seasonal variations in sulfate, nitrate and chloride in the greenland ice sheet: relation to...

Atmospheric Enuironment Vol. 23, No. 11, pp. 2403-2493, 1989. 0004-6981/89 $3.00+0.00

Printed in Great Britain. 0 1989 Pergamon Press plc

SEASONAL VARIATIONS IN SULFATE, NITRATE AND CHLORIDE IN THE GREENLAND ICE SHEET: RELATION

TO ATMOSPHERIC CONCENTRATIONS

C. I. DAVIDSON, J. R. HARRINGTON, M. J. STEPHENSON, M. J. SMALL, F. P. BOSCOE and R. E. GANDLEY

Departments of Civil Engineering and Engineering and Public Policy, Carnegie Mellon University, Pittsburgh, PA 15213, U.S.A.

(First received 16 May 1988 and infinalform 6 February 1989)

Abstract-Samples from three snowpits near Dye 3 in South Greenland have been used to study seasonal variations in contaminant transport from the atmosphere to the Ice Sheet. The snowpits cove; the years 1982-1987. The samples have been dated bv comuarina 6’*0 values with meteorological data from Dve 3. Airborne concentrations of SOi- over the I& Sh&t ha; been estimated for the dates corresponding to each snowpit sample by statistically analyzing data from several air monitoring stations throughout the Arctic, and computing average values from the appropriate stations. Seasonal variations in concentrations in air, concentrations in snow, and mass-basis scavenging ratios (concentration in snow divided by concentration in air) have been identified. Results indicate that concentrations of SOi- in the air show a strong peak in late February, resulting from long-range transport of mid-latitude anthropogenic emissions, while those in the snow show a broad peak in January, February and March with smaller seasonal variation overall. The smaller variation in the snow is attributed in part to the effect of riming, which results in more efficient scavenging during warm weather when airborne concentrations are low. The importance of riming is also supported by the annual cycle in scavenging ratio which peaks in mid-summer coincident with maximum temperatures. In agreement with previous estimates, dry deposition appears to account for lO-30% of the total SOi- in the snow. Concentrations of NO; in the snow show a strong peak in summer, natural material from the stratosphere as well as anthropogenic emissions transported from the mid-latitudes may be responsible. Concentrations of Cl- in the snow are maximum in January, with relatively high concentrations during October through March and a smaller peak in July. The winter peak is believed to reflect long- range transport (LRT) of marine aerosol from north Atlantic storms, while the summer peak is attributed to seaspray from nearby coastal Greenland. Riming also may influence the seasonal variations in NO; and Cl- in the snow.

Key word index: Sulfate, nitrate, chloride, precipitation scavenging, scavenging ratio, dry deposition, ice, snow, Arctic, Greenland.

1. INTRODUCTION

Several investigators have examined concentrations of chemical species in glacial snow and ice as a means of studying changes in atmospheric concentrations. Al- though a large number of contaminants have been examined, recent efforts have focused on acid SOi- and NO;. For example, Neftel et al. (1985) and Mayewski et al. (1986) have reported that concentrations of both of these species in Greenland snow have increased over the past several decades. The trends are attributed to increasing SO:- and NO; emissions from anthropogenic activities in the mid-latitudes.

While gross changes in atmospheric levels are known to be reflected in ice cores, a quantitative link between airborne concentrations and the glacial record has not yet been established. This is because the mechanisms of transport from the atmosphere to the snow surface are extremely complex. Wet deposition may dominate during certain times of the year at some locations. This category includes nucleation of particles during the formation of cloud droplets or ice

crystals, and scavenging of gases and particles by hydrometeors within and below clouds. Dry deposition may dominate at other times and locations. This involves wind eddy transport from the free atmosphere to the boundary layer above the snow surface, followed by Brownian diffusion, interception, or iner- tial transport across the boundary layer. Sedimenta- tion by gravity may influence particles that are sufficiently large. Other effects such as electrostatic or phoretic forces may also be important. The overall rates of transport by wet and dry deposition are dependent upon the mechanisms involved ahd are likely to be highly variable with space and time. A further complication is that the upper layers of snow may be redistributed by the wind, and contaminant migration may occur within the snowpack. The ways in which wet and dry deposition influence the glacial record have been discussed in detail elsewhere (Davidson, 1989).

In order to improve our understanding of air-to- snow transport, we previously conducted simultaneous air and snow sampling at Dye 3 on the south-

M mtt-.I 2483

central Greenland Ice Sheet (Davidson er ul., 1981. 1985). These studies enabled estimates of wet and dry deposition rates for SO:---, NO; and several trace metals for the time periods studied. Although inter- esting results were obtained, the experiments produced only a limited amount of data because of the difficulty of conducting real-time sampling during precipitation events in the Arctic.

In an effort to obtain a larger database, a pilot-scale snowpit study was conducted near Dye 3 in 1984 (Mayewski et al., 1987, Davidson et al., 1987a). Forty- eight samples were collected continuously down the snowpit wall from layers 5 cm thick. The samples were analyzed for SOi- and other anions, and also for 6 ‘*O; the layers were dated by comparing the 6 “0 values with pr~ipitation and temperature records at Dye 3. The resulting time series data for SOi- concentrations in the snow over a 2-year period were compared with airborne SOi- concentrations measured at coastal Greenland sites. This procedure yield- ed information on variations in wet and dry deposition rates of this species at Dye 3. It was found that precipitation scavenging ratios varied from m lO@-200 in winter to -20@400 in summer, and that dry deposition was responsible for roughly i&30% of the observed SOi- in the snow. The key assumption in the procedure was that airborne SOi- concentrations at Dye 3 are determined mainly by large scale meteorolo~cal phenomena. Therefore one would expect seasonal variations in airborne SOi- at the coastal Greenland sites to resemble those at Dye 3.

It was acknowledged in the pilot study that scavenging ratios thus determined could not be interpreted in a strict quantitative sense. This is because of expected differences in airborne concentrations at Dye 3 and the coastal Greenland sites, due to distances between the sites, differences in elevation, and locat sources at the coast such as the nearby marine influence. It was emphasized that only the seasonal pattern in scavenging ratios, and not the absolute values, could be used to help interpret glacial record data.

