1 Introduction

Model-based projections of climate change for the marine parts of the Baltic Sea region are sparse compared to studies for the atmosphere. The first assessment of climate change in the Baltic Sea area (BACC Author Team 2008) concluded that ‘the mean annual sea surface temperatures could increase by some 2–4 °C by the end of the twenty-first century. Ice extent in the sea would then decrease by some 50–80 %. The average salinity of the Baltic Sea is projected to decrease between 8 and 50 %. However, it should be noted that these oceanographic findings are based upon only four regional scenario simulations using two emissions scenarios and two global models’. At that time, only one study addressing uncertainties was available and this was based on a larger ensemble of 16 simulations to project Baltic Sea salinity by the end of this century (Meier et al. 2006c).

The reason why only a few Baltic Sea scenario simulations have been performed compared to scenario simulations for the atmosphere might be that the former are computationally demanding due to the high horizontal and vertical grid resolutions required. Furthermore, from the ocean perspective, the dynamical downscaling technique used to investigate changes in climate at regional scales requires coupled atmosphere–ice–ocean models (Döscher et al. 2002; Räisänen et al. 2004) instead of uncoupled regional climate model (RCM) simulations (Kjellström et al. 2011; Nikulin et al. 2011) because sea-surface temperature (SST) and sea-ice concentration fields from general circulation models (GCMs) used as surface boundary conditions in RCMs might be biased due to the coarse resolution of GCMs. Such biases would affect model sensitivity and hence the results in climate projections (Meier et al. 2011d). To reduce computational demands, scenario simulations were performed for selected time slices only (e.g. Madsen 2009) which requires the application of the ‘delta’ or ‘delta change ’ approach (e.g. Meier 2002).

Since the first assessment of climate change in the Baltic Sea basin (BACC Author Team 2008), the number of relevant scenario simulations has increased considerably. In particular, a large number of scenario simulations were carried out during 2009–2011 within the ECOSUPPORT project (advanced modelling tool for scenarios of the Baltic Sea ECOsystem to SUPPORT decision making, The following items characterise the new simulations compared to the first BACC assessment:

  • The horizontal resolution of atmosphere and ocean model components increased to typically less than 25 and 3.6 km, respectively (e.g. Meier et al. 2011b).

  • New model versions of GCMs and RCMs were used.

  • The results and assumptions of the Fourth Assessment Report (AR4) of the Intergovernmental Panel on Climate Change (IPCC 2007) were used instead of those from the IPCC’s Third Assessment Report (IPCC 2001).

  • Multi-model ensemble modelling was introduced to estimate uncertainties due to biases in the Baltic Sea models (e.g. Meier et al. 2011b).

  • Instead of time slices often combined with the delta approach (e.g. Meier 2006), transient simulations (1960–2100) were performed (e.g. Neumann 2010).

  • Coupled physical–biogeochemical models were used (Neumann 2010; Meier et al. 2011a, b, c, 2012a, b, c; Eilola et al. 2012; Neumann et al. 2012).

This chapter relies on literature concerning climate change scenario simulations for the Baltic Sea between 2007 and (March) 2012. The focus is on projected changes in water temperature, salinity , sea ice, storm surges , and wind waves . Changes in marine biogeochemical variables such as oxygen , nutrients , and phytoplankton (chlorophyll ) concentrations are addressed in Chaps. 18 and 19. The graphics included in this chapter (with the exception of Fig. 13.6) show ensemble mean changes projected for 2069–2098 relative to a baseline/reference of 1978–2007 in transient scenario simulations calculated with three coupled physical–biogeochemical models for the Baltic Sea following Meier et al. (2011b, 2012c). The Baltic Sea models are forced with RCM data driven by two GCMs at the lateral boundaries and two greenhouse gas (GHG) emission scenarios (SRES A1B and A2, see Nakićenović et al. 2000). The atmospheric forcing fields of the RCM were analysed by Meier et al. (2011d). Results at six monitoring stations representing different sub-basins of the Baltic Sea are reported in this chapter. The locations of the monitoring stations are depicted in Fig. 13.1.

