Introduction
Measurements of tropospheric halogen species are an area of increasing
research interest due to the ability of halogens to destroy tropospheric
ozone (O3) (Tossell, 2003; Holmes et al., 2009; Hynes et al., 2009), oxidize
atmospheric mercury (Holmes et al., 2009; Hynes et al., 2009), and modify
oxidative capacity (Parella et al., 2012). Most assessments of the impacts
of halogen chemistry are based on measurements of halogen oxides (bromine
monoxide, BrO, and iodine monoxide, IO), since these radicals are typically
found at higher concentrations throughout the troposphere than the
corresponding halogen atom radicals. Many of these studies take place in the
planetary boundary layer (PBL), given that this region of the atmosphere is
easily accessible from measurements located at the surface, and is also the
most directly impacted by anthropogenic activities. However, halogen
chemistry in the free troposphere (FT), albeit more challenging to measure,
has the potential to affect an even larger air volume and mass. In
particular, the colder temperatures of the free troposphere accelerate the
bromine oxidation of gaseous elemental mercury (GEM) (Donohoue et al.,
2006). Satellite-borne measurements represent a powerful resource for
assessing global distributions and tropospheric vertical column densities (VCDs)
of BrO (GOME – Van Roozendael et al., 2002; GOME-2 – Theys et al.,
2011; Sihler et al., 2012). However, satellite retrievals rely on
assumptions made about the vertical distribution of BrO, and uncertainties
in these assumptions can lead to over- or underpredictions in the derived
tropospheric VCD. The most direct method for measuring trace gas vertical
distributions is through the use of aircraft (Prados Ramon et al., 2011;
Volkamer et al., 2015) or balloons (Fitzenberger et al., 2000; Pundt et
al., 2002; Dorf et al., 2006). However, this type of measurement is costly
and potentially impractical if the goal is to establish long term trends in
the FT. Ground-based measurements are typically more straightforward to
deploy and maintain for extended periods of time, but optimizing
ground-based capabilities to observe the FT remains an area of active
research (Schofield et al., 2006; Theys et al., 2007; Hendrick et al., 2007;
Coburn et al., 2011). Specifically, ground-based multi-axis differential
optical absorption spectroscopy (MAX-DOAS) measurements are uniquely suited
for this type of study since this technique also assesses vertical
distributions, and derived VCDs can be directly compared with models and
satellites. Additionally, the DOAS retrieval allows for the detection of not
only BrO, but also other trace gases that have significant impacts on the
chemical cycling of bromine species in the atmosphere, such as NO2 and
some volatile organic compounds (VOCs). However, measurements of FT BrO from
ground-based MAX-DOAS are not straightforward for several reasons:
(1) stratospheric BrO represents a large portion of the measured signal and
creates a background that has to be accounted for when attempting to assess
the FT, (2) ozone absorption structures are strongly present in the same
wavelength region as BrO and can create interferences due to stratospheric
ozone absorption, in particular at high solar zenith angles (SZAs) (Aliwell
et al., 2002; Van Roozendael et al., 2002); and (3) the sensitivity of this
technique peaks at the instrument altitude and decreases with increasing
altitude. Recent advances with testing stratospheric BrO profiles in
atmospheric models (Liang et al., 2014) provide opportunities to properly
account for point 1 by assimilating information from atmospheric models.
Furthermore, retrievals that avoid SZAs larger than 70∘ do not suffer from 2,
and certain measurement geometries retain information about the FT. Figure S1
(Supplement) depicts the box air mass factors (bAMFs), which represent
the sensitivity of the slant column density (SCD) measurement geometry to
BrO concentrations at different altitudes, for two pointing directions (or
elevation angles, EA = 25 and 90∘ upwards) at several
SZA; at SZA < 70∘ the sensitivity of these EAs peaks
between 2 and 15 km. A more comprehensive view of the bAMFs for different EAs
over a wider SZA range is shown in Fig. S2.
Van Roozendael et al. (2002) compared ground-based and balloon-borne
measurements to VCDs of BrO from the space-borne Global Ozone Monitoring
Experiment (GOME) and found all platforms were consistent with a rather
widespread tropospheric BrO VCD of 1–3 × 1013 molec cm-2 once
appropriate radiative transfer effects were taken into consideration.
Salawitch et al. (2005) and Theys et al. (2011) also report satellite
derived tropospheric BrO VCDs (GOME and GOME-2, respectively) for the
midlatitudes of 2 × 1013 and 1–3 × 1013 molec cm-2,
respectively. Ground-based measurements (Theys et al., 2007;
Coburn et al., 2011) in the midlatitudes have reported BrO VCDs of
1–2 × 1013 molec cm-2 that are comparable to the findings from
satellites. Volkamer et al. (2015) recently reported 1.6 × 1013 molec cm-2
BrO VCD in the tropics measured by limb observations from
aircraft. All of these studies point to the widespread presence of BrO in
the FT, corresponding to a VCD of 1–3 × 1013 molec cm-2. Based on
these reports, tropospheric BrO could account for 20–30 % of a total BrO
VCD ∼ 5–6 × 1013 molec cm-2 as seen from satellite
(Van Roozendael et al., 2002; Theys et al., 2011), and significantly impacts
the lifetime of tropospheric O3 and atmospheric GEM (Wang et al., 2015).
Total mercury wet deposition in the US for 2013. The highest
levels of Hg deposition are observed in the southeastern US, where no local
and regional sources are located immediately upwind. The black square
indicates the measurement location.
Atmospheric Hg in the southeastern US
Mercury in the atmosphere exists in three forms: gaseous elemental mercury
(Hg0, GEM), gaseous oxidized mercury in the form of either Hg2+ or
Hg1+ (GOM), and particle-bound mercury (PBM). Understanding the
processes that cycle mercury between its various forms
(GEM ↔ GOM ↔ PBM) is of great importance, because this speciation
controls the deposition of mercury to the environment – i.e., GOM and PBM are
more readily removed from the atmosphere via wet and dry deposition than GEM
(Lindberg and Stratton, 1998; Bullock, 2000). Once deposited, biological
processes can methylate Hg2+ to form the neurotoxin methyl mercury,
which bio-accumulates in fish. Enhancement factors for methyl mercury of up
to 106 relative to water have been measured in predatory fish tissues
(Schroeder and Munthe, 1998; Selin et al., 2010).
