1. Introduction
Over the last three decades, strong efforts have been made to recover deep ice cores from the Greenland ice sheet to provide long-term records of chemical and isotopic signatures over the last climatic cycle. success of these efforts has made it possible to examine the natural and anthropogenically induced long-term changes of global climate and atmospheric chemistry (Reference RobinRobin, 1983, Reference Finkel, Lansway and ClausenFinkel and others, 1986; Reference DansgaardDansgaard and others, 1993; Reference Fuhrer, Neftel, Anklin and MaggiFuhrer and others, 1993; Reference Hanson and SaltzmanHanson and Saltzman, 1993).
However, the interpretation and intercomparison of isotopic (and chemical) deep ice-core records is not always unambiguous, due to the geographical location of each individual drill site. Local climatic and overall meteorological conditions influence precipitation processes and transport pathways of chemical species and the water vapour, which are together with source patterns responsible for the isotopic and chemical signature of the snow-pack.
So far, little information has been drawn from the large-scale geographical distribution of isotopic and chemical parameters in Greenland firn and their systematic trends, which are also crucial for the interpretation of the long-term records recovered. For example, Reference DansgaardDansgaard (1964) deduced a linear relation between δ18O and the condensation temperature at high-altitude sites on the Greenland ice sheet.
Furthermore, only little deuterium excess (d = δD - 8δ18O) data for the Greenland ice sheet are available. This hydrological parameter, however, is believed to reflect evaporation conditions in the source area and the advection history of water precipitated over Greenland (Reference Johnsen, Dansgaard and WhiteJohnsen and others, 1989, Reference FisherFisher, 1992). Aside from the long-term climatic records mentioned above, glacio-meteorological parameters (temperature and snow accumulation) may reveal recent climatic changes. In cold firn areas, firn temperature will react with an increase paralleling atmospheric warming (Reference HaeberliHaeberli, 1990). Also, increased snow deposition in the accumulation areas, due to higher water-vapour content of the advected air masses at higher atmospheric temperatures, may reflect global warming (Reference Morgan, Goodwin, Erheridge and WookeyMorgan and others, 1991).
Against this background investigation of glacio-meteorological and isotopic firn parameters it is important
-
1. to resolve spatial and temporal changes in climatologically relevant parameten such as temperature and accumulation;
-
2. to determine the temperature-dependence of accumulation and isotopic content by investigating and thus “simulating” different climatological regimes along the line investigated;
-
3. to use the geographical distribution of accumulation and isotopic (as well as chemical) parameters to identify sources and transport pathways;
-
4. to test the representativeness of deep drill sites with respect to the overall geographical and meteorological conditions on the ice sheet.
The reconstruction of the Expédition Glaciologique Internationale au Groenland (EGIG) line by the Institut für Vermessungskunde, Technische Universität Braunschweig, Germany, in 1990–92, allowed us to carry out a glaciological surface study investigating for the first time both glacio-meteorological and isotopie/chemical parameters along a complete west-east transect through central Greenland. Furthermore, the EGIG line, representing the only existing transect previously investigated glaciologically, allows the determination of possible temporal changes in accumulation and fim temperature by direct comparison (Reference Anklin, Stauffer, Geis and WagenbachAnklin and others, 1994) Seasonal and geographical variations of the parameters investigated in this study are further of fundamental importance as base data for later comparisons.
This paper essentially deals with the temporal variation and geographical distribution of firn temperatures, annual accumulation rates and the isotopic content of the firn along the EGIG line (results of chemical investigations will be published separately elsewhere).
2. Methods
2.1. Field activities
This study is based on the activities during two field seasons (in 1990 and 1992) along a traverse through central Greenland In 1990 (further on referred to as EGIG-west), the field party started out at the Greenland Ice Core Project (GRIP) ice-core site (T99) following the ice divide to Crête (T43). From Crête the field party moved westward along the EGIG line to TO1 (Fig. 1). Along this route, shallow firn core drillings (~8m), pit studies and firn-temperature measurements were made at ten sites (Fig. 1). The expedition of the field season 1992 (EGIG-east) again started at GRIP, moved to Crete and then followed the eastern part of the EGIG line to Caecilia Nunatak (CNN), from where it returned the same year. During this second season, shallow firn cores (~8 m) were drilled and snow pits excavated at eight sites. In addition, high-precision firn-temperature profiles were taken at 17 sites down to a depth of 15 m using the VAW/ETH-Ztirich steam-drilling equipment (Reference LaternserLaternser 1994).
Map of Greenland showing the drill sites investigated along the EGIG line.
The two field seasons differed significantly in the technique of temperature measurement and sampling procedure, hence for data comparison special care was taken to eliminate systematic effects affecting the temperature and the isotopic and chemical (to be published) content of the samples. For direct comparison between the two seasons, firn cores were drilled at sites T99 and T43 both in 1990 and 1992.
2.2. Temperature measurements
For the temperature measurements in 1990, thermistors (Unicurve, 1000Ω) with an overall accuracy better than ±3°C have been used. Due to technical drilling problems and time constraints of the traverse schedule, cores could not be drilled down to the planned depth of 10 m at all sites. Therefore, all temperature measurements were consistently made at 5 m depth, where the seasonal temperature oscillation still systematically affects the observed value.
