While laboratory experiments suggest that hydrogen peroxide (HOOH) on polar snow grains should be completely photolyzed during summer, field measurements show a continuous, multiyear record of HOOH in the snowpack. To understand this discrepancy, we apply a snow parcel model to follow HOOH in the snowpack at Summit, Greenland. Taking into account snowfall and variations in actinic flux, we evaluate the impact of three photochemical factors on HOOH preservation: the quantum efficiency of HOOH photodestruction, HOOH recycling through organic compounds, and HOOH photoproduction from snowpack chromophores. We find that 60%–100% of deposited HOOH is preserved in the first year, with less preservation of HOOH deposited during summer, and that the OH produced from HOOH photolysis likely contributes strongly to transforming snowpack bromide and organic compounds. Our findings suggest that photochemistry plays an important role in HOOH loss in snows at Summit and other polar sites, complementing temperature-dependent physical processes.
 Hydrogen peroxide (HOOH) is an important constituent in polar snow and ice, where it is a tracer for the oxidative capacity of past atmospheres [Hutterli et al., 2003; Legrand and Mayewski, 1997]. Field measurements have found a continuous record of HOOH from the surface snow into the firn and glacial ice [Legrand and Mayewski, 1997; Sigg and Neftel, 1991], indicating that snow grain HOOH is stable enough to last until it is buried to beneath the photic zone by new snowfall. Because HOOH in newly deposited snow is generally supersaturated with respect to the atmospheric boundary layer, thermally driven snow-to-air transfer is an important loss mechanism for snowpack HOOH that moves snow grains toward air-ice equilibrium with a time scale on the order of weeks or more [Bales et al., 1995]. However, photochemistry might also be an important loss mechanism. For example, at Summit, Greenland, the photic zone for HOOH is approximately 27 cm (defined as 2 times the e-folding depth of j(HOOH) [Galbavy et al., 2007]), which allows for 5 months of photochemistry before snow grain HOOH is buried beneath the photic zone, based on a typical snow accumulation of 5 cm month−1 [Steffen and Box, 2001]. Laboratory experiments indicate that the photolytic lifetime of HOOH in polar surface snow during the summer should be on the order of weeks to a few months [Chu and Anastasio, 2005], suggesting that HOOH in summer surface snow at Summit should be completely destroyed.
 How can we reconcile the discrepancy between this short calculated photolytic lifetime for HOOH and field measurements of HOOH in snow beneath the photic zone? Several factors may extend the photolytic lifetime of snow grain HOOH to beyond what is predicted from laboratory studies. First, snowpack HOOH is primarily found in bulk ice [Jacobi et al., 2002], where quantum efficiencies for HOOH photodegradation are approximately 2 times lower than previously determined values for HOOH in liquid-like regions (LLRs) [Beine and Anastasio, 2011; Chu and Anastasio, 2005]. Second, photolysis of HOOH is the major source of snow grain hydroxyl radical (OH) [Chu and Anastasio, 2005], which can react with organic compounds in ice to recycle HOOH [Hullar and Anastasio, 2011]. Third, illumination of light-absorbing compounds (chromophores) in snow and ice can produce HOOH [Hullar et al., 2012]. However, none of these explanations alone sufficiently extends the photolytic lifetime of HOOH to match field observations.
 Our goal in this current work is to examine whether a combination of these photochemical factors might explain the persistence of HOOH in the field. We track HOOH in deposited snow using a condensed-phase snow parcel model that includes the three photochemical factors as well as diel and seasonal changes in actinic flux and the attenuation of light by snowfall. While we use Summit conditions as the baseline for our model, we also perform sensitivity studies to examine conditions for other polar sites.
 Our snow parcel model follows a hypothetical volume of snow (a “snow parcel”) and monitors its condensed-phase HOOH concentration through time. In order to focus on the effect of photochemistry on snow-phase HOOH, the model does not include air-snow transfer of HOOH. Below we describe the major points of our model; additional details are in the supporting information (Figures S1–S9 and sections S1–S7).
