Journal of Geophysical Research: Atmospheres

Hydroxyl radical and NOx production rates, black carbon concentrations and light-absorbing impurities in snow from field measurements of light penetration and nadir reflectivity of onshore and offshore coastal Alaskan snow

Authors


Abstract

[1] Photolytic production rates of NO, NO2 and OH radicals in snow and the total absorption spectrum due to impurities in snowpack have been calculated for the Ocean-Atmosphere-Sea-Ice-Snowpack (OASIS) campaign during Spring 2009 at Barrow, Alaska. The photolytic production rate and snowpack absorption cross-sections were calculated from measurements of snowpack stratigraphy, light penetration depths (e-folding depths), nadir reflectivity (350–700 nm) and UV broadband atmospheric radiation. Maximum NOx fluxes calculated during the campaign owing to combined nitrate and nitrite photolysis were calculated as 72 nmol m−2 h−1 for the inland snowpack and 44 nmol m−2 h−1 for the snow on sea-ice and snowpack around the Barrow Arctic Research Center (BARC). Depth-integrated photochemical production rates of OH radicals were calculated giving maximum OH depth-integrated production rates of ∼160 nmol m−2 h−1 for the inland snowpack and ∼110–120 nmol m−2 h−1 for the snow around BARC and snow on sea-ice. Light penetration (e-folding) depths at a wavelength of 400 nm measured for snowpack in the vicinity of Barrow and snow on sea-ice are ∼9 cm and 14 cm for snow 15 km inland. Fitting scaled HULIS (HUmic-LIke Substances) and black carbon absorption cross-sections to the determined snow impurity absorption cross-sections show a “humic-like” component to snowpack absorption, with typical concentrations of 1.2–1.5 μgC g−1. Estimates of black carbon concentrations for the four snowpacks are ∼40 to 70 ng g−1 for the terrestrial Arctic snowpacks and ∼90 ng g−1 for snow on sea-ice.

1. Introduction

[2] It has been widely demonstrated that snowpack acts as a photolytic source of gaseous species that can be subsequently released to the atmosphere. Fluxes of NO, NO2 and HONO have been observed from snow cover [Beine et al., 2001, 2002, 2003, 2008; Dibb et al., 2004; Grannas et al., 2007; Honrath et al., 1999, 2000a, 2002; Jones et al., 2000, 2001; Wang et al., 2008], with laboratory studies demonstrating that the source of NOx (NO + NO2) is through the photolysis of nitrate and nitrite [Anastasio and Chu, 2009; Boxe and Saiz-Lopez, 2008; Chu and Anastasio, 2003; Cotter et al., 2003; Couch et al., 2000; Dubowski et al., 2001, 2002; Honrath et al., 2000b]:

display math
display math

Snowpack may be a highly-oxidizing medium through the photoproduction of hydroxyl radicals from the photolysis of hydrogen peroxide (2), of nitrite (3) and nitrate anions (1).

display math
display math

[3] Hydroxyl radicals, OH, are very reactive and will react with snowpack chemical impurities, and may release gaseous products from the snowpack. Hydroxyl radicals are implicated in the production of acetaldehyde, formaldehyde and oxygenated organic compounds in the snow, [Anastasio et al., 2007; Couch et al., 2000; Dassau et al., 2002; Grannas et al., 2004, 2007; Shepson et al., 1996; Wang et al., 2008]. Some of these species are subsequently released to the interstitial air of the snow and to the atmosphere. [Couch et al., 2000; Grannas et al., 2007; Hutterli et al., 2003, 2004; Jacobi et al., 2004; Shepson et al., 1996].

[4] Field measurements of photoformation rates of OH radicals in snow at Summit, Greenland show that photolysis of hydrogen peroxide is the main source, with a small contribution from photolysis of nitrate (reaction (1)) [Anastasio et al., 2007] and nitrite (reaction (3)) [Bock and Jacobi, 2010; Thomas et al., 2011]. Modeling studies have demonstrated that 93–99% of OH radical production in snowpack is likely to be due to hydrogen peroxide photolysis [Anastasio et al., 2007; France et al., 2007]. The importance of OH radicals in snowpack chemistry has been recently demonstrated through chemical box modeling [Bock and Jacobi, 2010; Thomas et al., 2011], showing that the concentration of OH radicals in the quasi-liquid layer surrounding snow grains is a rate controlling factor in the release of Br2 into the interstitial air [Abbatt et al., 2010]; similar reactions also occur in sea-salt particles [e.g., George and Anastasio, 2007]. Sea-ice has also been demonstrated to be a source of gaseous bromine through the reaction of sea-ice with ozone [Oum et al., 1998]. Mixing ratios and measurements of reactive gas-phase halogens have been compared to modeled values with good agreement, giving evidence that emissions from snow covered surfaces may be responsible for gas-phase halogens in the boundary layer [Domine and Shepson, 2002; Thomas et al., 2011].

[5] Photolytic production rates of NO2 and OH radicals within snowpack have been calculated previously for several terrestrial snowpack environments by some of the authors of this work, including Arctic, midlatitude and Antarctic snowpacks [Anastasio et al., 2007; Beine et al., 2006; Chu and Anastasio, 2007; Fisher et al., 2005; France et al., 2007, 2010; King and Simpson, 2001; Simpson et al., 2002], but the Alaskan site of Barrow provides an opportunity to measure optical properties of coastal snow both on land and on sea-ice.

[6] The work described in this paper reports field measurements of e-folding (light penetration) depths and nadir reflectivity [Duggin and Philipson, 1982] of Alaskan snowpacks on land and on sea-ice. Light penetration (e-folding depths) are defined as in (4) below.

display math

where Id is the irradiance at a depth d within the snowpack, d′ is the initial depth into the snowpack (d is a deeper depth than d′), ε(λ) is the asymptotic e-folding depth (the depth at which irradiance becomes 1/e (∼37%) of its initial value) for a specific wavelength. Note, both d and d′ are usually greater than a few cm into the snow to ensure the measured irradiance is diffuse only.

[7] Measurements of nadir reflectance, e-folding depth and snow stratigraphy (including snow density) are used to determine light absorption and scattering cross-sections from snow-atmosphere coupled radiative-transfer calculations. The scattering and absorption cross-sections are used to calculate spherical irradiance (“actinic flux”) in the snowpack as a function of solar zenith angle and depth to allow photolysis rate coefficients of reactions (1) and (2) to be calculated. The variation of the absorption cross-section with wavelength is due to changes in the absorption cross-section of ice and the absorption due to impurities within the snow. The absorption cross-section of ice is well studied [Warren and Brandt, 2008], so the absorption spectrum of light-absorbing impurities can be determined. The variation of absorption cross-section due to impurities of the snowpacks with wavelength is compared to absorption cross-sections of known absorbing species in the snowpack in an attempt to identify and quantify the amount of absorbers.

[8] The measurements of snowpack optical and physical properties made at Barrow, Alaska were part of the larger OASIS (Ocean-Atmosphere-Sea Ice-Snowpack) international field campaign.

[9] The aims of the work presented here were as follows.

[10] 1. Measure the optical properties of Barrow snowpacks by measuring light penetration depth, surface nadir reflectance and the snowpack stratigraphy.

[11] 2. Calculate wavelength-dependent cross-sections for light absorption due to impurities and for scattering from measurements of e-folding depth and snowpack surface reflectivity.

[12] 3. Calculate in-snow production rates of NO, NO2 and OH radicals for the duration of the campaign as an estimation of the potential flux of NO2 or NO from the snowpack.

[13] 4. Determine the identity and amount of light-absorbing impurities in the Alaskan snowpack from the absorption cross-section determined from field measurements of the snowpack.

