Photolytic modification of seasonal nitrate isotope cycles in East Antarctica

. Nitrate in Antarctic snow has seasonal cycles in its nitrogen and oxygen isotopic ratios that reflect its sources and atmospheric formation processes, and as a result, nitrate archived in Antarctic ice should have great potential to record atmospheric chemistry changes over thousands of years. However, sunlight that strikes the snow surface results in photolytic nitrate loss and isotopic fractionation that can completely obscure the nitrate’s original isotopic values. To gain insight into 10 how photolysis overwrites the seasonal atmospheric cycles, we collected 244 snow samples along an 850 km transect of East Antarctica during the 2013–2014 CHICTABA traverse. The CHICTABA route’s limited elevation change, consistent distance between the coast and the high interior plateau, and intermediate accumulation rates offered a gentle environmental gradient ideal for studying the competing pre- and post-depositional influences on archived nitrate isotopes. We find that nitrate isotopes in snow along the transect are indeed notably modified by photolysis after deposition, and drier sites have 15 more intense photolytic impacts. Still, an imprint of the original seasonal cycles of atmospheric nitrate isotopes is still present in the top 1–2 m of the snowpack and likely preserved through archiving in glacial ice at these sites. Despite this preservation, reconstructing past atmospheric values from archived nitrate along CHICTABA and in similar transitional regions remains will remain a difficult challenge without having an independent proxy for photolytic loss to correct for post-depositional isotopic changes. Nevertheless, nitrate isotopes should function as a proxy for snow accumulation rate in such 20 regions if multiple years of deposition are aggregated to remove the seasonal cycles, and this application can prove highly valuable in its own right.


Introduction
Nitrate (NO3 -) is one of the most prevalent ions in Antarctic snow and ice, arriving as an end product of the atmospheric oxidation of nitrogen oxides (NOx = NO + NO2) in wet or dry deposition of nitric acid (HNO3) or particulate nitrate (p-NO3 -) 25 (Neubauer and Heumann, 1988;Wolff, 1995;Röthlisberger et al., 2000;Savarino et al., 2007;Frey et al., 2009;Shi et al., 2018b). Because the isotopic ratios of nitrogen and oxygen in atmospheric NO3reflect differences in the original sourcing of the NO3and its atmospheric reaction history, a long-term NO3archive could reveal how the atmosphere's oxidative capacity and chemical reaction pathways have changed over time (Legrand et al., 1999;Michalski et al., 2005;Wolff et al., 2007;Alexander et al., 2009;Kamezaki et al., 2019). Despite its paleoenvironmental potential, NO3has been difficult to 30 interpret in ice cores because post-depositional processes in the uppermost snowpack often result in substantial mass loss and isotopic changes Grannas et al., 2007;Frey et al., 2009;Erbland et al., 2013;Meusinger et al., 2014;Traversi et al., 2014;Geng et al., 2015). Before the paleoenvironmental potential of NO3can be fully realized, we require an improved understanding on how the isotopic values in NO3are altered during the archiving process from in the snowpack from the atmosphere atmospheric source to the snowpack and finally tointo eventual glacial ice. 35 Atmospheric NO3sampled 1-10 m above the snow surface in Antarctica has clear annual cycles in concentration and isotopic values related to seasonal changes in NO3source and formation reaction pathways (Wagenbach et al., 1998;Savarino et al., 2007Savarino et al., , 2016Frey et al., 2009;Erbland et al., 2013;Ishino et al., 2017;Winton et al., 2020). Through wet or Formatted: Space After: 1.2 line dry NO3deposition, these annual cycles are transferred to with the NO3present onto the snow surface. After deposition, NO3photolysis, HNO3 volatilization, and physical snow mixing can alter and obscure these cycles, but post-depositional 40 NO3processes are largely restricted to a shallow (i.e., 0.1-1.0 m) surface layer of the snowpack where light can penetrate, interstitial air can exchange with the atmosphere, and snow can be eroded and mixed by wind (e.g., Grannas et al., 2007;Wolff et al., 2002;Röthlisberger et al., 2002;Frezzotti et al., 2002;Libois et al., 2014;Scarchilli et al., 2010;Picard et al., 2019). After NO3in a snow layer is buried beneath this "active zone" by additional snow accumulation, it is believed to be generally nonreactive and stable. 45 As a result, the magnitude of post-depositional mass loss and isotopic changes relative to the initial depositional values is heavily controlled by the speed at which NO3is buried, i.e., the local surface mass balance (SMB, equivalent here to "net accumulation rate"). At very high SMB sites near the Antarctic coast, NO3is rapidly buried, and the original chemical nature of the atmospheric NO3is largely preserved through the burial process. At very low SMB sites, in contrast, it may take several years for NO3to be buried below the zone of active post-depositional processes, and NO3observed in ice cores 50 and snow pits at dry interior Antarctic stations has such substantial isotopic changes and extreme mass loss that the original depositional values of NO3 − are completely obscured (Freyer et al., 1996;Frey et al., 2009;Erbland et al., 2013;Shi et al., 2015). Most of Antarctica, however, falls between these two SMB extremes environments (Agosta et al., 2019), and archived NO3concentration and isotopic profiles throughout Antarctica likely exhibit a gradient between full preservation of the atmospheric NO3characteristics and the complete post-depositional loss of these characteristics due to overwhelming 55 post-depositional changes. Snow and ice from intermediate SMB sites can thus offer valuable insight into exactly how postdepositional processes interact with and change the initial isotopic chemistry of NO3that is deposited in Antarctica.
We present here NO3data of snow samples taken during the CHICTABA ("Chemical-physical analyses of snow and firn for determining accumulation in Terre Adélie and Aurora Basin North") traverse across a relatively wetter and lower elevation region of the East Antarctic Plateau in austral summer 2013-2014. The NO3data include NO3mass fractions 60 (ω(NO3 -)), isotopic ratios ( 15 NNO3 and  18 ONO3, where  = − 1, with R denoting the 15 N/ 14 N or 18 O/ 16 O isotopic ratios of NO3 -, reported relative to the standards N2-Air (Mariotti, 1983) and Vienna Standard Mean Ocean Water (VSMOW) (Baertschi, 1976), respectively), and the oxygen isotope anomaly (Δ 17 ONO3, where Δ 17 ONO3 = δ 17 ONO3 − 0.52 × δ 18 ONO3) (Thiemens and Heidenreich, 1983). The sites sampled along this traverse have climatology and SMB intermediate to the coast and interior plateau, and thus the NO3offers an important link between existing studies focused on these those While the highest  15 NNO3 values coincide with the late winter peak in  18 ONO3 and Δ 17 ONO3 values, the lowest  15 NNO3 values occur in spring (Oct-Nov), 1-2 months before the minima in  18 ONO3 and Δ 17 ONO3. Additionally, a minor secondary peak in atmospheric  15 NNO3 has also been observed at Dome C in January (Figure 1b) Erbland et al., 2013;Winton et al., 2020). 80 These annual cycles have been attributed to changes in NO3sourcing and reaction pathways related to the distinctly different extreme environments of polar summer and winter. During daytime, photolysis can be a significant local source of NO3when ultraviolet solar radiation converts NO3in the snowpack into NOx gases that then ventilate upward into the atmosphere and oxidize back into HNO3 Erbland et al., 2015;Winton et al., 2020). In polar winter, however, the limited or complete lack of sunlight largely prevents photolysis from occurring, and atmospheric NO3over 85 Antarctica in winter is thought to be largely supplied through long-distance transport from lower latitudes Lee et al., 2014;Shi et al., 2018b;Walters et al., 2019). Substantial influx of this low latitude NO3 − is limited in winter by the intense Antarctic polar vortex, and, NO3concentrations in winter are very low as a result. During the coldest conditions in late winter and early spring, stratospheric denitrification through polar stratospheric cloud sedimentation supplies a small amount of NO3with relatively high  15 NNO3,  18 ONO3, and Δ 17 ONO3 values to the troposphere above 90 Antarctica (Fahey et al., 1990;Van Allen et al., 1995;Santee et al., 2004;Savarino et al., 2007;Ishino et al., 2017;Shi et al., 2022a). This stratospheric supply produces a small observed increase in NO3concentration and contributes to the annual peaks in isotopic values ( Figure 1). Additionally, because ozone (O3) transfers its anomalously high Δ 17 O value to NO3 − when it is involved in NOx cycling, the higher Δ 17 ONO3 values observed in Antarctic winter are attributed to this NO3 − being sourced from lower latitudes and the stratosphere where O3 oxidation is more important (Alexander et al., 2009;Savarino et 95 al., 2016;Ishino et al., 2017).Additionally, as ozone (O3) transfers its anomalously high Δ 17 O value to NO3when NOx is oxidized through O3 pathways, higher Δ 17 ONO3 values are favored in the dark polar winter when O3 oxidation does not compete with alternative photochemical oxidation pathways (Alexander et al., 2009;Savarino et al., 2016;Ishino et al., 2017).
With the return of intense sunlight in spring, photolysis will convert much of the NO3that has accumulated in the near 100 surface snowpack through winter into NOx which is rapidly re-oxidized into HNO3 upon reaching the atmosphere Davis et al., 2004Davis et al., , 2008Grannas et al., 2007;Jacobi and Hilker, 2007;Erbland et al., 2015;Winton et al., 2020;Barbero et al., 2021). This new source of "recycled" NO3produces a rapid rise in atmospheric NO3concentration in November, with some NO3possibly supplied by additional recycled HNO3 transported from upwind regions of Antarctica Shi et al., 2018a). The recycled NO3has isotopic values lower than the mean atmospheric NO3 − 105 values due to strongly negative isotopic fractionation factors during NO3photolysis Erbland et al., 2013;Berhanu et al., 2014Berhanu et al., , 2015Shi et al., 2015) and incorporation of oxygen atoms from local water sources (snow and water vapor  18 O = −20-−80 ‰, Δ 17 O ≈ 0 ‰) during re-oxidation (McCabe et al., 2005;Erbland et al., 2013;Winton et al., 2020).
Sunlight also triggers additional oxidation pathways for NO3formation through HOx, ROx, and H2O2 that lack the anomalous Δ 17 O value of O3 (i.e., their NO3product has Δ 17 O = 0), and Δ 17 ONO3 values are expected to decline in summer 110 as these pathways compete with the O3 pathway (Alexander et al., 2009;Savarino et al., 2016;Ishino et al., 2017). Several unknowns still exist regarding the atmospheric NO3budget for Antarctica, however,based on disagreements between field observations and model predictions for isotopic values and photolytic constants, and this the atmospheric NO3budget for Antarctica remains an active field of research (e.g., Savarino et al., 2016;Walters et al., 2019;Barbero et al., 2021).