In this study, we extend the methods developed in the pilot snowpit program to two additional snowpits near Dye 3. One snowpit was excavated during July 1986, while the other was excavated during June 1987. Results are presented for SO:-, NO; and Cl- over the S-year period represented by all three snowpits.

The paper includes several sections. First, we discuss the snowpit sampling and dating procedures. Then we examine airborne concentration data for SO:-, NO; and Cl- from several Arctic sites, and consider the extent to which these data can be assumed representative of concentrations at Dye 3. Finally, we present concentrations of the three anions in the snowpit samples and compute scavenging ratios for SOi- based on the air and snow data. The results are interpreted by comparison with data from previous air and snow sampling efforts at Dye 3 and at other sites in the Arctic.

2. SNOWPIT SAMPLING AND DATING

The 1986 and 1987 snowpits were constructed 23 km southwest of Dye 3 at an elevation of 2560 m, near the route to the pilot study snowpit which was 40 km from Dye 3. The snowpits were located approxi- mately 1 km apart. In 1986, samples were collected continuously down the snowpit wall from layers 5 cm thick over a total depth of 2.25m. From each layer, two replicate samples were obtained for ion chromatography analysis, two for (5’“O analysis, and two for density determination in the field. Tn 1987, samples were collected continuously down the snowpit wall from layers 2 cm thick over a total depth of 2.9 m. From each layer, five samples were obtained for ion chromatography analysis, two for 6 ’ 8O analysis, and two for density determination in the field. To assist with dating of the snowpits, a limited number of I f samples were obtained from 10 cm layers (1986 snowpit) and from 5 cm layers (1987 snowpit) for analysis of Chernobyl debris. In addition, the snowpits were mapped stratigraphically. The field sampling procedures used in both years were the same as those used in the pilot snowpit study (Mayewski et al., 1987).

The samples collected for ion chromatog~phy analysis were kept frozen from the time of collection in the field until just before analysis at Carnegie Mellon University (CMU). A Dionex ion chromatograph with a fast run separator column was used to analyze the samples for SO:-, NO; and Cl‘. Each individual sample was subjected to at least two analyses to assess analytical uncertainty. The fi “0 samples were melted in the field immediately after collection. The meltwater from each sample was used to fill a small bottle in order to minimize headspace that might affect the oxygen isotope ratios in the sample. The fi I80 samples were analyzed by the Geophysical Isotope Laboratory at the University of Copenhagen (Dangaard et al.,

1973). Thesamples collected for analysis of Chernobyl debris were melted in the field and acidified. They were analyzed for L34Cs and i37Cs by intrinsic germanium y-ray detection at the University of Chicago. Strict contamination procedures were followed in all phases of field and laboratory work throughout the project (Davidson ef al., 1985, 1987a; Mayewski et al., 1987).

For each of the snowpits, the samples have been dated by comparing a curve of 6t80 vs depth in the snow with a curve showing average air temperature vs cumulative precipitation at Dye 3. Details of the procedure are discussed in the pilot study results (Davidson et al., 1987a). The only difference in procedure between the two studies is that average air temperature during each snowstorm has been used here, resulting in better agreement with the 6 ’ “0 data; average air temperature for all days from one snowstorm to the next was used in the pilot study. This method of comparing 6 * 8O data with air temperatures can provide only approximate dating. The range of dates established for each layer is uncertain by at

Seasonal variation in sulphate, nitrate and chloride 2485

least 2-3 weeks and possibly more than a month, as described earlier (Davidson et al., 1987a).

Results are shown in Fig. 1, which displays the 6 “0 and air temperature data. Note that a continuous record of 5 years is provided by the three snowpits, which cover June 1982-June 1984, June 1984-August 1986, and July 1984-June 1987, respectively. For each snowpit, the stratigraphic mapping showed melt layers coinciding with peaks in the 6 ‘*O values, verifying the locations of summer snow.

Concentrations of 134Cs and 13’Cs were at or near background in all except for one layer in the 1986 snowpit and two layers in the 1987 snowpit. In the 1986 snowpit, elevated concentrations were found in the layer 10-20 cm below the surface; values for lJ4Cs and 13’Cs were 2.0f0.8 and 6.2 & 1.4 pC kg-‘, respectively. In the 1987 snowpit, elevated concentra-

tions were found in layers 95-M cm and 100-105 cm below the surface. The former layer had concentrations of 0.69 + 0.14 and 2.3 + 0.2 pC kg- ’ for the two radionuclides, respectively. The latter had concentrations of 0.27kO.05 and 1.5kO.15 pC kg-‘, respectively. When these values arc used with layer thickness and snow density data, the total deposition of each ra~onuclide to the Ice Sheet at the location of sampling can be determined. It is of interest that the total deposition as determined from the 1986 snowpit is much greater than that determined from the 1987 snowpit. For ‘34Cs, the difference is a factor of 3.6. For ‘37Cs, the difference is a factor of 2.7. These differences may be due in part to variations in precipitation over small spatial scales: the amount of radioactive material deposited is a strong function of the amount of

precipitation occurring at a particular site when the

I u I

ioo I I i t-40 200 10 100 200 300 I

I Depth of Snowpit Sahple Below Surface, cm

I 4’ ‘. I . 1 ,‘I.‘ 11 ’ I , 1 0 1 IdO 200’ sob 0 '1od

/ Snow Accumulation Backward in Time, cm / 1 / I

-40 0 100 200 300

Depth of Snowpit Sample Below Surface, cm

t‘, , .“‘,’ ,.- 0 ’ 100’ ’ 200 ’

Snow Accumulation Backward in lime, cm

Fig. 1. 6 ‘*O as a function of depth in the snowpits, compared with air temperature and snow accumulation data from meteorological records at Dye 3. The snow accumulation axis has been scaled to optimize agreement with the 6lsO data; scaling is uniform between each pair of arrows. Note that the surface of the snow during excavation of the 1986 snowpit corresponds to the 88 cm depth in the 1987 snowpit. The arrows marked with a %,, refer to the date of the Chernobyl accident and to layers of snow containing Chernobyl

debris.

