Detection of iodine monoxide in the tropical free troposphere

Abstract

Atmospheric iodine monoxide (IO) is a radical that catalytically destroys heat trapping ozone and reacts further to form aerosols. Here, we report the detection of IO in the tropical free troposphere (FT). We present vertical profiles from airborne measurements over the Pacific Ocean that show significant IO up to 9.5 km altitude and locate, on average, two-thirds of the total column above the marine boundary layer. IO was observed in both recent deep convective outflow and aged free tropospheric air, suggesting a widespread abundance in the FT over tropical oceans. Our vertical profile measurements imply that most of the IO signal detected by satellites over tropical oceans could originate in the FT, which has implications for our understanding of iodine sources. Surprisingly, the IO concentration remains elevated in a transition layer that is decoupled from the ocean surface. This elevated concentration aloft is difficult to reconcile with our current understanding of iodine lifetimes and may indicate heterogeneous recycling of iodine from aerosols back to the gas phase. Chemical model simulations reveal that the iodine-induced ozone loss occurs mostly above the marine boundary layer (34%), in the transition layer (40%) and FT (26%) and accounts for up to 20% of the overall tropospheric ozone loss rate in the upper FT. Our results suggest that the halogen-driven ozone loss in the FT is currently underestimated. More research is needed to quantify the widespread impact that iodine species of marine origin have on free tropospheric composition, chemistry, and climate.

Reactive iodine impacts atmospheric chemistry in several ways. Catalytic reaction cycles involving iodine atoms and iodine monoxide (IO; I_x = I + IO) destroy tropospheric ozone, which is a primary source for OH radicals (1, 2). Halogens contribute ∼45% of the ozone loss in the remote tropical marine boundary layer (MBL) (2–4). IO further affects the oxidative capacity of the atmosphere through fast reactions with HO₂ radicals and the resulting changes in HO_x (HO_x = OH + HO₂) (1, 2). Iodine also affects NO_x (NO_x = NO + NO₂) by oxidizing NO to NO₂ (1–4). Additionally, bromine atom recycling by IO increases ozone destruction and mercury oxidation rates in the MBL, resulting in higher mercury deposition rates to ecosystems and increased availability to the food chain (2, 5, 6). Finally, in coastal regions, the formation of ultrafine aerosol particles from iodine oxides can be a source of cloud condensation nuclei that can modify Earth’s albedo and thus, the radiative budget of the atmosphere (2, 7).

Oceans are the main source of iodine to the atmosphere. Most current knowledge of iodine sources and chemistry is based on measurements in the MBL (3, 8–13). IO observations at coastal MBL sites primarily link iodine sources to macroalgae (8–10). More recent studies have measured IO at open ocean sites (3, 11–13), suggesting that there might be reactive iodine chemistry over much of the open ocean. Emissions of reactive iodine species over the remote ocean remain poorly understood (11, 14) but are currently thought to be associated with primary production from biological sources (2, 14–16). Satellite maps of IO over tropical oceans (17) support the notion that iodine sources are mostly related to biological activity (2). However, the quantitative use of satellite IO data to infer iodine sources over oceans is currently limited by missing information about vertical distributions, because the satellite sensitivity varies over the height of the air column. Furthermore, the relevance of halogen chemistry for ozone loss rates is assessed using atmospheric models that cannot be validated because of the lack of halogen radical observations in the tropical free troposphere (FT). First model estimates suggest that the combined effect of iodine and bromine species could lead to 10% depletion of tropospheric ozone per year (18); the largest impact of the halogen-driven ozone loss is expected in the middle to upper troposphere. To date, there are no aircraft measurements of IO. In this study, we report IO observations in the tropical troposphere. Based on our vertical profiles, we discuss implications for the understanding of iodine sources and the relevance of the observed IO concentrations for tropospheric ozone loss rates.

Results and Discussion

Measurements.

