2.2.1 Past climate

For the Holocene climate evolution, palaeo-pollen data with a decadal resolution, reconstructing seasonal temperature and precipitation changes compared to preindustrial climate (Mauri et al., 2015), were analysed (Fig. 3). More accurate tree ring data, resolving annual summer mean temperatures, are available for the past millennium (Luterbacher et al., 2016) and have been discussed here (Fig. 4). For further details, the reader is referred to Sect. 3.1.

Figure 3Orbital forcing (irradiance) at 60^∘ N in January and July and the anomalies of reconstructed seasonal temperature and precipitation compared to preindustrial climate in the Baltic Sea region at the Mid-Holocene Optimum (6000 BP). Data source: irradiance data by Laskar et al. (2004) and temperature and precipitation data by Mauri et al. (2015).

Figure 4Reconstructed spring–early summer air temperature deviations in the Baltic Sea region (land areas in the box 0–40^∘ E × 55–70^∘ N; deviations from the 20th century mean). The record is smoothed by a 30-year low-pass filter. The approximate uncertainty range has been estimated here from the data provided by the original publication at interannual and grid cell scale. Data source: Luterbacher et al. (2016).

2.2.2 Present climate

Historical station data of sea level pressure and SST were used to calculate climate indices such as the NAO (sea level pressure differences; Fig. 5) and the AMO (SST anomalies; Fig. 6), describing decadal to multidecadal variability in the large-scale atmospheric circulation. Furthermore, selected records of variables such as air temperature (Fig. 8), river runoff (Fig. 10), land nutrient inputs (Fig. 11; Table 10), glacier masses (Fig. 12; Table 11), maximum sea ice extent (Fig. 14), ice thickness data (Figs. 15 and 16), length of the ice season (Fig. 17), sea level (Fig. 24), and gridded datasets of air temperature, e.g. the land-based CRUTEM4 data (Jones et al., 2012; Fig. 7; Table 9), and of precipitation, e.g. Copernicus data (Fig. 9), were analysed.

Figure 5Normalized winter (December through March – DJFM) mean NAO index during 1821/1822–2018/2019. Red – positive; blue – negative; black – 10-year running mean. The normalisation is the (data mean (data))/standard deviation (data). Data source: https://crudata.uea.ac.uk/cru/data/nao/nao.dat (last access: 17 February 2022).

Figure 6Normalized annual mean AMO index during 1882–2018. Red – positive; blue – negative; black – 10-year running mean. The normalisation is the (data mean (data))/standard deviation (data). Data source: https://climexp.knmi.nl/data/iamo_hadsst_ts.dat (last access: 17 February 2022).

Figure 7Annual and seasonal mean near-surface air temperature anomalies for the Baltic Sea basin for 1871–2020. The baseline period is 1961–1990. Blue and red show the Baltic Sea basin region north and south, respectively, of 60^∘ N. Dots are the individual years, and smoothed curves show the variability on timescales longer than 10 years. Data source: CRUTEM4v dataset (Jones et al., 2012).

For the Baltic Sea, intensive environmental monitoring started more than 100 years ago. In 1898, an agreement between various Baltic Sea countries on simultaneous investigations on a regular basis at a few selected deep stations was signed, and in 1902 the International Council for the Exploration of the Sea (ICES) started its work. Examples from the national monitoring programmes for water temperature (Figs. 18–20) and salinity (Figs. 21 and 22) are shown, illustrating climate variability and climate change of the Baltic Sea.

In addition, some institutes, such as the Swedish Meteorological and Hydrological Institute (SMHI), provide environmental/climate indices, e.g. averaged sea level station data corrected for land uplift (Fig. 23) and hypoxic and anoxic areas (Fig. 25).

Since 1979, satellite data have become available, complementing traditional Earth observing systems and having the advantage of a spatially high resolution (e.g. Karlsson and Devasthale, 2018).

Atmospheric reanalysis products, i.e. the combination of model data and observations (e.g. NCEP/NCAR, ERA40, ERA-Interim, ERA5, and UERRA), were important for calculating the water and energy budgets of the Baltic Sea region (BACC Author Team, 2008; BACC II Author Team, 2015). More recently, ocean reanalysis products have also been developed (e.g. Liu et al., 2017, 2019; Axell et al., 2019) and were, for instance, used for the evaluation of models (e.g. Placke et al., 2018).

Furthermore, various gridded datasets for North Sea SSTs exist and were compared (Fig. 34).

All datasets presented here are publicly online available. For further details on various datasets, the reader is referred to Rutgersson et al. (2022).

Table 9Linear surface air temperature trends (in Kelvin per decade) for the period 1878–2020 over the northern (>60^∘ N) and southern (<60^∘ N) Baltic Sea drainage basin (1878–2020 is selected for comparison with Rutgersson et al., 2014, with an equally long time period). Trends are significant at p<0.05. Data are from the updated CRUTEM4v dataset (Jones et al., 2012). Data source: CRUTEM4v dataset (Jones et al., 2012).

Download Print Version | Download XLSX

Table 10Average (2013–2017) riverine and coastal nutrient inputs (10³ t N (P) yr⁻¹) to the major basins of the Baltic Sea. For abbreviated basin names, see Fig. 11. TN and TP are total nitrogen and total phosphorus inputs, respectively. Data source: updated from Savchuk (2018); see https://www.researchgate.net/publication/358726567_Long-term_times-series_of_the_annual_N_and_P_pools_in_the_major_basins_of_the_Baltic_Sea_1970-2020 (last access: 6 March 2022).

Download Print Version | Download XLSX

Figure 8Homogenised annual and seasonal mean temperature in Stockholm during 1756–2018. Each coloured bar shows the annual mean temperature, in red or blue, depending on whether the temperature is above or below the average during the reference period 1961–1990. The black curve represents smoothed 10-year mean temperatures. Source: https://bolin.su.se/data/stockholm-historical-temps-monthly (last access: 17 February 2022); Anders Moberg, Stockholm University, Sweden.

Figure 9Mean annual precipitation over land in millimetres per year (hereafter mm yr⁻¹) in the Baltic Sea catchment area during 1950–2018. Blue is the whole catchment area, green is the area north of 59^∘ N, red is the area south of 59^∘ N, black is the 10-year running mean, and the bars are the annual sum. Data source: http://surfobs.climate.copernicus.eu/dataaccess/access_eobs.php#datafiles (last access: 17 February 2022). Trends are as follows: 1.44 mm yr⁻¹ (total), 1.51 mm yr⁻¹ (north), 1.37 mm yr⁻¹ (south), and significant at 99 %, using the phase-scrambling method (Kniebusch et al., 2019b).

Figure 10Area-weighted river runoff anomalies relative to 1960–1990 (in percent) from Sweden to the Baltic Sea. The black solid curve denotes Gaussian filtered data with a standard deviation of 3 years. Data source: https://www.smhi.se (last access: 5 March 2022).

2.2.3 Future climate

For the BEARs, regionalisation with the help of dynamical downscaling of general circulation or global climate models (GCMs) or ESMs from CMIP3 and CMIP5 analysed by IPCC (2014b, 2019b) are assessed. The scenario simulations of CMIP5 are driven by greenhouse gas concentration scenarios and the Representative Concentration Pathways, RCP2.6, 4.5, and 8.5 (see Sect. 1.5).

Uncoupled atmospheric regional climate simulations for the 21st century from the EURO-CORDEX framework, calculated with several RCMs and global ESMs, were analysed by Christensen et al. (2022), and the conclusions are summarised here. The choice of working with regional climate model projections, downscaling a limited subset of all available CMIP5 GCMs, implies that the resulting ensemble may not represent all available data properly. Previous studies of parts of the 72-member EURO-CORDEX RCP8.5 ensemble (a sparsely filled GCM–RCM matrix with 11 RCMs downscaling 12 GCM projections in total) assessed here, and presented in more detail by Christensen et al. (2022), illustrated this hypothesis.

By investigating 18 of these RCM simulations (eight RCMs downscaling nine GCMs), Kjellström et al. (2018) found that the nine-member GCM ensemble showed a lower temperature response for northern and eastern Europe compared to the entire CMIP5 ensemble. In addition, it was found that the RCMs can – to some degree – alter the results of the driving GCMs (as also discussed by Sørland et al., 2018). In a more recent study, Coppola et al. (2021) investigated a 55-member ensemble with the same 11 RCMs downscaling the same 12 GCMs as assessed by Christensen et al. (2022). They compared the 55-member ensemble to the driving 12 GCMs and concluded that the RCMs modify the results. In their analysis, Coppola et al. (2021) also considered a set of 12 CMIP6 GCMs, finding that these show a stronger warming signal than the 12 CMIP5 GCMs. This was related to the higher equilibrium climate sensitivity in several global models of the new generation.

Furthermore, coupled atmosphere–sea ice–ocean simulations for the Baltic Sea and North Sea regions with one so-called regional climate system model (RCSM, Dieterich et al., 2013, 2019; Bülow et al., 2014; Wang et al., 2015; Gröger et al., 2015, 2019, 2021b) driven by eight ESMs and three greenhouse gas concentration scenarios, i.e. RCP2.6, 4.5, and 8.5, were compared with atmosphere-only RCM simulations by Christensen et al. (2022). In this study, we present figures of these internally consistent results from the coupled atmosphere–sea ice–ocean scenario simulations, e.g. for air temperature and precipitation (Fig. 26; Tables 12 and 13) and for SST (Fig. 30; Table 14). The state-of-the-art coupled modelling is discussed by Gröger et al. (2021a). For further details about the comparison between coupled and uncoupled scenario simulations, the reader is referred to Christensen et al. (2022).

Table 11Mass balances for the Swedish glaciers Storglaciären, Rabots glaciär, Mårmaglaciären, and Riukojietna. General references are given as footnotes in connection with balance years and long-term monitoring intervals, respectively. Selected specific references are given as footnotes in connection with the glacier names and also include the neighbouring glaciers, Kårsa and Kebnepakte. Data source: SITES data portal at https://data.fieldsites.se/portal/ (last access: 10 February 2022); World Glacier Monitoring Service, 2021; Swedish Infrastructure for Ecosystem Science, 2021a, b, c).

^a World Glacier Monitoring Service (2020). ^b World Glacier Monitoring Service (2021). ^c Blunden and Arndt (2015). ^d Hartfield et al. (2018). ^e Mercer (2016). ^f Holmlund and Holmlund (2019). ^g Kirchner et al. (2019). ^h Brugger and Pankratz (2015). ⁱ Williams et al. (2016).

Download Print Version | Download XLSX

Table 12Air temperature (T_2 m) changes (degrees Celsius) between 1976–2005 and 2069–2098, averaged over each season, and the annual mean over the Baltic Sea catchment area and over the Baltic Sea calculated from nine dynamically downscaled ESM simulations. In addition to the ensemble mean change, the 5th and 95th percentiles indicating the ensemble spread are listed (in parentheses). Data source: Gröger et al., 2021b.

Download Print Version | Download XLSX

What are novel compared to the assessment by the BACC II Author Team (2015) are the high-resolution projections of glacier masses including Scandinavian glaciers (Hock et al., 2019). These results are reproduced in Fig. 28.

Oceanographic regional climate model projections for the Baltic Sea until 2100, driven by the atmospheric surface fields of the above-mentioned RCSM by Dieterich et al. (2019), have been developed and analysed by Saraiva et al. (2019a, b) and Meier et al. (2021a, 2022). In Meier et al. (2021a, 2022), global sea level rise was also considered, which is a driver of the Baltic Sea climate variability that was previously neglected (see Hordoir et al., 2015; Arneborg, 2016; Meier et al., 2017). Here, we compare the latest scenario simulation results by Saraiva et al. (2019b) with previous projections by Meier et al. (2021b, 2022) for, e.g. SST, sea surface and bottom salinities, sea level (Fig. 31), bottom oxygen concentration (Fig. 32), and Secchi depth (Fig. 33),

For further details about the latest oceanographic regional climate model projections for the Baltic Sea, the reader is referred to Meier et al. (2022).

3.1.1 Key messages from previous assessments

Climate variations may be triggered by changes in drivers external to the climate system or may be due to internal processes that reflect the nonlinear, chaotic interactions between the different components of the climate system. The analysis of past climate variations is, therefore, useful for two purposes. One is to estimate the response of climate to changes in the external forcing. The second is to better understand the mechanisms of internal climate variations. Since future climate change will include a mixture of both types of climate variations, the analysis of past climate variations is also necessary for better estimations of future climate change.

The past climate of the Baltic Sea region can be reconstructed from palaeo-pollen and dendroclimatological records with different time resolutions and degrees of accuracy. Palaeo-pollen data in lake sediments give information about the dominant plant species of a certain period. Combining the environmental ranges of those species in terms of annual maximum temperatures, minimum temperatures, and total annual precipitation allows an approximate reconstruction of past climate conditions over the past millennia, with time resolutions of a few decades (e.g. Kühl et al., 2002). Dendroclimatological data of tree ring widths, wood density, and sometimes also carbon and oxygen isotopic composition in tree rings can be dated as exactly as at annual scales.

As described by the BACC II Author Team (2015), the climate history of the Baltic Sea region during the Holocene, i.e. the last 12 000 years, involved very large climate changes – much larger than those during the 20th century. These climate changes were caused by strong changes in external forcing factors, in particular the Earth's orbit. These changes first brought about a warming that terminated the Last Ice Age about 13 000 years BP, then caused a period of very warm temperatures (∼3^∘ above preindustrial levels) centred around 6000 years BP (the Holocene Thermal Maximum), followed by a slow temperature decline towards preindustrial levels. During this long period, other shorter-lived climate events, with durations of a few centuries, caused abrupt drops in temperature. These events, e.g. the Younger Dryas (12 000 years BP) or the 8.2 K event (8200 years BP), were possibly related to abrupt changes in the North Atlantic circulation, when sudden melting of portions of the remnants of the North American ice sheet disturbed the circulation of the North Atlantic Ocean and disrupted the poleward heat transport.

In general, annual precipitation is believed to have changed with the slow multicentennial-scale changes in temperature. Warmer periods, in particular the Mid-Holocene Optimum, tended to be wetter, although the regional heterogeneity may have been larger than for temperature.

Following the end of the last glaciation, the coastlines of the Baltic Sea underwent changes due to the interplay between the rising global sea level and the local rebound of the Earth's crust after the disappearance of the Fennoscandian ice sheet. The weight of this ice sheet depressed Fennoscandia by about 500 m, and its slow, viscous rebound continues until today, with a rate of about 10 mm yr⁻¹ at the northern Baltic Sea coast. Due to this interplay, the Baltic Sea experienced periods of open or closed connections to the North Sea that governed the transport of salinity and heat and the nature of the Baltic Sea ecosystems (Groß et al., 2018).

The climate evolution during more recent historical times – the past 1000 years (Sect. 3.1.3) – can be reconstructed with better accuracy and higher time resolution due to better dendrochronological data availability. These data show the imprint of the Medieval Warm Period (approx. 900–1350 AD), the Little Ice Age (approx. 1550–1850 AD), and the Contemporary Warm Period (1850–present) on the Baltic Sea region. These periods were likely caused by long-term internal climate variability and changes in the external forcing (volcanic eruptions and solar radiation) and, during the Contemporary Warm Period, also by the increase in anthropogenic greenhouse gases (Hegerl et al., 2003).

In the Baltic Sea, this succession of warm–cold–warm temperatures was accompanied by changes in the deepwater oxygen content, with low oxygen conditions in warmer periods (see Sect. 3.1.2). The reasons for these oxygen variations are still not fully understood but may be relevant for the future should future warming also cause lower oxygen concentrations.

3.1.2 New palaeoclimate reconstructions

Since the publication of the BACC II report (BACC II Author Team, 2015), new reconstructions of the evolution of the European climate over the Holocene and over the past millennium have become available. Like previous reconstructions, the new ones are based on palaeo-pollen data and now comprise summer and winter temperatures and summer and winter precipitation. They are available for 1000 year time segments (Mauri et al., 2015). The reconstructions of the late spring–summer temperature evolution over the past millennium are based on dendroclimatological data, as in the previous reconstructions, but they are now based on wood density measurements, which reflect the slow climate variations better than tree ring width. These reconstructions are available for western Europe from 755 CE onwards (Luterbacher et al., 2016). In this study, only the results for a regular geographical box approximately covering the Baltic Sea region are discussed.

In addition, new regional climate simulations since the publication of the BACC II report have demonstrated the connections between the Baltic Sea and North Atlantic climates on multidecadal timescales better (Schimanke et al., 2012; Schimanke and Meier, 2016; Börgel et al., 2018, 2020; Kniebusch et al., 2019a).

3.1.3 Holocene climate evolution

The picture of the Holocene climate evolution from the BACC II report (BACC II Author Team, 2015) is essentially confirmed, but the regional details are now clearer (Mauri et al., 2015). Between 7000 and 5000 years BP, the Baltic Sea region (especially the western Baltic Sea) experienced a period with summer temperatures about 2.5–3 ^∘C warmer than in the preindustrial reference period (before the 20th century). However, according to these reconstructions, the eastern Baltic Sea region (Finland and the Baltic states, i.e. Lithuania, Latvia, and Estonia) did not experience a Mid-Holocene Optimum in summer, when temperatures were similar to the preindustrial period. In contrast, winter temperatures showed a clear Mid-Holocene Optimum over the whole Baltic Sea region, lasting about 8000–4000 years BP, with winter temperatures roughly 3 ^∘C warmer than during the preindustrial period. In the eastern Baltic Sea, winter temperatures were even slightly higher, especially between 6000–5000 years BP. As a result, annual mean temperatures during the millennia of the Mid-Holocene Optimum were generally warmer than in the preindustrial period. This warming was limited to the winter in the eastern Baltic Sea, where the amplitude of the annual temperature cycle was clearly lower than in the preindustrial period.

The warm temperatures in the Baltic Sea region during the Mid-Holocene Optimum are not surprising and basically agree with the previous review (BACC II Author Team, 2015). They also agree with evidence from regions farther north, indicating that the Arctic Ocean in summer may have been ice-free during this period (Jakobsson et al., 2010). These findings do not contradict the anthropogenic effect on climate observed during recent decades. During the Mid-Holocene Optimum, the orbital configuration of the Earth was different and favoured warmer temperatures at northern high latitudes, especially in summer, as explained later. For the analysis of climate impacts on ecosystems, it is relevant that high latitudes were exposed to warm temperatures and reduced sea ice cover just a few millennia ago. However, at that time, temperature changed at a much slower pace – around 2–3 ^∘C over several millennia – compared to the present and projected rates of about 2 ^∘C in just a few decades.

For precipitation, the new reconstructions give a regionally more nuanced view of climate evolution during the Holocene. The BACC II report (BACC II Author Team, 2015) indicated that warmer climates were generally more humid. The new reconstructions (Mauri et al., 2015) modulate this vision and constrain the wetter conditions to the eastern Baltic Sea region, both in summer and winter seasons, with a stronger signal in winter. Precipitation anomalies in the eastern Baltic Sea region were of the order of + 1–2 mm per month relative to preindustrial climate. In the western Baltic Sea region, the Mid-Holocene Optimum tended to be slightly drier than the preindustrial reference period, both in summer and winter, with precipitation deficits of the order of 1–2 mm per month.

The main external forcing that drove the millennial climate evolution over the Holocene period is the changing orbital configuration of the Earth (the so-called Milankovitch cycles), as explained in the BACC-II report (BACC II Author Team, 2015), and especially the variation in the time of the year of the perihelion (when the Earth is nearest to the Sun). The perihelion is now at the beginning of January, but ∼ 10 000 years BP, it was in July. This changes the seasonal distribution of solar insolation and determines the rate of melting of winter snow and its possible survival into the next winter. The solar insolation at 60^∘ N at the top of the atmosphere during the Holocene, derived from Laskar et al. (2004), is depicted in Fig. 3. The shift of the perihelion from summer to winter diminishes summer insolation – and, in principle, summer temperature – and increases winter insolation during the past few millennia. The long-term evolution of temperatures would, however, not be a linear response to the long-term evolution of the seasonal insolation. For instance, the presence or absence of ice sheets may influence the timing of the response to increasing insolation during the early Holocene, delaying the Holocene temperature maximum with respect to the annual insolation maximum. In wintertime, the insolation is rather weak, so its effect may be overwhelmed by other factors, such as changes in the large-scale atmospheric and oceanic heat transports.

For the Fifth IPCC Assessment Report (IPCC, 2014b), the Mid-Holocene climate was simulated with 14 global ESMs within CMIP5 (Schmidt et al., 2011). These models were essentially the same as those used for future climate projections, although, in some cases, with a few simplifications required by limitations in computer power and by the long timescales involved. These models were driven by known external forcings, including the orbital forcing. The common evaluation of these simulations with reconstructions helps us to interpret the reconstructions and sheds light on model limitations. An important aspect in this comparison was that the spatial resolution of global models was relatively coarse, about 2×2 degrees longitude × latitude, so that smaller details within the Baltic Sea region cannot be properly represented.

The simulations showed some agreements with the reconstructions but also clear, not-yet-resolved disagreements (Mauri et al., 2014). In summer, all models showed temperatures 2–3^∘ warmer during the Mid-Holocene Optimum than in the preindustrial climate (Fig. 3). However, no model showed the gradient with clearer warming in the western Baltic Sea, as seen in the reconstructions. For wintertime, the disagreement was much clearer. Whereas reconstructions show a clear warming over the whole region, the 14 models displayed widely varying patterns of temperature change. Only three models agreed with the reconstructions. For precipitation, the models disagreed with the west (wet) to east (dry) dipole shown by the reconstructions for summer and winter precipitation (Fig. 3). Not a single simulation showed this pattern of summer precipitation change, and in general, the simulated precipitation deviations were much smaller than in the reconstructions. This disagreement regarding temperature (especially in winter) and precipitation, known as the Mid-Holocene conundrum, is not unique to the Baltic Sea region and was also found for the Mediterranean (Mauri et al., 2014; Liu et al., 2014). Errors in the applied external (orbital) forcing can be ruled out, as this forcing can be accurately calculated during this period. The reasons for the disagreement are still unknown. They may involve the influence of chaotic internal climate variations (unlikely over such long timescales), model deficiencies, or reconstruction inaccuracies.

