A comparison of Arctic BrO measurements by chemical ionization mass spectrometry and long path-differential optical absorption spectroscopy



[1] A measurement intensive was carried out in Barrow, Alaska, in spring 2009 as part of the Ocean-Atmosphere-Sea-Ice–Snowpack (OASIS) program. The central focus of this campaign was the role of halogen chemistry in the Arctic. A chemical ionization mass spectrometer (CIMS) performed in situ bromine oxide (BrO) measurements. In addition, a long path-differential optical absorption spectrometer (LP-DOAS) measured the average concentration of BrO along light paths of either 7.2 or 2.1 km. A comparison of the 1 min observations from both instruments is presented in this work. The two measurements were highly correlated and agreed within their uncertainties (R2 = 0.74, slope = 1.10, and intercept = −0.15 pptv). Better correlation was found (R2 = 0.85, slope = 1.04, and intercept = −0.11 pptv) for BrO observations at moderate wind speeds (>3 m s−1 and <8 m s−1) and low nitric oxide (NO) mixing ratios (<100 pptv). The improved agreement is likely due to the elimination of periods when the spatial distribution of BrO is inhomogeneous. The detection limit obtained for the CIMS was 2.6 pptv (3σ) for a 4 s integration period, and the estimated uncertainty was ∼30%. The detection limits for the LP-DOAS ranged from 0.7 to 5 pptv (3σ) depending on the level of ambient light and the chosen light path, and the estimated systematic error was 10%. The agreement between the CIMS and LP-DOAS is excellent and demonstrates the capability of both instruments to selectively and accurately measure BrO with high sensitivity.

1. Introduction

[2] Bromine plays an important role in tropospheric ozone depletion events (ODEs). Barrie et al. [1988] first observed that filterable bromine (f-Br) strongly anticorrelates with ozone (O3). Hausmann and Platt [1994] reported the first measurements of BrO during ODEs in the Arctic at Alert, Canada. Frequent surface ODEs have been observed in the springtime polar marine boundary layer in a variety of locations such as Barrow, Alaska [Oltmans, 1981; Oltmans and Komhyr, 1986; Oltmans and Levy, 1994], Alert, Canada [Bottenheim et al., 1986, 2002], Ny-Ålesund, Svalbard [Lorenzen-Schmidt et al., 1998; Tuckermann et al., 1997] and Halley Bay, Antarctica [Jones et al., 2006]. A bromine radical catalyzed cycle, including Br and BrO, was proposed to destroy O3 in the troposphere [Barrie et al., 1988; Hausmann and Platt, 1994; Fan and Jacob, 1992; McConnell et al., 1992]. The relevant mechanisms are summarized in the review of Simpson et al. [2007]. The self reaction of BrO is often considered the rate limiting step for O3 destruction as it destroys odd oxygen and regenerates active bromine [Sander et al., 2006; Fan and Jacob, 1992]. Consequently, accurate measurements of BrO at low pptv levels are needed to quantify the active bromine concentrations and the O3 destruction rate catalyzed by bromine chemistry.

[3] Although evidence of bromine chemistry inducing surface O3 depletion events is strong, the sources of bromine and the bromine activation mechanisms remain unclear [Simpson et al., 2007]. In addition, the horizontal and vertical distribution of BrO in the troposphere is uncertain, especially where BrO levels are expected to be low even though the impact of a few pptv of BrO can be significant. Satellite data suggest that the average global daytime tropospheric BrO is of the order 0.5–2 pptv [Sinnhuber et al., 2005; Richter et al., 2002]. A series of DOAS data [Fitzenberger et al., 2000; Friess et al., 1999; Hendrick et al., 2007; Schofield et al., 2004] also indicate that a background level (∼1 pptv) of BrO exists in the free troposphere. Moreover, recent studies indicate that the derivation of tropospheric BrO column densities from satellite may need to be reassessed [Salawitch et al., 2010; Theys et al., 2009]. Consequently, in situ measurements of BrO with high sensitivity and time resolution will be useful to address these issues.

[4] Understanding the mercury cycle in the polar boundary layer and potentially other locations requires knowledge of BrO mixing ratios. Schroeder et al. [1998] observed that gaseous elemental mercury (GEM) undergoes rapid decreases in concentration during an ODE in the Arctic springtime. GEM was generally thought to be very stable (lifetime = ∼6–24 months) in the atmosphere [Schroeder and Munthe, 1998]. However, these results indicated that relatively benign GEM could be oxidized by halogen species to more reactive and soluble mercury compounds which can deposit to the surface or on aerosol [Steffen et al., 2008].

[5] The connections between the observed GEM depletion and simultaneous O3 loss in the Arctic have been studied in field campaigns in Barrow, Ny-Ålesund and other Arctic sites [e.g., Lindberg et al., 2002; Berg et al., 2003], where reactive gaseous mercury (RGM) anticorrelated with GEM during GEM depletion events. At this time the role of O3 and hydroxyl radical (OH) as oxidants of mercury (Hg) is not well established as the rate constants for these species may be too small to be important [Hall, 1995; Sommar et al., 2001]. However, Br atoms have been demonstrated to effectively oxidize Hg (i.e., convert GEM into RGM) [Ariya et al., 2002; Donohoue et al., 2006]. Br atom levels are usually estimated from BrO measurements and the ratio calculated assuming photochemical steady state (Br/BrO ratio ranges from 0.07 to 1 when O3 ranges from 40 ppbv to 3 ppbv) [Zeng, 2005]. Thus, accurate measurements of BrO concentrations are needed to estimate the oxidation rate of elemental mercury by bromine atoms and further understanding of the atmospheric chemistry of mercury.

