1 Introduction

Saimaa ringed seal (Pusa hispida saimensis) is one of the most endangered pinnipeds in the world with a population estimate of around 480 individuals (Metsähallitus 2023). This southern subspecies of ringed seal is an endemic to Finland and only lives in Lake Saimaa area. Active conservation measures have turned the small population to slow growth in the last decades (Kunnasranta et al. 2021). However, it is still under threat of extinction and climate change is one of the main risks for the recovery of the population.

The breeding success of ringed seals depend on sufficient ice and snow conditions (Kovacs et al. 2012). The Saimaa ringed seals build their lairs into the wind-driven snowdrifts on the lake ice at shorelines with direct access through ice to water (Sipilä 1990). The seals haul out in subnivean lairs and in February–March give birth to a single pup in the shelter of the lair. Birth lair provides protection for the pup against predators and harsh climate conditions. During mild winters when pups are without the lair, over one third of them may die before weaning (Auttila 2015). In addition to the height of the snowdrift, the layered dense structure of snow is important for the lair to last longer without collapsing. Time and varying weather conditions are needed to form sufficiently large snowdrifts and their dense layered structure. The female and the pup use the lair until the end of the nursing season in mid-May or until it melts, when nursing continues on ice and after the ice break-up on terrestrial habitat (Niemi et al. 2013; Kunnasranta et al. 2021).

To compensate for the negative effects of climate change to the pup survival of the Saimaa ringed seal, novel methods have been developed and implemented. Man-made snowdrifts have been piled up at shorelines of Lake Saimaa since 2014 (Auttila et al. 2014), and altogether over 300 pups have been born in those snowdrifts. However, there has already been winters when ice and snow cover has not been sufficient even for man-made snowdrifts. Therefore, testing and development of floating artificial nest boxes has started, and the seals have accepted the nests as lairs (Kunnasranta et al. 2022). Despite the compensation methods, the reduction of ice and snow cover and the shortening of the ice/snow covered season poses a significant risk for the long-term survival of the seal (Kunnasranta et al. 2021).

Studies based on long time series of ice phenology (since 1885) and lake ice thickness (since 1917) observations in Lake Saimaa show remarkable changes in the ice cover already during the last century. The maximum ice thickness decreased 1.5 cm/decade in 1925–2002 period and the ice duration decreased 1.3 and 2.2 days/decade in two different observation sites during the twentieth century (Korhonen 2006). The trends in ice variables in later 30 years period (1976–2005) were steeper than in the last 100- and 150-year periods generally in Northern Hemisphere (Korhonen 2006; Benson et al. 2012). In Lake Saimaa the decreasing trends of ice thickness and duration have continued in the last two decades. New records in late freeze-up, early break-up dates and thinner ice cover were set several times already in the beginning of the twenty-first century (Finnish Environment Institute 2023).

Modelling of ice cover in Finland has concentrated on sea ice modelling, but the same models have been applied in lake ice modelling as well. Saloranta (2000) developed a numerical one-dimensional sea ice thermodynamic model for simulation of ice thickness, snow and snow-ice. This model was used as an integrated part of a process based Multi-year Lake simulation model (MyLake) (Saloranta and Andersen 2007). The HIGHTSI model was initially developed for modelling seasonal snow and sea ice thermodynamics (Launiainen and Cheng 1998) but has also been used for lake ice modelling in several studies (Semmler et al. 2012; Yang et al. 2012, 2013; Cheng et al. 2014). Snow transportation models have been used for simulating snowdrift formation for terrestrial, hydrological and sea ice applications (e.g. Walter et al. 2004; Liston and Elder 2006; Liston et al. 2018). However, there is still lack of studies concentrating on lake ice and snow of large lakes and especially formation of the snowdrifts essential for the Saimaa ringed seal.

