Observed versus simulated OH reactivity during KORUS-AQ campaign: Implications for emission inventory and chemical environment in East Asia

Constraining the total amount of reactive trace gases in the troposphere is crucial for understanding their photochemistry, which is primarily associated with reactions among hydroxyl radicals (OH) (Levy, 1971). Hence, the OH reactivity (OHR), which is the reciprocal of the OH lifetime, can be used to measure total trace-gas loadings in any given environment (Kovacs and Brune, 2001). Comparisons between the observed total OHR (tOHR) and calculated OHR (cOHR) based on trace-gas observations at a given location enable critical evaluations of our ability to constrain reactive trace gases.

Since Kovacs and Brune (2001) first demonstrated ambient observations of tOHR based on an in situ flow reactor, there have been consistent reports of a “missing OHR fraction,” defined by Equation 2, where cOHR is calculated from summing the individual trace-gas reactivities as follows:

Here, $KOH−X$ is the biomolecular reaction rate constant (cm⁻³ molecule⁻¹ s⁻¹) between OH and the measured concentration of each species [X] in molecules cm⁻³.

The cOHR reported in previous studies ranges from 10% to 95% of the tOHR (Yang et al., 2016). Moreover, several large cities in North America, such as Nashville, TN, and New York, have reported missing OHR fractions of 10%–30%, with NO₂ and hydrocarbons representing the largest contribution to the reactivity (Kovacs et al., 2003; Ren et al., 2003a; Ren et al., 2003b). Even larger missing OHR fractions are consistently reported from biogenic volatile organic compound (BVOC)-dominant environments such as forest ecosystems. For example, Di Carlo et al. (2004) and Nakashima et al. (2014) reported missing OHR fractions of 30%–50% in a hardwood forest and ponderosa pine forest in the United States, respectively. Additionally, Nölscher et al. (2012) and Sinha et al. (2010) found missing OHR fractions of up to 60%–90% in a boreal forest in Finland, which was dependent on air temperature. Missing OHR of 5%–84% has also been reported in pristine rain forests such as the Borneo or Amazon rainforests (Edwards et al., 2013; Nölscher et al., 2016).

Several studies on remote marine boundary layers (MBLs) with relatively small loadings of reactive gases have also reported substantial missing OHR fractions. For example, Mao et al. (2009) presented a vertical profile of tOHR (median 4.0 ± 1.0 s⁻¹) during the Intercontinental Chemical Transport Experiment-B; however, the comprehensive NASA DC-8 observational airborne dataset could only account for 48% of the tOHR, primarily through CO (50%–70%), oxygenated volatile organic compounds (OVOCs, 15%–20%), and CH₄ (5%–15%). Conversely, the contributions of NO_X and nonmethane hydrocarbons (NMHC, e.g., ethane, ethene, and propane) to cOHR were relatively insignificant (less than 10%). Their study concluded that the assessed missing OHR fraction likely originated from the oxidation products of unmeasured VOCs, as a strong correlation (R² = 0.58) was observed between the missing OHR fraction and the concentration of formaldehyde (CH₂O). Thames et al. (2020) also measured tOHR (mean 1.9 s⁻¹) and the missing OHR fraction (approximately 30% of tOHR) in the remote MBL during the NASA Atmospheric Tomography (ATom) campaign. Similar to Mao et al. (2009), the major contributions to cOHR were CO (30%–40%), CH₄ (19%–24%), methyl hydroperoxide (5%–16%), and aldehydes (11%–12%); thus, they concluded a similar origin for the missing OHR fraction.

Substantial missing OHR fractions have also been reported in East Asia. For example, unaccounted VOCs, including OVOCs from anthropogenic and biogenic emissions, were found in Tokyo, Japan, which resulted in missing OHR fractions of 10%–30% (Chatani et al., 2009; Kato et al., 2011; Yoshino et al., 2012). In Seoul, South Korea, Kim et al. (2016) reported significant contributions of CO and NO_X to the tOHR at an urban site. Kim et al. (2016) also observed a substantial missing OHR fraction (approximately 70%) under the strong influence of BVOCs in suburban forests. Furthermore, Brune et al. (2022) compared tOHR and OHR calculated using the NASA Langley Research Center photochemical box model with master chemical mechanism (MCM) version 3.3.1 and reported a missing OHR fraction of 12%–30% in the planetary boundary layer (PBL) during the Korea–United States Air Quality (KORUS-AQ) campaign (Crawford et al., 2021).

Several attempts have been made to reconcile these missing OHR fractions by calculating the nonobserved oxidation products of VOCs with a 0-D box model or regional and global photochemical model frameworks. For example, Chatani et al. (2009) reconciled up to 60% of the missing OHR fraction for the Tokyo metropolitan area by scaling input emissions and regional background boundary layer averages using the Weather Research and Forecasting model coupled with Community Multiscale Air Quality modeling system. Edwards et al. (2013) reconciled approximately 50% of the missing OHR fraction in a tropical rainforest in southeast Asia using isoprene oxidation products based on box model simulations. However, explaining OHR with model results requires abundant caution because of potential uncertainties in the primary emissions and chemistry of missing species, which may hinder the realistic simulation of nonobserved reactive trace gases, particularly OVOCs (Kim et al., 2015; Kaiser et al., 2016).

Therefore, the aim of this study is to holistically examine airborne tOHR values observed during the KORUS-AQ campaign through a comparison with cOHR values calculated using a box model and a 3-D chemical transport model (CTM). As reported by Crawford et al. (2021), the study area includes various anthropogenic and biogenic emission sources as well as occasional pollution transport from the continent. Additionally, the Yellow Sea is located along the transport pathway. Therefore, our comprehensive approach provides a unique evaluation opportunity to constrain tropospheric reactive gases in a complex chemical environment comprising various emission sources in East Asia. We focus on 3 different regions: the Seoul Metropolitan Area (SMA, population 25 million); a suburban forest area, where BVOCs account for substantial trace-gas reactivity (Kim et al., 2021); and the MBL of the Yellow Sea, which is influenced by both transboundary and local emissions. Despite its significance for global air quality, the East Asian region is one of the least studied areas worldwide.