[6] As BrO is a short-lived radical species with low atmospheric abundance, accurate and reliable detection of BrO is quite challenging. The most developed and commonly used technique for ground based BrO measurement is long path-differential optical absorption spectroscopy (LP-DOAS). Column abundances of trace gases integrated along a path of several kilometers [Platt, 1994] are determined using specific narrow-band absorption structures in the ultraviolet and visible spectral regions. This technique has been used to measure BrO at: Alert, Canada [Hausmann and Platt, 1994]; Ny-Ålesund, Spitsbergen [Tuckermann et al., 1997]; Halley Station, Antarctica [Saiz-Lopez et al., 2007]; aboard the research ice-breaker in the Amundsen Gulf, Arctic Ocean [Pöhler et al., 2010], and in many other locations. LP-DOAS can provide measurements of O3, IO, ClO, NO2, OClO, OIO, and I2 and other species as well as BrO [Tuckermann et al., 1997; Saiz-Lopez et al., 2007; Stutz et al., 2002]. In recent years, cavity enhanced absorption techniques have been developed to provide DOAS measurements with much higher spatial resolution [e.g., Ball et al., 2004; Langridge et al., 2006]. A review of principles and applications of LP-DOAS is given by Platt and Stutz [2008].

[7] Satellite sensors that examine the same spectral features as LP-DOAS can be used to map the global BrO column density [Wagner and Platt, 1998; Chance, 1998; Richter et al., 1998]. However, there are challenges associated with deriving the tropospheric distribution of BrO from satellite data, especially in the vertical dimension. One difficulty in satellite measurements of BrO is that the tropospheric column must be obtained by subtracting the stratospheric column from the total column [Theys et al., 2009; Richter et al., 1998]. This can lead to significant errors where the stratospheric column is not accurately known [Salawitch et al., 2010; Theys et al., 2009].

[8] The first in situ measurement of tropospheric BrO was by chemical conversion/resonance fluorescence (CC/RF) [Avallone et al., 2003]. This technique was originally developed for measurement of BrO in the stratosphere [Brune et al., 1989; Avallone et al., 1995]. Due to the quenching of the fluorescence signal and the absorption of the fluorescence signal by oxygen and water, the detection sensitivity of this instrument decreases as altitude decreases [Toohey et al., 1990]. The uncertainty of BrO measurements by the CC/RF technique during ARCTOC'96 in the troposphere was typically about 5 pptv [Avallone et al., 2003]. The overall accuracy of BrO measurements was estimated to be about −30%/+50% (2σ) for ARCTOC'96 and ±35% (2σ) at Alert 2000 [Avallone et al., 2003].

[9] In recent years, methods to measure halogens such as BrO, Br2, Cl2, and ClNO2 by CIMS have been developed [e.g., Finley and Saltzman, 2008; Kercher et al., 2009; Neuman et al., 2010]. The CIMS methods to measure bromine species from the NASA DC-8 and NOAA P3 during the research flights in the Arctic [Neuman et al., 2010] are very similar to those used in this work. However, the BrO levels measured during these flight campaigns were always below 10 pptv [Neuman et al., 2010], which was lower than expected from satellite data [Salawitch et al., 2010]. Therefore, comparison of the CIMS with the well established LP-DOAS method is important to validate the ability of the CIMS to measure BrO. In this work, the comparison of BrO measurements by CIMS and LP-DOAS in Barrow, Alaska during the OASIS 2009 campaign is presented. Both techniques are described, and ancillary measurements are examined to analyze the instrument comparison.

2. Methods

2.1. Measurement Site

[10] The CIMS and LP-DOAS measurements of BrO were carried out from 3 March to 14 April 2009 at Barrow, Alaska (71°19′ N, 156° 39′ W), during the Ocean-Atmosphere-Sea-Ice–Snowpack (OASIS) campaign. A major focus of the OASIS campaign was to study halogen chemistry and its impact on the Arctic environment. The main research site included two trailers and the Barrow Arctic Research Center (BARC) which was located ∼800 m from the Chukchi Sea to the northwest, ∼15 km from the Beaufort Sea to the northeast and ∼5 km to the northeast of Barrow. The prevailing wind direction was from the northeast (i.e., from the Arctic Ocean) during the measurement period which minimized, but did not eliminate, the impacts of local emissions. The layout of the CIMS and LP-DOAS instruments is shown in Figure 1. The CIMS was installed in the northwest corner of trailer 2 with an inlet approximately 1.5 m above the snow surface. The LP-DOAS telescope was located at the northeast side of the BARC building. Two retro-reflectors (summer-camp and hangar) were installed at a distance of 3623 and 1074 m from the telescope and provided a light path of 7246 and 2148 m, respectively (black arrows in Figure 1). The height above the surface of both LP-DOAS light paths was ∼2 m and varied little with horizontal position. The CIMS instrument was located about 200 m to the southeast of the LP-DOAS telescope. Most of the other chemical measurements were located in one of the two trailers including a 4 channel chemiluminescence instrument used to measure O3, NO, NO2, and NOy [Weinheimer et al., 1998]. The NO observations from this instrument were used as the primary indicator of local pollution.