 Using the Tropospheric Ultraviolet-Visible Model [Madronich and Flocke, 1998], we calculated actinic fluxes for conditions at Summit, Greenland, over a 26 week period beginning with 21 June 2011 (section S2). Figure S8 shows the diel and seasonal variation in actinic flux. For our baseline simulation, we assume that snow falls continuously at the rate of 60 cm yr−1, without seasonal variation [Steffen and Box, 2001], but we also examine variations in accumulation rate and the effect of snowfall seasonality (section S3).
 Actinic flux, which decreases with depth in the snowpack, is calculated using snow depth, surface actinic flux, and e-folding depth (section S3). While the e-folding depth depends on grain size, shape, and snow impurities [Domine et al., 2008], we have used a baseline e-folding depth of 13.3 cm measured for j(HOOH) at Summit [Galbavy et al., 2007] and run additional simulations at other e-folding depths to account for variations in snowpack density and light absorption.
 We determined the rate constant for HOOH photolysis, j(HOOH)z, for each hour in the model using the appropriate temperature (T) and depth (z) of the snow parcel using
where NA is Avogadro's number, Iz,λ is the actinic flux at depth z, ΦT is the quantum efficiency for HOOH photolysis, and εT,λ is the molar absorptivity for HOOH (section S4). As described in section S4 of the supporting information, we used three different values for the HOOH quantum efficiency, referred to by their values at 258 K: Φ258 = 0.34 for HOOH in liquid-like regions of ice [Chu and Anastasio, 2005] and 0.19 or 0.15 in bulk ice [Beine and Anastasio, 2011]. We estimated the temperature at depth in the snowpack by using 16 years of hourly surface temperature data for Summit [Steffen et al., 1996] and modeling the snowpack as a conductive half-space (section S5).
 The second photochemical factor we examined is the reaction of photo-produced OH with organics to regenerate HOOH [Hullar and Anastasio, 2011]. Including this recycling, the rate of photolytic destruction of HOOH, Rd(HOOH)′, is
where RE is the recycling efficiency (i.e., 2 times the HOOH yield from the reaction of OH with an organic compound [Hullar and Anastasio, 2011]). Here we used a typical ice RE value of 0.10 and a maximum value of 0.24 [Hullar and Anastasio, 2011]; because the temperature dependence of RE is unknown, we did not adjust RE for temperature.
 The final photochemical factor we examined is the photoformation of HOOH via illumination of snowpack chromophores. The average rate of formation of HOOH (Rf(HOOH)) in illuminated polar snows (normalized to surface snow under summer solstice conditions at Summit, Greenland, at –5°C) is 5.3 ± 5.0 nM h−1, while the upper bound is 12 nM h−1 [Hullar et al., 2012]. We assumed that the photoformation rate of HOOH in the snowpack scales linearly with the amount of light, i.e., we scaled Rf(HOOH) to the actinic flux in the snowpack. We do not adjust Rf(HOOH) with T since its temperature dependence has not been measured. The net rate of change of HOOH concentration [R(HOOH)] in the model is calculated by
Based on measured and modeled HOOH concentrations at Summit [Anastasio et al., 2007; Hutterli et al., 2003], we set the initial HOOH concentration in fresh surface snow to be 10, 10, 3.0, and 5.0 μM for summer, fall, winter, and spring, respectively.
 Figure 1a shows the midday HOOH photolysis rate constants in snow newly deposited at the surface on 21 June [j(HOOH)0] and as it becomes buried in the snowpack on subsequent days over the course of a year [j(HOOH)z]. j(HOOH)0 does not exactly follow the seasonal variation of actinic flux but instead forms a slightly asymmetric curve because of seasonal variations in temperature (section S5). Due to snowfall burying our modeled snow parcel, the ratio j(HOOH)z/j(HOOH)0 drops to 0.5 after approximately 60 days (when z = 10 cm) and is 0.1 at approximately 180 days (z = 30 cm); both rate constants approach zero in the winter, with j(HOOH)0 reaching a minimum of 1.3 × 10−11 s−1 on day 182. During the following spring, j(HOOH)z in our snow parcel rises slightly but remains very low due to the accumulated snow cover.