Snow skin layer NO3chemistry 115
The seasonal variability of NO3in the snowpack's "skin layer" (i.e., the uppermost 2-6 mm layer of loose snow grains) generally follows that of the local atmospheric NO3 - (Figure 1Figure 1e-h). This similarity is because skin layer NO3is in a close exchange with atmospheric NO3 -, being sourced from recently deposited atmospheric NO3and also supplying NO3to the atmosphere through photolysis during sunlit times. Spatially across Antarctica, skin layer ω(NO3 -) is generally higher at drier and more inland regions Erbland et al., 2013;Shi et al., 2015Shi et al., , 2018b.Although despite atmospheric 120 NO3 − concentrations showing far less spatial variability Frey et al., 2009;Shi et al., 2022a). The higher ω(NO3 -) observed in the skin layer at drier sites is attributed to increased local NO3 − deposition from photolytic recycling as well as the fact that drier sites will dilute the NO3 − less when NO3 − deposition rates are similar across Antarctica (Erbland et al., 2013;Shi et al., 2018b;Winton et al., 2020).observations are uncommon outside of a few scientific stations, higher skin layer ω(NO3 -) values have been observed at drier and more inland regions Erbland et al., 125 2013;Shi et al., 2015Shi et al., , 2018b.
Some differences between atmospheric and skin layer values do exist, however. Notably,  15 NNO3 values in the skin layer are 5-15 ‰ higher than the atmosphere, possibly due to isotopic fractionation as atmospheric HNO3 gas adsorbs onto the snow surface (Erbland et al., 2013;Winton et al., 2020). Additionally, the NO3oxygen isotopes in the skin layer are consistently higher than those observed in atmospheric NO3 - (Erbland et al., 2013;Winton et al., 2020), and this unexpected discrepancy 130 is unexpected and currently unexplained and puzzling. This difference is greatest in the early winter, when  18 ONO3 and Δ 17 ONO3 values can be up to 20 ‰ and 10 ‰ higher, respectively, in the skin layer than the atmosphere. Full annual skin layer observations of ω(NO3 -) and NO3isotopes are only available to datehave until recently been only available from Dome C (Figure 1Figure 1e-h) Erbland et al., 2013;Winton et al., 2020), but a recent record from Zhongshan station suggests that oxygen isotopic values at coastal sites may match more closely between the atmosphere and 135 snow surface (Shi et al., 2022a). Additional data from Zhongshan station and other sites will allow us to better judge the representativeness of the Dome C data with regards to the broader Antarctic environment. and it is thus unfortunately not known if the patterns observed at Dome C are representative for other Antarctic sites with higher SMB Erbland et al., 2013;Winton et al., 2020).   (Erbland et al., 2013;Winton et al., 2020). Atmospheric NO3 − (a--d) was collected over week-long periods with a high-volume air filter located 5 m above the snow surface, and snow surface samples (e-h) were taken every 1-7 days from the 2-6 mm thick skin layer in the clean sector outside Concordia Stationstation. Individual 145 points represent individual samples, and the thick colored lines represent the monthly mean values with the 95 % confidence interval of the mean shown as colored shading. Note that the units for NO3 − concentration is ng m −3 for atmospheric NO3 − (a) and ng g −1 for the snow surface NO3 − (e). A dashed line representing the atmospheric NO3 − concentration multiplied by 10 is included in (a) for better observation of the annual variation pattern.

