Air-snow exchange of HNO3 and NOy at Summit, Greenland

,


Introduction
Deposition of aerosol NO 3-(p-NO3-) plus gaseous HNO 3 (denoted TIN herein for total inorganic nitrate) is generally believed to be the major atmospheric sink for total reactive nitrogen (NOy = NO + NO 2 + HNO 3 + p-NO 3-+ N205 + organic nitrates), [e.g., Logan, 1983;Platt, 1986].Since NO and NO 2 regulate the production and destruction of 03 and play central roles in the complicated set of reactions determining the mixing ratios of OH and HO 2 radicals (thereby strongly impacting all of the main oxidants in the troposphere), ice core records of NO 3-(assumed to reflect deposition of TIN) accumulation could provide critical insight into past photochemical and especially oxidative states of the troposphere.In other words, if we understood how ice core records of NO 3-accumulation are reflecting reactive N oxide concentrations and cycling in the overlying atmosphere, we would b•e able to provide powerful constraints for We have shown previously that during summer NO3-is the dominant ion in snow accumulating at Summit, Greenland (the site where the Greenland Ice Sheet Project 2 and Greenland Icecore Project deep drilling programs have recently produced the longest ice core records of NO3-accumulation possible for the northern hemisphere), with NO 3-concentrations exceeding those of SO4--' (the next most abundant ionic species) by a factor of 5-10 [ Dibb et al., 1994].In air just above the snow we found that HNO 3 represents the dominant fraction of TIN at this site during summer but that the concentrations of TIN in the atmosphere are nearly always less than or equal to those of aerosol-associated sulfate (p-S04=) [Dibb et al., 1994].It is not immediately obvious that TIN should be incorporated into snow more than 10 times more efficiently than p-SO4--', so there appears to be a problem that could be described in two different ways depending on the point of view; (1) snow NO 3-is very high in relation to atmospheric TIN, or (2) snow SO4--is low in relation to p-SO4 =.Comparisons of SO4 = scavenging ratios (concentration in precipitation/concentration in aerosol) from a range of different sampling programs [Davidson, 1989] indicate that Summit is not characterized by inefficient removal of p-SO4 =, so we have interpreted these observations as indicating that there is too little TIN to readily explain the high concentrations of, NO 3' in Summit snow during summer [Dibb et al., 1994].
At least two hypotheses that are not mutually exclusive can explain the combination of very high NO 3' concentrations in surface snow and low TIN concentrations in the atmosphere above the snow.First, rapid loss of HNO 3 from near-surface air by deposition to the surface could readily account for these observations.Second, N species in addition to TIN could somehow be incorporated into snow and converted to NO3'.(Conversion to NO 3' could occur in the snowpack, or when samples are melted for analysis.)The first of these hypotheses is consistent with the prevailing view that TIN deposition is the only significant sink of tropospheric reactive N oxides and source of NO 3-in snow.The second directly questions the standard assumption regarding the source of NO 3' in snow but does not necessarily imply that deposition of additional N species to Greenland snow represents a sink of any importance on the global scale.In this paper we present concurrent measurements of concentrations and gradients of HNO 3, concentrations and fluxes of NOy, and concentrations and inventories of NO 3-in surface snow, designed to illuminate the air/snow exchange of reactive nitrogen at Summit in summer.