In 1992, however, firn temperatures were measured in steam-drilled boreholes with previously calibrated Fenwal thermistors at depths of 3,0, 5-0, 7.0, 9.0, 10.0, 11 0, 12.0, 13.0, 140 and 15.0 m.This 15 m firn temperature, T15 m is closely related to the mean annual firn temperature, Tm which itself is a proxy parameter for the mean annual surface-air temperature, as long as no significant surface meltwater refreezing causes release of latent heat. The boreholes for the temperature measurements were drilled on the way from Dome GRIP to Caecilia Nunatak. Immediately after drilling, they were stabilized at the surface by Insertion of a 1 m long plastic pipe with a removable lid and left open while the traverse continued. Temperatures were finally measured on the return. The time lag between drilling and measurement was several weeks for most of the drill sites which allowed for complete dissipation of the thermal disturbance caused by the steam drilling. Only for sites near the ice margin, the expedition schedule was too tight for complete adjustment. For these sites, extrapolation formulas were applied to gain an “undisturbed” temperature value. In most cases the accuracy was better than ±Q.1°C, often even better than 0.05°C (for more details concerning the measuring and correction techniques for the temperature values along EGIG-cast, see Reference LaternserLaternser (1994)).To eliminate the seasonal influence on the 1990 data, the temperature offset ΔT = T(5m, t)— T15m between the temperature T(5m,t) (measured at 5 m depth and on the date of measurement t) and T15m were calculated.
The time- and depth-dependent temperature profile in the firn may be approximated by (Reference Carslaw and JaegerCarslaw and Jaeger, 1959)
where Ts
is the annual temperature amplitude, z is the measured depth, ω is the frequency of the annual temperature variation 
with k the firn diffussivity and ∈ the phase shift of this oscillation relative to the calendar date t. Using the measured temperature profiles of 1992, the parameters Ta
, ∈ and a mean diffusivity k could he numerically fitted (see Fig. 2) for each site Τa
and ∈ at site T99 were shown to be in good agreement with values from the automated weather station at GRIP (Reference GundestrupGundestrup, 1993), and k with values for firn diffusivities given by Reference LliboutryLliboutry (1964–65). To calculate ΔT for the sites of EGIG-west, climatologically comparable sites (equal altitude, similar latitude, and thus similar temperature amplitude and comparable firn stratigraphy) of the eastern part were assigned to the western sites. The temperature offsets ΔT for the eastern sites were calculated using Equation (1) at the measuring date of 1990 and then subtracted from the 5 m firn temperature value of the corresponding western sites to gain an undisturbed 15 m temperature value. The error introduced is difficult to quantify, but the variation of ΔT between different sites is primarily dependent on t and less so on the fitted parameters Ta
, ∈ and k, So the overall accuracy of T15m
, along EGIG-west may he estimated to be better than ± 0.5°C.
Firn-temperature profile at site 753. Plotted are the measured (rectangles) and fitted (solid line) values for the deviation ΔT from the 15 m Firn temperature T15 m dependent on the measured depth z.
The corrected 5 m firn temperatures are listed in Table 1 together with the 15 m firn temperatures measured at the firn-core drill sites along EGIG-east (for a complete listing of the Urn temperature measurements made in 1992, see Reference LaternserLaternser (1994)). The 15m temperatures obtained for EGIG-west certainly do not show the very high accuracy of the temperature measurements in the eastern part. Therefore, they are not used to resolve temporal changes in the firn temperature, which are of small magnitude For the determination of spatial variations in the firn temperature, however, the accuracy of the corrected values is sufficient.
Climatological and isotopic parameters determined at the sites along the EGIG linD: 15 m firn temperatures together with estimated accuracy in °C, average annual accumulation rate in cm a−1 water equivalent, mean summer and winter accumulation m cm a−1 water equivalent, average annual water-weighted mean of δD and the deuterium excess d in % together with standard deviations and the number of years n covered by the firn cores
2.3. Snow-accumulation rates and snow characteristics
For every drill site, water-equivalent depth scales were obtained by using the corresponding average density values of the single core sections and snow-pit samples measured. At sites T05 and T41 no snow pits were excavated, so there density profiles lack the first, 1.5 m. The gap was filled by using the density data of the adjacent sites T09 and T43, respectively. Because this correction only affects the top 0.5 m water equivalent, and the adjacent sites show very close climatological conditions, the error introduced is assumed to he very small.
Determination of annual accumulation rates in this study was done by counting annual layers of seasonally varying tracers in the firn. For the cores drilled in 1992, δ18 profiles (together with profiles of major ion concentrations) were measured at high resolution (~8)samples year−1 to resolve seasonal variations (Fig 3), whereas in 1990 the seasonally varying H2O2 concentration-had already been determined in the field in collaboration with the University of Bern (Reference Anklin, Stauffer, Geis and WagenbachAnklin and others, 1994) using a fluorimetric method after Reference SiggSigg (1990) that proved to he a reliable tool for in-situ accumulation determination. In this study, the summer maxima of the profiles were used to identify single years. In addition, to distinguish summer and winter accumulation, the δ 18O–(H2O2), profiles were divided into summer and winter half-years, with their limits being defined by the mean value between adjacent summer maxima and winter minima (Fig. 3). This method, however, only works if the profiles of the tracers which are regarded as temperature-proxy parameters follow closely the sinusoidal annual temperature variation. This condition seems to be sufficiently obeyed by all firn cores; however, thé relative position of the limits between maxima and minima varies. To interpret the seasonal pattern, therefore, only averaged summer and winter accumulation values are considered.