We have measured vertically resolved concentrations of the atmospheric gases IO, water vapor (H₂O), and oxygen dimers (O₄) and spectral irradiance over much of the tropospheric air column above the remote tropical Pacific Ocean. Gases are measured simultaneously by the University of Colorado Airborne Multi-AXis Differential Optical Absorption Spectroscopy (CU AMAX-DOAS) instrument (19). Spectral irradiance was measured by the HIAPER (High-performance Instrumented Airborne Platform for Environmental Research) Airborne Radiation Package (HARP) to obtain cloud optical thickness. Both instruments were mounted aboard the National Science Foundation/National Center for Atmospheric Research (NSF/NCAR) GV aircraft (Methods and SI Text). There are no previous aircraft measurements of IO. Spectral proof of IO detection at 0.3, 1.6, and 9.5 km is shown in Fig. 1A. These spectra provide unambiguous evidence for the presence of IO in the tropical remote MBL, transition layer (TL), and FT and the changing IO abundance with altitude (Fig. 2).

Fig. 1.

Spectral proof of the detection of IO along the flight track. (A) The measured IO signals at 0.3 (MBL), 1.6 (TL), and 9.5 km (FT) altitude are overlaid on the noise level of the instrument and show the unique (fingerprint) absorption of IO as it varies with altitude. Spectra were recorded between 00:49 Coordinated Universal Time (UTC) and 01:02 UTC at 158.6° W and 6.9°–8.0° N. SCDs and rms noise values are SCD(9.5 km) = 0.7 ± 0.14 × 10¹³ molecules/cm² (molec/cm²), RMS(9.5 km) = 1.1 × 10⁻⁴; SCD(1.6 km) = 1.4 ± 0.16 × 10¹³ molec/cm², RMS(1.6 km) = 1.3 × 10⁻⁴; and SCD(0.3 km) = 2.0 ± 0.16 × 10¹³ molec/cm², RMS(0.3 km) = 1.2 × 10⁻⁴. The fit uncertainty is indicated by the SCD error. (B) The flight track is overlaid on a GOES-11 IR satellite image (Geostationary Operational Environmental Satellites, www.ncdc.noaa.gov/gibbs) from January 30, 2010 at 00:00 UTC; it shows a high rising cloud cover (dark blue and green) that is indicative of deep convection. Locations where IO was detected in the FT are shown; altitudes below 1.8 km are shaded orange. During the beginning and end of the flight, no high-sensitivity spectra were recorded.

Fig. 2.

Vertical profiles of IO and H₂O (in volume mixing ratio; A and B) and potential temperature and aerosol extinction (C). Profiles were retrieved for two different locations labeled descent and ascent (Fig. 1B). Open symbols denote data points below detection limit. The corresponding descent and ascent IO VCDs are 2.91 ± 0.78 × 10¹² and 2.49 ± 0.72 × 10¹² molec/cm². Error bars indicate the retrieval uncertainty.

Spectra were recorded by collecting photons from scattered solar light along well-defined lines of sight (varying angles forward of the plane) on January 29, 2010 during a research flight conducted to show CU AMAX-DOAS performance from Kona, HI to the south of the Hawaiian archipelago (i.e., between 2° and 19° N and 145° to 160° W). The flight targeted air masses inside the MBL as well as in the FT. Air in the FT was probed in various distances downwind of tropical deep convective activity near the Equator. The flight track is shown overlaid on an IR satellite cloud image in Fig. 1B together with significant (above detection limit) measurements of IO slant column densities (SCDs; i.e., integrated concentrations of all photon paths along the line of sight) in the FT (here, above 1.8 km). Lower altitudes have been excluded to emphasize the fact that IO was found in significant concentrations over large spatial scales in the FT.