Chernobyl plume passed Redistribution of the surface snow by wind also may have affected concentrations of Chernobyl debris in the Ice Sheet (Dibb, 1989).

The layers in the snowpits that contained Cherno- byl radioactivity are marked with a “c” on the 6 l*O curves in Fig. 1. The date of the Chernobyl accident in late April 1986 is indicated on the air temperature curves in the figure. The difference in positions of the “Cs” on the 6’*0 and air temperature curves for each snowpit reflects the time of transport of Chernobyl emissions to Greenland as well as uncertainty in the dating procedures. Details of the atmospheric transport from Chernobyl to the Greenland Ice Sheet have been discussed by Davidson et al. (1987b).

3. ESTIMATION OF AIRBORNE CONCENTRATIONS

Only a limited number of airborne con~ntration me~urements of SO:-, NO; and Cl- have been obtained at Dye 3. However, a number of monitoring stations throughout the Arctic have been providing airborne data for several years. These data show that a definite annual cycle exists in Arctic airborne concentrations. In this section, we explore similarities in the statistical features of the datasets from several of these monitoring stations. If high correlations are found among the datasets, it may be assumed that gross features such as the magnitude and phase of the annual cycle reflect Arctic-wide phenomena, and thus may be applied qualitatively to Dye 3.

Airborne concentration data have been obtained from three sites in the Canadian Arctic (Alert, Igloolik and Mould Bay), three sites on the coast of Greenland (Nord, Thule and Godhavn), and two sites in the Scandinavian Arctic (Ny-Alesund and Bjsrnoya) as part of established sampling networks. The networks in these three areas are described by Barrie and Hoff (1985), Heidam (1984), and Joranger and Ottar (1984), respectively. Locations of these sites relative to Dye 3 are shown on a map presented previously (Davidson er al., 1987a). Data are available beginning in 1979 or 1980 and continue for 3-7 years, depending on the station. Concentrations of SOi- are available from all eight sites, while NO; data are available only from the Canadian stations. Cl- data are provided by the Canadian and coastal Greenland sites.

The original data cover daily, 3-day, 4-day, or l-week sampling intervals. Time series analysis of the data show a deterministic component characterized by a strong annual cycle as well as a stochastic component at each site (Davidson et al., 1987a). Since the stochastic component is most likely due to local influences occurring on time scales of a few days or less, an averaging time sufficiently long to minimize these local influences is needed. Furthermore, the snowpit dating technique provides resolution at time intervals of 2-3 weeks at best. For these reasons, the data for each site have been used to obtain 2-week average values.

The similarity of the airborne data at the different sites was evaluated by computing least squares regressions between the natural logarithms of the 2-week average values at each pair of stations. Logs have been used since the overall data tend to follow a lognormal distribution, and the seasonal variation of each time series appears to follow a log sinusoid curve (see below). The number of data pairs in each correlation ranges from 61 to 124.

For SO:-, the regressions show r2 values of 0.6-0.9 for all pairs of stations except those including either Bjornsya or Godhavn (r2 =0.3-0.5), and a value of 0.4 for Ny-dilesund vs Thule. Bjornoya is most likely influenced by different air masses than the remaining Arctic sites due to proximity to the European con- tinent. The Godhavn dataset has local contamination problems in summer when concentrations are small (Heidam, 1984) and has been excluded from the pilot snowpit study (Davidson et af., 1987a). For the present work, Bjornsya and Godhavn have been excluded, except for later use of the Godhavn values to assess marine influence. Data for each 2-week interval from the remaining six sites have been averaged to produce the airborne concentration time series shown in the center of Fig. 2. Only data from January 1982 onward are shown. The full time series for concentration C, beginning in 1979 has been used to derive best-fit sinusoid curves of the form y= a + A cos(wt + (p), with y = C,, y = Cz for several values of n, and y = log C,. The highest correlation is obtained for the last of these formulations (r’ =0.78). This best-tit log sinusoid curve is used to extend the time series forward past the end of the data (May 1987) to match the most recent snowpit samples in mid-June. Note that the curve peaks in late February. It is of interest that a best-fit log sinusoid curve through airborne SOi- data from Barrow, Alaska (Rahn and Shaw, 1982) shows a very similar pattern, with a peak in late February nearly coincident with the six-station average curve.

For NO;, least-squares linear regressions between the three pairs of Canadian stations yield r2 values of 0.5-0.7. The concentrations for each 2-week interval at the three sites have been averaged; the best-fit log sinusoid curve through the average values yields r2 = 0.68, with a peak in late February. The values of r*

for the three pairs of stations suggest that large scale meteorology is of primary importance in determining seasonal variations in concentration. This is further substantiated by the annual concentration maximum nearly coincident with that of SO:-. However, the cellulose filters used for sampling at the Canadian sites are believed to be unreliable for measurement of NO; due to artifact problems (Barrie, personal communi- cation). Thus the maximum airborne NO; concentration may indeed occur in late winter, but the relative contributions of various gaseous and particulate NOj species are unknown, as are the absolute values of total NO; concentration. For this reason, the airborne NO; data have not been used to compute scavenging ratios

0 1500

I

P 1000 F g a VI 500

0

Fig. 2. Measured concentrations in the snowpit samples, estimated concentrations in the air, and computed scaveng-

ing ratios for SO:-.

1982 1963 1984 1985 1986 1987 range of dates determined for each layer. The difference between the two analyses of each sample from the 1986 and 1987 snowpits averages 3% of the mean value for SO:- and NO, and 10% for Cl-. The coefficient of variation (standard deviation of the sample divided by the arithmetic mean of the sample) of the five samples for each layer in the I987 snowpit averages 20% for SO:-, 18% for NO; and 65% for Cl-. These uncertainties reflect the difficulty of collec- ting snow within well-defined 2-cm layers, where the concentrations may be quite different from one layer to the next. By far the greatest variability in adjacent layers is observed for Cl-.

