2.1 CHACHA campaign and location

The CHACHA field campaign took place from mid-February to mid-April of 2022 and was based out of Utqiaġvik, Alaska, which is located at the northernmost point of Alaska between the Chukchi and Beaufort seas. A ground-based station was set up to study turbulence at the Earth's surface and monitor ozone mixing ratios throughout the campaign. Two aircraft were used for the campaign, a University of Wyoming King Air aircraft and the Purdue ALAR aircraft, which was equipped with instrumentation to study gas-phase and particulate halogens, particle sizes, and cloud droplets. However, this paper only utilizes data from ALAR, which was equipped with a best air turbulence (BAT) probe that provided 50 Hz wind and pressure measurements (Garman et al., 2006, 2008), two roughly 2 Hz temperature probes, a 2B Technologies model 205 ozone monitor with 10 s time resolution, and a Grimm particle counter (model 1.109, briefly described in Peterson et al., 2017) that recorded size-resolved distributions every 6 s. Utqiaġvik is also the location of an Aeronet station located at the North Slope of Alaska Atmospheric Radiation Measurement (ARM) facility. These data were critical to characterize the impacts of particle light scattering on our remote sensing observations. Among the numerous other instruments housed at the ARM site is a total-sky imager with data on total cloud cover (Flynn and Morris, 2015), which can provide useful context for the conditions under which observations were made.

As Utqiaġvik is located at the northern tip of Alaska, ALAR flew over sea ice to the west, north, and east and over tundra to the southwest, south, and southeast. Few flights went further south than Atqasuk, Alaska, where the topography starts to rise, so the ground elevation was close to sea level for most observations. Regardless, all altitude plots shown throughout this work are above ground level (a.g.l.). The town of Deadhorse, Alaska, which sits in the middle of the Prudhoe Bay oil field, is located roughly 320 km southeast of Utqiaġvik, and there are many facilities in the area that emit pollutants. NO_x emissions in the area were dominated by two specific facilities (the Central Compressor Plant and Central Gas Facility) during the campaign as observed with the HAIDI nadir spectrometer. As these facilities are in close proximity at less than 1 km apart, Prudhoe Bay will be shown as a single point source centered between these facilities throughout this study.

2.2 HAIDI and dSCD retrieval

In this study, HAIDI was configured with two spectrometers having charge-couple device (CCD) detectors that measure light from either a nadir whisk-broom scanner or a forward near-limb push-broom imager (detailed in General et al., 2014). Both spectrometers recorded incoming light from roughly 301–408 nm, allowing for the observation of NO₂, BrO, and O₄, which is the collision-induced absorption of O₂. This study only uses observations from the near-limb spectrometer, which recorded incoming light over the 256 CCD channels that were binned into four different elevation angles of −3.6, −1.9, 0, and +1.6^∘ relative to the aircraft central axis, each with a field of view of 3.3^∘ full width at half maximum as measured in the field, though the most upward and downward views had some truncation at the edges. The telescopes were fixed with respect to the aircraft, and thus the viewing angle varied with the aircraft pitch angle.

HAIDI utilizes the differential optical absorption spectroscopy (DOAS) technique, detailed in Platt and Stutz (2008). Individual trace gases are quantified by identifying their narrowband absorption in measured scattered solar spectra. The logarithm of recorded spectra relative to a reference spectrum, taken in-flight with minimum background trace gas concentrations (more information in the Supplement), is fitted with a linear combination of trace gas absorption cross-sections and a polynomial that accounts for the spectrally broader features of Rayleigh and Mie scattering (see DOAS book for details, Platt and Stutz, 2008). The specific fit parameters and cross-section references used to retrieve BrO, NO₂, and O₄ differential slant column densities (dSCDs) and an example fit retrieval are shown in the Supplement (Table S1, Fig. S1).

The coefficients of the DOAS fit routine are dSCDs. A slant column density (SCD) describes the total number of trace gas molecules along light paths that lead from the sun through atmospheric light-scattering events to the observing telescope. The measurement is differential because the reference spectrum has some amount of trace gas absorption in it, and the dSCD represents the difference in column abundance from this reference. The observed dSCD is thus a function of the amount of trace gas in the atmosphere, location of the trace gas, solar geometry, measurement geometry, and aerosol light scattering. A radiative transfer model is used to account for the solar and measurement geometry, especially for near-limb observations where the sensitivity to different levels of the atmosphere varies by orders of magnitude.

The utility of an AMAX-DOAS instrument is that the combination of observations with different viewing geometries provides information on the vertical structure of the detected trace gas around the flight altitude, as these different viewing geometries provide variable sensitivity to different parts of the atmosphere. For instance, a limb observation (tangent to the Earth) is most impacted by trace gas absorption at the flight altitude, whereas an upward viewing angle will have higher sensitivity to the atmosphere above the aircraft, and vice versa. Radiative transfer model results describe how trace gases at each altitude impact the observations to determine a corresponding trace gas vertical profile that fits the observed dSCDs.