Climate change has significant effect on ice phenology. The long-term time series and ice modelling results show that the ice freeze-up date occurs approximately one week later and break-up date one week earlier, when temperature increases 1 °C, while changes in ice thickness are also dependent on changes in precipitation and snow cover (Brown and Duguay 2010; Leppäranta 2023). Dibike et al. (2011) studied the effect of climate change on ice phenology and lake water thermal structure with MyLake model using a Canadian Global Climate Model projection based on SRES A2 emission scenario in the Northern Hemisphere from baseline period of 1960–99 to 2040–79. They found that the freeze-up was delayed 5–20 days and break-up advanced slightly more, 10–30 days, while maximum lake ice thickness decreased 10–50 cm. The changes in snow depth on ice and snow-ice varied depending on the geographical location.

The aim of this study is to deepen the understanding of the impacts of climate change on formation of snowdrifts on lake ice and consequent impacts on breeding habitat conditions of the Saimaa ringed seal. The observed changes in ice duration and ice thickness in past two decades and effect of mild winters on the higher pre-weaning pup mortality are shown as the early signs of the breeding habitat decline of the seal. A lake ice model of Watershed Simulation and Forecasting System (WSFS-Ice) was developed to simulate the ice cover evolution on Lake Saimaa concentrating on the snow conditions and snowdrift formation on ice. The main focus is on the impacts of climate change on ice growth and snow conditions on ice during the Saimaa ringed seal pupping season between 1st of February and 20th of March.

2 Materials and methods

2.1 Study area and materials

Lake Saimaa is the largest lake in Finland and fourth largest natural freshwater lake in Europe. It is located in the Vuoksi watershed in south-eastern Finland (Fig. 1). Lake Saimaa area is approximately 4 280 km2 and mean depth is 10.0 m. Lake Saimaa consists of several connected lakes (Southern Saimaa, Ukonvesi, Pihlajavesi, Puruvesi, Haukivesi, Enonvesi, Pyyvesi, Orivesi and Pyhäselkä) with approximately the same water level. The lake is characterized by numerous islands (13 710) and islets and the long winding shoreline with length of 14 850 km (Kuusisto 1999; Finnish Environment Institute 2023).

Fig. 1
figure 1

Location of Lake Saimaa and ice observation sites

Full size image

The Saimaa ringed seal can only be found in Lake Saimaa. Around half of the seal population inhabits Lake Pihlajavesi and Lake Haukivesi, which are in central parts of the lake Saimaa (Metsähallitus 2023). As a result of the recovery of the population, the seals have returned to areas from which they had previously disappeared because of human activity. Nowadays the seals can be found anywhere in Lake Saimaa.

There are four national ice observation sites located in different parts of Lake Saimaa (Fig. 1). The longest time series since 1885 are available for ice freeze-up and break-up dates in Lauritsala and Haukivesi observation sites. In Pihlajavesi the long time series of freeze-up and break-up dates since 1911 include some data gaps since year 2000. In Pyhäselkä only a few observations are available since 1960.

Ice thickness is typically measured at 10- or 15-day intervals. Snow depth on ice is normally measured at the same time but includes some data gaps. The longest time series is available in Lauritsala since 1917. In Haukivesi and Pihlajavesi ice thickness has been observed since 1961 and in Pyhäselkä since 1992.

The climate of Lake Saimaa is temperate. In the 1981–2010 period the mean annual temperature was 4.8 °C in southernmost Lauritsala, 4.3 °C in Haukivesi and 3.8 °C in northernmost Pyhäselkä observation sites. Annual precipitation was on average approximately 660 mm, of which 30% was snow. In 1981–2010 mean freeze-up date in Lauritsala was on 11th of December and in Haukivesi one week earlier (4th of December), mean break-up date in Lauritsala was in the end of April (26th of April) and in Haukivesi almost two weeks later (8th of May). Mean maximum ice thickness in Lauritsala was 45 cm, in Haukivesi 50 cm and in Pyhäselkä 62 cm.

The long-term time series and trend analysis of ice phenology and ice thickness observations were studied by Korhonen (2006) in the beginning of this century. However, in last two decades, there has been significant decrease in ice thickness and ice duration in Lake Saimaa, which have impacted the breeding habitat of the ringed seal. Therefore, the main findings related to ice observations are summarized in the beginning of the results section.