150
2.3 Post-depositional processes affecting NO3 -During burial, several post-depositional processes can alter the values of skin layer NO3 -. Past studies of buried NO3on the interior East Antarctic Plateau have highlighted photolysis as the primary post-depositional process that affects NO3in East Antarctic snow, resulting in substantial NO3mass loss that can reach > 90 % reduction at dome summits. The NO3remaining in snow at depthafter mass loss shows marked increases in  15 NNO3 values due to a negative photolytic isotopic 155 fractionation factor for nitrogen. Although fFractionation factors for  18 ONO3 and Δ 17 ONO3 are theoretically predicted to be negative, and therefore oxygen isotopic values of remaining NO3 − should increase in a similar fashion tolike  15 NNO3 after photolysis . , actual observations of NO3in Antarctic snow conditions revealed thatHowever, NO3 − at sites with clear photolytic mass loss produces typically has lower  18 ONO3 and Δ 17 ONO3 values lower than atmospheric values . This discrepancy has been explained as the NO3incorporating and exchanging isotopically lighter 160 oxygen from local water reservoirs (e.g., snow and interstitial water vapor) through a cage effect during re-oxidation after of photolytic conversion to NOxproducts (McCabe et al., 2005;Erbland et al., 2015). There is no significant similar reservoir of Formatted: Space After: 6 pt exchange for nitrogen, and as a result, the net effect of photolysis and re-oxidation produces so-called "apparent" fractionation constants that are negative for  15 NNO3 and positive for  18 ONO3 and Δ 17 ONO3 (Röthlisberger et al., 2002;Wolff et al., 2002;Blunier et al., 2005;Grannas et al., 2007;McCabe et al., 2007;e.g., Frey et al., 2009;Winton et al., 2020). As 165 sunlight is rapidly attenuated beneath the snow surface, photolytic loss is restricted to the photic zone (i.e., the 0.1-1.0 m deep zone that light can penetrate and sustain photochemical reactions) and is most pronounced in the uppermost few centimeters of the snowpack Zatko et al., 2013;Erbland et al., 2015;Winton et al., 2020).
Although photolysis dominates post-depositional changes to NO3 -, other factors can also play minor roles. Wind can physically mix snow bearing NO3from different seasons or years which mayand blur pre-existing NO3cycles. 170 Additionally, the development and migration of surface features like dunes and sastrugi can result in wildly variable hyperlocal accumulation rates on short timescales (0.5-5 yr) and across very short distances (< 5 m), even if the mean SMB for the broader region stays constant. These phases of erosion and deposition can result in NO3cycles that appear stretched or compressed relative to expectations from regional SMB or even create stratigraphic unconformities with missing periods of deposition (Frezzotti et al., 2002;Scarchilli et al., 2010;Gautier et al., 2016;Picard et al., 2019). NO3volatilization can 175 also be a source of NO3mass loss in Antarctic snow, but it is largely restricted to the warmest coastal regions of Antarctica and is believed to have little isotopic fractionation impact (Erbland et al., 2013;Shi et al., 2019). Finally, downward NOx transport and reoxidationre-oxidation of photolytic NOx within the firn may also occur, but as of yet this process is poorly attested and significant impacts appear to be largely restricted to very dry interior sites (SMB < 40 kg m −2 a −1 ) (Akers et al., 2022). Once snow is buried beneath the depth where post-depositional processes are active, NO3it is assumed to be 180 practically chemically inert (especially for ω(NO3 -) and  15 NNO3) and physically immobile Erbland et al., 2013;Shi et al., 2015;Noro et al., 2018), aside from volcanic H2SO4-driven NO3displacement with no changes to isotopic compositions (Wolff, 1995;Röthlisberger et al., 2002;Jiang et al., 2019).
Once snow is buried beneath the depth where post-depositional processes are active, it is assumed to be practically chemically inert (especially for ω(NO3 -) and  15 NNO3) and physically immobile Erbland et al., 2013;Shi et 185 al., 2015;Noro et al., 2018), aside from volcanic H2SO4-driven NO3displacement (Wolff, 1995;Röthlisberger et al., 2002;Jiang et al., 2019). The overall impact of these post depositional effects on NO3 − in Antarctic snow and ice varies strongly depending upon local SMB (Shi et al., 2015Akers et al., 2022). At sites with very high SMB, such as near the Antarctic coast, post-depositional effects have little time to alter NO3 − , and the NO3in ice cores likelyshould preservees atmospheric NO3values relatively well in a manner following Late Holocene ice core NO3reported from similarly high 190 SMB Greenland (Hastings et al., 2004;Fibiger et al., 2013). For much of drier inland Antarctica, in contrast, it may take 2-10 years for NO3to reach thisreach the "archived zone" beneath the range of post-depositional effects, and the combined effects of the post-depositional processes here typically overwhelm and obliterate any NO3seasonal cycle variability (Erbland et al., 2013;Shi et al., 2015). Photolytic impacts, in particular, are sensitive to SMB in East Antarctica with a strong linear correlation observed spatially between  15 NNO3 and the reciprocal of SMB In past studies, photolytic changes to 195  15 NNO3 exhibit a linear correlation with the reciprocal of SMB (Akers et al., 2022)(e.g., Erbland et al., 2013;Noro et al., 2018). For much of inland Antarctica, it may take 2-10 years for NO3to reach this "archived zone", and the combined effects of the post-depositional processes typically overwhelm and obliterate any NO3seasonal cycle variability (Erbland et al., 2013;Shi et al., 2015). Higher SMB generally leads to better preservation of the original atmospheric chemistry of NO3because faster accumulation archives NO3in deep glacial ice more quickly. At sites with very high SMB, such as near the 200 Antarctic coast, NO3in ice cores likely preserves atmospheric NO3values relatively well in a manner following ice core NO3reported from similarly high SMB Greenland (Hastings et al., 2004;Fibiger et al., 2013). Changes in insolation, total column ozone, and snow optical properties also can leave imprints on the isotopic values of NO3 − by affecting the photolytic rate, but the greater photolytic sensitivity to SMB changes tends to overwhelm and obscure their impact (Zatko et al., 2016; Formatted: Not Superscript/ Subscript Winton et al., 2020;Akers et al., 2022;Cao et al., 2022;Shi et al., 2022b). Still, these other photolytic factors remain 205 enticing targets for paleoenvironmental reconstruction. (Zatko et al., 2016;Winton et al., 2020;Akers et al., 2022;Cao et al., 2022;Shi et al., 2022b)

Methods
We sampled snow for NO3analysis in Nov-Dec 2013 at 23 sites along the CHICTABA traverse (Table 1Table 1 Each snow sample was melted at room temperature at in Concordia sStation, Dome C, Antarctica,, and NO3concentrations of the melted samples were determined on aliquots by a colorimetric method with a detection limit of 0.5 ng g −1 and precision < 3 % Erbland et al., 2013). Melted samples were immediately passed through an anionic 225 exchange resin (Bio-Rad™ AG 1-X8, chloride form), and the resulting trapped NO3eluted with 2 x 510 ml of NaCl 1 M solution. These concentrated samples were then frozen and shipped to the Institut des Géosciences de l'Environnement (IGE), Grenoble, France, for isotopic analysis. Once re-melted, NO3in these samples was converted to N2O with a strain of the denitrifying bacteria Pseudomonas aureofaciens that lacks the ability to reduce N2O into N2. The N2O was thermally decomposed into O2 and N2 on a 900° C gold surface, separated by gas chromatography with a GasBench II™, and oxygen 230 and nitrogen isotopic ratios measured on a Thermo Finnigan™ MAT 253 mass spectrometer (Sigman et al., 2001;Casciotti et al., 2002;Kaiser et al., 2007;Morin et al., 2009). Isotopic effects from this analysis were corrected using the calibration regressions based on standards of international reference materials USGS 32, USGS 34, and USGS 35 processed and analyzed along with each set of samples Morin et al., 2009), Morin et al., 2009).
Standards and samples strictly follow an identical treatment, having the same liquid volume, bacterial culture, and water 235 isotope composition. Isotopic valuesand are reported relative to the N2-Air and VSMOW standard references (Baertschi, 1976;Mariotti, 1983). T, and the root mean square errors of of calibration regressionsstandards run alongside our for these samples over four analytical runs were ±0.7-1.1 ‰ for δ 15 NNO3, ±0.8-2.3 ‰ for δ 18 ONO3, and ±0.2-0.4 ‰ for Δ 17 O NO3. For statistical results reported throughout this paper, uncertainties are given as 95 % confidence intervals unless otherwise stated, and statistical significance is identified as p-values < 0.05. 240