Vertical trace gas concentration profiles of IO and H₂O were derived from one aircraft descent and ascent, because maximum vertical information is gained from measurements during aircraft altitude changes. Concentration profiles were retrieved between 0.3 and 10 km at two different locations indicated as descent and ascent in Fig. 1B. Fig. 2 A and B shows the vertical profiles expressed as volume mixing ratio [IO in parts per trillion (ppt) by volume: 1 ppt = 10⁻¹² ≅ 2.5 × 10⁷ molec/cm³ for T = 298 K and P = 1,013 mbar; our conversion is based on aircraft temperature and pressure measurements]. The vertical distributions of IO and H₂O exhibit different but strong variations with altitude. We find IO mixing ratios of 0.5 ± 0.06 ppt inside the MBL that decrease by about a factor of five in the FT and remain largely detectable over the entire probed air column. Our vertical profiles establish that most of the IO vertical column is located above the MBL. When the concentration of IO is integrated over altitude, then the partial vertical column densities (VCDs) above 800 m account for about two-thirds of the overall tropospheric column amounts (65.3%; average of both profiles) (Table 1). The water mixing ratio retrieved by CU AMAX-DOAS for the ascent profile ranges from 2.4 ± 0.3% to 0.25 ± 0.2% and agrees well with 2.3 ± 0.4% to 0.26 ± 0.02% measured by a frost point hygrometer on the plane. This good agreement confirms the accuracy of our retrievals in complex radiation fields below cirrus clouds (Figs. S1 and S2).

Table 1.

Comparison of averaged retrieved IO VCDs and modeled satellite SCDs (×10¹² molec/cm²)

VCD	Cloud cover (%)	SCD/VCD bias
Total
2.71	0	4.25/1.5
2.71	20	4.34/2.5
2.71	40	4.38/3.8
MBL
0.94	0	0.99 (23.3%)
0.94	20	0.60 (13.8%)
0.94	40	0.39 (8.9%)
TL
0.76	0	1.17 (27.5%)
0.76	20	1.29 (29.7%)
0.76	40	1.36 (31.1%)
FT
1.01	0	2.09 (49.2%)
1.01	20	2.45 (56.5%)
1.01	40	2.63 (60.0%)

The VCD bias expresses the overestimation of the VCD as a factor when the measured SCD signal is converted into a VCD based on an MBL-only vertical profile assumption.

The retrieved IO profiles in both locations show surprising similarities over the full tropospheric column, whereas water vapor measurements are significantly different above 2.5 km. The elevated water vapor indicates influences of convective outflow in the descent profile. By contrast, the ascent profile is markedly drier. Fig. 2C shows potential temperature (based on aircraft temperature and pressure measurements) and aerosol extinction derived from AMAX-DOAS O₄ measurements (SI Text and Tables S1 and S2). The top of the MBL is marked by an inversion of about a 3 K change in potential temperature, indicated by arrows in Fig. 2C, for both the descent and ascent profiles. During ascent, the potential temperature is not constant within the MBL but decreases continuously to the ocean surface, suggesting a stratified MBL. The TL is decoupled from the MBL and capped by a stronger temperature inversion of around 9 K at an altitude of about 1.8 km. The high altitude of the capping inversion is indicative of a decoupled TL (20, 21). Wood and Bretherton (20) find that, after the potential temperature inversion height exceeds 1 km, a TL is often decoupled from the near-surface air below, particularly in far offshore locations such as our location within the trade wind belts. Consistent with this finding, our aerosol extinction profiles are relatively constant below 800 m and decrease rapidly to background levels through the TLs in both profiling locations. Water vapor also shows a clear decrease near the top of the MBL and throughout the TL. The gradient in water vapor is particularly strong for the ascent profile, where fewer scattered clouds were present than during descent (Fig. S3). These observations suggest that MBL, TL, and FT are separate dynamical and chemical regimes, which is further supported by meteorological modeling.

Meteorological Modeling.

Air mass back trajectories were calculated with the weather research and forecast model (WRF) (22) (Methods). WRF was used to model MBL heights (23), which for our study area, are found to range between 500 and 1,000 m. Fig. S4A shows that boundary layer air at both profiling locations originates consistently from easterly low-level trade winds, with no apparent free tropospheric influences for about 50 h (Fig. S4B). Flow in the TL is also from the east (Fig. S4C). Time-resolved back trajectories initiated at 1.5 km altitude show that there has been almost no dynamic transport from altitudes below 800 m for the past 10–14 h (Fig. S4 D and E), indicating that air in the TL has not been in contact with the ocean surface for about one-half of 1 d. Air masses probed above 2 km are of a different origin for the ascent and descent profiling sites. Starting at 3.5 km, a bifurcation of trajectories is observed near 8° N (Fig. S5 B–F). Free tropospheric air masses probed to the south of 8° N (descent) were influenced by convective outflow at multiple altitudes, whereas free tropospheric air masses north of 8° N (ascent) had not been in contact with the MBL for at least 60 h (Fig. S5 I–L). Based on the cumulative observational and modeling evidence, we conclude that air inside the MBL is in direct contact with the ocean surface only at the descent location, whereas TL air is essentially decoupled from the MBL in both profiling locations (Fig. S3). We define MBL here as the layer extending from the surface to about 800 m, the TL refers to air between 800 m and about 1.8 km, and FT air is above 1.8 km. Implications for potential IO precursors and lifetimes in these distinctly different compartments of the atmosphere are discussed below.