2.3 Porpoising

As each individual observation is most sensitive to the area around the flight altitude, we combined observations at different altitudes to increase the amount of information we could retrieve on the vertical trace gas profile, as demonstrated in Baidar et al. (2013). To maximize our retrieved information content, the aircraft often maintained a constant heading while varying altitude between a minimum altitude (usually < 100 m) and a maximum altitude (> 600 m), as seen in Fig. 1 in a method known as “porpoising” (Gerber et al., 2013). By maintaining the heading, the radiative transfer complications of comparing multiple observations are greatly simplified, as the solar geometry is relatively unchanged. We then combined all observations from a porpoise (down and up) to retrieve a single high-resolution BrO and NO₂ profile. The mean horizontal distance that the aircraft traveled during these profiles was roughly 35 km. We therefore sacrifice horizontal resolution for vertical resolution. This choice likely has little impact for species that do not exhibit small-scale horizontal features, for example when quantifying background BrO profiles which should be relatively constant over such length scales (Simpson et al., 2017). However, this method would be limited in quantifying gases with sharp spatial gradients, such as near-source power plant NO₂ plumes, and other gases affected around such sharp gradients.

Figure 1Example flight pattern for ALAR during the CHACHA campaign with the (above ground level – a.g.l.) altitude profile shown in panel (a) and the flight map shown as a function of time in panel (b). The black dots denote the location at the center of each porpoise. By changing altitude and maintaining heading, the radiative transfer considerations are simplified, allowing for easier and more accurate retrieval of high-resolution profiles of BrO and NO₂ mixing ratios.

Download

2.4 Optimal estimation

An optimal estimation algorithm is used to determine mixing ratio profiles from AMAX-DOAS observations based on the sensitivity of the measured parameter (dSCD) to the state parameter we wish to retrieve (mixing ratio).

The details of the optimal estimation, or inversion, algorithms used in this study are described in the Supplement, including the standard mathematics from Rodgers (2000). These algorithms require radiative transfer modeling to describe how light reaches the four telescopes and thus how gases at different altitudes impact the observed dSCDs for the different viewing angles.

For this work, we used VLIDORT version 2.8 (Spurr et al., 2022), which is a 1-D radiative transfer model that calculates single-scattering processes in a full-spherical geometry and multiple-scattering processes with a first-order quasi-spherical approximation. The VLIDORT atmosphere used in this work was manually initialized and consisted of 78 layers up to 48 km altitude, with variable grid spacing ranging from 50 m for altitudes below 2 to 4 km spacing at the top of the grid. The rest of the model initialization is described in the Supplement. VLIDORT is used to calculate trace gas box air mass factors (BAMFs, Wagner et al., 2007), which describe the horizontal path-length enhancement of light through each modeled altitude grid cell (Hönninger et al., 2004) and are used as the forward model matrix (K) row vectors for the BrO profile optimal estimation (Baidar et al., 2013).

However, in order to accurately model how light reaches the telescope, it is particularly important to properly represent particle extinction in the atmosphere, as this greatly impacts how light travels through the lower atmosphere. The aerosol optical properties are calculated with a size-resolved bulk Mie code coupled to VLIDORT, with the input parameters described in the Supplement, and the particle vertical profile is retrieved using O₄ dSCD observations (Frieß et al., 2019). The vertical profile of O₄ depends on the square of O₂ concentration and thus on temperature and pressure, meaning that the vertical profile of O₄ is relatively well known. The observed O₄ dSCDs are reliant on the vertical profile of O₄ concentration and how light travels through the atmosphere. In a cloud-free atmosphere this is mainly a function of aerosol particle extinction, so these dSCD observations can be used to retrieve the vertical profile of the aerosol extinction coefficient.

For the O₄ optimal estimation inversion to determine the particle extinction profile, we utilized the Levenberg–Marquardt method (Levenberg, 1944; Marquardt, 1963), which is an iterative damped least-squares optimization method particularly useful for solving nonlinear inverse problems. The particle profile optimal estimation requires VLIDORT to calculate the impact of particle extinction on incoming light with and without O₄ absorption (Eq. 7 in the Supplement). The details of the optimal estimation algorithm are described in the Supplement.

To save computational time, we randomly selected 200 observations, roughly 5 % of an average 4141 observations per flight, from the different porpoises throughout each flight for use in this retrieval of the flight's average aerosol particle profile. Therefore, a constant particle extinction profile is assumed for each flight, as the mostly clear-sky nature of the days when ALAR flew is associated with strong surface temperature inversions that result in a stable atmosphere. These 800 data points, due to the four forward viewing angles, result in excellent particle profile retrievals, with a mean 15 degrees of freedom (DOF). An example of the particle profile averaging kernel matrix and retrieved particle extinction is shown in Fig. 2, where the averaging kernel represents the fraction of the retrieved information at each altitude that is due to observations at all levels. Measured and modeled O₄ dSCDs can be seen in Fig. S3, where modeled dSCDs fit all porpoise observations with an R² of 0.824.