250
REMA raster cell-size. The SMB values for the SMB profile were bilinearly interpolated at 1 km intervals from the original 35 km MAR output grid. MAR SMB uncertainty is included on (d) as a shaded zone around the profile line, but can beis difficult to see due to its relatively small extentsize. In line with previous studies (Blunier et al., 2005;Shi et al., 2015), εapp values are calculated as the slope of a linear regression through Eq. (1): where R0 and Rf denote isotopic ratios in the initial and remaining NO3and ωf denotes the mass fraction of remaining NO3 -.
This equation can also be written with delta notation: 270 where δ0 and δf denote the desired isotopic species in delta value notation versus a chosen standard (e.g.,  18 ONO3 vs. VSMOW). For the subset of skin layer sites that had a paired with the 1 m depth layer samples, this regression is simple as it only has two points. In the pits, however, the regressions did not capture well the broader multi-annual photolytic trend due largely to the limited number of seasonal cycles recorded per pit and due to the irregular magnitude peaks of the ω(NO3 -275 ) cycle, which contributed large outlier points. We therefore created "pseudo-depth layer samples" for each of the five pits to represent an annually-averaged NO3 − value from below the photic zone. These pseudo-depth layer samples are the ω(NO3 -)weighted means of ω(NO3 -),  15 NNO3,  18 ONO3, and Δ 17 ONO3 for the deepest complete full seasonal cycle observed for P1-P4 and the deepest three complete cycles combined for P5.We therefore created "pseudo-depth layer samples" for each of the five pits by calculating the ω(NO3 -)-weighted means of ω(NO3 -),  15 NNO3,  18 ONO3, and Δ 17 ONO3 for the deepest complete 280 full seasonal cycle observed in the pit data for P1-P4 and the deepest three complete cycles for P5.
The εapp values produced in this manner give insight into the isotopic fractionation processes at work, but they also have limitations that are important to recognize. Namely, the most accurate εapp determinations require many samples taken over a full photic zone profile to compensate for seasonal and environmental variability in ω(NO3 -) and  15 NNO3 values (Shi et al., 2015). This is of particular importance at sites where annual snow accumulation is greater than the NO3 − sampling 285 resolution, as is the case for our CHICTABA samples. Because our 1 m depth samples were taken as the mixed aggregate of a layer only 5-10 cm thick, each 1 m depth sample collects only part of a complete annual NO3 − cycle. Assuming that the odds of the exact seasonal timing sampled by each 1 m depth sample is stochastic, our individual εapp values should be viewed as having wide uncertainty with regards to the true site εapp value, but εapp values averaged across our dataset should reflect accurate regional εapp values. 290 To examine spatial relationships in NO3 − with SMB along the CHICTABA transect, we calculated Llinear regressions between NO3 − variables and local SMB were calculated for both skin layer samples and 1 m depth layer samples (including the five pit pseudo-depth layer samples) to examine spatial relationships in NO3 − with SMB along the transect. Following the relationships defined in Akers et al. (2022), regressions were performed as ω(NO3 -) or ln(δf + 1) versus SMB −1 . The SMB 295 values used in these regressions were the mean annual MAR output for the period 2011-2013 (i.e., the three years preceding sampling). This period was chosen because three years of snowfall at the CHICTABA sites is roughly equal to 1 m of accumulation and compaction. Additional regressions were calculated using the mean annual MAR output for the full data coverage period of 1979-2021 and for the sole year 2013 to determine if the choice of SMB data period substantially affected results. We again assume that any seasonal bias introduced by the 1 m depth sampling technique would be 300 stochastic and that conclusions drawn from observations integrating all sites are generally accurate but admittedly more imprecise than if the individual 1 m depth samples had integrated full annual cycles. Statistical calculations and figure production were performed using the R programming language with packages tidyverse, lubridate, RColorBrewer, gridExtra, cowplot, raster, rts, ncdf4, RMisc, and HMisc. QGIS was used for spatial analyses and map creation using data produced here or cited in image captions, and with Adobe Illustrator used for finalization of graphicfigures figures. 305

Results
In total, 234 individual snow samples were analyzed for ω(NO3 -) and NO3isotopic ratios (Figure 3Figure Table   S1). The pits have 2-2.5 cycles in the top 100 cm with drier sites containing more cycles per unit depth. For the deeper P5, we observe five complete cycles over the total 201 cm depth. Linear regressions of NO3 − variables with depth reveal that 325 ω(NO3 -) has statistically significant negative slopes at P3-P5 (p < 0.01) while  15 NNO3 has a significant positive slope only at P5 (p = 0.02). Both  18 ONO3 and Δ 17 ONO3 have statistically significant negative slopes at all pits except P1 (p < 0.01). In the absence of other supplemental geochemical data, the residuals of the Δ 17 ONO3 regression with depth were used to identify seasonal cycles with positive residuals representing colder months and negative residuals representing warmer months ( Figure 4). This seasonal identification is based on NO3 − monitoring data from Dome C (Figure 1) and previously reported 330 seasonal Δ 17 ONO3 cycles linked to snow  18 O variability in a snow pit (Shi et al., 2015).
We investigated how the cycle timing of NO3variables were interrelated by correlating values after we removed linear trends with depth (i.e., we correlated the residuals of the linear regressions). Values for  18 ONO3 and Δ 17 ONO3 values are wellcorrelated (r = +0.72, p < 0.01), as is typically observed for NO3 -. The ω(NO3 -) has a moderate negative correlation with Formatted: English (United States) p = 0.43). This difference in correlation strength seems unusual since  18 ONO3 and Δ 17 ONO3 are so strongly correlated, but it appears to arise because the  18 ONO3 cycle is slightly more irregular and offset from the  15 NNO3 cycle compared with the Δ 17 ONO3 values (Figure 4Figure 4). Additionally, Δ 17 ONO3 values tend to peak higher than  18 ONO3 values when coinciding 340 with the highest  15 NNO3 values (e.g., P2: 75 cm, P3: 35 cm, P4: 55 cm), and these shared extreme values promote a stronger correlation. The reason for these small differences between  18 ONO3 and Δ 17 ONO3 is not presently clear but may be due to  18 ONO3 values being theoretically directly affected by photolytic mass loss while Δ 17 ONO3 is not. Unfortunately, the impact of a theoretical fractionation of oxygen isotopes by photolytic mass loss is poorly constrained due to competing effects from oxygen atomic exchange during NO3 − re-oxidation, which we examine in more detail later. 345   Table S1). Dashed lines represent regressions whose f-statistic p-value < 0.05, and dotted lines represent regressions whose fstatistic p-value ≥ 0.05. Gray shaded backgrounds indicate inferred seasonal cycles (darker = colder months of ~May-Oct, lighter = warmer months of ~Nov-Apr) based primarily on when residuals of the Δ 17 ONO3 regression are positive (i.e., Δ 17 ONO3 peaks).

360
Measurement uncertainties in isotopic values are displayed as colored shaded zones around the stepped lines, but are too small to be visible on most data.
Across the full 1979-2021 dataset, we find interannual SMB variability to be very high, but the spatial pattern of variability  Table S2).
Only some of the NO3 − variables have statistically significant linear relationships with the 2011-2013 SMB −1 values ( Figure  375 5Figure 5, Table 2Table 2). The  15 NNO3 values decrease with higher SMB in both the skin layer and 1 m depth layer samples (Figure 5Figure 5b), but only the skin layer regression has a statistically significant f-statistic (p < 0.01, n = 23). For oxygen isotopes, only  18 ONO3 in the 1 m depth samples has a statistically significant regression with SMB −1 (f-statistic p = 0.04, n = 14) with higher isotopic values associated with greater snow accumulation rates (Figure 5Figure 5c). The ω(NO3 -) and Δ 17 ONO3 values do not have statistically significant relationships with SMB −1 (Figure 5a,d) in either the skin layer or 1 m 380 depth layer, (Figure 5a, d). although Although the not reaching our defined level of statistical significance, we note that nonzero slope relationships with SMB −1 can also be observed in  18 ONO3 in the skin layer and in  15 NNO3 and Δ 17 ONO3 at skin layer1 m depth. Δ 17 ONO3 does display a trend of increasing value with higher SMB (f-statistic p = 0.10, n = 23).  Table S3).
Apparent fractionation constants for each of the isotopic ratios are generally consistenthave high variability across all sites (Table 3Table 3). This variability is likely due in part to the sampling methodology where the skin layer sample will have summer values (as we collected it in summer), but the 1 m depth samples reflect a random sampling from a buried seasonal cycle. While this reduces the precision of our overall εapp estimate, general conclusions can be drawn from the range of εapp 395 values as well as their measures of central tendency. The 15 εapp values are all negative and range between -65.6 ‰ and -24.8 ‰, with a mean value of -39.7 ± 6.1 ‰. Fractionation constants for oxygen isotopes are positive except at two sites, but smaller in magnitude than that of the nitrogen isotopes: 18 εapp values range between −11.7 ‰ and +15.9 ‰ (mean: +5.0 ± 4.2 ‰) and 17 Εapp range between −5.1 ‰ and +5.3 ‰ (mean: +1.2 ± 1.7 ‰). Fractionation constants do not have statistically significant linear regressions with either SMB or SMB −1 . 400

405
direction of expected changes to NO3 − variables due to photolysis and associated re-oxidation is indicated by colored arrows. Regression coefficientsCoefficients and statistics for displayed regressions are given in Table 2Table 2.

