3.1. Structure, stratigraphy and oxygen-isotopic characteristics

A granular upper layer and a columnar lower layer are characteristic structural features of most of the cores. The granular ice layer often consisted of fine-grained layers (mm-sized crystals, denoted g in Figs 2–4) and a type of ice termed transition ice (Fig. 5, with cm-sized crystals, denoted g/c in Figs 2–4), which is an intermediate state resulting from oscillations between the granular- and columnar-ice growth modes (see Reference Weeks, Gow, Kosloff and Digby-ArgusWeeks and others, 1990). Figure 5 shows the structure of transition ice. The total ice thickness varied between 17 and 72 cm, and the snow depth between 0 and 40 cm, both showing a significant increase in thickness towards the north (Fig. 6b).

Fig. 2.

Structural properties and vertical profiles of salinity and δ ¹⁸O for ice cores from the Bothnian Bay. The ice cores were divided into granular ice (g), a mixture of granular and columnar ice (g/c), termed transition ice, and columnar ice (c). The vertical scale depends on the individual ice-core length.

Fig. 3.

Same as Figure 2, but for ice in the Bothnian Sea.

Fig. 4.

Same as Figure 2, but for ice in the Archipelago Sea and the Gulf of Finland.

Fig. 5.

A vertical thin section of transition ice, an intermediate granular–columnar ice type (see Reference Eicken and LangeEicken and Lange, 1989), photographed between crossed polarizers. Section is from depth 5–14 cm at site BS-3. The scale on the left is in centimeters.

Fig. 6.

(a) Relationship of δ ¹⁸O in snow, under-ice water and columnar ice (average value for columnar ice in each core) with latitude; least-squares linear regression lines included, with R ² = 0.63, 0.82 and 0.65 for snow, water and columnar ice, respectively. (b) Relationship of ice and snow thickness with latitude; least-squares regression lines included, with R ² = 0.94 and 0.83 for ice and snow thickness, respectively.

There were large variations in the amount of the different ice types: columnar ice constituted 0–84% of the ice sheet, the average being 40.2%, whereas granular layers (g and g/c) constituted 16–100% (mean 59.8%). Ice termed transition ice, defined as intermediate columnar granular ice by Reference Eicken and LangeEicken and Lange (1989), constituted on average 28.2% of the total ice thickness. The average values were computed according to the absolute method described in Reference JeffriesJeffries (1997). The absolute method sums the total amount (in meters) of a particular ice type measured in each core and then divides that total by the total length of the cores (in meters) to give an absolute percentage for that ice type. All average values given hereafter are computed according to this method.

The relative contribution of different ice types showed no significant regional variability (Figs 2–4), the clearest exception being the significant contribution of transition ice in the Bothnian Bay (Fig. 2). The contribution of granular and columnar ice showed large variability, and no clear differences between the different basins were observed.

Major exceptions to the characteristic two-layer ice structure were observed at sites BS-4 and BB-1. At BS-4 and BB-1 the bottommost layer was composed of fine-grained granular ice. This layer could be due to the generation of frazil ice in nearby open water, which had been driven below the fast ice at these sites. At BS-4 the fast-ice edge was relatively close, and new ice formation was observed some days before the sampling, making this assumption plausible. Another possibility is refreezing of surface meltwaters when they come into contact with seawater at sub-zero temperatures. However, the under-ice water salinities did not indicate this as a plausible explanation, and the δ ¹⁸O values were close to that of columnar ice, indicating a sea-water origin for these layers. This basal granular layer, however, did not contribute significantly to the total ice thickness at either BS-4 or BB-1, although it indicates that the generation of frazil ice in open leads at the fast-ice edge can also contribute to the sea-ice thickness.