Widespread IO in the FT.

Roughly one-third of the IO vertical column is located in the FT. This fraction is estimated conservatively, and could be up to 30% higher because of a temperature dependence of the IO absorption cross-section (24). Our observations are consistent with recent first measurements of IO in the lower subtropical FT by means of mountain top MAX-DOAS (Canary Islands at 2.4 km altitude) that estimate 0.2–0.4 ppt IO between 1 and 10 km with limited vertical resolution (25). Previous measurements of the atmospheric column abundance of IO in midlatitudes over the continental United States (Kitt Peak Observatory in Arizona at 2 km altitude) show 0.12 ppt of IO in the stratosphere (26). These ground-based direct sun column observations constrain the sum of tropospheric and stratospheric IO column amounts, but they do not provide altitude information. The average FT vertical column of our observations is 1 × 10¹² molec/cm². A similar tropospheric VCD over land would have created a signal of the same order of magnitude as the reference noise reported for the Arizona measurements (26). Therefore, free tropospheric IO could be present over the continental United States and would be consistent with our measurements. In the context of these studies, we conclude that IO is likely a component of lower free tropospheric air on global scales. Balloon-based measurements in the tropics and mid and high latitudes reported IO upper limits of 0.1 ppt in the upper FT and lower stratosphere (27, 28). Our measurements establish that this mixing ratio is present over most of the free tropospheric air column and extended spatial scales.

Relevance for Satellite Retrievals.

The marine atmosphere over the open ocean is still one of the most poorly probed atmospheric environments on our planet. Satellite measurements are, therefore, a particularly useful source of data over remote oceans. To convert satellite-retrieved IO SCDs into VCDs, knowledge about vertical profiles is required. Current atmospheric models cannot provide accurate profile information because of gaps in our understanding of the source mechanisms of IO over oceans (11, 14). We have simulated the satellite view of our measured vertical profiles by using a radiative transfer model. Satellite SCDs assuming a nadir viewing direction from space were calculated for different TL cloud fractions (SI Text and Table S3). The comparison of IO FT VCDs (average of both profile locations) and corresponding satellite SCDs (Table 1) reveals that FT VCDs account for roughly one-third of the total columns but contribute about 50% to the measured satellite SCD signal because of increased satellite sensitivity to the FT vs. the MBL. This contribution increases with cloud cover, reflecting the effect of increased sensitivity of satellites to partial columns located above a region of high albedo (i.e., above clouds) and the partial shielding of gases in the boundary layer by clouds. At a moderate cloud cover of 40%, only about 9% of the IO satellite signal originates from within the MBL.

The conversion of measured satellite SCDs into VCDs is often based on the assumption that trace gases are only located in the MBL (2, 11, 13, 17). Based on our profiles, the satellite VCDs could be lower by a factor of 1.5–3, depending on cloud fraction (expressed as VCD bias in Table 1). This uncertainty reveals the importance of independent vertical profile information to convert SCDs into VCDs. However, if there is, indeed, a widespread presence of IO in the FT as our analysis suggests, current satellite differential SCD (dSCD; i.e., retrieved SCD with respect to a reference spectrum) measurements may also indicate lower limits of the IO abundance, as an unknown amount of IO present over the region used to record a satellite reference spectrum (17) would reduce the retrieved satellite dSCDs. Notably, the simulated satellite SCDs from Table 1 are comparable with current satellite dSCD detection limits over oceans, which are on the order of 4–8 × 10¹² molec/cm² (corresponding to rms noise levels of 1–2 × 10⁻⁴ for single acquisitions) (17). This effect could, thus, be of similar magnitude but opposite sign than the VCD bias caused by assumptions on vertical distributions (Table 1). Improved signal to noise from averaging existing satellite data in space or time is a very promising method to further investigate the possible global presence of IO. Advances in satellite IO measurements also mean an increasing need for an independent assessment of the measured dSCDs and vertical trace gas distributions, which is now possible from aircraft.