2. Methods

[14] The majority of the basic fieldwork and modeling methodologies used here have been described in detail previously [e.g., Beine et al., 2006; Fisher et al., 2005; France et al., 2007, 2010, 2011a]. Only a brief description will be given here, with emphasis given to the new and improved sections of the methodology.

2.1. Field Methods

[15] Snowpacks within 1–15 km of the coastal Barrow Arctic Research Centre (BARC) (71.32063°N, 156.6748°W) were investigated during the spring of 2009 as part of the Barrow OASIS (Ocean-Atmosphere-Sea Ice-Snowpack) campaign. Approximately 50 test snowpits were dug, with 15 snowpacks studied optically at locations around the OASIS field site and 4 snowpacks of the 15 (with stratigraphic sequences representative of the different snowpack types) chosen for detailed analysis: (1) a soft snowpack close to the BARC, (2) a hard snowpack close to the BARC, (3) a snowpack 15 km inland, and (4) a snowpack on sea-ice near Point Barrow. Snowpits at these four locations were deeper than average [see Domine et al., 2012] to facilitate optical measurements. Furthermore, as also detailed by Domine et al. [2012], the snowpack stratigraphy was variable at the 1 m horizontal scale.

[16] At each of the 15 sites, a snowpit was dug ∼1 m wide by ∼1 m length, down to the ground or to the sea-ice, ensuring the snow was not contaminated by human or animal influence. At each site, measurements of light penetration depth, surface reflectance and stratigraphy were undertaken as described below.

2.1.1. Light Penetration (e-Folding) Depth

[17] Light-penetration depth was measured via a set of simultaneous irradiance measurements, after the approach of King and Simpson [2001]. Probes consisted of bundles of six fiber optics encased separately in 1/4 inch diameter stainless-steel tubing and with a cosine diffuser on the end of each fiber. The probes were 50 cm long and 49 cm was inserted into the snowpack. Probes were placed horizontally into the snowpack with 5–20 cm horizontal separation between the fibers, and with approximately 3–5 cm vertical separation between different probes, i.e., six irradiance measurements were taken concurrently at different depths in the same snowpack. The detector comprised 6 independent spectrometers (Ocean Optics USB2000) assembled in a single field-portable housing and operated from batteries. The snow pits were not back-filled with excavated material during irradiance measurements for calculation of e-folding depth because this would not reproduce the irradiance field within the snowpit owing to the mechanical properties of the snow. Previously, France et al. [2011a] have crudely argued that only 1% of the diffuse light measured by the deepest fiber optic probes has originated from the snowpit wall with the rest originating from the snow surface. For the shallowest probes the diffuse light measured by the probes from the snowpit wall is significantly less [France et al., 2011a].

[18] The 6-spectrometer instrument is capable of recording spectral irradiance from 190 nm to 1100 nm at a resolution of < 1 nm simultaneously. The optical properties of the fiber optics constrained the effective wavelength range of measurements within snowpack from 350 nm to 700 nm. The 6-spectrometer instrument removes the need to calibrate each individual measurement of in-snow irradiance with a downwelling atmospheric irradiance measurement as the measurements at all depths are effectively simultaneous, in contrast to previous studies [e.g., Beine et al., 2006; Fisher et al., 2005; France et al., 2011b]. Dark spectra were recorded in the field by capping the fiber optic probes to allow measurement (and the removal) of electrical noise. The spectrometers were used at ambient temperature, which significantly reduces the electrical noise. The sensitivity of each fiber and spectrometer were calibrated relative to each other by simultaneously measuring the intensity of the solar radiation above the snowpack with all 6 fiber optic probes pointing at the same target. The determination of an e-folding depth does not require an absolute calibration of the fiber optic probes or spectrometer efficiency, just the relative calibration between fibers and spectrometers. A wavelength calibration for each spectrometer was performed using a mercury-argon lamp in the field and an intensity calibration for each fiber was performed using a NIST traceable halogen light source to monitor fiber optical transmission for any decay. The signal-to-noise ratio of the spectrometers is typically better than 300:1.

2.1.2. Nadir Reflectivity

[19] The reflectance measurements were carried out using a portable nadir reflectance method [Duggin and Philipson, 1982], previously employed in midlatitude and polar environments to measure the reflectivity of the snowpack surface [Beine et al., 2006; Fisher et al., 2005]. The method uses two spectroradiometers (GER 1500 s) mounted upon tripods, one to measure radiance of a reference panel and one to measure the radiance of the snow surface simultaneously. Simultaneous measurement removes any influence of changing overhead sky conditions. The spectroradiometers were calibrated in the field by simultaneous measurement of the reflectance of the standard reference panel with both spectroradiometers before and after measurements of the snowpack. All measurements of snowpack reflectance were taken close to solar noon and during stable sky conditions. At each site, a transect of 10–15 reflectance measurements (within ∼1–10 m) was averaged.

[20] The snowpack stratigraphy was measured using the guidelines from the international classification of seasonal snow [Fierz et al., 2009], with measurements of snow density made every 5 cm using a snow cutter (volume = 284 cm3) and temperature-depth profiles of the snowpack recorded using a NIST traceable temperature probe.

2.2. Radiative-Transfer Modeling and Photolysis Calculations

2.2.1. Determining Absorption and Scattering Coefficients

[21] Wavelength-dependent cross-sections of light scattering, σscatt(λ), and absorption due to impurities, σabs+ (λ), were determined for the four Barrow snowpacks from measurements of light penetration (e-folding depth) and surface reflectance by the method of Lee-Taylor and Madronich [2002]. Briefly, this involves performing radiative-transfer calculations of reflectance and e-folding depths for a range of combinations of σscatt and σabs+, and interpolating to find unique solutions for σscatt (λ) and σabs+ (λ) that satisfy the field measurements of both reflectivity and e-folding depth at each wavelength. Radiative-transfer calculations were performed using the TUV-snow (Tropospheric Ultraviolet and Visible-snow) model [Lee-Taylor and Madronich, 2002], a discrete-ordinates [Stamnes et al., 1988] coupled atmosphere-snow model running 8 streams with a pseudo-spherical correction. The model configuration in the current study used 106 snowpack levels (with 1 mm spacing in the top 0.5 cm and 1 cm spacing for the rest of the 1 m snowpack) and 80 atmospheric levels spaced at 1 km intervals, clear skies, no atmospheric aerosol, an Earth-Sun distance (based upon the day of measurement), and overhead ozone column (from the day of measurement [McPeters et al., 1998]). The snow asymmetry factor, g, was set to 0.89 based on a Mie calculation of a 100 micron particles [Wiscombe and Warren, 1980]. As in previous work we have not adjusted the size of the snow grain. An under-snow albedo of 0.1 was specified. The recommended absorption cross-section for ice from Warren and Brandt [Warren and Brandt, 2008] was used, and linearly interpolated from 200 nm to 400 nm where the absorption is too small to be reliably measured at present.

2.2.2. Modeling Snowpack Absorption By Impurities

[22] Plotting the cross-section owing to absorption by impurities versus wavelength for each snowpack results in the absorption spectrum of all the light-absorbing impurities in the snowpack. The absorption spectrum is a summation of all of the absorbers in the snowpack. The absorbers in the visible wavelengths are probably black carbon, brown carbon (including HULIS) and dust [Anastasio and Robles, 2007; Doherty et al., 2010; France et al., 2011b; Grenfell et al., 2011; Warren, 1984; Warren and Clarke, 1990; Warren et al., 2006]. Previous work assumed that absorption due to impurities was due to black carbon [Lee-Taylor and Madronich, 2002]. However, in the work presented here the range of wavelengths (350–600 nm) studied allows identification of the light-absorbing impurities by their different absorption spectra.