Firn-core profiles of seasonally varying tracers: (a) H2O2. and the accompanying δ18O profile at site T21 of EGIG-west; (b) Ô>80 and d profile at site T47 of EGIG-east. Apart from the consecutive summer maxima (S) and winter minima (W), the limits between single half-years (grid lines) are plotted.
In addition to snow-pit sampling at each drill site, further detailed pit studies were done at 21 places along EGIG-east, where the stratigraphy was observed and both temperature and density profiles were measured. Within the dry-snow and upper percolation zones, the surface layer was generally built up of small, rounded grains, possibly topped by precipitation or decomposing and fragmented precipitation particles from recent snowfalls. This Layer containing last winter’s accumulation on average reached a depth of about 60–80 cm. Below that the crystals often showed a sharp change to facetted or even cup-shaped forms of considerably larger size marking last year’s summer horizon. This depth was usually characterized by a significant “dent” in the density profile. Further down, the crystals generally remained facetted, sometimes showing mixed forms with rounded grains again. Depending on the individual site and the depth of the corresponding pit, a second summer usually less distinct than the previous one, could be found at depths of around 120–160 cm. Fine “ice skins” caused by either melt or wind polish were encountered throughout all profiles. Closer to the ice margin (lower percolation and wet-snow zone), the thickness of the layers grew considerably and the occurrence of refrozen firn and actual wet grains became more frequent (for a detailed presentation of all snow-cover profiles along EGIG-east, see Reference LaternserLaternser (1994)).
2.4 δ18O, δD and deuterium excess
All samples were measured for δ18O and δD by mass spectrometry at the Institut für Umweltphysik, Heidelberg. Two cores were isotopically analysed at the GSF Forschungszentrum für Umwelt und Gesundheit, Institut für Hydrologie, Neuherberg. The current overall accuracy of these measurements is 0.1% for δ18O and 2.0% for δD, leading to an accuracy for the deuterium excess d = δD – 8δ 18O of 2.2% for every single sample.
In 1992, the firn cores and snow-pit samples were brought back frozen to Heidelberg using the GRIP cooling facilities. For the 1990 field season, however, it was not possible to get this logistic support, so samples could melt during transport from Jacobshavn to Heidelberg. Therefore, special care was taken to preserve seasonal resolution of the isotopic and chemical profiles and to prevent chemical contamination and water-vapour exchange with the surrounding air, which alters signature of the samples.
To do so, the drilled firn cores were subdivided into seasonal sub-samples in the field according to the H2O2profile (4–8 samplesyear see also Figure 3), thoroughly decontaminated for later chemical analyses and sealed in polyethylene (PE) bags, which were pre-cleaned with highly purified water. The samples from each site were then put into a larger closed PE bag. Liquid water within this PE bag (as a result of occasionally occurring sample leakages) built up a separate water-vapoui atmosphere with identical average isotopic content as the samples themselves, reducing water-vapour exchange with the ambient air significantly. To quantify possible changes in the isotopic content of the melted samples, various laboratory experiments were made, simulating the possible effects of water-vapour exchange and/or loss through the walls and seals of the sample PE bags. These experiments proved that thoroughly sealed samples, packed in the way described above, were (compared to the overall accuracy of the measurement) unaffected in their isotopic content (
, 
and 
). Only a few of those samples subject to extensive vapour diffusion through intentionally punctured walls or seals of the bag showed significantly lowered deutenum excess values but only slightly altered δ18
and SδD values. Therefore, all samples subject to obvious leakages and with negative deuterium excess values were excluded from further analysis. The part of affected samples varied between 5 and 20% for the different drill sites. Using the reduced data set, the average annual water-weighted mean of δ18, δD and d for sites T99 and T43 could be calculated and statistically tested against the unaffected values for the drilling in 1992 at the same sites. For the parallel drillings at sites T99 and T43 both t-test and the parameter-free Wilcoxon test showed no significant difference (p = 5%) for the average of δ18O,, δD and even for the more sensitive deuterium excess between the two field seasons. The findings of these experiments and tests prove that the reduced data set h reliable.
The δl8O, δD and d profiles (see Fig. 3) show vanations between each annual cycle, so for further interpretations mean annual cycles were calculated wherever temporal resolution was high enough (<6 samples year−1) To do so, the depth interval between summer maxima and winter minima in Figure 3 were divided into three equidistant sub-intervals, leading to a temporal resolutron of 2 months, if an homogeneous accumulation rate throughout the year is assumed. To calculate the average annual cycle, the water-weighted mean for every generated 2 month interval was determined and then averaged over all years covered.