At site BB-2, the whole core consisted of granular ice, fine-grained on top, and large irregular crystals (10–20 mm) in the bottom layer, which could be regarded as transition ice (g/c in Fig. 2). Similar thick layers of large irregular crystals were observed at sites BB-3 and BB-4 (see Fig. 2). For comparison, in the Antarctic, transition ice has been associated with nearby leads and a rough hydrodynamic regime (see Reference Eicken and LangeEicken and Lange, 1989, table 1, defined as intermediate c/g). The δ ¹⁸O values of transition-ice layers did not differ significantly from those in columnar-ice layers within individual sea-ice cores (paired t test, P = 0.17), indicating a sea-water origin for transition ice. On the other hand, the average δ ¹⁸O for fine-grained granular layers had significantly more negative values than transition or columnar ice in individual sea-ice cores (paired t test, P < 0.05). The salinity of fine-grained granular layers was significantly higher than in either columnar or transition ice (paired t test, P < 0.05), whereas transition ice and columnar ice had similar salinities (paired t test, P = 0.40) in individual sea-ice cores. Both the oxygen isotopic properties and salinity therefore indicate that the origin of transition ice (g/c ice in Figs 2–4) is from downward growth at the ice–water interface (see also Reference Eicken and LangeEicken and Lange, 1989). Possible mechanisms would be: (1) incorporation of frazil ice into the growing ice sheet; (2) enhanced growth rates due to changes in atmospheric conditions; and (3) changes in the hydrodynamic regime at the ice–water interface (Reference Eicken and LangeEicken and Lange, 1989). All mechanisms seem plausible, and further studies are needed to decipher the origin of this ice type in the Baltic Sea, since its contribution, especially in the Bothnian Bay, is very significant. However, the presence of transition ice could indicate that the fast-ice edge, with open leads, was close by at some stages of ice growth, and affected the hydrodynamic regime beneath the ice cover at those sites. A relatively small amount of transition ice (denoted g/c in Figs 3 and 4) was found in the Bothnian Sea, Archipelago Sea and Gulf of Finland. This is probably because land-fast ice was found only in rather sheltered locations, in the archipelago, in skerries, bays or embayments where the growth conditions are more static.

A characteristic feature of the δ ¹⁸O profiles is that the most negative values are encountered in the topmost granular layers. The δ ¹⁸O values generally increase downwards in the ice, ranging from −16.2‰ to −5.9‰ (Figs 2–4). In snow, δ ¹⁸O ranged between −17.4‰ and −8.3‰ (mean −12.2‰); the range is somewhat smaller than for snow in the Weddell Sea (Reference Eicken, Lange, Hubberten and WadhamsEicken and others, 1994). There is a trend of increasing δ ¹⁸O values in both columnar ice and snow going southwards (Fig. 6a), which correlates with increasing surface water salinities (e.g. Reference Haapala and AleniusHaapala and Alenius, 1994) and the north–south gradient in air temperature. Surface water δ ¹⁸O values follow a north–south trend similar to that for snow and columnar ice (Fig. 6a).

The least negative values in columnar layers are often encountered in the bottommost parts where the fractionation factor would be greatest due to slower growth rates when the ice gets thicker. On the other hand, the influence of under-ice spreading of riverine waters during the winter would tend to decrease the δ ¹⁸O in surface waters; consequently δ ¹⁸O in the columnar layers would decrease as the ice season progresses. Decreasing δ ¹⁸O values toward the bottom of the ice cover, below the highly negative surface layers, are observed only at BB-2 and AS-1 (Figs 2 and 4); the most distinct decrease is at BB-2, where the crystal structure is very different to that at the other sites. Also the salinity profile at site BB-2 indicates an increasing fresh-water contribution during the ice season (Fig. 2). However, the ice salinity is also affected by lower growth rates, i.e. the lower the growth rate the more salt is expelled from the ice. The under-ice water salinity profile, however, showed no distinct fresh-water layer beneath the ice at BB-2.

3.2. Identification of snow ice, frazil ice and transition ice

To differentiate between snow ice and frazil ice in sea ice requires stable-isotope data since the granular crystal texture of these ice types is very similar (Reference Lange, Schlosser, Ackley, Wadhams and DieckmannLange and others, 1990). Several different isotopic criteria were used by Reference Jeffries, Worby, Morris and WeeksJeffries and others (1997) to identify snow-ice layers; some of them have been adopted and modified to the Baltic Sea conditions. The use of average values of the whole study area in the Baltic Sea would tend to either over- or underestimate the contribution of snow and snow ice, since there is rather large spatial variability in the isotopic composition of both sea water (e.g. Reference Punning, Vaikmäe and MäekiviPunning and others, 1991) and precipitation in the study region (see IAEA, Isotope Hydrology Information System, the ISOHIS Database, accessible at: http://isohis.iaea.org). Conditions can differ significantly between different sampling sites, as seen in Figure 6a, so use of the properties of snow, ice and under-ice water at individual sites is more reliable than use of the average values over the whole study area, as in Antarctica by, among others, Reference Jeffries, Worby, Morris and WeeksJeffries and others (1997).