Snow depth in snowdrifts close to the shorelines of small islands and islets has been systematically observed at 10 snow lines in Haukivesi and Pihlajavesi by the Centre for Economic Development, Transport and the Environment for Southeast Finland since 2017. Snow depths in the snow lines are observed in the beginning of March (between March 4th and March 13th) each year. The snow depth is measured at five to six-meter-long snow lines every half meter from the shore. The selected snow lines represent potential seal nesting sites, but the known nesting sites were not included in the measurements to avoid disturbance.

2.2 Lake ice model

The WSFS-Ice model is part of the Watershed Simulation and Forecasting System (WSFS), which is based on the nation-wide hydrological model for forecasting floods and droughts in Finland and has been developed in the Finnish Environment Institute (Syke) (Vehviläinen 1992; Vehviläinen et al. 2005; Menberu et al. 2024). The meteorological input variables of the WSFS hydrological model including air temperature, precipitation, wind speed, relative humidity and cloudiness are provided by the Finnish Meteorological Institute. The WSFS model covers whole Finland including transboundary catchment. The WSFS-Ice model can also be used in other regions with seasonal ice-cover, if input variables and freeze-up dates are available. The model timestep is one day and the code is written in Fortran language. The detailed description of the model and the parameter values used are given in Online Resource 1.

The one-dimensional WSFS-Ice model was developed to simulate congelation ice thickness, snow cover on ice, slush layer and snow-ice formation (Fig. 2). In the model simulation congelation ice grows in ice-water interface and releases energy, which conducts through the ice and snow cover, when the surface energy balance is negative, and the slush layer has not formed. The slush layer forms if the snow weight on ice is large enough to raise water above the ice-snow interface. When the surface energy balance is negative, the slush layer freezes and forms the snow-ice layer (e.g. Leppäranta 1983; Saloranta 2000; Saloranta and Andersen 2007; Kirillin et al. 2012). The calculation of net radiation is divided to shortwave and longwave radiation and turbulent heat fluxes to sensible and latent heat. In wintertime the surface heat balance is typically dominated by longwave radiation and during melting season by shortwave radiation (e.g. Jakkila et al. 2009; Kirillin et al. 2012). The turbulent heat fluxes are significant, especially in windy and cloudy conditions.

Fig. 2
figure 2

An illustration of lake ice and snowdrift formation in winter (a) and ice melting in spring (b)

Full size image

Snowdrift formation adds another dimension to the WSFS-Ice model. The snowdrift forms to the lee side of the islet (Fig. 2), or close to a cliff or a forest. The snow drifting and settling is modelled for a 5 m high artificial islet hill slope using a model developed by Walter et al. (2004). The modelled results are compared with the average in-situ snow line observations. The modelled hill does not represent the real conditions (see Sipilä 1990) in nature, but it provides information about the amount of snow drifting and settling in different years. The snowdrift model is used for studying what kind of weather conditions form larger snowdrifts close to the shore and how much the size of the snowdrift varies between years and in the future due to climate change.

Melting of the ice cover begins from the snow layer on ice, then melts the snow-ice layer and finally the congelation ice. Solar radiation melts the congelation ice from the edges of the ice crystals rotting the congelation ice to candled ice formation. Part of the solar radiation penetrates through the ice warming the water body under ice and causing convective mixing (Fig. 2). The convective currents and warming water increase the heat flux from the water and affect the melting of the ice cover (e.g. Bengtsson 1996; Mironov et al. 2002; Hughes et al. 2024).

2.2.1 Model calibration and validation

The WSFS-Ice model performance was examined in four observation sites located in different parts of Lake Saimaa for calibration 1981–2010 and validation 2011–23 periods. The model was calibrated using ice and snow observations of all sites. The root mean square deviation (RMSD) and bias were separately calculated for ice thickness, snow-ice thickness, snow depth on ice, ice freeze-up and break-up dates. The modelled snowdrift formation was compared with the average snow line observations.

2.3 Climate scenarios

The climate scenarios used in this study are based on the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC 2013). The scenarios are based on three representative concentration pathways (RCPs) of atmospheric greenhouse gas concentrations (van Vuuren et al. 2011). The numbers of the acronyms RCP2.6, RCP4.5 and RCP8.5 represent the radiative forcing (in W/m2) of each greenhouse gas concentration pathways by the end of the twenty-first century.