Figure 2Example retrieval of the average particle extinction profile from 30 March, with the averaging kernels in panel (a) (DOF of 17.6) and the extinction profile in panel (b). The a priori extinction is based on the profile shape from the Grimm particle number density from throughout the flight scaled to Aeronet observations at the ARM facility. VLIDORT modeled how light at 361 nm reached our telescopes to perform the O₄ inversion and thus retrieves the 361 nm particle extinction. VLIDORT uses a bulk Mie code to determine aerosol optical properties at other wavelengths. The impact of Rayleigh scattering is shown in red, indicating that the particle extinction and molecular extinction are comparable at the surface for this day.

Download

The retrieval of a BrO profile is more straightforward than for O₄ and requires a simple linear inversion that is only performed once per profile. VLIDORT was constrained for each individual flight with the particle extinction profile calculated from the O₄ inversion and operated at a wavelength of 350 nm to calculate BAMFs for each flight for all four forward viewing angles by adding the mean aircraft pitch angle to the relative viewing angles for each individual observation. All BrO dSCDs from the four viewing angles are combined for each individual porpoise maneuver and used to retrieve a single high-resolution BrO profile, with the standard optimal estimation mathematics (Rodgers, 2000) as described in the Supplement.

Figure 3 shows the high confidence in the retrieval below the maximum flight altitude. For this profile, the aircraft flew between 64 and 967 m above ground level, and all averaging kernels in this range are close to 1. Further, the averaging kernels peak sharply at their own altitude and generally decrease by roughly 90 % at the two nearest altitudes (i.e., ±50 m), indicating that the retrieved mixing ratio at each altitude is primarily impacted by the measurement sensitivity to that same altitude and thus that this retrieval has a high vertical resolution, similar to the 50 m model resolution. The degrees of freedom (DOF, trace of the averaging kernel matrix) indicate how many independent pieces of information we can retrieve from each profile. For this case, we see that 16.7 DOF corresponds to highly independent retrievals below roughly 850 m (due to the 50 m grid spacing). All retrieved profiles were filtered based on confidence (DOF > 7.5), spatial resolution (horizontal porpoise distance < 100 km), and flight altitude (minimum altitude < 300 m), resulting in 245 BrO profiles from the full campaign (from 27 February to 16 April 2022).

Figure 3Example BrO profile retrieval from the fifth aircraft profile on 30 March, with the averaging kernels in panel (a) showing very good resolution and a high total DOF (16.7). The a priori profile is shown in black and the retrieved profile and uncertainty are shown in blue in panel (b). This vertical profile spanned a horizontal distance of 33.45 km.

Download

3.1 Average BrO profile

Using both sides of the aircraft profiles (down and up) results in very well-resolved BrO profiles throughout the lower atmosphere, as shown in Fig. 3, with a mean DOF of 18 throughout the campaign. This sampling also includes some profiles where only the descent was used in cases in which there was no immediately proceeding ascent, or vice versa. However, these profiles often lead to good retrieval results as well, with a mean DOF of 16.

The focus of this study is to improve the understanding of reactive bromine chemistry of the springtime Arctic. Therefore, the median BrO profile from the full campaign is of particular interest. Figure 4 summarizes all BrO profiles with a box-and-whisker plot. The median BrO mixing ratio is shown as the center line with the boxes depicting the interquartile range of the observations and the 5th and 95th percentile intervals as whiskers at each altitude.

Figure 4A box-plot summary of all BrO profiles from throughout the CHACHA field campaign. The box plot indicates the median BrO mixing ratio at each altitude (orange), with the edges of the box denoting the 25th and 75th quantiles and the whiskers denoting the 5th and 95th quantiles. In red is the mean BrO box profile, which demonstrates the quick decrease in the BrO mixing ratio with altitude.

Download

We retrieved a median BrO mixing ratio of 29 pmol mol⁻¹ (mean = 30 pmol mol⁻¹, SD = 13 pmol mol⁻¹) in the lowest grid cell (from 0–50 m), and BrO quickly decreases above the Earth's surface, likely due in part to the typical thermodynamic stability of the atmosphere in the springtime Arctic that inhibits vertical mixing (Bradley et al., 1992; Creamean et al., 2021; Peterson et al., 2015). Above the surface, BrO is roughly constant near 10 pmol mol⁻¹ for the next 150 m, possibly due to a residual layer above the common surface inversions. Above 225 m, BrO decreases to the a priori value of 1 pmol mol⁻¹ with mean values comparable to the standard deviation. However, it is clear from the width of both the boxes and whiskers that there is much variability in the BrO profile below 400 m altitude, where the inter-quantile range of the observations is greater than 3 pmol mol⁻¹ and greater than 15 pmol mol⁻¹ at the Earth's surface.