Many of our observation sites are potentially influenced by under-ice spreading of riverine waters, and δ ¹⁸O values of the surface waters could then change or fluctuate during winter. However, under-ice salinity profiles indicated that none of the sampling sites had a distinct under-ice freshwater plume/layer during sampling. The northernmost sites (e.g. BB-6 and BB-5) have surface water salinities as low as 0.1 PSU year-round, so δ ¹⁸O values are not subject to drastic changes during winter. Observations at freeze-up and late winter (under-ice waters) in the Bothnian Bay indicate that surface water δ ¹⁸O values in the study area fluctuate about 1‰ between the start of freezing and the end of winter (Fig. 7). Values tend to be higher at freeze-up time, and decrease slowly during winter. However, this occurs only in areas where the river input has an effect (see Reference Alasaarela and MyllymaaAlasaarela and Myllymaa, 1978). Assuming that the linear relationship between salinity and δ ¹⁸O is a result of a two-component mixture, one component with high salinity (the water entering the Baltic Sea from the North Sea) and one with zero salinity (river inflow and precipitation), the δ ¹⁸O value for the freshwater component in the study area in the Bothnian Bay is about −12.3‰, estimated from the regression equations in Figure 7 (Reference Fröhlich, Grabczak and RozanskiFröhlich and others, 1988). The largest variability in δ ¹⁸O values in Figure 7 is in waters with a salinity close to zero, i.e. in the vicinity of the river mouths. The large scatter is mainly attributed to differences in the location of the drainage basins. There are larger differences between different rivers than between seasonal values within the rivers in the study area, based on our limited dataset. Also Reference Fröhlich, Grabczak and RozanskiFröhlich and others (1988) noted that the water in the Baltic is not strictly a simple two-component mixture with two end-members, the zero-salinity input and the high-salinity input from the North Sea, but that there are regional variations in the zero-salinity input.

Fig. 7.

Relationship of δ ¹⁸O with salinity in surface waters in the Bothnian Bay sampled in March 2000 (under-ice water), November 2000 (surface water during freeze-up) and March 2001 (under-ice water). Least-squares linear regression lines and equations included for whole dataset (solid line) and for under-ice samples (dashed line).

One potential influence of river waters spreading on sea water could be to produce frazil ice, when more saline water below the freezing point of fresh water would cool down the river water and produce ice from double diffusion processes (see, e.g., Reference McClimans, Steen and KjeldgaardMcClimans and others, 1979). Evidence for such ice formation has yet to be presented for the Baltic Sea, and is probably constrained to the very mouth areas of rivers. Ice produced in this manner would have a lower isotopic signature than ice of ambient sea-water origin and could potentially be assumed to be snow ice. This would give a faulty estimate for the snow contribution. A prerequisite for this would then be that river water, close to or at freezing temperatures, would be present at the sampling sites in the early stages of freezing. As noted above, no evidence for river water as distinct layers below the ice was observed at the time of sampling or during a survey prior to freeze-up in 2000, which also rules out the possibility of such ice formation in the early stages of ice growth. Earlier studies indicate that distinct river-water plumes form only under a prolonged ice cover (Reference Alasaarela and MyllymaaAlasaarela and Myllymaa, 1978). Frazil production due to double diffusion processes is probably present in close proximity to river mouths. However, observations of such ice-formation mechanisms are yet to be presented for the northern Baltic Sea.

The isotopic approach requires that the isotopic fractionation during ice formation is taken into account (Reference Jeffries, Worby, Morris and WeeksJeffries and others, 1997). Since we do not know the fractionation during ice formation, especially during snow-ice formation, we apply three different isotopic criteria to distinguish snow-ice layers from other ice types. The division between frazil ice and snow-ice layers by visual inspection is difficult, if not impossible. However, since the transition ice (denoted g/c in Figs 2–4) was shown to have a sea-water origin (see section 3.1), those layers are not considered to have any snow contribution. Only layers denoted as g in Figures 2–4, i.e. fine-grained granular-ice layers, are considered to potentially have a snow contribution. Three isotopic criteria, described below, are used to determine ice layers as snow ice or frazil ice.

Criterion 1 assumes that any granular-ice layer (denoted g in Figures 2–4) with δ ¹⁸O values less than , the sea-water δ ¹⁸O value at each site, has a snow contribution (cf. criterion 1, Reference Jeffries, Worby, Morris and WeeksJeffries and others, 1997). This criterion assumes no fractionation during snow-ice formation and can be regarded as a minimum estimate for the snow-ice contribution.