A better understanding of the processes controlling atmospheric mercury
oxidation, and therewith removal, is particularly relevant for regions that
experience high levels of mercury deposition, such as the southeastern
United States (SE US). Figure 1 shows a map of the total mercury wet
deposition in the US from 2013 (http://nadp.sws.uiuc.edu/maplib/pdf/mdn/hg_dep_2013.pdf). The high deposition levels experienced in the
SE US cannot be explained by regional anthropogenic sources of mercury
alone, which are mainly located within the Ohio River valley, where the
prevailing winds carry emissions northeast. Additionally, several modeling
studies (Zhang et al., 2012; Nair et al., 2013) have shown that
meteorological patterns above the SE US greatly influence the wet
deposition of mercury and that these processes are linked with deep
convective activity. This indicates that a regional emission–deposition
pattern is most likely not the major source–receptor relationship for
mercury entering the environment over Florida, in the SE US, meaning that
other possibilities, such as enhanced atmospheric oxidation followed by
deposition, need to be explored.
Overview of the MAX-DOAS measurements for the week surrounding the
case study day (9 April, highlighted with blue box). Also included are
O3 measurements from a monitor collocated with the MAX-DOAS instrument
(label “CU”, black trace) and a monitor located at a Mercury Deposition
Network (MDN) site ∼ 30 km northwest of the EPA site (label
“OLF”, red trace), and wind direction measurements (grey trace) from a
site near the EPA facility along with HgII measurements from the MDN
site (scaled by a factor of 10, orange trace).
Experimental section
Atmospheric conditions
The case study during 9 April 2010 provided optimal conditions for assessing
the ability of a ground-based MAX-DOAS instrument to measure FT trace gases
(see Sect. S1 in the Supplement for a brief overview of instrumentation and
measurement site). Figure 2 shows a time series of trace gas differential
slant column densities (dSCDs) of BrO, IO, NO2, and O4 for the week
surrounding the case study day, with 9 April highlighted by the blue box.
The IO measurements are assimilated and used in the modeling portion of this
study (Sect. 3.4), while the NO2 measurements give an indication on the
amount of influence from anthropogenic activities in the lowest layers of
the BL. This day provides an excellent case study for two reasons:
(1) consistent shape of the O4 dSCDs across elevation angles, as well as the
clear splitting between the values, is a good indicator for a cloud-free day
and (2) the relatively high O4 dSCD values (compared with other days)
indicate a low aerosol load, enabling the instrument to realize longer
light paths (increased sensitivity due to fewer extinction events), and an
unobstructed view of the FT. An inspection of webcam pictures for the
instrument site proved the day to be free of visual clouds, and a precursory
look at the aerosol load confirmed the low values. Figure 2 also contains
in situ O3 measurements (from both the US EPA site and a nearby Mercury
Deposition Network (MDN) site), oxidized mercury measurements (see below),
as well as wind direction measurements from a WeatherFlow, Inc. monitoring
station located in Gulf Breeze, FL near the US EPA site.
Oxidized mercury measurements (HgII = GOM + PBM) at the Pensacola MDN
site during study period (see Edgerton et al., 2006, for a detailed
description of the site and instrumentation) are also shown in the bottom
panel of Fig. 2 (also Fig. 9, Sect. 3.4). On 9 April, HgII
concentrations at the MDN site were rising from near zero on 8 April (due to
rain) to peak values of 15–40 pg m-3 on the following days, which is
above average for the season. In prior years, average daily peak
concentrations at this site in spring were 15 pg m-3, which is higher
than during any other season (Weiss-Penzias et al., 2011; Nair et al.,
2012). Observed GEM concentrations are persistently around 1.4 ng m-3
throughout early April, as expected for this season, and therefore not
shown. It should be noted that recent studies have found that measurements
of HgII using KCl denuders can be influenced by atmospheric conditions
and other trace gases (McClure et al., 2014) and these external factors can
lead to an underestimation of HgII.
External model overview
The Whole Atmosphere Community Climate Model version 4 (WACCM4) (Garcia
et al., 2007; Marsh et al., 2013) has been extensively evaluated for its
representation of the stratosphere, including stratospheric BrO (Eyring et al., 2010). The model does not represent tropospheric bromine sources
from very short lived species (VSLSs, bromocarbons) and using this model in
this study is an active choice to ensure that a priori information about
tropospheric BrO represents a lower limit (see Sect. 2.4.1). However,
CHBr3 and CH2Br2 concentrations are fixed at the cold point
and add about 5–6 pptv stratospheric Bry (stratospheric Bry
loading is 21–22 pptv). In this work, WACCM is run with specified (external)
meteorological fields. This is achieved by relaxing the horizontal winds and
temperatures to reanalysis fields. The reanalysis fields used are taken from
the NASA Global Modeling and Assimilation Office (GMAO) Modern-Era
Retrospective Analysis for Research and Applications (MERRA) (Rienecker et
al., 2011). The horizontal resolution is 1.9∘ × 2.5∘
(latitude × longitude), with 88 vertical levels from the surface to the
lower thermosphere (-140 km). Other WACCM model outputs used are HCHO,
temperature, and pressure vertical profiles. Time-synchronized BrO vertical
profiles from WACCM are used as a priori inputs to the inversion
(tropospheric VCDs ∼ 1 × 1013 molec cm-2), while HCHO
profiles are used as input for the box model utilized in this study (see
Sect. S4). Temperature and pressure profiles were used to
construct a molecular density profile in order to convert from concentration
(output units of the inversion) to volume mixing ratios (VMRs). Model data
were generated specifically for this case study in order to best represent
that atmospheric composition at the time of the measurements.
The GEOS-Chem global chemical transport model (CTM) is used to provide a regional
and seasonal context for the DOAS observations and their relevance to mercury chemistry. GEOS-Chem (v9-02, http:www.geos-chem.org) is driven by
assimilated meteorology from the NASA Goddard Earth Observing System (GEOS-5).
Simulations here have 2∘ × 2.5∘
horizontal resolution at 47 vertical layers for bromine and all species
except mercury, which have 4∘ × 5∘ horizontal
resolution. The bromine chemical mechanism, described by Parrella et al. (2012),
includes marine bromocarbon emissions from Liang et al. (2010) and
debromination of sea salt aerosols (Sander et al., 2003; Yang et al., 2005).
Tropospheric bromine concentrations from GEOS-Chem are merged with
stratospheric bromine from GEOS-CCM (Liang et al., 2010) to produce a
complete atmospheric column (tropospheric VCD of 1.5 × 1013 molec cm-2).
