Introduction

It is well known that many parameters in the Greenland ice sheet display a seasonal variability. The oxygen isotope composition of the ice, δ¹⁸O, varies seasonally and is used for absolute dating of the deep Greenland ice cores by counting of individual annual layers. The amount of soluble and insoluble impurities in the ice also shows a seasonal variation (Reference SteffensenSteffensen, 1988). It is generally believed that crystal growth is related to the impurities in the ice (e.g. Reference Alley, Perepezko and BentleyAlley and others, 1986; Reference Alley and WoodsAlley and Woods, 1996), and therefore expected that the crystal sizes also vary seasonally. To our knowledge, a seasonal variation of crystal properties has not been reported until now.

The long-term variability in ice crystal properties, texture and fabric has been established for several deep Greenland ice cores (Reference GowGow and others, 1997; Reference Thorsteinsson, Kipfstuhl and MillerThorsteinsson and others, 1997; Reference Wang, Thorsteinsson, Kipfstuhl, Miller, Dahl-Jensen and ShojiWang and others, 2002). Often the dimensions of the samples used in fabric and texture studies are chosen in order to fit an instrument, rather than to cover a certain time interval in the core. A significant seasonal variability in crystal properties could imply scatter around the general trend, as observed, for example, in the upper 500 m of the Greenland Icecore Project (GRIP) core (Reference Thorsteinsson, Kipfstuhl and MillerThorsteinsson and others, 1997). If the seasonality occurs, one could consider sampling in such a way that the results would not be biased by the effect.

The purpose of this work is to quantify the seasonal and the interannual variability in crystal properties in a section from the NorthGRIP ice core at about 301 m depth. North-GRIP is a deep drilling site located in North Greenland at 75.10° N, 42.32° W with an annual accumulation of 19.5 cm ice equivalent (Reference Dahl-JensenDahl-Jensen and others, 2002). We compare the results with δ18O and the concentration of chemical impurities in the ice, and discuss which impurities chiefly influence the crystal properties at NorthGRIP.

Methods

A continuous profile of vertical thin sections was obtained from the NorthGRIP ice core in the depth interval 300.85– 301.94 m. This depth is within the normal grain-growth regime, where crystal areas increase linearly with time and important polygonization has not yet initiated (Reference SvenssonSvensson and others, 2003). Inspection of the electrical conductivity measurements (ECM) shows that the sampled ice has no detectable volcanic markers. The profile consists of eleven 10610 cm₂ thin sections, prepared in the freezer at the University of Copenhagen using standard techniques (Reference Hansen, Svensson, Wang and SteffensenHansen and others, 2002). The orientation of the ice core with respect to the core axis is lost when the core is being retrieved, because the drill rotates on its passage through the borehole; therefore, the orientation of the thin sections with respect to this axis is unknown. The logging of the core ensures, however, that the thin sections prepared for this study all have the same (unknown) orientation. In order to produce optimal results, the thin sections have a maximum thickness of 0.5 mm.

The thin sections were measured at the University of Copenhagen on a new automatic c-axis analyzer, developed in Australia by D. S. Russell-Head. The instrument produces digital images of the sample at a high resolution of 20 pixel mm_–1. The thin section is placed on an xy-stage between two crossed linear polarizers, and images of the section are taken with a camera for each 7.5° rotation of the polarizers. A deviator prism placed between the sample and the upper polarizer enables imaging of light that has traversed the sample at an angle. Crystals change in brightness and color as the polarizers are rotated; from these data, the c-axis orientation is calculated for each individual pixel in the sample images with accuracy better than 5°. The measurement time is approximately 2 hours for a 10 x 10 cm₂ sample.

From the digital images the outline of the individual crystals in the thin sections is drawn by an automatic method and by manual editing (Reference Hansen, Svensson, Wang and SteffensenHansen and others, 2002). The high resolution allows crystals as small as 0.2 mm₂ to be detected. Only complete crystals are included in the analysis, i.e. crystals that intersect with the border of the sample are discarded. After being identified, the crystals are numbered and each crystal is assigned a c-axis orientation equal to the mean c-axis orientation of all the pixels within the crystal. Sometimes the c-axis orientation cannot be determined, if a crystal is too small or if two crystals overlap significantly.