Criterion 2 takes into account that some fractionation will occur during snow-ice formation, and fine-grained granular layers with δ ¹⁸O values below (cf. criterion 2, Reference Jeffries, Worby, Morris and WeeksJeffries and others, 1997). The observed value for isotopic fractionation during columnar-ice growth in the Gulf of Finland has been around 2‰ (Reference KawamuraKawamura and others, 2001; derived as the difference between δ ¹⁸O value of sea water and bottommost ice sections analyzed). The maximum fractionation during fresh-water ice formation, at zero growth rate, is assumed to be close to 3‰ (e.g. Reference O’NeilO’Neil, 1968; Reference Lehmann and SiegenthalerLehmann and Siegenthaler, 1991). However, for a 4‰ NaCl solution, Reference Craig and HomCraig and Hom (1968) found a fractionation factor of 2.65‰. Furthermore, ice grown in natural conditions will not necessarily be formed in isotopic equilibrium, which depends on freezing rate and water mixing at the ice–water interface (Reference Lehmann and SiegenthalerLehmann and Siegenthaler, 1991). For fractionation between sea water and sea ice, a value of 2.70% seems to apply during the slowest growth rates realized under natural conditions (Reference Craig and HomCraig and Hom, 1968; Reference Melling and MooreMelling and Moore, 1995; Reference Eicken and JeffriesEicken, 1998). Since the fractionation processes during snow-ice formation have not been studied in great detail, we assume that fractionation is not as close to equilibrium as during columnar-ice formation, because snow-ice layers usually grow faster, and the fractionation will be less than that for columnar ice. Therefore we choose a value for fractionation that is between the observed value for Baltic Sea ice and zero fractionation. This represents an intermediate estimate for the contribution of snow ice.

Criterion 3 gives a plausible upper limit for the contribution of snow ice. Any granular-ice layer (denoted g in Figs 2–4) with δ ¹⁸O values less than δ ¹⁸ O, the average δ ¹⁸O value for columnar ice at each site, has a snow contribution (cf. criterion 3, Reference Jeffries, Worby, Morris and WeeksJeffries and others, 1997). However, this assumption probably underestimates the amount of frazil ice, because isotopic fractionation during congelation (columnar-ice) growth is higher than during frazil-ice formation. Criterion 3 provides a plausible maximum estimate for the thickness contribution of snow-ice layers, and a minimum estimate for the amount of frazil ice. This assumption was also used by Reference KawamuraKawamura and others (2001) to estimate the snow contribution in sea ice in the Gulf of Finland.

Of the above isotopic criteria those using values compared to the sea-water value at each site, i.e. criteria 1 and 2, would tend to underestimate, rather than overestimate, the amount of snow ice. This is because if the δ ¹⁸O of sea water at the sites decreased during winter, as would be the case if the observational sites were influenced by runoff from land, one would use lower δ ¹⁸O values of sea water as reference than would have been incorporated into the sea-ice cover earlier in the season.

The results for the thickness contribution of snow ice and granular ice using the different criteria described above are given in Table 2. Based on the above criteria, 23.3–31.6% of the total ice thickness is composed of snow ice (i.e. a mixture of sea water and snow), and 31.6% can be regarded as the best (i.e. intermediate) estimate. The results using criteria 2 and 3 give identical results, because all transition ice layers are excluded from the identification process. Criterion 2 already incorporates all fine-grained layers, indicating that no frazil ice is present. In individual cores, the snow-ice thickness contribution ranged between 6% and 73% (criterion 2). The contribution of granular ice (frazil and transition ice) is of the same order as the contribution of snow ice, with transition ice having a substantial contribution in the Bothnian Bay, and frazil ice having a negligible contribution except for criterion 1 (8%).

Table 2.

Contribution of granular-ice (frazil ice and transition ice) and snow-ice layers (%) to the total ice thickness averaged over all sites

3.3. Snow fractions

Structural analyses alone do not accurately quantify the amount of snow ice and the amount of snow incorporated into the ice sheet. For such information, δ ¹⁸O data have been used to distinguish between different ice types or to determine the contribution of snow to the sea-ice cover in Antarctic regions (e.g. Reference Gow and EpsteinGow and Epstein, 1972; Reference Lange, Schlosser, Ackley, Wadhams and DieckmannLange and others, 1990; Reference Eicken, Lange, Hubberten and WadhamsEicken and others, 1994; Reference Jeffries, Shaw, Morris, Veazey and KrouseJeffries and others, 1994; Reference JeffriesJeffries, 1997; Reference Kawamura, Ohshima, Takizawa and UshioKawamura and others, 1997). In order to quantitatively estimate the fraction of snow in granular-ice layers and sea-ice cores, they all assumed that only granular-ice layers had a contribution from snow. A simplified model developed by Reference Jeffries, Shaw, Morris, Veazey and KrouseJeffries and others (1994) from the original model by Reference Lange, Schlosser, Ackley, Wadhams and DieckmannLange and others (1990) is:

where f _s is the snow fraction (by mass) and f _sw the seawater fraction (by mass) of the granular-ice section in question and δ _s is the δ ¹⁸O value for snow at each site (−8.3 to −17.4‰ (Fig. 6)). For the δ _sw value, we use the sea-water value at each site, with 1 added , which allows for some isotopic fractionation during the formation of snow ice (see Reference Jeffries, Worby, Morris and WeeksJeffries and others, 1997), and δ is the δ ¹⁸O value for the granular-ice section for which the snow fraction is being determined.