The GEOS-Chem BrO vertical profile is also used as input to the
box model in order to assess sensitivity to differences between measured and
modeled BrO vertical distributions on mercury oxidation; GEOS-Chem O3
and NO2 profiles are also used as input to the box model. Additionally,
the GEOS-CCM profile is utilized as an a priori during the inversion of the
MAX-DOAS measurements due to its representation of the stratosphere while
excluding most bromine chemistry in the troposphere (lower limit of bromine
in the FT, tropospheric VCD of 3.5 × 1012 molec cm-2). The mercury
simulation is configured as described by Zhang et al. (2012), which includes
GEM and two HgII species: GOM and PBM. Anthropogenic mercury emissions
are from the US EPA National Emission Inventory (NEI) 2005 and National
Pollutant Release Inventory (NPRI) 2005 inventories over North America
(Zhang et al., 2012), which are adjusted to account for HgII reduction
in power plant plumes (Landis et al., 2014), and elsewhere from the Global
Emission Inventory Activity (GEIA) (Streets et al., 2009; Corbitt et al.,
2011). Emissions and cycling from oceans and the biosphere are also included
(Selin et al., 2008; Soerensen et al., 2010). Atmospheric GEM is oxidized by
bromine (Holmes et al., 2010) using updated kinetic rate coefficients from
Goodsite et al. (2012) and bromine concentrations from GEOS-Chem. HgII
partitions between GOM and PBM (Rutter and Schauer, 2007; Amos et al.,
2012), which are both quickly scavenged by precipitation or dry deposited
but also susceptible to reduction in cloud water. The bromine and mercury
simulations are both spun up for at least 1 year. The model is then
sampled hourly during April 2010 over Pensacola.
DOAS BrO retrieval
Following is a brief description of the parameters and settings used for the
DOAS analysis of BrO for this study. A series of sensitivity studies on the
BrO retrieval determined the optimal wavelength window for the analysis in
this study to be 338–359 nm and include a fifth order polynomial.
Sensitivity studies, which varied the fit window, polynomial order, and
intensity offset, indicate that the chosen fit settings estimate BrO dSCDs
conservatively, as is discussed in more detail in Sect. S2.
The reference cross sections used in the DOAS retrieval (WinDOAS software;
Fayt and van Roozendael, 2001) were O3 (at 223 and 243 K; Bogumil
et al., 2003), NO2 (at 220 and 297 K; Vandaele et al., 1998),
O4 (at 293 K; Thalman and Volkamer, 2013), HCHO (Meller and
Moortgat, 2000), BrO (Wilmouth et al., 1999), and a Ring spectrum (Chance
and Spurr, 1997) calculated using the software DOASIS (Kraus, 2006) at 250 K
for the reference used in the analysis. Additionally, a constant intensity
offset was included in the fit, but limited to a range (±3 × 10-3)
determined by an estimation of the upper limit for the correction of
spectrometer stray light. Details of the retrieval for other trace gases
can be found in Table 1 along with an overview of the BrO retrieval settings
listed here. A single zenith reference from a low SZA of the case study day
was used to analyze all spectra (referred to as a fixed reference analysis).
This spectrum was taken at ∼ 18:01 UTC on 9 April 2010 (∼ 23∘ SZA).
Overview of DOAS settings.
Cross section,
Reference
BrO window
IO window
NO2 window
O4 window
parameter
(338–359 nm)
(415–438 nm)
(434–460 nm)
(437–486 nm)
Polynomial order
5
5
3
5
O3 T = 223 K
Bogumil et al. (2003)
X
X
X
X
O3 T = 243 K
Bogumil et al. (2003)
X
X
X
X
NO2 T = 220 K
Vandaele et al. (1998)
X
X
X
X
NO2 T = 297 K
Vandaele et al. (1998)
X
X
X
X
O4 T = 293 K
Thalman and Volkamer (2013)
X
X
X
BrO
Wilmouth et al. (1999)
X
HCHO
Meller and Moortgat (2000)
X
IO
Spietz et al. (2005)
X
CHOCHO
Volkamer et al. (2005)
X
H2O
Rothman et al. (2005)
X
X
Inversion method
The retrieval of the BrO VCDs and vertical profiles for this study is a
three-step process: (1) aerosol profiles are determined based on DOAS
measurements of O4 (see Sect. S3), (2) derived aerosol
profiles are used in a radiative transfer model (RTM) to calculate weighting
functions for BrO; and (3) weighing functions are used in an optimal
estimation inversion of the DOAS dSCDs to derive VCDs and vertical profiles
(Sect. 2.4.1). Additionally, a method is presented here for determining the
residual amount of BrO contained in the reference spectrum (SCDRef,
Sect. 2.4.2). The relationship between SCDRef and the measured dSCDs is
shown as Eq. (1).
SCD=dSCD+SCDRef
Both dSCDs and SCDs were used as input to the inversion, and sensitivity
tests are presented to assess the impact of the SCDRef value on the
derived vertical profiles and VCDs.
Trace gas inversion
Weighting functions calculated at 350 nm (for BrO) are used in an optimal
estimation (OE) inversion (Rodgers, 2000) to determine the trace gas vertical
profiles from the MAX-DOAS dSCD measurements, as given by Eq. (2).
xr=xa+Axt-xa
Here xr represents the retrieved profile, xa is the a priori profile
assumption, A is the averaging kernel matrix (AVK), and xt is the true
atmospheric state (represented by the MAX-DOAS dSCD or SCD measurements
here). The AVK gives an indication of where the information in xr is
coming from, i.e., information from xt (measurements) vs. xa (a priori
assumption). The trace of this matrix gives the degrees of freedom (DoFs) of
the retrieval and indicates how many independent pieces of information are
contained in the retrieval (see Sect. 3.1).
One important aspect in this study is the choice of the altitude grid used
for both the radiative transfer calculations and the inversion. We used a
grid of varying layer thickness: 0.5 km from 0 to 2 km, a single 3 km layer
from 2 to 5 km, and 5 km layer thickness from 5 to 50 km. This grid is chosen to
effectively combine the information from multiple altitudes into a single
grid point to ensure reasonable peaks in the averaging kernels at increasing
altitude, where the MAX-DOAS measurements have limited vertical resolution.
For the BrO inversion, four different a priori profiles are used in order to
assess the robustness of the inversion (see Fig. 5, Sect. 3.2, which shows
three of the a priori profiles along with their corresponding a posteriori
results). Two of the a priori profiles are based on WACCM model output, the
third used data from the CU Airborne MAX-DOAS (AMAX-DOAS) instrument
collected during the Tropical Ocean tRoposphere Exchange of Reactive halogen
species and Oxygenated VOC (TORERO) 2012 field campaign (Volkamer et al.,
2015), and the fourth is a mean profile from GEOS-CCM. The a priori profiles
from WACCM are (1) direct model output for the time of each MAX-DOAS scan
throughout the day (i.e., different a priori profile for each MAX-DOAS
scan) and (2) the profiles from (1) increased by 40 % (in VMR space). The
a priori profile for the TORERO case is derived from data collected during
research flight 12 (RF12; Volkamer et al., 2015), which also closely represents
the tropical average BrO profile (see Fig. 2 in Wang et al., 2015). The
GEOS-CCM profile is a daytime zonal mean at 30∘ N. The TORERO and
GEOS-CCM profiles are used to invert each MAX-DOAS scan, i.e., the same
profile is used for each scan inversion. This is in contrast to the WACCM
a priori profiles, which change throughout the day, due to the availability
of time-resolved output. For reference, diurnal variations in the WACCM
model output for BrO vertical distributions are shown in Fig. S3a, while Fig. S3b shows the corresponding tropospheric
and total VCDs from these profiles. The a priori profile error used as input
to the OE is constructed based on upper limit values (in units of VMR)
expected throughout the troposphere; this is set at 1 pptv for altitudes
0–2 km, 3 pptv for 2–5 km, and 4 pptv for 5–20 km (except for the GEOS-CCM case
where this altitude range is set to 6 pptv to account for the extremely low
values found in the lower FT of the a priori). The a priori error is
constrained to 40 % for the stratospheric portion of the profile (20–50 km),
based on the assumption that the WACCM and GEOS-CCM profiles in this
region of the atmosphere accurately represent the true atmospheric state,
but allowing for sufficient room to reflect contributions of very
short-lived organo halogen species (VSLS) to stratospheric BrO (Liang et al., 2014).