Implications for Iodine Sources.

An increasing body of evidence from laboratory experiments (29–32), field observations (11, 12, 14), and modeling studies (11, 14) suggests that very short-lived polyhalogenated iodocarbons, such as diiodomethane (CH₂I₂; photolytic life time of 2–10 min), bromoiodomethane (CH₂IBr; 1–2.5 h), or chloroiodomethane (CH₂ICl; 2.4–8 h), as well as molecular iodine (I₂; 15 s) contribute significantly to the iodine source flux in the MBL and are needed to sustain elevated IO abundances in the remote MBL (2, 16, 33). Our airborne observations of ∼0.5–0.6 ppt IO in the central Pacific MBL are slightly lower but generally consistent with ∼1.5 ppt IO at Cape Verde Islands in the tropical Atlantic ocean (3, 11), ∼0.9 ppt from ship observations over the Eastern Pacific (12, 13), and up to a few ppt over upwelling areas a few hundred kilometers from the Peruvian coast (12). However, in our vertical profiles, the larger share of the IO VCD is located above the MBL. In particular, elevated IO in the decoupled TL and aged FT air cannot be explained by polyhalogenated iodocarbons and I₂, which react inside the MBL. Only iodocarbons like methyl iodide (CH₃I; lifetime of 5–6 d) or ethyl iodide (C₂H₅I; 4 d) and possibly, aerosols are precursors that are long-lived enough to carry iodine into the remote upper tropical FT by means of tropical deep convective transport pathways. Among these compounds, CH₃I is the most abundant (2, 15).

Distributions of chlorophyll-a (Chl-a) in the surface ocean are currently being used to scale very short-lived marine iodocarbon emission (16, 18). Positive correlations between satellite IO and Aqua/Moderate Resolution Imaging Spectroradiometer (MODIS) satellite-derived Chl-a (SI Text and Figs. S6 and S7), indeed, seem to provide some evidence for the relevance of biological sources over the Eastern Pacific, but they are in contrast to recent ship-based measurements that show the lowest I_x in areas of high Chl-a (i.e., show a negative correlation between Chl-a and I_x in the MBL) (13). Our profiles suggest that scattered light satellite IO signals over tropical oceans primarily indicate FT-IO. This finding implies that satellites could provide information that is essentially decoupled from the ocean surface (Table 1). IO in the FT could, thus, help explain why IO columns measured from satellites and ships scale differently with Chl-a observations. The spatial disconnect between atmospheric IO and the Chl-a concentration, however, poses questions about which iodine emissions can legitimately be scaled using Chl-a distributions. Notably, some of the highest CH₃I concentrations observed are found over the Eastern Pacific ocean (34), where deep convection also provides a transport pathway into the FT. The apparent correlation between satellite IO and Chl-a could possibly be because of the coupled effect of biological sources producing IO precursors with a longer lifetime and dynamical impacts on their distribution.

Interestingly, in the ascent profile, IO decreases by only 24% (±24%) from the MBL to the TL. For comparison, the TL is depleted by 60% (±10%) in aerosol extinction and contains 79% (±14%) less water vapor than the MBL; an even higher chemical depletion is expected for the short-lived iodine precursors (11, 14). If the decrease in aerosol extinction and water can be taken as an indicator for the dilution of long-lived iodocarbons, the IO source flux from known precursors in the TL is likely much lower (factor of 3–15) than in the MBL. Furthermore, the I_x lifetime with respect to irreversible uptake of IO and HOI to aerosol surfaces is on the order of 1 h (for a typical aerosol surface area of 10 mm²m⁻³ in the TL). In contrast, a 24% decrease in IO over 12 h (see above) suggests a much longer effective IO lifetime (∼44 h). These I_x lifetimes might be conservatively estimated. We conclude that, to sustain the elevated IO concentrations in the TL, an efficient IO regeneration mechanism must be operating. We hypothesize that iodine recycling from aerosols back to the gas phase sustains IO concentrations in the TL. Aqueous surfaces containing iodide are known to release I₂ and IO on reaction with ozone (32, 35); additionally, the multiphase reaction of HOI with dissolved halides could contribute to recycling of iodine back to the gas phase, analogous to bromine chemistry (36). Iodine is also an abundant component of FT aerosols (37). Based on the similarity of our IO vertical profiles in the FT, we speculate that the recycling from aerosols back to the gas phase could further extend the IO effective lifetime in the FT. There are currently no simultaneous measurements of IO and iodocarbons in the FT.