The total snow fraction (by mass) of the entire ice thickness in each core, F _m, was obtained according to:

where f _s is the snow fraction and h _i is the thickness of individual snow-ice layers in the core for which F _m is being determined. h _tot is the total length of the ice core. Several snow-ice layers were encountered in the cores, because the topmost fine-grained layer was usually divided into several sections along stratigraphic boundaries.

Snow-fraction (F _m) values resulting from applying Equations (1–3) to granular-ice layers, which have been identified as snow-ice layers using criteria 1 and 2 (see section 3.2), are shown in Table 3. On average, 20.7% of the ice in the study region consisted of snow using the intermediate criterion. Applying the above equation to snow-ice layers identified by criterion 3 gives the same values as criterion 2. The minimum estimate (criterion 1) gives an average snow fraction of 18.3%.

Table 3.

The snow fraction, F _m (%), at the different sites derived with the isotopic mass-balance model (Equations (1) and (2)) applied to snow-ice layers as identified by criteria 1 and 2 (see text)

f _s values for individual snow-ice layers ranged from 3% to 213%, with an average of 54% for those layers with values < 100%. Those above 100% were corrected to a value of 100% when the F _m values were computed, even though a value of 100% or more indicates that the whole layer consisted of snow only, which is not usually plausible unless superimposed ice is formed from melting snow. Only at one site, AS-1, is this a plausible explanation for f _s values > 100%. In the rest of the cases, the δ _s value at that site was higher than the δ ¹⁸O values for the corresponding granular-ice layers. This implies that temporal variability in δ ¹⁸O values for the snow should be taken into account in order to improve the estimate for the δ _s value for snow included in the ice cover. However, using the average δ ¹⁸O values for snow at all sites, or for each basin, in Equations (1) and (2), instead of the individual value for each site, does not improve or change the results very much. This indicates that the temporal variability of the snow properties is the most important parameter to consider when improving the estimates presented here. There can be large seasonal fluctuations in the isotopic signature of precipitation in the Baltic Sea area (Reference Punning, Vaikmäe and MäekiviPunning and others, 1991). However, there are virtually no data on the δ ¹⁸O of snow or precipitation in the Gulf of Bothnia or along the coasts of Finland (see, e.g., Reference Fröhlich, Grabczak and RozanskiFröhlich and others, 1988; Reference Punning, Vaikmäe and MäekiviPunning and others, 1991), which regrettably undermines an effort to present more exact figures here. Therefore our maximum estimate (20.7%) should be considered as a plausible upper limit for the snow contribution to the sea-ice mass balance in the study region.

Applying the methods used in this study to the data presented by Reference KawamuraKawamura and others (2001), the estimates for the snow fractions in the Gulf of Finland in winter 1999 decrease to 20% at maximum (compared to 26% in Reference KawamuraKawamura and others (2001)). The method used by Reference KawamuraKawamura and others (2001), described as criterion 3 in this study, tends to overestimate the snow contribution in the snow-ice layers (Reference Jeffries, Worby, Morris and WeeksJeffries and others, 1997). Furthermore Reference KawamuraKawamura and others (2001) used a δ _sw value assuming that the fractionation during snow-ice formation is as high as during columnar-ice growth, which is probably not the case since snow-ice growth is usually faster than growth at the ice–water interface. Therefore the fractionation is not as close to equilibrium during snow-ice formation.

The temporal variability of the snow properties (δ ¹⁸O and density) should also be better taken into account, and in the Baltic Sea area there is a need for more systematic measurements of the isotopic composition of sea water, river inflow and precipitation (including snow) (see Reference Fröhlich, Grabczak and RozanskiFröhlich and others, 1988), especially in the Gulf of Bothnia where data are virtually non-existent. These observations would be valuable in sea-ice studies, as well as in more detailed studies of the water balance in the region (Reference Fröhlich, Grabczak and RozanskiFröhlich and others, 1988). There is also a need for more detailed information about isotopic fractionation during consolidation of mixtures of snow and sea water, in order to be able to present more accurate estimates of the snow contribution to sea-ice mass balance, not only in the Baltic Sea area, but also in areas where isotopic methods have been widely used (e.g. in Antarctic sea ice).