A similar method is also used to derive IO profiles (used as input to the
modeling in Sect. 3.4). The IO inversion uses two a priori profiles: (1) a
moderate IO VMR in the BL (0.2–0.5 pptv) and decreasing to very low levels
(< 0.1 pptv) throughout the rest of the troposphere and (2) a profile
from recently reported AMAX-DOAS measurements (Dix et al., 2013; Volkamer et
al., 2015). A summary of the a priori profiles, example a posteriori
profiles, the average (for the case study) profile, and the diurnal
variation in the VCD for IO is found in Fig. S4.
Conceptual sketch of the SCDRef retrieval (a) and
resulting SCDRef values (black trace, b). The sensitivity of the
a posteriori BrO VCD (red trace) to SCDRef is also shown (b).
Determination of SCDRef
The BrO profile at the recording time of the fixed reference spectrum is
included in the analysis by estimating the reference SCD (SCDRef) for
BrO, which is then added to the dSCDs at off-axis EAs according to Eq. (1).
Initial sensitivity studies revealed that some MAX-DOAS scans from the case
study day are better suited for producing consistent values for SCDRef
than others. The final choice of the reference spectrum is informed from
comparing the SCDRef, determined using the iterative method presented
here, to the SCDRef predicted by the WACCM a priori profile (see
Sect. 2.4.1, profile 2), which assumes a median BrO abundance. The ratio of SCDRef
divided by the BrO VCD corresponds to the AMF of the reference spectrum,
which in the case of our case study was ∼ 1.2. The following
criteria are applied to select a reference spectrum: (1) the SCDRef
needs to be consistent with the AMF for the a posteriori BrO profile, as
well as (2) be within ±1 × 1013 molec cm-2 of the WACCM
profile increased by 40 % (see Sect. 2.4.1). Figure S5 shows
the results from this approach for multiple zenith spectra (potential
references) on the case study day, and thus illustrates the need for active
measures, such as the above, to build internal consistency between the BrO
SCDRef, BrO VCD, AMF, and forward RTM calculations of the a priori and
a posteriori BrO profiles. The results presented in this paper are produced
from a single reference spectrum (see Sect. 2.3); however, other references
that pass the above quality assurance criteria would not alter our
conclusions. Selecting another reference spectrum that meets the
aforementioned criteria only increases the variability in the derived VCD by
less than 9 × 1012 molec cm-2 (for any given single scan), and
changing the a priori assumption has also only a small effect on the VCDs
(±1.5 × 1012 molec cm-2 from average value).
Iterative approach
The value of SCDRef is determined here by
iteratively running the optimal estimation inversion on the MAX-DOAS scan
containing the reference spectrum; this process is shown as a flow chart in
Fig. 3a. The initial value for SCDRef is determined from the
a priori profile, and subsequent SCDRef values are determined from the
a posteriori profile of the inversion returned from the previous iteration.
For each iteration only SCDRef changes; the a priori profile remains
constant. After multiple iterations this method converges on a SCDRef
that minimizes the differences between the measured and forward calculated
scan SCD inputs. This value is then used as input for the inversion of the
rest of the MAX-DOAS scans throughout the day. Figure 3b depicts
the iterative refinement of the SCDRef and corresponding VCD for the
reference chosen as most ideal for this case study.
Characterizing the retrieval averaging kernel (AVK) and degrees of
freedom (DoF). (a) AVK output for an inversion of BrO SCDs (at
SZA = 25∘) as a function of altitude. The colored traces represent the
individual altitude grids of the inversion accounting for SCDRef; the
thick black trace is the diagonal of the AVK matrix; the thick red trace is
the diagonal of the AVK matrix from the inversion not accounting for
SCDRef. Panel (b) shows the difference between the two AVK diagonals as
a function of altitude.
BrO profiles for a single MAX-DOAS measurement scan at SZA
∼ 23∘ around solar noon. Panel (a) contains three of
the a priori profiles tested (dashed lines) and the respective a posteriori
profiles (solid lines); colors correspond to use of WACCM (black), GEOS-CCM
(orange), and TORERO RF12 (blue) as a priori information. Panel (b) contains
the AVK from the inversion using the WACCM profile as the a priori, and
gives an indication of the amount of information coming from the
measurements as a function of altitude.
Results and discussion
Characterizing the BrO profile retrieval
Figure 4 compares the AVK resulting from the OE inversion of BrO SCDs
constructed by accounting for SCDRef (addition of this value to the
measured dSCD output from the DOAS analysis prior to inversion) vs. not
accounting for SCDRef (where SCD = dSCD and with appropriate
modification of the weighting functions to reflect this treatment). The
largest differences are seen at low SZA in the FT. Figure 4a shows the AVK
matrix, which gives an indication as to how much information can be derived
from the measurements at specific altitudes, as a function of altitude for
the inversion (at SZA = 25∘) accounting for SCDRef
(different colored thin lines) along with the diagonal of this matrix (thick
black line) and the diagonal for the inversion not accounting for
SCDRef (thick red line). Figure 4b shows the difference
between the two diagonals as a function of altitude. Also listed in Fig. 4
are DoFs for the different retrieval methods, which, when SCDRef is
properly accounted for, demonstrate an increase of 0.5 (mostly coming from
the FT). For these reasons, we chose to include SCDRef in the inversion
of the measured BrO dSCDs in order to further maximize the sensitivity of
these measurements towards the FT. Additional information on the DoFs and
inversion RMS for these retrievals throughout the case study day is found
in Fig. S6. It should be noted that the presence of even
moderate aerosol loads would significantly impact the ability to derive
vertical profiles extending into the FT but column-integrated information
would remain intact.
Recently, Volkamer et al. (2015) reported significant sensitivity of ship
MAX-DOAS profiles to the magnitude of SCDRef over oceans. For several
case studies independent aircraft measurements above the ship show that a
significant trace gas partial column resides above the BL (IO and glyoxal).