Relevance for Atmospheric Chemistry.

Tropospheric ozone is a greenhouse gas and the primary source for OH radicals, which are an important sink for methane in the tropical atmosphere (38). Atmospheric models estimate that halogen-mediated ozone loss could deplete the tropospheric ozone column by 10% per year (18). These estimates are as of yet unconstrained by halogen radical observations in the FT and predict ∼0.02 ppt IO in the upper troposphere over the tropical Atlantic (18), which is a factor of five less IO than the upper limit reported for this region (28) and the FT–IO mixing ratio that we find over the Central Pacific ocean. The possibility of iodine recycling from aerosols could further extend the effective iodine lifetime and add an additional ozone loss pathway that is not yet considered in atmospheric models. We investigate the relevance of halogen-mediated ozone loss as a function of altitude using our IO vertical profiles and other flight observations to constrain a photochemical model (SI Text). Fig. 3A shows that the ozone loss rate is a strong function of altitude, whereas the iodine contribution to the overall ozone loss rate (Fig. 3A) is a strong function of ozone concentration (Fig. 3B). Fig. 3A assumes that the average ozone vertical profile as measured during the Pacific Exploratory Mission (PEM) Tropics field campaign (same area as our flight track during March of 1999) is representative for our case study and further assumes that 0.5 ppt bromine oxide (BrO) is present over the entire tropospheric air column (6, 18, 39–41) (SI Text). Under the rather low ozone concentrations observed during PEM tropics (Table S4), iodine chemistry determines 23%, 26%, and 11% of the overall ozone loss rate in the MBL, TL, and FT, respectively. Sensitivity studies (Fig. 3B and Fig. S8 A and B) show that, for ozone concentrations below 40–50 parts per billion (ppb), the fraction of iodine-induced ozone loss generally is around 10% (7–15%) over most of the tropospheric air column. Higher ozone is the primary reason why this fraction decreases by up to two orders of magnitude in the stratosphere (Fig. 3B), consistent with previous estimates (26). At constant O₃, our simulations show an increase in the OH radical concentrations caused by the IO + HO₂ reaction that is most prominent in the MBL (7.8%) and leads to increases of 4.6% and <0.5% in the TL and FT, respectively (SI Text and Table S5). The fraction of iodine-induced ozone loss is largely insensitive to photolysis frequencies and NO₂ (Fig. S8 C and F).

Fig. 3.

Ozone loss simulations. (A) The total ozone loss rate and percent contributions constrained by IO observations and simulated bromine, HO_x, photolysis, and NO_y chemistry during the ascent profile (base case simulation; conditions in SI Text). (B) Sensitivity of the percentage contribution of iodine-induced ozone loss to the ozone mixing ratio (base case, O₃-varied). The shaded orange line indicates base case ozone levels (A).

In our base case simulation (Fig. 3A), 0.5 ppt BrO and measured IO correspond to 0.61 ppt Br_x (Br_x = Br + BrO) and 0.23 ppt I_x (average mixing ratio below 10 km), which are responsible for 19% and 14% of the column average ozone loss rate below 10 km, respectively. Per X_x molecule (X = Br, I), iodine is about two times as efficient at destroying ozone than bromine. In absence of any BrO, the fraction of iodine-induced ozone loss would be slightly larger (Fig. S8D). Arguably, uncertain BrO concentrations in the FT currently limit our ability to quantify the impact of halogen-mediated tropospheric ozone loss (Fig. S8), and therefore, simultaneous observations of BrO and IO from research aircraft are desirable.