2.2.3. Calculating Photolytic Rate Constants and Fluxes With Clear and Diffuse Skies

[23] The spherical irradiance (“actinic flux”) in the snowpack is needed to calculate the photolytic rate constants for reactions (1)(3). Spherical irradiance was calculated using the TUV-snow model [Lee-Taylor and Madronich, 2002] at 1 nm intervals from 290 to 700 nm and at 106 calculation depths in the 1 m model snowpack, using the wavelength dependent snowpack optical properties σabs+ (λ) and σscatt(λ), determined from field data in section 2.2.2. Snowpacks thicker than 3–4 e-folding depths can be considered optically semi-infinite as over 95% of sunlight is attenuated by 3 e-folding depths [France et al., 2011a]. All the snowpacks studied in the field work were optically semi-infinite. Thus, for modeling photolysis rate constants and spherical irradiance in the snowpack, the actual recorded depth of the snowpack in the field or a larger depth can be considered without affecting the result. In this work all snowpacks are considered to be 1 m deep for comparison. All other conditions are as stated in section 2.2.1.

[24] Photochemical rate coefficients, J, for the photolytic reactions (1)(3) were calculated according to equation (5) for solar zenith angles between 0 and 90° for clear sky and diffuse sky conditions.

display math

where σ is the absorption cross-section of the chromophore (NO2, NO3 or H2O2), Φ is the quantum yield for photolysis, I is the spherical irradiance, T is the snowpack temperature, λ is the wavelength and θ is the solar zenith angle.

[25] The absorption cross-sections and temperature dependent quantum yields for reactions (1), (2) and (3) are from Chu and Anastasio [2003]; Chu and Anastasio [2005] and Chu and Anastasio [2007], respectively. For the calculation of photochemical rate constants, the average campaign overhead column ozone is used [McPeters et al., 1998]; the snowpack location, temperatures and densities are also shown in Table 1.

Table 1. Data Used As Inputs for the Modeling of Photolysis Rate Coefficients Using TUV-Snow for Each of the Barrow Snowpacks
SnowpackSnow Temperature (°C)Snow Density (g cm−3)LocationElevation (m)Average Column Ozonea (Dobson Units)
LongitudeLatitude
  • a

    Ozone conditions determined from the NASA TOMS program as an average of the campaign duration to 2 significant figures [McPeters et al., 1998].

Hard−150.39156.67243°W71.31896°N∼5460
Soft−180.38156.67634°W71.31987°N∼5460
Snow on sea-ice−240.40156.45239°W71.38469°N0460
Inland−280.30156.47471°W71.20259°N∼2460

[26] Photolysis rate coefficients were calculated for clear skies (no cloud or aerosol) and diffuse sky conditions. To obtain diffuse sky conditions a cloud layer 1 km above the ground, 100 m thick, with an optical depth of 16, an asymmetry factor of 0.85 and a single scattering albedo of 0.9999 was included. Depth-integrated production rates, or maximum fluxes (assuming that all the photoproduced NO2 or NO is liberated from the snowpack), F, are calculated using (6).

display math

where z is the depth into the snowpack, and [x] is the concentration of the chromophore, H2O2, NO2 or NO3.

[27] Equation (6) assumes a depth-independent concentration of chromophore. France et al. [2007] showed depth variation of chromophore concentration to be much less important relative to the depth-dependence of the spherical irradiance. A constant nitrate concentration in the snowpack of 3.9 μmol l−1 was used (an average concentration of more than 100 samples including all encountered snow types as shown in Figure 1a [Jacobi et al., 2012]). Measurements of [H2O2] in snowpack during the Barrow campaign were made at ∼500 locations in the Barrow area, with concentrations found to be fairly invariant with depth [Beine et al., 2012]. The calculated depth-integrated production rates of OH radicals used a concentration of H2O2 of 0.4 μmol l−1, the average concentration measured in the top 10 cm of snowpack [Beine et al., 2012]. Concentrations of NO2 were measured at a single site near the BARC over a period of 36 h in surface snow and at depth of 5 cm [Villena et al., 2012]. The average NO2 measured in these samples was 0.02 μmol l−1. The single-site average nitrite concentration is used for all of our snowpack calculations.

Figure 1.

(a) snowpack stratigraphy based on the notation of Fierz et al. [2009], (b) wavelength dependent snowpack nadir reflectance (markers every 10 data points for clarity), (c) wavelength dependent e-folding depths (markers every 10 data points for clarity), (d) snowpack density profiles and (e) snowpack temperature profiles for each characteristic snowpack. For the e-folding depths plotted in Figure 1c, the fiber optic probes were placed between the following depths: hard snow 6 cm to 19 cm, snow on sea-ice 4 cm to 20 cm, soft snow 3 cm to 28 cm and for the inland snowpack the probes were placed at depths between 7 cm and 31 cm.

2.2.4. Calculating Photolytic Rate Constants and Fluxes From the Snowpack for Sky Conditions During the OASIS Campaign

[28] Depth-integrated production rates or fluxes were calculated as a function of solar zenith angle for clear sky or thick cloud cover (i.e., completely diffuse). To calculate the depth-integrated production rates (fluxes) for the duration of the campaign required an identification of sky conditions as “clear sky” or “diffuse” and then a scaling of depth-integrated production rates (calculated from a solar zenith angle dependence) by actual downwelling UV atmospheric irradiance at the surface. The sky conditions were monitored using two continuous measurements: broadband, downwelling irradiance (wavelength 295 nm to 385 nm) using a TUVR (Tropospheric Ultraviolet and Visible Radiation) Eppley flat-plate radiometer sited 0.5 m above the snowpack approximately 800 m South of the BARC, and hemispheric sky images (and retrievals of fractional sky cover for periods when the solar elevation was greater than 10 degrees) using a total sky imager (Yankee Environmental Systems). The total sky imager was based at Atmospheric Radiation Measurement site at the North Slope, Alaska (71° 19′ 23.73” N, 156° 36′ 56.70” W) [Long and DeLuisi, 1998; Long et al., 2001], within a few kilometers of the measurement site and the Eppley TUVR. Every minute of the measurement campaign from Julian Day 60 to Day 90 was assigned as diffuse sky conditions or clear (diffuse and direct), depending on whether the sun's direct beam was occluded by cloud using the total sky imager [Long and DeLuisi, 1998; Long et al., 2001]. The radiative-transfer calculations of section 2.2.3 were repeated to calculate the ‘flat-plate’ broadband downwelling irradiance from 295 nm to 385 nm as measured by a surface TUVR Eppley radiometer for the diffuse and clear sky conditions described in section 2.2.3 as a function of solar zenith angle under exactly the same conditions as the photolysis rate constants were calculated. The calculated photochemical fluxes of NO2 and OH radicals were then scaled using,

display math
display math
display math

to produce photochemical production rates of OH radicals, NO2 and NO at a 1 min resolution for the whole of the Barrow campaign period. In equations (7), (8) and (9), ITUVR measured is the downwelling broadband irradiance for 295–385 nm measured by the Eppley TUVR and ITUVR modeled is the calculated downwelling broadband irradiance for 295–385 nm as a function of solar zenith angle and sky conditions. The quantity F(OH) is the depth-integrated production rate of hydroxyl radicals and F(OH)modeled is the depth-integrated production rate of hydroxyl radicals calculated by the radiative-transfer modeling of section 2.2.3. Note that the values of F(OH)modeled and ITUVR modeled correspond to the same solar zenith angle and sky conditions (i.e., clear or diffuse sky). The linear relationship of F(OH)modeled and ITUVR modeled for diffuse and clear sky conditions is demonstrated in section 3.