The climate models used in this study are distributed by the EURO-CORDEX initiative that produce ensemble climate simulations based on regional climate models (RCMs) forced by RCP pathways and global climate models (GCMs) from the Coupled Model Intercomparison Project Phase 5 (CMIP5) (Taylor et al. 2012 and Jacob et al. 2014). The climate models used consist of three RCMs using boundary conditions of four different GCMs. The combinations of the RCMs and GCMs (Table 1) have been selected to represent the average and extreme (warm, cold, wet and dry) climate change impacts and the range of uncertainty related to climate scenarios. All the combinations are available with all the three RCPs (2.6, 4.5 and 8.5).

Table 1 The six combinations of the global climate models and regional climate models used in this study
Full size table

Ruosteenoja and Jylhä (2021) have compared the CMIP5 projections in Finland with the latest climate scenarios of CMIP6, which have been used by the IPCC in the Sixth Assessment Report (IPCC 2021). According to their study, temperature increase in CMIP6 projections is similar in winter but stronger in summer months, and precipitation changes are in good agreement with previous projections in both seasons. However, in this study we decided to use RCMs based on CMIP5, because the difference in winter is small and and RCMs produce more accurate simulation of regional climate compared to GCMs, especially in the lake area in south-eastern Finland (Samuelsson et al. 2010).

The climate scenarios are simulated for three future periods 2010–39, 2040–69 and 2070–99. The control period is 1981–2010. In the different RCMs the range of annual temperature change from 1981–2010 to 2070–99 in the Lake Saimaa region varies between 0.9–2.2 °C in RCP2.6, 1.8–3.9 °C in RCP4.5 and 3.8–5.4 °C in RCP8.5. Annual precipitation change for the same periods ranges between -4.4–6.9% (RCP2.6), 2.0–14.6% (RCP4.5) and 8.4–25.0% (RCP8.5). The proportion of snowfall of the total precipitation decreases in all scenarios. In 1981–2010 the snowfall was 30.2% of total precipitation. In 2070–99 period the snowfall percentages ranges between 21.7–26.3% in RCP2.6, 18.2–23.5% in RCP4.5 and 12.3–16.4% in RCP8.5.

The effect of climate change on input variables of the model is carried out by a delta-change method shown in Fig. 3 (e.g. Arnell 1999; Prudhomme et al. 2003; Veijalainen 2012). The monthly mean changes in temperature and precipitation from control period to future periods are calculated from the climate scenarios. These changes are then added to the daily temperature and precipitation observations of the control period 1981–2010, which are used as inputs of the model. According to climate scenarios, cold temperatures in winter increase more than mild temperatures, and using of mean temperature changes could lead to unrealistically large decrease in snow and ice. Therefore, the changes in temperature distribution are taken into account by using a temperature dependent temperature change, in which temperature changes depend on temperature distribution calculated separately for each month (Andréasson et al. 2004; Veijalainen 2012). The monthly percentual changes of precipitation are calculated using three months moving window.

Fig. 3
figure 3

Description of the delta-change method used for projections of the climate change impacts on lake ice

Full size image

Net longwave and shortwave radiation are affected by the temperature and precipitation changes (i.e. air and surface temperatures and albedo, see Online Resource 1: Eq. 6, 7 and 9) but changes in CO2 concentration also affect the effective atmospheric emissivity. Therefore, the effect of CO2 concentration is taken into account according to the relative changes estimated for each RCP and time period (Fig. 3 and Online Resource 1: Eq. 8). Sensitivity analysis of the effect of wind speed, cloudiness and relative humidity on ice thickness, snow depth on ice and snowdrifts was performed and the results are shown in the Online Resource 2. Possible increase in cloudiness could lead to slightly faster decrease in ice thickness and snow depth, but changes in temperature and precipitation are the dominating factors in determining the changes in lake ice and snow.