3. Results and Discussion
3.1. Firn temperature
3.1.1. Temporal variations
Whiie the ablation zone of an ice sheet will react with increased melting on climatic warming, the accumulation zone will mainly show a change in the firn temperature accompanying the atmospheric temperature change (Reference HaeberliHaeberli, 1990). Hence, the measurement of this parameter allows determination of a possible climatic change on the ice sheet by comparison with previously measured firn temperatures in the area under investigation. The firn temperatures for the western part of the EGIG line (contrary to the 15 m firn-temperature values along EGIG-east) do not show the required accuracy for such a comparison, due to the 5 m firn-temperature currection carried out. Therefore, in Table 2 only the 1992 temperature values for sites T43 and T5 3 are listed together with values by Reference Quervainde Quervain (1969) for the time span 1959–64. Comparisons between these two measurements show no significant temperature change within the error limits (Reference LaternserLaternser, 1994). However, temperature profiles measured in deep boreholes in the Greenland ice sheet show distinct heat-flux anomalies caused by strong secular warming followed by slight cooling after about 1950 (Reference RobinRobin, 1983). The similarity of the 1992 and 1959-64 values could, therefore, possibly indicate recent re-warming of firn temperatures to the high levels previously reached around 1950. Comparison with later temperature measurements along the EGIG line will clarify whether a long-term warming trend persists.
Comparison of firn temperatures determined by Reference Quervainde Quervain (1969) during the time span 1959–64 with the firn temperatures measured in 1992 The given firn temperatures T(z, t) in °C for 1992 are corrected for the date of measurement and the measured depth in m of the values by de Quervain and others (Reference LaternserLaternser, 1994). Also given an the estimated measurement accuracies
3.1.2. Geographical distribution
The 15 m firn temperatures along the EGIG line show the expected linear relation between altitude and latitude (Fig. 4). The areas marked A and B in Figure sections of approximately equal altitude but different latitude. In addition, the geographical distribution of Τ15m along the EGIG line is plotted in Figure 5b, showing the temperature minimum is situated directly on the ice divide. Note that in Figure 5b distance to the ice divide parameterizes both altitude and latitude of the drill site.
Temperature-altitude dependence for all investigated sites along the EGIG line. Areas A and B indicate regions of approximately equal altitude but different latitude.
Geographical distributions of various parameters investigated along the EGIG line dependent on the distance from the ice divide x: (a) altitude above sea level h; (b) 15 m firn temperature T15 m; (c) average annual accumulation rate acc; (d) average annual water-weighted mean δ18O Of (e) average annual water-weighted mean of the dueterium excess d; (f) average annual chloride deposition flux FCl,
For the central dry-snow zone of EGIG-east (from site T99 via T43 to T53), the route of the EGIG line allows the decoupling of altitude and latitude effects for the 15m firn temperature by means of multiple regression. The corresponding partial temperature gradients are:
(h is the altitude and λ is the latitude of the drill site). Along the eastern slope from site T65 to T69, where the latitude of the traverse route remained almost constant, the 15 m firn temperature-altitude gradient is
and even higher in the lower percolation and wet-snow zone (-1.4–C (100 m)−1; Reference LaternserLaternser, 1994). These differences are attributed to increased latent-heat transport through the firn due to higher firn densities and occasionally occurring meltwater percolation (for a detailed description of the temperature and stratigraphical findings along EGIG-east, see Reference LaternserLaternser (1994)).
Due to the high intercorrelation of altitude and of the drill sites along the EGIG-west traverse, no unambiguous decoupling of the two parameters on the measured firn temperatures (from site T05 to T41) by multiple regression is possible. Linear regression leads to an overall (including latitude effects) 15 m firn temperature-altitude gradient of-0.94°C (100 m)−1 (r2 = 0.99). If the 15 m firn temperature-latitude gradient for the central-eastern part in Equation (3) is applied, one obtains a latitude-corrected 15 m firn temperature-altitude gradient of -0.87°C (100 m) −1 (r2= 0.99) for the western part of the EGIG line This temperature-latitude gradient in the west is higher than in the central-eastern dry-snow zone (Equation (2)) but lower than on the eastern slope from site T65 to T69 (Equation (4)). The differences result from averaging of different snow zones when applying linear regression on the whole western EGIG line. A more detailed investigation of the geographical temperature distribution along EGIG-west, however, cannot be attempted because of the substantial 5 m firn-temperature correction described in section 2.2.
3.2. Accumulation
3.2.1. Temporal variations
By stratigraphical dating of the ire cores (see section 2,3 ), we were able to derive the annual accumulation deposited in every firn core. In Figure 6, accumulation is plotted for the common time span covered by all cores (1989–93). Apart from the geographical trend (described in section 3.2.3),. Figure 6 shows distinct temporal features (valleys and ridges) along the whole EGIG line, revealing that the annual accumulation variability is a large-scale phenom-However, an overall temporal trend cannot be identified. This is also supported by Reference Anklin, Stauffer, Geis and WagenbachAnklin and others (1994), who, by means of comparison with earlier measurements along the western EGIG line, found no significant temporal change in the precipitation pattern over West Greenland during the last 30 years. For the eastern part, only a few previous accumulation measurements are available. These are listed in Table 3 together with the values determined in this study. The methods used by the authors listed in Table 3 to determine the average annual accumulation rate significantly differ from the δ18O method used in this study (see section 2.2). Reference Merlivat, Ravoire, Vergnaud and LoriusMerlivat and others (1973) gave tritium-based values indicating mean annual accumulations for the time span 1959–68. Reference Quervainde Quervain (1969) used a combination of snow-stake measurements and density together with hardness profiles at sites T47, T53 and T61, and Reference BoutronBoutron (1979) used snow stakes at site T46. The latter site is close to site T47 and is therefore used for comparison. The geodetic study by Reference SeckelSeckel (1977) gave mean annual snow thicknesses for the time span 1956–68. For comparison with our own data, these have to be converted into water equivalent values. Since no density measurements were made by Seckel, we use our density profiles from 1992. This procedure seems to be justified, since the mean annual temperature, and hence the snow texture at the drill sites, has not changed significantly since 1959–68 (see section 3.1.1). The same procedure was used by Reference Anklin, Stauffer, Geis and WagenbachAnklin and others (1994) for EGIG-west, resulting in good agreement of the converted values with the data determined in their study.