Figure 1.

The layout of the CIMS and LP-DOAS instruments during the OASIS campaign. The CIMS instrument and LP-DOAS telescope were located ∼800 m from the Chukchi Sea to the northwest and ∼5 km to the northeast of Barrow. Two retro-reflectors (summer-camp and hangar) were installed at distances of 3623 and 1074 m from the telescope and provided light paths of 7246 and 2148 m, respectively (black arrows). The CIMS instrument was about 200 m to the southeast of LP-DOAS telescope. The geometries of the CIMS instrument and LP-DOAS telescope (dashed rectangle area) are displayed in the inset. The CIMS instrument was installed in the northwest corner of Trailer 2 (black dot). The NO and O3 instrument was installed in Trailer 1. The telescope of the LP-DOAS instrument was located in northeast side of the BARC building (black dot).

2.2. CIMS

2.2.1. Instrument

[11] The CIMS used to detect halogen species at Barrow is very similar to that used to measure PANs and other species [Slusher et al., 2004; Kim et al., 2007; Nowak et al., 2006; Huey, 2007]. The inlet configuration is essentially identical to that used previously to measure HNO3 and NH3 [Huey et al., 2004; Nowak et al., 2006]. The design of the inlet is similar to that used to sample highly reactive gases such as OH [Eisele et al., 1997]. Consequently, only details relevant to the BrO measurements are described. The instrument and inlet configuration are shown in Figure 2.

Figure 2.

The instrument and inlet configuration of the CIMS system.

[12] The outer portion of the inlet was a 7.6 cm ID aluminum pipe that extended about 20 cm beyond the wall of the sampling trailer. A total flow of approximately 900 standard liters per minute (slpm, standard temperature = 273 K, standard pressure = 1.01 × 105 Pa) was maintained in the pipe with a blower (AMETEK Windjammer 116637–03). A portion of this flow (13.0 slpm) was sampled into a custom three-way valve, constructed of perfluoroalkoxy (PFA) Teflon, which connected the center of the pipe to the CIMS sampling orifice. Most of this flow (11.6 slpm) was exhausted through a mass flow controller in series with a small diaphragm pump, with the rest (1.4 slpm) entering the CIMS flow reactor through a 0.51 mm dia. orifice. The valve was maintained at a constant temperature of 40°C and could be automatically switched between two flow paths. The first path was equivalent to a 25 cm long, 0.65 cm ID, Teflon tube. The second configuration delivered ambient air through a glass wool scrubber to the CIMS to determine background levels. Finally, the output of either a Br2 (118 ng min−1) or Cl2 (183 ng min−1) permeation tube was periodically delivered to the upstream end of the Teflon valve to monitor the CIMS sensitivity toward halogens.

[13] The CIMS flow reactor was operated at 20 hPa with a total flow of 3.5 slpm maintained by a scroll pump. The total flow consisted of 1.4 slpm of ambient air and 2.1 slpm of ion source flow that consisted of N2 containing approximately 5 ppmv methyl iodide. The methyl iodide was delivered to the ion source by adding 10 standard cubic centimeters per minute (sccm) of a 0.1% mixture in nitrogen to the ion source flow. A flow of 20 sccm N2 was passed through a room temperature bubbler (∼20°C) containing deionized water and introduced directly to the flow reactor. Water vapor was added to the flow tube to increase the sensitivity to BrO and other halogens [Neuman et al., 2010].

[14] The primary reagent ion produced by this configuration is hydrated I which was used to detect a series of halogens using reactions R1R5. Due to collisional dissociation of the hydrated ions in the sampling process, described below, only the detected core ions are listed in the reactions. The collisional rate constants for the ion molecule reactions R2a and R3a are estimated by standard methods assuming that on average I is clustered with four water molecules [Su and Chesnavich,1982; Chesnavich et al.,1980; Johnson et al., 2010]. The rate constants for R1 and R4a are estimated relative to R2a from the sensitivity ratio of BrO and HOBr to Br2. The reaction rate constant of R5a is estimated relative to R3a from the sensitivity ratio of ClNO2 to Cl2.

equation image
equation image
equation image
equation image
equation image
equation image
equation image
equation image
equation image
equation image

A small portion (∼100 sccm) of the ion reactor flow was sampled through a 0.81 mm dia. orifice into a collisional dissociation chamber (CDC) maintained at 0.67 hPa with a molecular drag pump. An electric field of approximately 4 V cm−1 is applied in the CDC to produce energetic collisions to dissociate weakly bound water clusters. Ions at the exit of the CDC are guided by an octopole and sampled into a differentially pumped quadrupole mass spectrometer. The ions are filtered and detected with an electron multiplier detector (K+M) and counted as individual pulses. The signals are reported in Hz, which represents the number of ion counts per second. Each cluster ion resides at a unique mass, and the measurement of different bromine (and chlorine) isotopes provided a critical test of the measurement technique. A typical measurement cycle (15.5 s) monitored the masses listed in Table 1 for the indicated integration times. As a consequence, BrO was measured with a 6% duty cycle which gives a ∼4 s integration period for the CIMS data used to derive the 1 min average used in this work. On the other hand, the LP-DOAS has a duty cycle of ∼40% for BrO. The remaining time was used for background spectra and other wavelengths. The smaller duty cycle for the CIMS is partially responsible for the larger variability in the CIMS data.