In contrast to previous studies, which have investigated VOCs for significant emission sources (Simpson et al., 2020) or vertical profiles of OHR and trace gases for specific ground sites (Kim et al., 2021), we extend our analysis to all flight tracks during the KORUS-AQ campaign for intercomparison with a 3-D global CTM, GEOS-Chem, routinely used for regional air quality studies. The observed trace-gas dataset derived from the NASA DC-8 airborne laboratory is used to assess the cOHR and evaluate the GEOS-Chem simulation results for the sOHR. An observationally constrained 0-D box model is used to calculate the unmeasured oxidation products of numerous VOC precursors, which allow us to diagnose the potential influence of nonobserved species on the highly reactive chemical environment of this Asian megacity. We then discuss the degree of unaccounted trace gases in the model frameworks and the corresponding implications for regional and global air quality. All abbreviations used in this study are defined in Table 1.

KORUS-AQ was a multi-platform international field campaign conducted in Korea between May and June 2016 (Crawford et al., 2021). In this study, we used the observational dataset from the NASA DC-8 airborne laboratory, which is available to the public (https://www-air.larc.nasa.gov/cgi-bin/ArcView/korusaq; DOI: http://dx.doi.org/10.5067/Suborbital/KORUSAQ/DATA01, accessed July 1, 2022). Peterson et al. (2019) categorized the campaign into 4 meteorological periods: dynamic (May 1–16) with complex aerosol vertical profiles, stagnant (May 17–22) under a persistent anticyclone, transport (May 25–31) with low-level transport and haze development, and blocking period (June 1–7) (Rex, 1950). Simpson et al. (2020) summarized the individual meteorological conditions for 20 DC-8 flights during the campaign.

During the campaign, tOHR was measured using the airborne tropospheric hydrogen oxide sensor (ATHOS) with an error of ±0.64 s⁻¹ and 90% confidence (Mao et al., 2009; Thames et al., 2020). For comparison, the cOHR was calculated from the observed trace-gas dataset and laboratory kinetics data. The relevant reaction constants for the observed trace gases were adapted from a comprehensive MCM version 3.3.1 (http://mcm.york.ac.uk/) (Saunders et al., 2003; Jenkin et al., 2015). Uncertainties associated with species observations and their reaction coefficients with OH are critical for calculating the cOHR. Such uncertainties are assessed based on the known observed uncertainty for species measurements and the experimental uncertainty of rate constants (usually approximately 10%–15%). In Section 3, we discuss the potential impact of these uncertainties on cOHR assessments.

The NASA DC-8 flying laboratory measures reactive gases including O₃, NO_X, NO_y, OH, HO₂, and speciated VOCs. The NCAR NO-NO_y instrument uses the chemiluminescence of NO via reaction with O₃ (Ridley and Grahek, 1990) to measure airborne O₃, NO_X, and NO_y. ATHOS (Brune et al., 1995) measures OH and HO₂ with a detection limit of 0.018 pptv and 0.2 pptv, respectively. Detailed quality assurance and control procedures are described in Thames et al. (2020). The fundamental identical instrumentation configuration described by Mao et al. (2009) was adopted for the KORUS-AQ campaign, with more careful characterization of offset calibrations described in Thames et al. (2020). Alkanes, alkenes, and aromatic hydrocarbons were sampled from whole-air samples and identified using gas chromatography with electron capture detection, mass spectrometry detection, and flame ionization detection with detection limits of 2–3 pptv (Colman et al., 2001). Measurements taken with a proton-transfer-reaction time-of-flight mass spectrometer (PTR-ToF-MS) (Müller et al., 2014) provided the concentrations of isoprene and its oxidation products (sum of MVK, MACR, and ISOPOOH), methanol, acetaldehyde, and acetone and propanal. Measurements of isoprene by PTR-MS tend to be higher than other measurements owing to interference from other C₅H₈ compounds in regions with heavy anthropogenic emissions (Vlasenko et al., 2009). Here, we assume that these other C₅H₈ compounds react with OH as quickly as isoprene, and their reactivities are included in cOHR. A compact atmospheric multispecies spectrometer (Richter et al., 2015; Fried et al., 2020) was used to measure formaldehyde. Table 2 summarizes the NASA DC-8 observations used in this study. Each observed species has different time resolutions; therefore, we used the 1-min merged dataset for subsequent data analysis. In addition, the photolysis frequencies were calculated using spectrally resolved charge-coupled device actinic flux spectroradiometers. The calibration protocol of each measurement is consistent with previous campaigns in which the NASA DC-8 was deployed, and the details of the calibration protocol can be found at https://airbornescience.nasa.gov/instrument/all (accessed July 18, 2022). The notated uncertainty and the limit of detection are mostly based upon 1-s average statistics. As we exclusively rely on the 1-min merged dataset for our presented data analysis, the lower limit of detection (LLOD) is likely way lower than the notated LLOD in Table 2 according to the Poisson distribution statistics.

We used a box model and a global 3-D CTM (GEOS-Chem) to conduct simulations for the KORUS-AQ campaign. The box model with observational constraints simulates nonobserved species, mainly VOC oxidation products, which may account for the missing OHR fraction of cOHR. The box model results were further compared with GEOS-Chem to provide a comprehensive understanding of the chemical scheme used in air quality simulations and insights into the missing OHR fraction of sOHR.