They find that while the BL VMR is insensitive to the value of SCDRef,
the partial tropospheric VCD can be impacted by up to a factor of 2–3.
During this study the impact of using BrO dSCDs, rather than SCDs, in the
inversion leads to an average percent difference in the BrO VCD derived for
each MAX-DOAS scan of ∼ 30 %. The average VCDs were
2.1 × 1013 and 2.3 × 1013 molec cm-2 (integrated from 0 to 15 km)
when using dSCDs or SCDs (only ∼ 10 % difference). This
reflects that the inversion based on dSCDs (not accounting for SCDRef)
produced highly variable FT VCDs throughout the case study day, and that
this variability is reduced for the SCD-based inversion.
The reduced variability in the BrO VCD compared to the other gases
investigated by Volkamer et al. (2015) is probably due to the fact that no
BrO was detected in the BL in either case study. For glyoxal and IO about
50 % of the VCD resides near the instrument altitude (Volkamer et al.,
2015). This BL contribution also adds offsets to the SCDs for the higher EAs
that “partially obstruct” the view of the FT, and makes FT partial columns
subject to larger error bars. The lack of BrO in the BL seems to simplify
measurements of FT partial columns from the ground. The Volkamer et al. (2015) case
studies in combination with our results thus suggest that MAX-DOAS
instruments, which measure FT partial columns of gases other than BrO,
should actively avoid complications from absorbers near instrument altitude,
for example by placing the instrument on mountaintops. The lack of aerosols
from high mountaintops has the additional benefit of increased sensitivity
to measure profiles for BrO and other gases.
BrO VCDs and vertical profiles
Figure 5 shows the results from the inversions using the three different
a priori profiles. Figure 5a shows the a priori profiles, and the
a posteriori vertical profiles corresponding to one scan at
∼ 23∘ SZA (around solar noon), as well as the median profile from
all profiles. A total of 45 profiles, corresponding to SZA < 70∘
for each a priori case, were combined to create the median BrO
vertical profile for the case study day (Fig. 5). The median profile was
used for the modeling results and discussion in Sect. 3.4. Figure 5b shows
the AVK (see Sect. 2.4.1) from the inversion using the un-modified WACCM BrO
output as the a priori profile. In an ideal scenario, the AVK for each layer
peaks near unity within that layer. The derived vertical profiles show only
slight dependence on the a priori profile with a maximum average difference
between a posteriori profiles of ∼ 0.6 ppt at altitudes < 15 km
for the entire day. These profile differences have only a
small effect on the free-tropospheric VCDs, i.e., < 20 % difference
for SZA < 70∘. The AVK peaks twice – once in the lowest
layer (from the lowest looking elevation angles) and again between 5 and
20 km – which reflects the combination of the optimization of the radiative
transfer grid and the measurement sensitivity. This is interpreted as an
indication that the DoFs remaining after accounting for the peak at
instrument altitude correspond to the FT. Figure S7 contains a
comparison of the a posteriori profile derived BrO dSCDs and the measured
dSCDs (Fig. S7a and b) and the corresponding RMS difference between measured
and calculated dSCDs in Fig. S7c.
This inversion procedure allows for the determination of the diurnal
variation in the BL (0–1 km), FT (1–15 km), and total VCDs for BrO. These
results, corresponding to the inversion utilizing WACCM output as the
a priori profile, are found in Fig. 6b along with the corresponding
DoFs (Fig. 6a). The errors bars on the tropospheric and stratospheric VCDs
reflect the range of derived values resulting from the use of different
a priori assumptions. The retrieved diurnal variation in the free-tropospheric (1–15 km) BrO VCD follows that of the total VCD with an average
of 53 % contribution to the total and varying between 47 and 57 % throughout
the day; this corresponds to a daytime average VCD of
∼ 2.3 ± 0.16 × 1013 molec cm-2, where the error reflects the
average error on the VCD as seen in Fig. 6b. The overall error on
the average VCD is ∼ 0.7 × 1013 molec cm-2 and takes
into account change in the VCD when assuming the maximum profile error given
by the OE inversion. This treatment is expressed in Eq. (3).
Totalerror=VCDPrf-VCDPrf+OEerr2,
where VCDPrf is the VCD calculated directly from the OE inversion
a posteriori, VCDPrf+OEerr is the VCD calculated from the OE inversion
a posteriori profile plus the OE error of that profile, and the factor of 2
reflects that the OE inversion error at each altitude covers oscillations
both higher and lower than the retrieved profile at that point.
Diurnal variation in the BrO vertical profiles represented as
partial VCDs for the BL (0–1 km), troposphere (0–15 km), total (0–50 km) (b),
and the total DoFs from the inversion for each profile in (a). The
error bars on the VCDs indicate the range of values retrieved using
the three different a priori profiles.
The tropospheric VCD of ∼ 2.3 ± 0.7 × 1013 molec cm-2
BrO derived from this study falls in the range of other currently
reported measurements, which span from ∼ 1 to 3 × 1013 molec cm-2
(Van Roozendael et al., 2002; Salawitch et al., 2005; Theys
et al., 2007, 2011; Coburn et al., 2011; Volkamer et al.,
2015; Wang et al., 2015). The observed tropospheric BrO VCD is consistent
with the VCD of 2.0 × 1013 molec cm-2 measured by the same
instrument during a previous study in the same location (Coburn et al.,
2011). This value is higher than other ground-based measurements in the
Southern Hemisphere tropics (Theys et al., 2007), which reported
tropospheric VCDs of 1.1–1.2 × 1013 molec cm-2. Aircraft profiles
over the Pacific Ocean have reported variable BrO VCDs ranging from 1.0 to
1.7 × 1013 molec cm-2 BrO, and a campaign average of
1.3 ± 0.2 × 1013 molec cm-2 BrO (Volkamer et al., 2015; Wang et al.,
2015). While it appears that the BrO VCD is highly variable, depending on
the location and time of measurement, the average BrO VCD measured by the
GOME-2 (Global Ozone Monitoring Experiment-2) satellite for January/February in the
tropics (30∘ N–30∘ S; longitudinal average) is 1.6 × 1013 molec cm-2,
in reasonably close agreement with ground and aircraft studies.
Much of the current knowledge of bromine chemistry in the upper
troposphere–lower stratosphere (UTLS) is currently based on balloon-borne
direct-sun BrO measurements (Pundt et al., 2002; Dorf et al., 2006), which
have found an order of magnitude lower BrO. This lower BrO is – at least in
part – due to atmospheric variability (see e.g., Wang et al., 2015; Schmidt
et al., 2016). Ground, aircraft and satellite measurements consistently
support elevated BrO in the FT. A reassessment of halogen chemistry in the
UTLS (bromine and iodine) seems warranted (Volkamer et al., 2015; Wang et
al., 2015; Saiz-Lopez et al., 2015).