3.2 BrO profile uncertainty

The uncertainty of the retrieved BrO profiles can be attributed to a combination of smoothing error, retrieval noise, and forward model parameter error (each described in the Supplement, Rodgers, 2000). The average BrO profile uncertainty peaks at the surface with a value of 5 pmol mol⁻¹. This is in part due to the fact that viewing angle uncertainty and particle extinction uncertainty have a larger impact on BAMF calculations near the surface, where light-path truncation by the surface is greatly affected by both parameters. The BrO profile uncertainty decreases with altitude to be less than 3 pmol mol⁻¹ above 250 m, at which point it linearly decreases to roughly 1 pmol mol⁻¹ above 1 km altitude. The mean BrO profile error is roughly equal to the mean BrO profile above 500 m altitude. Similarly, the standard deviation of the retrieved BrO profiles is larger than the mean uncertainty below 450 m and is over twice as large at the surface. Therefore, most of the variance observed in the BrO profiles in the lower atmosphere is significant.

3.3 Cluster classification of the BrO profiles

Previous BrO MAX-DOAS studies investigate BrO profile shapes through calculations of lower-tropospheric vertical column densities (LT-VCDs) and f₂₀₀, where the LT-VCD is the vertical BrO column density from 0–2 km and f₂₀₀ is the fraction of the LT-VCD below 200 m (Peterson et al., 2015; Simpson et al., 2017). Due to the aircraft measurements and the resulting increased DOF, we are able to study differences in BrO profile shapes with finer detail.

To characterize the large vertical variability of BrO, we used a K-means cluster analysis of the BrO profiles (MacQueen, 1965; Lloyd, 1982). A cluster analysis is an unsupervised algorithm that simply finds common BrO profile shapes that occurred throughout the campaign, and there are several tests to determine how many clusters to include. We utilized a silhouette score (Rousseeuw, 1987) that quantifies (on a scale from −1 to 1) how much better a profile's cluster explains its shape compared to the next closest cluster. The highest average silhouette score corresponded to two clusters. However, this resulted in many clusters with negative silhouette scores, indicating that they may be assigned to the wrong cluster. We therefore opted for the second-highest average silhouette score (Fig. S4), which also minimized negative silhouette score profiles and corresponded to four BrO clusters. The utility of the silhouette scores is that it ensures that no two clusters are too alike. If that were the case, the algorithm would suggest that the number of clusters be reduced and the two clusters would be merged together.

The four identified clusters are shown in Fig. 5, which are defined from most common to least common as low BrO, high surface BrO, high total BrO, and lofted BrO. The low-BrO case was identified 44 % of the time. Although there was still 20 pmol mol⁻¹ of BrO at the Earth's surface (between 0 and 50 m), there was very little (< 10 pmol mol⁻¹) above the surface on average. Similarly, the high-surface-BrO case (33 % of the retrievals) also resulted in less than 10 pmol mol⁻¹ above the surface, although there was roughly twice as much BrO at the surface compared to the low-BrO case. The high-total-BrO case was identified 17 % of the time and resulted in large observed BrO vertical columns due to a combination of high BrO at the surface and aloft. Least frequently (6 % of the retrievals), we observed lofted-BrO profiles, with 73 % of these 15 observations occurring on a single day (19 March 2022), which yielded in the largest observed BrO vertical columns.

Figure 5The four BrO profile clusters identified from the entire dataset. Profile clusters are shown in decreasing occurrence, with low BrO being the most common and lofted BrO the least common. Error bars denote the standard deviation of all clustered profiles around the mean profile. Each plot shows the number of observed profiles in each cluster along with the mean lower-tropospheric (< 2 km) vertical column density and the ratio of the LT-VCD below 200 m.

Download

The LT-VCD and f₂₀₀ observed here (Fig. 5) are consistent with previous literature. Simpson et al. (2017) also observed daytime surface BrO mixing ratios often ranging from 10–40 pmol mol⁻¹ near Utqiaġvik in spring of 2012. For a subset of this same data from roughly 18–24 March 2012, Peterson et al. (2015) regularly observed daytime BrO mixing ratios of roughly 15 pmol mol⁻¹ with a low LT-VCD (generally less than 2 × 10¹³ molec. cm⁻²) and a f₂₀₀ of roughly 50 %. This closely resembles the statistics of the low-BrO cluster. Peterson et al. (2015) also showed little relationship between surface BrO concentrations and LT-VCDs, which was also evident here where a doubling of the surface mixing ratio between the low-BrO and high-surface-BrO clusters led to only a small increase in the LT-VCD. Also in agreement with Peterson et al. (2015) is the fact that the highest LT-VCDs were observed when BrO was extended to higher altitudes (200–400 m) in the lower troposphere (lofted-BrO cluster).