Photolytic impacts observed in skin layer and 1 m depth samples
Our data reveal evidence of photolytic loss changes toof NO3in the photochemically active zone of the snowpack. The 430 mean  15 NNO3 of skin layer samples (−8.9 ± 3.3 ‰) falls is within the typical seasonal range (≈−20 40 to +20 ‰) observed in atmospheric NO3at both coastal and interior Antarctic stations Frey et al., 2009;Erbland et al., 2013;Winton et al., 2020;Shi et al., 2022a), suggesting that the skin layer NO3is recently deposited from the atmosphere and has experienced little to no photolytic mass losseffects. In contrast, the  15 NNO3 values at 1 m depth are 49 ± 11 ‰ higher on average than the skin layer  15 NNO3. This increase, combined with the average 71 ± 9 % ng g −1 drop in ω(NO3 -) from the 435 skin layer to the 1 m depth, strongly points to substantial photolytic mass loss Frey et al., 2009;Meusinger et al., 2014;Zatko et al., 2016). As further support, the range of 15 εapp values (−65.6 ‰ to −24.8 ‰) at the CHICTABA sites is comparable to both modeled and field-observed values previously reported for photolytic fractionation across interior Antarctic transects (−76.8 ‰ to −31.5 ‰) Erbland et al., 2013;Berhanu et al., 2015;Shi et al., 2015). While we acknowledge that HNO3 volatilization could beis not excluded as a minor source of mass loss at some 440 sites due to the wide range in fractionation factors, photolytic mass loss alone can explain our the observed findings without needing to invoke additional, non-fractionating mass loss from volatilization.
Photolystics-related impacts on oxygen isotopes are also present but more subtle. Our observed apparent isotopic fractionation factors for 18 εapp and 17 Εapp are comparable to NO3photolytic oxygen fractionation factors reported in other photolysis studies Erbland et al., 2013;Berhanu et al., 2015;Shi et al., 2015) that are a combination of 445 effects from both photolytic mass loss fractionation and oxygen exchange due to a cage effect during re-oxidation of photolytic products. As is expected from photolysis and the resulting NO3 − re-oxidation, mean values for  18 ONO3 and Δ 17 ONO3 along the CHICTABA transect are lower at 1 m than in the skin layer. However, the difference between the mean skin layer and 1 m depth values is much smaller than observed in  15 NNO3.In contrast to the  15 NNO3 results, neither  18 ONO3 nor Δ 17 ONO3 have large differences in mean value between the skin layer and 1 m depth samples. This is likely because theis 450 reflected in how the fractionation factorsapparent fractionation factors for  18 ONO3 and Δ 17 ONO3 are much closer to zero than the apparent fractionation factor for  15 NNO3 Erbland et al., 2013;Shi et al., 2015), and thus photolytic mass loss did not produce result in as large of a change in isotopic value for oxygen as for nitrogen. Still, mean values for  18 ONO3 and Δ 17 ONO3 are, as expected from photolysis, lower at 1 m than in the skin layer, although only the difference in  18 ONO3 means has a 95 % confidence interval that excludes zero. The difference in oxygen isotopic values between the skin 455 layer and 1 m depth is larger at drier sites (Figure 5c-d), likely because NO3is exposed to photolytic radiation for a longer time due to slower accumulation which serves to exaggerate the magnitude of isotopic change.

Annual nitrate cycle and photolytic evidence observed in pit samples
We interpret the cyclical variability of ω(NO3 -) and NO3in the depth profiles of the CHICTABA snow pits (Figure 4Figure  460 4) as a relic of the annual cycles observed in atmospheric and skin layer NO3 - (Figure 1Figure 1) that has been partially preserved through NO3deposition and initial burial. The cycles in  18 ONO3 and Δ 17 ONO3 are clear and well-synchronized in each pit, which allows us to differentiate between the winter darkness season (peaks) and summer sunlit season (troughs) Frey et al., 2009;Shi et al., 2015). Peaks in ω(NO3 -) generally coincide with the summer minima in the oxygen isotopic cycles due to enhanced deposition of recycled NO3 − (Figure 4Figure  monitoring at Dome C (Figure 1Figure 1) and in three snow pits reported in a previous study (Shi et al., 2015). However, a few minor peaks observed in winter (e.g., in P1 and P4) could represent NO3deposition from stratospheric denitrification.
Additionally, the annual ω(NO3 -) peak corresponding to summer 2012-2013 (i.e., the summer before sampling occurred) is particularly large relative to other ω(NO3 -) peaks in most pits. This may represent a particularly heavy local NO3deposition that year, although atmospheric and skin layer NO3monitoring at Dome C captured no unusually high NO3at that time 470 (Erbland et al., 2013;Winton et al., 2020). Overall, the range and cycles in ω(NO3 -) values observed in these CHICTABA pits are similar to those reported from pits with similar SMB on a transect from Zhongshan station to Dome A (Shi et al., 2018b).
Following that each complete oxygen isotopic cycle is equivalent to one year, the pits cover roughly 2-3.5 years of snow accumulation in the top 100 cm, with five years of accumulation at the 201 cm deep P5. This accumulation is similar to 475 rough estimates (P1: 2.0 yr; P2: 2.5 yr; P3: 2.9 yr; P4: 3.4 yr; P5: 7.0 yr) calculated from modeled SMB for 2011-2013 and snow density profiles taken from two shallow cores along the transect (where 1 m snow depth = 38.9 cm water equivalent and 2.25 m snow depth = 90.4 cm water equivalent). Differences between the modeled estimates and the dating from NO3 − oxygen isotope cycles could be due to interannual snowfall variability, surface roughness, and/or localized differences in snow density profiles. An example of a A surface roughness effect may explain the exceptionally broad  15 NNO3 peak and 480 lack of ω(NO3 -) spike in the upper 50 cm of P2, where as a localized high rate of drifted snow accumulation that might have "stretched" the typical cycle frequency. Otherwise, the general regularity of the isotopic cycles suggests that limited physical mixing or snow layer disturbance occurred after initial deposition.
Although photolysis only occurs during sunlit periods, it affects NO3deposited in all seasons. For the pit data, the cyclical patterns of  15 NNO3 and  18 ONO3 are offset 10-80 ‰ higher and 5-15 ‰ lower, respectively, compared to the mean seasonal 485 cycle values reported from the skin layer at Dome C (Erbland et al., 2013). Because it takes over two years for newly deposited NO3to be buried below 1 m along the CHICTABA traverse, NO3that is deposited in winter darkness will still be exposed to summer sunlight and partially photolyzed before being fully buried below the photic zone. We also find that NO3deposited in the late winter and early spring has the greatest  15 NNO3 increase relative to its corresponding seasonal skin layer values, with pit  15 NNO3 values of 50-100 ‰ compared to skin layer mean values of 10-30 ‰. The 15 N 490 enrichment maximum at this time can be expected because the NO3deposited during late winter and early spring will typically have been buried perhaps 5-20 cm beneath the surface by the time intense summer insolation returns. At this depth, the late winter/early spring NO3is still shallow enough to be readily photolyzed, but also deep enough that newly recycled, isotopically light NO3deposited onto the surface will not be mixed in. with previous observations where increased photolysis and its resulting oxygen exchange during NO3 − re-oxidation produce lower oxygen isotopic ratios (Erbland et al., 2013;Shi et al., 2015). However, it is notable that only P4 and P5 have visibly 500 increasing  15 NNO3 trends with depth as would be expected from cumulative photolytic mass loss. While the difference in mean isotopic value between the top and bottom of the pits is smaller and not as easily distinguishable for nitrogen as with the oxygen isotopes,  15 NNO3 values in P4 and P5 have visibly increasing trends with depth as would be expected from cumulative photolytic mass loss. This suggests that substantial oxygen exchange may be occurring regardless of photolytic mass loss, perhaps due to photolytic NOx being produced and re-oxidized in place without the ventilated transport that leads 505 to mass loss to the atmosphere. In this case of in situ photolysis and re-oxidation, no isotopic effect of photolysis would be observed in nitrogen, but there could be an isotopic change in oxygen due to the chance of an atomic exchange with the local snow grain matrix.
Possible oxygen isotopic changes not triggered by photolysis must also be considered. We expect photolysis to drive the greatest rate of isotopic change in the uppermost depths where radiation is strongest and increasingly less change toward the 510 bottom of the photic zone. We observe this in the  15 NNO3 where  15 NNO3 values greatly increase between each skin layer and ≈6-9 cm depth, even for the P2 and P3 pits where no clear additional photolytic change is present beneath this uppermost zone (Figure 4b). In contrast, the oxygen isotopes have a remarkably consistent rate of isotopic change with depth for P2-P5 (Figure 4c-d). Competition between photolytic mass loss fractionation and oxygen exchange isotopic effects is discussed in the following section as one possible explanation for this difference between nitrogen and oxygen profiles. However, the 515  18 ONO3 and Δ 17 ONO3 values in P5 continue to decline steadily from 100-201 cm. These depths are well beneath the photic zone, and therefore the NO3 − should be isotopically stable. No current mechanism in our current understanding of Antarctic NO3 − dynamics has been described for oxygen isotopic changes in the snowpack without photolysis, and it is difficult to make strong hypotheses or conclusions at this time in the absence of deeper and/or replicated pits. Further and more extensive field observations will be needed to clarify this uncertainty. 520 The relative timing of isotopic cycles in the pits has some small but important differences from the cycles observed in the atmosphere and skin layer at Dome C. As best seen in the P2-P5 pits, the  15 NNO3 cycle generally aligns in phase with oxygen isotopes, but with a slight offset so that the  15 NNO3 maxima and minima are 0-10 cm shallower (~0-3.5 months later) than the corresponding oxygen isotope cycles (Figure 4Figure 4b-d). The delayed  15 NNO3 minima, in particular, is unexpected because the early summer  15 NNO3 minima in atmospheric and skin layer NO3precedes the mid-summer minima 525 in oxygen isotopes by 1-2 months (Figure 1Figure 1b-d) Erbland et al., 2013;Winton et al., 2020). A similar "delayed" relationship between  15 NNO3 and  18 ONO3 can be observed in three snow pits sampled from the wetter section of the Zhongshan to Dome A traverse route (Shi et al., 2015), suggesting that this phenomenon is not unique to CHICTABA and may be typical for intermediate SMB regions of Antarctica.
This discrepancy between observations in snow pits versus the observations in the atmosphere and skin layer may be 530 explained by the seasonality of photolytic loss ( Figure 6). The early summer atmospheric  15 NNO3 minima is due to the photolytic production and subsequent re-oxidation of NOx with low  15 N from the snowpack NO3 -, and the skin layer NO3shares a similarly timed  15 NNO3 minima as the re-oxidized NO3is deposited back onto the surface (Figure 1). However, as this skin layer NO3is buried by additional snow, it will be exposed to sunlight in the photic zone for the entire summer season with subsequent photolytic losses and an increase in  15 NNO3 values.