The box model used in this study was the Dynamically Simple Model of Atmospheric Chemical Complexity (DSMACC, https://sites.google.com/site/dsmaccmanual/home) (Emmerson and Evans, 2009; Zhou et al., 2014; Gressent et al., 2016; Stone et al., 2018), driven by the Kinetic Pre-Processor (Sandu and Sander, 2006). The photolysis rates in DSMACC were calculated using the Tropospheric Ultraviolet Model version 4.6 (Madronich and Flocke, 1999) with modifications to incorporate it into the DSMACC. We conducted box model simulations with MCM version 3.3.1 and a chemical mechanism of GEOS-Chem v12.7.2 (Bey et al., 2001, DOI: http://dx.doi.org/10.5281/zenodo.3701669). The simulated results were used to estimate the potential contribution of nonobserved species to the OHR and evaluate the chemical mechanism of GEOS-Chem to reproduce the observed OHR.

The DSMACC uses a diurnal steady-state approach (Olson et al., 2012; Schroeder et al., 2020) to calculate the 1-min steady-state concentrations of nonobserved species with observationally constrained values as inputs for O₃, CO, water vapor, CH₂O, PAN, H₂O₂, HNO₃, NMHC, photolysis rates, temperature, and pressure for each NASA DC-8 flight. For the diurnal steady-state approach, we constrained NO_X and allowed NO/NO₂ to vary throughout the diurnal cycle. The photolysis rates in the DSMACC varied and were scaled by the observed J[O₃ → O₂ + O(¹D)] and J[NO₂ → NO + O(³P)] until a diurnal steady state was reached. All constrained species are listed in Table S1. We included DC-8 data as constraints for our box model simulations if the valid values were greater than 40% of each flight observation. Missing data for each flight were interpolated using nearby observations. Interpolated data accounted for 13%, 8%, and 3% of NO, O₃, and H₂O, respectively, and less than 1% of CH₄ and CO observations. For CH₂O, one of the noncritical species (Schroeder et al., 2020), 14% of the observations were interpolated. Otherwise, they were excluded from the constraints of each flight. We also excluded DC-8 flight points near the Daesan petrochemical complex, South Korea’s largest point emission source (126–126.88°E, 36.4–37.15°N), because a typical CTM cannot sufficiently represent highly polluted plumes from the complex because of its coarse spatial resolution. We also evaluated the capability of the box model by comparing the simulated OH, HO₂, and CH₂O concentrations with the observations.

The 3-D global CTM, GEOS-Chem v12.7.2 (https://geos-chem.seas.harvard.edu; DOI: http://dx.doi.org/10.5281/zenodo.3701669), was also used to simulate the KORUS-AQ campaign. The model results were sampled according to the NASA DC-8 flight tracts based on longitude, latitude, and pressure with a 1-min temporal resolution. A plane flight diagnostic (http://wiki.seas.harvard.edu/geos-chem/index.php/Planeflight_diagnostic) in GEOS-Chem was used for sampling. The model was driven by the Goddard Earth Observing System that assimilated meteorological data (GEOS-FP) from NASA’s Global Modeling and Assimilation Office (Bey et al., 2001, http://gmao.gsfc.nasa.gov/products/). Moreover, we used a nested model configuration of GEOS-Chem with spatial resolutions of 0.25° × 0.3125° and 47 vertical hybrid pressure-sigma levels up to 0.01 hPa for the East Asia domain (100–140°E, 15–55°N).

The boundary conditions for the nested simulations were obtained from the global simulation with a resolution of 2° × 2.5° every hour. We used the KORUSv5 inventory developed by Konkuk University for anthropogenic emissions from East Asia (Woo et al., n.d.). This inventory includes CO, NO_X, SO₂, NH₃, and VOCs emissions for the SAPRC-07 chemical mechanism (Carter, 2010), whose annual total emissions for the nested domain in 2016 were 1.9, 1.31, 0.34, 0.29, and 0.88 Tg/yr, respectively. The KORUSv5 inventory was adopted for GEOS-Chem after conversion from SAPRC-07 to GEOS-Chem chemistry, based on a study by Li et al. (2014). The Global Fire Emissions Database 4 inventory (van der Werf et al., 2010) and the Model of Emissions of Gases and Aerosols from Nature version 2.1 (Guenther et al., 2012) provided biomass burning and biogenic emissions during the campaign period. We then spatiotemporally sampled all simulation results along the flight tracks of NASA DC-8 for comparison with the observations, as discussed in Section 3.

Figure 1 presents the observed tOHR, cOHR, the missing OHR fraction of cOHR, and the contributions of CO and OVOCs to cOHR in the PBL (0–2 km) averaged during the KORUS-AQ campaign. In the PBL, elevated tOHR occurred in the SMA, the suburban forest area, and the MBL of the Yellow Sea. cOHR captures the spatial variation of tOHR moderately effectively, with a spatial correlation coefficient (R) of 0.5. The missing OHR fraction was typically in the range of 20% to 60% and did not show a significant correlation with tOHR (R = 0.1). The average values of spatially distributed tOHR, cOHR, and missing OHR fraction in the PBL were 5.2 ± 0.11 s⁻¹, 3.6 ± 0.08 s⁻¹, and 32.2% ± 1.25%, respectively. Brune et al. (2022) also calculated the missing OHR fraction with 5-min averages during the KORUS-AQ campaign, and their results showed good consistency with the missing OHR fraction calculated with 5-min averages in this study (not presented).

The emission profiles of SMA were consistent with those of other megacities driven by anthropogenic activities (Crawford et al., 2021). The Taeahn Peninsula, southwest of the SMA, has large point sources such as large-scale coal power plants (e.g., Taeahn Thermal Power) and the Daesan petrochemical complex, which may influence SMA air quality depending on the meteorological conditions (Crawford et al., 2021). We found that the contributions of CO and OVOCs to tOHR increased by up to 32% and 27% (Figure 1d and e) and were likely derived from anthropogenic trace gases and their oxidants (Simpson et al., 2020). The cOHR values in the SMA and Taehan Peninsula reproduced the elevated tOHR with a smaller missing OHR fraction than other areas because of the dominant contributions of well-identified anthropogenic trace gases such as CO and anthropogenic primary VOCs. However, the missing OHR fraction increased in the downwind regions of the SMA and Taehan Peninsula, likely because of the rapid conversion of primary anthropogenic gases to their oxidation products, which was not well captured by the model.