Summary of mercury reactions and rate coefficients used in the box model.
Reaction
Rate or equilibrium1
Reference
coefficient2
Hg0 + O3 → HgO + O2
3 × 10-20
Hall (1995)
HgO(g) ↔ HgO(aq)
Keq1
Rutter and Schauer (2007)
HgO(aq) → Hg(g)0
1.12 × 10-5
Costa and Liss (1999)
Hg0 + Br ⟶M HgBr
1.46 × 10-32 ⋅ T298-1.86 ⋅ [M]
Donohoue et al. (2006)
HgBr + M → Hg0 + Br + M
4.0 × 109 ⋅ exp-7292T
Goodsite et al. (2012)
HgBr(g) ↔ HgBr(aq)
Keq1
Rutter and Schauer (2007)
HgBr + Y3 → HgBrY
2.5 × 10-10 ⋅ T298-0.57
Goodsite et al. (2004)
→ Hg0 + Br2
3.9 × 10-11
Balabanov et al. (2005)
HgBrY(g) ↔ HgBrY(aq)
Keq1
Rutter and Schauer (2007)
HgBrY(aq) → Hg(g)0
1.12 × 10-5
Costa and Liss (1999)
HgBr + Y′4 → HgBrY′
1 × 10-10
Dibble et al. (2012)
HgBrY(g)′ ↔ HgBrY(aq)′
Keq1
Rutter and Schauer (2007)
HgBrY(aq)′ → Hg(g)0
1.12 × 10-5
Costa and Liss (1999)
1 Equilibrium coefficient is parameterized according to Rutter and
Schauer (2007): Keq = (SA - PM)/10((-4250/T)+10), where
SA is the specific aerosol surface area, and PM is the particulate mass;
2 rate coefficients are given in either cm3 molec-1 s-1
or s-1;
3 Y = Br, OH;
4 Y′ = HO2, NO2, BrO, IO, I.
Comparison with models
Global model results for BrO VCDs generally predict much lower columns than
observed in this study. Yang et al. (2005), using the p-TOMCAT model, found
midlatitude values of ∼ 0.4 × 1013 molec cm-2 for
spring/summer months and Parrella et al. (2012), using GEOS-Chem, reported
values of < 1 × 1013 molec cm-2 for the entire year for
the Northern Hemisphere midlatitudes. The value from Parella et al. (2012)
is slightly more comparable to that predicted by WACCM for the case study
day (∼ 1 × 1013 molec cm-2, see Fig. S3).
Wang et al. (2015) reported sensitivity in the predicted BrO to updated
heterogeneous chemistry of bromine (Ammann et al., 2013). Recently, Schmidt
et al. (2016) incorporated the enhanced heterogeneous
chemistry (EHC) scheme into GEOS-Chem. The EHC bromine chemistry mechanism
produces higher average tropospheric BrO columns, which are more consistent
with our observations. They found that average tropospheric BrO
concentrations are ∼ 50 % more than predicted from the
Parrella et al. (2012) bromine mechanism; this is similar to the findings
from this study, where the tropospheric VCD predicted by GEOS-Chem (which
uses the Parrella et al. (2012) bromine chemistry) is
∼ 30 % lower than the measured VCD. Notably, different representations of
halogen chemistry can result in very different BrO vertical profile shapes
(Sander et al., 2003; Fernandez et al., 2014; Long et al., 2014; Liang et
al., 2014), despite an apparently good agreement (∼ 1 × 1013 molec cm-2)
in the BrO VCD at tropical latitudes. The ability to
predict BrO vertical profiles is particularly relevant to predict
distributions of oxidized mercury in the lower troposphere (see Sect. 3.4.2).
Mercury modeling
A detailed description of the box model utilized in this study to assess the
oxidation pathways of mercury is found in Sect. S4. Table 2
gives an overview of the reactions controlling the cycling of atmospheric
mercury included in the box model and Fig. S8 contains the
vertical profiles of modeled and measured parameters used as input to the
box model. In general, the mercury modeling scheme employed in this study
follows that as set forth in previous works (Selin et al., 2007; Holmes et
al., 2009, 2010; Wang et al., 2015).
Mercury oxidation rates
The primary finding from a comparison of the Hg0 oxidation rates for
the two radical species tested in this study (Br and O3) is that the
reaction with Br dominates the overall rate throughout the troposphere for
the conditions tested by this case study, independent of initial BrO
vertical profile used. The column integral rate is 7.8 × 104 molec cm-2 s-1
for O3, while the Br rates are 3.5 × 105 and
3.7 × 105 molec cm-2 s-1 for the BrO vertical profiles from the
MAX-DOAS measurements and GEOS-Chem, respectively. Preliminary tests included
chlorine (Cl) radicals as an additional oxidant using a vertical
distribution estimated for the tropical troposphere (Wang et al., 2015).
However, this reaction was deemed unimportant, and is thus not included
here, due to the column integral oxidation rates being factors of 18 and 5
lower than column rates for Br and O3, respectively. The reaction rates
from Br are at least a factor of 3 greater than the contribution from O3.
Box model results of the rate of mercury oxidation as a function
of altitude for two species: (1) ozone (red); and (2) bromine radicals
(solid blue: MAX-DOAS; dashed pink: GEOS-Chem) (a); the corresponding
lifetimes are found in (b). The black dashed line at 4 km shows where
measurement sensitivity starts to drop because of the decreasing amount of
BrO (the measured parameter) in the lower layers of the atmosphere.
The vertically resolved rates are shown in Fig. 7a for both
reactions: Br (solid blue and dotted pink traces representing the BrO
profiles from the MAX-DOAS measurements and GEOS-Chem, respectively, used to
derive the Br radical concentrations) and O3 (red trace), reflecting
the contributions of these reactions at different altitudes. Also included
in Fig. 7b are the corresponding lifetimes of GEM (as a function of
altitude) against oxidation by the two molecules. Only in the lowest layers
of the atmosphere do the rates of oxidation through reaction with O3
become comparable or greater than those of the reaction with Br. However, it
should be noted that for altitudes < 4 km, where the rate of reaction
with O3 dominates, the BrO VMR is < 0.03 pptv, while the O3
VMR is ∼ 50 ppbv, and in cases where reaction with Br
dominates (still at altitudes < 4 km) the BrO VMR is still
< 0.3 pptv. The BrO vertical profile from the MAX-DOAS instrument below 4 km
contains the highest amount of uncertainty because BrO is essentially zero
within the variability of the measurements. The column integral rates of
oxidation based on the MAX-DOAS measurements and the BrO profile from
GEOS-Chem are actually quite similar (only ∼ 5 %
difference), which is most likely due to the additional bromine chemistry
included in this model (see Sect. 2.2) as compared to other models (e.g.,
GEOS-CCM); however, the discrepancies in the vertical distribution of BrO
lead to a mismatch in the altitudes at which the oxidation can occur. For
both BrO profiles, the reaction with Br dominates above 4 km (indicated by
the black dashed line in Fig. 7). Altitudes above 4 km are responsible for
∼ 72 % of the column integral rate of GEM oxidation (through
reaction with Br) based on the MAX-DOAS measurements, while that value
decreases to ∼ 56 % for the bromine profile from GEOS-Chem.