3. Results

[29] The results are presented in three sections: field measurements, snowpack optical coefficients determined from the field measurements, and the calculated production rates of NO2, NO and OH radical within the snowpack. Data from the four snowpacks are presented: Soft and hard snows (prevalent around Barrow and BARC), snow on sea-ice, and inland snow (∼ 15 km inland from Barrow).

3.1. Field Results

[30] Measurements of snowpack stratigraphy, snowpack depth, e-folding depth, surface reflectance, snowpack density and temperature are shown in Figure 1 for each of the four individual, representative snowpacks. The snowpack stratigraphy in Figure 1 illustrates the variability between snowpacks, as well as the similarities: in general the stratigraphy featured basal depth hoar, intermediate layers of faceted crystals, and top layers comprised of windpacks and/or recent unsintered wind-drift, often topped by a millimetric diamond dust layer [Domine et al., 2012]. A few very thin melt-freeze crusts were also sometimes observed at several depths. The snow on sea-ice was essentially similar to that on land, but the depth and stratigraphy were much more variable. Melt-freeze layers were also more frequent (this may have been caused by supercooled droplets generated by nearby open leads, which froze on the snow surface). Depth hoar layers in all the snowpacks are formed through snowpack metamorphism due to the presence of a strong temperature gradient and are typical of the snowpack of Alaska's Arctic Coastal Plain [Sturm and Liston, 2003].

[31] The nadir reflectivity of the snowpack across the visible spectrum is similar for three of the snowpacks: the inland snowpack, soft snowpack and the snow on sea-ice. The peak reflectance occurs at a wavelength of 600 nm, a longer wavelength than that of clean snow, as reflectance measurements in central Antarctica suggest peak reflectance in the region of 470–520 nm [France et al., 2011b]. Peak reflectance at 470 nm is seen only in the hard snowpack at Barrow. A shift in maximum albedo has been previously noted by Warren et al. [2006] and mostly attributed to the presence of dust or organic matter. The differences in reflectivity between Barrow snow and the cleaner Antarctic snow [France et al., 2011b] are due to absorbing impurities within the snowpack. The impurities absorb light over the visible and near UV wavelength range. The e-folding depths, reported in Figure 1c, are largest for the inland snowpack, with an e-folding depth of ∼15 cm at a wavelength of 400 nm, and smallest in the soft snowpack, ∼10 cm at 400 nm. The e-folding depths of the snowpacks at the BARC (hard and soft snowpack) and on the sea-ice (only 1–2 km offshore away from the settlement) are essentially the same.

3.2. Optical Coefficients of Barrow Snowpack

[32] The wavelength-dependent absorption and scattering coefficients σscatt(λ) and σabs+ (λ) determined for the Barrow snowpacks are shown in Figures 2 and 3, respectively. Table 2 reports these values relative to previously determined snowpack coefficients [Beine et al., 2006; Fisher et al., 2005; France et al., 2010; Lee-Taylor and Madronich, 2002] at a wavelength of 400 nm.

Figure 2.

Wavelength dependence of the scattering coefficient, σscatt, for each of the four snowpacks investigated at Barrow. All snowpacks demonstrate a broadly invariant relationship of scattering coefficient with wavelength. Determined using the conditions described in Table 1.

Figure 3.

Black Carbon (BC) and HULIS (Humic Like Substances) absorption cross-sections fitted to total snowpack absorbance due to impurities for each of the four Barrow snowpacks. The lines with round markers are absorptions due to impurities in the snowpack derived from field measurements, the dashed line is black carbon absorption and the thick solid line is combined black carbon and HULIS absorption.

Table 2. A Comparison of Optical Coefficients Previously Determined for Arctic and Northern Hemisphere Snowpacks at a Wavelength of 400 nma
StudySnow Descriptionσscatt (m2 kg−1)σabs+ (cm2 kg−1)
  • a

    Note the values of σabs+ and σscatt for the four Barrow snowpacks studied are very similar.

  • b

    Modeling of optical coefficients conducted by Lee-Taylor and Madronich [2002], not in the original measurement study.

Grenfell and Maykut [1977]Arctic Summer dryb6.47.3
Arctic Summer meltingb1.17.8
King and Simpson [2001]Arctic spring windblownb6−304−25
Fisher et al. [2005]Midlatitude windslab - melting11
Midlatitude windslab - dry2−51−2
France et al. [2010]Fresh Ny-Ålesund snowpack16.72.7
Melting Ny-Ålesund snowpack0.819.8
France et al. [2011a]Ny-Ålesund – old windpack9.51.4
Ny-Ålesund – fresh windpack7.75.3
Ny-Ålesund – marine influenced203.4
Ny-Ålesund – glacial accumulation zone25.50.5
This workBarrow – soft snowpack2.011
Barrow – hard snowpack1.711
Barrow – inland snowpack1.89
Barrow – snow on sea-ice1.715

[33] The typical error on the scattering coefficient has been calculated to be ∼1 m2 kg−1, and therefore the scattering coefficient can be considered effectively constant across the wavelength range 350 to 600 nm, justifying the assumption of wavelength-independent σscatt in the original work by Lee-Taylor and Madronich [2002]. The values of σscatt at 400 nm are ∼2 m2 kg−1, and are similar to those for midlatitude dry windslab [Fisher et al., 2005] and both the absorption and scattering coefficients are similar to snowpacks previously determined for summer Alaskan snow on sea-ice in Lee-Taylor and Madronich [2002] using data from Grenfell and Maykut [1977] as shown in Table 2.

[34] The absorption cross-sections of snowpack impurities shown in Figure 3 all demonstrate a general trend of decreasing absorption with increasing wavelength from 350 nm to 550 nm and a smaller increase from 550 nm to 600 nm with the exception of inland snow. The inland snow has a lower absorption cross-section at longer wavelengths than the other snowpacks. Comparison of absorption cross-sections at a wavelength of 400 nm (Table 2) shows that the Barrow snowpacks are highly absorbing when compared to other dry Arctic snowpacks. Lee-Taylor and Madronich [2002] assumed that both σabs+ and σscatt could be modeled as constant with wavelength, however the value of σabs+ is highly variable with wavelength in the UV-visible wavelength range.

3.3. Snowpack Absorption By Impurities

[35] Figure 3 demonstrates that the majority of absorption can be fitted by a linear combination of absorption cross-sections of particulate black carbon and particulate HULIS. The black carbon absorption spectrum was calculated from a Mie calculation assuming spherical black carbon particles in the snow of 0.1 μm radius, 1 g cm−3 density and a refractive index of 1.8–0.5i, as in work by Warren and Wiscombe [1980, 1985]. The black carbon cross-section is shown in Figure 4. The absorption spectrum for the HULIS absorber was taken from Figure 4 of Hoffer et al. [2006]. The linear combination of absorption cross-sections was fitted to the experimental data in Figure 3 by eye. The HULIS and the black carbon absorption spectra from Hoffer et al. [2006] and Warren and Wiscombe [1980] are presented as per unit mass of carbon, thus it is possible to crudely estimate the amount of black carbon and HULIS in the snowpack. The amounts of HULIS and black carbon in the snowpack determined in this work are presented in Table 3. These values are considerably larger than the amounts of impurities measured by chemical extraction in the windpack and windblown snowpacks at Barrow, which were in the range of 2–17 ng g−1 for black carbon, water insoluble organic carbon of 30–200 ng g−1 and 30–360 ng g−1 of dissolved organic carbon (D. Voisin et al., Carbonaceous species and humic-like substances (HULIS) in Arctic snowpack during OASIS field campaign in Barrow, submitted to Journal of Geophysical Research, 2012). These estimates depend upon many factors and are discussed in section 4.4.