3.1.4 The past millennium

For shorter periods closer to the present, like the past 1 or 2 millennia, the data available for reconstructing past climate are denser and more accurate. Abundant dendroclimatological information is available, dated to the exact year, in contrast to the uncertain decadal-scale dating of palaeo-pollen data. Recently, temperature reconstructions for western Europe, spatially resolved and approximately covering the last 1200 years, have become available (Luterbacher et al., 2016) and are presented here in some detail for the Baltic Sea region. These data are based on the analysis of wood density in tree rings. Wood density is more sensitive to the growing season temperature than tree ring width. In addition, tree ring width variations usually contain multidecadal-scale variations that are too weak, even when the year-to-year variations in temperature may be well captured. This makes wood density a better proxy for temperature reconstructions at these latitudes.

Figure 4 shows the reconstructed growing season temperature (spring–early summer) for the period 755–2000 CE, averaged over the Baltic Sea region, based on the European reconstructions by Luterbacher et al. (2016). The reconstructed spring–early summer temperature displays warmer conditions around 950 CE, confirmed also by the previous pollen-based reconstructions (Mauri et al., 2015), and colder conditions between 1200 and 1850 CE, followed by the recent warming. This temperature evolution confirms that presented by the BACC II Author Team (2015). The temperatures in the Medieval Warm Period and the Contemporary Warm Period (mid-20th century) are similar within their respective error bars. According to these reconstructions, the Little Ice Age was, on average, about 0.8^∘ colder than the 20th century.

There is no new analysis of the causes of this temperature evolution specific to the Baltic Sea region. For Europe as a whole, for which the reconstructions display a similar temporal pattern, the main identified forcings were volcanic activity – more intense during the Little Ice Age and weaker during the Medieval Warm Period – and solar activity, with roughly the reverse temporal signal (Luterbacher et al., 2016). In the context of industrialisation since the 19th century, anthropogenic greenhouse gases have become the dominant forcing.

The CMIP5 project also included simulations of the past millennium with ESMs, although with fewer models than for the Mid-Holocene. These simulations have been compared with the temperature reconstructions for Europe, yielding agreement in general. However, for the Baltic Sea region, the simulated temperature changes tend to be smaller than those reconstructed, especially for the transition between the Medieval Warm Period and the Little Ice Age, with a modelled temperature difference of only ∼0.2^∘ (compare with Fig. 4 by Luterbacher et al., 2016).

Climate fluctuations are driven not only by the external forcings but also by the chaotic internal dynamics of the Earth system. Regional climate simulations indicate that North Atlantic temperature variability influences Baltic Sea temperatures (Kniebusch et al., 2019a) and precipitation (Börgel et al., 2018). North Atlantic temperatures tend to fluctuate internally at multidecadal timescales, i.e. the AMO, which influences the atmospheric circulation of the Baltic Sea region. Furthermore, the interaction between internal modes of climate variability has recently been identified as a key driver for the state of the Baltic Sea. Internal fluctuations in the North Atlantic are likely to influence the spatial position of the NAO, affecting the regional importance of this climate mode for the Baltic Sea (Börgel et al., 2020).

Climate simulations also indicate an impact of internally driven climate variability on the frequency of wind extremes. In the present climate, the wintertime wind regime in the Baltic Sea is linked to the NAO, but at the longer timescales of the preindustrial period, variations in wind extremes appear related neither to the mean wind conditions nor to the external climate forcings (Bierstedt et al., 2015). In the recent centuries, the main driver of trends of wind extremes over land appears to be land use changes such as de- and reforestation (Bierstedt et al., 2015; Gröger et al., 2021a, and references therein).

An important question is how North Atlantic variations can influence the state of the Baltic Sea, especially its oxygen conditions, since freshwater input and water temperature (less strongly) affect the stratification of the water column and, therefore, the exchange of oxygen between the surface and deeper layers. Temperature also modulates algal blooms and, thus, dissolved oxygen when bacteria use oxygen to decompose dead algae. An analysis of sediment cores indicated that the Mid-Holocene Optimum, the Roman Period (2000 years BP), and the Medieval Warm Period were all periods of oxygen deficiency at the bottom of the Baltic Sea. Low oxygen conditions are also observed during the Contemporary Warm Period, unique in their extent on a 1000 year perspective (Norbäck Ivarsson et al., 2019, and references therein). Hence, factors other than temperature, like nutrient input into the Baltic Sea, can also affect oxygen conditions, and thus, the reasons for those hypoxic phases during the past millennia are not yet completely clear (Schimanke et al., 2012). It had been suggested that agricultural nutrient input was large enough to influence oxygen conditions during the Medieval Warm Period already, and perhaps it was also modulated by changes in river runoff due to the described climate fluctuations (Zillén and Conley, 2010). However, a detailed analysis of new sediment records find little evidence of anthropogenic eutrophication before the industrial period (Norbäck Ivarsson et al., 2019; Ning et al., 2018; van Helmond et al., 2018). In view of the large temperature increases projected for this region in the next decades, further study of the influence of climate on oxygen conditions is warranted.

3.2.1 Atmosphere

3.2.1(1) Large-scale atmospheric circulation

Long-term trends in NAO could not be detected (e.g. Deser et al., 2017; Marshall et al., 2020). For the period 1960–1990, a positive trend in NAO, with more zonal circulation, mild and wet winters, and increased storminess in central and northern Europe was found (Hurrell et al., 2003; Gillett et al., 2013; Ruosteenoja et al., 2020). However, from the mid-1990s to the early 2010s, there was a tendency towards more negative NAO indices, i.e. a more meridional circulation and more cold spells in winter (Fig. 5).

There is no consensus on how strongly the interannual NAO variability is forced externally (Stephenson et al., 2000; Feldstein, 2002; Rennert and Wallace, 2009). Several external forcing mechanisms have been proposed, most prominently SST (Rodwell et al., 1999; Marshall et al., 2001) and sea ice in the Arctic (Strong and Magnusdottir, 2011; Peings and Magnusdottir, 2016; Kim et al., 2014; Nakamura et al., 2015). However, other authors (Screen et al., 2013; Sun et al., 2016; Boland et al., 2017) found no dependence on sea ice extent. Furthermore, the impact of changes in the Arctic on midlatitude dynamics is still under debate (Dethloff et al., 2006; Francis and Vavrus, 2012; Barnes, 2013; Cattiaux and Cassou, 2013; Vihma, 2017). However, Scaife and Smith (2018) suggested that the atmospheric circulation in climate models might not be sensitive enough to changes in, for instance, sea surface conditions.

A weakening of the zonal wind, eddy kinetic energy, and amplitude of Rossby waves in summer (Coumou et al., 2015), as well as an increased waviness of the jet stream associated with Arctic warming (Francis and Vavrus, 2015) in winter have been identified, which may be linked to an increase in blocking frequencies. Blackport and Screen (2020) argued that previously observed correlations between surface temperature gradients and the amplitude of Rossby waves have broken down in recent years. Therefore, previously observed correlations may have to be reinterpreted as internal variability. On the other hand, it has been shown that observed trends in blocking are sensitive to the choice of the blocking index, and that there is a huge natural variability that complicates the detection of forced trends (Woollings et al., 2018), compromising the robustness of observed changes in blocking.

With ongoing global warming, the Arctic will warm faster than the rest of the Earth. This decrease in the poleward temperature gradient will tend to weaken the westerlies and increase the likelihood of blockings. On the other hand, maximum warming (compared to other tropospheric levels) will occur just below the tropical tropopause due to the enhanced release of latent heat, which tends to increase the poleward gradient, strengthen upper-level westerlies, and affect the vertical stability, thus altering the vertical shear in midlatitudes. It is not clear which of these two factors will have the largest effect on the jet streams (Stendel et al., 2021).

The atmospheric circulation over Europe naturally varies significantly on decadal timescales (Dong et al., 2017; Ravestein et al., 2018). Proposed drivers for these circulation changes include polar and tropical amplification, stratospheric dynamics, and the AMOC (Haarsma et al., 2015; Shepherd et al., 2018; Zappa and Shepherd, 2017). The attribution of drivers is more straightforward for local changes, in particular for the soil moisture feedback, for which an enhancement of heatwaves due to a lack of soil moisture has been demonstrated (Seneviratne et al., 2013; Teuling, 2018; Whan et al., 2015). Räisänen (2019) found only a weak effect of circulation changes on the observed annual mean temperature trends for 1979–2018 in Finland, but circulation changes have considerably modified the trends in individual months. In particular, circulation changes explain the lack of observed warming in June, the very modest warming in October in southern Finland, and about a half of the very large warming in December.

As part of its natural variability, the North Atlantic warmed from the late 1970s to 2014 (Fig. 6). Recently, the AMO began transitioning to a negative phase again (Frajka-Williams et al., 2017). Palaeoclimate reconstructions and model simulations suggest that the AMO might change its dominant frequency over time (Knudsen et al., 2011; Wang et al., 2017). The impact of the AMO on climate is, however, independent of its frequency (Börgel et al., 2018, 2020). Its influence on regional climate has been analysed in several studies (Enfield et al., 2001; Knight et al., 2006; Sutton and Hodson, 2005; Ting et al., 2011; Casanueva et al., 2014; Ruprich-Robert et al., 2017; Peings and Magnusdottir, 2014), with some dealing with the Baltic Sea (Börgel et al., 2018, 2020; Kniebusch et al., 2019a). Kniebusch et al. (2019a) suggested that the influence of the AMO on the warming of Baltic Sea SSTs during 1980–2008 might have been at least as strong as that induced by humans (IPCC, 2014b).

3.2.1(2) Air temperature

A significant increase in surface air temperature in the Baltic Sea region during the last century has been shown previously (e.g. BACC Author Team, 2008; Rutgersson et al., 2014; BACC II Author Team, 2015). The temperature increase was not monotonous but accompanied by large multidecadal variations that divided the 20th century into the following three main phases: (1) warming from the beginning of the century until the 1930s, (2) slight cooling until 1960s, and (3) a distinct warming during the last decades of the time series that has continued also during 2014–2020 (Figs. 7 and 8; Table 9).

Linear trends of the annual mean temperature anomalies during 1878–2020 were 0.10 ^∘C per decade north of 60^∘ N as well as south of 60^∘ N in the Baltic Sea region. This is larger than the global mean temperature trend and slightly larger compared to the earlier BACC reports. Over the Baltic Sea surface, air temperature trends were smaller than over land. During 1856–2005, surface air temperature over the Baltic Sea increased by 0.06 and 0.08 ^∘C per decade in the central Baltic Sea and in Bothnian Bay, respectively (Kniebusch et al., 2019a).

There is a large variability in the annual and seasonal mean temperatures, particularly during winter, but the warming is seen for all seasons (being the largest during spring in the northern part of the region).

Both daily minimum and daily maximum temperatures have increased. A decrease in the daily temperature range (DTR) has been observed in many regions of the world, but there is no clear signal for the entire Baltic Sea region (see for example, Jaagus et al., 2014, for a DTR analysis of the Baltic states).

These changes have also resulted in seasonality changes in that the growing season has lengthened by about 5 d per decade in the period 1965–2016 (Cornes et al., 2019). From this follows that the cold season has become shorter.

Extreme air temperatures can be high or low, but extended periods of extreme temperatures (spells or waves) are often the most influential. When averaged over land areas, warm spell duration increased during recent decades (Rutgersson et al., 2022). For some regions, the annual number of days defined to belong in warm spells increased from 6–8 to 14 during recent decades. Along with more frequent and longer warm spells came decreases in the frequency, duration, and severity of cold spells, based both on observations (Easterling et al., 2016) and model results. The length of the frost season and the annual number of frost days also decreased (Sillmann et al., 2013).

3.2.1(3) Solar radiation and cloudiness

Multidecadal variations in solar radiation at the Earth's surface, called dimming and brightening, have been observed in Europe and other parts of the world, particularly in the Northern Hemisphere (Wild et al., 2005, 2017; Wild, 2012).

One of the world's longest time series of global radiation, i.e. incoming solar radiation at the Earth's surface, is from Stockholm, where measurements started in 1922. Recently, a first attempt to homogenise this time series was made by Josefsson (2019). No significant trend was found over the whole time series, but there were large variations over 1–3 decades. Other long time series of global radiation in northern Europe are from Potsdam, Germany, (Wild et al., 2021), and Tõravere, Estonia, (Russak, 2009). All three time series show a minimum in global radiation around the mid-1980s. Then a clear increase or brightening of about 5 %–8 % followed, until at least 2005. Before the 1980s minimum, there was a period of dimming at all stations but with differences in the details. In Potsdam, there was rather stable dimming with decreasing solar radiation from the late 1940s until the mid-1980s. In Tõravere, the dimming period started around the mid-1960s, while in Stockholm the dimming phase started around 1950 but with a temporary interruption with increased solar radiation around 1970.

Current 20th century reanalyses models provide results for surface solar radiation. However, most of them fail to capture multidecadal surface radiation variability in central and southern Europe (Wohland et al., 2020). The CERA20C reanalysis, which shows best results for central and southern Europe, still gives questionable results over Scandinavia, showing a weak increase instead of a decrease in surface solar radiation during the presumed dimming period before 1980.

Satellite data allowing analyses of cloudiness and solar radiation at the Earth's surface have been available since the early 1980s. For Europe, important work has been done within the EUMETSAT Satellite Application Facility on Climate Monitoring (CM SAF). Several satellite data records have been validated and used in climate studies (e.g. Urraca et al., 2017; Pfeifroth et al., 2018). At the highest latitudes of the Baltic Sea region, there are, however, larger inaccuracies (Riihelä et al., 2015) or often no data at all due to the low standing Sun and slant viewing geometry from the satellites.

The satellite data only cover the latest brightening period observed at ground-based stations in Europe. While the geographical patterns of global average cloud conditions agree well among several satellite cloud datasets, there are clear differences in the distribution and size of trends (Karlsson and Devasthale, 2018). However, there seems to be consensus on a decreasing trend in total cloud fraction of about 1–2 % per decade over the Baltic Sea region during 1984–2009.

Recent CM SAF satellite products on solar irradiance at the Earth's surface, the SARAH-2 and CLARA-A2 datasets, both agree well with station data, according to Pfeifroth et al. (2018). In many cases, this holds both for climatological averages and for trend detection. The average trend for the period 1983–2015 is about + 3 W m⁻² per decade, both at the stations closest to the Baltic Sea and in the SARAH-2 dataset. The three long-term stations mentioned above are all used as reference stations for the satellite data validation. For example, the ongoing monitoring at stations spread all over Sweden show an average increase of about 8 % (corresponding to + 4 W m⁻² per decade) from 1983 until 2005–2006 (SMHI, 2021). In later years, the solar radiation levelled off, but, inter alia, the extremely sunny year of 2018 in northern Europe contributed to keeping the trend increasing over time.

The multidecadal variations in the solar radiation at the Earth's surface were most probably caused by a combination of changes in cloudiness and in anthropogenic aerosols. Which of the two drivers gives the largest contribution is still an open question and might differ among regions. Aerosol concentrations over northern Europe decreased during the brightening period from the mid-1980s onwards (Ruckstuhl et al., 2008; Russak, 2009; Markowicz and Uscka-Kowalkowska, 2015; Glantz et al., 2019). Russak (2009) considered changes in cloudiness caused by variations in atmospheric circulation to be the most important factor in Estonia, but aerosol changes also played a role. In an early study of the modern radiation measurements in Sweden, the strong increase in solar radiation 1983–1997 was also accompanied by a clear decrease in total cloud cover, especially during the half-year of summer (Persson, 1999). The satellite datasets SARAH-2 and CLARA-A2 were both derived using an aerosol climatology as input. This underlines the important role of changes in cloudiness for surface solar radiation. Stjern et al. (2009) also stressed the importance of the contribution of clouds and the atmospheric circulation for dimming and brightening periods in northern Europe.

Other studies, e.g. Ruckstuhl et al. (2008) and Wild et al. (2021), concluded that aerosol effects under clear skies is the main contributor to the multidecadal variations in solar radiation in central Europe. Aerosol-induced multidecadal variations in surface solar radiation could be expected also over oceans (Wild, 2016), but long-term measurements are lacking. The interaction between aerosols and clouds, the indirect aerosol effects, also needs to be better understood and quantified.

3.2.1(4) Precipitation

During the 20th century in the Baltic Sea region, changes in precipitation were spatially more variable than for temperature (BACC II Author Team, 2015). Irregularly distributed precipitation measurement stations make it difficult to determine statistically significant trends and regime shifts. Sweden has shown an overall increasing trend in precipitation since the 1900s, in particular since the mid-20th century (Chen et al., 2020). In Finland, the overall increase detected for 1961–2010 is neither regionally consistent nor always statistically significant (Aalto et al., 2016). The same holds for the Baltic states (Jaagus et al., 2018). In the south of the Baltic Sea region, changes were small and not significant. Nevertheless, precipitation averaged over the Baltic Sea catchment area has increased since 1950 due to an increase in winter (Fig. 9). The number of heavy precipitation days is largest in summer. Compared to southern Europe, precipitation extremes in the Baltic Sea region are not as intense, with the 95th percentile of wet day precipitation amounts typically ranging from 8 to 20 mm (Cardell et al., 2020). Extreme precipitation intensity increased during the period 1960–2018. An index for the annual maximum 5 d precipitation (Rx5d) shows significant increases of up to 5 mm per decade over the eastern part of the Baltic Sea catchment (EEA, 2019b). The change is more pronounced in winter than in summer.

3.2.1(5) Wind

In situ observations allow direct analysis of winds, in particular over the sea (e.g. Woodruff et al., 2011). However, in situ measurements, especially over land, are often locally influenced, and inhomogeneities make the straightforward use of such data difficult, even for recent decades. Therefore, many studies use reanalyses rather than direct wind observations. But the analysis of storm track activity for longer periods using reanalysis data suffers necessarily from inaccuracies associated with changing data assimilation and observations before and after the introduction of satellites, resulting in large variations in storm track changes across assessments (Wang et al., 2016; Chang and Yau, 2016). Another concern about wind trends in reanalyses, especially over land, is related to an appropriate consideration of the effects of land use changes on surface roughness.

Owing to inherent inhomogeneities and the large climate variability in the Baltic Sea region, it is unclear whether there is a general trend in wind speed in the recent climate. Results regarding changes or trends in the wind climate are strongly dependent on the period and region considered (Feser et al., 2015). Due to the strong link to large-scale atmospheric variability over the North Atlantic, conclusions about changes over the Baltic Sea region are perhaps best made in a wider spatial context, considering, inter alia, the NAO.

Recent trend estimates for the total number of cyclones over the Northern Hemisphere extratropics during 1979–2010 revealed a large spread across the reanalysis products, strong seasonal differences, and decadal-scale variability (Tilinina et al., 2013; Wang et al., 2016; Chang et al., 2012, 2016; Matthews et al., 2016). Common to all reanalysis datasets is a weak upward trend in the number of moderately deep and shallow cyclones but also a decrease in the number of deep cyclones, in particular for the period 1989–2010. Chang et al. (2016) reported a minor reduction in cyclone activity in the Northern Hemisphere summer due to a decrease in baroclinic instability as a consequence of Arctic temperatures rising faster than at low latitudes. Chang et al. (2012) also noticed that state-of-the-art models (CMIP5) generally underestimate this trend. In the Northern Hemisphere winter, recent studies reported a decrease in storm track activity related to Arctic warming (Ceppi and Hartmann, 2015; Shaw et al., 2016; Wills et al., 2019; Stendel et al., 2021).

Despite large decadal variations, there is still a positive trend in the number of deep cyclones (<980 hPa) over the last 6 decades, which is consistent with results based on the NCEP (National Centers for Environmental Prediction) reanalysis since 1958 over the northern North Atlantic (Lehmann et al., 2011). Using an analogue-based field reconstruction of daily pressure fields over central to northern Europe (Schenk and Zorita, 2012), the increase in deep lows over the region might have been unprecedented since 1850 (Schenk, 2015). However, for limited areas, the conclusions were rather uncertain.

The effect of differential temperature trends on storm tracks has been recently addressed, both in terms of upper tropospheric tropical warming (Zappa and Shepherd, 2017) and lower tropospheric Arctic amplification (Wang et al., 2017), including the direct role of Arctic sea ice loss (Zappa et al., 2018) and a possible interaction of these factors (Shaw et al., 2016). The remote and local SST influence has been further examined by Ciasto et al. (2016), who confirmed the sensitivity of the storm tracks to the SST trends generated by the models and suggested that the primary greenhouse gas influence on storm track changes was indirect, acting through its influence on SSTs. The importance of the stratospheric polar vortex for storm track changes has recently also received attention (Zappa and Shepherd, 2017).

3.2.1(6) Air pollution, air quality, and atmospheric nutrient deposition

Air pollution continues to significantly impair the health of the European population, particularly in urban areas. Brandt et al. (2013) estimated the total number of premature deaths due to air pollution in Europe in the year 2000 to be ∼ 680 000 yr⁻¹. Although this number was predicted to decrease to approximately 450 000 by 2020, it is still a matter of concern. Particulate matter (PM) concentrations were reported to be the primary reason for adverse health effects. Estimates indicated that PM_2.5 concentrations in 2016 were responsible for ∼ 412 000 premature deaths in Europe, due to long-term exposure (EEA, 2019a).

The state of air pollution is often expressed as air quality when human health is in focus. The ambient air quality in the Baltic Sea region is dominated by anthropogenic emissions, and natural emissions play only a minor role. These emissions have shown an overall decreasing trend in recent years (EEA, 2019a), as reflected in the ambient concentrations reported in the EMEP (European Monitoring and Evaluation Programme) status report (Fagerli et al., 2018). To quantify air quality, concentrations of certain gases and particulate matter are used as measures. The general conclusions in the field of air quality reported by the BACC II Author Team (2015) still hold today, i.e. that land-based emissions and concentrations of major constituents continue to decrease due to emission control measures, with the possible exception of certain emissions from the shipping sector. Sulfur emissions from shipping have continued to decrease strongly in the Baltic Sea from 2015, due to much lower limit values for the sulfur content of ship fuel in the emission control areas. A noticeable decrease in nitrogen emissions due to the newly (2021) implemented nitrogen emission control area (NECA) is expected in the next decade.

In Europe, the pollutants most harmful to human health are PM, nitrogen dioxide (NO₂), and ground-level ozone (O₃). About 14 % of the EU-28 urban population was exposed to O₃ concentrations above the EU target value threshold (EEA, 2019a). When compared to other European countries, air pollution was relatively low in Scandinavia, except for a few urban traffic hotspots, with annual mean NO₂ concentrations elevated to near the limit value (EEA, 2019b). In this comparison, northern Germany is located in the lower mid-field, while northern Poland is among the more polluted countries, especially with PM. Biomonitoring samples analysed for toxic metals by Schröder et al. (2016) tended to show lowest concentrations in northern Europe.