For Cl-, small positive correlations have been found among the data from all pairs of stations (r2 c 0.3). Examination of the data reveals frequent episodes of high concentration during both winter and summer at Igloolik, Mould Bay, Godhavn and Thule. In contrast, Alert and Nord show frequent periods of high concentration in winter, but uniformly low concentrations in summer. The difference in summer con~ntrations between these two categories of sites most likely results from the melting of sea ice near the first four locations; the ocean is perpetually frozen at Alert and Nerd. The high concentrations in winter at all six sites reflect LRT from oceanic areas farther south. Spectral analysis of the data supports this hypothesis, showing a much more pronounced annual cycle at Alert and Nord compared with the other sites. Best-fit log sinusoid curves indicate a late December peak at Alert (r2=0.35) and a late October peak at Nord (t2 =0.50).


These concentration patterns suggest that numer- ous factors influence airborne Cl- in the Arctic. Unlike SOi- and NO; where seasonal variations are influenced mainly by large scale meteorology, airborne Cl- levels depend on proximity to the ocean, sea ice cover, and local winds as well as long-range transport. Herron (1982) has compared Cl- in snow from several Arctic and Antarctic sites to show that Dye 3 is not influenced by seaspray to the same extent as sites at Iower elevations closer to the coast. Further- more, marine-derived SO:- (computed as 0.14 Cl-) averages only 3% of total SOi- for the snowpit samples in the current study, with very few values greater than 15%. This suggests that airborne data from coastal sites such as Igloolik, Mould Bay, God- havn, and Thule cannot be used to estimate airborne Cl- at Dye 3. The extent to which concentrations at Alert and Nord resemble those at Dye 3 is unknown due to differences in latitude and elevation. As a result, the airborne Cl- data have not been used to compute scavenging ratios.

4 CONCENTRATIONS OF ANIONS IN THE SNOWPIT AND

COMPUTATION OF SO- SCAVENGING RATIOS

Concentrations of SO:-, NO; and Cl- in the snowpit samples are graphed in Figs 2, 3 and 4, respectively. The concentrations shown represent the averages of l-5 samples collected from a particular layer. Data from all three snowpits are shown, representing 823 samples from 227.layers (some samples were lost in transit). The position of each datapoint with respect to the x-axis is at the midpoint of the

The concentrations of SO:- in the snow and air have been used to calculate the mass-basis scavenging ratio:

where C,=concentration in snow (ngg- “), C, = concentration in air (ng m - 3), and pa = density of air (gm- 3). Values of Wfor each snowpit layer are plotted at the bottom of Fig. 2. Four values are off-scale and therefore not shown.

It is significant that the above calculations use total rather than non-marine SO:-. The fraction of air-

Nitrate

1982 1983 1984 1985 1986 1987

Fig. 3. Measured NO; concentrations in the snowpit

samples.

Chloride

1982 1983 1984 1985 1986 1987

Fig. 4. Measured Cl- concentradons in the snowpit samples.

borne marine-derived SOi- at Igloolik, Mould Bay, Godhavn and Thule averages 18%, lo%, 19% and 6% of the total at these four sites, respectively, with some values exceeding 25% in summer. Marine- derived SOi- at Alert and Nord averages 3% of total SOf- with values rarely exceeding 15%. The marine influence at Ny-Wlesund is unknown since Cl- data are unavailable. When Godhavn is excluded, the overall average marine influence on the airborne data is small; as discussed above, marine-derived SOi- in the snowpit samples is also small. Therefore total SOi- data have been used in Equation(l).

5. DI§CU!SS~#N

The snow concentration data in Figs 2-4 have been analyzed in two ways. First, the data for each month have been averaged in order to assess seasonal variations. Second, the concentrations have been used to determine best-fit log sinusoid curves for comparison with the airborne concentration models using this function.

The monthly average concentrations in the snowpit samples are shown in Fig. 5. SOi- shows a broad peak during the first 3 months of the year, with a

Sulfate

Chloride !

Fig. 5. Measured concentations of SOi- , NO; and Cl- in the snowpit samples, averaged by month. The number of samples averaged within each month ranges from 7 to 40. The error bars show values one standard error of the mean above and below the arithmetic mean

concentration.

steady decrease continuing to the end of the summer. In contrast, NO; peaks in the summer. Cl- peaks in January, with smaller relative maxima in March, July

and October; there is considerable variability from month to month and within each month.

The best-fit log sinusoid curve for SOi- shows a peak in mid-March (r* =0.21). The timing of the peak is slightly different from that in Fig. 5, primarily because the sinusoid imposes symmetry on the annual cycle. The best-fit log sinusoid for NO; shows a peak in late May (r* =0.12). The Cl - data cannot be fit with a log sinusoid function (r* w 0). Despite the high r2 values when airborne SO:- and NO; data are fit to log sinusoid curves, as reported earlier, the low values for the snowpit samples imply that this function is not well-suited to describe concentrations in the snow. We now discuss possible reasons for the observed seasonal variations in Fig. 5. We also discuss the influence of dry deposition on the observed concentrations in the snowpit samples.

SOi- is known to be a major constituent of Arctic haze. The February/March annual maximum in Arc- tic haze is well documented, attributed to the polar front extending into the mid-latitudes in winter, pro- moting rapid atmospheric transport from anthropo- genie source regions to the Arctic. Also important are the relatively low precipitation rates during the Arctic winter which result in minimal scavenging of atmospheric ~ont~nants during long-range transport


(Rahn and Shaw, 1982). Trajectory analysis and air sampling at Dye 3 have shown that these phenomena influence the south Greenland Ice Sheet (Davidson et al., 1985).

The seasonal variation in airborne SOi- in the Arctic is roughly preserved in the Greenland Ice Sheet. Concentrations in the air peak in late February, while those in the snow show a broad maximum at about the same time. However, the amplitude of the variation in the snow is much smaller than that in the air. The difference in amplitudes is partly the result of the process of riming during the formation of snow. This refers to the growth of ice crystals, which eventually form snowflakes, by collisions with liquid water droplets in clouds (rime) as well as by diffusion of water vapor. Rime contains far greater concentrations of atmospheric contaminants than unrimed ice crystals formed solely by water vapor diffusion (Borys et al., 1983). Thus one would expect precipitation scavenging by rimed snow, which occurs during relatively warm weather, to be more efficient than scavenging by unrimed snow. Since airborne concentrations are minimum in summer, the seasonal variation in riming will result in more uniform concentrations in snow throughout the year.