Table 1. Summary of Monitored Masses, Ionized Products, Neutral Reactants, and Integration Time for CIMS Measurements During the OASIS Campaign
Mass (amu)IonNeutralIntegration Time (ms)Comments
147IH218ONA1000Hydrated reagent ion isotope; the signal of IH216O at 145 amu was saturated
7979BrMultiple1000Significant channel for Br2 that is more important at lower H2O levels
222I79BrOBrO1000Channel enhanced at higher H2O levels
289I 81Br81BrBr21000 
285I 79Br79BrBr21000 
208I35ClNO2ClNO21000Species other than ClNO2 also present at 208 amu; HOBr detected as IBr at 208amu
210I37ClNO2ClNO21000Background not well determined
62NO3NO3, N2O5, ClNO3500 

2.2.2. Background Determination

[15] Instrumental background signals were measured regularly, and subtracted from the total signals to determine BrO mixing ratios. The background signal at the mass used for BrO detection as well as other halogen species was determined periodically (approximately every 40 min) by scrubbing ambient air with glass wool. The custom made Teflon valve was used to switch between the two flow paths. Laboratory tests [Neuman et al., 2010] have shown that many inorganic halogens (Br2, HOBr, BrO, Cl2, and HCl) can be efficiently removed from a gas stream via contact with glass wool. The effectiveness of halogen removal by the glass wool was confirmed every few days by adding either Cl2 or Br2 upstream of the scrubber. However, we found that ClNO2 is essentially unreactive toward glass wool and a background measurement for this species could not be provided by this method. The ion reactor pressure decreased slightly (∼26.7 Pa or 1.3%) during background measurements. In order to diminish this effect, after 29 March the ion reactor was actively pressure controlled by modulating the ion source flow to maintain constant pressure. The water vapor concentration in the flow tube was not found to change significantly during these background measurements, as the signal due to I(H2O) was found to be essentially constant. Thus, the modulation of the ion source flow to maintain constant pressure did not alter halogen sensitivity but slightly increased the stability of the background measurement. The median absolute difference between successive background measurements was 0.89 pptv before pressure control and 0.77 pptv after pressure control. A typical example of the raw BrO signal (224 amu) with a series of background measurements is shown in Figure 3.

Figure 3.

Typical raw signal of I81BrO (224 amu) (black line) and I79Br81Br (287 amu) (black line with cross markers) from the OASIS campaign. (top) The gray dots show the average I79Br81Br (287 amu) signal during an addition of the Br2 standard. (bottom) The gray squares illustrate the average I81BrO (224 amu) signal during background measurements (shaded areas).

2.2.3. Calibration

[16] Because a portable, constant calibration source for BrO has not been developed, Br2 and Cl2 permeation tubes (Kin-tek) were used as the primary calibration standards for the OASIS campaign. Br2 and Cl2 were detected by the association reactions (R2a and R3a). A known amount of Br2 (1.47 ppbv) or Cl2 (5.18 ppbv) was added to the inlet every 2 h to track the sensitivity of the observed halogen species including BrO. A typical example of the raw signal from a Br2 calibration is shown in Figure 3.

[17] The emission rates of the permeation tubes were measured every few days by conversion to I3 in aqueous solution. This was accomplished by passing the output of the permeation tube in a 10 sccm flow of nitrogen through an aqueous solution of KI (2% w/w KI, 1 mM phosphate buffer, pH = 7). Br2 or Cl2 quantitatively oxidizes I(aq) to form I3(aq) via the following net reactions [Wu et al., 1963; Kazantseva et al., 2002].

equation image
equation image

The resulting I3 was quantified by optical absorption at 352 nm [Finley and Saltzman, 2008]. The quantitative conversion of Br2 to I3 was also confirmed in the laboratory by ion chromatographic measurements of Br. The average emission rate of the Br2 permeation tube for the OASIS campaign was 118 ng min−1 with a standard deviation of 5%. The average emission rate of the Cl2 permeation tube for the OASIS campaign was 183 ng min−1 with a standard deviation of 7%. The sensitivity for the detection of Cl2 relative to Br2 at the most abundant isotopes (mass 197 and mass 287) was determined to be 1.6 ± 0.1. The measurement of this ratio allowed the use of either permeation tube as the primary standard during OASIS. The Br2 permeation tube was used as the primary standard for most of the campaign. The Cl2 permeation tube was used from 28 March to 4 April 2009.

[18] The relative rates of reactions R1 and R2a were measured before the campaign. BrO was generated from Br2 by the reaction of O (3P) with Br2 in excess O3. O3 and Br2 were flowed through an oven upstream of the inlet to the CIMS. When the oven was heated to ∼350°C, BrO was produced by the following series of reactions.

equation image
equation image
equation image

By using a large excess of O3, each Br2 molecule was quantitatively converted to two BrO molecules. The sensitivity of BrO relative to Br2 is measured to be 0.47 ± 25% which is very similar to that determined by Neuman et al. [2010] using a microwave discharge to make Br atoms from Br2 [Orlando et al., 1991]. We have found that for a given set of conditions the relative sensitivity of the CIMS to different compounds is generally stable within 10% [Kim et al., 2007]. The uncertainty in the measurement over a range of experimental conditions is estimated to be 25%. Combining this with the estimated uncertainties in the Br2 and Cl2 permeation rate of 5% and 7%, and an estimated uncertainty in the Cl2 to Br2 sensitivity ratio of 6%, the total uncertainty of the BrO measurement is estimated to be 29% using Br2 as the calibration gas and 34% using Cl2 as the calibration gas.