Contributions to air pollution and pollutant deposition in coastal areas by shipping can be substantial. Major pollutants from shipping are SO₂, NO_x, and PM (including black carbon). The BACC II assessment estimated emissions from shipping in the Baltic Sea region. Several studies have recently been published, including Jonson et al. (2015), Claremar et al. (2017), and Karl et al. (2019b), which used chemistry transport models to predict ambient concentrations from known emissions. They showed, as expected, that the highest air pollution concentrations due to ship exhaust are found near major shipping lanes and harbours, but also that considerable concentrations of NO₂ and PM reach populated land areas. This effect is pronounced in the southwestern Baltic Sea area (Quante et al., 2021). Exact numbers from such modelling should still be interpreted with care, as shown by Karl et al. (2019a), who compared output from three state-of-the-art chemistry transport models for the Baltic Sea area.

The most important recent change in shipping emissions in the North Sea and Baltic Sea are due to the 2015 strengthening of the fuel sulfur content limit for the sulfur emission control areas (SECAs), by lowering the maximum allowed sulfur content from 1 % to 0.1 %. Model calculations indicated large reductions in sulfur deposition in countries bordering these two sea areas after the implementation of the lowered sulfur limit (Gauss et al., 2017). Barregard et al. (2019) estimated the contribution of Baltic Sea shipping emissions to PM_2.5 before 2014 and after 2016, when the new SECA regulation of marine fuel sulfur was implemented. These authors also estimated human exposure to PM_2.5 from shipping and its health effects in the countries around the Baltic Sea. They concluded that PM_2.5 emissions from Baltic Sea shipping, and resulting health impacts, decreased substantially after the 2015 SECA regulation. Population exposure studies estimating the influence of shipping emissions for selected Baltic Sea harbour cities were published for Rostock, Riga, and Gdańsk–Gdynia by Ramacher et al. (2019) and for Gothenburg by Tang et al. (2020). Ramacher et al. (2019) found that shipping emissions strongly influence NO₂ exposure in the port areas (50 %–80 %), while the average influence in home, work, and other environments is lower (3 %–14 %) but still with strong influence close to the ports. It should, however, be noted that the reduction in sulfur emissions to the atmosphere by the use of new cleaning techniques (e.g. open-loop scrubbers) can increase the risk of acidification and marine pollution (Turner et al., 2017, 2018).

Johansson et al. (2020) published a first comprehensive assessment of emissions from leisure boats in the Baltic Sea. While the modelled NO_x and PM_2.5 emissions from leisure boats are clearly lower than those from commercial shipping, these first estimates suggest that carbon monoxide (CO) emissions from leisure boats equal 70 % of the registered shipping emissions, and non-methane volatile organic carbon (NMVOC) emissions equal 160 %. It should be noted that most of the leisure boat emissions occur in summer and often occur near areas for nature conservation and tourism. Most of these emissions can be attributed to Swedish, Finnish, and Danish leisure boats, but the leisure boat fleet has the potential for large future increases also in Russia, Estonia, Latvia, Lithuania, and Poland.

Air pollution leads to environmental degradation by affecting natural ecosystems and biodiversity. Ground-level ozone (O₃) can damage crops, forests, and other vegetation, impairing growth and reducing biodiversity. According to a recent study by Proietti et al. (2021), for the period 2000–2014, no statistically significant trend in the O₃ mean concentration in northern Europe could be identified due to the large internal variability. The annual mean ozone concentration is reported to be slightly below 35 ppb (parts per billion), as compared to 43 to 45 ppb in the Mediterranean region, for which a significant decreasing trend is found. The exposure index AOT40 (sum of the hourly exceedances above 40 ppb for daylight hours during the growing season) significantly declined in all European regions except for northern Europe, for which a positive but not significant trend is seen. At the nation level, among the six European countries showing a positive trend were Denmark, Germany, and Sweden (Proietti et al., 2021). A clear difference in trends between rural sites and other station typologies is found for Europe for the period 2000 to 2017. That is, for traffic sites, a substantial increase in annual mean O₃ concentration was observed, in contrast to rural stations, for which a slight decrease was found (Colette and Rouïl, 2020). Regarding the monitored population exposed to a large number of days with high ozone concentrations, all countries in Europe showed a decrease from 2000 to 2014 (NDGT60 >25 d yr⁻¹; Fleming et al., 2018).

Harmful exposure and impacts of air pollutants on ecosystems are assessed using the concept of critical loads (CLs; Nilsson and Grennfelt, 1988). The CL is the total amount of pollutants that an ecosystem can tolerate without risking unacceptable damage. The most harmful air pollutants in terms of damage to ecosystems, in addition to O₃, are reduced and oxidised nitrogen. It is estimated that about 62 % of the European ecosystem area is still exposed to high levels of NO_x, leading to exceedances of CLs for eutrophication of open water bodies in all European countries in 2016 (EEA, 2019a). Hotspots of exceedances of CLs for acidification in 2016 were the Netherlands and its borders with Germany and Belgium, southern Germany, and also the Czech Republic. However, most of Europe including the Baltic Sea region did not exceed the CLs for acidification (EEA, 2019a).

Since the 1980s, the total nitrogen deposition on the Baltic Sea has decreased substantially due to an overall reduction in European emissions, but emission and deposition reductions have stalled since the mid-2000s (Colette et al., 2015; Gauss et al., 2021). Atmospheric phosphorus deposition remains highly uncertain in amount and trends (HELCOM, 2015; Kanakidou et al., 2018; Ruoho-Airola et al., 2012).

Air quality and climate interact in several ways. On the one hand, air pollutants can affect climate both directly and indirectly by changing the radiative balance of the atmosphere. On the other hand, climate change alters meteorological conditions, which may affect the concentrations of air pollutants via several pathways, since the air quality is strongly dependent on weather (Jacob and Winner, 2009). The effects of important meteorological and climate variables on surface O₃ and PM were discussed in a comprehensive review by Doherty et al. (2017). The connection between high temperatures and increased ground-level ozone concentrations is well established. Increases in temperature related to climate change (i.e. during heatwaves) are expected to lead to higher ozone concentrations in certain regions with the required precursor concentrations. Other important meteorological factors influencing air pollution concentrations are a possible change in the number of midlatitude cyclones and in the number of occurrences of, and the duration of, stagnant weather conditions (Jacob and Winner, 2009).

3.2.2 Land

3.2.2(1) River discharge

The total river discharge to the Baltic Sea is approximately 14 000 m³ s⁻¹ (Bergström and Carlsson, 1994). This is substantially more than the direct net precipitation (precipitation minus evaporation) on the Baltic Sea itself, which has been estimated at 1000–2000 m³ s⁻¹ (Meier and Kauker, 2003; Meier and Döscher, 2002; Meier et al., 2019d; see also the discussion by Leppäranta and Myrberg, 2009). In other words, most of the fresh water entering the Baltic Sea comes from the terrestrial part of the catchment. Therefore, the freshwater input to the Baltic Sea cannot be described entirely with only climatic parameters. Non-climatic drivers of runoff include river regulation by dams and reservoirs, land use changes in the catchment, and water uptake for irrigation. Although dams are known to have altered the seasonality of discharge (e.g. McClelland et al., 2004; Adam et al., 2007; Adam and Lettenmaier, 2008), they do not seem to be responsible for annual discharge changes. In the long-term, net precipitation over the catchment area and river runoff are strongly correlated (Meier and Kauker, 2003).

Figure 11Long-term dynamics (1970–2017) of annual nitrogen (a) and phosphorus (b) riverine inputs to the major Baltic Sea basins. BB – Bothnian Bay; BS – Bothnian Sea; BP – Baltic proper; GF – Gulf of Finland; GR – Gulf of Riga; DS – Danish straits; KT – Kattegat. Time (in years) is on the horizontal axis. Data source: updated from Savchuk (2018); see https://www.researchgate.net/publication/358726567_Long-term_times-series_of_the_annual_N_and_P_pools_in_the_major_basins_of_the_Baltic_Sea_1970-2020 (last access: 6 March 2022).

Figure 12Cumulative mass loss for four Swedish glaciers, i.e. Storglaciären (since 1946), Rabots glaciär (since 1982; no data for 2004 and 2007 and, hence, interpolated), Riukojietna (since 1986; data for 2004 interpolated), and Mårmaglaciären (since 1990; no data for 2020). Data source: SITES data portal at https://data.fieldsites.se/portal/ (last access: 10 February 2022); World Glacier Monitoring Service, 2021; Swedish Infrastructure for Ecosystem Science, 2021a, b, c).

For the period 1850–2008, the total river discharge from the Baltic Sea catchment area, reconstructed from observations (Bergström and Carlsson, 1994; Cyberski et al., 2000; Hansson et al., 2011; Mikulski, 1986) and hydrological model results (Graham, 1999), showed no statistically significant trend but a pronounced multidecadal variability, with a period of about 30 years (Meier et al., 2019d). Furthermore, summed river flow observations in the period 1900–2018 (Lindström, 2019) and a historical reconstruction of the annual river discharge for the past 500 years showed no statistically significant trend either (Hansson et al., 2011). However, river runoff from northern Sweden, part of the catchment area of the Bothnian Bay, has significantly increased since the 1980s compared to 1911–2018 (Lindström, 2019).

There are indeed substantial regional and decadal variations in the river flow. Stahl et al. (2010) studied the near-natural rivers of Europe over the period 1942–2004 and found a clear overall pattern of positive trends in annual streamflow in the northern areas. Kniebusch et al. (2019b) also identified a statistically significant positive trend in the river discharge to the Bothnian Bay for 1921–2004. In Estonian rivers, regime shifts in annual specific runoff corresponded to the alternation of wet and dry periods (Jaagus et al., 2017). A dry period started in 1963–1964, followed by a wet period from 1978, with the latest dry period commencing at the beginning of the 21st century.

For the period 1920–2005, positive trends in streamflow at stations of a pan-Nordic dataset dominate annual mean, winter, and spring figures, whereas summer trends are statistically not significant (Wilson et al., 2010). A clear signal of earlier snowmelt floods and a tendency towards more severe summer droughts in southern and eastern Norway were found.

The observed temperature increases have affected streamflow in the northern Baltic Sea region for 1920–2002 in a manner corresponding well to the projected consequences of a continued rise in global temperature in terms of increasing winter time discharges (Hisdal et al., 2010). However, the regional impacts of precipitation change on both the observed and projected changes in streamflow are still unclear, as the combined effects of changes in precipitation and temperature are still not well known (Stahl et al., 2010).

In the northern Baltic Sea region, all the way south to the Gulf of Finland, runoff is strongly linked to the climate indices air temperature, wind, and rotational circulation components. In the southern region, runoff is associated more with the strength and torque of the cyclonic or anticyclonic pressure systems (Hansson et al., 2011).

In the Baltic states, changes in streamflow over the 20th century showed a redistribution of runoff over the year, with a significant increase in winter and a tendency for decreasing spring floods (Reihan et al., 2007; Sarauskiene et al., 2015; Jaagus et al., 2017). A similar winter trend has also been found for the reconstructed river discharge to the entire Baltic Sea since the 1970s (Meier and Kauker, 2003).

For the period 1911–2010, a trend of observed annual maximum daily flows in Sweden could not be detected (Arheimer and Lindström, 2015). However, in particular, the annual minimum daily flows in northern Sweden considerably increased in the period 1911–2018 (Lindström, 2019). Analysing a pan-European database, Blöschl et al. (2017) showed that river floods over the past 5 decades occurred earlier in spring due to (1) an earlier spring snowmelt in northeastern Europe, (2) delayed winter storms associated with polar warming around the North Sea, and (3) earlier soil moisture maxima in western Europe.

3.2.2(2) Land nutrient inputs

The Baltic Sea catchment area of 1.7 × 10⁶ km², which is more than 4 times larger than the sea surface area of the Baltic Sea (see Fig. 1), is populated by over 84 × 10⁶ inhabitants. Stretching between 49–69^∘ N and 10–38^∘ E, the catchment exhibits significant gradients in both natural (precipitation, river discharge, temperature, etc.) and anthropogenic (population density and occupation and agricultural and industrial development, etc.) environmental factors. These factors change both in time (phenological changes, long-term trends, and lags due to land cover processes) and space (north–south gradients in climate and land use and east–west gradients in socioeconomic features and climate), thus determining heterogeneity and variation in land nutrient inputs that drive long-term eutrophication of the Baltic Sea (Savchuk, 2018, and references therein; Kuliński et al., 2021).

Estimates of nutrient inputs had been attempted since the 1980s (e.g. Larsson et al., 1985; Stålnacke et al., 1999) and are now being compiled within a permanent process of the HELCOM Pollution Load Compilation (PLC; e.g. Sonesten et al., 2018). However, these data officially reported to HELCOM by the participating riparian states have been, and are still, suffering from gaps and inconsistencies. Therefore, the best available estimates have been reconstructed in attempt to both fill in such gaps and correct possible sources of inconsistencies (Savchuk et al., 2012; Svendsen and Gustafsson, 2020). For long-term studies of the Baltic Sea ecosystem, a historical reconstruction of nutrient inputs since 1850 has been made available (Gustafsson et al., 2012).

According to HELCOM (HELCOM, 2018a; Savchuk, 2018) and updated estimates (Sonesten et al., 2018), substantial reductions in land nutrient inputs, comprising riverine inputs and direct point sources at the coast, have been achieved since the 1980s (Fig. 11; Table 10). Since there are no statistically significant trends in annual river discharge (Sect. 3.2.2(1)), these reductions are attributed to socioeconomic development, including the expansion of the wastewater treatment and the reduction in atmospheric nitrogen deposition (Gauss et al., 2021) over the entire Baltic Sea drainage basin and not to climate-related effects (HELCOM, 2018a; Svendsen and Gustafsson, 2020). As an example, the coastal point sources of total nitrogen (TN) and total phosphorus (TP) decreased 3- and 10-fold, respectively, compared to the 1990s (Savchuk et al., 2012) and today contribute fewer nutrients to the Baltic Sea than they did in 1900 (Savchuk et al., 2008; Kuliński et al., 2021).

Agriculture is the main source of anthropogenic diffuse nutrient inputs, which comprise 47 % of the riverine nitrogen and 36 % of the riverine phosphorus inputs (Sonesten et al., 2018). In turn, mineral fertiliser dominates the anthropogenic nutrient inputs to the Baltic Sea catchment, in particular in its intensely farmed southern part (Hong et al., 2017). However, during 2000–2010, only about 17 % of the net anthropogenic nitrogen and only about 4.7 % of the net anthropogenic phosphorus input to the catchment were exported with rivers to the sea (Hong et al., 2017). While denitrification might have removed part of the nitrogen applied in agriculture, the remaining phosphorus has accumulated in the drainage basin. A global budget estimated that agriculture has increased the soil storage of phosphorus in the drainage basin by 50 Mt during 1900–2010 (Bouwman et al., 2013), and a regional approach calculated an increase by 40 Mt during 1900–2013 (McCrackin et al., 2018). However, McCrackin et al. (2018) estimated that about 60 % of these phosphorus inputs were retained in a stable pool and did not contribute noticeably to the riverine export. About 40 % accumulated in a mobile pool with a residence time of 27 years. McCrackin et al. (2018) suggested that leakage from this mobile legacy pool is the, though slowly declining, dominant source of present riverine phosphorus inputs.

3.2.3 Terrestrial biosphere

Previous assessments

Referring to the comprehensive review of climate-related changes in terrestrial ecosystems in the first BACC report (BACC Author Team, 2008), Smith et al. (2008), concluded that climate change during the preceding 30–50 years had already caused measurable changes in terrestrial ecosystems in the Baltic Sea region, e.g. an advancement of spring phenological phases in some plants, upslope displacement of the alpine treeline, and increased land surface greenness in response to improved growth conditions and a richer CO₂ supply. But as nearly all ecosystems in the region were managed to some extent, the climate impacts might be alleviated or intensified by human interventions, e.g. by choosing favourable tree species in forestry. The observed trends were expected to continue for at least several decades, assuming that continued future increases in the atmospheric CO₂ concentration will cause continued warming.

In the second BACC report (BACC II Author Team, 2015), climate effects on terrestrial ecosystems were less in focus, with a section related to forests and natural vegetation in the chapter on environmental impacts on coastal ecosystems, birds, and forests (Niemelä et al., 2015) and as part of the chapter on socioeconomic impacts on forestry and agriculture (Krug et al., 2015). On the other hand, the second BACC report also considered anthropogenic land cover changes as a driver of regional climate change (Gaillard et al., 2015). Niemelä et al. (2015) concluded that the observed positive effects of climate change on forest growth would continue, in particular for boreal forest stands that benefitted more than temperate forest stands. The species composition of natural vegetation in the Baltic Sea region was expected to undergo changes, with a predominantly northward shift of the hemiboreal and temperate mixed forests. Terrestrial carbon storage was likely to increase in the region, but land use change could play an important modifying role, affecting this storage both positively and negatively. Krug et al. (2015) concluded that there were regional differences in how the vulnerability and productivity of forestry systems were affected by climate change, with the southern and eastern parts of the Baltic Sea region likely to experience reduced production and the northern and western parts increased production. Gaillard et al. (2015) found no indication that deforestation in the Baltic Sea region since 1850 could have been a major cause of the observed climate warming.

Acknowledging the importance of the land component in the climate system, the IPCC recently published its special report entitled “Climate Change and Land” on climate change, desertification land degradation, sustainable land management, food security, and greenhouse gas fluxes in terrestrial ecosystems (IPCC, 2019b). This is because land, including its water bodies, provides the basis for human livelihoods and wellbeing through primary productivity, the supply of food, fresh water, and multiple other ecosystem services.

Biophysical and biogeochemical interactions

The land surface and its terrestrial ecosystems in the Baltic Sea region interact with the atmosphere and are, thus, coupled to the local and regional climate. These interactions determine the exchanges of heat, water, and momentum between the land surface and the atmosphere via biophysical processes, the exchange of greenhouse gases, e.g. CO₂, CH₄ and N₂O, and emissions of black carbon from forest fires, aerosol precursors, e.g. biogenic volatile compounds, or organic carbon aerosols via biogeochemical processes altering the atmospheric composition.

The nature of the biophysical and biogeochemical interactions between the land surface and the atmosphere and their effects on climate are studied by comparing the effects of forests with that of open land, e.g. grassland, pastures, or cropland (e.g. Bonan, 2016). Details on the nature of these feedbacks are given in Gröger et al. (2021b).

On average, across the globe, forests absorb atmospheric CO₂ and, thus, reduce net radiation and have a cooling effect on climate. In equilibrium, forest ecosystems, are expected to be carbon neutral, with carbon loss through phenological turnover, mortality, and decomposition over large areas on average, matching plant productivity. CO₂ fertilisation is considered a strong driver for the terrestrial carbon sink, but demographic recovery following past land use, e.g. afforestation and replanting of harvested forest stands, likely provides an equally important explanation for net carbon uptake by forests in industrialised regions of North America, Europe, and Asia (Pugh et al., 2019). Deforestation, on the other hand, can lead to a release of carbon and to an increase in net radiation and, thus, has a warming effect on climate (Friedlingstein et al., 2020, and references therein). In contrast to the biophysical effects and some of the other biogeochemical interactions, the biogeochemical interactions associated with the carbon cycle have global impacts and operate at very long timescales.

Anthropogenic land use and land cover changes

Using remote sensing data, Jin et al. (2019) investigated recent trends in springtime plant phenology in the Baltic Sea region and the sensitivities of phenological trends to temperature and precipitation, in spring, winter, and summer. Considering the entire region and combining all vegetation types, the authors found an advancement of the growing season by 0.30 d yr⁻¹ over the period 2000–2016. The advancement was particularly strong for evergreen needleleaf forests (0.47 d yr⁻¹) and weaker for cropland and grassland (0.14 d yr⁻¹). Evergreen needleleaf forests, together with deciduous broadleaf forests, dominate the northern part of the Baltic Sea region, while the southern part is mainly grassland and cropland. Jin et al. (2019) found that the most important driver of the advancement of the growing season is spring mean temperature, with an advancement rate of 2.47 d (^∘C)⁻¹ of spring warming when considering the entire area and all vegetation types. In this study, spring was defined as the 3 months preceding the start of the growing season. Spring drying could further increase the advancement rate by a 0.18 d cm⁻¹ decrease in precipitation. While the sensitivity of the start of the growing season to climate conditions in spring is comparable for the entire Baltic Sea region, the sensitivity to climate conditions in the previous summer and winter seasons differs between the northern and southern parts of the region. In both seasons, an increase in the mean temperature was found to advance the growing season in the southern part of the Baltic Sea region but to delay the start of the growing season in the northern part, which is in contrast to the spring warming. These sensitivities were markedly stronger for summer than for winter. These trends in plant phenology in spring result in changes of the land cover early in the year, thus affecting climate through biophysical and biogeochemical interactions with the atmosphere.

Considering the mechanisms for the delayed start of the growing season in Fennoscandia in relation to a warming in the preceding summer and winter seasons, Jin et al. (2019) evaluated the effects on plant dormancy. A winter warming could, for instance, prolong the chilling accumulation required to break winter dormancy of trees. Later summer temperatures, on the other hand, could affect bud dormancy initiation, while reduced soil moisture associated with higher summer and autumn temperatures could delay the leafing and flowering of plants. However, the details of the mechanisms involved have not yet been explored.

Observations reveal a local impact of changes in forest cover on near-surface temperatures, due to biophysical effects that depend on the geographical latitude, roughly separating the boreal regions and the temperate zone of the Northern Hemisphere. When investigating the effects of small-scale clearings at sites in the Americas and Asia, M. Zhang et al. (2014) found, on both continents, that annual mean temperatures cooled over open land north of about 35^∘ N and warmed south of this latitude. Changes in forest cover have, however, different effects on daily minimum and daily maximum temperatures. M. Zhang et al. (2014) found that the warming effect over open land south of 35^∘ N was related to an increase in daily maximum temperatures, with little change in daily minimum temperatures, while the cooling effect to the north was due to a decrease in daily minimum temperatures. Lee et al. (2011) showed consistent results for North America, where the cooling effect of 0.85±0.44 ^∘C over non-forested areas north of 45^∘ N was due to a decrease in daily minimum temperatures associated with the reduced roughness length. At night, open land cools more than forests, regardless of the geographical latitude. This is confirmed by Alkama and Cescatti (2016), who analysed the impacts of recent losses in forest cover on near-surface and land surface temperatures in the boreal zone. For both, the authors found cooling trends in daily minimum and warming trends in daily maximum temperatures in response to deforestation and opposite tendencies after afforestation. These effects were somewhat stronger for land surface temperatures than air temperatures.

