2.1. Data sources and quality

The research area, Tvärminne Storfjärden, is located on the southwest Finnish coast, on the east side of Hanko Peninsula. Hydrographic data were collected three times per month, starting in mid-1926. Water temperature and salinity were measured at six different depths (0 m, 5 m, 10 m, 15 m, 20 m and 30 m). Measurements at 15 m depth started only in 1956. Data collection was interrupted from March 1940 to May 1942, when the region was occupied by the Soviet Union. Temperature was always measured by reversing thermometers. Water samples from each depth were collected into sample bottles and analyzed for salinity in the laboratory of the Finnish Institute of Marine Research (FIMR; fused into the Finnish Meteorological Institute (FMI) in 2009). After the 1950s, the methodology to determine water salinity changed from chemical titration to electrical conductivity measured with laboratory salinometers, increasing the accuracy to ±0.01‰ (Reference EmeryEmery and Thomson, 2001). Data used for our study were provided by FMI (Table 1).

2.2. Methods

After basic quality control and statistical analysis of the hydrographic data, we analyzed and plotted the seasonal variability over the past 85 years. In addition, the annual cycles of thermocline and the halocline depths were estimated from the steepest gradient of the mean temperature and salinity, respectively. We also estimated the seasonal variability of the heat content of the water body, taking the sea-ice formation and break-up into account. For monthly changes in heat content of the liquid water component (Q _w), we used

where p w is the sea-water density, c _w is the specific heat of the sea water, Tw is the vertically averaged water temperature, t is time and H is the water depth. To account for the ice, we then calculated the latent heat lost or gained during the ice formation or ice melt (Qi):

where p_i = 917kgm ^{−– 3} is the ice density, L is the latent heat of freezing and h is the ice thickness. Changes in heat content due to ice temperature variations were small and are ignored here. Since no ice thickness data were available in Tvärminne, we approximated the ice thickness from a degree-day model (Reference ZubovZubov, 1945):

where a = 2n _i /p _i L « 14.7 x 10–¹⁴m² (°Cs)^–2 (re_i is the thermal conductivity of ice), FDD is the freezing-degree-days (s), FDD₀ is a small correction term for the surface heat balance, and d«0.1 m is the insulation efficiency of the atmospheric surface layer. This derives from a simple analytical approach to the ice growth problem based on the following assumptions: (1) thermal inertia is ignored, (2) solar radiation is ignored, (3) oceanic heat flux is ignored, and (4) air temperature is known (Leppäranta and Myrberg, 2009). However, assumption (3) can be biased and snow accumulation is omitted from the model. An empirical modification was developed by fitting the parameter a to a nearby ice station in Hanko, where ice data were available from 1971 to 2011. The empirically estimated coefficient a* normally varies between a/2 and a. In this case, a*>7.1 x 10–¹⁴m² (°Cs)^–2 was quite low, most likely due to the snow accumulation on the top of the ice sheet. To calculate FDD, we used air temperature data from Tvärminne and Hanko.

Table 1.

Data used in the FIMR/FMI analysis

Finally, the freshwater budget was estimated based on the salt content (Q _s) of the water column:

where S is the vertically averaged salinity of the water column. Thus, Q _w +Q _i and Q _s are the true changes in the heat and salt content, respectively, and further investigations are needed to identify the cause of changes in heat and salt content.

3.1. Annual cycles

Figures 2–4 show the seasonal variability of the water temperature, salinity and density, respectively, based on the long-term monthly averages at each depth. The greatest variations of the hydrographic properties occurred closer to the surface since they were driven primarily by the atmosphere. The long-term averaged monthly surface water temperatures varied between 0.2°C and 16.7°C. The freezing point at the site was –0.3°C. The minimum temperature was observed in February, and the maximum in August. Near the bottom (30m depth), the temperature variation was smaller, from 0.6°C to 8.8°C. The minimum occurred in March and the maximum in September, with a lag of 1 month compared to the surface due to vertical diffusion of heat. Thermal stratification began after the ice break-up in spring. The largest temperature gradients occurred in August, while the stratification vanished after October, just before ice formation.

Fig. 2.

Annual cycle of water temperature at different depths, Tvärminne Storfjärden, 1927–2011.

Fig. 3.

Annual cycle of water salinity at different depths, Tvärminne Storfjärden, 1927–2011.

Fig. 4.

Annual cycle of water density (σ_T ) at different depths, Tvärminne Storfjärden, 1927-2011.

The long-term averaged surface salinity varied between 3.6 and 6.2 psu. Minimum values were observed in March and maximum in October, mainly due to vertical convection and advection. A remarkable drop in surface salinity, from 6.1 psu to 3.6 psu, occurred after December. This was most likely due to a freshwater layer originating from Pojo Bay, further inland, and that was held below the ice sheet during winter. One reason for this was reduced mixing due to the landfast ice barrier. In addition, circulation under the ice was probably weak so that there was limited exchange between the coastal waters and the central GoF. After surface salinity reached the March minimum, it started to increase at a constant rate of ∼1 psu month ^–1, reaching almost 5.5 psu in May. From May to October the rate of increase was lower (∼0.15 psu month^-1) and thereafter the salinity started to decrease towards the winter, along with the development of the winter stratification.