Previous studies based on surface measurements at urban sites reported much higher tOHR values (approximately 5–20 s⁻¹) than those observed in this study (Kovacs et al., 2003; Ren et al., 2006; Sinha et al., 2008; Williams et al., 2016). For example, Kim et al. (2016) reported a daily average of approximately 15 s⁻¹ at a ground site near the Seoul City center. The relatively low tOHR values observed over the SMA in this study could reflect the rapid oxidation of highly reactive gases in the shallow part of the PBL, which complicates sampling by an airborne platform such as NASA DC-8 (Kim et al., 2021).

In the suburban forest area, elevated OHR occurred despite the relatively smaller contributions of anthropogenic emissions in this region than in the urban area (Figure S1). We also noted increased contributions of BVOCs to cOHR in this region, likely because of the high reactivity of BVOCs, such as isoprene and monoterpenes. Indeed, during the KORUS-AQ campaign, ground-based observations of tOHR in the Taehwa forest revealed substantially high reactivity inside the forest canopy (Kim et al., 2021). In addition, we found a positive correlation (0.3) between the missing OHR fraction and the concentrations of typical biogenic VOCs, such as isoprene, methyl vinyl ketone, and methacrolein. This mild positive correlation is consistent with a report by Sanchez et al. (2020), demonstrating that unidentified/unmeasured BVOC oxidation products contribute to missing OHR along with anthropogenic NMHC oxidation products in the suburban forest area.

In the MBL of the Yellow Sea, elevated tOHR was captured by the NASA DC-8, as shown in Figure 1. This was likely driven by pollution outflow from China (May 25–31) and South Korea (June 5) because most of the flights that flew over the Yellow Sea aimed to capture continental outflows during the campaign. More details about flight tracks during the campaign can be found on the KORUS-AQ website (https://www-air.larc.nasa.gov/missions/korus-aq/docs/KORUS-AQ_Flight_Summaries_ID122.pdf, accessed July 1, 2022). Thus, the following findings may represent the Yellow Sea under polluted conditions caused by continental outflows. We found an increase in CO contributions to OHR, which became a tracer for the pollution outflow of anthropogenic sources from the continent. Furthermore, the missing OHR fraction in the MBL of the Yellow Sea increased by up to approximately 60%, exhibiting a moderate correlation with CO (R = 0.5) and OVOCs (R = 0.4), which implied that the missing OHR fraction derived from unmeasured anthropogenic trace gases or their photochemical products. Fried et al. (2020) also provided evidence of highly localized emission plumes of CH₂O and other constituents over the Yellow Sea on June 5 caused by localized sources believed to be ship plumes (see Fried et al., 2020; figures S13 and S14).

In the free troposphere, the average values of spatially distributed tOHR, cOHR, and missing OHR fraction were 2.2 ± 0.06 s⁻¹, 1.1 ± 0.03 s⁻¹, and 49.4% ± 0.93%, respectively (Figure 2 and Figure S2). The tOHR and cOHR values in the free troposphere were generally lower than those of the PBL, with a high OHR over the SMA and MBL of the Yellow Sea. However, the missing OHR fraction in the free troposphere was larger than that in the PBL. The elevated tOHR in the free troposphere over the Yellow Sea was more pronounced than that in the SMA, reflecting pollution outflow from the continent. However, the cOHR over the ocean was considerably lower than the tOHR. This discrepancy suggests the significant contribution of unmeasured long-lived species from continental outflow or the accumulation of unaccounted oxidation products in the free troposphere. The missing OHR fraction accounted for a larger proportion of OHR (20%–80%) in the free troposphere than in the boundary layer, indicating the significance of unidentified trace gases contributing to OHR in the free troposphere.

The meteorological patterns during the KORUS-AQ campaign led to variations in the tOHR and missing OHR fraction under local or transboundary influences (Figures S3 and S4). A notable signal relating to tOHR enhancement and a missing OHR fraction was observed during the transport period, especially in the free troposphere. The transportation of some nonobserved oxidation products likely resulted in a missing OHR fraction during the transport period. In addition, during the blocking period, NASA DC-8 flew over local point sources, causing significant enhancement of the OHR. However, NASA DC-8 flew different flight tracks during each meteorological period to capture various signals, such as transboundary transport and local emissions. Thus, intercomparisons of the OHR between different meteorological periods would be meaningful only for identical flight tracks of NASA DC-8 (over the SMA). This would require a further case study to isolate the impact of meteorological conditions on OHR.

Figure 3 presents the profiles of median tOHR and cOHR for 500-m altitude bins during (a) the entire campaign, (b) over land, and (c) over the ocean. We also show the contributions of individual species to different colors of cOHR. As discussed above, tOHR was highest in surface air and generally decreased with altitude. The vertical distribution over land (Figure 3b) illustrates a similar vertical variation for cOHR and tOHR, but a substantial underestimate for cOHR in the PBL. The gap between tOHR and cOHR decreased as altitude increased and fell within the measurement uncertainty of tOHR at altitudes higher than 5 km. Among the trace gases, CO and OVOCs were the 2 most significant contributors to cOHR. The contribution in the free troposphere was 33%–48% for CO and 14%–28% for OVOCs. In the PBL over land, contributions of NO_X, BVOCs (ISOP+MVK+MACR), and hydrocarbons to cOHR were substantially increased by up to 9%, 6%, and 7% compared to those of the ocean (3%, 2%, and 3%), respectively. It is worth noting that the contribution of NO_X increased by up to approximately 20% near the surface. However, this contribution was substantially lower than that determined by Kim et al. (2016), who found that NO_X comprised up to 55% of OHR in the center of Seoul. There are 2 explanations for this difference: (1) the NO_X level decreased rapidly as NO_X was transported out of the city center toward suburban and rural regions (Crawford et al., 2021) because of fast oxidation, and (2) rapid oxidation also occurred in the vertical scale, as discussed previously (Kim et al., 2021). The relatively weaker emissions of reactive trace gases over the MBL can be attributed to the lower tOHR and cOHR, as illustrated in Figure 2c. Interestingly, tOHR and cOHR values were higher at altitudes between 1 km and 2 km than on land and at the surface of the MBL, indicating successful DC-8 samplings of transboundary pollutants from China and Daesan.