535
IIn contrast, while NO3deposited toward the end of summer may not initially have  15 NNO3 values as low as in early summer, this NO3will experience far less photolytic-inducing radiation before winter darkness and will likely be buried and protected relatively deep in the photic zone before the next summer begins. In this manner, the late summer  15 NNO3 values could end up as the lowest  15 NNO3 values simply because they are photolytically elevated the least from initial atmospheric values. Likewise, the minimum values in pit oxygen isotope cycles may be shifted slightly earlier in the summer because 540 photolysis re-oxidation of photolytic products lowers  18 ONO3 and Δ 17 ONO3 values through oxygen atomic exchange. Thus, we would observe the oxygen isotopic minima occurring before the nitrogen isotopic minima in the pit profiles, despite the atmospheric and skin layer cycles not exhibiting this pattern.  for nitrogen due to photolytic fractionation (orange arrows) but decreases values for oxygen due to oxygen atomic exchange (blue arrows). Because the typical amount of photolytic activity experienced by NO3 − deposited on the snow surface (orange solid curve) also changes seasonally in a cycle not aligned with the skin layer isotopic value cycles, photolysis will enhance or subdue the 550 existing skin layer isotopic cycle differently for  15 NNO3 than for  18 ONO3 and Δ 17 ONO3. This produces the observed cyclical offsets between nitrogen and oxygen isotopes (Figure 4), even if the magnitude of isotopic value changes due to photolysis (i.e., the size of the arrows) is the same for all isotopic species at a given point in the cycle.

Links between  15 NNO3 and SMB
The linear relationships between NO3variables and snow accumulation rateSMB (Figure 5Figure 5) match what is expected 555 based on photolysis-dominated NO3dynamics on the East Antarctic Plateau. While not all the regressions are statistically significant at p < 0.05, their combined evidence supports increased photolysis with lower SMB. At drier sites, NO3will remain within the photic zone for a longer period due to slower snow accumulation, and as a result the NO3will experience more photolysis before being buried in the archived zone (Akers et al., 2022). For the 1 m depth layer samples,  15 NNO3 values increase while  18 ONO3 and Δ 17 ONO3 values decrease with lower SMB (Figure 5Figure 5b-d), which reflects the 560 negative apparent isotopic fractionation factor of nitrogen with NO3photolysis and the positive apparent fractionation factors for oxygen (Erbland et al., 2013;Shi et al., 2015).
An improved sampling method for the 1 m depth samples might produce stronger and more precise linear regressions with SMB −1 for all isotopic ratios. Each seasonal isotopic cycle typically covers 30-50 cm depth in the upper snowpack as observed in the pit records ( Figure 4). However, each 1 m depth sample taken along the CHICTABA transect likely 565 represents only part of an annual isotopic cycle because our sampling methodology mixed snow from only a 5-10 cm thick layer at 1 m depth. If a seasonal maximum or minimum happened to fall at 1 m depth, the resulting  15 NNO3,  18 ONO3, and Δ 17 ONO3 values could be offset from the true annual mean value by 20-50 ‰, 10-20 ‰, and 5-6 ‰, respectively (Figure 4).  The skin layer samples also show an increase in  15 NNO3 and decrease in  18 ONO3 with lower SMB (Figure 5Figure 5b-c) 575 despite not having much photolytic mass loss that would drive this pattern. Instead, Tthis spatial relationship between NO3 -isotopes 15 NNO3 and SMB in the skin layer likely results from NO3recycling Winton et al., 2020), where some of the NO3deposited on the skin layer is derived from re-oxidized photolytic NOx ventilated from the local snowpack. Because NO3in the snowpack beneath the skin layer has higher  15 NNO3 and lower  18 ONO3 and Δ 17 ONO3 values at drier sites due to increased photolytic mass loss, the isotopic ratios of photolyzed NOxNox products and resulting re-580 oxidized NO3coming from the snowpack will also share similar isotopic relationships with SMBtend to have higher  15 NNO3 values at drier sites. Nitrate at drier sites also experiences more NO3recycling (Erbland et al., 2013Winton et al., 2020) which drives skin layer NO3isotopes to be closer to the high  15 NNO3 and low  18 ONO3 that we observe deeper in the snowpack. Additionally, the skin layer NO3 − would sit at the surface for a slightly longer period at the drier sites than the wetter sites, potentially also giving a slightly greater photolytic imprint on skin layer  15 NNO3 for sites with lower SMB.

585
This shared spatial relationship with SMB for both the skin layer and 1 m depth NO3 - 15 NNO3 samples might be seen as evidence that the NO3 -isotopic 15 NNO3 values at depth 1 m are simply preserving an already existing spatial relationship in NO3isotopes present in the skin layer. If photolytic impacts were indeed the same at all sites, regardless of SMB, we would expect the slope of the 1 m depth samples to match the slope of the skin layer samples, because the degree of isotopic fractionation per unit depth would be the same at every site. However, comparing the spatial regressions of the skin layer 590 samples to the 1 m depth samples reveals that the isotopic differences between the two sample sets is greater at drier sites (Table 2). For nitrogen, both skin layer and 1 m depth samples have higher  15 NNO3 values as SMB decreases, but the  15 NNO3 values in the 1 m samples increase at a greater rate than the skin layer samples (i.e., the magnitude of the regression's slope is greater for the 1 m depth dataset than for the skin layer dataset) ( Table 2). As a result, the greater photolytic action at drier sites enhances and exaggerates the pre-existing  15 NNO3 trend with SMB observed in the skin layer.