Additional differences between column oxidation rates for the MAX-DOAS and
GEOS-Chem BrO profiles exists at altitudes < 4 km, where the MAX-DOAS
measurements find no BrO and GEOS-Chem predicts up to 0.75 pptv. Similar to
the findings of this study, other publications have put upper limits on
marine boundary layer (MBL) BrO ∼ 0.5 pptv (Gomez-Martin et
al., 2013; Wang et al., 2015; Volkamer et al., 2015).
Box model results for the scavenging of the HgBr adduct as a
function altitude for two different reaction schemes: traditional (a)
and revised (b). Panel (c) contains the ratio of the total rates (black
traces in (a) and (b) to show the enhancement in the rate of the
scavenging reaction when other reactants are taken into consideration.)
Mercury oxidation pathways
Figure 8 shows the results of the “traditional” (Fig. 8a) and “revised”
(Fig. 8b) HgBr scavenging schemes (see Sect. S4 for descriptions)
on the rate of GEM oxidation as a function of altitude. In each panel,
contributions of individual molecules (colored lines) are shown along with
the total removal rate (black lines). Figure 8c shows the vertical profile
of the ratio of the “revised” total rate to “traditional” total rate,
which demonstrates the enhanced oxidation of HgBr when considering the
additional scavenging reactions. In the “traditional” model the percent
contributions to the column integrated rate of oxidation of HgBr are
57.5 and 42.5 % for OH and Br, respectively, and in the “revised”
model the percent contributions are as follows: 71.3 % (NO2),
21.3 % (HO2), 4.3 % (BrO), 1.4 % (OH), 1.1 % (Br); 0.5 %
(IO), and < 0.1 % (I). Note that mercury oxidation is initiated by
reaction between Br radicals and GEM in both reaction schemes, and the
additional scavenging reactions in the “revised” scheme primarily increase
the overall rate of oxidation at altitudes where HgBr decomposition is fast.
The greatest enhancement is seen below 8 km, where the overall rate of
oxidation is ∼ 100 times faster, primarily because of the
reaction of HgBr with NO2 and HO2.
Figure 8 also illustrates the increased number of species produced from the
additional oxidation mechanisms, some of which may have physical and
chemical properties that differ from the two products of the “traditional”
mode, which are also products in the “revised” scenario but are present at
much lower concentrations. In the “traditional” scenario at 1 km, the
scavenging products HgBrOH and HgBr2 account for 96 and 4 % of
the total HgBrX, respectively, and these values drop to 0.5 and
≪ 0.1 % in the “revised” scenario, where HgBrNO2
accounts for 73 % of HgBrX. In the “revised” scenario, HgBrNO2
remains the major product throughout the atmosphere, but at free-tropospheric
altitudes HgBrHO2 also contributes significantly at
37 %, compared to HgBrNO2 at 52 %. There are currently no
observations of the molecular composition of GOM with which to evaluate
these simulated product distributions.
Mercury lifetime with respect to oxidation
The above oxidation rates correspond to a minimum lifetime of GEM with
respect to oxidation by bromine radicals of ∼ 20 days in the
FT (based on the MAX-DOAS measurements). The average total (Hg0 + Br
and Hg0 + O3) tropospheric column lifetimes are 38 and 54 days
(altitudes > 4 km) for the scenarios including BrO profiles from
the MAX-DOAS measurements and GEOS-Chem, respectively, where in both cases
the contributions from the reactions with O3 are the same and the
differences are owing to the differing amounts of Br radicals. This is much
shorter than the currently expected atmospheric lifetime on the order of
several months. However, the box model only accounts for partitioning of the
GOM species between the gas phase and aerosols; once they are in the aqueous
phase they can be photoreduced to GEM, which can then subsequently return to
the gas phase (Costa and Liss, 1999), thus extending the effective lifetime
significantly beyond that calculated above. The kinetic coefficients for
reactions involving HgBr come mainly from quantum chemical calculations,
which have significant uncertainties, so the simulated GEM lifetime could be
extended by reducing the rate coefficients within their uncertainties.
Another possible mechanism is the photodissociation of HgBrX products
containing species that have significant absorption cross sections in the
ultraviolet–visible (UV/Vis) region of the electromagnetic radiation
spectrum, e.g., HgBrNO2 and HgBrHO2, which could reproduce HgBr. This
HgBr could then thermally decompose to re-form GEM, or be oxidized again.
Our observations are consistent with previous findings (Wang et al., 2015)
that establish GEM as a chemically highly dynamic component of the FT. It is
expected that GOM species will also go through additional processes in the
aqueous phase, which could significantly impact the ultimate fate of the mercury.
Observed (black solid trace) and simulated (blue traces)
concentrations of oxidized mercury (HgII = GOM + PBM) during
April 2010 (9 April highlighted). Model results from GEOS-Chem (solid) include a
sensitivity test, labeled FT (dashed trace), in which North American
anthropogenic emissions are zero and no GEM oxidation occurs in the lower
troposphere (> 700 hPa); therefore all HgII in the “FT”
model is derived from oxidation of in the free troposphere or long-range transport.
Atmospheric implications
The rapid oxidation of mercury in the lower FT is of potential relevance in
the SE US, where there have been several studies linking deep convective
activity to the elevated levels of mercury found in rainwater (Guentzel
et al., 2001; Landing et al., 2010; Nair et al., 2013). The bioaccumulation
of methyl mercury in fish tissues is particularly relevant in this region,
where it has been deemed unsafe to eat fish harvested from many lakes in the
region (Engle et al., 2008; Liu et al., 2008). Wet deposition measurements
of mercury exceed what can be explained through regional sources in the
southeast. In fact, Guentzel et al. (2001) estimated that < 46 %
of the mercury deposited in Florida was a result of local emissions, while the
other > 50 % was attributed to long-range transport of mercury
in the atmosphere; the transported fraction may have increased since those
data were collected in the 1990s because regional US mercury emissions have
declined while global emissions have risen. This attribution, coupled with
mountaintop and aircraft studies locating elevated levels of GOM in the FT
(Swartzendruber et al., 2006; Faïn et al., 2009; Swartzendruber et al.,
2009; Lyman and Jaffe, 2012; Brooks et al., 2014; Weiss-Penzias et al., 2015,
Shah et al., 2016), strongly suggests the presence of a global “pool” of
mercury in the upper atmosphere that contributes to widespread mercury
deposition on a local to regional scale.