Figure 4 presents the detailed contributions of the OVOCs, aromatics, and other NMHC to the cOHR averaged during the KORUS-AQ campaign. It should be noted that CH₂O was the most significant contributor to PBL (Figure 4e and f). However, acetaldehyde was the most significant contributor to MBL (Figure 4g), reflecting the direct emission of acetaldehyde from oceanic sources (Millet et al., 2010). Acetaldehyde was more important in the free troposphere, followed by methanol and CH₂O (Figure 4a–c). Isoprene contributions increased in the PBL over land by up to approximately 4.1% but decreased in the MBL with distance from onshore vegetation because of its short lifetime (1–2 h at [OH] = 2.0 × 10⁶ molecules cm⁻³). However, a small isoprene contribution was noted over the ocean, which cannot typically be simulated by CTMs owing to low oceanic isoprene emissions in the model.

As discussed above, we identified a significant missing OHR fraction during the KORUS-AQ campaign. In this section, 2 additional factors were considered to reconcile these missing OHR fractions. First, it should be noted that we determined wide ranges of rate coefficients between the major trace gases and OH, which cannot be explained by the nominal uncertainty in rate coefficients that typically ranged from 10% to 30%. As shown in Figure S5, CO had the largest range of reported rate coefficients, which likely resulted in significant uncertainty in cOHR values owing to the prevalence of CO in the atmosphere. Furthermore, this uncertainty may be significant because the contribution of CO generally increased with altitude. The rate coefficients of CH₃OOH also varied significantly; however, their impact on cOHR was relatively small because of the low CH₃OOH concentration. Although the rate coefficient ranges of trace gases from inland sources and their oxidants, including NO₂, isoprene, and OVOCs, were relatively narrow, their omnipresence and high concentrations in the PBL may have resulted in significant differences in cOHR depending on the rate constants used.

This point is well demonstrated in Figure 5, which shows a correlation plot between tOHR and cOHR with color-coded altitude. tOHR was higher at lower altitudes, primarily within the PBL with a high loading of trace gases such as CO, NO₂, acetaldehyde, CH₂O, and methanol. Once we account for the range of previously reported rate coefficients, the uncertainty of cOHR was much larger than that assessed using the uncertainty derived from each rate coefficient. Although this may help reconcile the missing OHR fraction, cOHR still systematically underestimated tOHR.

Second, we conducted box model simulations to reconcile the missing OHR fractions. The simulation calculated the steady-state concentrations of unmeasured oxidation products for observationally constrained VOCs. We then evaluated the performance of the box model DSMACC by comparing the observed and simulated profiles of OH, HO₂, and CH₂O (Figure 6). The box model reproduced the observed vertical profiles of OH and HO₂ with an observational uncertainty of 25%. However, the CH₂O at high altitudes (>4.5 km) was substantially overestimated by the model. This may be caused by the significant number of invalid data points of VOC observations, accounting for approximately 68% of observation points, which likely aggravated the constraint quality at high altitudes (>4.5 km). Schroder et al. (2020) also conducted box model simulations during the KORUS-AQ campaign using the observationally constrained NASA Langley Research Center with a 1-s merge dataset, the results of which were in good agreement with our box model results (not presented).

Then, we assessed the OHR of potentially unmeasured VOC oxidation products (bOHR). Figure 7 presents the spatial distribution of bOHR and the ratio of bOHR to tOHR in the PBL and free troposphere during the KORUS-AQ campaign. The mean values of bOHR and the ratio of bOHR to tOHR were 0.48 ± 0.03 s⁻¹ and 8.9% ± 0.31% in the PBL and 0.20 ± 0.02 s⁻¹ and 9.3% ± 1.10% in the free troposphere. The greater enhancement of bOHR in the PBL over land than in the MBL and free troposphere is attributed to the higher levels of primary VOCs. Indeed, the most conspicuous enhancements occurred near strong VOC emission sources, such as the PBL of the Daesan Petrochemical Complex, with an enhancement of up to 3.5 s⁻¹. This bOHR enhancement was also observed in the ratio of bOHR to tOHR, with a 25%–75% percentile range of 6%–11% in the PBL and 5%–9% in the free troposphere.

bOHR was highly correlated with OVOCs (0.59) and other hydrocarbons (0.54). The box model results in Table 3 show that the highest bOHR contributors included MCATEC1OOH, MGLYOX, and NCRES1OOH, likely because of anthropogenic or biogenic precursors. This supports the enhancement of bOHR near strong VOC emission sources. Although analyzing the box model results was rather difficult owing to the complex chemistry, the bOHR values consequently allowed us to reconcile 32.2%–24.2% of the missing OHR fraction in the PBL and 49.4%–41.5% of the missing OHR fraction in the free troposphere. Because our box model simulations did not consider gas–aerosol partitioning, these values could be an upper limit for reconciling the missing OHR fraction with bOHR.