Spatial and temporal variation of the annual accumulations for the common time span covered by all firn cores
Comparison of average annual accumulation rates by Reference BoutronBoutron (1979), Reference Merlivat, Ravoire, Vergnaud and LoriusMerlivat and others (1973), Reference SeckelSeckel (1977) and Reference Quervainde Quervain (1969) with the values determined in this study together with their standard deviations. In addition to the accumulation rates in cm a−1 water equivalent, the corresponding time spans are listed
Comparison of the values for sites T47 and T53 shows no significant difference between the authors except the values of Seckel, which are ~20% lower. The values of Seckel for these sites, however, are also inconsistently low compared to his accumulation rates for the adjacent sites in the west (T41 and T43). For sites T61 and T66, the values given by Reference Merlivat, Ravoire, Vergnaud and LoriusMerlivat and others (1973) agree very well with our recent data. However, Reference Quervainde Quervain (1969) gave an accumulation rate for site T61 which is significantly higher.
Summarizing this comparison for the eastern part of the EGIG line, we conclude that the average annual accumulation rate most likely has not changed during the last 30 years. However, the sparse data available for this region is not free from ambiguities, and therefore does not allow a final statement. The data determined in this study are therefore of great importance as a reference for future measurements along the EGIG line.
3.2.2. Geographical distribution
The geographical distribution of the mean annual accumulation rate is plotted in Figure 5c, A first inspection allows one to subdivide the area under investigation into three different regions:
-
1. A high-accumulation plateau (~45cm a−1 water equivalent) from site T05 to site Τ17. At site T09, accumulation is slightly reduced compared to sites T05 and T13, probably due to surface undulations (Reference HempelHempel, 1994) that occur at this site, which locally may result in a partial loss of the annual accumulation by wind drift.
-
2. Steadily decreasing accumulation rates from site T21 (~43 cm a−1. water equivalent) to site T4 3 (−25 cm a−1water equivalent) and further on to site T99 (~20 cm a−1 water equivalent).
-
3. Constant accumulation rates in the eastern part of the EGIG line (sites T47-T66, ~23–17 cm a−1 water equivalent).
This overall geographical distribution along the EGIG line supports the picture of a main water-vapour trajectory along the western slope of central Greenland from west to east, resulting in decreasing accumulation due to gradual water-vapour loss by cooling during ascent and a precipitation shadowing lee sides of the ice divide in the east. Air masses entering Greenland from the east, however, seem to be substantially reduced in their water-vapour content, because of heavy precipitation during their ascent over the steep coastal range. Thus, they are unable to produce high accumulation rates on top of the ice sheet. Comparison with the accumulation map for Greenland compiled by Reference Ohmura and ReehOhmura and Reeh (1991) (based on data from various expeditious and values provided by meteorological stations along the Greenland coast) shows good agreement in the central part of the ice sheet and slightly lower values along the western and Pastern slope in this study which, however, are still within the standard deviation.
In addition to the mean annual accumulation, the corresponding values of the summer and winter half-years defined by the H2O2 and δ18O profiles (see section 2.3) are listed in Table 1. In the west, accumulation during the winter half-year is 20–40% higher than in the summer half-year. Although one cannot define precisely the annual accumulation maximum, this rough seasonal distribution does not support the view of Ohmura and Reeh, who proposed a precipitation maximum on the western slope in summer. However, their conclusion is based on the precipitation measurements at meteorological stations at sea level. Therefore, this would indicate a significant difference between coastal and inland sites with respect to their seasonal precipitation patterns.
Further inland, the excess accumulation decreases to the ice divide. In the eastern part, precipitation during the summer and the winter half-years is of comparable magnitude. This temporal and spatial distribution supports cyclonic influence, which is greatly enhanced in the winter half-year, as being responsible for the main parts of the annual accumulation. Mean weather charts compiled by Reference LiljequistLiljequist (1970) indicate that these cyclones enter Greenland through Davis Strait, causing substantial snowfalls on the western slope. Precipitation terminates when these systems pass the ice divide.
This cyclonic influence also explains the geographical and seasonal distribution of the chloride deposition. Marine trace elements in the firn, such as chloride and sodium, originate from sea-salt dispersion over the ocean and are transported to the central Greenland ice sheet via the free troposphere (Reference SteffensenSteffensen, 1988). The chloride input into the free troposphere is greatly dependent on the storm, hence cyclonic, activity in the source region, the chloride content of the lower troposphere in the vicinity of the Greenland coast, and on local sea-ice conditions.