3 Results

3.1 Observed changes towards milder winters

Shortening ice duration and thinner ice thickness are clear signs of changing climate. According to the ice observations (Finnish Environment Institute 2023), mean ice duration decreased 14 days and mean ice thickness during the pupping season of the Saimaa ringed seal (1st of February – 20th of March) decreased 7.4 cm in Lauritsala observation site between 1981–2000 and 2001–23 periods. Increasing inter-annual variation has led to more frequent mild winters. The ice freeze-up and break-up dates have been observed in Lauritsala and Haukivesi sites since 1885. In both stations at least five out of ten latest freeze-up dates and six out of ten earliest break-up dates have been observed in the last two decades. Ice thickness has been observed in Lauritsala since 1917. The six thinnest mean ice thicknesses and four out of ten thinnest mean snow depths during the pupping season have been observed in the last two decades.

Mild winters cause higher risk for the increased pre-weaning pup mortality. For instance, in winters 2006–07 and 2019–20 the pup mortality was above 20% (Metsähallitus 2023). The record shortest (41 days) and second shortest (79 days) ice durations were observed in Lauritsala in 2019–20 and in 2006–07, respectively. The record thinnest mean ice thickness and snow depth on ice during the pupping season were observed in 2020 and second thinnest in Haukivesi site in 2007.

3.2 Model performance

The model performance for calibration 1981–2010 and validation 2011–23 periods are summarized in Table 2. The WSFS-Ice model RMSD of ice thickness varies between 6.8–10.9 cm in the calibration period and between 7.8–8.2 in the validation period in different locations (Table 2). The RMSD of snow depth on ice varies between 5.1–8.5 cm in calibration and between 3.8–7.4 cm in validation period in different ice observation sites. In Pihlajavesi observation site all results are not shown due to lack of observations in the validation period.

Table 2 Model root mean square deviation (RMSD) and bias for ice thickness, snow-ice thickness, snow depth, ice freeze-up and break-up days in the calibration (1981–2010) and validation (2011– 23) periods
Full size table

The model underestimates ice thickness on average by 5.8 and 5.3 cm in Pihlajavesi and Pyhäselkä observation sites, respectively (Table 2). The maximum bias of the snow depth on ice is ± 3.1 cm (in Haukivesi). On average the ice freeze-up day is estimated well, but the interannual variation is large and in some years the difference between modelled and observed freeze-up days can be more than two weeks. The ice break-up day is modelled more accurately and RMSD is between 3–8 days in the calibration and between 4–16 days in the validation period.

3.3 Snowdrift formation

Snow drifting and settling was modelled in a 5 m high islet hill slope. The modelled snowdrift formation on the artificial islet was compared with mean values of the 10 snow line observation sites in Lake Haukivesi and Lake Pihlajavesi. In the lee side of the modelled artificial slope the simulated snow accumulates similarly as on average in the snow line observations (Fig. 4).

Fig. 4
figure 4

Modelled and observed average snow depths in snowdrifts close to the shore

Full size image

The snow depth in the snowdrift mainly depends on the accumulated snowfall after the freeze-up of the lake. Wind speed during the snowfall correlates with the amount of drifted snow, increases the snow depth and density of the snowdrifts. The modelled and observed snow depths show that the snowdrifts were remarkably smaller in 2017, 2018 and 2020 (Fig. 4). Years 2017 and 2020 were mild and occasional melting periods decreased the snow depths in the snowdrifts. In 2018 and 2020 the lake froze-up late in the end of December. In 2020 snowdrifts were remarkably smaller than in other years, due to lack of snow.

3.4 Climate change impacts

Climate change has a significant effect on ice phenology, ice thickness and snow depth on the ice cover. The following results are shown only for Lake Haukivesi, where the snowdrift observations were conducted. According to the model results with climate scenarios of the RCP4.5 pathway, mean ice thickness decreases on average by 13 cm and snow depth by 9 cm from 1981–2010 to 2070–99 during the seal pupping season 1st of February – 20th of March. Ice freeze-up dates occur approximately two weeks later and ice break-up dates almost one month earlier. Mean depth of the snowdrifts decreases approximately to half from 1981–2010 to 2070–99. The density of the snowdrifts is projected to decrease due to later formation of the snowdrifts and shorter time for snow compaction (Table 3).