Fig. 13.1
figure 1

Bottom topography in the Baltic Sea and selected Swedish and Finnish monitoring stations at Anholt East (AE), Bornholm Deep (BY5), Gotland Deep (BY15), LL07 in the Gulf of Finland, SR5 in the Bothnian Sea and F9 in the Bothnian Bay

2 Water Temperature

Projected change in seasonal and annual mean ensemble average SST is shown in Fig. 13.2. The projected change is greatest in the Bothnian Bay and Bothnian Sea during summer and in the Gulf of Finland during spring. According to the mini-ensemble, which is forced by the SRES A1B and A2 scenarios, summer SST could increase by about 2–4 °C in the southern and northern Baltic Sea, respectively. The north–south gradient in SST change is caused at least partly by the ice-albedo feedback, which increases SST sensitivity in the northern Baltic Sea (Meier et al. 2011d).

Fig. 13.2
figure 2

Projected change in seasonal (a DJF, b MAM, c JJA, d SON) and e annual mean ensemble average SSTs for 2069–2098 relative to a baseline of 1978–2007. See Meier et al. (2012a)

The projected change in annual mean temperature at the six monitoring stations (Fig. 13.1) is 2–3 °C (surface water layers) and 0–2 °C (deep-water layers) (Fig. 13.3). At Anholt East in the Kattegat , temperatures in the deep-water remain unchanged due to the assumption in the scenario simulations that the vertical temperature and salinity profiles at the lateral boundaries in the Kattegat or Skagerrak (depending on the ocean model used) do not change over time. Sensitivity experiments showed that artificially fixing the boundary data does not affect the Baltic Sea interior (Meier 2002). This is because in the model even during major Baltic inflow events the inflowing water volume is smaller than that of the Kattegat surface layer. Because at the time scale of inflows, no water from outside the model domain is advected through the Danish straits into the Baltic Sea and because at this time scale, the heat budget of the Kattegat surface layer is controlled by vertical heat fluxes, water masses that flow into the Baltic Sea are usually realistically simulated.

Fig. 13.3
figure 3

Ensemble average vertical profiles and projected change in temperature for 2069–2098 relative to a baseline of 1978–2007 at the monitoring stations at a Anholt East (AE), b Bornholm Deep (BY5) and c Gotland Deep (BY15), and in the d Gulf of Finland (LL07), e Bothnian Sea (SR5) and f Bothnian Bay (F9) (for locations, see Fig. 13.1): observations (green), baseline 1978–2007 (black). The range in variability is indicated by the ± one standard deviation band around the ensemble average of model results (dotted lines) or observations (grey shaded area), see Meier et al. (2012b)

In all sub-basins, the surface layer is projected to warm more than the deep-water. Although this increasing temperature differential would cause an increase in vertical stratification, the projected salinity change is more significant in terms of water density than the projected temperature change. Hence, scenario simulations suggest that the vertical stratification between surface and deep-water layers would actually decrease (see Sect. 13.3).

The magnitude and spatial patterns of change in water temperature are relatively similar when different Baltic Sea models use the same atmospheric forcing (Meier et al. 2011a, b, 2012a, c). Even with atmospheric forcing from different RCMs or different GHG emission scenarios (SRES A2, B2, A1B, B1), temperature changes are similar (Meier 2006; Neumann 2010; Neumann and Friedland 2011; Gräwe et al. 2013) although temperature changes in the B1 scenario appear smaller than those in the other scenarios and a stabilising tendency at the end of the scenario simulation is observed (Neumann 2010; Neumann and Friedland 2011).