RCMs have been used to investigate the biophysical effects of changes in forest cover on climate in Europe. Strandberg and Kjellström (2019), for instance, used simulations with the Rossby Centre Atmosphere (RCA) RCM to assess the climate effect of maximal afforestation or deforestation in Europe, focusing on seasonal mean temperatures and precipitation, and daily temperature extremes. Maximum afforestation and deforestation were inferred from a simulation with the LPJ-GUESS dynamical vegetation model, providing a map for potential natural forest cover for Europe in equilibrium with present-day climate (Gröger et al., 2021a). To simulate maximum afforestation, present-day land cover classes, which represent considerable agricultural activity in Europe (particularly in western, central, and southern Europe) were replaced by the potential natural forest cover. In the case of deforestation, on the other hand, the potential natural forest cover was converted to grassland in the model.

The simulations indicated that afforestation in Europe generally increased evapotranspiration, which, in turn, led to colder near-surface temperatures. In western, central, and southern Europe, the cooling in winter due to afforestation was between 0.5 and 2.5 ^∘C. The cooling effect was somewhat stronger in summer, exceeding 2.5 ^∘C in large parts of western and southeastern Europe. Deforestation had the opposite effect, with warmer near-surface temperatures due to decreased evapotranspiration, typically in the range between 0.5 and 2 ^∘C in western and central Europe and reaching up to 3 ^∘C in southeastern Europe. In regions with low evapotranspiration, however, changes in the surface albedo were relatively more important for temperatures. During summer, warming by deforestation affected the entire Baltic Sea region (in the range between 0.5 and 1.5 ^∘C), while the cooling associated with afforestation only affected its southern part. Over parts of Scandinavia, afforestation actually resulted in a slight warming of about 0.5 ^∘C. In winter, the cooling effect of afforestation was only evident over the southern part of the Baltic Sea region, while deforestation had no effect on winter temperatures.

Strandberg and Kjellström (2019) found relatively strong biophysical effects of afforestation or deforestation in Europe on daily maximum temperatures in summer. Deforestation markedly increased daily maximum temperatures over the entire Baltic Sea region (typically between 2 and 6 ^∘C), while afforestation lowered daily maximum temperatures in the southern part of the region by about 2 to 6 ^∘C and slightly increased temperatures over parts of Scandinavia. In contrast to its cooling effect on mean winter temperatures, afforestation lead to a warming of the daily minimum temperatures in the southern part of the Baltic Sea region in the range between 2 and 6 ^∘C.

In a similar study with the regional model REMO, Gálos et al. (2013) investigated the biophysical effects of afforestation in Europe on climate and also compared these effects to the climatic changes expected from future global warming. Potential afforestation was implemented by specifying deciduous forest cover at all vegetated areas that were not covered by forests at the end of the 20th century, mainly in western, central, and eastern Europe. Given the strong historical deforestation in central Europe, there is the potential for rather extensive afforestation in the southern part of the Baltic Sea region but less potential in the northern part, i.e. Scandinavia and Finland. The results indicated a cooling effect of the re-established forests in boreal summer exceeding 0.3 ^∘C, mainly related to increased evapotranspiration from the trees, in combination with intensified fluxes of latent heat due to stronger vertical mixing. The stronger latent heat fluxes also enhanced precipitation by more than 10 % in some regions. These effects of potential afforestation counteracted the projected future changes in climate, i.e. somewhat reduced the magnitude of the pronounced future warming and markedly reduced the future drying in the boreal summer. In some cases, such as in northern Germany, the enhanced precipitation was found to completely offset the drying effect of future warming. More recent results (Meier et al., 2021b) have used rain gauge data to estimate precipitation changes induced by land cover change. Meier et al. (2021b) created a statistical model to show that reforestation of agricultural land can increase precipitation locally, especially in winter, and were able to separate the effects on both local and downwind precipitation regionally and seasonally. They also found that climate-change-induced summer precipitation reductions could be offset by reforestation, with a particularly strong effect in southwestern Europe. However, their analyses also indicate small precipitation increases in the Baltic Sea region, relative to a baseline scenario with no land cover change, consistent with the results of Gálos et al. (2013).

In a more regionalised study, Gao et al. (2014) applied an RCM to investigate the biophysical effects of peatland forestation in Finland before (1920) and after drainage (2000s). In Finland, as in other northern European countries, vast areas of naturally treeless or sparsely tree-covered peatland were drained for timber production in the second half of the 20th century. The total peatland area of Finland was estimated to be 9.7 × 10⁶ ha in the 1950s, but at the beginning of the 21st century, the area of peatland drained for forestry was estimated to 5.5 × 10⁶ ha. The authors found that the peatland forestation caused warming in spring, i.e. during the snowmelt season, and slight cooling in the growing season (May through October). The spring warming was mainly caused by decreased surface albedo and the cooling in the growing season by increased evapotranspiration.

3.2.4 Cryosphere

3.2.4(1) Snow

In the Baltic Sea region, snow cover is an important feature that greatly affects cold season weather conditions. It is characterised by very high interannual and spatial variability. Snow cover is a sensitive indicator of climate change, and its variations are closely related to air temperature in many regions. General climate warming is expected to reduce snow cover. Several thaw periods now interrupt snow cover, making it less stable. Total winter snowfall in northern Europe is projected to decrease but is still expected to increase in mid-winter in the very coldest regions (Räisänen, 2016).

Previous climate change assessments demonstrated a number of snow cover trends in recent decades in the Baltic Sea region (BACC Author Team, 2008; BACC II Author Team, 2015). A decrease in snow cover was observed in the south, while an increase in snow storage and the duration of snow cover was detected in the northeast and in the Scandinavian mountains. The spring snowmelt has become earlier in most parts of the region. As a result, the spring maximum river discharge has become smaller and earlier, shifting in many regions from April to March.

Recent investigations confirm these results. Snow cover in the Northern Hemisphere has decreased since mid-20th century (IPCC, 2014a), as also shown by satellite measurements (Estilow et al., 2015). The largest decline in the extent of snow cover has been observed in March–April and also in summer. Using the satellite-based National Oceanic and Atmospheric Administration (NOAA) Climate Data Record (CDR) for the period 1970–2019, it was shown that the annual snow cover fraction has reduced over most areas of the Northern Hemisphere by up to 2 % per decade (Zhu et al., 2021). Thereby, the annual snow cover area has reduced by 2×10⁵ km² per decade.

In 1980–2008, the snow cover duration in northern Europe decreased by about 3–7 d per decade, and the trend was significant at many stations (Peng et al., 2013). Most of the reduction happened in spring, with the end date of snow cover 5 d earlier per decade, on average (Peng et al., 2013). Snow cover variability over Europe is closely related to temperature fluctuations, which, in turn, are determined by large-scale atmospheric circulation during the cold season (Ye and Lau, 2017). A recent study of European snow depth data in 1951–2017 demonstrated an accelerated decrease after the 1980s (Fontrodona Bach et al., 2018), with an average decline, excluding the coldest climates, of 12.2 % per decade for mean snow depth and 11.4 % per decade for its maximum. A decreasing trend in snow density was detected in the eastern Baltic Sea region in 1966–2008 (Zhong et al., 2014).

In Poland, rather large changes in snow cover parameters were found for 1952–2013 (Szwed et al., 2017). The duration of snow cover decreased in almost the whole country, but this change is mostly not statistically significant. The total reduction in snow cover duration was 1–3 weeks over the 62 years, but the mean and maximum snow depths did not change. The start date of snow cover has not changed, but the end date moved slightly earlier (Szwed et al., 2017). In a recent study, statistically significant decreasing trends in snow cover duration and of in snow depth, based on 40 Polish stations in 1967–2020, were found (Tomczyk et al., 2021). The trend values for the number of days with snow cover were from −3.5 to −4.9 d per decade.

The snow cover regime at 57 stations in the eastern Baltic Sea region (Lithuania, Latvia, and Estonia) in 1961–2015 was analysed by Rimkus et al. (2018). The mean decrease in snow cover duration was 3.3 d per decade and was statistically significant at 35 % of the measuring sites, mostly in the southern part of the region. There were no trends in maximum snow depth. An earlier study for Lithuania found similar results (Rimkus et al., 2014).

A detailed study of snow cover data at 22 stations in Estonia during the period 1950–1951 to 2015–2016 revealed remarkable decreasing trends (Viru and Jaagus, 2020). Snow cover duration decreased significantly at 16 stations, and the mean decrease was 4 d per decade. Start dates for permanent snow cover had a non-significant tendency to occur later. Permanent snow cover had a statistically significant trend to end earlier at almost all stations. There were no overall trends in maximum snow depth in Estonia in 1951–2016 (Viru and Jaagus, 2020).

Significant decreases in snow depth parameters were found in Finland in recent decades (Aalto et al., 2016; Luomaranta et al., 2019). Regional differences were substantial. In 1961–2014, the largest decrease in snow depth occurred in the southern, western, and central parts of Finland in late winter and early spring. In northern Finland, a decrease in snow depth was most evident in spring, with no change in the winter months, even though the amount of solid precipitation was found to increase in December–February (Luomaranta et al., 2019). Winter mean snow depth (Jylhä et al., 2014) and the annual maximum snow depth (Lehtonen, 2015) has decreased significantly at many stations in southern Finland. At the same time, the annual maximum snow depth has not changed in Finnish Lapland (Lépy and Pasanen, 2017; Merkouriadi et al., 2017).

For the century 1909–2008, a general decrease was detected in many snow cover parameters at three stations in different parts of Finland (Irannezhad et al., 2016). A sharp decline in annual peak snow water equivalent has been detected since 1959. The period of permanent snow cover shortened by 21–32 d per century. However, it should be noted that the study of Irannezhad et al. (2016) was based on a temperature index snowpack model using daily temperature and precipitation as input rather than directly using snow observations.

A decline in snow cover parameters was found also in Norway for 1961–2010 (Dyrrdal et al., 2012; Rizzi et al., 2017).

3.2.4(2) Glaciers

A recent basic inventory of Scandinavian glaciers is available through the Randolph Glacier Inventory (RGI Consortium, 2017; Pfeffer et al., 2014), a collection of digital outlines of the world's glaciers prepared to meet the needs of the Fifth IPCC Assessment Report (IPCC, 2014b; Vaughan et al., 2014). It has since been updated in support of the IPCC Special Report on the Ocean and Cryosphere in a Changing Climate (IPCC, 2019a; Hock et al., 2019). Of the 3417 Scandinavian glaciers reported in the Randolph Glacier Inventory (v6.0), 365, with a combined area of ca. 360 km², lie within the Baltic Sea drainage basin, as defined by Vogt et al. (2007, 2008), and are all in the Scandinavian mountains. The combined glacier volume estimate (following Farinotti et al., 2019) for the Scandinavian glaciers reported in Hock et al. (2019) is 0.7±0.2 mm global sea level equivalent or 254±72 Gt. The rate of mass loss for all Scandinavian glaciers (not only in the Baltic Sea drainage basin) for the period 2006–2015 is 2±1 Gt yr⁻¹, corresponding to a negligible 0.01±0.00 mm yr⁻¹ of global sea level rise equivalent, yet with potential importance for local streamflow (Hock et al., 2019; Zemp et al., 2019). With a 90 %–100 % likelihood, atmospheric warming is the primary driver of glacier mass loss (Hock et al., 2019).

Most of the 365 glaciers in the Baltic Sea drainage basin are in Sweden (ca. 72 % by number; ca. 75 % by area), with the remaining ones in Norway. Both Sweden and Norway do nationally coordinated glacier monitoring, with the most recent results summarised in the Global Glacier Change Bulletin (GGCB) no. 3 (World Glacier Monitoring Service, 2021; World Glacier Monitoring Service, 2020) as, among other factors, glacier mass balance changes. A glacier mass balance year usually covers the period from 1 September to 31 August in the subsequent year. The GGCB includes four Swedish glaciers, two of which (Rabots glaciär and Storglaciären in the Kebnekaise massif, which has the world's longest continuous mass balance record, starting in 1945–1946) are so-called reference glaciers. This means that their dynamics are not dominated by non-climatically driven dynamics such as calving or surging, and that more than 30 years of ongoing measurements are available (Fig. 12). In Table 11, mass balances of the Swedish GGCB glaciers, following the World Glacier Monitoring Service (2021), are summarised. None of the Norwegian GGCB glaciers are in the Baltic Sea drainage basin.

These increasingly negative mass balances with time coincide with globally increasing air temperatures, with the latest 6 years, 2015–2020, being the warmest since instrumental recording began (World Meteorological Organization, 2020). Regional and local deviations in mass loss from that anticipated by long-term global warming are, however, expected. For example, slightly positive mass balances were observed for Mårmaglaciären, Storglaciären, and Riukojietna in 2016–2017 as a result of a cold summer (Swedish Infrastructure for Ecosystem Science, 2020; World Glacier Monitoring Service, 2021).

The ice summit of Kebnekaise Sydtopp (south peak) lost its status as Sweden's highest in September 2019 and 2020 when, due to melting of its ice-covered summit, its elevation dropped below that of the rocky, non-ice-covered Kebnekaise Nordtopp (north peak; Stockholm University – Department of Physical Geography, 2017, 2019, 2020).

3.2.4(3) Permafrost

The drainage basin of the Bothnian Bay is characterised by low mean annual air temperatures and includes boreal forest and mountain ecosystems, as well as large peatland areas. In this region, permafrost – ground frozen for at least 2 consecutive years – exists in high alpine environments and in peatlands. While almost all Baltic Sea drainage basin permafrost is found in the Bothnian Bay catchment, some isolated occurrences of permafrost are found in the upper reaches of alpine rivers farther south (mainly the Umeälven headwaters). Permafrost is a thermal state of ground, rock, soil, or sediment and occurs in regions with low mean annual air temperatures. Temperature is the strongest control on permafrost, but thin winter snow depth also favours permafrost aggradation and stability. In alpine permafrost, insolation is an important factor. In the upper Torne river catchment, incoming shortwave summer radiation causes a difference in the altitude of alpine permafrost from 850 m a.s.l. (above sea level) on shaded slopes to 1100 m a.s.l. on south-facing slopes (Ridefelt et al., 2008). In lowland areas, permafrost is more likely to form and persist in peatlands, where the low thermal conductivity of peat insulates the ground from warm summer air (Seppälä, 1986). Mire complexes with palsas, elevated peat mounds with an ice-rich permafrost core, are the most common form of lowland permafrost in the Baltic Sea drainage basin (Luoto et al., 2004). Palsa mires in the Baltic Sea basin are predominantly found in regions with a mean annual air temperature of <−3 ^∘C and low mean annual precipitation (often <450 mm), based on the 1961–1990 climate period (Fronzek et al., 2006).

Earlier maps of the Northern Hemisphere permafrost extent showed relatively extensive Fennoscandian permafrost, especially in alpine regions (Brown et al., 1997). Recent advances in permafrost modelling reveal a more nuanced picture, where permafrost is spatially patchy, persisting at high elevations and in lowland regions with low precipitation and large expanses of peat plateaus (Obu et al., 2019; Gisnås et al., 2017). Consistent with the observation of permafrost warming and thawing (Biskaborn et al., 2019), models indicate substantial permafrost losses in recent decades. High-resolution modelling (1 km pixels) driven by remotely sensed land surface temperature showed that ca. 6200 km² of permafrost in 1997 was reduced to 4800 km² in 2018 (Obu et al., 2020). Most of this loss is modelled alpine permafrost (here defined as >700 m a.s.l.), decreasing from 4700 to 3700 km², while lowland permafrost has decreased from 1500 to 1100 km² (Fig. 13).

Figure 13Modelled permafrost extent of alpine (>700 m a.s.l.) and lowland permafrost for the years 1997–2018 in the Baltic Sea drainage basin. Data source: Permafrost data are from Obu et al. (2020), the extent of the catchment is from Hannerz and Destouni (2006), and the elevation data are from the U.S. Geological Survey (USGS) Global Multi-resolution Terrain Elevation Data 2010 (GMTED2010). Analyses are performed at 1 km resolution in an equal area projection.

3.2.4(4) Sea ice

Introduction

Sea ice is an essential indicator of climate change and variability in the Baltic Sea region. Not only does the existence of sea ice indicate the general severity of the winter due to its close correlation with winter air temperature, but, in addition, parameters such as the annual maximum sea ice extent of the Baltic Sea (MIB), duration of ice season, and maximum thickness of level ice have been monitored regularly in the Baltic Sea since the late 19th century. The MIB was observed even earlier.

The BACC II Author Team (2015) concluded that all sea ice observations demonstrated large interannual variations, but with a long-term, statistically significant, trend to milder ice conditions that is projected to continue in future (Fig. 14). In this section, we review recent ice climate research in the Baltic Sea and provide updated figures on sea ice trends and projections that largely confirm previous conclusions.

Figure 14The graph on the left shows the annual maximum sea ice extent of the Baltic Sea (MIB) in kilometres squared (km²) during 1720–2020. Blue bars show the annual mean, and red lines show the 15-year running means. The dashed bars represent the error range of the early observations (Vihma and Haapala, 2009). The error range of the 30-year moving average is indicated by two red curves, which converge into one when high-quality data became available. The graphs on the right show the 30-year distribution functions of MIB during 1909–1938 and 1991–2020. Data sources: https://www.eea.europa.eu/data-and-maps/daviz/maximum-extent-of-ice-cover-3#tab-chart_1 (last access: 17 February 2022); Finnish Meteorological Institute (https://en.ilmatieteenlaitos.fi/ice-season-in-the-baltic-sea, last access: 17 February 2022).

An important indicator of advancing climate change in the Baltic Sea region is that two of the latest six ice winters have been extremely mild (2015–2020). In winter 2015, the Bothnian Bay was not fully covered by ice (with a MIB of 51 000 km²), which is the first such extreme winter observed with certainty. The winter of 2020 was even milder, with a MIB of only 37 000 km², and the lowest value in a time series that began in 1720.

Sea ice conditions in the Baltic Sea

Recently, on average, the northern sea areas of the Baltic Sea are ice covered every year from December to May. During the mildest winters, only the Bothnian Bay (in a few years even only partially) and coastal zones of other basins are ice covered. In the past, the entire Baltic Sea was ice covered only in the most severe winters, e.g. 1940, 1942, and 1947 (Vihma and Haapala, 2009).

In the fast ice regions near the coast, sea ice grows by thermodynamic processes only. Maximum sea ice thickness in the fast ice regions typically amounts to 40–70 cm in the Bothnian Bay and 10–40 cm in the Bothnian Sea, Gulf of Finland, and Gulf of Riga. Ice thickness is regularly monitored at tens of fast ice sites.

Observations of sea ice thickness in drift ice are much more limited. A recent study combined all airborne electromagnetic ice thickness measurements in the Bothnian Bay and derived the first estimate of the basin-scale ice thickness distribution in the Baltic Sea (Fig. 15; Ronkainen et al., 2018). An important finding of that study was that mean ice thickness in drift ice regions is greater than the thickness of fast ice and also greater than the ice thickness indicated on the ice charts. As expected, the data showed large interannual variability, but temporal and spatial coverage was not sufficient for conclusions on changes in drift ice thickness.

Figure 15Sea ice thickness distribution in the Bothnian Bay, as estimated from helicopter electromagnetic measurements during February–March in five winters (a–e), and its five-winter average (f), also shown as a red line in panels (a–e). Source: Ronkainen et al., 2018.

Individual ice ridges caused by compression and shearing of ice drift can be 30 m thick. The largest gradients in ice motion are found in the coastal zone (Leppäranta et al., 2012), where mean ice thickness over several square kilometres can be 1–3 m (Ronkainen et al., 2018).

In some circumstances, sea ice can accumulate towards the coast and cause spectacular on-shore ridges or a ride-up of several hundred metres from the shore to the land, causing damage to built structures (Leppäranta, 2013). Such events have been observed in exposed coastal regions, where the stability of fast ice can be overcome by combinations of storms, currents, and water level (Leppäranta, 2013). In the Bothnian Bay, such events can occur regardless of the severity of ice seasons. In the southern Baltic Sea, on-shore ice has been common during severe winters. During the last 10 years, such events were observed in 2010, 2011, 2012, and 2019 (Girjatowicz and Łabuz, 2020).

Observed changes

Long-term changes of the MIB (Seinä and Palosuo, 1996; Niskanen et al., 2009) are shown in Fig. 14. The trend of the MIB during the last 100 years (1921–2020) is − 6400 km² per decade. This is almost twice the trend reported by the BACC II Author Team (2015), based on the period 1910–2011. Since 1987 no severe and no extremely severe ice winters and since 2012 only average, mild, or extremely mild winters have been observed. The latter sea ice conditions explain the accelerated trend after 2011.

The recent 30-year period (1991–2020) is definitely the mildest since 1720 (Uotila et al., 2015). The probability distribution of the MIB has shifted towards low values, with severe winters being very rare (Fig. 14). The 30-year mean MIB is now 139 000 km², and the winter of 2021, with a MIB of 127 000 km² on 15 February 2021 (Jouni Vainio, FMI, personal communication), was close to this mean. During the second-mildest 30-year period (1909–1938), the average MIB was 184 000 km², and during the last 100 years it was 182 000 km². The shape of the MIB probability distribution has changed, also indicating a change in sea ice extremes. According to the ice season classification (Seinä and Palosuo, 1996), the recent 30-year period includes only one severe ice winter (2011) and 13 mild ice winters.

Present ice conditions differ from the past to the extent that Rjazin and Pärn (2020) even suggested defining this change as a regime shift. They analysed changes in sea ice extent and air temperature in the Baltic Sea in 1982–2016, using a method of splitting the time series in two and concluded that a regime shift towards milder ice conditions occurred in 2006–2007.

Other studies complement these and BACC II conclusions. Kiani et al. (2018) examined the influence of atmospheric changes on ice roads between Oulu and Hailuoto. They used air temperature data to calculate freezing and thawing degree days and found that freezing degree days decreased and thawing degree days increased significantly during 1974–2009. As a consequence, the ice road season started later and ended earlier.