2.2.4. CIMS Performance

[19] The CIMS instrument performance is assessed by examining the detection sensitivity and the background signal levels to estimate a limit of detection for BrO. The sensitivity was determined from the response to a standard addition of Br2. The typical sensitivity of the CIMS for Br2(R2a) during OASIS was 12 Hz per pptv. This corresponds to a typical sensitivity for BrO of 6 Hz per pptv. The median background signal at the BrO mass was 69 Hz which is equivalent to ∼12 pptv of BrO. Because BrO was monitored for 4 s out of every minute by CIMS, the standard deviation of the background measurement due to counting statistics is calculated to be 0.70 pptv for a 1 min integration period. The absolute difference between successive 1 min average background measurements was essentially normally distributed with a median of 0.86 pptv for the whole data set and with a median of 0.78 pptv after pressure control was instituted. This indicates that variance of the background due to factors other than counting statistics such as electrical noise, flow turbulence, etc., is small and that the CIMS detection limit will improve with averaging. We estimate the 3σ limit of detection as 2.6 pptv for a 4 s integration period which is calculated by multiplying the variance of individual determinations of the background signal by a factor of three. As most of the background variance is due to counting statistics, the detection limit may decrease to 1 pptv for a 1 min integration period when the BrO measurement time is increased to 60% of the duty cycle.

[20] The natural isotopic abundances of 79Br and 81Br are 50.69% and 49.31%, respectively. The total raw signal in ambient mode at I79BrO (222 amu) is plotted against the signal at I81BrO (224 amu) from 18 March to 14 April 2009 in Figure 4. The correlation is excellent, R2 = 0.85, but if only random error due to counting statistics was present, the correlation would be higher for this data set (maximum R2 = 0.96). This indicates that 10% of the observed variation is due to other sources such as a small interference in mass 222 (see below). An equally weighted, bivariate regression analysis gives a slope of 1.11 and an intercept of −0.84 Hz (−0.14 pptv). These results strongly indicate that a species containing one bromine atom is detected at these masses and there are no large interferences present. Observations during the campaign indicate that mass 222 did have a small interference during pollution events and for that reason mass 224 was used to derive BrO levels.

Figure 4.

The raw ambient signal at I79BrO (222 amu) versus I81BrO (224 amu). The correlation coefficient is 0.85 (R2) with a slope of 1.11 and an intercept of −0.84 Hz or −0.14 pptv.

2.3. LP-DOAS

2.3.1. Instrument

[21] Long path-differential optical absorption spectroscopy quantifies the average concentration of trace gases along a well-defined light path using their individual absorption cross sections [e.g., Platt and Stutz, 2008]. The LP-DOAS instrument used in this study was previously deployed aboard the Amundsen RV in Spring 2008 and described by Pöhler et al. [2010]. See section 2.1 and Figure 1 for details about the setup of the light paths at Barrow.

[22] The active part of the LP-DOAS instrument consisted of the following components: Radiation from a broadband light source (Xenon arc lamp, usually Osram XBO 75, Osram XBO500 between 19 and 30 March) was coupled into a single 600 μm mode-mixing quartz fiber using a fused silica lens. The exit end of the single fiber was connected to six transmitting 200 μm fibers at one arm of a Y bundle (total length 10 m). At the bottom of the Y, which was placed close to the focal point of a telescope mirror (30 cm diameter, 1.5 m focal length), these six fibers surrounded a single 200 μm receiving fiber. The pointing of the telescope could be adjusted by stepper motors. The receiving fiber, the other arm of the Y, led to an Acton 300i spectrometer equipped with a CCD camera. For stray-light reduction, a BG24A filter or a UG5 filter (both manufactured by Schott, 2 mm thickness) was placed behind the entrance slit of the spectrometer.

[23] The light path through the atmosphere could be blocked using an aluminum diffuser plate placed at about 1 mm distance in front of the single end of the Y bundle. This “shortcut” configuration was applied after each measurement and was used to record the light source reference spectrum I0. In addition, a “background spectrum” for each spectrum was obtained with the artificial light source blocked. This was necessary to eliminate a possible signal offset due to scattered sunlight entering the telescope. The offset, as well as the dark-current signal of the CCD electronics, was also accounted for in the evaluation procedure.

2.3.2. Measurement Regime

[24] Varying spectral structures originated from fluctuations of the light source and of ambient illumination (scattered sunlight), which was an issue during the day due to high surface albedo close to the retro-reflector arrays. In order to minimize these effects, atmospheric absorption (I) spectra and reference (I0) spectra, as well as their respective background signal, were recorded in quick succession. The light path had to be manually adjusted depending on meteorological conditions. During periods of poor visibility due to fog, clouds, mirages, or blowing snow, the short light path was chosen instead of the long light path.

2.3.3. Evaluation and Error Analysis

[25] In order to obtain the concentrations of BrO by LP-DOAS at a similar time resolution as for CIMS, individual measurements (integration time 1 to 50 s depending on signal strength) were used for the spectral analysis. The dispersion of the spectrometer was calibrated using the position of mercury emission lines.