On the case study day, HgII (= GOM + PBM) concentrations at the nearly
Pensacola MDN site reached 25 pg m-3 around midday and nearly 40 pg m-3
on 10 April (Fig. 9). High HgII events in Pensacola are
frequently consistent with emissions from a nearby coal-fired power plant;
however about 25 % of such events have significant contributions from the
FT (Weiss-Penzias et al., 2011). The high HgII : SO2 ratios
(10–20 pg m-3 ppb-1) and high NOy : SO2 ratios
(3.5–5.2 ppb ppb-1) recorded on 9–10 April make the power plant an unlikely source
of the HgII on these days (Fig. S9); for comparison,
Weiss-Penzias et al. (2011) reported HgII : SO2 and
NOy : SO2 ratios of 3.5 and 1.0, respectively, in power plant
plumes. Like HgII concentrations, ozone concentrations and HgII
dry deposition in Florida also peak in spring, all of which are consistent
with a significant source of surface HgII being from the upper
troposphere (Lyman et al., 2009; Gustin et al., 2012). We use the GEOS-Chem
model to further probe the sources of this HgII. Figure 9 shows that
the model generally reproduces the day-to-day variability of HgII
observations during April 2010 (but not extremes), such as the low
concentrations during 7–8 April and relative maxima on 5 and 10 April. In
both the model and observations, HgII concentrations rise abruptly in
the morning, consistent with entrainment of HgII aloft and unlike the
other combustion tracers (e.g., NOy, SO2) that reach high
concentrations at night (Fig. S9). The model does not reproduce
the abrupt drop in HgII around 18:00 CST, however, which may be partly
due to local sea breeze circulations that are evident in wind observations
but not simulated at the 4∘ × 5∘ model resolution. In
addition, mixing depth errors in the driving meteorology are known to affect
other species, particularly at night (Lin and McElroy, 2010; McGrath-Spangler and
Molod, 2014). To assess the contribution of the FT to surface HgII
during April 2010, we conduct an additional model simulation with zero
anthropogenic emissions in North America and no mercury redox chemistry in
the lower troposphere (> 700 hPa). The simulation is initialized
from the base run on 1 April. Due to its fast deposition, all boundary-layer
HgII in the sensitivity simulation after about 1 day originates from
oxidation of GEM in the free troposphere or, less likely, from
intercontinental transport of anthropogenic HgII. Figure 9 shows there
is little difference between the base and sensitivity simulations, meaning
that the FT is the main source of boundary-layer HgII in the model and
that variability in the FT component explains most of the day-to-day
HgII variability. Thus, the 3-D model shows that conditions on 9 April
are favorable for HgII transport to the boundary layer, exactly when
the DOAS observations find that substantial amounts of BrO are present in the FT.
The GEM oxidation mechanism in the 3-D model corresponds to the
“traditional” scheme in the box model and bromine concentrations in
GEOS-Chem are lower than recent aircraft observations (Volkamer et al.,
2015), so the GEM oxidation may be faster than simulated. Greater HgII
production in the FT might help correct the model's 20 % low bias in
mercury wet deposition over the southeastern US (Zhang et al., 2012), but this
would depend on the rate of any compensating reduction reaction, as
discussed above. The findings of this study indicate that the amount of
bromine located in the FT can be sufficient to quickly oxidize GEM.
Conclusions
We show the benefits of determining SCDRef to maximize the sensitivity
of ground-based MAX-DOAS measurements to detect BrO in the FT and to improve
the overall consistency of time-resolved BrO tropospheric VCDs. The
retrieval can also be applied to other trace gases. Knowledge of SCDRef
allows the derivation of one vertical profile for each MAX-DOAS scan
throughout the day (SZA < 70∘) and the assessment of the
diurnal variation of the partial BrO vertical columns. Our retrieval is complementary
to previous studies that have characterized the stratosphere using zenith-sky measurements
under twilight conditions (Theys et al., 2007; Hendrick et al., 2007). Here, we minimize
the influence of O3 absorption and the contribution of stratospheric BrO to the overall
BrO signal by using BrO dSCDs measured at low SZA and CTM output to constrain
stratospheric BrO. The FT VCDs reported here are in good agreement
with the previously cited values for BrO, with the average BrO FT VCD
(∼ 2.3 × 1013 molec cm-2) falling within the range
reported by other studies (1–3 × 1013 molec cm-2) (see Wang et
al., 2015, and references therein). These measurements all point to the presence
of background amounts of BrO in the FT that are larger than current models
predict. The method employed here would further benefit from deployments at
mountaintop sites, where aerosol shielding of the FT can be overcome and
tropospheric vertical distributions assessed more frequently.
The presented box model studies indicate that, for the conditions
probed,
bromine radicals are the dominant oxidant for atmospheric GEM throughout the
FT above the studied region. Given the similarities between the vertical
profiles of BrO derived in this study and other profiles measured in the
tropical FT (Volkamer et al., 2015; Wang et al., 2015), the results from our
case study may apply more broadly, though past aircraft and modeling studies
have reported significant variability in BrO and Bry (Wang et al.,
2015; Schmidt et al., 2016). The drivers for such variability deserve
further investigation. Our results confirm that mercury is rapidly oxidized
by bromine, and is a chemically highly dynamic species in the atmosphere. The
chemical lifetime of GEM is ∼ 45 days in the tropical FT based
on calculations and vertical profiles presented in this study; longer GEM
global lifetimes should thus be regarded to indicate “effective
lifetimes”, i.e., they are the result of rapid chemical cycling of GOM back to
GEM (e.g., photoreduction; Pehkonen and Lin, 1998; Gustin et al., 2002; Tong
et al., 2013). Mercury measurements during our study period show high
surface HgII concentrations that likely originate in the FT, meaning
that we have observed substantial BrO columns under conditions favorable for
HgII transport to the boundary layer. Additionally, this study suggests
that the experimental observation of elevated GOM in the FT may be linked to
our incomplete understanding about tropospheric bromine sources
(Swartzendruber et al., 2006; Faïn et al., 2009; Lyman and Jaffe, 2012;
Wang et al., 2015) and indicates that conditions exist where the amount of
bromine located in the FT above the coastal regions of the SE US is
sufficient to quickly oxidize GEM to GOM. This GOM can then be
wet-deposited, and can help explain the observed elevated mercury wet deposition pattern
in this region. Our results highlight the need to understand BrO vertical
profiles in the FT, and represent them in atmospheric models to understand
the location where mercury is oxidized in the atmosphere, and is available
for wet and dry deposition. More studies are needed to test and represent
the bromine sources in atmospheric models; test atmospheric GOM abundances
by field data; and clarify the chemical identity, the global distribution,
and dry and wet removal processes of GOM.