The water density (σT) values were between 2.8 and 5.3. As with the water temperature and salinity, the greatest variations occurred at the surface. There were two distinct surface density minima, in March and August. The surface minima act as seasonal barriers to convection. This could have implications for both the hydrography and ecology of the water body. In addition, it affects the ice season length, with ice formation occurring sooner due to the higher freezing point of the fresh water and with a delay in ice break-up due to the limited oceanic heat flux. In the winter, when the temperature of the water column was uniform with depth, density had a very similar distribution to salinity. In the summer, the density distribution changed as it was influenced by the strong thermal stratification. There was neutral stability in the water column throughout the year.

Figure 5 plots the seasonal variability of the water body temperature and salinity, minimum air temperature, precipitation and monthly maximum ice thickness. Due to the larger heat capacity of water than of air, and the winter phase changes, the average standard deviation (4.6°C) was ∼5°C less in water than in air (9.2°C). Seasonal distribution patterns were similar in air and water. However, water temperature values lagged by 1 month compared to the atmosphere throughout the year. As already noted, changes in water salinity occurred mainly at the surface. The depth-averaged salinity minimum in the water body occurred in April during ice melt, driven mainly by the surface layer. Seasonal variation of precipitation was also analyzed for 1927-2011. There was an increase in precipitation towards summer, coinciding with an increase in surface salinity. Thus, the increase in surface salinity was mainly associated with the wind forcing and advection from the outer GoF, where salinity is between 5.5 psu at the surface and 8 psu at the bottom (Reference BockBock, 1971). Favorable winds push saline water from the central to the coastal GoF. Here easterly winds pushed surface waters towards the Finnish coast, while westerly winds drove upwelling that may also have brought saline water to the study site. Sea-ice thickness reached a maximum in March, ∼0.33 m on average, and the ice was melted completely in April.

Fig. 5.

Seasonal variation of the vertically averaged water temperature (dashed line), minimum air temperature (red line), surface salinity, precipitation and maximum ice thickness in Tvärminne.

Monthly statistical information on water temperature and salinity in Storfjärden is listed in Table 2 and Table 3. The minimum, maximum, mean and standard deviation values are taken from the monthly average profiles. The maximum standard deviation of the temperature values was observed in August and the minimum in February and March, when the water was ice-covered.

The long-term averaged gradients of temperature, salinity and density were examined (Fig. 6). The salinity gradient was positive from the surface toward the bottom, while the temperature gradient was negative, indicating net heat and freshwater fluxes, respectively, to the atmosphere and from the land. The temperature gradient was ∼–0.1°Cm–¹. Taking the turbulent diffusion coefficient as D>10^–4m²s–¹ (e.g. Reference TennekesTennekes and Lumley, 1972), the mean heat loss became ∼40Wm^{– 2}, which must have been compensated by advection:

Fig. 6.

Long-term annual average gradient of temperature, salinity and density vs depth, Tvärminne Storfjärden, 1927–2012.

The halocline was at a shallow depth, held by a freshwater flux from Pojo Bay to the north and brackish water inflow from the central GoF. The annual cycles of the thermocline and the halocline were estimated by the steepest gradients between different depths. From January to August, the halocline was prominent mainly in the first 5 m. The underlying water masses did not interact directly with the wind-induced and convective mixing. Especially during the ice season, wind did not affect the surface water. The salinity gradient below the halocline was almost constant with depth. The thermocline was located mainly between 5 and 10m, but did not peak strongly in the mean distribution. After September, the surface water cooled, the water body became mixed and thermal stratification vanished.

In addition, we estimated the annual behavior of the water body heat content in Storfjärden, including the ice formation and break-up (Fig. 7). The total heat flux (black line) indicated heat loss in the autumn-winter months (September-February) mainly to the atmosphere, resulting in ice formation. In the spring-summer months (March-August) the water body gained heat and balanced the winter losses. The grey dotted line represents the heat fluxes due to ice formation and break-up. Heat loss from the water body to the atmosphere contributes to ice formation, while heat gain to the water body drives ice melting. There was a clear peak in heat input during ice melting in April (∼40 Wm^{– 2} ) . The atmospheric heat flux melted the ice in April and contributed to the heat content of the water column. The peak was significant because ice break-up is a much faster process than ice formation. The mean annual ice heat balance (grey dotted line) is always zero; however, the contribution of ice break-up to the heat budget was important because it occurred in a short period.

Fig. 7.