[26] The spectral analysis was performed in the wavelength interval between 315.5 nm and 348.0 nm and included the cross sections of the trace gases listed in Table 2. Usually, the trace gas column densities are determined by fitting a linear combination of the narrow-band trace gas absorption structures to the measured optical density spectra τ = −ln(I/I0) using a nonlinear Levenberg-Marquardt algorithm [Kraus, 2004]. In this study, however, the standard evaluation scheme needed to be altered in order to overcome a still relatively large residual background signal. The disadvantageous influence of the high intensity of scattered sunlight on the background signal was amplified by a malfunction of the camera software, which slightly altered the exposure time between measurements depending on the illumination of the CCD. Thus, in addition to a set of trace-gas absorption spectra and the initial intensity ln(I0), three further correction spectra were used to fit the measurement spectrum ln(I). The additional correction spectra consisted of the logarithm of the actual atmospheric background spectrum ln(BG) as well as the squared spectra ln(BG2) and ln(I02), which represent the second term of the Taylor expansion of the exponential Beer-Lambert Law. This approach improves the modeling of nonlinear artifacts within the otherwise linear DOAS algorithm and was first published by Puķīte et al. [2009]. The calculated optical density as well as the absorption cross sections adapted from the literature were filtered using a binomial high-pass filter with 1000 iterations prior to the analysis. A third-order polynomial accounted for residual broadband structures. Finally, the amplitude of the residual spectrum was used as a quality criterion in order to filter unsuccessful fit results due to poor signal quality.

Table 2. Literature Cross Sections Included in the Presented Spectral Evaluation
O3Burrows et al. [1999]241K
BrOWilmouth et al. [1999]228K
O4Greenblatt et al. [1990] 
HONOStutz et al. [2000]298K
NO2Burrows et al. [1998]241K
HCHOMeller and Moortgat [2000]298K
OClOBogumil et al. [2003]293K
SO2Bogumil et al. [2003]243K

[27] The spectral analysis yields the column density of each fitted absorber. The path-averaged concentration of a particular trace gas is then calculated by dividing its column density by the optical path length. To account for remaining systematic structures in the residual, the measurement error (σ) of the Levenberg-Marquardt analysis error is multiplied by a factor of 1.8 according to Stutz and Platt [1996]. In accordance to the evaluation of the CIMS, the detection limit is estimated to 3σ. This leads to an optimum nighttime detection limit of 2 × 107 molecules cm−3 (0.7 pptv) for the long light path and 6 × 107 molecules cm−3 (2 pptv) for the short light path. During the day, however, due to larger interferences with sunlight scattered into the light path, the detection limits were usually higher depending on the viewing conditions and light source used. Mean daytime detection limits are estimated to be 5 × 107 molecules cm−3 (2 pptv) and 1.5 × 108 molecules cm−3 (5 pptv) for long and short light paths, respectively.

2.3.4. Systematic Errors

[28] It is important to consider systematic errors, in particular when comparing two different instruments measuring the same parameter. While the DOAS retrieval is a self-calibrating procedure, the largest contributions to systematic error come from the absolute calibration of cross sections reported in the literature. During the measurements of laboratory cross sections it is crucial to determine the exact concentration of the target trace gas [Wilmouth et al., 1999; Fleischmann et al., 2004]. The differential cross sections of BrO differ significantly within the literature. Therefore a systematic error of 10% for the retrieved BrO concentration is a realistic assumption [Dorf, 2005].

3. Results and Discussion

[29] The CIMS and the LP-DOAS instruments were operated independently to detect BrO from early March to mid-April during the OASIS campaign. BrO measurements from both instruments and O3 observations from a chemiluminescence instrument from the National Center for Atmospheric Research (NCAR) are shown in the time series plot in Figure 5. All data are averaged to a 1 min time basis from 18 March to 14 April 2009. The gaps in the CIMS data are either due to power outages or instrument malfunction. Due to the low BrO duty cycle for the CIMS, the background variance mostly due to counting statistics is 2.6 pptv for the 1 min data average. This largely statistical variance can explain many of the cases when the BrO observed by CIMS is more variable than the LP-DOAS. Both instruments show clear diurnal profiles of BrO with maxima in the daytime and no evidence for significant nighttime levels. The highest BrO concentrations measured by the CIMS and the LP-DOAS are 41 and 42 pptv, respectively. These levels are comparable to the maximum BrO detected in the Amundsen Gulf (41 pptv) by Pöhler et al. [2010]; in Ny-Ålesund, Spitsbergen (∼30 pptv) by Tuckermann et al. [1997]; in Alert, Canada (∼ 30 pptv) by Hönninger and Platt [2002] and in Halley station, Antarctica (∼20 pptv) by Saiz-Lopez et al. [2007].

Figure 5.

Time series of CIMS (red) and LP-DOAS (blue) BrO observations, O3 measurements (green), and j values of Br2 (gold) from 18 March to 14 April 2009.