Figure 4.

Mass absorption cross-section for black carbon. Filled circles are the absorption cross-section used in the work described here based on a Mie calculation described in the text. The unfilled squares represent values of the black carbon absorption cross-section measured and reviewed by Bond and Bergstrom [2006] or measured by Adler et al. [2010]. The shaded band represents the value recommended by Bond and Bergstrom [2006] for the black carbon mass absorption cross-section. The values used in this work are in agreement with the values reviewed in the literature by Bond and Bergstrom [2006].

Table 3. Estimated Concentrations of Black Carbon and HULIS in the Snowpacks Around Barrow, Derived From Fitting Linear Combinations of Black Carbon and HULIS Absorption Cross-Sections to Absorption Cross-Section of Snowpack Impurities in Figure 3a
Snowpacke-FoldBlack Carbon (ng-C g−1)HULIS (μg-C g−1)
λ = 350 nm (cm)λ = 400 nm (cm)
  • a

    Note that the concentration is per unit mass of carbon in HULIS or black carbon per gram of snow. The concentration of HULIS in the snowpack should be treated as an upper limit and to represent all light-absorbing snowpack impurities. Values of e-folding depth are included for comparison.

Hard710701.2
Soft7.59701.4
Snow on sea-ice7.59901.5
Inland1215411.2

3.4. Photolytic Rate Constants and Fluxes As a Function of Solar Zenith Angle and Snowpack Depth

[36] Photolysis rate coefficients, J, for the photolysis of hydrogen peroxide, nitrate and nitrite within the four Barrow characteristic snowpacks were calculated for 106 depths within each snowpack for 30 separate solar zenith angles between 0 and 90° (equally spaced over Cos θ). Figures 5, 6 and 7 plot contours of equal photolysis rate coefficients for reactions (1), (2) and (3) respectively versus depth and solar zenith angle. The photolysis rate coefficient for the production of OH radicals, NO2 or NO in the snowpack may be interpolated from the plot for any depth in the top 1 m of snow and for any solar zenith angle.

Figure 5.

Photolysis rate coefficients (s−1) as a function of depth and solar zenith angle for the photolysis of H2O2 to produce two hydroxyl radicals (reaction (2)) for each of the four Barrow snowpacks, determined using the conditions described in Table 1 under clear skies.

Figure 6.

Photolysis rate coefficients (s−1) as a function of depth and solar zenith angle for the photolysis of NO3 to NO2 and OH (reaction (1)) for each of the four Barrow snowpacks. Values were determined using the conditions described in Table 1 under clear skies.

Figure 7.

Photolysis rate coefficients (s−1) as a function of depth and solar zenith angle for the photolysis of NO2 to NO and OH (reaction (3)) for each of the four Barrow snowpacks. Values were determined using the conditions described in Table 1 under clear skies.

3.5. Depth-Integrated Production Rates (Fluxes) of NO, NO2 and OH Radicals for the OASIS Field Campaign

[37] Photochemical depth-integrated production rates (fluxes) of OH radicals, NO2 and NO were calculated for each of the four snowpacks for every minute of the OASIS campaign and are presented in Figures 8 and 9. The variation in calculated photolytic production of NO2 between the Barrow snowpacks is less than a factor of 2, with maximum potential fluxes of NO2 on Julian Day 90 of 17 nmol m−2 h−1 for the inland snowpack, 15 nmol m−2 h−1 for the hard snowpack and 13 nmol m−2 h−1 for the soft snow and 14 nmol m−2 h- for the snow on sea-ice. The maximum depth-integrated in-snow production rate of OH radicals for Day 90 is 160 nmol m−2 h−1 for the inland snowpack, 110 nmol m−2 h−1 for the soft snowpack and 120 nmol m−2 h−1 for the snow on sea-ice and hard snowpack. The maximum photolytic production rate of NO from the photolysis of nitrite is 30 nmol m−2 h−1 for the soft snowpack and 55 nmol m−2 h−1 for the inland snowpack and 36 nmol m−2 h−1 snow on sea-ice and 39 nmol m−2 h−1 for the hard snowpack. The maximum in-snow NOx production rate is therefore 72 nmol m−2 h−1 for the inland snowpack and 44 nmol m−2 h−1 for the soft snowpack. The total OH radical production rate is a summation of OH radicals produced through nitrate, hydrogen peroxide and nitrite photolysis according to reactions (1), (2) and (3) respectively.

Figure 8.

Depth-integrated production (maximum fluxes) of NOx and NO2 in the snowpack for the duration of the Barrow OASIS campaign, assuming all in-snow photolytic production of NOx from nitrate and nitrite is liberated from the snowpack. The top line in each graph is total NOx photolytic production rates (NO + NO2) and the lower line is NO2 photolytic production rates for each snowpack; note that the corresponding left and right hand side axes have different ranges. The depth-integrated production rates (maximum fluxes) are calculated with a concentration of 3.9 μmol l−1 of nitrate and 0.02 μmol l−1, an average of snow measurements during the OASIS campaign, with no depth dependence. Concentrations of chromophores are for melted snow.

Figure 9.

Total depth-integrated production rates of OH radicals from the photolysis of hydrogen peroxide, nitrite and nitrate in the snowpack for the duration of the Barrow OASIS campaign. The depth-integrated production rates are calculated with a concentration of 0.4 μmol l−1 of hydrogen peroxide, 0.02 μmol l−1 of nitrite and 3.9 μmol l−1 of nitrate. Concentrations of chromophores are for melted snow.

[38] Scaling the depth-integrated production rate (or flux) of NO2, NO and OH radicals by downwelling TUVR downwelling irradiance (295–385 nm broadband) measurement is only valid while F(NO), F(NO2) or F(OH) is proportional to the downwelling broadband UV irradiance measured by the Eppley TUVR. Figure 10 demonstrates that for diffuse conditions and clear sky conditions between solar angles of 51–90°, there is an approximately linear relationship between depth-integrated production rate of OH radicals, NO2 or NO and downwelling broadband UV irradiance. As a wider point, depth-integrated production rates (fluxes) of NO2, NO and OH radicals can be predicted from the TUVR downwelling irradiance measurements using Figure 10 for the snowpacks and solar zenith angles studied here. Similar relationships may be possible for other snowpacks and the relationships shown here are only for these snowpacks.

Figure 10.

Demonstrating the linear relationship and validity of scaling NO, NO2 and OH depth-integrated production rates of NO, NO2 and OH radicals using broadband UV measurements (295–385 nm). Each point is a solar zenith angle between 90° and 51°, Minimum solar zenith angle was ∼66° during the OASIS campaign. The snowpack was the hard snow using conditions as described in Table 1. Similar relations exist for the other three snowpacks.

4. Discussion

4.1. Comparison With Previous Work

[39] Previous measurements of e-folding depths in snowpack at Barrow gave values of 8 cm and 26 cm for spring and summer respectively [Rowland and Grannas, 2011]. Those results were obtained using a solid-state chemical actinometry method and e-folding depths calculated using only 3 depth measurements of in-snow actinometry with the wavelength peak of the action spectrum at ∼340 nm [Rowland and Grannas, 2011]. The shortest wavelength at which e-folding depths are reported in this work is at 350 nm, with e-folding depths for four Barrow snowpacks of 7–12 cm. Thus the values of measured e-folding depths presented here are consistent with those of Rowland and Grannas [2011] for Spring snowpacks. Previous analysis of the error in determining e-folding depths using a fiber optic probe placed horizontally into a snowpack determined an uncertainty in e-folding depth of ± 20% [France et al., 2011a]. As this study recorded irradiance at 6 depths concurrently rather than at the 4 depths consecutively used in the uncertainty analysis, it is expected that the error of 20% can be considered a very conservative maximum. The three snowpacks within the vicinity of Barrow (hard, soft and snow on sea-ice) can all be effectively described as the same (optically) within uncertainty, whereas the inland snowpack has a 50% longer e-folding depth at a wavelength of 400 nm.