Merkouriadi and Leppäranta (2014) analysed ice thickness and freezing and break-up dates collected at Tvärminne Zoological Station, at the entrance to the Gulf of Finland. They found a decrease of almost 30 d in the ice-covered period and a reduction of 8 cm in annual maximum ice thickness in the last 40 years. Laakso et al. (2018) used observations from the Utö Atmospheric and Marine Research station during 1914–2016 and concluded that the length of the ice season has decreased from 10–70 d before 1988 to 0–35 d after 1988 in the northern Baltic Sea proper. Figures 16 and 17 show level ice thickness at Kemi and Loviisa and the length of the ice season at Kemi, Loviisa, and Utö, respectively. All graphs show statistically significant decreasing trends, except the level ice thickness at Kemi station, which was probably influenced by snow cover or changes in measurement location.

3.2.4(5) Lake ice

The recent change in ice phenology is probably the single most important climatically induced alteration in lake environments within the Baltic Sea catchment. The new literature demonstrates, almost unanimously, significant changes towards earlier ice break-up, later freeze-up, and shorter duration of ice cover across the Baltic Sea catchment, apart from the coldest climate regime in Lapland. The available centennial data indicate that the ice cover duration has decreased by several days per century, whereas the intensified warming in recent decades has produced a similar change per decade (Efremova et al., 2013; Filazzola et al., 2020; Kïaviòš et al., 2016; Knoll et al., 2019; Korhonen, 2019; Lopez et al., 2019; Nõges and Nõges, 2014; O'Reilly et al., 2015; Ptak et al., 2020; Sharma et al., 2016, 2020; Wrzesiński et al., 2015). However, very few lakes have responded only weakly to the warming trend, such as Lake Peipus in Estonia, probably due to increasing snowfall. In individual years, a positive wintertime NAO seems to be an important factor causing a short ice cover duration. Among the main properties that affect the ice cover of individual lakes are size, depth, and shoreline complexity.

3.2.5 Ocean and marine sediments

3.2.5(1) Water temperature

The main driver of annual mean water temperature variations and long-term changes is air temperature (Meier et al., 2019c, d; Kniebusch et al., 2019a; Dutheil et al., 2021). The Baltic Sea water temperature has risen the fastest at the sea surface (Meier et al., 2022). With time, the heat spreads downward through different processes, such as lateral inflows, vertical down-welling, and diffusion, and eventually the whole water column warms up, with the smallest trends in the cold intermediate layer between the thermo- and halocline (Meier et al., 2022).

Since the 1980s, marginal seas around the globe have warmed faster than the global ocean (Belkin, 2009), and the Baltic Sea has warmed the most (Belkin, 2009). Climate change and decadal variability led to an annual mean, area-averaged increase in Baltic SST of + 0.59^∘ per decade for 1990–2018 (Siegel and Gerth, 2019) and of + 0.5^∘ per decade for 1982–2013 (Stramska and Białogrodzka, 2015). Both figures were derived from satellite data. In accordance with earlier investigations (BACC Author Team, 2008; BACC II Author Team, 2015), SST variability in winter can be linked to the NAO (Stramska and Białogrodzka, 2015). However, the spatial maps of SST trends by Stramska and Białogrodzka (2015) differ from those by Lehmann et al. (2011), perhaps because of the differing horizontal resolution of the satellite data products. Linear trends for the Baltic Sea during 1982–2012 of 0.41^∘ per decade are slightly larger than 0.37^∘ per decade for the North Sea (Høyer and Karagali, 2016).

Figure 16Level ice thickness at Kemi, Finland, and Loviisa, Finland, during 1912–2019. Dots are the annual mean, and lines are the linear trend with 95 % confidence intervals. Data are available from the ice service of the Finnish Meteorological Institute, Helsinki, Finland, upon request (https://en.ilmatieteenlaitos.fi/ice-conditions, last access: 6 March 2022).

Figure 17Length of the ice season in days at Kemi, Loviisa, and Utö (Finland) during 1890–2019. Dots are the annual mean, and lines are the linear trend with 95 % confidence intervals. Data are available from the ice service of the Finnish Meteorological Institute, Helsinki, Finland, upon request (https://en.ilmatieteenlaitos.fi/ice-conditions, last access: 6 March 2022).

Using monitoring data, Liblik and Lips (2019) found that the upper layer has warmed by 0.3–0.6^∘ per decade and the sub-halocline deep layer by 0.4–0.6^∘ per decade in most of the Baltic Sea during 1982–2016. The total warming in the whole Baltic Sea at all monitoring stations was 1.07^∘ over 35 years, approximately twice that of the upper 100 m in the Atlantic Ocean.

During 1856–2005, the reconstructed Baltic Sea average annual mean SST increased by 0.03 and 0.06^∘ per decade in the northeastern and southwestern areas, respectively (Kniebusch et al., 2019a). The largest SST increase trends were found in the summer season in the northern Baltic Sea (Bothnian Bay). Bottom water temperature trends were smaller than SST trends, with the largest increase in the Bornholm Basin. Independent monitoring data support the results of the long-term reconstruction (Figs. 18 and 19; see also Meier et al., 2019c, d). The largest SST warming occurred in summer (May to September), while trends in winter were smaller (Kniebusch et al., 2019a; Liblik and Lips, 2019).

Figure 18Annual mean values of deseasonalised daily sea surface (left column) and bottom (right column) temperature (red dots) at seven monitoring stations during 1877–2018. For the location of the stations, see Fig. 2. The grey lines show the period when every station has data for every year (1954–2018). Data source: for Figs. 18, 19, 21, and 22, ICES data (https://www.ices.dk/data/data-portals/Pages/ocean.aspx, last access: 21 February 2022) for temperature and salinity (bottle data; i.e. from specific depths) were used. Postprocessing of the data was done, following Radtke et al. (2020), in order to overcome possible seasonal biases due to missing values in the observations. Therefore, gaps were statistically filled using GAMMs (general additive mixed models), taking the seasonality into account.

Figure 19Seasonal mean sea surface and bottom temperature values during 1877–2018 at Gotland Deep (BY15). Blue, green, yellow, and orange show winter, spring, summer, and autumn, respectively. The grey lines show the period when every station has data for every year (1954–2018).

During the more recent period of 1978–2007, the annual mean SST trend was 10 times higher, with a mean of 0.4^∘ per decade (Kniebusch et al., 2019a). Trends increased more in the northeastern areas than in the southwestern areas and exceeded the contemporary trends in air temperature. See also the MARNET station data at Darss Sill and Arkona Deep in the southwestern Baltic Sea (Fig. 20).

Figure 20Annual mean sea surface temperature anomalies, with respect to the reference period of 2002–2018, of deseasonalised measurements at the monitoring station BY2 (for the location, see Fig. 2) and the MARNET stations (Darss Sill and Arkona Basin) during 1980–2018. Data source: Oceanographic Data Center at the Federal Maritime and Hydrographic Agency ([email protected]).

The seasonal ice cover clearly plays an important role in the Baltic Sea by decoupling the ocean and the atmosphere in winter and spring. Hence, the large trends in air temperature in winter were not reflected by the SST trends because the air temperature was still below the freezing point. During the melting period, the ice albedo feedback led to larger trends in SST than during the ice-covered period because of a prolonged warming period of seawater.

It has been suggested that the accelerated warming in 1982–2006 might partly be explained by a dominance of the positive phase of the AMO (Kniebusch et al., 2019a). Furthermore, historical eutrophication redistributed the heat in the ocean by warming the surface layer more than the underlying layers, in particular during spring and summer, because the increased water turbidity caused an enhanced absorption of sunlight close to the sea surface. However, modelling studies suggest that the historical eutrophication had no impact on SST trends (Löptien and Meier, 2011).

The summer of 2018 was the warmest on instrumental record in Europe and also the warmest summer in the past 30 years in the southern half of the Baltic Sea (Naumann et al., 2019), with surface water temperatures 4–5^∘ above the 1990–2018 long-term mean. This heatwave was also observed in the bottom temperatures (Humborg et al., 2019). However, systematic studies on changes in Baltic Sea marine heatwaves are not available.

3.2.5(2) Salinity and saltwater inflows

During the last decade, many new insights have been gained about the salt balance of the Baltic Sea and the dynamics of inflow and mixing processes. The Major Baltic Inflow (MBI) in December 2014, in particular, triggered new investigations and is by far the most intensively observed and modelled inflow event. The pathways and timing of the inflowing water were tracked by observations (Mohrholz et al., 2015), and numerical modelling (Gräwe et al., 2015) could reproduce the salt mass and volume of the inflow, as calculated from observations. The inflow was found to be barotropically (pressure) controlled in the Danish straits but dominated by baroclinic (density stratification) processes on the pathway farther into the Baltic proper. At the Bornholm Gat and the Słupsk Furrow ,the water exchange showed the clear two-layer flow pattern of an estuarine circulation. The inflow-related studies were underpinned by theoretical work based on the famous Knudsen relation (Knudsen, 1900) for estuarine exchange flow and its extension to total exchange flow by Burchard et al. (2018). The contribution of the inflowing saline water to the spatial distribution of salt and the total salt budget depends essentially on mixing with ambient brackish waters. In the course of the inflow path, the character of the mixing process between the deep salty layer and the brackish water above changes with increasing depth and decreasing current velocities. In the entrance area, the mixing is dominated by the entrainment of brackish water into the eastward-spreading saline bottom water due to turbulence generated by shear instability. The sills between the consecutive Baltic basins are particular mixing hotspots (Neumann et al., 2017). In the deeper basins of the Baltic proper, boundary mixing driven by the interaction of currents and internal waves with the topography and mixing processes at sill overflows contribute to the upward salinity flux (Reissmann et al., 2009). Mixing in the eastern Gotland Basin was investigated during the Baltic Sea Tracer Release Experiment (BATRE). Using an inert tracer gas, the basin-scale vertical diffusivities were estimated to 10⁻⁵ m² s⁻¹, whereas the locally inferred diffusivities in the basin's interior were 1 order of magnitude lower (Holtermann et al., 2012). This finding holds also for the inflow of saline water in course of an MBI (Holtermann et al., 2017). The interior mixing is often controlled by double diffusive convection that leads to a typical staircase-like vertical stratification structure (Umlauf et al., 2018). The crucial role of boundary mixing at the basin rim was confirmed by Holtermann and Umlauf (2012) and Lappe and Umlauf (2016). Both studies identified near-boundary turbulence as the key process for basin-scale mixing. The main energy sources for boundary mixing are basin-scale topographic waves, deep rim currents, and near-inertial waves.

The temporal statistics of barotropic saline inflows were reviewed by Mohrholz (2018). In contrast to earlier investigations, Mohrholz (2018) found no long-term trend in inflow frequency but a pronounced multidecadal variability in 25 to 30 years. Lehmann and Post (2015) and Lehmann et al. (2017), who studied the frequency and intensity of large volume changes in the Baltic Sea due to inflows, likewise could not find a long-term trend. The distinction between MBIs and smaller inflows is artificial and does not correspond to the frequency distribution of the inflows, which shows an exponential decrease in frequency with increasing inflow intensity (Mohrholz, 2018). The classical MBIs are only responsible for about 20 % of the total salt input, while the rest is accounted for by medium and small inflows with much less pronounced interannual variability.

Palaeoclimate simulations nearly covering the recent millennium (Schimanke and Meier, 2016) have provided new insights into the long-term behaviour of the mean salinity of the Baltic Sea. In accordance with previous historical reconstruction studies (Schimanke and Meier, 2016; Meier and Kauker, 2003), Schimanke and Meier (2016) identified river discharge, net precipitation, and zonal winds as the main drivers of the decadal variability in Baltic Sea salinity. However, their relative contributions are not constant. Extreme periods with strong salinity decrease for about 10 years occurred once per century. Thus, the long stagnation period from 1976 to 1992 was obviously a rare but natural event, although its extreme duration might be caused by anthropogenic effects. The Baltic Sea salinity also has a natural centennial variability. Based on the same numerical simulations, Börgel et al. (2018) could show a strong coherence between the AMO climate mode and river runoff on timescales between 60 and 180 years. Accordingly, the Baltic Sea salinity and the AMO are correlated, probably due to the dominating impact of river discharge on salinity. The river runoff leads salinity changes by about 20 years during the entire modelling period of 850 years. Schimanke and Meier (2016) reported a similar lag of 15 years between river runoff and Baltic Sea salinity.

According to model results, multidecadal variations in runoff (Gailiušis et al., 2011; Meier et al., 2019d) explain about half the long-term variability in volume-averaged Baltic Sea salinity (Meier and Kauker, 2003). Radtke et al. (2020) found that the direct dilution effect was only responsible for about one-fourth of the multidecadal variability and proposed a link between river runoff and inflow activity. Furthermore, they found that the influence of vertical turbulent mixing is small. Saltwater inflows contribute to the multidecadal salinity variability, in particular for the bottom layer salinity. The positive trend of river runoff in the northern catchment area led to a significant increase in the north–south salinity gradient in the Baltic Sea surface water layer (Kniebusch et al., 2019b). The causal relationship between the trends in runoff and horizontal salinity gradient was confirmed by numerically sensitivity experiments. Additionally, their model simulations revealed a multidecadal oscillation of salinity, river runoff, and saltwater inflows of about 30 years, consistent with the long-term observations.

From observations during 1982–2016, Liblik and Lips (2019) detected decreasing surface (see also Vuorinen et al., 2015) and increasing bottom salinities, but no long-term trend in the total salt budget were found (see Fig. 21). Both temperature and salinity contribute to a strengthening of the vertical stratification. Enhanced freshwater fluxes combined with higher deepwater salinities intensify the vertical density gradient throughout the year.

Figure 21Annual mean values of deseasonalised daily sea surface (left column) and bottom (right column) salinity (blue dots) at seven important stations during 1877–2018. The grey lines show the period when every station has data for every year (1954–2018).

3.2.5(3) Stratification and overturning circulation

A direct consequence of increasing stratification is that mixing between well-ventilated surface waters and badly ventilated deepwater weakens, making the Baltic Sea vulnerable to deoxygenation of bottom waters (Conley et al., 2002). An increase in seasonal thermal stratification (e.g. Gröger et al., 2019) can additionally lower the vertical nutrient transport from deeper layers to the euphotic zone, thereby limiting nutrient supply and potentially affecting algal and cyanobacterial blooms, at least at the species level. The latter potential effect has not yet been thoroughly investigated. However, the hypothesis was supported by the results of Lips and Lips (2008), who found a correlation between cyanobacteria bloom intensity in the Gulf of Finland and the frequency of upwelling events along both coasts.

Since the start of regular salinity measurements at the end of the 19th century, the saline stratification has been dominated by sporadic inflows from the adjacent North Sea and variations in river discharge (Fig. 22). While no long-term trends could be demonstrated in Baltic Sea salinity during 1921–2004 (Kniebusch et al., 2019b) or in halocline depth during 1961–2007 (Väli et al., 2013), a trend towards increased horizontal salinity difference between the northern and southern Baltic Sea was found during 1921–2004 (Sect. 3.2.5(2)). Furthermore, vertical stratification increased in most of the Baltic Sea during 1982–2016, with the seasonal thermocline strengthening by 0.33–0.39 kg m⁻³ and the perennial halocline by 0.70–0.88 kg m⁻³ (Liblik and Lips, 2019).

Figure 22Difference between bottom and surface salinity as a measure for the vertical stratification (blue dots) during 1877–2018. Only time steps when both values were available are considered. The grey lines show the period when every station has data for every year (1954–2018).

Sensitivity studies with a numerical model suggested that the basin-wide overturning circulation will decrease if the climate warms or when river runoff increases but will tend to increase if global sea level rises (Placke et al., 2021). However, historical multidecadal variations in the overturning circulation are mainly wind driven. Multidecadal variations in neither river runoff nor saltwater inflow have had an impact, according to Placke et al. (2021).

3.2.5(4) Sea level

For the era of continuously operated satellite altimetry, absolute mean sea level (relative to the reference geoid) increased in the Baltic Sea. Available estimates vary, depending on the exact period considered, but are broadly consistent with or slightly above the global average (3–4 mm yr⁻¹; Oppenheimer et al., 2019; Nerem et al., 2018). For the period 1992–2012, Stramska and Chudziak (2013) estimated an increase of 3.3 mm yr⁻¹, and for the period 1993–2015, Madsen et al. (2019) estimated an increase of 4 mm yr⁻¹ in the Baltic Sea absolute mean sea level. For the period 1886–1889 to 2018, the analysis of Swedish mareograph data suggest a sea level rise of about 1-2 mm yr⁻¹ (Figs. 23 and 24). Passaro et al. (2021) showed that the increase is not uniform across the Baltic Sea but varies between about 2 mm yr⁻¹ in the western Baltic Sea and more than 5 mm yr⁻¹ in the Gulf of Bothnia for the period 1995–2019. The acceleration of the sea level rise in the Baltic Sea was studied by Hünicke and Zorita (2016). They found that present acceleration is small and could only be detected through spatial averaging of observations.

Figure 23Annual mean sea level changes in centimeters for 14 Swedish mareographs since 1886. The data are corrected for land uplift. The grey line shows a smoothed curve. Data source: https://www.smhi.se (last access: 5 March 2022) at the Swedish Meteorological and Hydrological Institute, Norrköping, Sweden, compiled by Berit Recklebe, Leibniz Institute for Baltic Sea Research Warnemünde, Rostock, Germany.

Figure 24Annual mean sea level in Stockholm during 1774–2018. Grey bars show the historic time series from Ekman (1988, 2003). Red is the SMHI data (RH2000, 1889–2018), and blue is the trend computed for 1774–1884 (estimated land uplift of 4.9 mm yr⁻¹) and extrapolated until 2018. The SMHI data were adjusted to the historical data, considering the mean difference during the overlapping time period, to make both time series comparable. The sea level rise of the SMHI data were corrected by estimated land uplift amounts of 1.13 mm yr⁻¹ during 1889–2018. Data source: https://www.smhi.se (last access: 5 March 2022) at the Swedish Meteorological and Hydrological Institute, Norrköping, Sweden.

Sea level changes relative to the coast are more complex, since land is rising in the northern Baltic Sea by up to about 10 mm yr⁻¹ and sinking in the southern Baltic Sea by about 0.5 mm yr⁻¹, relative to the Earth's centre of mass (Hünicke et al., 2015; Groh et al., 2017). In addition to the global mechanisms (thermal expansion due to warming and land–ice melting), sea level changes in the Baltic Sea are also affected by the changes in atmospheric circulation, water inflow from the North Sea, and changes in the freshwater budget (river runoff, precipitation, and evaporation). Precipitation and river runoff are linked to westerly winds and affect salinity and the salinity gradient across the Baltic Sea (Kniebusch et al., 2019b) and, thus, the sea level height and its gradient. Stronger-than-normal westerly winds are associated with increased transport across the Danish straits, which lead to an increase in the Baltic mean sea level. The correlations between the sea level height and westerly wind are higher in the eastern and northern parts and lower in the southern and western parts of the Baltic Sea. Westerly winds in the region became more intense from the 1960s to the early 1990s but weakened somewhat thereafter (Feser et al., 2015). Over longer periods, no significant long-term trend is detected (Feser et al., 2015).

The Baltic mean sea level shows a pronounced seasonal cycle, with a minimum in spring and maxima in late summer (in 1900–1930) or winter (in 1970–1998). According to Hünicke and Zorita (2008), the amplitude of the seasonal cycle increased over the 20th century. Other authors found different periods without systematic long-term trends (Barbosa and Donner, 2016) or even regional decreases (Männikus et al., 2020).

The Baltic sea level extremes are caused by strong atmospheric cyclones that are sometimes in interaction with seiches (Neumann, 1941; Wübber and Krauss, 1979; Matthäus and Franck, 1992; Wolski et al., 2014; Soomere and Pindsoo, 2016; Girjatowicz, 2004; Leppäranta, 2013; Orviku et al., 2011) or, more seldom and on smaller scales, by wind-induced meteotsunamis (Pellikka et al., 2020). Cyclones associated with strong onshore winds can cause coastal storm surges over several hours to about 1 d (Wolski and Wiśniewski, 2020). If the pathway of the cyclones is aligned along the west–east direction, they may also increase the volume of water in the Baltic Sea. Major inflow events are associated with volumes changes corresponding to an increase in the Baltic sea level of about 24 cm (Matthäus and Franck, 1992) at timescales of about 10 d or longer (Soomere and Pindsoo, 2016). Under such conditions, even moderate wind and wind surges may lead to coastal sea level extremes (e.g. Weisse and Weidemann, 2017). Extreme sea levels over a predefined threshold become more frequent with rising mean sea level (Pindsoo and Soomere, 2020). In addition, model results and an analysis of the observations indicate that changes in atmospheric forcing are responsible for the long-term increases in storm surges in some localised areas of the eastern Baltic Sea (Ribeiro et al., 2014). The presence of sea ice impedes the development of extreme sea levels by shielding the ocean surface from forcing by the wind. Coastal ice protects the coast from erosion by extreme sea levels, but a pileup of ice and ice growth down to the sea bottom can also result in the transport of bottom sediment when the ice drifts out to sea (Girjatowicz, 2004; Leppäranta, 2013; Orviku et al., 2011).

Storm surges caused by strong onshore winds represent a substantial hazard for the low-lying parts of the Baltic Sea coast and, in particular, the southwestern parts (Wolski et al., 2014), the Gulf of Finland (e.g. Suursaar and Sooäär, 2016), the Gulf of Riga (e.g. Männikus et al., 2020), and the Gulf of Bothnia (Averkiev and Klevanny, 2010). The highest surges were reported for the Gulf of Finland (about 4 m in 1824 in St. Petersburg) and the western Baltic Sea (more than 3 m in 1871; Wolski and Wiśniewski, 2020). For the Gulf of Riga and the western Baltic Sea, values around 2 and 1–1.5 m are frequent, respectively (Wolski and Wiśniewski, 2020). The 100-year storm surges are higher (up to 2.4 m) at the inner end of the basins, farthest away from the Baltic proper, than in the centre of the Baltic Sea (up to 1.2 m). No consistent long-term trend for an increase in extreme sea levels relative to the mean sea level of the Baltic Sea has been found, which is in agreement with earlier assessments (BACC II Author Team, 2015). This finding is supported by palaeoclimate model studies that show no influence on extreme sea levels in the North Sea in warmer climate periods compared to colder periods (Lang and Mikolajewicz, 2019) and by recent studies of sea level records that suggest a pronounced decadal to multidecadal variability in storm surges relative to the mean sea level (Marcos et al., 2015; Marcos and Woodworth, 2017; Wahl and Chambers, 2016). Although Ribeiro et al. (2014) argued for an increase in annual maximum sea level during 1916–2005, especially in the northern Baltic Sea, and attributed the trends to changes in wind, these results were likely affected by the long-term internal variability. Furthermore, extreme sea levels in the Gulf of Finland, especially in Neva Bay, are very sensitive to the position of storm tracks (Suursaar and Sooäär, 2007).