Table 3 Modeled mean ice thickness and snow depth during the pupping season of the Saimaa ringed seal (1st February – 20th March) and mean ice freeze-up, break-up dates, ice duration, first and last snow days and snow duration days in the RCP4.5 scenario
Full size table

Earlier snowmelt leads to earlier break-up of the seal’s lairs. The model results show that snow melts before the end of the pupping season in more than half of the years in the 2040–69 and 2070–99 periods with RCP4.5 and RCP8.5 (Fig. 5 and Fig. 6). During the mildest winters in the 2040–69 period and especially in the 2070–99 period the ice cover is very thin in the beginning of the pupping season in early February and melts before the pupping season has ended. Inter-annual variability continues to increase in the coming decades. Maximum ice thickness and snow depth on ice decrease significantly less than mean and minimum thicknesses. Only with the RCP8.5 pathway maximum ice thickness decreases significantly in the 2070–99 period (Fig. 5).

Fig. 5
figure 5

The simulated daily mean, minimum and maximum ice thickness and snow depth on ice in Lake Haukivesi in the control period 1981–2010 and in the 2010–39, 2040–69 and 2070–99 periods with RCP4.5 (left) and in the 2070–99 period with RCP2.6 and RCP8.5 (right)

Full size image
Fig. 6
figure 6

Daily probability of the presence of ice and snow on ice in Lake Haukivesi in the control period and different future periods with RCP4.5 (left figures) and with different RCPs in the 2070–99 period (right figures)

Full size image

The probability of ice and snow existence (based on the ratio of the number of years with ice- and snow-cover) decreases in the future periods (Fig. 6). Minimum ice thickness in the beginning of the pupping season (1st of February – 20th of March) of the Saimaa ringed seal is 18 cm in the control period 1981–2010 and 11 cm in the 2010–39 period with RCP4.5. Probability of no ice days in the end of the pupping season varies in different RCMs between 0–17% in 2040–69 and 3–20% in the 2070–99 period with RCP4.5. With RCP8.5 probability of no ice days in the end of the pupping season varies between 23–47% in the 2070–99 period, whereas with RCP2.6 the minimum ice thickness during the pupping season is 8 cm.

Already in the 2010–39 period there is an increasing probability of lack of snow on ice during the pupping season of Saimaa ringed seal, especially due to earlier snowmelt in March. The probability of no snow conditions on average in February increases to 7–18% in the 2040–69 period with RCP4.5. In the 2070–99 period the probability of no snow conditions with RCP2.6 is 6–15% and with RCP8.5 23–50% in the different RCMs.

Climate change increases the inter-annual variability of ice thickness and snow depth on ice during the twenty-first century (Fig. 7). Maximum ice thickness and snow depth on ice during the seal pupping season remain almost the same or may even increase according to some climate models (CNRM-A and MPI-SMHI) in the RCP4.5 pathway, but the mean and minimum values decrease significantly.

Fig. 7
figure 7

Median, interquartile range, minimum and maximum ice thickness and snow depth on ice and snow depth in snowdrift during the pupping season of Saimaa ringed seal in Lake Haukivesi in the control period 1981–2010 and with RCP4.5 in different future periods (left) and with different RCPs in the 2070–99 period (right)

Full size image

The future development of greenhouse gas emissions and carbon sinks has a clear impact on the changes in ice and snow conditions. With RCP2.6 the changes projected for the 2070–99 period are even smaller than with RCP4.5 for the 2040–69 period. With RCP8.5 the mean snow depth during the pupping season is less than 10 cm almost every year in the 2070–99 period (Fig. 7).

Median snow depth in the snowdrifts close to the shoreline decreases from 36 to 17 cm by the 2070–99 period (Fig. 7). The main factors affecting the snow depths in the drifts are the later freeze-up of the lake, smaller accumulated snowfall after the freeze-up, larger rain/snowfall ratio and increased number of melting periods during the ice season.