Most of the scenario simulations for the Baltic Sea were performed with regionally limited ocean models with lateral boundaries in the Kattegat or Skagerrak. Due to the proximity of the lateral boundary, results for the Kattegat are not reliable. Instead, projections for the North Sea should be employed to study changes in the Kattegat and Skagerrak (Ådlandsvik 2008; Holt et al. 2010). Only Madsen (2009) investigated both shallow seas—the Baltic Sea and the North Sea—simultaneously and found that warming is greater in the Baltic Sea than that in the North Sea.

Holt et al. (2010) found in the A1B scenario that the shelf sea regions warm substantially more than the open ocean, by 1.5–4 °C depending on location. These results agree with observed warming trends in the Baltic Sea, North Sea and other shelf seas (Belkin 2009).

3 Salinity

Results of the multi-model ensemble simulations by Meier et al. (2011b, 2012c) indicate that projected changes in sea-surface salinity are small in the northern and eastern Baltic Sea (smallest in the Bothnian Bay) and greatest in the Danish straits region, especially in the Belt Sea (Fig. 13.4). The latter are due to a shift in the fronts within the transition zone. Seasonal change in sea-surface salinity is projected to be minimal.

Fig. 13.4
figure 4

Projected change in seasonal (a DJF, b MAM, c JJA, d SON) and e annual mean ensemble average sea-surface salinity for 2069–2098 relative to a baseline of 1978–2007. See Meier et al. (2012a)

The projected reductions in salinity in the Bornholm Basin and Gotland Basin are almost constant with depth and amount to 1.5–2 g kg−1 in the ensemble mean (Fig. 13.5). However, change in the deep-water is greater than that in the surface layer in these sub-basins. In more weakly stratified regions such as the Gulf of Finland or Bothnian Bay, the difference in the projected change for salinity in surface and bottom layers is even greater causing a reduction in vertical stability. These results are consistent across the three Baltic Sea models. The projections of halocline depth are most uncertain in the Gotland Basin , as indicated by the standard deviation among the ensemble members (Fig. 13.5).

Fig. 13.5
figure 5

Ensemble average vertical profiles and projected change in salinity for 2069–2098 relative to a baseline of 1978–2007 at the monitoring stations at a Anholt East (AE), b Bornholm Deep (BY5) and c Gotland Deep (BY15), and in the d Gulf of Finland (LL07), e Bothnian Sea (SR5) and e Bothnian Bay (F9) (for locations, see Fig. 13.1): observations (green), baseline 1978–2007 (black). The range in variability is indicated by the ± one standard deviation band around the ensemble average of model results (dotted lines) or observations (grey shaded area), see Meier et al. (2012b)

In the ensemble presented, the salinity changes projected are driven by changes in run-off which is projected to increase by 15–22 % (Meier et al. 2012b). The latter figures are estimated from the difference between precipitation and evaporation over land calculated from the RCM output directly (Meier et al. 2012b). If a hydrological model is used to calculate run-off from the same atmospheric forcing, run-off changes are smaller and increase by 4–13 % (Arheimer et al. 2012). These discrepancies illustrate the uncertainty in hydrological modelling.

In the ensemble presented by Meier et al. (2012b), changes in wind speed play only a minor role in the salinity changes in contrast to earlier findings by Meier (2006). In the latter scenario simulations driven by ECHAM4/OPYC3, monthly mean increases in wind speed of almost 30 % were found during February (Räisänen et al. 2004). Hence, salinity projections remain uncertain in accordance with earlier results by Meier et al. (2006c) owing to the uncertainty in wind speed projections over the Baltic Sea region (Kjellström et al. 2011; Nikulin et al. 2011). However, there is generally a tendency for both increased mean (Kjellström et al. 2011) and extreme (Nikulin et al. 2011) wind speeds over the sea.