4.1 Sampling bias considerations

It is important to note that the findings of this study have a clear-sky bias because ALAR could only fly in the absence of clouds to avoid icing, and the porpoising maneuver requires good visibility near the ground. This selection bias is made clear by the results of the aerosol particle inversion from each flight. Throughout the campaign, the mean 361 nm aerosol optical depth (AOD) retrieved from the O₄ inversion was 0.13 (±0.05). On average, the highest particle extinction was found at the Earth's surface (0.3 km${}^{-\mathrm{1}}\pm$ 0.4 km⁻¹) and quickly decreased with altitude. Above 450 m, the average particle extinction was relatively constant at 0.05 km⁻¹ (±0.04 km⁻¹) until roughly 1.5 km altitude.

These observations appear to be typical for this time and location based on data from the Aeronet station in Utqiaġvik. Aeronet observations (Holben et al., 2001) from March and April from 2010 to 2022 showed an average 340 nm AOD of 0.15 with a standard deviation of 0.06, which is very similar to the campaign average retrieved from HAIDI. A similar particle extinction profile shape was observed in 2020 during MOSAIC (Ansmann et al., 2023) with the highest springtime extinction at the surface and decreasing with altitude, albeit for a different Arctic location (> 85^∘ N, northeast of Greenland).

Our clear-sky sampling bias becomes apparent when considering observations from a total-sky imager at the ARM facility in Utqiaġvik which observes cloud fraction throughout the year, often with 30 s resolution (Flynn and Morris, 2015). For daytime observations during March and April from 2010–2022, 56.4 % of observations had less than 20 % cloud cover, and this was slightly higher in 2022 at 62.9 % of observations. However, for flight days during CHACHA, 81.6 % of observations had less than 20 % cloud cover. There were also much fewer overcast days in the CHACHA flight day subset. So although clear-sky conditions were relatively common in Utqiaġvik, they are still overrepresented in this dataset. Aeronet data from 2010–2022 show 86.3 % of days with a 340 nm AOD of less than 0.2, although, as Aeronet is a direct-sun radiometer, these results may also have a clear-sky bias.

Despite the clear-sky bias of this study, several research flights occurred on days preceding and following blowing snow events. So although no HAIDI observations directly probed halogen chemistry under such conditions (Jones et al., 2009), our observations should have some connection to these conditions that can be investigated in the future.

4.2 Profiles indicate snowpack as an important source of reactive bromine

Throughout the campaign, we often retrieved the highest BrO mixing ratios at the Earth's surface. This near-snowpack peak is clear in the median retrieved profile (Fig. 4) and in the clustered profile shapes (Fig. 5), where 94 % of the profiles fit into the first three clusters that have the highest BrO mixing ratios at the snowpack surface. This profile shape underscores the importance of multiphase snowpack surface chemistry for reactive bromine in the springtime Arctic, as well as the atmospheric stability of the area during this time of year. The high mixing ratios at the Earth's surface likely occur due to numerous recycling pathways of different reservoir species that produce Br₂ on the snow surface (Pratt et al., 2013; Custard et al., 2017; Halfacre et al., 2019; Wang et al., 2019). The reactive bromine is then trapped near the ground by atmospheric stability (Bradley et al., 1992; Peterson et al., 2015).

However, BrO mixing ratios were not always highest at the Earth's surface, as shown by the lofted-BrO cluster (Fig. 5). HAIDI observations were largely made under conditions with little cloud cover, no inclement weather, and good visibility near the ground (Sect. 4.1). For one of the few days (19 March) on which we observed a small amount of haze (AOD = 0.21), we retrieved multiple lofted-BrO profiles.

4.3 Lofted-BrO profiles on 19 March 2022

The cluster analysis revealed that 73 % of the lofted-BrO cases occurred on a single day (19 March 2022 shown in Fig. 6). This day was noted as being hazier than a typical flight day (visibly seen in Fig. S6), and our particle extinction inversion supports that (Fig. 7). The 361 nm AOD retrieved for this flight (0.21 ± 0.03) was the second-highest of the campaign and more than a standard deviation higher than the campaign average (0.13, standard deviation of 0.05). The largest particle extinction for this day was retrieved at 225 m above the ground, and the average of the retrieved BrO profiles for this day also peaked at the same altitude with a maximum mixing ratio of 25 pmol mol⁻¹. This vertical correlation could signal that the BrO observed on this day may have advected from the Chukchi Sea to the west and been maintained through multiphase chemistry on airborne particles. Alternatively, this relationship could also be caused by increased mixing from the Earth's surface, lofting particles several hundred meters above the ground.