The importance of riming is suggested by monthly average values of SOi- scavenging ratio shown in Fig. 6. Much greater values are seen to occur in summer: the scavenging ratios average 690 + 70 (arithmetic mean+ 1 standard error of the mean) for the months of June, July and August, but only 200 +_ 20 for the remaining months. In comparison, Scott (1981) has reported values in the range 150-3000 for rimed snow and 20-400 for unrimed snow for SO:- over Lake Michigan. Scavenging ratios resembling the Dye 3 winter values have been reported for SOi- and other sub-pa species in the Arctic during times when the snow is expected to be unrimed. For example, Semb et al. (1984) report an average value of about 300 for SO:- on Spitsbergen during February and March. Noone and Clarke (1988) report values for elemental C averaging about 100 during March and April in nothern Sweden. Barrie et al. (1985) give an average value of 240 for the scavenging of airborne acidity in

Sulfate

01 , , ) , , Jan Feb Mar Apr May Jun Jul Aug Sep Ott Nov Dee

Fig. 6. Computed scavenging ratios for SO:-, averaged by month. The error bars are defined in the same manner

as in Fig. 5.

the Canadian Arctic for a dataset weighted toward winter conditions. The importance of riming is also supported by measurements at Dye 3 during May 1985 as part of the present study, when a heavy coating of rime formed on a fresh snow surface. The concentration of SOi- in the rime averaged 330 ng

g -l, compared with 95 ng g-’ in the snow. As discussed in the introduction, the scavenging

ratios are problematic in that the true airborne concentrations over the Ice Sheet are unknown. A particular problem is that Dye 3 is at a much higher elevation than the sea-level stations providing air monitoring data. To assess this problem, Fig. 7 shows airborne concentrations of SOi- plotted as a function of date during the year from 25 air sampling measurements at Dye 3 (Davidson et al., 1985), including some previously unpublished values. Several years of spring /summer data are represented. Also shown is the appropriate segment of the log sinusoid curve representing a best fit to the data in the center of Fig. 2. For this limited set of data, the concentrations at Dye 3 are consistent with seasonal variations observed at the sea-level sites. It is also noteworthy that the Dye 3 airborne concentrations average 0.6 f 0.08 (arithmetic mean f 1 standard error of the mean) of the sea-level log sinusoid values at corresponding dates. The tend- ency toward lower concentrations at higher elevations is in agreement with data collected during the first Arctic Gas and Aerosol Sampling Program (AGASP- I). For example, Hansen and Rosen (1984) report aerosol sulfur concentrations of 500-1000ngm-3 at elevations of 24 km, with an average of 1000 ng m- 3 at ground level. Radke et al. (1984) have measured SOi- concentrations averaging 1800ngm-3 in haze layers up to 5km above the surface, compared with 2600 ng mV3 at the ground. These data are from Bar- row, Alaska in March-April 1983. In contrast, there is recent evidence from AGASP-II that some particularly polluted haze layers aloft may contain contaminant concentrations in excess of those at the surface. It must also be acknowledged that concentrations at Dye 3 may be quite different from those at comparable elevations over sea-level sites, due to the influence of Ice Sheet topography. All of this information suggests that although the seasonal variations in airborne concentration and scavenging ratio at Dye 3 are qualitatively described by Figs 2 and 6, the absolute values of either parameter cannot be established with confidence.

Some of the airborne concentrations shown in Fig. 7 have been obtained during storms where SOi- levels in the snow have also been measured. All of these storms involved clouds that were at or near the surface. The resulting scavenging ratios computed from Equation (1) for simultaneous or near simultaneous air and snow sampling are shown in the lower part of the figure. Although there is a considerable amount of scatter, the seasonal pattern is consistent with that shown in Fig. 6, namely greater scavenging ratios during warmer weather.

2490 C‘. 1. DAVID%)& rl ul.

1200 , I

JUI *w

3000

.g

z 2000

F '6 E 2 1000 fJ4 I -. . .

. . , - 0, : I . I I 1

Mav JUll JUl Aw

Fig. 7. Measured SOi- concentrations in the air, and computed scavenging ratios based on simultaneous and near-simultaneous air and snow sampling at Dye 3. The data are from separate experiments between 1979 and 1987. The segment of the best-fit log sinusoid curve representing the airborne data in Fig. 2 is

shown in the upper graph.

Seasonal variations in NO;

The strong summer NO; peak in Fig. 5 is consistent with data from ice cores collected previously in Greenland. For example, Risbo et’ al. (1981) report marked seasonal variations in NO; in the Crete ice core. In addition, Herron (1982) has used summer NO; peaks in the deep Dye 3 core to define annual layers.

It is of interest that the airborne NO; data from the three Canadian Arctic sites show minimum concentrations in summer. If the seasonal variations in airborne NO; at Dye 3 resemble those at the Canadian sites, the maximum concentrations in the snow occur at a time of minimum concentrations in the air. This may reflect greatly increased scavenging in relatively warm weather, more than compensating for the low airborne levels. Alternatively, the airborne concentrations at Dye 3 in summer may be greater than expected on the basis of the Canadian site data. Finkel et al. (1986) offer two hypotheses for the summer NO; peaks in the snow based on possible high summer airborne concentrations at Dye 3. The first is that local phenomena such as photochemical NO; aerosol produc- tion may be prevalent in summer. The second hypothesis is that NO; may be reaching Dye 3 through high altitude transport, perhaps from the stratosphere. The latter hypothesis appears particularly plausible con-

sidering the high elevation of Dye 3: NO, from the stratosphere as well as NO, produced by lightning are believed to be important sources of NO; in remote regions (Logan, 1983). Neftel et al. (1985) have suggested the importance of these sources in influencing NO; in the Greenland Ice Sheet, while Legrand and Delmas (1986) conclude that mid-latitude lightning is responsible for much of the NO; in Antarctic snow. Nevertheless, enhanced scavenging by riming is likely to be significant, as the May 1985 Dye 3 measurements show 380 ng g-’ NO, in rime but only 180 ng g ’ in snow.