2.1.
Atmospheric Sampling 2.1.1.HNO3.Nitric acid in the gas phase at Summit was measured with the mist chamber sampling technique and subsequent ion chromatographic analysis, as described previously by Dibb et al. [1994Dibb et al. [ , 1996]].The mist chamber sampler concentrates soluble gases from the sampled airstream into a small volume of ultrapure water.A Teflon filter, in a custom made holder, is attached directly to the inlet of the sampler to exclude particles.Such a prefilter can cause either positive (due to dissociation of particulate NH4NO 3 trapped on the filter) or negative (due to reaction with basic particles on the filter) artifacts in the derived HNO 3 concentrations.For the case of Summit during summer we suspect that any bias is likely to be toward higher values, since the levels of sea salt and dust are extremely low [Bergin et al., 1995;Kuhns, 1997].It is not possible to estimate how large any contribution from dissociation of NH4NO 3 might be, but we assume that such a positive bias is generally very small on the basis of aerosol sampling at Summit that rarely found particulate NO 3' to be detectable [Dibb et al., 1994;Bergin et al., 1995;Kuhns, 1997].Of course, the dissociation of NH4NO 3 would yield low values for particulate NO3', so this is a circular argument.However, our main finding is that measured HNO 3 levels are problematically low, and correcting for any positive bias would accentuate this.
During the 1994 season all samples were collected 1.5 m above the snow surface at the base of an 18 m tall sampling tower.A total of 745 samples integrating for intervals of 30-50 min were collected between May 10 and August 10.In 1995, two mist chamber sampling systems were operated in parallel throughout the April 24 to July 8 sampling season.
One of the samplers was kept at 1.5 m above the snow for each of 402 sample collection intervals.The second system was deployed at 7 m above the surface to measure gradients during 230 of the intervals and at the 1.5 m height during all other intervals to establish the level of agreement between samplers when measuring the same air.The mist chamber that remained at 1.5 m ( the "stationary mist chamber" herein) was located 3 m away from the sampling tower that supported the other system ("mobile mist chamber") and the Harvard sampling inlets and sonic anemometer described below.
During 1994, mist chamber sampling was conducted around the clock during several 3-4 day long intensive sampling periods and for at least 12 hours on most other days.In 1995 the bulk of the sampling was conducted between 1000 and 2400.Concentrations of HNO 3, plus CH3COOH and HCOOH, were determined in a field laboratory within 12 hours (usually much less) of collection during both seasons.Errors in the measurement of sampled air volumes and in the ion chromatographic determination of concentrations in the stripping solution dominate uncertainty in the calculated atmospheric mixing ratios.We estimate this uncertainty to be of the order of 15-20% for the operating conditions at Summit, increasing to about 50% when the atmospheric mixing ratio of HNO3 falls below 0.1 nmol m-3 STP.Hampshire.They were melted in small batches immediately prior to analysis, which occurred within 2 months of the completion of the summer field seasons.HNO 3 concentration, with confidence in the implied gradients increasing as the concentration ratio departs farther from 1.02.

Nitrate in Snow
Nitrate in surface snow during the summer also shows considerable short-term variability (Figures la and 2a) that is not often readily related to variations in HNO3 just above the snow.

Carboxylic Acids in the Atmosphere
Both acetic (CH3COOH) and formic (HCOOH) acids are present in surface-level air at Summit in much higher concentrations than HNO3 (Table 1; also compare Figure 3 with Figures lb and 2b).Temporal variations of the two carboxylic acids are tightly linked during both background periods and passage of plumes (Figure 3).Correlation coefficients between CH3COOH and HCOOH were 0.81 and 0.92 for all samples collected 1.5 m above the surface in 1994 and 1995, respectively.The sources of carboxylic acids over the Greenland ice sheet are poorly understood, but this important question cannot be addressed herein.For the present purposes it is noteworthy that the concentrations of HCOOH and CH3COOH were comparable between the summers of 1994 and 1995 (Table 1).
Season-long summaries for the 2 years are compared to the 2 month long period of overlap.Concentrations are in nanomoles per gram.
samples collected in the 1992-1996 seasons is in progress [Slater et al., 1996].The focus of the present paper will be restricted to NO 3-in several intervals where air-snow comparisons are valid.
A key feature of the seasonal overview presented in Table 2 is that NO 3-is clearly the dominant soluble ionic species in summer snow.Mean and median concentrations of NO 3exceed those of the next most abundant ions (NH4+ and SO4 =) by more than a factor of 3.6.On average, NO 3-concentrations were more than twofold higher than the sum of NH4+ and SO4 = and exceeded the sum of all other measured ions.It should also be noted that the mean and median concentrations of NO 3differed only of the order of 10% between the two seasons (Table 2).,,,i,,,,1,,,,i,,,,i,,,,i,,,,i,,,,i,,,•l,,,,r,,,,   o