Chemistry samples in 5 cm resolution have been analyzed for the concentrations of Ca²⁺, Na⁺, K⁺, Mg²⁺, Cl^–, NO_{3 –} and SO₄₂– in the ice by ion chromatography at the University of Copenhagen as described in Reference SteffensenSteffensen (1997). Because the δ₁₈O, the chemistry and the thin-section samples are not obtained from the same ice, there may be an offset in depth between the various measurements of the order of millimeters.

Results

A total of about 13500 crystals have been detected in the 11 thin sections. Thecrystal areas fall in the range 0.2–80mm², and the mean crystal area of all crystals is 6.51 mm². The crystal outlines of three thin sections are shown in Figure 1.

Fig. 1. Crystal outlines of three thin sections and the corresponding average crystal area at 1cm resolution (thick step curve). The 5 cm band of smaller crystals below 301.00 m depth appears during spring of AD 471 (see Fig. 2).

Figure 2 shows the average crystal areas, the degree of orientation of c axes, the Ca²⁺ concentration and the “back-diffused” δ¹⁸O signal of the ice (personal communication from S. J. Johnsen, 2001) on a depth scale. The“back-diffused” δ¹⁸O signal is an enhanced, deconvoluted δ¹⁸O profile, which takes into account the damping of the seasonal signal due to diffusionof water molecules in the firn (Reference JohnsenJohnsen, 1977). The seasonal variability is thus more exposed in the “back-diffused” δ¹⁸O than in the measured δ¹⁸O profile. From reference horizons at other depths in the ice and from counting of annual layers, the sampled ice is known to cover the years AD466–471 (personal communication from H.B. Clausen, 2001).

Fig. 2. The temporal variability of various parameters. (a) Degree of orientation of c-axes (R). (b) Back-diffused δ18O (personal communication from S. J. Johnsen, 2001) with indication of individual years (personal communication from H. B. Clausen, 2001). (c) Mean crystal area. (d) Concentration of Ca2+ in the ice. Vertical lines indicate spring minimums in crystal area. The resolution is 2.5 cm except for Ca2+, which is at 5 cm resolution.

Each crystal is assigned a depth in the ice corresponding to the position of its center of mass. When grouped into 2.5 cm depth intervals, the crystal areas show a strong sea-sonalvariabilityof 430% deviation from the average value of 6.7 mm² (the average of the mean crystal areas within the 2.5 cm intervals) (Fig. 2). The crystal areas are not directly correlated with δ¹⁸O, but in ice deposited during each spring a local minimum in crystal areas is formed. Although the seasonal variability in crystal areas does not reproduce from year to year, the band of smaller spring crystals appears very consistently. According to the dating of the core, the minor dip in crystal areas at around 301.3 m depth does not, however, correspond to a spring season. The interann-ual variability in crystal area is significantly lower than the seasonal variability.

The c-axis orientation has been determined for 97% of the identified crystals in the samples. The overall degree of orientation of the c axes is described by the R parameter (see, e.g., Reference Wang, Thorsteinsson, Kipfstuhl, Miller, Dahl-Jensen and ShojiWang and others 2002), so R = 0% corresponds to randomly distributed c axes, and R =100% corresponds to a single maximum with all c axesbeing vertical. For all c axeswe find R = 35.50%, in accordance with the general R level determined at this depth (Reference SvenssonSvensson and others, 2003). At a 2.5 cm depth resolution, the R parameter shows some fluctuation around the average value (Fig. 2), but there is no correlation to any of the other ice-core parameters. We thus conclude that the fabric does not exhibit seasonal variability. In Figure 3a, the R parameter is shown as a function of crystal area. Some scatter is seen around the average level, but with no systematic dependence on crystal area.