Although all studies based on dynamical modelling suggest that in a future climate, Baltic Sea salinity will decrease or remain unchanged compared to the present-day climate (Meier 2006; Meier et al. 2006c, 2011b, 2012b; Neumann 2010), Hansson et al. (2011) claimed that run-off from the total Baltic Sea catchment would decrease if air temperature rises. Hansson et al. (2011) reconstructed river run-off to the Baltic Sea for the period 1500–1995 using a statistical approach with air temperature and atmospheric circulation indices as predictors. They found that over the past 500 years, the total river run-off to the Baltic Sea showed no significant long-term trend but decreased slightly in response to the observed rise in temperature, at a rate of 3 % per 1 °C rise (see also Chap. 5, Fig. 5.3). However, their approach considerably underestimates interannual variability. In addition, changes in future climate that by the end of the century are projected to be outside the range of decadal variability are very likely not comparable with changes in past climate because past anthropogenic warming is small and other drivers may have controlled the air temperature –run-off relationship.

Studies of both past and future climates suggest that increased total run-off would increase the ventilation of the upper halocline due to weakened stratification causing improved oxygen conditions in the upper deep-water (e.g. Gustafsson and Omstedt 2009; Meier et al. 2011b). Despite changing halocline depth, stratification changes in the Baltic proper due to increased freshwater supply are expected to be minor (Meier 2005). Increased freshwater supply would drive an increased recirculation of brackish surface waters and consequently reduced saltwater fluxes into the Baltic Sea. For the hydrography of the north-western European continental shelf, Holt et al. (2010) found that the strength of seasonal stratification may increase by about 20 % on the shelf, compared with 20–50 % in the open ocean. The former being controlled by temperature and the latter by salinity.

Hordoir and Meier (2011) found increased spring and summer stratification under a warmer climate, with the changes greater in the northern Baltic Sea than the southern. The north–south gradient might be partly explained by the salinity-dependent temperature of maximum density. In the present climate, winter temperatures in the Baltic Sea are often below the temperature of maximum density such that warming during spring causes thermal convection (see also Chap. 7, Sect. 7.1). In a future climate, temperatures are expected to be typically higher than the temperature of maximum density and thermally induced stratification may start without prior thermal convection (Hordoir and Meier 2011). In particular, in the northern Baltic Sea, these stratification changes might affect vertical nutrient fluxes and thus the intensity of the spring bloom in future climate.

Gräwe and Burchard (2011, 2012) and Gräwe et al. (2013) studied local changes in the western Baltic Sea with a high-resolution model. They found no significant trend in potential energy measuring the competition between stratification and mixing . They also found no clear tendencies in the projected change in saltwater transport for either medium or major inflow events.

4 Sea Ice

Whether there is a reduction in sea-ice cover in the future depends mainly on the projected change in air temperature over the Baltic Sea in winter; other drivers such as wind are considered less important (e.g. Tinz 1996; Meier et al. 2004a, 2011d; Meier 2006; Jylhä et al. 2008; Neumann 2010). Hence, the projected changes in sea-ice cover depend on the GHG emission scenario, the GCM and the Baltic Sea model used. Both dynamical modelling and statistical modelling suggest that the relationship between annual maximum ice extent and winter mean air temperature changes is nonlinear (Meier et al. 2004a; Jylhä et al. 2008), that is, even under a warmer future climate, sea ice is likely in the northern Baltic Sea (Fig. 13.6).

Fig. 13.6
figure 6

Sea-ice extent as function of time for 1961–2007 and 1961–2100 in hindcast and scenario simulations, respectively (left panels): observations (red), model results (black). The mean seasonal cycles for 1980–2007 are also shown (right panels). The three rows show results from RCAO-ERA40, RCAO-ECHAM5-r3-A1B and RCAO-HadCM3-ref-A1B using a horizontal resolution of 50 km for the atmosphere model (Meier et al. 2011d)

For instance, Jylhä et al. (2008) examined the Baltic Sea ice cover using a statistical model that related annual maximum ice extent to wintertime coastal temperatures. They found that all model simulations, irrespective of the forcing scenario (SRES A2, B2 scenarios) and driving RCM (seven RCMs of the PRUDENCE project), produce considerably milder sea-ice conditions. However, they used only one driving GCM and concluded that a larger number of GCMs as drivers of the RCMs is likely to have resulted in wider ranges in sea-ice estimates than those in their study (see Fig. 13.6). This could explain why Jylhä et al. (2008) found more frequent unprecedentedly mild years for 2070–2100 than in the scenarios by Meier et al. (2004a). Nevertheless, all new scenario simulations indicate a strong decrease in sea-ice extent in agreement with earlier studies summarised in the first BACC assessment (BACC Author Team 2008).