3.2.5(5) Waves

Instrumental wave measurements in the Baltic Sea have been made since the 1970s, first as measurement campaigns and then, since the 1990s, as continuous monitoring (e.g. Broman et al., 2006; Tuomi et al., 2011). As the spatial coverage of wave measurements is still quite sparse, and there are long-term data from few locations only, wave hindcasts have become a valuable tool for estimating the Baltic Sea wave climate (e.g. Björkqvist et al., 2018). Lately, satellite altimeter measurements have been used to estimate the changes in the Baltic Sea wave climate (Kudryavtseva and Soomere, 2017).

Hindcast studies (e.g. Räämet and Soomere, 2010; Tuomi et al., 2011; Björkqvist et al., 2020) are in good agreement with wave measurements and estimate the annual mean significant wave height (SWH) in the open sea areas of the Baltic Sea at 0.5–1.5 m. Wave growth in the Baltic Sea is hampered by the shape and small size of the basins. The highest mean values are recorded in the Baltic proper, with the longest and widest fetches. The gulfs have less severe wave climates (Björkqvist et al., 2020). In addition, wave growth in the northern Baltic Sea in winter is limited by the seasonal ice cover, leading to considerably lower mean and maximum values of SWH in the northernmost Gulf of Bothnia and the easternmost Gulf of Finland.

Although the measurement and hindcast periods have been quite short so far, some studies have also analysed trends in the Baltic Sea SWH. For example, Soomere and Räämet (2011) and Kudryavtseva and Soomere (2017) suggested that the trend in SWH has been increasing since the 1990s, but the results are site specific and, so far, rather inconclusive.

The seasonal wave climate is driven by the wind climate. The highest mean and maximum values of SWH are reached in autumn and winter, while summer typically has the mildest wave climate. In subbasins with a long ice season and a large ice extent, such as the Bothnian Bay, the seasonal variation in the SWH is slightly different, since the waves are damped by the ice.

So far, the northern Baltic proper holds the record measured value of Baltic Sea SWH. In December 2004, a SWH of 8.2 m was measured by the northern Baltic proper wave buoy, with a highest individual wave of ca. 14 m (Tuomi et al., 2011; Björkqvist et al., 2018). As the spatial coverage of the wave measurements has increased and milder ice winters have also allowed late autumn and even winter measurements in the northern parts of the Baltic Sea, 8 m SWH has also been measured in the Bothnian Sea in January 2019. Björkqvist et al. (2020) estimated a return period of 104 years for this event in the present climate. Hindcast statistics have suggested that even higher maximum values between 9.5–10.5 m may occur in areas, and the times for which wave buoy measurements are not available (Soomere et al., 2008; Tuomi et al., 2011; Björkqvist et al., 2018).

3.2.5(6) Sedimentation and coastal erosion

The glacial isostatic adjustment and eustatic sea level change impose a first-order control on Baltic Sea coastal landscape change (Harff et al., 2017). In the subsiding southern Baltic Sea region, wind-driven coastal currents and waves are the major drivers for erosion and sedimentation, especially along the sandy and clayey sections of sandy beaches, dunes and soft moraine cliffs (Zhang et al., 2015; Harff et al., 2017).

Owing to spatial variation in aero- and hydrodynamic conditions (winds, waves, and alongshore currents) and underlying geological structure (lithology and sediment composition), a diversity of morphological patterns have developed along the Baltic Sea coast. Because of the dominant westerly winds that blow 60 % of the year (Zhang et al., 2011) and a sheltering by land in the west, wind waves are larger in the southeastern Baltic Sea than in the southwest. As a result, sediment transport and dune development are more active and dynamic along the southeastern coast. Thus, the largest coastal dunes are found along the Polish coast, with wave lengths of >100 m and heights of >20 m (Ludwig, 2016), while dunes along the German coast normally have wave lengths less than 60 m and heights below 6 m (e.g. Lampe and Lampe, 2018). Under conditions favourable for wind-driven sand accumulation along sandy Baltic Sea coasts, a typical cross-shore profile features one or several foredune ridges, generally with a height of between 3 and 12 m above the mean sea level (Zhang et al., 2015; Łabuz et al., 2018). At the backshore, behind the established foredune ridges, drifting or stabilised dunes in transgressive forms, mainly parabolic or barchanoid types, are commonly developed. The source of sediment for dune development includes fluvioglacial sands from eroded cliffs, river-discharged sands, and older eroded dunes (Łabuz, 2015).

Because the wind–wave energy increases from west to east, so does erosion along the Baltic Sea coast. The mean annual erosional rate of the soft moraine cliffs and sandy dunes along the northern side of the southwestern Baltic Sea coast (southern Sweden and Denmark) is 1–2 m yr⁻¹, which is larger than 0.4–1 m yr⁻¹ along the southern side of the southwestern Baltic Sea coast (Germany). Erosion along the southern Baltic Sea coast increases eastward, with a mean annual rate of 0.5–1.5 m yr⁻¹ in Poland and 0.5–4 m yr⁻¹ in Latvia, Lithuania, and Russia (BACC II Author Team, 2015). Severe coastal erosion in the Baltic Sea region is often caused by storms. The maximum storm-induced erosion increases eastward from 2–3 m yr⁻¹ at the southwestern Baltic Sea coast (southern Sweden, Denmark, and Germany) to 3–6 m yr⁻¹ along the Polish coast and ∼10 m yr⁻¹ along the coast of Lithuania and Russia (Kaliningrad). Each storm can erode soft Latvian cliffs 3–6 m, with a maximum of up to 20–30 m locally. Many sandy beaches along the Gulf of Finland have recently been severely damaged by frequent storm surges, despite extensive protective measures (BACC II Author Team, 2015).

The prevailing wind–wave pattern controls the spatial variations in not only coastline change rates but also submarine morphologies (Deng et al., 2019). In the southwestern Baltic Sea coast where wind–wave energy is relatively small, nearshore submarine morphology is generally featured by smooth transition from beach dunes/moraine cliffs to deeper water perturbed by one or two longshore bars. Morphological perturbations (e.g. the number and amplitude of longshore bars and rip current channels) become increasingly larger toward the east due to increased wind–wave energy. The wave incidence angle also impacts the nearshore submarine morphology, with, in general, a smaller angle leading to a larger morphological heterogeneity, i.e. a larger amplitude of perturbations. The amplitude of near-shore morphological perturbations may significantly affect coastal erosion because rip currents act as an efficient conduit for offshore sediment transport, despite the fact that they only occur sporadically along the Baltic Sea coast (Schönhofer and Dudkowska, 2021).

The sediments eroded from the soft moraine cliffs are composed of grain sizes from clay to pebbles. The fine-grained sediments are mostly transported outwards to the deeper seafloor (i.e. the Baltic Sea basins), either suspended in the water column or in a concentrated benthic fluffy layer (Emeis et al., 2002). Eroded fine-grained sediments from the moraine cliffs have been found to contribute to a major portion (40 %–70 %) of the Holocene deposits in the muddy Baltic Sea basins (Porz et al., 2021). Coarser material, such as sand, stays mostly nearshore, partly in the water and partly transported onto the beach and the dunes (Deng et al., 2014; Zhang et al., 2015).

3.2.5(7) Marine carbonate system and biogeochemistry

Studies summarised in BACC II showed that nitrate and phosphate concentrations in the winter surface water of the Baltic proper had increased by a factor of about 3 in the second half of the 20th century and reached a peak between 1980 and 1990. This change was consistent with the enhanced nutrient inputs to the Baltic Sea and caused eutrophication in the affected basin. Based on the available CO₂ system data, it has been estimated that the net ecosystem production in the Baltic Sea has increased since the 1930s by a factor of about 2.5. The increase in net ecosystem production and poor ventilation of the deepwater layers due to the permanent stratification of the water column caused significant expansion of the anoxic and hypoxic areas in the Baltic Sea. Since the 1980s, nutrient inputs to the Baltic Sea have decreased. This led to a decrease in winter surface water nitrate concentrations. No decrease was, however, observed for phosphate, due to the long residence time of P in the Baltic Sea and reduced P storage in oxygen-deficient sediments by binding to Fe-oxyhydroxides.

3.2.5(7.1) Oxygen and nutrients

In the Baltic Sea, hypoxia (oxygen deficiency) and even anoxia has expanded considerably since the first oxygen measurements in 1898 (Gustafsson et al., 2012; Carstensen et al., 2014a; Fig. 25). In 2016, the maximum hypoxia area was about 70 000 km², which is almost the combined area of Belgium and the Netherlands, whereas it was presumably very small or even absent 150 years ago (Carstensen et al., 2014a, b; Meier et al., 2019c, d). Hypoxia was caused mainly by increasing land nutrient inputs and atmospheric deposition that led to eutrophication of the Baltic Sea (Andersen et al., 2017; Savchuk, 2018). The impacts of other drivers like observed warming and eustatic sea level rise were smaller but still important (Carstensen et al., 2014a; Meier et al., 2019c). On annual to decadal timescales, halocline variations also had considerable influence on the hypoxic area (Conley et al., 2002; Väli et al., 2013).

Figure 25The extent of hypoxic (<2 mL O₂ L⁻¹) and anoxic (<0 mL O₂ L⁻¹) bottom water (in 10⁴ km²) in the Baltic proper, Gulf of Finland, and Gulf of Riga during regular cruises in August–October during 1960–2020. Source: https://www.smhi.se (last access: 5 March 2022) at the Swedish Meteorological and Hydrological Institute, Norrköping, Sweden.

Besides its detrimental effects on biota, hypoxia is responsible for the redox alterations of nitrogen and phosphorus integral stocks reaching in the Baltic proper hundreds of thousand tonnes annually. The dissolved inorganic nitrogen (DIN) pool is being depleted by denitrification, while the dissolved inorganic phosphorus (DIP) pool is increasing due to phosphate release in the water and sediment anoxic environments (e.g. Savchuk, 2013, 2018, and references therein). The resulting changes of nitrate and phosphate concentrations at the upper boundary of the halocline also affect neighbouring gulfs and export to the Bothnian Sea and Gulf of Finland waters with elevated phosphorus concentration (Rolff and Elfwing, 2015; Lehtoranta et al., 2017; Savchuk, 2018).

Despite the decrease in land nutrient inputs after the 1980s, the extent of hypoxia in the Baltic Sea remains unaltered. This is due to the long response time of the system to reductions in N and P inputs. According to recent computations, the residence times for TN and TP in the water and sediments of the Baltic Sea combined are 9 and 49 years, respectively (Gustafsson et al., 2017; Savchuk, 2018). Furthermore, recently observed oxygen consumption rates in the Baltic Sea are higher than earlier observed, counteracting the vertical oxygen supply and natural ventilation by oxygen-rich saltwater intrusions from the North Sea (Meier et al., 2018b). Although sediments are still the most important sinks of oxygen in the Baltic Sea, the increased rates of oxygen consumption were largely driven by water column processes with, for instance, bacterial nitrification as the most prominent process. Also, zooplankton and higher trophic level respiration were suggested to contribute more to oxygen consumption than 30 years ago (Meier et al., 2018b). However, the importance of the latter processes is still unknown. The present total oxygen consumption rate in the water column below 60 m depth in the Baltic proper, Gulf of Riga, and Gulf of Finland is estimated to be about 5 times that in the period 1850–1950 (Meier et al., 2018b).

Hypoxia also remains an important problem in the Baltic Sea coastal zone (Conley et al., 2011). Coastal hypoxia most often has an episodic or temporary character and is driven by the seasonal variations in organic matter supply, advective transport, and water column stratification (Carstensen and Conley, 2019). The latter is mostly caused by seasonal temperature changes, but changes in vertical stratification may, in some areas, be due to occasional inflows of saltier water and lower winds during summer. The coastal regions most affected by hypoxia are estuaries in the Danish straits and parts of Swedish and Finnish archipelagos located in the Baltic proper and Gulf of Bothnia (Conley et al., 2011). In contrast, hypoxia is rare along the southern and southeastern coastline (from Poland to Estonia), due to enhanced water circulation, as well as in the less productive coastal zone of the northern Baltic Sea. Despite the recently reduced nutrient inputs, bottom water oxygen concentrations have improved only in a few coastal ecosystems that have experienced the largest reductions (for instance, in the Stockholm Archipelago). In most of the 33 coastal sites evaluated by Caballero-Alfonso et al. (2015), oxygen conditions have deteriorated, especially along the Danish and Finnish coasts. This finding was explained as being a coupled effect of climate-change-related drivers, especially warming, which reduces oxygen solubility in water and strengthens thermal stratification, and a delay of the system in responding to nutrient reduction.

N and P are removed from the Baltic Sea by burial in sediments, but much N is also lost by denitrification (Gustafsson et al., 2017). Coastal regions constitute an efficient nutrient filter (Almroth-Rosell et al., 2016; Asmala et al., 2017) that remove about 16 % of N (by denitrification) and as much as 53 % of P (by burial) delivered from land (Asmala et al., 2017). The filter effect of the coastal zone is, however, highly diverse. Denitrification rates are highest in lagoons that receive large inputs of nitrate and labile organic material, while P is most efficiently buried in archipelagos (Carstensen et al., 2020). Additionally, Hoikkala et al. (2015) argued that dissolved organic matter (DOM) plays an important role for nutrient cycling in the Baltic Sea, since more than 25 % of bioavailable nutrients in riverine inputs and surface waters can be in organic form.

Furthermore, the exchange of nutrients between the coastal zone and the open sea (Eilola et al., 2012) and the role of MBIs for the phosphorus cycling (Eilola et al., 2014) were analysed. Eilola et al. (2014) concluded that the overall impact of MBIs on the annual uplift of nutrients from below the halocline to the surface waters is small because vertical transports are also comparably large during periods without MBIs. Instead, phosphorus released from the sediments between 60 and 100 m depth under anoxic conditions contributes to the eutrophication.

The cycling between nutrients and phytoplankton biomass was studied by Hieronymus et al. (2018), who found a regime shift between nutrient-limited phytoplankton variations before 1950 and a less nutrient-limited regime after 1950, with a larger impact of other variations such as those in water temperature.

3.2.5(7.2) Marine CO₂ system

The marine CO₂ system in the Baltic Sea is greatly influenced by the production and remineralisation of organic matter, as well as inputs of organic and inorganic carbon from land (Kuliński et al., 2017). The combination of all these factors makes the Baltic Sea pH and partial pressure of CO₂ (pCO₂) highly variable in space and time (Carstensen and Duarte, 2019). The Baltic Sea surface water is in almost permanent pCO₂ disequilibrium with the atmosphere throughout the year (Schneider and Müller, 2018). In spring and summer, the surface seawater is undersaturated with respect to atmospheric CO₂ as a consequence of biological production and the shallowing mixed layer depth. Thus, seawater pCO₂ typically has two minima corresponding to the spring bloom and the mid-summer nitrogen fixation period. In autumn and winter, pCO₂ increases due to the shifting balance between autotrophy and heterotrophy and the entrainment of deeper CO₂-rich waters.

Remineralisation of terrestrial organic matter plays an important role in shaping pCO₂ fields in the Baltic Sea. Kuliński et al. (2016) found that about 20 % of the dissolved organic carbon (DOC) delivered from the Vistula and Odra rivers is bioavailable, while Gustafsson et al. (2014) even estimated that 56 % of allochthonous (originating outside the Baltic Sea) DOC is remineralised in the Baltic Sea. High inputs of terrestrial organic matter that is subsequently partially remineralised in seawater turned the basins most affected by riverine runoff (Gulf of Bothnia, Gulf of Finland, and Gulf of Riga) to net CO₂ sources to the atmosphere in the period 1980–2005. This outgassing was more than compensated by the high CO₂ uptake in the open Baltic proper. In 1980–2005, the whole Baltic Sea was found to be, on average, a minor sink for atmospheric CO₂, absorbing 4.3±3.9 g C m⁻² yr⁻¹, with the rate of atmospheric CO₂ exchange highly sensitive to the inputs of terrestrial organic matter (Gustafsson et al., 2014).

The high seasonal variability in pCO₂, enhanced by eutrophication, causes large seasonal fluctuations in surface water pH, amounting to about 0.5 in the central Baltic Sea (Kuliński et al., 2017) and even more, often exceeding 1, in productive coastal ecosystems (Carstensen and Duarte, 2019; Stokowski et al., 2021). Furthermore, low total alkalinity (A_T; measures buffer capacity), prominent in the northern basins, makes the Baltic Sea potentially vulnerable to ocean acidification (OA), i.e. pH decrease caused by rising pCO₂ in the atmosphere and, thus, also in seawater. However, Müller et al. (2016) showed that A_T in the Baltic Sea has increased over time, which may partly be due to increasing inputs from land (Duarte et al., 2013). The highest trend, 7.0 µmol kg⁻¹ yr⁻¹, found in the Gulf of Bothnia, almost entirely mitigates the pH drop expected from rising pCO₂ in the atmosphere alone. In the southern Baltic Sea, the A_T increase is lower (3.4 µmol kg⁻¹ yr⁻¹) and reduces OA by about half. High seasonal pH variability, increasing A_T, and variable productivity imply that OA is not measurable in the central and northern Baltic Sea. In the Danish straits, where no A_T increase has been detected (Müller et al., 2016), a mean pH decrease of 0.004 years⁻¹ was identified in coastal waters in the period of 1972–2016 (Carstensen et al., 2018), which is approximately twice the ocean trend.

Recent studies showed that the Baltic Sea CO₂ system functions differently from the open ocean waters. These differences include a large CO₂ input from the remineralisation of terrestrial organic matter (Gustafsson et al., 2014), which is a considerable contribution by organic alkalinity (Kuliński et al., 2014; Ulfsbo et al., 2015; Hammer et al., 2017), A_T generation under hypoxic and anoxic conditions (Gustafsson et al., 2019; Łukawska-Matuszewska and Graca, 2018), and a borate alkalinity anomaly (Kuliński et al., 2018), making modelling a challenge. Due to the insufficient understanding of the processes involved, state-of-the-art biogeochemical models cannot yet reproduce the positive A_T trend in the Baltic Sea.

3.2.6 Marine biosphere

3.2.6(1) Pelagic habitats

3.2.6(1.1) Microbial communities

Microbial communities respond to increases in SST and river runoff that enhance metabolism and augment the amount of substrate available for bacteria. By using a long time series from 1994 to 2006, an increased input of riverine dissolved organic matter (DOM) in the Bothnian Bay and Bothnian Sea was shown to suppress phytoplankton biomass production and shift the carbon flow towards microbial heterotrophy (Wikner and Andersson, 2012; Paczkowska et al., 2020). Berner et al. (2018) presented further evidence of changes in marine microbial communities.

3.2.6(1.2) Phytoplankton and cyanobacteria

The phytoplankton growing season has become considerably prolonged in recent decades (Kahru et al., 2016; Groetsch et al., 2016; Hjerne et al., 2019; Wasmund et al., 2019). In the Baltic proper, the duration of the growing season, arbitrarily indicated by a threshold of 3 mg Chl m⁻³ of a satellite-derived chlorophyll, has doubled from approximately 110 d in 1998 to 220 d in 2013 (Kahru et al., 2016). In the western Baltic Sea, it now extends from February to December (Wasmund et al., 2019). Wasmund et al. (2019) analysed data on chlorophyll a and microscopically determined biomass from 1988–2017 and found an earlier start of the growing season, which correlated with a slight increase in sunshine duration in spring, and a later end to the growing season, which correlated with warmer water in autumn. The shifts in the spring and autumn blooms led to a prolongation of the summer biomass minimum. However, time series were rather short (30 years), and trends in irradiance might be caused by internal variability (see Sect. 3.2.1(3)). The spring phytoplankton communities have shifted from a preponderance of early-blooming diatoms to dominance by later-blooming dinoflagellates (Wasmund, 2017; Wasmund et al., 2017) and the autotrophic ciliate Mesodinium rubrum (Klais et al., 2011; Hällfors et al., 2013; Hjerne et al., 2019), perhaps due to reduced ice thickness and increased winter wind speed since the 1970s (Klais et al., 2013). Wasmund (2017) suggested that the decline in the ratio between diatom and dinoflagellate biomasses between 1984 and 1991 was caused by warmer winters. Confidence in these results is, however, low.

In summer, the amount of cyanobacteria has increased, and the phytoplankton biomass maximum, which in the 1980s was in spring, is now in July–August. This shift has been explained by a complex interaction between warming, eutrophication, and increased top-down pressure (Suikkanen et al., 2013). There are, however, different opinions concerning the relative effects of eutrophication and climate on changes in phytoplankton biomass and community composition. In the long-term data, results vary according to area and species group (Wasmund et al., 2011; Groetsch et al., 2016). Some studies saw evidence of eutrophication effects modified by climate-induced variations in temperature and salinity (Hällfors et al., 2013; Olofsson et al., 2020). Others found no explanation for the gradual change in community composition and concluded that the Baltic Sea phytoplankton community is not in a steady state (Olli et al., 2011; Griffiths et al., 2020).

Cyanobacteria accumulations derived from satellite data for 1979–2018 show both short-term (2–3-year) oscillations and decadal-scale variations (Kahru and Elmgren, 2014; Kahru et al., 2018; Kahru et al., 2020). Cyanobacteria accumulations in the Baltic proper were common in the 1970s and early 1980s but rare during 1985–1990. They increased again starting in 1991 and especially since 1998. In the 1980s, the annual chlorophyll maximum in the Baltic proper was caused by the spring diatom bloom but has in recent decades shifted to the summer cyanobacteria bloom in July; the timing of this bloom has also advanced by about 20 d, from the end to the beginning of July (Kahru et al., 2016).

In the Baltic proper, the hypoxia-induced decrease in N : P ratio and increase in phosphate pool left over in the surface layer after the spring bloom led to intensification of the vicious circle (Vahtera et al., 2007), further augmented by increasing water temperature, and resulted in a conspicuous expansion of the surface diazotrophic cyanobacteria accumulations covering, in the 21st century, 150 000–200 000 km² (e.g. Kahru and Elmgren, 2014; Savchuk, 2018, and references therein). Although mechanisms of the interannual oscillations of 2–3 years remain unexplained (Kahru et al., 2018), there is a strong correlation of the accumulations with hypoxia-related biogeochemical variables and water temperature at the decadal scale of 5–20 years (Kahru et al., 2020). In the Bothnian Sea, the decreased nitrogen import and increased phosphorus import from the Baltic proper has shifted the nutrient balance and made cyanobacteria accumulations a permanent feature (Kahru and Elmgren, 2014; Kuosa et al., 2017) and probably also increased production and sedimentation (Ahlgren et al., 2017; Kahru et al., 2018, 2020; Kuosa et al., 2017; Lehtoranta et al., 2017; Rolff and Elfwing, 2015; Savchuk, 2013; Vahtera et al., 2007).