Interestingly, most of the iodine-induced partial column ozone loss occurs in the TL (40%), with MBL and FT contributing 34% and 26%. The TL is a unique chemical environment characterized by moderately warm temperatures, the presence of water, and aqueous (low viscosity) aerosols and clouds. Entrainment of ozone from the FT often leads to higher ozone concentrations here than in the MBL (42), and the lack of contact with the ocean surface can extend the lifetime of reactive species like IO, thus increasing its relevance for HO_x, ozone, and mercury chemistry. The chemical state of the TL remains poorly understood and warrants increased attention in future field studies. Simultaneous measurements of IO and its precursors are now possible from research aircraft, and they hold great potential to shed light on the relative importance of organic iodine precursors and iodine recycling pathways from aerosols as sources for IO in the FT and advance our understanding of iodine chemistry in the global atmosphere.

Methods

We measured solar scattered light spectra with the CU AMAX-DOAS instrument (19) on board the NSF/NCAR GV research aircraft (HIAPER) during a 9-h flight on January 29, 2010. The descent and ascent profiles were flown from 00:47 to 01:03 Coordinated Universal Time (UTC) and 01:40 to 02:25 UTC (January 30, 2010) at solar zenith angles of 39.0°–43.2° and 51.8°–64.2°, 6.6° N/158.7° W to 8.1° N/158.5° W, and 10.6° N/158.0° W to 14.2° N/157.2° W (a distance of about 400 km apart). The profile retrieval uses a two-step process: (i) DOAS analysis of spectra to retrieve trace gas SCDs and (ii) profile retrieval from inverse radiative transfer modeling (SI Text). Based on measurement errors and vertical information content of the SCDs, absolute detection limits in the FT for our profile retrieval are 0.06 ppt for IO and 0.05% for H₂O. Partial column amounts were integrated over altitude for the MBL (0–800 m), TL (800–1,800 m), and FT (above 1.8 km).

The HARP is a comprehensive atmospheric radiation suite used to measure spectrally resolved in situ actinic flux and irradiance. Up-welling and down-welling irradiance was measured from 300 to 2,400 nm at 1 Hz. Cloud optical thickness was derived from either transmitted or reflected irradiance at 500 nm.

The radiative transfer model McArtim (43) was used for the interpretation of our profile retrievals and calculation of the satellite view of our profiles. The profile retrievals used McArtim as the forward model combined with an in house–developed inversion algorithm to account for the light path dependency of the measured SCDs. Atmospheric constraints were provided by avionics temperature and pressure profiles, O₄ columns measured by CU AMAX-DOAS, and cloud optical depth inferred from HARP (SI Text). Satellite SCDs were simulated using McArtim-calculated weighting functions and representative settings for current instruments measuring solar scattered light [e.g., Global Ozone Monitoring Experiment (GOME), Scanning Imaging Absorption Spectrometer for Atmospheric Chartography (SCIAMACHY), Ozone Monitoring Instrument (OMI), and GOME-2] (SI Text).

Two separate model runs were conducted with WRF (22): (i) a 60-h forecast with 30-km horizontal grid spacing and (ii) a 24-h forecast with increased horizontal resolution (12 km) to better resolve local deep convection. The model was initialized at 12:00 UTC on January 27, 2010 for the 60-h run and 00:00 UTC on January 29 for the 24-h run. The domain was centered at 9° N, 156° W and covered 20° in each direction for the longer run and 8° for the shorter period. Initial conditions were based on National Centers for Environmental Predictions–Global Forecast System (NCEP-GFS) analyses. Trajectories were computed from the WRF output for a grid of points along the flight track.

The ozone loss rate calculations were performed using a photochemical box model in which modeled species reach steady state over multiple days. The model conceptually follows the work by Crawford et al. (44), and the chemical mechanism includes both gas-phase and heterogeneous reactions of iodine and bromine species based on works by Ordóñez et al. (16), Sommariva et al. (45), and Parrella et al. (39). Details and sensitivity studies are in SI Text.