Fig. 3. (a) Degree of orientation ofcaxes (R) for area classes containing 4100 crystals (points) and a linear fit (curve). (b) Crystal area distribution of all measured crystals (step curve) and afitted log-normal distribution (continuous curve) with indication of the log-normal mode μ = 3.40 mm₂ and the log-normal standard deviation σ =3.05. (c) Same, but for the depth interval 301.00–301.05m, which contains relatively small crystals (see Fig 1). (d) Same, but for the depth interval 301.10–301.15 m with large crystals.

As shown in previous studies of Greenland texture (Reference Thorsteinsson, Kipfstuhl and MillerThorsteinsson and others, 1997; Reference Hansen, Svensson, Wang and SteffensenHansen and others, 2002), the crystal area distribution is close to a lognormal distribution. Our results confirm this for the full ensemble of identified crystals (Fig. 3b), as well as for two subsets, from depth intervals with “small” and “large” crystals (Fig. 3c and d, respectively). Although the total distribution in Figure 3b is indeed close to a lognormal distribution, a slight asymmetry is noticed with a “shoulder” of smaller crystals centred around 1mm² (see Reference SvenssonSvensson and others, 2003, for discussion). The mean of the lognormal fit to the total distribution is 3.40 mm², and the lognormal standard deviation is 3.05.

Discussion

The mean crystal areas are compared to the concentration of chemical impurities of Ca²⁺, Na⁺, K⁺, Mg²⁺, Cl^–, NO^3– and SO^42– at a 5 cm resolution, which is the resolution of the chemistry measurements (Table 1). Although this resolution is limited in resolving the fast transitions in crystal sizes, which occur on a cm scale (Fig. 1), the crystal areas show quite strong anticorrelation to several of the ionic species. We find that generally the best correlation with crystal areas is obtained when applying the logarithm to the concentration of impurities. The strongest anticorrelation appears for Ca²⁺, but Mg²⁺, Cl^–, K⁺ and Na⁺ also correlate quite strongly with crystal areas, whereas NO^3– and SO^42– show poor correlation. The crystal areas are plotted against log (Ca²⁺) in Figure 4. They are also compared to the ECM signal and to the δ¹⁸O of the ice at a 2.5 cm resolution, but here no correlation is observed (Table 1).

Fig. 4. Crystal area vs logarithm of Ca²⁺concentration in the ice at a 5 cm resolution (points) and a logarithmic fit (curve). Area = –0.97 log (Ca²⁺) + 8.31 (R² =0.58).

Table 1. Correlation coefficients between the mean crystal area and the logarithm of impurity concentrations of the ice at a 5 cm resolution, and between the mean crystal area and ECM and “back-diffused” δ¹⁸O of the ice at a 2.5 cm resolution. 95% confidence level is 0.37 (N = 20)

The concentration of insoluble impurities (dust) in the ice has not been measured, but both Ca²⁺ and Mg²⁺ are major constituents in dust, and Ca²⁺ is often used as a proxy for the dust concentration of the ice (Reference Steffensen, Clausen, Hammer, Legrand and AngelisSteffensen, 1997). Because of the high anticorrelation between crystal areas and both Ca²⁺ and Mg²⁺, it is likely that the dust concentration of the ice plays a major role in controlling the crystal size variability on a seasonal time-scale.

Conclusions

A continuous 1.1m long section from around 301 m depth of the NorthGRIP ice core has been investigated for crystal properties and for impurity concentrations. A strong sea-sonalvariationin crystal areas is observed, with the smallest crystals appearing in spring, when the concentration of Ca²⁺ also has its maximum. This result suggests that it may be possible to establish an empirical relationship between crystal areas and dust concentration of Greenland Holocene ice. Thecrystal area distribution is very close to a lognormal distribution. The fabric of the ice does not display a detectable variation with seasons, nor does it vary as a function of crystal sizes. We thus conclude that the fabric of the ice develops independently of the texture and of the chemical impurities in the investigated ice.