The modelled average snowdrift depth below 20 cm during the pupping season of the ringed seal can be used as an indicator of a risk for higher pre-weaning pup mortality. During the pupping season, the modelled snowdrift depth is below 20 cm in four years in the control period 1981–2010. In the 2010–39 period the minimum depth in snowdrift is below 20 cm in 28% of the years with RCP4.5, in the 2040–69 and 2070–99 periods 50% and 60% of the years, respectively. With RCP2.6 mean snowdrift depth during the pupping season is less than 20 cm in 42% of the years and with RCP8.5 in 86% of the years in the 2070–99 period.

4 Discussion

4.1 Current breeding conditions and importance of the ice cover and snow on ice

Negative effects of mild winters have already been seen in the breeding habitat of the Saimaa ringed seal. More frequently occurring mild winters have shortened ice and snow duration, decreased snow depth on ice and reduced snowdrift formation during the pupping season. The snowdrifts are essentially important for the success of the breeding of the seal. The subnivean lairs provide shelter from harsh weather conditions and predators, having an important role for the pup survival also after birth. Saimaa ringed seals use lairs until they melt (Kunnasranta et al. 2021), and there are observations where red fox (Vulpes vulpes) have killed a seal pup that has been on ice in March (Auttila 2015). Thus, earlier snowmelt increases the threats to the pups in the beginning of their life.

In recent years man-made snowdrifts and nest boxes have been implemented as a conservation measure (Auttila et al. 2014; Kunnasranta et al. 2022). During mild winters when there have not been sufficient snowdrifts for the seal’s lairs or lairs have collapsed too early, the observed pre-weaning pup mortality has been near 30% (Auttila 2015), whereas the pre-weaning mortality with good snow conditions is ca. 10% (Sipilä 2003). The recovery of the seal population has been slow, and the population growth has temporarily stopped due to high pre-weaning pup mortality in years with poor snow conditions, like in 2020 (Kunnasranta et al. 2021; Metsähallitus 2023).

4.2 Impact of climate change on ice and snow

The projected climate change impacts on ice phenology are in good agreement with the previous studies focused on estimating the changes in the ice freeze-up and break-up dates. According to Leppäranta (2023) the freeze-up and break-up dates both show climate sensitivity of one week per 1 °C. Dibike et al. (2011) found 5–20 days delayed freeze-up and 10–30 days advanced break-up dates in the Northern Hemisphere from the baseline 1960–1999 period to the 2040–79 period. Blagrave and Sharma (2023) projected on average 13 days later freeze-up dates and 20 days earlier ice-off dates for lakes in the Midwestern and Northeastern United States by 2070–99. Our results show on average 16 days later freeze-up (range 5–32 days with all scenarios and RCPs) and 27 days earlier break-up dates (range 9–49 days) in the 2070–99 period.

Effects of climate change on maximum ice thickness and snow depth on ice vary in different geographical locations depending on the changes in temperature and snow accumulation. Possible increase in snow accumulation may increase snow-ice formation and total ice thickness, even though increasing temperature decreases the ice growth rate (Korhonen 2006; Dibike et al. 2011; Leppäranta 2023). Korhonen (2006) found positive trends in maximum ice thickness in Northern and Central Finland explained by the increased snow-ice formation, but negative trends in Southern Finland. Dibike et al. (2011) projected increased snow depth and snow-ice thickness in cold climate conditions in North-Eastern Asia but decrease in most parts of North America and Eastern Europe including Finland. In our results ice thickness, snow depth on ice and snowdrifts decrease due to climate change, but the inter-annual variation increases (see Table 3 and Fig. 7).

4.3 Effect of future projections on the seal’s breeding habitat

According to our results, the natural breeding habitat of the Saimaa ringed seal is significantly deteriorating and may even disappear in more frequent milder winters. The formation of the snowdrifts depends mainly on the snow accumulation. The later freeze-up and increased rainfall-snowfall ratio during the ice season decreases the snow accumulation and the depth of the snowdrifts. Due to delayed freeze-up of the lake, there is also less time for the formation of sufficient snowdrifts and their dense layered structure, which is essential for the lairs to last as long as possible without collapsing. Possible, but still uncertain changes in wind speed could have a significant effect on the snowdrifts. The amount of drifting snow depends on the wind speed during the snowfall and wind also compacts and improves the snowdrift structure.