Although environmental differences and the altitude at which measurements were performed likely differentiate the contributions of nonobserved species in our work from those in previous studies, this level of reconciliation from unmeasured oxidation products corresponds to that of previous studies conducted on the ground. For example, Edwards et al. (2013) used a box model with MCM version 3.2 and reported that unmeasured oxidants from BVOCs could account for up to 30% of the missing OHR fraction of cOHR in a southeast Asian tropical rainforest during the Oxidant and Particle Photochemical Processes project. Furthermore, Whalley et al. (2016) reconciled 6%–33% of the missing OHR fraction in London during the Clean Air for London campaign by considering higher VOCs and their intermediates in a box model.

The improved vertical distribution of tOHR and cOHR by incorporating bOHR is presented in Figure 8. The fraction of cOHR is shown with light bars, where dark bars indicate the bOHR contribution. The associated uncertainty in cOHR was attributed to the rate constants of MCM chemistry (−11% to 14%), measurement uncertainty (−7% to 7%), and the rate constants of various references (−12% to 11%) on a campaign mean basis. Acetone and propanal were detected in the same channel as the PTR-ToF-MS analytical system (de Gouw and Warneke, 2007). We then applied the upper limit of the propanal to acetone ratio (approximately 10%; de Gouw and Warneke, 2007; Li et al., 2015) to account for the impact of propanal, which is much more reactive with OH. The uncertainty ranges resulted in cOHR variations from −33% to 35%, driven mainly by the dominant contributors in cOHR and their changes according to the reaction constants with OH, which became the most significant uncertainty in the cOHR assessment. Consequently, the aggregated uncertainty range of cOHR + bOHR and the measurement uncertainty range of tOHR overlapped at almost all altitudes.

One potential caveat of our attempt to reconcile the missing OHR fraction from the oxidation products calculated using the box model is that it is impossible to estimate the potential contributions from transported oxidation products. This is particularly true for the assessment in the free troposphere, as most of the primary VOCs emitted from the surface are likely to be oxidized. However, their relatively long-lived oxidation products are likely to be transported to the free troposphere. The oxidation products, which represent a contribution to bOHR of more than 10%, have a relatively short lifetime (0.24–1.66 h), effectively preventing significant transport from the PBL to the free troposphere. However, some of the top 10 contributors in Table 3 have a long lifetime (e.g., NCRES1OOH, 44 h; CH₃CO₂H, 85 h; MALANHY, 55 h) for transport in the free troposphere. Therefore, we can speculate on the potential contributions of these compounds from long-range transport, which requires further investigation.

In this section, we highlight the impact of the fundamental differences in treating the chemical complexity of VOC oxidation chemistry between the frameworks of box models and global CTMs, such as GEOS-Chem, owing to computational limitations. For example, GEOS-Chem uses the lumped species of ALK4, PRPE, MTPA, and XYLE for ≥C4 alkanes, ≥C3 alkenes, α/β-pinene, and m/o/p-xylene, respectively. This parameterization leads to a discrepancy between the calculated and simulated reactivity of these species.

Therefore, we evaluated the effect of species lumping on sOHR (Figure 9). Specifically, we calculated the nonlumped and lumped reactivity: the former was calculated using measured concentrations and their rate coefficients; the latter was calculated using measured but lumped concentrations and the rate coefficient used in GEOS-Chem. The largest reactivity underestimation in GEOS-Chem caused by species lumping occurred with ALK4 (−34%), followed by PRPE and MTPA with −10% and −14%, compared to the nonlumped reactivities (Figure 9b and c). However, the model overestimated the reactivity of XYLE by 29% compared to the nonlumped reactivity of m/o/p-xylene (Figure 9d). The misrepresentation of species reactivity caused by the lumping scheme may be particularly pronounced in urban and suburban environments, where the emission of these chemicals occurs.

Figure 10 presents the spatial distribution of the sOHR and its missing OHR fraction. For the sOHR calculations, the simulated concentrations and reaction rates in GEOS-Chem were used without observational constraints. The average values of spatially distributed sOHR and its missing OHR fraction were 2.7 ± 0.08 s⁻¹ and 47.7% ± 1.25% in the PBL and 0.9 ± 0.03 s⁻¹ and 61.0% ± 0.93% in the free troposphere, respectively. The sOHR captures the spatial variation of tOHR in the free troposphere with a spatial correlation coefficient of 0.7 in the free troposphere. However, sOHR in the PBL only showed a spatial correlation coefficient of 0.4, which was likely caused by underestimation of the sOHR over the ocean. Furthermore, sOHR only accounted for 30%–40% of tOHR, with a higher missing OHR fraction than that determined for cOHR (Figure 10c and d compared to Figure 1c and Figure 2c) during the KORUS-AQ campaign. Kim et al. (2016) also determined a similar magnitude missing OHR fraction with GEOS-Chem in Seoul during the MAPS campaign, with considerable contributions of NO_X, BVOCs, and hydrocarbons to tOHR. However, Travis et al. (2020) also used GEOS-Chem but reported a missing OHR fraction of 30%–40% during the NASA atom field campaign, which was lower than that determined in this study. In the case of the NASA ATom campaign (Thompson et al., 2022), measurements were conducted over remote regions such as the Atlantic and Pacific oceans with less inland influence, resulting in fewer uncertainties in anthropogenic and biogenic trace gases and their oxidation products. This environmental disparity implies that the missing OHR fraction during the KORUS-AQ campaign was likely caused by uncertainties in the anthropogenic emission inventory.

The missing OHR fraction of sOHR reached 74% in the PBL and 83% in the free troposphere, showing much larger mean values than those of cOHR. Although spatially very similar to those of cOHR, the sOHR values underrepresent cOHR values in the PBL and free troposphere. This problem was attributed to the underestimation of trace-gas concentrations in GEOS-Chem. As shown in Table 4, GEOS-Chem underestimated almost all species. The low CO (approximately 28%) and CH₂O (approximately 31%), which contributed the most to OHR, directly led to lower sOHR than tOHR and cOHR in both the PBL and free troposphere.