Although new sea-ice models with explicitly resolved sea-ice categories for ridged and rafted ice (Haapala et al. 2005) have been applied in scenario simulations for the Baltic Sea, results for ice categories are not yet published.

5 Storm Surges

Past sea-level variability, trends and possible drivers are addressed in Chap. 9. Future changes in the mean sea level of the Baltic Sea caused by large-scale drivers such as steric expansion , geoid changes, melting of mountain glaciers and ice caps and changes in the mass balance of the Greenland and Antarctic ice sheets are discussed in Chap. 14. This chapter focuses on changes projected in regional sea level caused by changes in the regional wind field. A few scenario simulations for the Baltic Sea that consider other regional drivers such as changes in sea-ice cover and regional thermosteric and halosteric changes are also available (e.g. Madsen 2009; Hünicke 2010; Gräwe and Burchard 2012).

Tidal amplitude in the Baltic Sea is small. However, tidal amplitudes in the Skagerrak, particularly for the semi-diurnal constituents, are significant. On the European Shelf, future sea-level rise is projected to cause non-negligible increases and decreases in the amplitude of the principal lunar, semi-diurnal tidal constituent (M2) depending on the assumed large-scale sea-level rise (Pickering et al. 2012; Müller et al. 2013).

For the Baltic Sea, results of the transient (1960–2100), multi-model ensemble simulations by Meier et al. (2011b, 2012c) indicate that at the end of the century, changes in mean sea-surface height are projected to be greatest during spring and up to 20 cm in coastal areas of Bothnian Bay (Fig. 13.7). The annual mean sea-surface height is projected to reach a maximum of 10 cm. The large spring signal is caused by one of the three Baltic Sea models and is related to the earlier melt of sea ice in a future climate. In model results by Meier et al. (2012a), only mean sea-level changes caused by regional wind changes and in one model even by steric effects are considered and neither the large-scale sea-level rise nor land uplift is included and has to be added to compile plausible future sea-level scenarios (Meier et al. 2004b). The changes projected in sea-level extremes (or storm surges) from the scenario simulations by Meier et al. (2011b) have not yet been analysed.

Fig. 13.7
figure 7

Projected change in seasonal (a DJF, b MAM, c JJA, d SON) and e annual mean ensemble average sea-surface height for 2069–2098 relative to a baseline of 1978–2007, see Meier et al. (2012a)

An evaluation of the earlier scenario simulations by Meier et al. (2004b) or Meier (2006) indicates that during the control period, both mean seasonal cycles and 100-year surge events in the Baltic Sea are simulated well relative to observations. However, extreme sea-level events in the western Baltic Sea and Danish straits were considerably underestimated, probably because the regional topography was not sufficiently resolved using a Baltic Sea model with a relatively coarse horizontal resolution of about 11 km only (Meier 2006). To address this shortcoming, Gräwe and Burchard (2012) used a high-resolution local model for the western Baltic Sea (with a spatial resolution of about 1 km), nested into a RCM in a dynamical downscaling approach, which also took into account baroclinic effects. They found that the quality of surge heights in their model was significantly better than in the driving model. Taking available global sea-level rise scenarios and simulated regional wind speed changes, Gräwe and Burchard (2012) found that sea-level rise has greater potential to increase surge levels in the Baltic Sea than does increased wind speed. Using the SRES A1B scenario and assuming a sea-level rise of 50 cm and a change in mean wind speed of approximately 4 %, resulted in projected surges with a 100-year return period for 2071–2100 that were greatest at the stations Lübeck, Koserow and Gedser in the western Baltic Sea and up to 2.7 m compared to simulated surge heights of less than 2.1 m for 1961–2000. However, the relative impact of changing wind speed on sea-level extremes might be greater for stations in the eastern Baltic Sea, for example in St Petersburg, depending on the driving GCM used (Meier 2006). Furthermore, Gräwe and Burchard (2012) found that changes in storm surge height in the scenarios can be consistently explained by the increase in mean sea level and variation in wind speed, supporting earlier approaches by Meier et al. (2004b) and Meier (2006).