Discussion
It has long been supposed that deposition of HNO 3, by the processes of scavenging in snow and ice fog falling onto the surface and direct dry deposition to the snow surface, accounts for the majority of NO 3-measured in polar snow and ice, with only minor contributions from p-NO 3-[e.g., Legrand andDelmas, 1986, 1988;Steffensen, 1988;Laj et al., 1993;Wolff 1995].This hypothesis was supported by recent aerosol sampling campaigns at Summit during the summer, which generally found that concentrations of p-NO3-were extremely low while concentration of other ions, particularly SO4 =, were much higher, yet NO 3-was overwhelmingly dominant among soluble ions in the snow [e.g., Bergin et al., 1995].Determination of HNO 3 concentrations at Summit by several techniques has confirmed that HNO 3 is usually the dominant fraction of TIN (TIN = HNO 3 +p-NO3-) during summer, but the concentrations of TIN are quite low (Table 3).Dibb et al. [1994] pointed out that the ratio TIN/p-SO4--in the atmosphere just above the snow at Summit is generally <1, while the NO3'/SO4 = ratio in fresh summer snow often exceeds 10.The extensive sampling conducted in 1994 provided more insights into the magnitude and interrelationships of shortterm variations in the concentrations of HNO 3 and NO 3-in air and snow, respectively, at Summit.On many days there was a marked diurnal variation in HNO3, with peaks most common in early afternoon and minima during the late evening and early morning hours.Binning all data by time of day shows a subdued, but statistically significant, midafternoon peak (Figure 5a).The nighttime minima may reflect depletion of H NO3 beneath the strong, surface-based inversion that characterizes the Summit site during summer when winds are light [Dibb et al., 1992;Bergin et al., 1996].Recovery of HNO 3 concentrations during daytime could reflect downward mixing of free tropospheric air that had not lost HNO3, as the inversion lifts and weakens.However, the concentrations of carboxylic acids just above the snow show a similar diurnal pattern, with the potentially critical difference that their peak concentrations, and initial increase from nighttime minima, occur several hours before those of HNO3 (Figure 5b).Such an offset in timing would not be expected if downward mixing of air from aloft was the primary process responsible for daytime peaks of all three gases.

An intuitively attractive explanation for the simultaneous observation of relatively high NO 3-concentrations in surface
We have previously shown that the carboxylic acids can be rapidly lost from surface snow at Summit and have suggested that degassing of these acids could be an important proximate source for air just above the snow [Dibb et al., 1994].Thus it appears that we must consider degassing of carboxylic acids from the snow as a potential contributor to their daytime peak concentrations and cannot immediately rule out a similar surface source of HNO 3 on timescales of hours.Simultaneous sampling of air and snow was expected to provide some constraints on the magnitude of HNO 3 flux into or out of the snow.During the first 2 days of an intensive sampling period June 9-12, very clear daytime HNO 3 peaks with amplitudes of about 1 nmol m-3 were observed (Figure 6b).If we assume that all of this HNO 3 came out of, then redeposited to, the surface layer of snow, there should be changes in the NO 3-inventory of the surface layer that are antiphase with those in the gas phase, yet little or no change was seen in NO 3-inventory during this period (Figure 6a).However, a 1 nmol m-3 increase of HNO 3 in a well-mixed 100 m column below the inversion would require a loss of only 0.01 nmol NO 3-cm-2 from the snow over a period of a few hours.Inventory changes this small are nearly impossible to detect given the inherent small-scale spatial variability of surface snow [Dibb, 1996], as reflected in the standard devi•ation of adjacent replicates in Figure 6a.A more realistic scenario might relax the assumption of a well-mixed column (allowing for decreasing concentrations higher above the presumed surface source), which would require an even weaker source, such that losses could never be identified by repeated sampling of the surface snow layer.We conclude that even high-resolution (every 3 hours) sampling of surface snow does not help to discriminate between the snow and downward mixing of air aloft as the source of daytime HNO 3 increases at Summit.