In Figure 5f, the average annual chloride-deposition flux 
is plotted (Cli
is the annual water-weighted chloride concentration, accii
is the annual accumulation of year i and n is the number of years covered by the firn core). Here, Fcl
shows a rapid decrease from the west coast to the ice divide, which significantly exceeds a decrease solely produced by the decline of the accumulation rate Towards the eastern coast, however, Fcl
remains constant. This indicates that the main cyclonic influence driving marine-aerosol species to the ice sheet enters Greenland from the west.
The seasonal variation of the chloride concentration (Fig. 9c) shows a maximum at all EGIG sites during spring, which is attributed to the enhanced cyclonic activity over the Atlantic Ocean during this season. Similar findings also hold for other drill sites throughout inland Greenland (Reference Mayewski, Spencer, Lyons and TwicklerMayewski and others, 1987; Reference Davidson, Harrington, Stephenson, Small, Boscoe and GandleyDavidson and others, 1989; Reference BeerBeer until others, 1991). Aerosol cuncentrations at coastal sites (Reference HeidamHeidam, 1981), however, show a chloride maximum in summer caused by sea-salt uptake from the local ocean surface, probably being suppressed during early winter/spring due to local sea-ice coverage. This difference in the seasonal chloride deposition patterns of coastal and inland sites leads to the conclusion that the lower troposphere, causing the concentrations at sea-level altitude, does not affect efficiently the snow chemistry on top of the Greenland ice sheet.
To predict changes in annual accumulation rates in Greenland for altered climatic conditions, the relationship between accumulation and temperature is required. The average accumulation rate, as well as the isotopic content of the firn, is greatly dependent on the condensation temperature Tc. This temperature, however, is not necessarily equal to the mean annual surface temperature, due to the often prevailing temperature inversion above the ice sheet. Unfortunately, no adequate data for the inversion strength over the Greenland ice sheet are available. For Antarctica, Reference Phillpot, and ZillmanPhillpot and Zillraan (1970) stated a linear relationship between Tc and Tm. Furthermore, most of the accumulation over Greenland is caused by cyclonic systems, which destroy any existing inversion layer These single precipitation events therefore greatly influence the average and the geographical distribution of the condensation temperature Precipitation-weighted temperatures therefore should be used 1990). Due to the lack of detailed climatological temperatures along the EGIG line, the firn-temperature data determined in this study constitute the only representative recent temperature data available. Therefore, we prefer an empirical approach, based on the climatological division of the area investigated described above, which describes the relationship between the 15 m fun temperatures and the average annual accumulation rates.
The westernmost part of the EGIG line (sites TO5-T17) in Figure 7 shows decreasing 15 m firn temperatures from —18° to -24°C but no significant spatial change in the accumulation rate. The central western slope (sites T21-T43 and further on to site T99), however, shows a linear decrease of the accumulation rate, which accompanies a decline of T15m from -24° to -32°C. Linear regression for this section yields
Accumulation-temperature dependence for all sites investigated along the EGIG line.
Since precipitation essentially depends on the water-vapour pressure at saturation, being mainly a function of temperature, this geographical temperature relationship of the accumulation rate can also be mapped on to temporal temperature changes, provided that the circulation pattern over the Greenland ice sheet and, hence, the history of the advected air masses have not changed. Thus, we can conclude that the annual accumulation rate in this area linearly reacts to occurring temperature variations. Note that Equation (6) only holds west of the ice divide, while in the eastern part accumulation is constantly low even at the easternmost site T66, which is close to the coast. Here, the existence of the geographical barrier constituted by the coastal range and possible radiative cooling by the prevailing sea-ice coverage results in significant reduction of the water-vapour content before air masses advected from the east are able to penetrate on to the ice sheet.
3.3. Isotopic signature
3.3.1. Temporal variation of δ18O
Apart from the direct comparison of different temperature measurements, the temporal variation of the δ18O record allows identification of significant temporal trends in the atmospheric temperature on the ice sheet. The temporal distribution of the annual water-weighted mean of δ18O does not show any significant temporal changes along the EGIG line during the common time span covered hy all firn cores. Due to the shortness of this time span (6years), this is just an indication. Statistical testing (Shapiro Wilk test, p. = 5 %) supports the assumption of a common unimodal normal distribution for the annual water-weighted means of each core, which cover time spans between 5 and 21 years. Thus, we conclude, provided that the advection history of the precipitation-delivering air masses, i.e. the water-vapour trajectory, remained equal, that the mean annual condensation temperature has not significantly changed during the time span covered by the firn cores.
3.3.2. Geographical distribution δ18O and δD
δ18O and δD, as proxy parameters for the mean annual temperature, are of special interest for the use of deep ice cores as climatic records. For such an interpretation, however, knowledge of the temperature-dependence as well as the geographical dependence of δ18O are crucial. Reference Johnsen, Dansgaard and WhiteJohnsen and others (1989) derived an empirical linear relationship between δ18Ο and Tm for high-altitude sites in central Greenland, which is used for the interpretation of the GRIP ice-core isotopic record
An overall linear regression of the average water-weighted annual mean of δ18O and the 15 m firn temperature determined in this study leads to δ18O = (0.69 ± 0.03 T15m -(13.16 ± 0.90) (%), r 2 = 0 97. This is close to the relationship given by Reference Johnsen, Dansgaard and WhiteJohnsen and othen (1989).