In addition to the impact of changing winds on sea levels, Madsen (2009) analysed steric effects using a coupled North Sea and Baltic Sea model. The author found maximum halosteric changes in annual mean sea level of about 6.5 cm in the Baltic Sea. Hence, locally generated steric effects will be small compared to those of the NE Atlantic Ocean, basically because of the relatively small water depths. Changes in sea level from dynamical adjustment of global and regional changes might be greater but have not been studied in detail.

Hünicke and Zorita (2008) found an increase in sea-level amplitude (i.e. the difference between mean winter and spring sea level) and suggested that this could be due to the long-term trend in seasonal precipitation. The results fit well with the projections by Hünicke (2010) who applied a statistical downscaling approach to the output of five GCMs. The author found that using both sea-level pressure and precipitation as the predictor resulted in significant future change in the sea level of Baltic Sea. Although Hünicke’s estimates are lower, they are of the same order of magnitude as the projected rise in global sea level.

The implications of sea-level rise for various coasts have been analysed by several authors. For instance, impacts on the Estonian, Polish and Danish coasts were discussed by Kont et al. (2008), Pruszak and Zawadzka (2005, 2009) and Fenger et al. (2008), respectively. For further details, see Chap. 20.

6 Wind Waves

Since the first BACC assessment (BACC Author Team 2008), when scenario simulations with simplified wave models only were available (Meier et al. 2006a, b), advanced wave models have been applied for the Baltic Sea (e.g. Weisse and Günther 2007) and projections of the future wave climate have been made both at the Swedish Meteorological and Hydrological Institute (Kriezi and Broman 2008) and at the Helmholtz-Zentrum Geesthacht (Groll and Hünicke 2011).

In a recent study by Dreier et al. (2011), a statistical model based on wind–wave correlations and wind data derived from RCM simulations forced by ECHAM5/MPI-OM and two GHG emission scenarios (SRES A1B, B1) was used to perform scenario simulations of the future wave climate along the German Baltic Sea coast. The results indicate only a small increase in overall wave energy input in a future climate. However, the frequency of wave directions from the west could increase by 3.5 % compared to present-day conditions.

7 Conclusion

Recent studies confirm the findings of the first assessment of climate change in the Baltic Sea basin (BACC Author Team 2008), namely that under all GHG emission scenarios used (covering the range between SRES B1 and A2), water temperature is projected to increase significantly and sea-ice cover to decrease significantly. Warming would enhance the stability across the seasonal thermocline .

Although one study claimed an increase in salinity in the Baltic Sea under a warmer climate, most studies project decreased salinity and reduced stability across the permanent halocline due to higher, spatially integrated run-off from land. No clear tendencies in saltwater transport were found. However, the uncertainty in salinity projections is likely to be large due to biases in atmospheric and hydrological models.

Although wind speed is projected to increase over sea, especially over areas with diminishing ice cover, no significant trend was found in potential energy (measuring the competition between stratification and mixing). In accordance with earlier results, it was found that sea-level rise has greater potential to increase surge levels in the Baltic Sea than does increased wind speed.

In contrast to the first BACC assessment (BACC Author Team 2008), the findings reported in this chapter are based on multi-model ensemble scenario simulations using several GHG emissions scenarios and Baltic Sea models. However, it is very likely that estimates of uncertainty caused by biases in GCMs are still underestimated in most studies.