Figure 6Identified BrO profile clusters for the research flight on 19 March 2022. The flight path is shown in red, beginning in Utqiaġvik and heading east along the coast of the Beaufort Sea before turning to the west around Prudhoe Bay (indicated with a star) and returning back over the tundra. The markers indicate the identified cluster and are plotted at the location of the aircraft profile, with the mean wind speed and direction from ALAR meteorological data denoted with the wind barbs. Lofted-BrO profiles were identified everywhere throughout the flight, except at Utqiaġvik.

Download

Figure 7Aerosol particle and BrO inversion results from the flight on 19 March 2022. The 361 nm particle extinction profile (a) shows the highest extinction at 225 m above the ground, and the mean retrieved BrO profile (b) peaks at the same altitude. Individual BrO profiles from this day (c) indicate some profiles with BrO and/or reactive bromine lofted nearly 1 km above the ground. Error bars for the particle extinction are from the inversion uncertainty; error bars for the BrO profile are the standard deviation of the retrieved BrO profiles from this day.

Download

For a single profile on this day, we also observed over 10 pmol mol⁻¹ of BrO nearly 1 km above the ground associated with higher particle extinction at this same level, which could be due to stratified layers of the atmosphere carrying advected particles from which reactive bromine is activated on those surfaces. However, high mixing ratios of BrO lofted this far above the ground were rarely observed in the retrieved profiles.

Lofted BrO has been observed in several prior studies (Peterson et al., 2017; Simpson et al., 2017; Frieß et al., 2023). Many studies since Frieß et al. (2011) have suggested that bromine recycling could occur on aerosol particles based on an observed correlation between BrO and aerosol particle extinction or particle surface area concentration in the case of Peterson et al. (2017) during March 2012 near Utqiaġvik. However, Peterson et al. (2017) also observed a distinct BrO plume nearly 1 km above the ground clearly distinct from the boundary layer, whereas the lofted BrO observed here was often much closer to the ground (Fig. 7).

4.4 Spatial and surface dependence of BrO profiles

The four BrO clusters allow us to identify spatial changes in BrO profiles, and this is apparent for the flight on 1 April 2022 (Fig. 8). For this flight, ALAR first flew east over the sea ice out of Utqiaġvik, turned south upwind of Prudhoe Bay, and then turned back to the west with two transects downwind of Prudhoe Bay before flying over the tundra while returning to Utqiaġvik. Throughout this flight, the aircraft flew vertical profiles of the lower atmosphere (e.g., Fig. 1). Figure 8 shows the flight path for this day, with the location of the markers indicating the location of the aircraft profiles and with the markers indicating the cluster into which the retrieved profile fit. On this day, there was a clear north–south gradient where observations over sea ice were exclusively high-surface-BrO cases and observations over tundra were often high-total-BrO cases. This finding could be explained by a slightly deeper boundary layer over the tundra with higher wind speeds near the Earth's surface (Fig. 9).

Figure 8Identified BrO profile clusters for a research flight on 1 April 2022. The flight path is shown in red, beginning in Utqiaġvik and heading east over the Beaufort Sea before turning south and returning back west over the tundra. The markers indicate the identified cluster and are plotted at the location of the aircraft profile, with the mean wind speed and direction for each porpoise from ALAR meteorological data denoted with the wind barbs. The star indicates the location of the two largest CO₂ emitters in the Prudhoe Bay oil field (from EPA Greenhouse Gas Reporting Program, https://ghgdata.epa.gov/ghgp, last access: 29 April 2022).

Figure 9(a) The 1 April meteorology from in situ instrumentation on ALAR showing the average potential temperature for the high-surface-BrO cases (blue) over the sea ice and the high-total-BrO cases (red) over the tundra. The average wind speed for these porpoises is shown in panel (b). The lower minimum porpoise altitude over sea ice explains the different minimum altitude of the two cases. Error bars indicate the standard deviations of the parameters.

Download

The exception to this clear spatial gradient on 1 April is the area directly downwind of the Prudhoe Bay region (indicated with a star in Fig. 8). For these three observations, we identified low-BrO cases, likely due to direct interaction with fresh NO_x emissions from the Prudhoe Bay area that were also observed by Custard et al. (2015) and in agreement with the results of Wang and Pratt (2017), who found an impact on BrO with just 100 pmol mol⁻¹ of NO_x.

NO₂ profiles were retrieved using the same method and same spectra as the BrO profiles to identify BrO profiles impacted by NO_x pollution. Throughout the campaign, we identified 12 profiles that appear to be impacted by elevated NO_x, with average retrieved mixing ratios of NO₂ over the lowest 500 m ranging from 0.5 to 1.4 nmol mol⁻¹. As our detection limit for average NO₂ mixing ratios was roughly 0.22 nmol mol⁻¹, these profiles clearly contained more than background levels of NO₂. Of these 12 cases impacted by NO_x pollution, 11 were identified as low-BrO cases (of the total 110 low-BrO cases).