[40] The measured e-folding depths can be converted to liquid equivalent e-folding depths in order to facilitate comparison between different snowpacks [Lee-Taylor and Madronich, 2002; Warren, 1982], using equation (12).

display math

where ρ is the density, ε is the measured e-folding depth and εliq is the liquid equivalent e-folding depth.

[41] The liquid equivalent e-folding depths for the snowpacks in Barrow are in the range 3.8–4.5 cm (for a wavelength of 400 nm) which are a little larger than liquid equivalent e-folding depths previously reported for Arctic snow on sea-ice of 2–3 cm [Grenfell and Maykut, 1977], but are within the large range of liquid equivalent e-folding depths for Northern hemisphere snows (∼1.5 cm [King and Simpson, 2001] to ∼16 cm [Fisher et al., 2005]). The snowpacks near to the Barrow Science Centre and the snowpack on sea-ice have a liquid equivalent e-folding depth of ∼4 cm at a wavelength of 400 nm, which is comparable to springtime measurements made at Ny-Ålesund [France et al., 2011a; Gerland et al., 1999]. Previously it has been stated that 85% of photochemistry occurs in the top 2 e-folding depths of snowpack [King and Simpson, 2001], therefore for the Barrow snowpacks 85% of the photochemistry occurs within the top 14–24 cm of snow cover using e-folding depths measured at 350 nm. It should be noted that the e-folding depths will depend upon the season [France et al., 2010] and the metamorphic history of the snowpack, and that the above comparison does not take this into account.

[42] The nadir reflectivity of the snowpacks at Barrow was relatively consistent between snowpacks, between 0.82 and 0.85 at 400 nm, and between 0.84 and 0.91 at 500 nm. The reflectivity measurements typically have an uncertainty of 0.04 (two standard deviations), with the variation at each site likely due to changes in surface topography and localized impurities. Previous work investigating albedo uncertainty suggests that a 2° slope can lead to a 10% change in albedo depending upon illumination angle [Grenfell et al., 1994]. Previous analysis of reflectance of dry snow at Barrow recorded a value of 0.92 at a wavelength of 400 nm [Grenfell and Maykut, 1977] and clean Antarctic snow a value of 0.98 at a wavelength of 400 nm [France et al., 2011b]. As discussed in the next section, the low reflectivity of the Barrow snowpacks is attributable to highly absorbing black carbon and humic material within and on the snowpack.

4.2. The Effect of Grain Size on the Determination of σscatt or σabs+

[43] In previous publications by the main (U.K) authors of this work the effect of snow grain size on derived values of σscatt or σabs+ has not been considered in a systematic manner because the values of σscatt and σabs+, along with the asymmetry parameter, g, were used to calculate irradiances and photolytic rate constants in the snow. In the work described here the absorption cross-section of the impurities in the snowpack are derived and it is prudent to assess whether the grain size of the snow affects the value of the absorption cross-section derived for the impurities in the snowpack. As described by Lee-Taylor and Madronich [2002], the only grain-size-dependent quantity used in the radiative-transfer calculation that is not empirically fitted to the measured albedo and e-folding depth (values of σscatt and σabs+ are empirically fitted) is the asymmetry parameter, g. Inspection of the Mie calculation for 100–2000 μm diameter ice spheres by Wiscombe and Warren [1980, Figure 4] demonstrates that the value of the asymmetry parameter may be bracketed by values between 0.885 and 0.895 for the wavelengths considered in the work presented here. As a sensitivity study of the effect of the asymmetry parameter on the values of σscatt or σabs+ the radiative-transfer calculations to empirically fit σscatt and σabs+ for the hard snowpack described in Figure 1 were repeated with values of the asymmetry parameter, g = 0.880, 0.885, 0.890, 0.895 and 0.900. The results are displayed in Table 4 for solar wavelengths of 400, 500 and 600 nm. Table 4 demonstrates the values of σscatt and σabs+ derived are insensitive to the value of asymmetry parameter.

Table 4. Values of σscatt and σabs+, Calculated Empirically By Fitting the Reflectance, and e-Folding Depth for the Hard Snowpack Using Different Values of the Asymmetry Factor ga
Asymmetry Parameter, gλ = 400 nmλ = 500 nmλ = 600 nm
σscatt (m2 kg−1)σabs+ (cm2 kg−1)σscatt (m2 kg−1)σabs+ (cm2 kg−1)σscatt (m2 kg−1)σabs+ (cm2 kg−1)
  • a

    Note that the values of σscatt and σabs+, derived are insensitive to the value of g. Values are reported for the solar wavelengths of 400, 500 and 600 nm.

0.8801.511.01.79.52.514.5
0.8851.611.21.89.72.714.5
0.8901.711.51.89.52.714.5
0.8951.811.51.99.52.914.5
0.9001.811.72.09.53.014.5

[44] As a further test of the procedure to determine values of σscatt and demonstrate that σscatt is sensitive to the grain size of the snowpack the authors used the procedure outlined in section 2.2.1 to fit σscatt to the albedos (calculated by radiative-transfer) contained in Figure 7 of Wiscombe and Warren [1980]. Figure 7 of Wiscombe and Warren [1980] contains semi-infinite direct beam albedo data as a function of wavelength, grain size (100 and 1000 μm) and black carbon content (50, 500 and 500 ng g−1). Values of σscatt were determined for solar wavelengths of 300, 325, 350, 375, 400, 450 and 500 nm. Values of σscatt were found to be 20 ± 3 m2 kg−1 for the small grained snowpack (for all black carbon concentrations) and 1.9 ± 0.1 m2 kg−1 for the large grained snowpack (for all black carbon concentrations) over the wavelengths 300–500 nm.

[45] In previous work by France et al. [2010], stratigraphic snowpack data was used to calculate the irradiance and photolytic rate coefficients in separate windpack layers in an Antarctic snowpack. The different windpack layers had slightly different optical properties in the radiative-transfer calculations owing to slightly different grain size and light-absorbing impurity content. For windpack layers similar to those in Figure 1 the very slight change to irradiance-depth profiles calculated from the radiative-transfer calculations was not worth the increase in substantial computational effort. The typical uncertainty in σscatt owing to a 5% change in snowpack density is ± 5%7% and the typical uncertainty in σabs+, black carbon or HULIS owing to a 5% change in snowpack density is ± 9%6%.

4.3. The Scattering Cross-Sections of Barrow Snowpack

[46] The values of the scattering cross-section may be considered to be smaller than expected, but a sensitivity analysis of the radiative-transfer modeling process to determine absorption and scattering cross-sections yielded no large changes in the values of σscatt or σabs+ for small changes in albedo, e-folding depth or the asymmetry parameter, g. An expectation of greater values of σscatt(λ) is due to a possible relationship between specific surface area of snow and scattering cross-sections [Domine et al., 2008]. Domine et al. [2008] state there is a mathematical relationship between snow specific surface area (SSA) and the scattering cross-section, (σscatt) and Kokhanovsky and Zege [2004, p. 1594, equation (22)] describe this relationship. Similar values of scattering cross-sections were recorded for coastal snowpacks similar to Barrow in Antarctica [Beine et al., 2006]. Domine et al. [2012] have derived values of SSA from snowpack reflectivity in the near IR of ∼30–40 m2 kg−1 for surface snow, which would suggest scattering cross-sections of 15–20 m2 kg−1. However, the results in Figure 2 and Table 2 demonstrate typical values of σscatt ∼ 2 m2 kg−1 were derived from the measurements of e-folding depth and surface reflectance, and a re-investigation of the modeling process to calculate σscatt(λ) revealed no errors or processes that could cause a large change in σscatt(λ) for a small change in the modeling parameters. An almost identical study using identical techniques by the same authors at DOME C in Antarctica found values of SSA of ∼31 m2 kg−1 [Gallet et al., 2010], which would suggest σscatt values of ∼15–16 m2 kg−1, and values of σscatt of 14–24 m2 kg−1 were determined for surface snows [France et al., 2011b]. Thus the relationship between SSA and σscatt appears to be valid for snow at Dome C, but not valid for Barrow. One obvious difference between these two studies is the amount of light-absorbing impurities in the snowpack as the snowpacks studied at Barrow during this campaign are very dirty relative to the very clean snowpacks measured at DOME C. Scattering and absorption are independent quantities.