Finkel et al. (1986) note that NO; seasonal variations in the Ice Sheet at Dye 3 have changed in recent years. While a fairly distinct summer peak is observed prior to the 195Os, more recent samples show an increase beginning in late winter with a broad peak extending into summer. The authors attribute this change to LRT of anthropogenic NO; during winter and spring. Similar although less dramatic changes are seen for SO:-. Results of the present study suggest that scavenging is not as efficient during the winter and spring periods of long-range transport as during the summer months. Thus the broad SOi- and NO; peaks during the first half of the year observed by Finkel et u[. probably reflect generally increasing scavenging efficiencies coincident with decreasing airborne concentrations.

Seasonal variations in Cl-

The Cl- snow data in Fig. 5 may be interpreted considering the airborne concentrations at Alert and Nord. As discussed earlier, these two sites show frequent periods of high concentration in winter, primarily between October and March. It is of interest that airborne Cl- peaks several weeks earlier than airborne SOi- and NO;; this is probably due to LRT of marine aerosol from north Atlantic storms in late fall and early winter. Cl- emissions within the Arctic are minimal at this time since the seas are frozen. Later in winter, when seas in the Arctic are still frozen, the storms move farther south as the Arctic air mass becomes more stable (Rahn, personal communi- cation). This results in decreasing airborne concentrations through the late winter and spring. The high concentrations in the snow in winter probably reflect the high airborne concentrations resulting from LRT. However, the smaller peak in summer seen in Fig. 5 may be due to seaspray emissions from nearby coastal Greenland locations. No Cl- data are available from the May 1985 Dye 3 samples to assess the possible importance of riming, although this mechanism may enhance scavenging of Cl- in summer.

It must be cautioned that the Cl- data presented here are rather uncertain. This is evidenced by the large coefficients of variation in the analyses of replicate samples taken from each layer, and by the significant variability in concentrations of Cl- in adjacent layers. Therefore the conclusions regarding this species are especially tentative.


The influence of dry deposition

The discussion of the importance of riming implies that scavenging by nucleation and by in-cloud droplet growth are dominant mechanisms by which contaminants are incorporated into the snow. Several investigators have suggested the importance of such mechanisms in the polar regions (Junge, 1977: Shaw, 1980; Barrie, 1985; Borys, 1989). However, dry deposition may also be significant, particularly when precipitation rates are low and unrimed snow predominates.

Following the procedure in the pilot study (David- son et al., 1987a), the concentration of a contaminant in a snowpit sample due to dry deposition is determined as:

c Vddry cat sdry = -

PP

where v&r is the dry deposition velocity (cm s- ‘1, t is the time of surface exposure (s), pI1 is the snow density (gem-“), and d is the depth of each sample (cm). Both vddry and C, are assumed to be constant over the exposure period. The fraction of a particular species in each sample due to dry deposition is computed simply

as Csdry/Cs~

Equation (2) has been applied to the SO:- data from the three snowpits using vddry = 0.05 cm s- l. The true dry deposition velocities are unknown, but this value has been used previously for sub-~ species over snow surfaces (Davidson et al., 1987a; Davidson and Wu, 1989). Results show wide variations in C&&$, ranging from near zero to > 1. This reflects uncertainties in the dates of each sample, since long surface exposures thought to apply to one layer may actually apply to an adjacent layer. The problem is minimized by applying Equation (2) to longer periods involving several layers. When used with annual accu- mulations in the snowpits (June of one year to June of the next year), the values of C,,,& range from 0.2 to 0.5, with an overall average of 0.3.

These results indicate that dry deposition may have some influence on SO:- concentrations in the Ice Sheet. If values of vddry are in range 0.02405 cm s- 1 as assumed in the pilot snowpit study, dry deposition accounts for roughly l&30% of the total SOi- in the snowpits, consistent with previous estimates (David- son et al., 1985,1987a). Precipitation rates are lower at more northern sites in the Arctic, and thus dry deposition is expected to be > l&30% at those sites. This is consistent with the observations of Herron 619823, who noted an inverse relationship between SO:- concentrations and snow accumulation rate at several Arctic and Antarctic sites; of the 11 sites considered, Dye 3 had the highest accumulation rate and the lowest SOi- concentration. Herron reasoned that the total SOi- flux should be relatively insensitive to precipitation rate where dry deposition is important, leading to lower concentrations at sites with higher accumulation rates. He attributed his findings to two possible dry de~sition mechanisms, namely gaseous

reactions with freshly fallen snow and gas-to-particle conversion followed by deposition at the snow surface.

It must be noted that precipitation is uneven throughout the year at Dye 3, and thus the importance of dry deposition is likely to vary with season. The snowpit records indicate that only 30% of the annual precipitation falls during December-May. The relatively low pr~ipitat~on rates, small amount of riming, and high airborne con~ntrations of contaminants all favor the importance of dry deposition at this time of the year. Because most of the Greenland Ice Sheet has lower accumulation rates and is expected to have less riming than Dye 3 (Hammer, 1985), dry deposition may be more significant at other ice coring sites in Greenland.

Uncertainties in data interpretation

Much of the inte~retation in this paper is based on airborne ~on~ntrations from widely separated monitoring stations throughout the Arctic. There are several problems with this approach. One difficulty is the assumption that concentrations measured at these sites can be used to represent the concentrations at Dye 3; as noted earlier, the distances between sites and the elevation of the Ice Sheet are sources of concern. Furthermore, Rahn (1981) has pointed out that precipitation-bearing air masses may have different orig- ins and thus different concentrations than long-term average air samples which are mostly from dry periods. Another problem is that cont~inants in the snow reflect scavenging at the elevations of precipitating clouds where airborne concentrations may be quite different from those at the surface. Pacyna and Ottar (1988) report high aerosol concentrations in summer as well as winter at elevations over 2 km in the Norwegian Arctic. Thus it is possible that the small seasonal variations in the SOi- snow concentration in part reflect small variations in airborne concentration at higher elevations. Because the true airborne con~ntrations at cloud level over the Ice Sheet are unknown, the SO;- scavenging ratios given here should not be considered as quantitative. Rather, they are intended solely for the purpose of identifying general seasonal variations in air-to-snow transport.