[30] Similar diurnal BrO profiles were also detected by LP-DOAS in coastal Antarctica [Saiz-Lopez et al., 2007], at the Dead Sea [Hebestreit et al., 1999] and in the Amundsen Gulf [Pöhler et al., 2010]. The OASIS data are consistent with BrO behaving as a short-lived photochemically produced species. It is interesting to note that some model studies have predicted comparable diurnal profiles of BrO with maxima of ∼30 pptv by assuming an initial bromide concentration of ∼50 pptv [Fan and Jacob, 1992; Evans et al., 2003]. These studies predict the highest BrO concentrations near sunrise and sunset with a local minimum at near noon. This behavior is apparent in the diurnal profiles observed by Pöhler et al. [2010] in the Amundsen straits. However, the BrO measurements at Barrow did not show the same diurnal pattern. Consequently, this data set will provide an interesting opportunity to further test models of halogen chemistry.

[31] BrO levels do not have a simple relationship with O3. Midday (0900 LST to 1800 LST) BrO concentrations are listed in Table 3 as a function of O3. The lowest average BrO mixing ratios were observed at low O3 which is consistent with decreasing the formation rate of BrO via reaction R10. The highest BrO levels were observed between 5 and 30 ppbv of O3 and were associated with the “edges” of ozone depletion events (e.g., 25, 30, and 31 March 2009) or regions of partial ozone depletion (10−13 April 2009). Lower BrO concentrations at high O3 levels (>30 ppbv) were also observed, consistent with little halogen chemistry having impacted the air mass. Mixing can also influence the relationship between BrO and O3: during midday there is some increased turbulent mixing that may increase O3 by mixing down from aloft, and dilution of the surface-mediated bromine chemistry that may result in lower BrO concentrations.

Table 3. Daytime (0900 LST to 1800 LST) BrO Concentrations Observed by CIMS as a Function of O3
O3 (ppbv)BrOmean (pptv)BrOmedian (pptv)BrOstd (pptv)Data Points

[32] The unfiltered BrO data from the CIMS and LP-DOAS are well correlated (R2 = 0.74) (Figure 6, top). A weighted, bivariate regression analysis was performed on the data, where the weighting of CIMS and DOAS data was based on the estimated errors described above. The regression analysis yielded a slope of 1.10 and an intercept of −0.15 pptv with a linear correlation coefficient (R2) of 0.74. The median ratio of LP-DOAS data to CIMS data (when [BrO] > 10 pptv) is 1.09 with a standard deviation of 0.34. BrO measured by the LP-DOAS was larger than CIMS observations by approximately 10% with no significant offset. With over 15000 data points and dissimilar spatial scales and duty cycles, such a strong correlation indicates that the atmosphere is reasonably homogeneous over the spatial scale of the experiment (∼4 km) and constant on the measurement time scales (1 min) and that both instruments accurately determine BrO mixing ratios.

Figure 6.

The 1 min LP-DOAS observations plotted against CIMS observations with (top) all data and (middle) filtered data. (bottom) The 10 min LP-DOAS observations plotted against CIMS observations with filtered data. The filtered data exclude points when NO is larger than 100 pptv and wind speeds are less than 3 m s−1 or more than 8 m s−1. The correlation coefficients (R2) for the 1 min unfiltered and filtered data are 0.74 and 0.85, respectively. A bivariate fit of all 1 min data gives a slope of 1.10 and an intercept of −0.15 pptv. A bivariate fit of 1 min filtered data gives a slope of 1.04 and an intercept of −0.11 pptv. The correlation coefficient (R2) for the 10 min filtered data is 0.88. A bivariate regression analysis of the 10 min data yields a slope of 1.01 and an intercept of −0.27 pptv.

[33] The correlation of the BrO measurements is even better, R2 = 0.85 (number of points N = 5347) (Figure 6, middle), if the data are filtered to exclude points when NO is larger than 100 pptv and wind speeds are less than 3 m s−1 or more than 8 m s−1. Note that, due to the lack of NO data after 12 April 2009, the BrO data are not filtered to exclude points at high NO mixing ratio after 12 April 2009. We did not find a clear relationship between the CIMS and LP-DOAS agreement and other variables such as solar radiation or humidity. Using a bivariate regression analysis as above, the slope of the correlation is 1.04 with an intercept of −0.11 pptv. The median ratio of the LP-DOAS to the CIMS data (when [BrO] > 10 pptv) is 1.08 with a standard deviation of 0.29. This analysis probably provides the best comparison of the instruments and indicates that both instruments are measuring BrO within their stated uncertainties.

[34] The correlation of BrO measurements is further improved (R2 = 0.88) when the measurements are averaged to a 10 min time base and filtered to exclude points when NO is larger than 100 pptv and wind speeds are less than 3 m s−1 or more than 8 m s−1 (Figure 6, bottom). This is consistent with some of the variability in the CIMS data being due to counting statistics. The same bivariate regression analysis is applied weighted by the estimated errors. The slope of the correlation is 1.01 with an intercept of −0.27 pptv. The median ratio of the LP-DOAS to the CIMS data (when [BrO] > 10 pptv) is 1.08 with a standard deviation of 0.27 for the 10 min data.

[35] To illustrate the difference between the two measurements as a function of mixing ratio, the normalized difference (ND) of the CIMS and LP-DOAS data is plotted versus the CIMS observations in Figure 7 (filtered 1 min data). The normalized difference is defined by the following equation.

equation image

The average ND for each interval and the standard deviation are shown as black dots and bars, respectively. There is no evident bias above 5 pptv of BrO. Below this level the plot becomes scattered as expected given the detection limits of both instruments.

Figure 7.