Experimental evidence supports the idea that climate change can and will drive changes in the pelagic production from both ecological and evolutionary perspectives (Sommer et al., 2012; Hattich et al., 2022), and a thorough review of benthic–pelagic coupling in the Baltic Sea demonstrates ecosystem-wide consequences of altered pelagic primary production (Griffiths et al., 2017).

3.2.6(1.3) Zooplankton

Several studies have confirmed that marine copepod species have declined in abundance since the 1980s, while euryhaline or limnetic, often small species, have increased (Hänninen et al., 2015; Suikkanen et al., 2013; Kortsch et al., 2021). The observed decline of marine taxa has been linked to the reduction in surface water salinity since the 1980s (Vuorinen et al., 2015), whereas the increase in brackish water taxa has been positively influenced by the temperature increase, either directly or indirectly (Mäkinen et al., 2017). Small-scale effects on individual species may affect reproductive success and, hence, influence both populations and communities (Möller et al., 2015).

3.2.6(2) Benthic habitats

3.2.6(2.1) Macroalgae and vascular plants

Long-term changes in Baltic Sea macroalgae and charophytes have been attributed to changes in salinity, wind exposure, nutrient availability, and water transparency (Gubelit, 2015; Blindow et al., 2016; Rinne and Salovius-Laurén, 2020), and biotic interactions may also play a role (Haavisto and Jormalainen, 2014; Korpinen et al., 2007). The long-term decrease in water transparency from 1936 to 2017 has been estimated to have reduced seafloor areas in the northern Baltic Sea favourable for Fucus spp. by 45 % (Sahla et al., 2020). Overall, it is expected that climate change and its interaction with other environmental factors (e.g. eutrophication) will cause complex responses and influence carbon storage in both macroalgae and vascular plants in the Baltic Sea (Jonsson et al., 2018; Takolander et al., 2017; Röhr et al., 2016; Perry et al., 2019; Salo et al., 2020; Bobsien et al., 2021).

3.2.6(2.2) Zoobenthos

Soft-sediment benthic communities depend on variables that are influenced by climatic variability. On the southwestern coast of Finland, amphipods have been replaced by Baltic clam Limecola balthica and the invasive polychaetes of Marenzelleria spp., a change that is attributed to an increase in near-bottom temperature and fluctuations in salinity and oxygen (Rousi et al., 2013). Variations in zoobenthos in the Åland archipelago during 1983–2012 were associated with a salinity decline (Snickars et al., 2015), and the effects related to climate change have acted as drivers for the long-term progression of zoobenthic communities (Rousi et al., 2019; Weigel et al., 2015; Ehrnsten et al., 2020).

3.2.6(3) Non-indigenous species

Numerous non-indigenous species have gained a stronghold in the Baltic Sea ecosystem during the past few decades, and in many cases, these species have wider tolerance ranges than the native ones, thus making them highly competitive under changing climate including warmer, and possibly less saline, water, which is further impacted by other drivers such as eutrophication. The ecological impacts of these species may vary from filling vacant ecological niches to potentially outcompeting native species and, thus, influencing the entire food web structure and functioning (Weigel et al., 2015; Griffiths et al., 2017; Ojaveer et al., 2017).

3.2.6(4) Fish

Sprat and herring in the Baltic Sea are influenced by multiple factors, including fisheries, predation, food availability, and climatic variations. Sprat has benefited from the seawater warming (Voss et al., 2011). In 1990–2020, sprat populations were affected both by climate and top-down control, i.e. fisheries and predation by cod (Eero et al., 2016). In the 1980s, overfishing and a partly climate-change-induced decline in suitable spawning habitat (reproductive volume) interacted to drastically reduce the cod population (Hinrichsen et al., 2011; Casini et al., 2016), with cascade effects on its main prey, sprat and herring, and zooplankton (Casini et al., 2008).

The various effects of temperature and salinity on sprat and cod also resulted in a spatial mismatch between these species, which contributed to an increase in sprat stocks (Reusch et al., 2018). The freshening of the Baltic Sea surface water, with the associated decline in marine copepods (Hänninen et al., 2015), contributed to a halving of the weight-at-age of a 3-year-old herring, from 50–70 g in the late 1970s to 25–30 g in 2000s (Dippner et al., 2019).

Among coastal fish, pikeperch (Sander lucioperca) has recently expanded its distribution northwards along the coasts of the Bothnian Sea, apparently aided by warmer waters (Pekcan-Hekim et al., 2011). For many coastal piscivores (perch, pike, and pikeperch) and for cyprinids, coastal eutrophication is, however, equally or more important than climate (Bergström et al., 2016; Snickars et al., 2015). Long-term studies illustrate that it is hard to disentangle abiotic and biotic interactions, e.g. between fish and their food (benthos), and climate-related drivers thus appear significant on a multidecadal timescale across a large spatial scale (Törnroos et al., 2019).

3.2.6(5) Marine mammals

During the 20th century, the marine mammals of the Baltic Sea experienced large declines in abundance because of hunting of seals, by-catch of porpoises and exposure to harmful substances causing reduced fertility.

The breeding distributions of the ice-breeding seals in the Baltic Sea have evolved with ice coverage, with the seals breeding where and when ice optimal for breeding occurs. Breeding ringed seals need ice throughout their relatively long lactation period (>6 weeks) and also use ice as moulting habitat. Ringed seals prefer compact or consolidated pack ice as it provides cavities and snowdrifts suitable for the construction of the lairs (most importantly, the breeding lair; Sundqvist et al., 2012).

The implementation of specific management and protection measures has had a profoundly positive influence on the populations of several Baltic Sea mammal populations, particularly seals (Reusch et al., 2018). Reusch et al. (2018) attributed these changes also to reduced exposure to harmful substances and initial increases in overall fish stocks as a consequence of eutrophication, followed by a reduction in several commercial stocks due to overfishing.

Specific climate-change-related impacts on seals are difficult to detect, although reconstructions of historical habitats since the last glaciation have been attempted for some seal species (Ukkonen et al., 2014). Along with warmer winters, the availability of suitable breeding ice for ringed seals in the Bothnian Bay is decreasing (Sect. 3.2.4(4)). The breeding success of ringed seal was probably reduced by the exceptionally mild winter of 2007–2008 (Jüssi, 2012), and several similar or even milder ice winters have followed (Ilmatieteen laitos, 2020). The winters of 2019–2020, 2007–2008, and 2014–2015 are the mildest in the annual ice cover statistics for the Baltic Sea (Uotila et al., 2015). The southern breeding populations of the ringed seal in the Gulf of Finland, the Gulf of Riga, and Archipelago Sea are already facing the challenges of milder winters; the ice-covered area during the breeding season has been reduced, and overlying snow for breeding lairs has been absent from the southern areas of the ringed seal breeding range in most winters of the past decade (Ilmatieteen laitos, 2020). Thus, the only available breeding ice in the Gulf of Finland in 2020 was found very near St. Petersburg (Halkka, 2020).

For grey seals, the lower availability of suitable breeding ice in its core distribution area has led to more breeding on land in areas where drift ice used to be found (Jüssi et al., 2008). Grey seals are known to gather to breed in certain sea areas, regardless of the winter severity, so some land colonies may become overpopulated. As an example, in 2016, 3000 grey seal pups were born on three islets in the northern Gulf of Riga.

Flooding of seal haul-outs due to sea level rise will first occur in the southernmost Baltic Sea, where the relative sea level rise will be most rapid (EEA, 2019c), and haul-out sites are mainly low sand or shingle banks. In Kattegat, relative sea level rise is estimated to be lower, and while the haul-outs in the western part are low sand and shingle banks, similar to those in the southern Baltic Sea, haul-outs in eastern Kattegat are skerries with a higher profile. In the central and northern Baltic Sea, haul-outs are mainly skerries, and here, the relative sea level rise is estimated to be low or even negative in the 21st century (EEA, 2019c). Yet, in recent history, winter storm surges have been observed to flood grey seal breeding colonies and push limited ice with ringed seal pups onto shore in the Gulf of Riga and Pärnu Bay. In the southern Baltic Sea and western Kattegat, increasing sea levels may turn parts of larger islands or previously inhabited islands into suitable seal haul-outs, but this is hard to project and depends, among other things, on the future management and protection of such areas.

Sea levels in the southern Baltic Sea have been rising at up to 3 mm yr⁻¹ since the 1970s (EEA, 2019c), and the available haul-out areas have, thus, seen reductions already. However, during this time, the relevant harbour and grey seal populations have been recovering at high rates from past depletion (Häkönen et al., 2018), with no documented or suspected effects of rising sea levels published.

The only cetacean resident in the Baltic Sea, the harbour porpoise (Phocoena phocoena), is a widespread species and seems to be rather tolerant of different temperatures and habitats. There are harbour porpoise populations in the waters around Greenland and along the coast of the Iberian Peninsula, Morocco, West Sahara, and in the Black Sea. However, the Baltic proper harbour porpoise population has been shown to differ genetically (Lah et al., 2016) and morphologically (Galatius et al., 2012) from neighbouring populations, which may imply local adaptations of which we are currently unaware. Sea ice limits the range available to the Baltic proper harbour porpoise population, since they need to come to the surface to breathe every 1–5 min. Hence, a decreasing ice cover is likely to increase the available range for the population. However, a change in the prey community resulting from climate-change-related factors could potentially have serious effects on this critically endangered (Hammond et al., 2008) population of a small whale, which is dependent on constant access to prey (Wisniewska et al., 2016).

3.2.6(6) Waterbirds

The winter distribution of many waterbirds has extended northwards in response to the increase in temperature and the decreasing extent of sea ice cover. This can be observed as an overall increase in the winter abundance of waterbirds because part of the population of some species (mainly diving ducks) that formerly wintered further to the southwest now remain in the Baltic Sea (Pavón-Jordán et al., 2019). Many species show decreasing trends in abundance in the southern parts of their wintering ranges (typically in western and southern Europe) but increases near the northern edge of their distribution, typically in the Baltic Sea region (MacLean et al., 2008; Skov et al., 2011; Aarvak et al., 2013; Lehikoinen et al., 2013; Pavón-Jordán et al., 2015; Nilsson and Haas, 2016; Marchowski et al., 2017; Fox et al., 2019). Similar shifts are seen in species that traditionally wintered in the Baltic Sea, but currently show declining wintering numbers there, as part of the population now winters in the White, Barents, and Kara seas (Fox et al., 2019).

Although the community composition changes rapidly, the changes are not fast enough to track the thermal isocline shifts (Devictor et al., 2012; Gaget et al., 2021). How species respond to changes in winter temperature seems, however, to be highly species or group specific (Pavón-Jordán et al., 2019). Many species now winter closer to their breeding areas, shortening the migration distances (Lehikoinen et al., 2006; Rainio et al., 2006; Gunnarsson et al., 2012).

Mainly owing to milder spring temperatures and related effects on vegetation and prey, many waterbirds migrate earlier in spring (Rainio et al., 2006) and, hence, arrive earlier in the breeding area (Vähätalo et al., 2004); some also start breeding earlier (van der Jeugd et al., 2009). Delayed autumn migrations have also been noted, but their relation to climate change is less clear (Lehikoinen and Jaatinen, 2012).

Earlier loss of sea ice was found to improve the pre-breeding body condition of female common eiders, leading to increasing fledging success in offspring (Lehikoinen et al., 2006). On the other hand, algal blooms promoted by higher seawater temperature have, in some cases, caused low quality in bivalve prey for common eiders, leading more birds to skip breeding (Larsson et al., 2014). Warmer seawater in winter also increases the energy expenditure of mussels, thus directly reducing their quality as prey for eiders (Waldeck and Larsson, 2013).

Most Baltic Sea waterbird species are migratory and affected by climate change also outside the Baltic Sea region, in the Arctic (breeding season), and in southern Europe and western Africa (wintering; Fox et al., 2015). This is important, given that climate warming is most pronounced in the Arctic and northern Eurasia and also above average in southern Europe and northern Africa (Allen et al., 2018).

3.2.6(7) Marine food webs

The entire marine food web of the Baltic Sea has been greatly impacted by climate-change-related drivers that have altered the physical environment and the physiological tolerance limits of several species, by causing the micro-evolution of Baltic Sea species and by interactive effects of climate change with other environmental drivers, such as eutrophication and hypoxia/anoxia (Niiranen et al., 2013; Wikner and Andersson, 2012; Schmidt et al., 2020; Pecuchet et al., 2020).

Integrated approaches encompassing all of the ecosystem components discussed above are needed in order to understand and manage the links among large-scale and long-term climate effects. These are driven by synergistic interactions of climate-change-related physical and chemical drivers with other factors, such as eutrophication or large-scale fisheries, which complicate human adaptation to the changing marine ecosystem (Niiranen et al., 2013; Blenckner et al., 2015; Hyytiäinen et al., 2021; Stenseth et al., 2020; Bonsdorff, 2021).

3.3.1 Atmosphere

3.3.1(1) Large-scale atmospheric circulation

Continued greenhouse-gas-induced warming is the key driver for future climate change, and changes in atmospheric circulation are relatively less important (IPCC, 2021a). However, as the regional climate in the Baltic Sea region is strongly governed by the large-scale circulation of the atmosphere, it is important to also consider changes in it when assessing future regional climate (e.g. Kjellström et al., 2018).

In the future, the NAO is very likely to continue to exhibit large natural variations similar to those observed in the past. In response to global warming, it is likely to become slightly more positive on average (Knudsen et al., 2011). Trends in the intensity and persistence of blocking remain to be determined (IPCC, 2014b). The AMO is expected to be very sensitive even to weak global warming, thus shortening the timescale of its response and weakening its amplitude (Wu et al., 2018; Wu and Liu, 2020). This will likely reduce the decadal variability in SSTs in the Northern Hemisphere. Recent studies indicate a degree of decadal predictability for blocking and the NAO influenced by the AMO (Athanasiadis et al., 2020; Wills et al., 2018; Jackson et al., 2015).

3.3.1(2) Air temperature

Table 12 lists the air temperature changes over the Baltic Sea catchment area and the Baltic Sea calculated from an ensemble of regional coupled atmosphere–ocean simulations (Gröger et al., 2021b). Due to the ice/snow albedo feedback, warming is larger in winter than in summer, and the land is warming faster than the Baltic Sea (Fig. 26a). Due to its proximity to the Arctic, the Baltic Sea region, including both land and sea, is warming faster than the global mean figures (Sect. 1.5.1(1)). The surface air temperature increase is expected to be largest in the northern Baltic Sea region, especially in winter. These statements are true for both uncoupled atmosphere (Christensen et al., 2022) and coupled atmosphere–ocean regional climate simulations (Gröger et al., 2019, 2021b).

Figure 26(a) Ensemble mean 2 m air temperature change (degrees Celsius) between 1970–1999 and 2070–2099 for winter (December through February; upper panels) and summer (June through August; lower panels) under RCP2.6, RCP4.5, and RCP8.5. Panel (b) is the same as panel (a) but for precipitation change (percent). In total, eight different dynamically downscaled Earth system model (ESM) simulations are used. Data source: Gröger et al., 2021b.

For RCP2.6, the global annual mean surface air temperature change between 1976–2005 and 2069–2098, averaged over the simulations, is 1.0^∘C. The corresponding global changes for RCP4.5 and RCP8.5 are 1.9 and 3.5^∘C, respectively. Over land in the Baltic Sea region, the warming is larger in each of the three scenarios, amounting to 1.5, 2.6 and 4.3^∘C, respectively (Table 12). Over the Baltic Sea, the increase is slightly smaller than over land (1.4, 2.4 and 3.9^∘C, respectively) but still larger than the corresponding global mean increase in surface air temperature. The latter result was expected and found in coupled atmosphere–ocean scenario simulations (Table 12) but not in all atmosphere-only runs (Christensen et al., 2022).

Extreme air temperatures

Changes in daily minimum and maximum temperatures have similar spatial patterns to the mean air temperature changes, although with greater warming for minimum temperature (Fig. 27). Such a decrease in the diurnal temperature range can have a number of explanations. Based upon an ensemble of CMIP5 GCMs, Lindvall and Svensson (2015) suggested that increasing downwelling longwave radiation due to larger greenhouse gas concentrations increased cloudiness at high latitudes, and changes in the hydrological cycle and changes in shortwave incoming radiation might be an explanation for the decreased diurnal temperature range. In addition to these drivers, the difference in the diurnal temperature range in northern Europe is small in winter, implying that differences between daily maximum and minimum temperatures also substantially depend on synoptic-scale variability and air mass origin. As a consequence, reduced temperature variability on synoptic timescales, resulting from reduced temperature gradients between the Atlantic Ocean and Eurasia, may be a reason for the decreased diurnal temperature range.

Figure 27From left to right, daily temperature range (DTR; in degrees Celsius) during historical (1981–2010) and future (2071–2100) periods and changes in daily temperature range, daily minimum, and daily maximum temperatures in RCP8.5, according to EURO-CORDEX scenario simulations (Christensen et al., 2022). The upper and lower panels show winter (December to February – DJF) and summer (June to August – JJA) means, respectively. Data source: Earth System Grid Federation (ESGF) network, http://esgf-data.dkrz.de (last access: 5 March 2022); see Christensen et al. (2022).

The number of hot spells is projected to increase, in particular in the southern Baltic Sea region (Gröger et al., 2021b). In coupled atmosphere–ocean simulations, the strongest increases in the annual mean number of consecutive days of tropical nights and the annual maximum number of tropical nights (with temperature above 20 ^∘C all night) in the Baltic Sea region were projected to occur over the open sea (Gröger et al., 2021b). In contrast, projections of tropical nights with atmosphere-only models show no significant change (Gröger et al., 2021b; Meier et al., 2019a). Due to the sea ice/snow albedo feedback, the largest decline in the number of frost days was projected to occur over the northeastern Baltic Sea region, i.e. northern Scandinavia and adjacent northern Russia (Gröger et al., 2021b). In addition to the magnitude of the warming that is affected by the sea ice/snow albedo feedback, the baseline climate may also play a role. Further southwest, where the winters are milder, there are fewer frost days to start with and, therefore, less room for a further decrease in the future.

3.3.1(3) Solar radiation and cloudiness

There are a few studies on projected future solar radiation over Europe. Global climate models of the CMIP5 generation indicated an increase in surface solar radiation, which will be the highest over southern Europe and decrease towards the north, but still with a slight increase over the Baltic Sea (Bartók et al., 2017; Müller et al., 2019). The RCP scenarios even include a decrease in aerosol emissions, which enhances the warming and the increase in solar radiation in the global climate models.

However, some regional climate models instead showed a decrease in surface solar radiation over the Baltic Sea area, in winter, by about 10 % over most of the catchment (Bartók et al., 2017; Christensen et al., 2022). This change was largely attributed to increasing future cloud cover, due to a more zonal airflow, and was accompanied by increased winter precipitation. Thus, there are large differences in the modelled surface solar radiation between global and regional models (Bartók et al., 2017). Unknown future aerosol emissions add to the uncertainty. In most RCM scenario simulations, the effect of changing aerosol concentrations was not considered (Boé et al., 2020a).

Global mean energy balance components have improved with every new climate model generation. For the latest CMIP6, models show good agreement for clear-sky shortwave energy fluxes in today's climate, both between models and between models and reference data (Wild et al., 2021). However, there are still substantial discrepancies among the various CMIP6 models in their representation of several of the global annual mean energy balance components, and the intermodel spread increases further on regional, seasonal, and diurnal scales. Thus, future changes in solar radiation and cloudiness remain highly uncertain, not least on the regional scale.

3.3.1(4) Precipitation

Precipitation in winter and spring is projected to increase over the entire Baltic Sea catchment, while summer precipitation is projected to increase in the northern half of the basin only (Christensen et al., 2022). In the south, summer precipitation is projected to change very little, although with a large spread between different models including both increases and decreases. The projected increase in the north is a rather robust feature among the regional climate models but with a large spread in the amount. Ensemble mean precipitation changes from coupled atmosphere–ocean simulations are summarised in Table 13. For the Baltic Sea catchment area, projected annual mean precipitation changes for the three RCP scenarios amount to 5 %, 9 %, and 15 % (Table 13) and are much larger than global averages (Sect. 1.5.1(2)). Over the Baltic Sea, the changes are similar to those over the land area (6 %, 8 %, and 16 %).

As expressed by the Clausius–Clapeyron equation, warming increases the potential for extreme precipitation due to intensification of the hydrological cycle associated with the growth of atmospheric moisture content. For northern Europe, regional climate models indicate an overall increase in the frequency and intensity of heavy precipitation events in all seasons (Christensen and Kjellström, 2018; Rajczak and Schär, 2017; Christensen et al., 2022, and references therein). The largest increase in the number of high precipitation days is projected for autumn. Changes in more extreme events, like 10-, 20-, or 50-year events, are less certain.

Changes in dry spells are another feature of an intensified hydrological cycle. Coppola et al. (2021) found an increasing number of consecutive dry days without precipitation for countries south of the Baltic Sea, while no changes were reported in the north.

3.3.1(5) Wind

In general, projected changes in wind speed over the Baltic Sea region are not robust among ESMs (Kjellström et al., 2018; Gröger et al., 2021b). However, Ruosteenoja et al. (2019) found, in CMIP5 projections, a slight but significant wind speed increase in autumn and a decrease in spring over Europe and the North Atlantic. Furthermore, over sea areas where the ice cover is projected to diminish on average, such as the Bothnian Sea and the eastern Gulf of Finland, the mean wind is projected to increase systematically because of a warmer sea surface and reduced stability of the planetary boundary layer (Meier et al., 2011c; Gröger et al., 2021b; Räisänen, 2017).