CO underestimation in East Asia is a well-documented problem that occurs not only in GEOS-Chem but also in other CTMs (Park et al., 2021). CH₂O is an oxidation product of hydrocarbons, and its low bias in GEOS-Chem is related to its direct emissions and secondary production from the oxidation processes of higher hydrocarbons, making it challenging to determine the cause of CH₂O underestimation. However, anthropogenic VOCs emissions in East Asia contain significant uncertainties, for example, Kwon et al. (2021) and Choi et al. (2022) suggested an increase in anthropogenic emissions during the KORUS-AQ campaign of up to 1.5–6.9 times and 56%, respectively. These uncertainties then propagate nonlinearly and influence the simulated concentrations and sOHR. Thus, it is necessary to extensively evaluate the anthropogenic emission inventory in East Asia to better represent the OHR in CTMs.

Unlike CO and CH₂O, Figure S6 shows that GEOS-Chem captured the elevated contributions of NO_X and other hydrocarbons in the PBL of the SMA and near the strong point sources, but slightly overestimated their inland contributions to sOHR compared to the measured ones. GEOS-Chem also reproduced an increased contribution of BVOCs in the PBL over the inland region, but overestimated the magnitude of the BVOC contribution. This high bias in the PBL over land slightly alleviated the underestimation of sOHR, but for incorrect reasons. However, it did not fully compensate for the missing OHR fraction of sOHR owing to the low CO and OVOCs. In addition, we found that GEOS-Chem tended to underestimate PBL heights over land in the nighttime and early morning, possibly resulting in simulated high BVOC concentrations. The sensitivity of OHR to the misrepresentation of PBL heights in the model requires further investigation.

In the MBL, GEOS-Chem failed to capture the increased contribution of oxygenated or biogenic VOCs and resulted in a worse simulation of sOHR. The failure over the MBL may be attributed to the fact that GEOS-Chem considers oceanic emissions but substantially underestimates them. Some studies have reported nonnegligible marine emissions of trace gases, including isoprene, acetaldehyde, and acetone, at the air–sea interface (Carpenter et al., 2012; Ciuraru et al., 2015; Dani and Loreto, 2017; Brüggemann et al., 2018). Improving these emissions in CTMs would simulate a better representation of sOHR in the MBL.

The vertical distributions of tOHR and sOHR are presented in Figure 11. Among the simulated trace gases, CO, OVOCs, and NO_X were the most significant contributors to sOHR. However, as shown in Figures S6 and S7 and Table 4, the contributions of CO and OVOCs were underestimated compared to the observations. This underestimate was worse over the MBL. Although this was also caused by the lower trace-gas concentrations in GEOS-Chem compared to the observations, it failed to reproduce the much lower sOHR over the MBL compared to cOHR (Figure 3c). There are 2 plausible reasons for this result. The first is artificial mixing in the grids in the 3-D CTM. Unlike in reality, the air mass in the model grid was assumed to be uniformly mixed, regardless of the grid size. Second, oceanic emissions were underestimated in GEOS-Chem, as discussed above. Currently, the separation and confirmation of these 2 explanations require further study before reaching a definitive conclusion.

To assess the uncertainty of GEOS-Chem in the OHR simulation derived from the limitations in constraining the oxidation products of VOCs in the studied environment, we repeated the box model run by replacing the MCM chemistry with the GEOS-Chem reaction mechanisms. Therefore, comparing the 2-box model outcomes allows us to quantify the consequences of employing simplified chemistry to assess the contributions from oxidation products.

Figure 12 shows that box-modeled OHR accounted for 50%–60% of tOHR, which is better than the results in Figure 11 owing to observational constraints. However, bOHR from the box model with GEOS-Chem chemistry (bOHR_GC) was lower than bOHR from the box model with MCM chemistry (Figure 13). This difference was likely caused by species lumping in GEOS-Chem, resulting in bOHR_GC values that were 63% and 72% lower than bOHR in the PBL and the free troposphere, respectively. Indeed, Table 5 shows smaller oxidant contributions from anthropogenic hydrocarbons than Table 3, reflecting the impacts of the simplified chemistry in GEOS-Chem.

Figure 14 shows a vertical profile of OH concentrations observed during the campaign and the ratios of simulated to observed OH concentrations. The results from the observationally constrained box model with GEOS-Chem chemistry are also shown. The models generally reproduced the observed OH concentrations. Although the sOHR values were 60%–70% lower than tOHR values, GEOS-Chem OH concentrations were approximately 28% higher than the observations. However, it is difficult to determine the possible causes of this discrepancy. This may not be surprising because the tropospheric OH level is highly buffered; thus, the lower simulated OHR causes slower OH recycling rates. However, this observation implies that a comparison between the model-simulated and observed OH for model evaluations may not be the best practice unless the model-simulated OHR is comprehensively examined.

To the best of our knowledge, this is the first study to comprehensively assess the spatial distributions of observed and simulated OHR in East Asia, where complex emission profiles challenge the sound understanding of regional air quality. We found that large anthropogenic emissions elevate tOHR near the sources and downwind regions, as well as further afield in the free troposphere. Compared to previous studies that measured tOHR at the surface in East Asia, tOHR based on aircraft observations appears to be lower because of the smaller contributions from primary anthropogenic trace gases, such as NO_X and hydrocarbons, which implies rapid oxidation or dilution of reactive anthropogenic trace gases at the surface.

cOHR based on trace-gas measurements from NASA DC-8 reproduces tOHR moderately effectively, with a spatial correlation coefficient (R) of 0.5 in the PBL and 0.7 in the free troposphere, but has a missing OHR fraction range of 10%–80%. We found that 10%–15% of the missing OHR fraction could be reconciled by nonobserved species calculated using the observationally constrained box model. In addition, we argue that the wide range of previously determined rate coefficients between ubiquitous trace gases and OH may cause substantial uncertainty in bOHR assessments. The uncertainty in varying rate constants leads to cOHR variations of −12% to 11% on a campaign mean basis. Nonetheless, we failed to close the reactivity budget, even with the inclusion of bOHR. Therefore, fundamental laboratory kinetic research should be reevaluated, even for chemical species that are well studied in the atmosphere.