The normalized difference (ND) of CIMS and LP-DOAS measurements versus CIMS measurements (filtered 1 min data). ND is defined as: ND = ([BrO]CIMS − [BrO]LP-DOAS)/([BrO]CIMS × [BrO]LP-DOAS)0.5. Individual ND values are shown as gray dots. The average ND for each interval and the standard deviation are shown as black dots and error bars. The horizontal axis has units of pptv.

[36] The decreased correlation between the CIMS and LP-DOAS data at high NO levels likely results from an inhomogeneous spatial distribution of BrO, caused at least in part by the spatial inhomogeneity of NO. Local variations of BrO due to high NO (such as due to generator plumes, etc.) can affect the in situ CIMS measurement more than the LP-DOAS which averages over a few kilometers. This effect is illustrated in Figure 8 (top), where the CIMS observations are much more variable than the LP-DOAS. The rapid decreases in the CIMS observations correspond to elevated NO which impacts BrO levels through the reaction of BrO and NO. In addition, other species such as NO2 that are likely to be present with the NO can also impact BrO. For example at 1006 LST in Figure 8, the NO level is approximately 430 pptv which corresponds to a lifetime of BrO due to reaction with NO of 3 s. The typical lifetime of BrO in this environment under clean conditions is of the order of a minute, primarily due to photolysis. Conversely, the LP-DOAS BrO observations were much less variable during the same time period (Figure 8). Most of the high-NO episodes occurred when the wind speed was lower than 2 m s−1 from the west. This suggests that the high NO might be from local sources such as power generators. These results indicate that the local pollution plumes could impact BrO levels on spatial scales smaller than the DOAS path length.

Figure 8.

(top) An example of a time period with CIMS observations (gray line) and elevated NO levels (black line). The LP-DOAS observations are less variable (black line with cross markers). (bottom) The [BrO]LP-DOAS/[BrO]CIMS values are binned according to NO level. The average [BrO]LP-DOAS/[BrO]CIMS ratio in each bin is shown as the height of the shaded boxes. The standard deviation is shown as the error bar and the number of points in each bin is plotted on the right-hand axis as an open circle.

[37] The effect of high NO on BrO ratios observed by the CIMS and LP-DOAS is further illustrated in Figure 8 (bottom). The LP-DOAS to CIMS ratios (when [BrO] > 1 pptv) are binned according to NO mixing ratios. The average value of [BrO]LP-DOAS/[BrO]CIMS in each bin is shown as the height of the shaded boxes. The average [BrO]LP-DOAS/[BrO]CIMS ratios increase from 1.3 to almost 5 as NO mixing ratios increase from <100 pptv to 2 ppbv. This is consistent with localized pollution plumes decreasing the BrO levels observed by the CIMS. However, most (85%) of the BrO data were observed when NO < 100 pptv, with a median LP-DOAS to CIMS ratio of 1.12.

[38] Low wind speeds (<3 m s−1) and high wind speeds (>8 m s−1) also contributed to the discrepancy between the CIMS and LP-DOAS measurements. These results are consistent with low wind speeds decreasing the homogeneity of the vertical and horizontal spatial distribution of BrO and high wind speeds inducing blowing snow and lower visibility. Moreover, potential bromine heterogeneous reactions on blowing snow surface may also impact the spatial distribution of BrO [Sjostedt et al., 2007; Jones et al., 2009].

4. Conclusion

[39] The observations of BrO by CIMS and LP-DOAS during the OASIS campaign were highly correlated and agreed within their uncertainties (R2 = 0.74, N > 15,000, slope = 1.10, intercept = −0.15 pptv). The agreement between CIMS and LP-DOAS measurements is best (R2 = 0.85, N = 5347, slope = 1.04 and intercept = −0.11 pptv) at low NO concentrations and moderate wind speeds (>3 m s−1 and <8 m s−1). These conditions favor a more homogeneous spatial distribution of BrO which allows the best comparison between the in situ CIMS and the spatially averaged LP-DOAS. The excellent agreement between the CIMS and LP-DOAS measurements demonstrates the capability of both instruments to accurately measure BrO and is extremely important to link future findings by CIMS to previous observations of bromine species using the DOAS technique. The comparison of BrO measured by in situ and long-path instruments indicates that BrO is often distributed homogeneously on spatial scales of up to at least ∼4 km, although local NOx emissions can cause large variability in BrO over small spatial scales. We believe that concerted applications of long-path observations and in situ methods such as CIMS will provide a more complete picture of halogen chemistry.


[40] This work is part of the international multidisciplinary Ocean-Atmosphere-Sea-Ice–Snowpack (OASIS) program and DFG project HALOPOLE (grant FR2497/2-1 AOBJ 544771), and financially supported by NSF grants ATM-0807702 and ARC-0806437 and by the German Research Council. We would like to thank the OASIS campaign organizers and the NCAR shipping department for logistical support. We also thank Ellery Ingall, Arsineh Hecobian, and Michael Nicovich for their help in certifying the emission rate of the halogen permeation tubes. Holger Sihler is funded by the International Max Planck Research School for Atmospheric Chemistry and Physics, Mainz (Germany). Thomas Wagner (MPI for Chemistry, Mainz, Germany) is gratefully acknowledged for fruitful discussions for the DOAS data. The National Center for Atmospheric Research is operated by the University Corporation for Atmospheric Research under the sponsorship of the National Science Foundation.