There are also uncertainties in the snowpit data. The S 1 8O patterns are similar for overlapping parts of the 1986 and 1987 snowpits, and there is reasonable consistency in the anion data of Fig. S. Nevertheless, sampling problems and analytic errors must be acknowledged, and there are uncertainties in the assign- ment of dates to each snowpit layer. Problems of redistribution of the snow by wind and contaminant migration within the snowpack may be important. The differences in total deposition of Chernobyl debris as reflected in the 1986 and 1987 snowpits may be indicative of such problems. Because of these uncertainties, interpretations of the snowpit data must be regarded as tentative.

Based on these uncertainties, a number of research topics may be suggested for improving our ability to

interpret the glacial record. Contaminant concentration data are needed for fresh snow on an event basis, with later snowpit sampling to identify changes occurring as the snowpack ages. Airborne concentration measurements at the height of precipitating clouds during snowfall are also needed. Micrometeorological data must be acquired at the same time to help assess scavenging mechanisms. Such field work, along with complementary laboratory studies and computer modeling efforts, are needed to improve our understanding of the fundamental processes by which atmospheric contaminants become incorporated in glaciers.

6. SUMMARY

Concentrations of SO:-, NO; and Cl- have been measured as a function of depth in three Greenland snowpits representing a 5-year period. The samples have been dated by comparing 6 ‘*O values with meteorological records at Dye 3. The resulting graphs of concentration vs time have been used to assess seasonal variations in transport from the atmosphere to the Ice Sheet. In addition, airborne concentrations of SO:- over the Ice Sheet have been estimated for the same time periods as the snowpit samples. This has been accomplished by statistically analyzing data from routine monitoring stations at several sites in the Arctic, and averaging data from the appropriate sites. The snow and air data have been used to compute scavenging ratios as a function of time.

Results show that SOi- concentrations in air have a strong peak in late February, while those in snow have a broad peak during January, February and March, with less seasonal variation. The airborne concentration peak is the result of LRT of anthropo- genie emissions when the polar front extends into mid- latitude source regions. The smaller amplitude of the SOi- seasonal variation in snow compared with air is attributed in part to the effect of riming: airborne SOi- is more efficiently scavenged by rimed snow than by unrimed snow. Since riming is more frequent in warm weather when airborne concentrations are low, SO:- concentrations in snow tend to be more uniform throughout the year. Scavenging ratios for SO:- are observed to peak in summer when temperatures are maximum.

NO; concentrations in the snow peak during the summer months. In contrast, airborne NO; concentrations measured in the Canadian Arctic peak in late February, coincident with the airborne SOi- peak. The summer maximum in the snow may reflect the influence of riming, resulting in efficient scavenging which more than compensates for possible low airborne NO; levels at Dye 3. However, the true airborne concentrations over the Ice Sheet may be much greater than those estimated on the basis of the Canadian monitoring stations; summer peaks in the snow may thus result from high airborne concentrations. The latter hypothesis is consistent with the high

elevation of Dye 3 and suggestion by others of high altitude sources of NO; and its precursors.

Cl- concentrations in the snow are relatively l+ during the Arctic winter, from October to March, with a dominant peak in January. There is also a weaker relative maximum in July. The high conr.,- trations in winter probably reflect LRT of seaspray emissions from the north Atlantic, at a time when seas in the Arctic are frozen but before the stable Arctic air mass has become established. The summer peak is attributed to seaspray generated in coastal Greenland areas. The effect of riming may also be important during the summer.

Dry deposition may influence concentrations of SO: _ in the Ice Sheet. The available data suggest that l&30% of the SOi- in the Dye 3 snowpits is due to dry deposition on an annual basis, although this mechanism may dominate during the dry winter and spring periods. In the higher latitudes of Greenland, where precipitation rates are lower and riming is less prevalent, dry deposition is likely to be more important.

Acknowledgements--We gratefully acknowledge the assistance of personnel at Dye 3, and of K. Kuivinen, B. Boiler, J. Litwak, C. Shaner, and K. Swanson of the Polar Ice Coring Office in helping with logistics associated with this project. Assistance with data analysis was provided by L. Hoffman. Data for the 1984 snowpit and suggestions for field opcra- tions were provided by P. Mayewski and M. Spencer of the University of New Hampshire. The S’*O analyses were conducted by N. Gundestrup of the Geophysical Institute, University ofcopenhagen. Analyses of lJ4Cs and ‘37Cs were conducted by M. Monaghan of the Department of Geophys- ical Sciences, University of Chicago. Airborne concentration data were provided by the Canadian Atmospheric Environ- ment Service (L. Barrie), the Danish Agency of Environ- mental Protection (N. Heidam), the Norwegian Institute for Air Research (B. Ottar and A. Semb), and the University of Rhode Island (K. Rahn and D. Lowenthal). Suggestions on the statistical analyses were provided by J. Kadane and R. Tsay. This work was funded by National Science Founda- tion grant DPP-8618223 and scholarships from the Richard K. Mellon and Claude W. Benedum Foundations.

REFERENCES

Barrie L. A. (1985) Atmospheric particles: their physical/ chemical characteristics and deposition processes relevant to the chemical composition of glaciers. Ann. Glacial. 7, lO@-108.

Barrie L. A. and Hoff R. M. (1985) Five years of air chemistry observations in the Canadian Arctic. Atmospheric Environ- ment 19, 1995-2010.

Barrie L. A., Fischer D. and Koemer R. M. (1985) Twentieth century trends in Arctic air pollution revealed by conduct- ivity and acidity observations in snow and ice in the Canadian High Arctic. Atmospheric Environment 19, 2055-2063.