The performance of GEOS-Chem in reproducing the OHR was evaluated against tOHR and cOHR. Compared with tOHR and cOHR, sOHR could only account for 30%–40% of tOHR. Specifically, GEOS-Chem substantially underestimated the observed CO and OVOC concentrations. Significant uncertainties in the anthropogenic emission inventory in East Asia largely account for the underestimation of simulated species and propagating uncertainties in the sOHR, implying the need for extensive validation of emission inventories in East Asia. To evaluate the performance of GEOS-Chem chemistry, we repeated the observationally constrained box model simulation. The outcome illustrated an improvement, but not in the degree of box-model-calculated OHR with MCM chemistry. The main cause appeared to be the lumping and simplification of reaction products in the reaction mechanism. Thus, careful consideration is required during the development of reaction mechanisms to correctly simulate not only the carbon budgets but also the reactivity budgets of VOC oxidation.

Thus, uncertainties in the anthropogenic emission inventory, particularly those associated with CO and OVOCs, directly cause underestimation of the OHR. Furthermore, we found that the secondary chemical production of OVOCs is challenging to quantify and evaluate. Thus, more effort is required to refine the emission inventories of anthropogenic emissions. In VOC photochemistry, a more careful lumping scheme that considers reactivity may be required. We also evaluated the ability of GEOS-Chem to reproduce the OH concentration against observed OH concentrations. Despite substantial underestimation of sOHR, GEOS-Chem reproduced the variation and magnitude of OH moderately effectively, indicating the slow recycling of OH, which may result in less O₃ and OVOCs in GEOS-Chem.

The observational dataset from KORUS-AQ used in this study is publicly available in the International Consortium for Atmospheric Research on Transport and Transformation (ICARTT) format from the data archive website (https://www-air.larc.nasa.gov/cgi-bin/ArcView/korusaq; DOI: http://dx.doi.org/10.5067/Suborbital/KORUSAQ/DATA01, accessed July 1, 2022). The 1-min merged dataset with ICARTT format is used (https://www-air.larc.nasa.gov/cgi-bin/ArcView/korusaq?MERGE=1#60_SECOND.DC8_MRG/, accessed July 1, 2022). GEOS-Chem version v12.7.2 can be downloaded from http://www.geos-chem.org or https://doi.org/10.5281/zenodo.3701669, accessed July 1, 2022.

The supplemental files for this article can be found as follows:

Table S1. Summary of constrained species from NASA DC-8 for box model simulations.

Figure S1. Spatial distributions of the contributions of observed trace gases to cOHR in the PBL.

Figure S2. Spatial distributions of the contributions of observed trace gases to cOHR in the free troposphere.

Figure S3. Spatial distributions of tOHR during meteorological patterns in the PBL (upper panels) and free troposphere (lower panels).

Figure S4. Spatial distributions of the missing OHR fraction during meteorological patterns in the PBL (upper panels) and free troposphere (lower panels).

Figure S5. Relative ranges (%) of previously determined rate coefficients between trace gases and OH with respect to values in the MCM version 3.3.1. Average values for individual gases are shown by closed circles.

Figure S6. Spatial distributions of the contributions of simulated trace gases to sOHR in the PBL.

Figure S7. Spatial distributions of the contributions of simulated trace gases to sOHR in the free troposphere.

We thank all participants of the KORUS-AQ team for their dedication during the field campaign and the agencies operating the measurements onboard DC-8 and ground sites. This work was supported by the National Research Foundation of Korea (NRF) No. NRF-2018R1A2B3004494 and NRF-2020H1D3A2A01060699. SK acknowledges support from U.S. National Science Foundation (2035225) and National Oceanic Atmospheric Administration (NA21OAR4310129). PTR-ToF-MS measurements aboard the NASA DC-8 during KORUS-AQ were supported by the Austrian Federal Ministry for Transport, Innovation and Technology (bmvit, ASAP – FFG). The PTR-MS instrument team (P. Eichler, L. Kaser, T. Mikoviny, M. Müller) is acknowledged for their support with field work and data processing. We would like to thank Editage (www.editage.co.kr) for English language editing.

This study was supported and grant funded by the National Research Foundation of Korea (NRF) No. NRF-2018R1A2B3004494 (HK, RJP) and No. NRF-2020H1D3A2A01060699 (HK, RJP, SK).

The authors have declared that no competing interests exist. Armin Wisthaler is an associate editor at Elementa. He was not involved in the review process of the article.

Contributed to conception and design: HK, RJP, SK.

Contributed to acquisition of data: All coauthors.

Contributed to analysis and interpretation of data: HK, RJP, SK, WHB, AW.

Drafted and/or revised the article: HK, RJP, SK, AF, AW.

Approved the submitted version for publication: All coauthors.

How to cite this article: Kim, H, Park, RJ, Kim, S, Brune, WH, Diskin, GS, Fried, A, Hall, SR, Weinheimer, AJ, Wennberg, P, Wisthaler, A, Blake, DR, Ullmann, K. 2022. Observed versus simulated OH reactivity during KORUS-AQ campaign: Implications for emission inventory and chemical environment in East Asia. Elementa: Science of the Anthropocene 10(1). DOI: https://doi.org/10.1525/elementa.2022.00030

Domain Editor-in-Chief: Detlev Helmig, Boulder AIR LLC, Boulder, CO, USA

Guest Editor: Frank Flocke, Atmospheric Chemistry Observations and Modeling Laboratory, NCAR Research Earth System Laboratory, Boulder, CO, USA

Knowledge Domain: Atmospheric Science

Part of an Elementa Special Feature: Korea-United States Air Quality (KORUS-AQ)