 Figure 1b shows the noontime rate of formation of HOOH [Rf(HOOH)], the recycling-corrected rate of HOOH photolysis (Rd(HOOH)′), and the net rate of HOOH formation [R(HOOH)] in our snow parcel beginning with its deposition on 21 June. Both Rf(HOOH) and Rd(HOOH)′ decrease with time because of the decreasing amount of sunlight in the snow parcel, a result of less actinic flux at the surface and more light attenuation by the accumulating snowpack above the snow parcel. The net rate of HOOH formation on snow grains, i.e., the sum of Rf(HOOH) and Rd(HOOH)′, is always negative here, showing that photochemistry is a net sink for HOOH in this modeled Summit snowpack. Because the three rates are light dependent, they all fall to essentially zero during the winter.
 Figure 2 shows the sensitivity of HOOH concentrations to individual photochemistry variables in the simulation relative to the baseline values. As seen in Figure 2a, varying the quantum efficiency for HOOH photolysis (Φ) within the range of experimentally measured values significantly affects HOOH preservation. Under the baseline conditions (Φ258 = 0.19, the laboratory value for HOOH in bulk ice), 59% of the HOOH deposited at the surface in June survives until the winter, while slightly more (68%) of HOOH is preserved if we use the laboratory quantum yield determined in bulk ice after several hours of illumination (Φ258 = 0.15). In contrast, only 33% of the initial HOOH is preserved if we use the quantum yield determined for HOOH photolysis in liquid-like regions of ice (Φ258 = 0.34).
 Figure 2b shows the impact of HOOH recycling. Changing the assumed recycling efficiency within the range of measured values [Hullar and Anastasio, 2011] results in only small changes in HOOH concentrations: Compared to the baseline value (RE = 0.10, which results in 59% of HOOH being preserved), assuming no recycling (RE = 0) results in 54% of HOOH being preserved, while the maximum RE value (0.24) only increases the preserved HOOH to 66%. Figure 2c shows that Rf(HOOH) has a stronger impact on HOOH preservation. The amount of HOOH preserved ranges from 41% with no photoformation, 59% with the baseline photoformation rate [Rf(HOOH)0 = 5.3 nM h−1], and 81% for the upper bound of HOOH photoformation (12 nM h−1). Based on these results, snowpack HOOH levels are sensitive to Rf(HOOH) and Φ but relatively insensitive to RE.
 So that our results can be roughly applied to sites other than Summit, in Figures 2d–2f, we examine the sensitivity of HOOH preservation on accumulation rate, e-folding depth, and snowfall seasonality. Figure 2d shows that a fivefold decrease in snowfall slightly decreases HOOH preservation, while a fivefold increase significantly raises the preservation percentage. Figure 2e shows that large differences in e-folding depth (efd), which is dependent upon snow density and impurity levels, have only a minor effect on the preservation percentage for snow deposited in summer. Finally, Figure 2f shows that a possible seasonal variation in snowfall rates at Summit has a minimal effect on preservation. We have also examined a simple, quasi-Antarctic scenario; as described in section S7, the results suggest that HOOH preservation in Antarctic snows is roughly similar to that at Summit.
 We turn now to the impact of model start date. Figure 3 shows simulation results when the initial snow parcel is deposited in summer, fall, winter, and spring. The 21 June starting date shows the widest variation of preservation fraction for different values of Φ, probably because this scenario exposes the parcel to the most light. The percentages of HOOH preserved 1 year after deposition are 33%, 59%, and 68% for the Φ258 = 0.34, 0.19, and 0.15 scenarios, respectively. The simulation beginning in the fall (21 September) first shows a small initial loss of HOOH, then a period of constant HOOH during the dark winter, followed by a small amount of additional HOOH loss during the subsequent spring and summer. In this case, 80% of HOOH is preserved after 1 year in the most photochemically active scenario (Φ258 = 0.34), while 91% and 92% are preserved for Φ258 = 0.19 and 0.15, respectively. The scenario starting on 21 December shows essentially complete preservation of HOOH over the subsequent year for Φ258 = 0.15 and 0.19 and efficient preservation (83%) for Φ258 = 0.34. In fact, modeled wintertime HOOH concentrations rise slightly because of photoformation; however, this minor effect is probably overestimated because Rf(HOOH) in our model is independent of temperature. Finally, the 21 March start date shows a pattern of HOOH loss similar to the 21 June scenario, with 55%, 80%, and 87% of the initially deposited HOOH preserved over the course of the year for Φ258 = 0.34, 0.19, and 0.15, respectively. As described in section S6, we estimate that our preservation percentages have a relative uncertainty of approximately ±30%.