4.2
The .By neglecting the resistances other than aerodynamic we calculate an upper limit for V a .Unfortunately, few data are available from the sonic anemometer during the mid June period of interest.However, throughout the season such upper limit estimates for Va at Summit were generally below 1 cm s -I and never exceeded 1.5 cm s-1.Furthermore, Johansson and Granat [1986] suggested that the other resistances can not be neglected when considering deposition of HNO 3 to cold snow, since they found that R c controlled V a at 0.4 cm s -1 for temperatures as warm as -8oC.Aerodynamic resistance became the controlling .factoronly when the snow approached the freezing point.It thus appears difficult to attribute the increasing NO3inventory in snow (Figure 7a) to deposition of HNO 3.
Clear midday peaks in HNO3 on June 13, 15, and 16 (Figure 7b) could reflect an upward flux during part of these days (see discussion of the 1994 case study above).We measured HNO 3 vertical gradients by operating mist chamber samplers at 1.5 and 7 m above the snow surface for much of 1995, including the mid-June interval in Figure 7.For part of each day the concentrations of HNO 3 were significantly higher at 1.5 than 7 m (Figure 7c), consistent with an upward flux.Intervals of apparent downward flux were also seen each day.It thus appears that HNO3 can be rapidly exchanged from airto-snow and snow-to-air.We acknowledge that these gradient measurements do not provide insight into the magnitude of HNO 3 flux into or out of the snow, but we are confident that they do reflect the direction of exchange.Accounting for the facts that HNO3 deposition was not continuous for the entire 6 days, and that there were periods of loss, would require V d Harvard NOy eddy correlation system and its operation are provided by J. W. Munger et al. (manuscript in preparation, 1998) (JWM, 1998 hereafter).Briefly, reactive N species are converted to NO by catalysis on a heated Au surface with H2, and the NO is detected by 03 chemiluminescence [Bakwin et al., 1994; Munger et al., 1996].These NOy measurements are made at 4 Hz, so that fluxes can be estimated from the covariance with vertical wind velocity measured with a sonic anemometer.The heated Au converter was mounted at 17.5 m on the tower, adjacent to the sonic anemometer.No inlet was used before the converter, so that line losses of sticky compounds such as HNO3 were minimized.Detectors, pumps, controllers, and data acquisition systems were all located in a covered trench at the foot of the tower.This system was operated at Summit for part of the 1994 season, but problems with icing of the NOy sampling line (due to oxidation of H 2 in the convertor) and local pollution (presumably mainly NO x from vehicles and a portable generator operated at the GISP2 drill dome within the clean air sector 500 m from the tower for several weeks) resulted in relatively sparse, discontinuous periods of valid data.The limited results from the Harvard instrument in 1994 will not be discussed herein.Additional heaters on the NOy sampling line, and more favorable wind directions during periods when other experiments needed to operate generators at the dome, were made during 230 thirty to fifty minute intervals between May 5 and July 8, 1995.Measurements of HNO 3 at the two heights (gradient mode) were made for slightly more than half these intervals.Occasional loss of data from the anemometer, mainly due to riming of the sensor during overnight ice fog events, and periods when wake effects from the tower invalidated the air motion data reduced the number of common intervals that included estimates of both HNO 3 gradients and NOy flux to 64. during the forementioned intensive sampling periods.In 1995 a single sampling was conducted each day, generally between 0800 and 1000.In both seasons three adjacent replicate samples were collected during each sampling event.All samples were melted in the high-density polyethylene (HDPE) collection bottles in the field laboratory; aliquots were transferred into HDPE sample tubes that fit our autosampler and immediately refrozen.These aliquots remained frozen during shipment back to our laboratory and during short-term storage in New