Borys R. D., Demott P. J., Hindman E. D. and Feng D. (1983) The significance of snow crystal and mountain-surface riming to the removal of atmospheric trace constituents from cold clouds. Proc. Fourth International Conference on Precipitation Scavenging, Dry Deposition, and Rewpen- don, Santa Monica, CA, 29 November-3 December, 1982, pp. 181-189. Elsevier, Amsterdam.

Seasonal variation in sutphate, nitrate and chloride 2493

Borys R. D. (1989) Studies of ice nucleation by Arctic aerosol on AGASP-II. J. afmos. Chem. (In press).

Dansgaard W., Johnson S. J., Clausen H. B. and Gundestrup N. (1973) Stable isotope glaciology. Med. on Gronland 187, l-53.

Davidson C. I. (1989) Mechanisms of wet and dry deposition of atmospheric contaminants to snow surfaces. In The ~~vira~e~ta~ Record in Glaciers and Ice Sheets (edited by H. Oeschger and C. C. Langway, Jr.), pp. 29-51. Dahlem Konferenzen, John Wiley and Sons, Chichester.

Davidson C. I., Chu L., Grimm T. C., Nasta M. A. and Qamoos M. P. (1981) Wet and dry deposition of trace elements onto the Greenland Ice Sheet. Atmospheric En- vironment 15, 142991437.

Davidson C. I., Santhanam S., Fortmann R. C. and Olson M. P. (1985) Atmospheric transport and deposition of trace elements onto the Greenland Ice Sheet. Atmospheric En- vironment 19, 2065-2081.

Davidson C. I., Honrath R. E., Kadane J. B., Tsay R. S., Mayewski P. A., Lyons W. B, and Heidam N. Z. (1987a) The scavenging of atmospheric sulfate by Arctic snow. Atmospheric Envjronment 21,871~882.

Davidson C. I., Harrington J. R., Stephenson M. J., Mon- aghan M. C., Pudykiewicz J. and Schell W. R. (1987b) Radioactive cesium from the Chernobyl accident in the Greenland Ice Sheet. Science 237, 633-634.

Davidson C. I. and Wu Y. L. (1989) Dry deposition of particles and vapors. In Acid Precipitation, Volume 3. Sources Deposition and Canopy interactions (edited by D. C. Adriano). Advances in Environmental Science Series, Springer-Verlag, New York (in press).

Dibb 1. (1989). The Chernobyl reference horizon (?) In the Greenland Ice Sheet. Geopkys. Res. Lett. (in press).

Finkel R. C., Langway C. C., Jr. and Clausen H. B. (1986) Changes in precipitation chemistry at Dye 3, Greenland. J. geopkys. Rex 91,9849-9855.

Hammer C. U. (1985) The influence on atmospheric composition of volcanic eruptions as derived from ice core analysis. Ann. Glaciot. 7, 125-129.

Hansen A. D. A. and Rosen H. (1984) Vertical distributions of particulate carbon, sulfur and bromine in Arctic haze and comparison with ground-level measurements at Barrow, Alaska. Geophys. Res. Lett. 11, 381-384.

Heidam N. Z. (1984) The components of the Arctic aerosol. Atmospheric Environment 18, 329-343.

Herron M. M. (1982) Impurity sources of F-, Cl-, NO; and SOi- in Greenland and Antarctic precipitation. J. gee- phys. Res. 87, 3052-3~0.

Joranger E. and Ottar B. (1984) Air pollution studies in the

Norwegian Arctic. Geopkys. Res. Lett. 11,365-368. Junge C. E. (1977) Processes responsible for the trace content

in precipitations. In Isotopes and Impurities in Snow and Ice, Proc. IUGG Symp., Grenoble, August-September 1975. IAHS Publ. 118, pp. 63-77.

Legrand M. R. and Delmas R. J. (1986) Relative contributions of tropospheric and stratospheric sources to

nitrate in Antarctic snow. Teiius 38 B, 236-249. Logan J. A. (1983) Nitrogen oxides in the troposphere: global

and regional budgets. J, geop~ys. Res. 88, 10,78~10,807. Mayewski P. A., Lyons W. B., Spencer M. J., Twickler M.,

Dansgaard W., Koci B., Davidson C. I. and Honrath R. E. (1986) Sulfate and nitrate concentrations from a south Greenland Ice Core. Science 232, 975-977.

Mayewski P. A., Spencer M. J., Lyons W. B. and Twickler M. (1987) Seasonal and spatial trends in South Greenland snow chemistry. Atmospheric Environment 21, 863-869.

Neftel A., Beer J., Oeschger H., Zurcher F. and Finkel R, C. (1985) Sulfate and nitrate concentrations in snow from South Greenland. Nature 314, 61 l-613.

Noone K. J. and Clarke A. D. (1988) Soot scavenging measur~ents in Arctic snowfall. Atmospheric environment 22,2773-2778.

Pacyna J. M. and Ottar B. (1988) Vertical distribution of aerosols in the Norwegian Arctic. Atmospheric Enuiron- ment 22,2213-2222.

Radke L. F., Lyons J. H., Hegg D. A., Hobbs P. V. and Bailey I. H. (1984) Airborne observations of Arctic aerosols-l. Characteristics of Arctic haze. Geaphys. Res. Lett. 11, 393-396.

Rahn K. A. (1981) Atmospheric, riverine and oceanic sources of seven trace constituents to the Arctic Ocean. Atmos- pheric Environment 15, 1507-1516.

R&n IL A. and Shaw G. E. (1982) Sources and transport of Arctic pollution aerosol: a chronicle of six years of ONR research. Navaf Res. Rev. 3, 3-26.

Risbo T., Clausen H. B. and Rasmussen K. L. (1981) Super- novae and nitrate in the Greenland Ice Sheet. Nature 294, 637639.

Scott B. C. (1981) Sulfate washout ratios in winter storms. J. appl. Met. 20, 619625.

Semb A., Braekkan R. and Joranger E. (1984) Major ions in Spitsbergen snow samples. Geophys. Res. Letters 11, 445-448.

Shaw G. E. (1980) optical, chemical and physical properties of aerosols over the Antarctic ice sheet. Atmospheric environment 19,2025-2031.

seasonal variations in sulfate, nitrate and chloride in the greenland ice sheet: relation to...

Documents