 Which of the three HOOH quantum efficiencies best reflects field conditions? Although some snowpack HOOH is present in liquid-like regions, most is present in bulk ice [Jacobi et al., 2002], suggesting that Φ258 = 0.19 is most appropriate since it was determined under conditions where HOOH is expected to be in bulk ice [Beine and Anastasio, 2011]. The higher quantum efficiency, Φ258 = 0.34, was determined in ice samples where HOOH is expected to be in LLRs [Chu and Anastasio, 2005]. While HOOH photolysis in LLRs is important in polar snow (especially since many other solutes will be present in this reservoir), in terms of the HOOH photolytic budget, natural conditions are probably best represented by Φ258 = 0.19, i.e., the solid lines in Figure 3.
 Based on our model results, photochemical processes are broadly consistent with the long lifetimes of HOOH observed in polar snow. In this section, we more specifically examine whether our results are consistent with previous model and field data from Summit. HOOH concentrations in snow at Summit have a seasonal cycle, with summertime maxima and wintertime minima [Anklin and Bales, 1997; Hutterli et al., 2003; Sigg and Neftel, 1991]. For HOOH deposited in snow during summer, approximately 60%–80% is preserved over the course of the next 2 years, with most of the decrease occurring in the first year [Anklin and Bales, 1997]. This loss has been explained by degassing of HOOH driven by temperature gradients [Anklin and Bales, 1997]. Hutterli et al.  found that a model incorporating only physical processes in the snowpack could reproduce HOOH concentration profiles measured at Summit: For snow parcels deposited in summer, fall, winter, and spring, approximately 60%, 100%, 100%, and 30%–60%, respectively, of the initial HOOH were preserved over the next year. Our model (with Φ258 = 0.19) gives similar results, with preservation percentages of 60%, 90%, 100%, and 80%, respectively. Thus, our model results are generally consistent with Summit measurements. The higher spring preservation in our model may be because we used a lower initial HOOH concentration (5 μM) compared to Hutterli et al.  (20–30 μM); as shown in Figure S9, the preservation percentage is higher at lower initial HOOH concentrations.
 Although the physical model of Hutterli et al.  can reproduce measured HOOH concentrations in the Summit snowpack, it does not exclude the possibility that photochemistry contributes to establishing these concentrations. While HOOH loss from snowpacks certainly involves physical exchange from snow grains to the atmosphere, our results indicate that photolysis also contributes in the photic zone. In other words, the Hutterli et al.  study suggests that physical processes control the final concentration of HOOH, while our study suggests that photochemistry is one of the mechanisms that moves the system to this final concentration.
