online measured species, consistent with the major role that IVOCs play in SOA formation in southern California, as suggested by Hayes et al. 54 .
The SOA formation potential at SOAS (0.90 µgCsm -3 ) is a little more than half of that during CalNex. More than half of this is from isoprene and monoterpenes (0.15 and 0.36 µgCsm -3 ). The dominance of monoterpene SOA is consistent with the analysis of observed SOA concentrations in the SEUS by Zhang et al. 55 . This is considerably less than the SOA formation potential of 1.4
µgCsm -3 reported for a ponderosa forest site 15 , however approximately half of the SOA formation potential assessed at that site was estimated to come from previously unmeasured S/IVOCs, some of which are not measured during SOAS. Sesquiterpenes, with their high SOA yields 56 , could also contribute to SOA in this region, however, as noted in Section 3.2, concentrations of these species are small during SOAS and even assuming 100% yield would contribute less than 0.1 µgCsm -3 to the SOA potential. Using an apportionment analysis of FIGAERO-CIMS, Lee et al. 57 also rule out any substantial contribution of sesquiterpene SOA to OA measured during SOAS.
SOA formation potential cannot be directly compared to the observed OA concentrations which reflect contributions from up-wind (both primary and secondary), as well as multi-phase chemistry. However, for context, we note that observed OA concentrations from CalNex (3.37 µgCsm -3 ) and SOAS (1.84 µgCsm -3 ) are both more than double the instantaneous SOA formation potential estimated at these sites. Furthermore, oxidation flow reactor (OFR) measurements during CalNex report an average of 10.5 µgsm -3 (~5.25 µgCsm -3 ) 58 , ~1.9 µgCsm -3 above ambient loading.
Our estimated SOA potential (including IVOCs) of 2.3 µgCsm -3 , is similar, but the OFR constraint may suggest that our yields for SOA potential are overestimated. Similar analysis of the SOAS OFR measurements suggest an enhancement over ambient of 1.2 µgCsm -3 , ~30% higher than our estimated SOA potential. Thus the application of estimated yields based primarily on vapor pressures (with the exception of aromatics and primary biogenics, for which we follow Hunter et al. 15 ) may overestimate SOA potential in an urban environment and underestimate SOA potential in a biogenic environment. Further exploration of how estimated SOA potential compares with OFR measurements at various sites, particularly with better constrained contributions from S/IVOCs, may provide constraints on assumed SOA yields.
Reactive organic carbon (ROC) is a key fuel for atmospheric chemical transformations and the production of secondary pollutants. In this study we contrast the mass concentrations, OH reactivity, and SOA formation potential of ROC in two very different environments in the United States: Southern California and the SEUS. This comparison allows us to explore how these simple metrics can be used to characterize and contrast the local atmospheric environment. We find that the average ROC mass concentration during CalNex is twice that observed during SOAS, with the former dominated by anthropogenic species, and the later characterized by biogenics. Despite these considerable differences, we find that the OHR is similar at the two sites; however, only about half of the cOHR observed during CalNex comes from organics. Thus, the ROC measured in Southern California is, on average, about a factor of 4 less reactive with OH than that in the SEUS. This could translate to much higher ozone production in the SEUS, however NOxlimitation and lower OH concentrations depress ozone production rates compared to Southern California. In addition, the lower OH in the SEUS extends the isoprene lifetime, increasing the observed reactivity 59 . Finally, the SOA formation potential for all the ROC species measured in CalNex is almost twice that estimated for SOAS, largely driven by aromatics. Thus we conclude that the ROC species in Southern California are more abundant, less reactive, and (due to the presence of elevated NOx and OH) more efficiently produce ozone and SOA than the ROC species present in the SEUS.
This analysis is based on a bottom-up characterization of ROC, by adding up the contributions of measured individual species. Given the absence of any top-down total carbon measurement, we cannot assess the degree of mass closure obtained with this suite of measurements. However, the total OHR measurements from these campaigns suggest that the bottom-up tally of cOHR is short by 14% (SOAS) and 31% (CalNex). This missing OHR likely represents some combination of missing primary species (e.g. sesquiterpenes and oxygenated VCP emissions) and unmeasured secondary products. We highlight throughout the potential role of intermediate volatility species, using as an example the offline measurements made during CalNex 33 . While these species likely make a modest contribution to mass concentrations, OHR, and ozone formation, they may well contribute more than a quarter of the SOA formation potential in that region. This illustrates the importance of characterizing the role of IVOC species in PM formation, particularly in urban environments. For SOAS, the missing reactivity is limited to the nighttime and is much more modest, perhaps within the uncertainty of the measurements. This raises the interesting question of how measurement and reaction rate uncertainties may dictate the limit of how OHR measurements can be used to constrain ROC.
Of the 112 different organic species measured during CalNex and SOAS, many are not routinely reported in field campaigns or simulated in models, however the importance of these constituents varies with location. For example, the chemical mechanism used in the GEOS-Chem global model represents the majority of the species reported during SOAS, and could represent (individually or lumped) 87-95% of the mass, OHR, and SOA formation potential reported here. This same chemical mechanism could represent 85% of the mass concentrations measured during CalNex, but only 80% of the SOA formation potential and 76% of the OHR. Thus, the more diverse suite of anthropogenic compounds emitted in urban environments may be challenging to represent in current models. The number of species routinely detected by new instrumentation deployed in the field continues to grow (e.g. 60 ), and as future campaigns likely report a greater fraction of the ROC in the atmosphere, it may become more challenging for simplified models to represent this diversity and the resulting chemistry. 1 and2). The calculated standard error (defined as 𝜃/√𝑁) on campaign average OHR is 0.32 s -1 and 0.28 s -1 respectively (too small to be shown on the plot).