[47] It is not possible to drastically alter the value of σscatt and replicate the measured e-folding depth and reflectivity measurements by (1) changing σabs+ by 20%, (2) varying the asymmetry parameter, g, within the limits suggested by Wiscombe and Warren [1980] or (3) varying nadir reflectivity or e-folding depth (i.e., measurement error) by amounts representing experimental error. An explanation that is occasionally proposed is the measured reflectivity and measured e-folding depth are for different snow layers with very different optical properties, as the measurements of e-folding depth and nadir reflectance are in different parts of the snowpack. Thus, it may be possible to have a thin (<1 cm) top layer of snowpack containing all the light-absorbing impurities, underlain by a clean snowpack. Experience and exploratory calculations with TUV-snow suggest such a condition would be obvious to the naked eye as a dirty top layer of a different color to the rest of the snowpack. No such layer was observed at Barrow and all 15 snowpits studied had similar values of σscatt (1.7–4 m2 kg−1 (λ = 400 nm)). To surmise, there is no reason to suspect that the method of determining σscatt is not robust. At the present time an explanation for the disagreement between SSA and σscatt for the snowpacks presented here is not available.

4.4. The Wavelength Dependence of Absorption in Barrow Snowpack

[48] The absorption spectrum of light-absorbing impurities in the snowpack is plotted in Figure 3. The absorption cross-section represents the total absorption of light-absorbing impurities whether they are internal or external to the snow grains. External light-absorbing impurities include particles such as soil or black carbon that were trapped or deposited since snow fall. An internal light-absorbing impurity is likely to have been part of the original snow fall or has been incorporated into the snow grain as a deposited gas or during snow metamorphism. Figure 4 also compares the mass absorption cross-section used in this and previous work with measurements of the mass absorption coefficients found in Table 6 of the review by Bond and Bergstrom [2006], and in work by Adler et al. [2010]. Figure 4 demonstrates the values of the mass absorption cross-section from Bond and Bergstrom [2006], and Adler et al. [2010] are similar with values used in this study. The very interesting result of this work is that a HULIS absorber and a black carbon absorber are required to explain the total snowpack impurity absorption, not just black carbon. In future it may not be possible to model UV-visible photolytic processes in the snowpack without considering HULIS and black carbon absorptions.

[49] Comparison of the four snowpacks in Figure 3 demonstrates that the coastal snowpacks have an absorption in the wavelength region 550–600 nm that is not present at the inland site and not accounted for by the HULIS or black carbon absorption spectrum used in this work [Hoffer et al., 2006]. Thus, for the three coastal snowpacks a third light-absorbing impurity may be required. Marine microbiology may be responsible for this absorption and a similar feature has been noted in the extracted HULIS spectrum by Voisin et al. (submitted manuscript, 2012) and melted snow samples by Beine et al. [2012]. The feature is not present in the inland snowpack and only sites close to the open lead at Barrow have this absorption feature. Continued investigation is underway, but at present a realistic absorption spectra for marine microbiological detritus in snowpack does not exist to the authors' knowledge. The amounts of black carbon predicted for Barrow snowpacks are much greater than in more remote regions of the Arctic, where average black carbon concentrations range from 3 ng g−1 (Greenland) to 26 ng g−1 (West Russia), with an Alaskan snowpack average of 9 ng g−1 [Doherty et al., 2010]. However, the derived absorption and scattering coefficients from the work presented here are similar to previously determined values for Alaskan snow on sea-ice from Lee-Taylor and Madronich [2002] using data from Grenfell and Maykut [1977]. Doherty et al. [2010] extracted and quantified black carbon in Barrow snow in April 2007 within 10 km of the snowpacks studied and shown in Figure 1. Lyapustin et al. [2010] have a photograph and reflectivity data of the snow sampled by Doherty et al. [2010] and describe it as “fresh snow with minimal redistribution by the wind.” The new unworked snow sampled by Doherty et al. [2010] is very different to the old windpacked snow described in this work in Figure 1. The windpacked snows in Figure 1 were characteristic of the OASIS campaign and have clearly had the opportunity to accrue more light-absorbing impurities through multiple wind events relative to the Doherty et al. [2010] sample. Doherty et al. [2010] highlight that they ignored samples from the lower 40% of some snowpacks to avoid biasing their samples with windblown soil. Such sampling measures were not possible or desired in the study presented here as the main aim of this work was to measure and model the optical properties of the Barrow snowpack during the OASIS campaign to allow photochemical production rates to be calculated. The initial aim was not to measure black carbon concentrations in the snowpack. Thus a comparison of the black carbon concentrations between snowpacks of Doherty et al. [2010] and those in Figure 1 is not sensible. The calculated amounts of black carbon in Barrow snow are comparable to values measured in East Arctic Russian snow (∼10 to 150 ng g−1), where sampling was restricted to within 100 km of cities and local sources of pollution could have become incorporated into the snowpack [Doherty et al., 2010]. Snow machine traffic and the local town pollution could be the sources of the high amount of black carbon in the Barrow area.

[50] Voisin et al. (submitted manuscript, 2012) reported concentrations of various brown and black carbon species in snow for some of the snowpacks studied in this work. The values reported here are considerably higher than those of Voisin et al. (submitted manuscript, 2012) where the carbon concentrations in the snowpack were determined by extracting carbon from the snowpack using various methods (e.g., SPE cartridge, filtering etc.) and measuring carbon content chemically. However, the values reported here are reported optically after making assumptions about the absorber identity. The majority of the disagreement between Voisin et al. (submitted manuscript, 2012) and the results shown in Table 3 may be owing to our simple approximations for a HULIS absorber and black carbon absorber in the snowpack. Bohren [1986] demonstrated that reasonable uncertainties in the shape, refractive index, and internal / external nature of black carbon with respect to the snow grain could yield changes in the black carbon absorption cross-section of factors of 2.2, 5 and 1.4 respectively, Using the refractive index of HULIS measured by Hoffer et al. [2006] and the mathematical approach of Bohren [1986] results in uncertainties in the HULIS absorption cross-section of factors of 1.8 and 1.3 for the shape and internal / external nature of the HULIS particles respectively. The uncertainty owing to refractive index of HULIS was not considered, as there is more uncertainty in the identity of the UV absorber than the value of its refractive index. There may be a small measure of disagreement between Voisin et al. (submitted manuscript, 2012) and values presented in Table 3 owing to how HULIS and BC are measured and defined. It is therefore important to understand what Figure 3 demonstrates: (1) that the total snowpack light-absorbing impurity is not constant over 350–600 nm, (2) that the light-absorbing impurities in snowpack can be fitted to a combination of HULIS and black carbon absorber and a third absorber based on marine microbiological detritus may be required, and (3) estimates of concentration of black carbon and HULIS in snow can be calculated from optical measurements, but these values depend upon the physical characteristics and location of the absorber. It should again be noted that the absorption by snowpack impurities has been fitted to a linear combination of black carbon and HULIS absorption spectra to demonstrate that the absorption is consistent with a mixture of light-absorbing snowpack impurities. It is quite possible that the material termed HULIS could be replaced by other forms of brown carbon, dust or other light-absorbing impurities. The concentration of HULIS in the snowpack should be treated as an upper limit and to represent all light-absorbing snowpack impurities. One advantage of the technique presented here is that the total absorption of light-absorbing impurities in the snowpack is determined without melting or significantly perturbing the snowpack.