Although each BrO cluster was observed above both sea ice and inland snowpack, it is important to consider the impact of the snowpack surface on the retrieved BrO profiles. Since most profiles peaked toward the Earth's surface, the type of surface under each profile could affect the recycling of reactive bromine and the BrO cluster that the profile fit into. To test this, all profiles were flagged as being retrieved over sea ice or inland snowpack based on the median location of the profile in relationship to the coastline. There were too few high-total-BrO and lofted-BrO cases to be included in this analysis, as their statistics would be skewed by flight location; i.e., all but two profiles were retrieved over tundra on 19 March when most lofted-BrO cases were observed. Therefore, we only consider low-BrO and high-surface-BrO cases.

Between the low-BrO and high-surface-BrO clusters, and after removing the profiles impacted by NO_x pollution, 87 profiles were retrieved over land and 90 were retrieved over sea ice. However, despite this roughly even sampling distribution, only 42 of 99 (42 %) low-BrO clusters were retrieved over sea ice, while 48 of 78 (62 %) high-surface-BrO profiles were retrieved over sea ice (Table 1), meaning that high-surface-BrO observations were overrepresented over sea ice and low-BrO cases were overrepresented over tundra. This spatial difference indicates a significant (p=0.015 based on Fisher's exact test) location difference between these two clusters that could be driven by bromide enhancement in snow above what was likely first-year sea ice just off the coast (Peterson et al., 2019).

Table 1Distribution of BrO clusters as a function of location. The surface type was determined based on the median location of each porpoise in relation to the coastline.

Download Print Version | Download XLSX

4.5 Meteorological conditions for BrO clusters

Considering the clear differences between the four BrO profile cluster shapes in Fig. 5, it is important to consider what factors, apart from surface type, influenced these profiles. To study this, we compiled data from the other instruments on ALAR and binned them in 25 m intervals for the same porpoises used for the clustered BrO profiles. The average profiles for the normalized potential temperature (the difference of the potential temperature profile and the profile's potential temperature at 500 m) and wind speed are shown in Fig. 10 for the four BrO profile clusters. The offset potential temperature is shown to isolate the stability of the lower atmosphere. These plots begin at 100 m altitude, as not all porpoises descended to the same altitude.

Figure 10The average meteorological conditions associated with the four BrO profile clusters. The normalized potential temperature (middle, offset by the potential temperature at 500 m) shows decreasing stability associated with increasing BrO columns. The variability of these parameters is not shown here for clarity, but is indicated with error bars in the Supplement (Figs. S6 and S7).

Download

The stability of the atmosphere clearly impacted the BrO profile shape, as the two most stable profiles with the largest increases in potential temperature with altitude corresponded to the low-BrO and high-surface-BrO clusters, where there was a large difference between the BrO mixing ratio at the Earth's surface and mixing ratios aloft. The high-total-BrO case corresponded to a more neutral atmosphere. Similarly, the lofted-BrO cases are associated with a relatively neutral atmosphere from roughly 100–300 m altitude. Because the temperature profiles did not extend to the ground, it is difficult to determine if this relates to turbulence extending from the surface or a residual layer above a stable surface layer. These lofted-BrO cases are also associated with lower wind speeds throughout the lower atmosphere, which could help maintain stratification of lofted aerosol layers. Back-trajectory analyses and nearby balloon-sounding data could be used to investigate these lofted profiles more closely, though this is beyond the scope of this study.

4.6 Associated ozone chemistry

It is also important to consider the implications of the four BrO profile clusters for ozone chemistry by studying the average ozone profiles associated with each cluster (Fig. 11). These profiles are calculated in the same manner as the meteorological parameters shown in the previous section. Also shown is the average particle extinction for each flight associated with the four clusters.

Figure 11The ozone mixing ratio (middle) and 361 nm particle extinction (right) profiles associated with the four BrO profile clusters. The variability of these parameters is not shown here for clarity but is indicated with error bars in the Supplement (Figs. S8 and S9).

Download

This figure reveals that there was little difference in the ozone profile above the Earth's surface for low-BrO cases compared to high-surface-BrO cases, likely because these two BrO mixing ratio clusters are only significantly different at the surface, which was not often measured in situ during porpoises. We also see that the lowest ozone mixing ratios are associated with the high-total-BrO cluster, indicating that these profiles were sampled during instances of active bromine chemistry. These three clusters correspond to very similar particle extinction profiles, which suggest few lofted particles on average during these cases and the surface-based nature of the reactive bromine chemistry and ozone depletion for the majority of our observations. The lofted-BrO cluster is associated with lower ozone mixing ratios between 200 and 500 m, with a profile that reflects the increased particle extinction at higher altitudes, indicating that these may be cases in which reactive bromine was activated by multiphase chemistry on the surface of the particles, impacting ozone mixing ratios aloft.