595
Similarly, the The oxygen isotope regressions with SMB alsoalso provide some evidence of greater photolytic mass lossactivity at drier sites. Making definitive conclusions from the oxygen isotope regressions is more difficult than for nitrogen isotopes because the uncertainty uncertainties of the 1 m depth layer regressions largely overlap and encompass the regressions for skin layer samples. Still, the regressions suggest that at sites with the highest SMB (180-200 kg m −2 a −1 ), there will not be a significant difference in the oxygen isotopic ratios between the skin layer and the 1 m depth layer while, 600 in contrast, the skin layer has higher  18 ONO3 and Δ 17 ONO3 values than 1 m depth at the driest SMB sites (110-130 kg m −2 a −1 ). This divergence is also expressed through the regression slopes where the 1 m depth samples have a more positive relationship with SMB than the skin layer samples (Figure 4c-d). )(McCabe et al., 2005Erbland et al., 2015) (Zatko et al., 2016)Compared to nitrogen isotopes, it thus appears that a greater degree of photolytic activity (i.e., a drier site) is needed to observe a clear divergence between skin 605 layer and 1 m samples for oxygen isotopic values. This is a reasonable observation because the apparent isotopic fractionation factors for oxygen isotopes are much smaller than for nitrogen, and we would expect based on these observations that photolytic impacts become obvious more quickly for  15 NNO3 than for  18 ONO3 or Δ 17 ONO3. However, the reduced photolytic impact in oxygen isotopes compared to nitrogen isotopes may seem surprising given that the isotopic trends with depth in the pit data are much clearer in the oxygen isotopes than  15 NNO3 (Figure 4).

610
The relatively limited photolytic signal in the oxygen isotopes is likely due in part to summer bias in our skin layer sampling whereas the 1 m depth samples draw from the full range of the annual cycle. During summer, skin layer NO3 − has maximum ω(NO3 -) and minimum isotopic values (Figure 1). An annual mean skin layer sample, however, would have lower ω(NO3 -) and higher isotopic values, although still weighted heavily toward summer values due to summer's much higher NO3 − concentrations. Adjusting our observed skin layer values to reflect annual values increases our calculated εapp values for all isotopic species by 3-10 ‰. This slightly weakens the observed 15 εapp values for nitrogen but more impactfully shifts the 18 εapp and 17 εapp of the oxygen isotopes to clearly positive values that better reflect our observations of oxygen isotopic change in the pit profiles.
The snow pit isotopic trends reveal another unusual characteristic that may also help explain why skin layer and 1 m depth sample values only diverge at drier sites for oxygen isotopes. In the snow pits,  15 NNO3 values rapidly increase in the 620 uppermost 5-10 cm coinciding with the rapid decline in ω(NO3 -) (Figure 4a-b). This follows our expectations as photolytic activity is concentrated near the surface due to rapid attenuation of solar radiation in the snowpack. However,  18 ONO3 and Δ 17 ONO3 values exhibit a steady decline throughout the entire 100-200 cm depth of the pits with no obvious signs of a greater rate of decline at shallow depths where photolytic activity should be strongest (Figure 4c-d). Re-examining the drivers of NO3 − oxygen isotopic change may help explain this inconsistency. 625 The seeming insensitivity of oxygen isotopic values to changing photolytic activity may instead reflect changes in the balance between two competing isotopic effects. Although it has not been experimentally observed, photolytic mass loss is theoretically predicted to have a direct isotopic fractionation effect on oxygen that would increase  18 ONO3 values in the remaining NO3 − , similar to  15 NNO3 . This is counter-balanced by an opposing isotopic effect resulting from oxygen exchange from a cage effect during NO3 − re-oxidation (McCabe et al., 2005). In uppermost 5-10 cm of the 630 snowpack, the proximity of the atmosphere makes it relatively easy for photolyzed NO3 − to be lost from the snowpack. This leads to the rapid change observed in  15 NNO3, but the lack of corresponding substantial change in  18 ONO3 suggests that the isotopic effect of mass loss fractionation is balanced by the competing effect from oxygen exchange in these uppermost depths.
Deeper within the snow, however, photolyzed NO3 − lacks this nearby interface with the atmosphere, and it is more likely that 635 photolytic products will re-oxidize back into NO3 − in place or somewhere within the photic zone. This increase in intrasnowpack NO3 − recycling will reduce photolytic mass loss fractionation, but oxygen exchange can still occur. The balance in competing isotopic effects will thus shift increasingly toward oxygen exchange with greater depth. Although photolytic activity and NO3 − recycling is decreasing with depth due to radiation attenuation, the increased dominance of the oxygen exchange effect appears to compensate for the decreasing radiation to produce the steady lowering in  18 ONO3 values.

640
Therefore unlike  15 NNO3, the greatest degree of isotopic change for  18 ONO3 should occur beneath the immediate uppermost snowpack layers once the oxygen exchange effect is predominant. Presumably, the quicker burial of NO3 − at wetter sites would limit the amount of oxygen exchange that could occur in the deeper photic zone, and thus we observe little difference in  18 O values between skin layer and 1 m depth samples. In contrast, the greater photolytic activity at drier sites would enhance the imbalance between competing isotopic effects and produce distinctly lower  18 ONO3 values at 1 m depth 645 compared to the surface.
This proposed concept works well to explain the patterns observed in  18 ONO3, but it struggles to fully explain the similar patterns also observed in Δ 17 ONO3. Unlike  18 ONO3, photolytic mass loss is not expected to affect Δ 17 ONO3 values (McCabe et al., 2005). Thus, there is no isotopic effect counter-balancing the oxygen exchange brought by NO3 − re-oxidation and the cage effect, yet we still observe an unusually steady lowering of Δ 17 ONO3 values with depth in the pit data. As previously 650 mentioned, periods with high photolytic mass loss observed in the pit data (as indicated by the highest  15 NNO3 values) often have Δ 17 ONO3 peaks that are higher than would be expected compared to the coinciding  18 ONO3 values. In other words, Δ 17 ONO3 values decline from skin layer values to a lesser extent than  18 ONO3 values during times of high photolytic mass loss, which is in fact the opposite expected from our proposed "balanced competing effects" concept and difficult to explain mechanistically. Overall, this suggests that substantial complexities and unknowns still exist with regards to photic zone 655 Formatted: Superscript

Formatted: Superscript
Formatted: Superscript processes and NO3 − dynamics at the snow-atmosphere interface in Antarctica and resolving these issues will be necessary to properly interpret NO3 − oxygen isotopes archived in Antarctic ice.
In contrast, at the driest SMB sites (110-130 kg m −2 a −1 ) the regressions show that  18 ONO3 and Δ 17 ONO3 values are notably higher in the skin layer than at 1 m depth. Compared to nitrogen isotopes, it appears that a greater degree of photolytic mass loss (i.e., a drier site) is needed to observe a clear divergence between skin layer and 1 m samples for oxygen isotopic values. 660 This is a reasonable observation because the isotopic fractionation factors for oxygen isotopes are much smaller than for nitrogen, and we would expect that photolytic impacts become obvious much quicker for  15 NNO3 than for  18 ONO3 or Δ 17 ONO3.
An improved sampling method for the 1 m depth samples might produce stronger and more precise linear regressions with SMB −1 . The methodology used to collect 1 m depth samples during CHICTABA was to mix snow in a 5-10 cm thick layer 665 at 1 m depth. However, each seasonal isotopic cycle typically covers 30-50 cm depth in the upper snowpack as observed in the pit records ( Figure 4). As a result, each 1 m depth sample taken along the CHICTABA transect likely represents only part of full isotopic cycle. If a seasonal maximum or minimum happened to fall at 1 m depth, the resulting  15 NNO3,  18 ONO3, and Δ 17 ONO3 values could be offset from the true annual mean value by 20-50 ‰, 10-20 ‰, and 5-6 ‰, respectively ( Figure   4). For example, although the oxygen isotopic values for 1 m depth samples at CHIC-18 and CHIC-20 are much higher than 670 expected (see high values near 130 kg m −2 a −1 in Figure 5), their values are similar to winter maximum values and may simply be a result of seasonally-biased sampling. Future sampling of 1 m depth samples should ideally mix snow from at least a 50 cm range (i.e., from 1.0 to 1.5 m depth) to reduce the chance of seasonal bias and provide more accurate ω(NO3 -) and NO3 − isotopic values.