The increasing temperature and rainfall events decrease snow albedo, which is typically the triggering factor of snowmelt onset (e.g. Surdu et al. 2015; Robinson et al. 2021). The earlier melting of snow and ice together pose a significant threat to the long-term survival of the Saimaa ringed seal. The median snowmelt day of all scenarios and RCPs is before the end of the pupping season already in the 2040–69 period, meaning that in half of the years snow melts before all pups have been born.

4.4 Climate change mitigation and adaptation possibilities

Mitigation of climate change is highly important for the preservation of the natural breeding habitat of ringed seals (Kovacs et al. 2012). The severe impacts of climate change have already been seen in the southern ringed seal populations, like in Lake Saimaa, Lake Ladoga and Baltic Sea. Although the Saimaa ringed seal has evolved adaptations to earlier snow and ice melting, like the pups' gray lanugo hair and the ability to also use a terrestrial habitat for molting, resting and nursing, the mild winters have caused increased pre-weaning pup mortality (Hyvärinen et al. 1995; Kunnasranta 2001; Niemi et al. 2013; Auttila 2015).

Winter 2019–20 was the mildest winter ever recorded in terms of ice and snow duration as well as ice thickness and snow depth observations during the pupping season of the Saimaa ringed seal. Even making sufficiently large man-made snowdrifts was impossible, due to the lack of snow. In the 2070–99 period the proportion of as mild or milder winters than 2019–20 varies between 1–13% with RCP2.6, 13–32% with RCP4.5 and 45–78% with RCP8.5 depending on the different RCMs.

The more frequently occurring mild winters pose a significant threat for the Saimaa ringed seal. If the lack of snow makes it impossible to even build man-made snowdrifts, the nest boxes are only possibility of the safe breeding conditions for the seals. In addition, other conservation acts, such as reducing by-catch mortality, securing low disturbance levels close to the seal’s breeding habitats and stable water levels during the ice-covered season, play important role in targeting to rapid increase of the seal population, which enables it to survive in the changing climate.

The reduction of ice and changes in snow cover significantly affect other ice-associated marine mammals as well (Kovacs et al. 2012). While this study focused on breeding habitat of the Saimaa ringed seal, similar modelling approach by applying lake or sea ice models could be used for studying the impacts of climate change on other ringed seal subspecies and other ice-associated species.

5 Conclusions

The emphasis of the development of the WSFS-Ice model was to improve the estimation of climate change impacts on the breeding habitat conditions of the endangered Saimaa ringed seal. The model results show significant decrease in ice and snow duration by the end of the twenty-first century. Mean ice thickness and snow depth on ice and snowdrifts decrease in all climate scenarios. The increasing inter-annual variability in ice thickness and snow depth on ice increases the number of years without sufficient snow cover for the seals lairs already in the near future. Especially the remarkable decrease in size of the shoreline snowdrifts, earlier snowmelt and ice break-up will significantly increase the pre-weaning mortality of seal pups endangering the population’s recovery and long-term existence.

The success of the climate change mitigation efforts plays an important role in the changes of the natural breeding habitat conditions of the Saimaa ringed seal. In the RCP2.6 pathway, which requires a rapid decline of the greenhouse gas emissions, the deterioration of the breeding habitat settles down in the middle of the twenty-first century. In the RCP4.5 pathway the years with lack of snow become more common already in the middle of the twenty-first century and in some years ice cover melts even before the end of the pupping season. In the worst-case scenario – the emissions continue to increase towards the end of the twenty-first century – the years without snow on ice and even years without ice increase and would require extensive conservation actions for maintaining the current ringed seal population in Lake Saimaa.

Our study clearly shows that the ongoing climate change poses a considerable threat to the Saimaa ringed seal. Therefore, conservation acts should be targeted at the most acute threats and a rapid population growth should be ensured because the larger population size would enable better prospects for its survival in a rapidly changing climate. This should be done by reducing the by-catch mortality to the minimum, securing low disturbance levels at the seals breeding habitat and a stable water level during the ice-covered season. In addition, it is important to continue making man-made snowdrifts and adopt artificial nest boxes as an active conservation act to mitigate the negative effects of climate change to the seals breeding success.