4.7 Satellite implications of surface-enhanced BrO

The frequency with which surface-peaking profiles were observed has significant implications for satellite observations for two reasons. The first is that UV-Vis satellite observations, even over bright surfaces, have lower sensitivity to trace gases at the surface than aloft (Eskes and Boersma, 2003). More importantly, the BrO mixing ratio at the Earth's surface appears to have little impact on the vertical column density. Figure 5 shows that the LT-VCD difference between the low-BrO and high-surface-BrO clusters is very small (roughly 20 %), even though their surface mixing ratios differ by a factor of 2. This similarity in BrO columns suggests that many high-surface-BrO conditions are likely not evident in satellite column BrO observations, which could explain some of the remaining difficulty in calculating tropospheric BrO from such measurements (Wales et al., 2023). Wales et al. (2023) also note an underestimation of surface ozone depletion at coastal sites that could be explained by these BrO profiles. These findings point out that ODEs and reactive bromine chemistry under relatively clear-sky conditions in the Arctic constitute a snowpack surface phenomenon and motivate the need for frequent MAX-DOAS and chemical ionization mass spectrometer (CIMS) observations in the stable boundary layer in the Arctic to identify these high-surface-BrO cases.

4.8 Porpoising with AMAX-DOAS

The porpoising flight pattern used during CHACHA resulted in considerably higher degrees of freedom (DOF) of the radiative transfer inversion than ground-based MAX-DOAS instruments can achieve. The improved vertical resolution is due to the combination of observations at different altitudes, so the increased DOF is to be expected. Ground-based MAX-DOAS instruments are particularly beneficial for providing higher temporal coverage than aircraft measurements, are less impacted by adverse weather, cost considerably less to operate, and provide more vertical information than ground-based and/or in situ observations. The combination of aircraft and ground-based MAX-DOAS observations can thus achieve both high vertical resolution and good temporal coverage. The profiles found here can be used as a priori profiles for ground-based observations to improve BrO retrievals.

Peterson et al. (2015) and Simpson et al. (2017) retrieved BrO profiles with an average of 2–3 DOF, and these works therefore reported two vertically defined parameters to represent the information they could retrieve. It is typical for ground-based MAX-DOAS instruments to retrieve 1–3 DOF for various trace gases (Hendrick et al., 2014; Vlemmix et al., 2015; Ortega et al., 2015; Ryan et al., 2023). One way to increase DOF is by raising the MAX-DOAS instrument above the surface, which could result in upwards of 5 DOF (Koenig et al., 2017). AMAX-DOAS observations are able to retrieve considerably higher DOF, as they can sample a larger portion of the atmosphere. For instance, the work of Baidar et al. (2013) reported an average DOF of 12 for their trace gas retrievals, with lower sensitivity to the Earth's surface than found in this work due to the challenges of flying at low altitude over populated areas. Similarly, Prados-Roman et al. (2011) noted 10 DOF for BrO profiles of the entire troposphere. In those two AMAX-DOAS studies, averaging kernels were near unity for the range of altitudes flown by the aircraft, indicating that the amount of information retrieved by the inversions could be limited by the grid spacing used. Peterson et al. (2017) utilized a finer grid spacing of 100 m and were able to retrieve up to 12.4 DOF for a profile with a maximum altitude of 1039 m (shown in the Supplement of that work). Our study was able to retrieve higher DOF than other AMAX-DOAS studies due to the finer vertical resolution of the modeled atmosphere, at the cost of computational time, resulting in finer retrieved detail through the lower atmosphere. This assertion is supported by Volkamer et al. (2015), who retrieved upwards of 20 DOF for trace gas profile retrievals. This was accomplished by profiling most of the troposphere and modeling with 500 m grid spacing. The averaging kernels in our work (Fig. 3) peak at values slightly lower than 1, indicating that the 50 m grid spacing used here is roughly the highest-resolution vertical spacing that we could use.

In all four AMAX-DOAS setups (Prados-Roman et al., 2011; Baidar et al., 2013; General et al., 2014; Volkamer et al., 2015), telescopes were mounted to research aircraft with some external instrumentation. This technique could be used with a less labor-intensive setup. For example, a telescope could be mounted on a window of a non-pressurized light aircraft. Similarly, drones could be used to profile the lower atmosphere with a small limb-viewing spectrometer–telescope combination. Considering the greatly improved vertical resolution of combining observations at different altitudes and the increased DOF retrieved with finer model grid spacing, this method could be expanded to quantify additional trace gases in the lower atmosphere (e.g., HCHO, CHOCHO, HONO, NO₂).