 An accurate understanding of the relative contributions of photochemistry and physical processes would require a coupled model incorporating bulk ice, LLRs, and air reservoirs for HOOH, which is beyond the scope of our effort here and, to the best of our knowledge, has not been done. We can, however, compare with results from Hutterli et al. , who measured HOOH concentrations in the top 2 cm of a Summit snowpack over the course of 12 days in summer and found an average loss rate of 3.3 × 1013 mlc m−2 s−1. Using the HOOH concentrations measured on day 1 from their study, the average loss rate in our model from photolysis is 1.4 × 1013 mlc-HOOH m−2 s−1 in the top 2 cm. This is approximately 40% of the loss measured by Hutterli et al., suggesting that photochemistry is significant. In this same study, the physical model predicted a total snow-to-air flux of 4.6 × 1013 mlc-HOOH m−2 s−1, which closely matched the average (±1σ) measured value of 4.9 (±0.9) × 1013 mlc m−2 s−1. While this agreement suggests that physical release is the dominant mechanism for snowpack HOOH loss, the uncertainty in the average measured flux is large enough (“…less than a factor of 2…” [Hutterli et al., 2001]) that there could also be a significant amount of HOOH photolysis. For these conditions, our model predicts a photolysis loss rate of 5.2 × 1013 mlc-HOOH m−2 s−1 in the photic zone of the snowpack. This is similar to the physical flux and too large to be fully accommodated in the uncertainty of the measured snow-to-air flux, suggesting that either we are overestimating the photolysis rate or there are mechanisms in the snowpack that reduce the net loss of HOOH from photolysis. We do not believe that the first possibility is the primary factor, based on previous measurements of OH production from HOOH photolysis in manipulated Summit surface snow [Anastasio et al., 2007]: After adjusting to the field conditions, our model gives an HOOH loss rate in surface snow that is roughly consistent with the measured OH rates, suggesting that our model effectively captures HOOH photochemistry. Instead, we posit that multiphase reactions in the snowpack recycle snow grain HOOH, thus reducing the net loss from photochemistry. For example, one such sequence is as follows: (1) HOOH photolysis on snow grains to make OH; (2) OH oxidation of an organic to make formaldehyde (HCHO), which degasses to the firn air; (3) photolysis of HCHO(g) to make HO2, which partitions to a snow grain; and (4) reaction of two HO2 on the snow grain to reform HOOH. This multiphase recycling might be much more efficient than the homogeneous, ice-phase recycling we have previously examined [Hullar and Anastasio, 2011].
 It is important to distinguish between physical and photolytic losses of snow grain HOOH because photolysis forms OH, which can oxidize organic compounds to volatile species that could be emitted to the atmosphere [Chu and Anastasio, 2005; Dolinova et al., 2006] and oxidize bromide to gaseous Br2 [Thomas et al., 2011]. Measured organic carbon concentrations in snow at Summit range from 44 to 123 µg C kg−1 [Hagler et al., 2007], which correspond to molar concentrations of approximately 0.4–1.2 μM, assuming an average molecular weight of 100 g mol−1. In our model (21 June start and Φ258 = 0.19), photolysis of HOOH produces the equivalent of approximately 1.2 μM OH in 5 days, comparable to the total concentration of snowpack organics. While the organic compounds and HOOH will not all be collocated, this simple comparison suggests that OH produced from snowpack HOOH photolysis is important for oxidizing and transforming organics in polar snowpacks, as previously noted [Anastasio et al., 2007; Gao et al., 2012].
 The photolytic lifetime of snow grain HOOH calculated from laboratory experiments suggests that HOOH should be completely destroyed in polar snowpacks during summer, a finding at odds with the continuous presence of snow grain HOOH from the surface to below the photic zone in the field. Our model shows that this discrepancy is resolved by using quantum efficiencies for HOOH determined in bulk ice along with the process of HOOH photoformation; this combination gives a photochemical loss that is consistent with field observations of HOOH in snow. Applying the model to Summit, Greenland, shows that photolysis is a significant sink for snow grain HOOH but that the photolysis rate decreases as the snow parcel is buried, until photolysis is negligible approximately 1 year after initial deposition. In agreement with field observations at Summit, our model predicts that HOOH deposited in the summer shows the greatest percentage loss over the course of a subsequent year, while HOOH deposited in winter is efficiently preserved. Our results suggest that photochemistry plays an important role in determining HOOH concentrations in snowpacks at Summit and other polar sites, complementing the well-known temperature-dependent physical processes. However, comparing our results with field measurements indicates that there are still some uncharacterized components of the HOOH system in polar snowpacks, possibly including multiphase reactions that recycle photolyzed HOOH.
 We are grateful for funding from the National Science Foundation (ANT-0636985 and ECS-1214121) and a UCD Donald Crosby Fellowship in Environmental Chemistry to T. H.
 The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.