4.5. NOx and OH Radical in-Snow Production Rates (Fluxes)

[51] The combined depth-integrated production rate of NO2 and NO is the upper-bound for the NOx flux from the snowpack, assuming all photoproduced NOx is able to escape from the snowpack. The maximum calculated noon-time NOx flux from the inland snow at Barrow for a solar zenith angle of 66° is 72 nmol m−2 h−1, with 30% and 70% from photolysis of nitrate and nitrite, respectively. The uncertainty in the depth-integrated production rates is approximately 20% [France et al., 2010].

[52] The photolysis of NO2 to produce NO is not usually considered as a significant source of NOx, but in the conditions at Barrow nitrite appears to dominate, contributing approximately 3 times more NOx than nitrate. Previously, analysis by Chu and Anastasio [2007] demonstrate that for OH production on ice grains (and therefore applicable to NOx in-snow production (reactions (1) and (3)), nitrate and nitrite photoproduction rates are comparable. A similar result was obtained for Antarctic snow during the CHABLIS campaign [Jones et al., 2001]. However, under the conditions in Barrow nitrite photolysis is favored relative to nitrate due to: (1) the large value of the product image relative to the value of image (2) the presence of large ozone column, which attenuates the shorter UV wavelengths relative to longer UV wavelengths. The maximum of the action spectrum (product of absorption cross-section and quantum yield) for nitrate photolysis is ∼302 nm, whereas the maximum of the action spectrum of nitrite photolysis is ∼355 nm. (3) the presence of a wavelength-dependent absorber in the snowpack, with increasing absorption at shorter wavelengths. Therefore, if the impact of nitrite photochemistry to form NO in the snowpack is also considered and added to the NO2 production rate to give a total NOx in-snow production rate, then the total in-snow production for the Barrow snowpacks of 44–72 nmol m−2 h−1 is larger than the maximum flux of 40 nmol m−2 h−1 NOx measured at Alert, Canada for a solar zenith angle of ∼66° [Beine et al., 2002]. The depth-integrated production rate of NO is dependent upon the assumption that the single site measurement of nitrite concentration in the snowpack is valid across the Barrow area.

[53] Snowpack emissions of NOx have previously been demonstrated to have an impact on the oxidative capacity of the lower troposphere [Bloss et al., 2007; Morin et al., 2008; Wang et al., 2008; Yang et al., 2002]. The depth-integrated production rates of NO2 presented here can only be considered to be maximum fluxes because some of the photoproduced NO2 in the snowpack is likely to be involved in some secondary chemistry within the snow matrix. It has previously been suggested that 30% of the NO2 is converted prior to release from the snowpack [Anastasio and Chu, 2009]. The release of NO2 from the snow cover to the atmosphere is likely to be at least partly temperature controlled [Boxe et al., 2006], but as sunlight-dependent fluxes of NOx have been already observed over colder snowpacks in Alert [Beine et al., 2002], it is unlikely that the NOx photoproduced at Barrow in the warmer snowpacks will be trapped within the snow microstructure to a large degree. The mechanism for movement of photoproduced NOx from snowpack to the atmosphere appears to be mostly influenced by windpumping at the surface of the snowpack and gas diffusion deeper into the snowpack [Thomas et al., 2011]. The complexity of the nitrate photochemical system in snow is discussed by Bock and Jacobi [2010].

[54] The increased black carbon (and HULIS) absorption in the coastal snowpacks relative to inland snow reduces the predicted amount of photolytic NOx production in the coastal snowpack by a factor of ∼1.7 relative to the cleaner inland snowpack (Figure 8). The effect of increasing black carbon concentration upon the Alaskan snowpack photochemistry is explored in detail by H. J. Reay at al. (Decreased albedo, light penetration depth and photolytical production of OH radicals and NO2 in Barrow snowpack: A scenario of increasing black carbon, submitted to Journal of Geophysical Research, 2012).

[55] The calculated depth-integrated production rates of total OH radical production from the photolysis of H2O2, NO3 or NO2IO in Barrow snowpacks are shown in Figure 9, demonstrating that the variation in snowpack absorption and scattering cross-sections between the Barrow snowpacks only causes a small variation of the in-snow photochemical production of OH by a factor of less than 2. The in-snow photochemical production of OH radicals may well be a driving factor in the formation and release from the snowpack of organic compounds that appear to have a photochemical source [Anastasio et al., 2007; Hutterli et al., 2004; Sumner et al., 2002]. The importance of OH radical production with respect to halogen release from snowpack is demonstrated through the dynamically coupled atmospheric-snow modeling performed by Thomas et al. [2011]. The relative contributions of OH production by H2O2, NO3 or NO2 are approximately 60%, 4% and 36%, respectively. It was previously calculated that nitrite and nitrate photolysis contributed a similar amount of OH radicals to the snowpack inventory [Chu and Anastasio, 2007], but the snowpack conditions in Barrow with large nitrite concentrations and relatively low H2O2 concentrations favor the production of OH radicals from nitrite (relative to nitrate) compared to previous calculations, but still a factor of ∼2 smaller than H2O2.

5. Conclusions

[56] The investigations into the optical properties of the snowpacks at Barrow and subsequent in-snow photochemical modeling have allowed a number of important conclusions to be drawn from the work:

[57] 1. It is important to accurately account for spectrally resolved absorption by both black carbon and non-black carbon impurities in the snowpack because non-black carbon impurities have a large absorption cross-section at the short solar wavelengths, responsible for photolytic reactions in the snowpack. The relative importance of the photolysis of nitrite (versus nitrate) as a source of NOx from the snowpack is increased when absorption of short wavelength solar radiation by non-black carbon impurities is considered.

[58] 2. Estimates of depth-integrated production rates can be scaled and approximately correlated with downwelling UV irradiance for individual snowpacks.

[59] 3. The importance of NO fluxes from the snowpack owing to nitrite photolysis may have been significantly overlooked during previous campaigns. From the calculations in this work, NO from nitrite photolysis is approximately three times larger than NO2 from nitrate photolysis. The contribution of nitrite to OH radical production in the snowpack is approximately half that of hydrogen peroxide.

Acknowledgments

[60] This work is part of the international multidisciplinary OASIS (Ocean-Atmosphere-Sea Ice-Snowpack) program. J.L.F. and M.D.K. thank NERC for support through grants NE/F010788/1 and NE/F004796/1, NERC FSF for support through grant 555.0608 and the RHUL research strategy fund. Funding for this work was also gratefully received from NSF grant ATM-0807702. H.J.R. thanks RHUL for support through the Thomas Holloway Scholarship scheme. D.V., H.W.J. and F.D. acknowledge financial support by the LEFE-CHAT program of CNRS-INSU. The participation of LGGE was funded by the French Polar Institute (IPEV) grant 1017 to F.D. The National Center for Atmospheric Research is sponsored by the National Science Foundation. All sky camera data were obtained from the Atmospheric Radiation Measurement (ARM) Program sponsored by the U.S. Department of Energy, Office of Science, Office of Biological and Environmental Research, Climate and Environmental Sciences Division.