However, detailed inspection of Figure 8, showing the corresponding linear relationship for δD over the whole EGIG line, suggests the data are better fitted by two individual regression lines for the western and the eastern parts. For the deuterium data, this leads to
δD-temperature dependence for all rites investigated along the EGIG line.
The δ18 O-temperature gradients ought to be directly related to the corresponding deuterium gradients in Equations (7) and (8) Linear regression of δ18O and T15m , after correction for inter-instrumental offsets between the different spectrometers used (see also section 3.3.3), leads to redundant temperature information gained by the δ18O and δD values
The existence of two separate regression lines for the areas east and west of the ice divide is attributed to different transport pathways of the water vapour precipitated over the Greenland ice sheet. Moist air masses from the west cause high accumulation in the western part, decreasing towards the ice divide, accompanied by a gradual depletion of the heavier isotope species during ascent (line 1 in Figure 8). When the air from the west passes the ice divide, precipitation stops. Remaining water vapour could only be further depleted in the heavier isotopes, and therefore cannot explain increasing δD values in the east. Line 2 in Figure 8 therefore must he caused by water vapour advected from the east, which is also gradually depleted in D and 18O due to cooling during ascent. These air masses, however, are low in their water-vapour content and therefore cannot produce high accumulation rates. Again, precipitation stops when the air passes the ice divide. The higher δD(δ18O)-temperature gradient in the western part, is probably due to a little more isoharically influenced water-vapour trajectory over the vast ice plains of the western ice sheet, while the geographic profile shows steep slopes at the eastern ice margin, which essentially lead to adiabatic cooling of the water vapour during ascent. The westernmost site T05 deviates from line 1 in Figure 8, due to summer melting of the snow cover, causing distortion of the strata in the corresponding isotopic record. Therefore, this site was excluded from the linear regression analysis.
The assignment of the ice divide sites to one primary water-vapour trajectory (site T43 to line 1, NST08 and site T99 to line 2) is not straight-forward, because mixing of vapour from the west and the east can cause a breakdown of the linear relationship between δD(δ18O) and Tm on the ice divide. Reference FisherFisher (1992), for example, proposed mixing of vapour of both trajectories when applying a multi-source isotopic model at site T43 (Crête). However, the outcome of his model indicates that site T43 receives two-thirds of its water from the east, an assumption not supported by Figure 8. Most likely, the assignment of ice-divide sites to one of the regimes governing the isotopic signature is dependent on local circulation conditions. The geographical distribution of δ18O is plotted in Figure 5d. Here, the average δ18O value for site T47 is ~1.2% lower than for site T43 (corresponding to —1.8°C by Equation (6)), while the mean annual temperature is 0.15°C higher. Reference Clausen, Gundestrup, Johnsen, Bindschadler and ZuallvClausen and others (1988) also found excessively depleted δ18O values in the Crête area east of the ice divide. This means that the geographical isotope minimum along the actual EGIG line lies alike to the accumulation minimum but contrary to the temperature minimum east of the ice divide.
3.3.3. Deuterium excess
Essentially independent information from δ18O and δD about sources and transport pathways of the water vapour precipitated over Greenland is provided by the deuterium excess d = δD - 8δ18O (Reference Johnsen, Dansgaard and WhiteJohnsen and others, 1989; Reference FisherFisher, 1990). This parameter represents an indicator of kinetic effects occurring during the individual stages of the hydrological water cycle (i.e. evaporation at the ocean surface, and in high-polar regimes, also sublimation at snowflake formation), due to the different diffusion coefficients of HDO and H2 18O vapour. Furthermore, the deuterium excess shows a distinct annual variation, essentially caused by the difference in the temperature-dependence of the equilibrium evaporation of HDO and H2 18O. Reference Johnsen, Dansgaard and WhiteJohnsen and others (1989) used the Rayleigh model of Reference Jouzel and Merlivatjouzel and Merlivat (1984), which also considered kinetic effects during snow formation, to calculate the deuterium excess distribution along assumed water-vapour trajectories from different Atlantic Ocean source regions to central Greenland. Applying this single-source model, the mean level of d and its seasonal variation in central Greenland is mainly dependent on the evaporation temperature Te while the slope of the d-δ18O curve along the trajectory and, hence, the geographical distribution of d over the Greenland ice sheet, depends on the initial water-vapour mixing ratio we at the site of evaporation. Sensitivity studies on the model by Reference Johnsen, Dansgaard and WhiteJohnsen and others (1989) showed that the relationship is constant for an initial water-vapour mixing ratio of ~15 g kg and that variations of Te of -1°C lead to an offset of the d-δ18O curve of ~2%. The outcome of all single-source models, however, is very sensitive to the tuning of the saturation history of the advected water vapour, which determines kinetic effects during snowflake formation. Therefore, Reference FisherFisher (1991) proposed mixing of different saturation histories, thus varying d to be responsible for thr measured deuterium excess.
In Figure 5e, the geographical distribution of the average annual water-weighted mean of d is plotted, showing no significant geographical trend along the EGIG line (average d = 12.7%). Also, the annual variation of d (Fig. 9b) is comparable at all sites investigated (d amplitude ≈ 4%, phase shift relative to δ18O ≈ 2months). Using the approach by Fisher, this would imply that the mean composition of water vapour of different saturation histories is approximately equal throughout central Greenland.