Conclusions 675
Our analysis of NO3in snow samples taken along the CHICTABA traverse transect reveals the environmental drivers of NO3concentration and isotopic variability at an unprecedented spatial resolution for a region of East Antarctica with intermediate SMB values (110-200 kg m -2 a -1 ). We find that seasonal geochemical cycles observed in atmospheric NO3are preserved in NO3buried in the snowpack. However, these cycles are clearly altered by post-depositional photolytic mass losschanges as shown by NO3 − isotopic values changes and calculated apparent isotopic fractionation factors that match 680 previous observations from elsewhere in Antarctica attributed to photolysis from elsewhere in Antarctica. We find no strong evidence that HNO3 volatilization or physical snow mixing substantially affected NO3after deposition. Additionally, we observe that the isotopic changes are greater at drier sites along the transect. This supports is consistent with photolysis as a causative factor in NO3isotopic change because slower burial rates at dry sites expose NO3to more cumulative photolytic radiation before the NO3is buried deeper beneaththan the reach of sunlight. 685 Because photolysis does not entirely wipe out the initial seasonal NO3cycles like it does at very dry sites in the Antarctic interior (e.g., Erbland et al., 2013;Shi et al., 2015), the interpretation of NO3is complicated in firn and ice cores from regions with intermediate SMB valuessimilar to the CHICTABA transect. If sampled at a high enough resolution, seasonal cycles in NO3concentration and isotopes may be recoverable far into the past, but these values are not representative of the exact NO3characteristics at the time of deposition. Photolysis will reduce ω(NO3 -) while increasing  15 NNO3 values and, to 690 a lesser degree, decreasing  18 ONO3 and Δ 17 ONO3 values from their initial atmospheric values. The amount degree of photolytic change is not likely consistent from year to year as it will depend strongly upon local SMB. Because regions in East Antarctica with intermediate SMB are generally found on the sloped transition between the high elevation interior plateau and low-lying coastal zone, katabatic winds drive intense irregular erosion and deposition of the snow surface (Frezzotti et al., 2002;Agosta et al., 2012). Additionally, and intrusions by atmospheric rivers and lower latitude moisture 695 Formatted: Space After: 6 pt Formatted: Space After: 6 pt bring infrequent but regular extreme accumulation events to these transitional regions (Gorodetskaya et al., 2014;Wille et al., 2021;Djoumna and Holland, 2021). This As a result, the regions have produces very high interannual SMB variability that will leads to very high interannual variability in interannual photolytic impacts that, and this variability makes it difficult or impossible to reconstruct precise initial atmospheric NO3characteristics at a seasonal resolution from NO3archived in firn and glacial ice. 700 However, relative to the interannual variability introduced by local SMB changes, interannual differences in initial mean atmospheric NO3isotopic values are likely to be relatively small, at least in the recent past. Atmospheric and skin layer NO3samples at Dome C are generally consistent year to year (Erbland et al., 2013;Winton et al., 2020), and atmospheric NO3observed at other sites have similar patterns and values (Wagenbach et al., 1998;Savarino et al., 2007;Frey et al., 2009). Regular sampling of atmospheric and skin layer NO3over one or more full years at an moderate intermediate SMB 705 site would greatly aid our comprehensive spatial understanding of NO3depositional dynamics, but unfortunately no permanent scientific stations exist in moderate intermediate SMB regions far from the coast. Atmospheric and skin layer NO3samples at Dome C are generally consistent year to year (Erbland et al., 2013;Winton et al., 2020), and atmospheric NO3observed at other sites have similar patterns and values (Wagenbach et al., 1998;Savarino et al., 2007;Frey et al., 2009). The most practical approach to NO3interpretation in firn and ice cores from intermediate SMB sites may be to 710 assume atmospheric NO3isotopic values can be considered "constant" when aggregated over multiple years. As a result, observed isotopic variability at this multiannual resolution will reflect changes in photolytic activity driven by local SMB, with stronger and more detectable effects at drier sites and more accuracy with more years of accumulation aggregated per sample.
Recognizing the importance of SMB in determining the isotopic composition of NO3 − may allow us to investigate other 715 drivers of isotopic change. Ice cores from intermediate accumulation regions can preserve seasonal ion and water isotope cycles well enough to produce highly precise chronologies (Buizert et al., 2015). Coupled with physical measurements of the ice core's volume and mass, we can model SMB based on physical changes in ice density and/or annual layer thickness (e.g., Fudge et al., 2016;Akers et al., 2022). This physical SMB reconstruction could then be used to remove the SMB signal from a parallel NO3 − isotope record, and the residual NO3 − isotopic variability should reflect past changes in other environmental 720 factors, such as insolation, total column ozone, snow optical properties, and atmospheric NO3 − sourcing and chemistry (Zatko et al., 2016;Cao et al., 2022;Shi et al., 2022b). This would be most effective for  15 NNO3 which has a more clear relationship with SMB (Akers et al., 2022) than  18 ONO3 or Δ 17 ONO3, but additional investigation into the mechanisms behind the apparent impacts of photolysis on oxygen isotopic composition is likely to provide valuable insight into past and present NO3 − dynamics as well. Additionally, ice cores taken from high SMB regions nearer the coast (i.e., regions with limited 725 photolytic mass loss) should better preserve the seasonal and interannual variability of atmospheric NO3and can provide an interesting comparison for ice core NO3records from drier inland settings.
Our NO3 − work as part of CHICTABA adds to the growing body of literature on NO3 − isotopes that point the way forward for future improvements to NO3 − interpretation in Antarctica. This knowledge is particularly critical for understanding the environmental changes archived in deep Antarctic ice cores, including new projects such as Beyond EPICA-Oldest Ice 730 (Lilien et al., 2021). Based on our CHICTABA findings and other recent studies (Erbland et al., 2013;Shi et al., 2015Shi et al., , 2018a, we highlight in particular the value of NO3 − isotopic profiles from snow pits in understanding the transition of NO3 − from the atmosphere into archived glacial ice. We argue for additional dedicated pit sampling of NO3 − isotopes with particular emphasis on extending profile depth below 1 m with paired chronological and snow density profiles to constrain SMB changes. Replication of pit profiles at individual sites will also improve our understanding of the natural range of local 735 spatial NO3 − variability. Expansion of atmospheric NO3 − monitoring beyond Dome C and Zhongshan stations will also help constrain spatial variability in seasonal NO3 − cycling. Finally, the potential spatial variability in snow optical properties and Formatted: Not Superscript/ Subscript photic zone depths remain one of the greatest unknowns in Antarctic NO3 − dynamics (France et al., 2011(France et al., , 2020Winton et al., 2020), and improved field observations and modeling will be required to precisely interpret NO3 − isotopic variability for paleoenvironmental reconstructions. (Akers et al., 2022)Ice cores taken from high SMB regions nearer the coast (i.e., regions 740 with limited photolytic mass loss) likely preserve the seasonal and interannual variability of NO3at deposition better and can provide an interesting comparison for ice core NO3records from drier inland settings. Overall, the NO3samples from the CHICTABA mission confirm the general understanding of NO3dynamics in East Antarctica that has developed in the past two decades and suggest that the understudied regions between the coasts and interior dome summits hold much untapped potential to improve our understanding of the Antarctic environment. The authors declare that they have no competing interests.