Monthly averaged heat fluxes from changes in water temperature and sea-ice formation and break-up in Tvärminne Storfjärden, 1927–2012. Positive values indicate heat fluxes from atmosphere to water.

Table 2.

Monthly statistics of the averaged water column temperature (°C) in Tvärminne Storfjärden

Fig. 8.

Monthly averaged freshwater budget in Tvärminne Storfjärden, 1927–2012. Red line represents maximum length of ice coverage. Positive fluxes correspond to salinity decrease, i.e. freshwater input. Negative fluxes correspond to salinity increase, i.e. saline water input.

In addition, we examined the seasonal trend of the surface salinity during its annual minimum (March). Over the 85 year record, March salinity increased significantly (by almost 2 psu), 1.2 psu higher than the long-term trend in the surface salinity (Reference Merkouriadi and LeppärantaMerkouriadi and Leppäranta, 2014a). The seasonal trend was also examined in October, when salinity was maximum at the surface. There was no trend indication for the surface salinity in October.

Fig. 9.

Spectral analysis of the surface salinity in Tvärminne Storfjärden.

The annual course of the freshwater budget was calculated based on the monthly salinity changes of the water body (Fig. 8). Positive values were observed mainly during the ice season and were probably associated with the limited mixing and water exchange with the central GoF. After the ice season, the freshwater budget turned negative. Since the GoF is a dilution basin by definition, the negative freshwater budget was linked with increased advection of more saline water from the Baltic proper.

Table 3.

Monthly statistics of the averaged water column salinity (psu) in Tvärminne Storfjärden

Finally, we estimated the power spectrum of the surface salinity in Tvärminne Storfjärden (Fig. 9). There was a strong peak in the 12 month cycle and a lower one in the 6month cycle, which was the second harmonic correction into the annual cycle. Very low frequencies show large levels due to the trend.

3.2. Seasonal long-term trends

In our previous work (Reference Merkouriadi and LeppärantaMerkouriadi and Leppäranta, 2014a) we examined the impact of sea ice on the annual cycle of the sea surface temperature (SST), by calculating the correlation of the SST and the number of real ice days. Real ice days are defined as the number of days of actual ice occurrence. Based on the significance of the correlation, the thermal memory of the system was estimated at 2-2.5 months. To examine the long-term trend of the amplitude of the annual climatological cycle, the average surface temperature in August and the maximum ice thickness in February are plotted together in Figure 10. There was an increasing trend in the August surface temperature of ∼1.5°C in 85 years, and the February maximum ice thickness decreased by almost 0.05 m. The linear regression was based on the least-squares estimation of the parameters. In a non-freezing water basin, when minimum and maximum temperature changes are of the same amount, the annual amplitude of the heat content is the same. In the case of sea-ice formation, phase changes need to be taken into account. An ice thickness decrease of 0.05 m means a heat content increase of 15 MJ m^–2corresponding to a water temperature that is 1.3°C higher in a 3 m layer. Thus, the annual amplitude had not essentially changed during the observations.

Fig. 10.

Interannual variability of the March surface salinity (top), August surface water temperature (middle) and February maximum ice thickness (bottom) in Tvärminne Störfjarden, 1927–2011.

3.3. Sea-ice influence

From the salinity distribution (Fig. 3) we noticed that greater variations occurred at the surface, especially in winter. In order to examine the role of ice formation, we grouped the years when the number of real ice days was >120 (1931, 1956, 1966 and 1970) and years without any ice (1975, 1992 and 1995), and compared the salinity distributions (Fig. 11 and 12). Real ice days refer to the number of days of actual ice occurrence. A large drop in surface salinity during winter (Fig. 3) was apparent in ice years, but almost vanished in ice-free years. This was probably associated with the weak mixing of the water body under the ice. During years with persistent ice cover, cold and fresh water entered Tvärminne from Pojo Bay in the north but did not become mixed, resulting in a very low-salinity layer just below the ice sheet.

Fig. 11.

Seasonal salinity distribution in years with extreme ice coverage (>120 days).

Fig. 12.

Seasonal salinity distribution in ice-free years.

The same comparison was also made for the temperature and the density distributions. During winter in ice-free years, the water-body temperature was between 0.5°C at the surface and 0.8°C at the bottom. In ice-covered years, the water temperature was negative and reached minimum values of -0.3°C at the surface and -0.15°C at the bottom. Ice formation of 0.3 m in the area corresponds to cooling of the water body by ∼0.6°C. The landfast ice cover prevents turbulent transfer of heat from the water. In open-water conditions turbulent heat losses can reach 50-100 Wm^{– 2}, but, in the case of an ice cover, heat must be transferred by molecular conduction, resulting in heat fluxes of 10–15 W m^–2. From the density distributions, we noticed stability in the water column in both ice-covered and ice-free years. In ice-covered years, minimum density values were observed in April, whereas in ice-free years they were observed in June.