On the basis of the findings of Johnsen and others the constancy of d along the whole EGIG line limits the original water-vapour mixing ratio at the site of evaporation to approximately 15 g kg−1. The uniformity of the spatial and temporal distribution of d also indicates the influence of possible water-vapour sources to he equal both on the western and the eastern slope of the central Greenland ice sheet. Pacific Ocean and North American source-area contributions, as stated by Reference Charles, Rind, Jouzel, Koester and FairbanksCharles and others (1994), which show significantly different values for we and Te , compared to the Atlantic Ocean, are therefore unlikely. Vapour from these sources is advected to die investigated from west to east and therefore their influence along the eastern slope of central Greenland is presumably efficiently blocked by the ice divide. The resolution of the model used by Reference Charles, Rind, Jouzel, Koester and FairbanksCharles and others (1994) is probably not sufficient to model the geographical distribution, since main barriers of the ice sheet and, hence, circulation on a synoptic scale cannot he taken into account appropriately.
Average annual variation of different isotopic and chemical parameters investigated at site T53 on EGIG-east: (a) δ18O; (b) deuterium excess d; (c) the water-weighted chloride concentration Cl.
The annual variation observed is in good agreement with the findings of Reference Johnsen, Dansgaard and WhiteJohnsen and others (1989) with respect to its amplitude and the phase shift relative to δ18O but our d level at site T47 (d= 12.2 ± 0.9%) is significantly higher than the value of Johnsen and others at site G (d = 8.9%), which is close to site T47. The average d value at site T99 (Summit) (d = 13.4%) is ~3% higher than the recent values determined at GISP2 (Reference Barlow, White, Barry, Rogers and GrootesBarlow and others, 1993). A systematic error in our study, due to the melted transport of samples from the western sites (as described in section 2.4), can be ruled out, because possible post-sampling evaporation would always lead to reduced d values. Note, in this context, that the average d values of the cores drilled at sites T43 and T99 in 1990 and 1992 are virtually identical. Also, a loss of fragile depth-hoar strata with lower deuterium-excess values during the drilling procedure is unlikely to he responsible for the offset, because detailed pit studies, which are not subject to loss of distinct strata, do not show significantly lower d levels either. Inter-instrumental differences in our data set further show the need for quality control and inter-comparison studies of the laboratories involved in isotopic ice-core analyses, to confirm the interpretation of this highly valuable hydrological tracer. So far, only very few deuterium-excess values are available for central Greenland and therefore no ultimate statement can be made However, implications derived in this paper from the geographical distribution of the deuterium excess hold Independently of the real d level. Assuming our d level is correct, and using the outcome of the model by Johnsen and others, the average d of 12.7% would imply a 2°C higher evaporation temperature Te thus moving the main source area of water precipitated over central Greenland further south.
4. Summary and Conclusions
The temporal variations of annual snow-accumulation rates and the annual water-weighted means of δ18O and δD do not reveal significant trends in climatic conditions over the EGIG line for the time span covered by the firn cores. Also, comparison of the accumulation rate in the eastern part with earlier measurements indicates no significant change during the last 30 years.
The geographical distributions of the average annual accumulation rates and the mean content suggest a primary water-vapour trajectory along the EGIG line from west to east, resulting in a precipitation shadowing east of the ice divide. Air masses advected from the east isotopically govern the eastern slope but are not able to produce high accumulations.
This is of relevance for the interpretation of central Greenland ice cores as climatic records So, the accumulation rate, which is thought to increase for warmer climates due to higher water-vapour content, will react differently east and west of the ice divide. While central-western sites probably show the expected increase, accumulation rates at the eastern sites probably do not react to temperature changes in a similar manner.
The isotope-climatological sub-division of the ice sheet into west and east also affects the interpretation of δD and δ18O as proxy temperature parameters in deep ice cores, dependent on their geographical position. A shift of the ice divide to the east, as modelled by Reference Anandakrishnan, Alley and WaddingtonAnandakrishnan and others (1994) for colder climates, for example, would move site T99 (recent summit) into the western area, which shows a δ18O offset relative to the eastern part. However, the magnitude of this effect (~1%) can only partially explain glacial-interglacial changes.
The geographical distribution and the seasonal variation of the deuterium excess are uniform over the whole EGIG line. This suggests the mean water-vapour contribution from different sources is approximately equal all over central Greenland throughout the year. Here, more detailed models have to be applied to reveal the advection and, therefore, precipitation history of the water vapour transported to Greenland.
Acknowledgements
Financial support for this study was provided by the Deutsche Forschungsgemeinschaft (DFG) and covering the field activities in 1990 by the Merck Stiftung and the Freundeskreis der Heidelberger Akademie der Wissenschaften. We also acknowledge the Institut für Vermessungskunde, Technische Universität Braunschweig, which was in charge of the logisitics of the traverses, the Physikalisches Institut, Universität Bern for the H2O2 measurements and the GSF Forschungszentrum für Umwelt und Gesundheit, Institut für Hydrologie, Neuherberg, for carrying out some of the δ18O and δD measurements. We should also like to thank D. A. Fisher and K. O. Munnich for very helpful comments on the draft of this paper and last, but not least, K. Geis and M. Anklin for their dedicated commitment by doing the glaciological field work during the 1990 traverse.