Figure 2 .
Figure 1.Concentrations of (a) NO3' in surface snow and (b) HNO3 1.5 m above the snow at Summit, Greenland, during summer 1994.
Concentrations As a result of the problems noted above, NOy data are available only for the 1995 season.The similar concentrations of HNO3, carboxylic acids, and NO 3-in snow between the two seasons suggest that 1995 was not an anomalous summer, so we assume the observations discussed below are representative of recent summers.The limited NOy concentration and flux data from 1994 are within the ranges observed in 1995, lending support to this assertion.During 1995 the concentration of NOy decreased by about a factor of 5 from early May to early July (Figure 2c).Decreasing NOy concentrations through the season drove an increase of the HNO3/NOy ratio from values well below 0.5% in early May to sustained values of about 1% in early July, with occasional short peaks in the range of 7-9 % (Figure 4).Mean and median values for the HNO3/NOy ratio at Summit in summer 1995 were 1.3% and 0.8%, respectively.

Figure 3 .Figure 4 .
Figure 3. Concentrations of acetic and formic acids 1.5 m above the snow at Summit in the (a) 1994 and (b) 1995 seasons.Note the offset of the two y axes to separate the curves.

*
References: 1, Silvente [1993]; 2, Dibb et al. [1994]; 3, this work; 4, Bottenheim et al. [1986]; 5, Bottenheim et al. [1993]; 6, Jaffe et al. [1991]; 7, Honrath and Jaffe [1992]; 8, Sandholm et al. [1992]; 9, Singh et al. [1994] decrease with altitude, but the levels at Summit are considerably lower than free tropospheric measurements made near the 3 km elevation of the ice sheet.As a result the ratio HNO3/NOy at Summit is also notably low, such that HNO3 accounts for a smaller fraction of reactive N at Summit than at other investigated locations.When measured, NO x has been a minor fraction of reactive N at all high northern sites (Table3), so we assume the same is true at Summit) is an upper limit estimate of the possible offset between NOy measurements and the concentration of reactive N oxides at Summit.The conclusions we will draw in this paper would not be negated even if the NOy data overestimated reactive N concentrations by several hundred parts per trillion by volume throughout the season.
summer snow within the first year after deposition suggests that snow may be a source of TIN at times rather than a continuous strong sink[Dibb et al., 1994].Second,Bergin et al. [1995] found that deposition of snow and ice fog could account for 99% of the NO 3-that accumulated in surface snow during summer 1993, leaving little room for a large contribution from dry deposition of HNO3.Further, modeling of the deposition of soluble species by ice fog suggested that PAN and related species could be making contributions to the flux of NO 3-in fog that rivaled the incorporation of HNO3[Bergin et al., 1996].WhileBergin et al. [1996] admit that their calculations are illustrative rather than quantitative, largely because of the complete lack of any direct measurements of PAN at Summit, they point out that similar processes in clouds forming snow could make PAN an even larger player in the snow NO 3

Figure 5 .
Figure 5. Mean concentrations of (a) HNO 3 and (b) carboxylic acids binned by time of day during the 1995 season.The bin widths were chosen to give approximately equal numbers of data points in each bin.The vertical error bars represent the standard error of the means, and the horizontal bars represent bin width.
Figure 7.A 1995 case study of relationships between (a) NO 3-in surface snow, (b) HNO 3 in air above the snow, and (c) HNO 3 gradients above the snow (see text).The same layer of snow remained at the surface throughout the period.The dashed line in Figure 7a is a linear least squares fit to the increase of inventory over time (0.07 nmol cm-2 d-l).

Figure 8 .
Figure 8.Time series of (a) HNO3 gradients and (b) NOy fluxes during the 1995 season with (c) a scatterplot for all intervals when both parameters were measured at the same time.

27.9 Season-long summaries for the 2 years are compared with the 2 month long period of overlap. Concentrations are in nanomoles
per cubic meter.Abbreviations are as follows: s.d., standard deviation; n.d., no data.*NOy measurements began May 5.