Projections of the future behaviour of extratropical cyclones are inherently uncertain because changes in several drivers result in opposite effects on cyclone activity. With global warming, the lower troposphere temperature gradient between low and high latitudes decreases due to polar amplification. Near the tropopause and in the lower stratosphere, the opposite is true, thus implying competing lower and upper tropospheric effects on changes in baroclinicity (Grise and Polvani, 2014; Shaw et al., 2016; Stendel et al., 2021). An increase in water vapour enhances diabatic heating and tends to increase the intensity of extratropical cyclones (Willison et al., 2015; Shaw et al., 2016) and contribute to their propagation farther poleward (Tamarin-Brodsky and Kaspi, 2017; Tamarin and Kaspi, 2017). The opposite is true in parts of the North Atlantic region, e.g. south of Greenland. For this region the north–south temperature gradient is increasing, as the weakest warming in the entire Northern Hemisphere is over ocean areas south of Greenland. North of this local minimum, the opposite is true. The increase in the north–south gradient over the North Atlantic may be responsible for some ESMs showing an intensification of the low pressure activity and, thereby, higher wind speeds over a region from the British Isles and through parts of north-central Europe in winter (Leckebusch and Ulbrich, 2004; Ulbrich et al., 2008). These projections have been confirmed by Harvey et al. (2012). They compared the ensemble storm track response of CMIP3 and CMIP5 model simulations and found that both projections show an increase in storm activity in the midlatitudes, with a smaller spread in the CMIP5 simulations. In contrast to CMIP3, the CMIP5 ensemble showed a significant decrease in cyclone track density north of 60^∘ N. Hence, pre-CMIP3 and CMIP3 studies showed a clear poleward shift of the North Atlantic storm track (e.g. Fischer-Bruns et al., 2005; Yin, 2005; Bengtsson et al., 2009), whereas the CMIP5 ensemble predicts only an eastward extension of the North Atlantic storm track (Zappa et al., 2013). The newest generation of models from CMIP6 resulted in significant reduction in biases in storm track representation compared to CMIP3 and CMIP5, but the response to climate change is quite similar compared to the previous assessments (Harvey et al., 2020). The eastward extension of the North Atlantic storm track seems to be a robust result as it is found in pre-CMIP3, CMIP3, and CMIP5 simulations (Feser et al., 2015). The response of CMIP6 models is similar to CMIP5 models, but it is considerably larger, probably due to the larger climate sensitivity in the CMIP6 models (Harvey et al., 2020). Furthermore, based on idealised simulations, Sinclair et al. (2020) found that warming would only lead to a slight decrease in the number and no change in the median intensity or lifetime of extratropical cyclones.

In summary, there is no clear consensus among climate change projections in how changes in the frequency and/or intensity of extratropical cyclones will affect the Baltic Sea region (Räisänen, 2017). However, in future climate, the frequency of severe wind gusts in summer associated with thunderstorms may increase (Rädler et al., 2019).

3.3.1(6) Air pollution, air quality, and atmospheric nutrient deposition

The main conclusions of the BACC II Author Team (2015) concerning projections of air quality in the Baltic Sea region still hold. The main factor determining future air quality in the region is regional emissions of air pollutants and not changes in meteorological factors related to climate change or in intercontinental pollution transport (e.g. Langner et al., 2012; Hedegaard et al., 2013).

Recent post-BACC II air quality modelling studies for the Baltic Sea area are Colette et al. (2013, 2015), Varotsos et al. (2013), Hendriks et al. (2016), and Watson et al. (2016). They concentrated mainly on particulate matter (PM) and ground-level ozone (O₃), which are the pollutants most likely to be affected by changing climate parameters. These studies agree with current-day air quality trends in that the Baltic Sea region, in general, is less exposed to air pollution than the rest of Europe.

Jacob and Winner (2009) showed that climate change is likely to increase ground-level ozone in central and southern Europe. In a meta-analysis, Colette et al. (2015) assessed the significance and robustness of the impact of climate change on European ground-level ozone based on 25 model projections, including some driven by SRES (Special Report on Emission Scenarios; Nakićenović et al., 2000) and RCP scenarios. They indicated that an increase in ground-level ozone is not expected for the Baltic Sea region. A latitudinal gradient was found from an increase in large parts of continental Europe (+5 ppbv – parts per billion by volume) but a small decrease over Scandinavia (up to −1 ppbv). Studies that explicitly compared the magnitude of projected climate and anthropogenic emission changes (Langner et al., 2012; Colette et al., 2013; Varotsos et al., 2013) all confirmed that changes in emission of ozone precursors (NO_x and VOCs) had the larger effect. For northern Europe, Varotsos et al. (2013) estimated that reductions in snow cover and solar radiation in a SRES A1B scenario lead to an ozone decrease of about 2 ppbv by 2050, as compared to present conditions.

Varotsos et al. (2013) stressed the importance of future biogenic isoprene emissions for ozone concentrations. In the 2050 climate, increases in ozone concentrations will be associated with increased biogenic isoprene emissions due to increased temperatures, whereas increased water vapour over the sea, and increased wind speeds, are associated with decreases. Hendriks et al. (2016) emphasised that isoprene emissions may increase significantly in coming decades if short-rotation coppice plantations are greatly expanded to meet the increased biofuel demand resulting from the EU decarbonisation targets. They investigated the competing effects of anticipated trends in land use, anthropogenic emissions of ozone precursors, and climate change on European ground-level ozone concentrations and related health and environmental effects by 2050. They found that increased ozone concentrations and associated health damage caused by a warming climate (+2 to 5 ^∘C across Europe in summer) might be more than the reduction that can be achieved by cutting emissions of anthropogenic ozone precursors in Europe. This is an example for the dominance of the climate change effect over a pollutant emission reduction effect, which is an exception to the statement made at the beginning of this section.

Orru et al. (2013, 2019) and Geels et al. (2015) studied the effect of climate change on ozone-related mortality in Europe. Orru et al. (2019) presented their results on a country level, including all Baltic Sea EU countries. They concluded that, although mortality related to ground-level ozone is projected to be lower in the future (mainly due to decreased precursor emissions), the reduction could have been larger without climate change and an increasingly susceptible population.

In parts of the Baltic Sea region, considerable air pollution is due to shipping. Ship traffic in the region is projected to increase over the coming decades, which could lead to larger emissions (i.e. NO_x and PM) than today unless stricter air quality regulations counter this potential trend. For the Baltic Sea, a nitrogen emission control area (NECA) will become effective in 2021. Karl et al. (2019a) designed future scenarios to study the effect of current and planned regulations of ship emissions and the expected fuel efficiency development on air quality in the Baltic Sea region. They showed that, in a business-as-usual scenario for 2040 (SECA-0.1 % and fuel efficiency regulation effective starting in 2015), the introduction of the NECA will reduce NO_x emissions from ship traffic in the Baltic Sea by about 80 % in 2040. The reduction in NO_x emissions from shipping translates to a ∼60 % decrease in NO₂ summer mean concentrations in a wide corridor around the ship routes. The coastal population of northern Germany, Denmark, and western Sweden will be exposed to less NO₂ in 2040 due to the introduction of the NECA. With lower atmospheric NO_x levels, less ozone will be formed, and the estimated daily maximum O₃ concentration over the Baltic Sea in summer 2040 will, on average, be 6 % lower than without the NECA. Compared to today, the introduction of the NECA will also reduce ship-related PM_2.5 emissions by 72 % by 2040 (compared to −48 % without the NECA). Simulated nitrogen deposition on the Baltic Sea decreases by 40 %–44 %, on average, between 2012 and 2040. A similar study by Jonson et al. (2019) estimated that the contributions of Baltic Sea shipping to NO₂ and PM_2.5 concentrations, and to the deposition of nitrogen, will be reduced by 40 %–50 % from 2016 to 2030, mainly as a result of NECA.

The pollutant concentrations reported in this section may drop to as yet unknown lower values if the shipping sector is successful in meeting the International Maritime Organization (IMO) target in greenhouse gas emission reduction of 50 % by 2050. This reduction is only possible if low carbon alternative fuels are introduced. Employing a high-energy efficiency, as it is already considered in scenarios used by Karl et al. (2019a), will not be sufficient. The new fuels will also lead to altered emissions of pollutants.

3.3.2 Land

3.3.2(1) River discharge

Climate change may have an influence on the seasonal river flow regime as a direct response to changes in air temperature, precipitation, and evapotranspiration (BACC II Author Team, 2015; Blöschl et al., 2017).

For areas in the northern Baltic Sea region presently characterised by spring floods due to snowmelt, the floods are likely to occur earlier in the year, and their magnitude is likely to decrease owing to less snowfall, shorter snow accumulation period, and repeated melting during winter. As a consequence, sediment transport and the risk of inundation are likely to decrease.

In the southern part of the Baltic Sea region, increasing winter precipitation is projected to result in increased river discharge in winter. In addition, groundwater recharge is projected to increase in areas where infiltration capacity is not currently exceeded, resulting in higher groundwater levels. Decreasing precipitation combined with rising temperature and evapotranspiration during summer is projected to result in the drying of the root zone, increasing demands for irrigation in the southern Baltic Sea region.

Projections with a process-oriented hydrological model suggested that, under the RCP4.5 and RCP8.5 scenarios, the total river flow during 2069–2098 relative to 1976–2005 will increase 1 %–21 % and 6 %–20 %, respectively, illustrating the large inherent uncertainty in hydrological projections (Saraiva et al., 2019a). According to these and previous projections, the increase in river flow will mainly take place in the north, while total river flow to the south will decrease (Stonevičius et al., 2017; Šarauskienė et al., 2017). Winter flow will increase due to intermittent melting (Stonevičius et al., 2017). Projected discharge changes attributed to increasing air temperature are reflected in observed trends (Sect. 3.2.2(1)), whereas changes attributed to increasing precipitation are not necessarily so (Wilson et al., 2010).

Since the publication of BACC II (BACC II Author Team, 2015), ensemble sizes of scenario simulations with hydrological models have increased, enabling the estimate of uncertainties in projections (e.g. Roudier et al., 2016; Donnelly et al., 2017). Donnelly et al. (2014) focused on projecting changes in discharge to the Baltic Sea by using a semi-distributed conceptual hydrological model for the drainage basin (Balt-HYPE) combined with a small ensemble of climate projections under the SRES A1B and A2 scenarios. Results showed an increased overall discharge to the Baltic Sea, with a seasonal shift towards higher winter and lower summer flows and diminished seasonal snowmelt peaks. Efforts were made to assess the uncertainty in the model chain, and the simulated changes were not larger in magnitude than the ensemble spread, highlighting the importance of such uncertainty assessments in effect studies to frame the quantitative model results.

Arheimer and Lindström (2015) studied future changes in annual maximum and minimum daily flows. Their projections suggested that snow-driven spring floods in the northern central part of Sweden may occur about 1 month earlier than today, and rain-driven floods in the southern part of Sweden may become more frequent. The boundary between the two flood regimes is projected to shift northward.

Past observations (see Sect. 3.2.2(1)) and future projections (e.g. Graham, 2004) suggest a temporal shift in the seasonality of the river discharge, with decreasing flow in spring/summer and increasing flow in winter. Global warming and river regulation due to hydropower production cause similar changes. However, in snow-fed rivers globally, the impact of climate change is projected to be minor compared to river regulation (Arheimer et al., 2017).

3.3.2(2) Land nutrient inputs

Projected changes in riverine discharge and nutrient inputs from the Baltic Sea drainage basin to coastal waters have been studied using a number of modelling frameworks in recent years. Projecting the regional effects of future climate and environmental change on hydrology and nutrient turnover poses challenges in terms of (i) the complex nature of the modelled system, including human influence on riverine nutrient inputs and transport processes alike, which necessitates long projection model chains and leads to uncertainty in modelled hydrologically driven responses, and (ii) the significance of changes in human behaviour, e.g. in terms of land management, population, or nutrient emissions from point sources, which adds complexity to the formulation of scenarios for future change on top of the climate change signal. Hydrological impact studies in the catchment area (and elsewhere), therefore, often explicitly use simplifying assumptions in order to reduce the complexity of the modelled system and to focus on certain aspects of impacts of projected changes.

Hesse et al. (2015) reported increasing discharges in a model study of the Vistula lagoon catchment, using a hydrological model (SWIM), which also allows for nutrient load assessment, and climate change impact modelling based on a climate model ensemble. On average, results showed decreasing trends for nitrogen and phosphorus inputs, but a wide range of projections with individual ensemble members.

Hägg et al. (2014) used a split-model approach to project changes in TN and TP inputs to Baltic Sea subbasins. Changes in discharge were estimated with a hydrological model (catchment simulation software – CSIM) combined with a climate projection ensemble, which sampled a range of climate model and emission scenario combinations. Inputs were then calculated with a statistical model, based on modelled discharges and population as a proxy for human nutrient emissions, combining population change assumptions with climate projections. Results showed a general trend towards higher nutrient inputs across the region as a result of climate change, and a significant (i.e. potentially trend changing) influence of human nutrient input reduction scenarios, particularly in the southern half of the Baltic Sea catchment basin.

Øygarden et al. (2014) used measurements in a number of small agricultural catchments to establish functional relationships between precipitation, runoff, and nitrogen losses from agricultural land and qualitatively related their findings to projected precipitation changes across the Baltic Sea drainage basin under climate change scenarios and to anthropogenic measures to counteract the climate-driven effects on nutrients. The analyses showed a positive relationship between runoff and nitrogen losses, as well as between rainfall intensity and nitrogen losses, but stressed the wide range of feedback loops possible between climate change effects and anthropogenic measures through management or policy changes. Such data-driven approaches avoid uncertainties related to effect model chains at the expense of direct basin-wide quantitative effect projections.

The potential effects of nutrient input abatement scenarios under climate change conditions were investigated by Huttunen et al. (2015) in a study of Finnish catchments draining to the Baltic Sea. A national nutrient load model (VEMALA) was combined with a mini-ensemble of climate models, and then a number of agricultural scenarios were derived, based on crop yield and policy changes, and an economic model (DREMFIA) was used to translate scenario assumptions to changes in the nutrient load model for evaluation of effects. On average, increased precipitation led to increased annual discharge and a shift from spring to winter peaks, with TN and TP inputs increasing with the discharge. Here, adaptation scenarios had less of an effect than climate change, with some regional variation, but significantly different load reductions were found among assessed adaptation strategies, leading to the conclusion that adaptation measures are important for overall climate change effect mitigation in the region.

The relative importance of management decisions for TN and TP load effects was studied also by Bartosova et al. (2019), using a hydrological model (E-HYPE) applied for the entire Baltic Sea drainage basin. The ensemble approach combined climate and socioeconomic pathways based on IPCC Fifth Assessment data (Zandersen et al., 2019), where socioeconomic changes were directly translated into changes of the effect model set-up. The influence of nutrient input abatement strategies were shown to be in the same magnitude range as climate effects, thus indicating the importance of effective nutrient input mitigation strategies for the region. In order to increase this efficiency, Refsgaard et al. (2019) developed and explored the concept of spatially differentiated measures for TN load reductions in the Baltic Sea drainage basin, based on the realisation that measures are not uniformly efficient over large areas and should, therefore, not be uniformly applied either.

3.3.3 Terrestrial biosphere

In the following, we focus on the European drought in 2018 to study the impact of very warm conditions on the terrestrial ecosystem and on projections for the terrestrial ecosystems in the Arctic because of the particularly strong climate warming in the Arctic and potentially strong feedbacks from the release of CO₂ and CH₄ in the northernmost part of the Baltic Sea region. Finally, we discuss mitigation scenarios for land use and land cover changes associated with the Paris Agreement.

Terrestrial ecosystems in the European drought year 2018

The summer of 2018 saw extremely anomalous weather conditions over Europe, with high temperatures everywhere, and low precipitation and high incoming radiation in western, central, and northern Europe (Peters et al., 2020). These extreme weather conditions resulted in severe drought (indicated by soil moisture anomalies) in western, central, and northern Europe, including the entire Baltic Sea region. The impacts of the severe drought and heatwave in Europe in 2018 were investigated in a series of papers, ranging from individual sites to the continental scale (Peters et al., 2020).

Graf et al. (2020) studied the effects of the 2018 drought conditions on the annual energy balance at the land surface, in particular the balance between sensible and latent heat fluxes, across different terrestrial ecosystems at various sites in Europe. Graf et al. (2020) found a 9 % higher incoming solar radiation compared to their reference period across the drought-affected sites. The outgoing shortwave radiation mostly followed the incoming radiation, with an increase of 11.5 %, indicating a small increase in surface albedo. The incoming longwave radiation, on the other hand, did not change significantly, indicating that the effects of higher atmospheric temperatures and reduced cloudiness cancelled out, while outgoing longwave radiation increased by 1.3 % as a result of higher land surface temperatures. Overall, the net radiation increased by 6.3 % due to the extreme drought conditions. As for the non-radiative surface energy fluxes, the sensible heat flux showed a strong increase by 32 %, while the latent heat fluxes did not change significantly on average. Graf et al. (2020) attributed the negligible effect on latent heat fluxes to the opposing roles of increased grass reference evapotranspiration on the one hand and soil water depletion, stomatal closure, and plant development on the other. Evapotranspiration increased where and when sufficient water was available and later decreased only where stored soil water was depleted. As a consequence, latent heat fluxes typically decreased at sites with a severe precipitation deficit but often increased at sites with a comparable surplus of grass reference evapotranspiration but only a moderate precipitation deficit. Consistent with this, peatlands were identified as the only ecosystem with very strong increases in latent heat fluxes but insignificant changes in sensible heat fluxes under drought conditions. Crop sites, on the other hand, showed significant decreases in latent heat fluxes.

Lindroth et al. (2020) analysed the impact of the drought on Scandinavian forests, based on 11 forest ecosystem sites differing in species composition, i.e. spruce, pine, mixed, and deciduous. Compared to their reference year, in 2018 the forest ecosystem showed a slight decrease in evaporation at two of the sites, was nearly unchanged at most sites, and increased at two sites with pine forest. At the same time, the mean surface conductance during the growing season was reduced by 40 %–60 %, and the evaporative demand increased by 15 %–65 % due to the warm and dry weather conditions. The annual net ecosystem productivity (NEP) decreased at most sites, but the reasons differed. At some sites, the NEP decrease was due to an increase in ecosystem respiration (RE), while at others both RE and the gross primary productivity (GPP) decreased, with the decrease in GPP exceeding that in RE. At six sites, the annual NEP decreased by over 50 g C m⁻² yr⁻¹ in 2018. Across all sites considered, NEP anomalies varied from −389 to +74 g C m⁻² yr⁻¹. A multilinear regression analysis revealed that the anomalous NEP could, to a very large extent (93 %), be explained by anomalous heterotrophic respiration and reduced precipitation, with most of the variation (77 %) due to the heterotrophic component.

Rinne et al. (2020) studied the effects of the drought on greenhouse gas exchange in five northern mire ecosystems in Sweden and Finland. Due to low precipitation and high temperatures, the water table sank in most of the mires. This led not only to a lower CO₂ uptake but also to lower CH₄ emissions by the ecosystems. In total, three out of the five mires switched from sinks to sources of CO₂. Estimates of the radiative forcing expected from the drought-related changes in greenhouse gas fluxes indicated an initial cooling effect due to the reduced CH₄ emissions, lasting up to several decades, followed by a warming caused by the lower CO₂ uptake. However, it is unknown whether these results can be generalised to all wetlands of the Baltic Sea region.

Terrestrial ecosystems in the Arctic region

Climate warming has been particularly strong at high northern latitudes, and climate change projections indicate that this trend will continue, due to the anticipated increase in anthropogenic climate forcing. This strong warming is expected to have major consequences for terrestrial ecosystems in Arctic and sub-Arctic regions.

Zhang et al. (2013) used the Arctic version of a dynamic global vegetation model (LPJ-GUESS, Smith et al., 2001), forced with a regionalised climate scenario (A1B anthropogenic emission scenario), to investigate land surface feedbacks from vegetation shifts and biogeochemical cycling in terrestrial ecosystems under future climate warming. They found marked changes in vegetation by the second half of the 21st century (2051–2080), i.e. a poleward advance of the boundary between forests and tundra, expansion of tundra covered with tall shrubs, and a shift from deciduous trees, e.g. birch, to evergreen boreal coniferous forest. These changes in vegetation were associated with decreases in surface albedo, particularly in winter due to the snow-masking effect, and with increases in evapotranspiration. The reduced surface albedo would tend to enhance the projected warming (positive feedback), while increased evapotranspiration would dampen it (negative effect). The terrestrial ecosystems continued to act as carbon sinks during the 21st century but at diminished rates in the second half of the century. The initial increase in carbon sequestration, due to a longer growing season and CO₂ fertilisation, could be reduced and eventually reversed by increased soil respiration and greater CO₂ release from increased wildfires. Peatlands were identified as hotspots of CH₄ release, which would further enhance the projected warming (positive feedback).

Using a regional ESM (RCA-GUESS; Smith et al., 2011) over the Arctic region, W. Zhang et al. (2014) investigated the role that the biophysical effects of the projected future changes of the land surface play for the terrestrial carbon sink in the Arctic region under a future climate scenario based on a high emission scenario (RCP8.5). In total, two simulations were performed to determine the role of the biophysical interactions, i.e. one with and one without the biophysical feedbacks resulting from the simulated climatic changes to the terrestrial ecosystems in the model. In both simulations, the Arctic terrestrial ecosystems continued to sequester carbon until the 2060–2070s, after which they were projected to turn into weak sources of carbon, due to increased soil respiration and biomass burning. The biophysical effects were found to markedly enhance the terrestrial ecosystem carbon sink, particularly in the tundra areas. These two opposing feedback mechanisms, mediated by changes in surface albedo and evapotranspiration, contributed to the additional carbon sequestration. The decreased surface albedo in winter and spring notably amplified warming in spring (positive feedback), while the increased evapotranspiration led to a marked cooling during summer (negative feedback). These feedbacks stimulated vegetation growth due to an earlier start of the growing season, leading to changes in woody plant species and the distribution of vegetation. In a later study, Zhang et al. (2018) found that these biophysical feedbacks play essential roles also in climate scenario simulations with weaker anthropogenic climate forcing.

Mitigation

The beneficial effects of carbon sequestration by forest ecosystems on climate change may be reinforced, counteracted, or even offset by management-induced changes in surface albedo, land surface roughness, emissions of biogenic volatile compounds, transpiration, and sensible heat flux (see above). Luyssaert et al. (2018) investigated the trade-offs associated with using European forests to meet the climate objectives in the Paris Agreement. The authors argued that a more comprehensive assessment of forest management as a strategy to achieve the goals of the Paris Agreement should go beyond the reduction in atmospheric CO₂ and, thus, of the radiative imbalance at the top of the atmosphere. They suggested two additional targets, namely that forest management should neither increase the near-surface temperature nor decrease precipitation because climate effects arising from the changes in the terrestrial biosphere would make adaptation to climate change more demanding. Analysing different forest management portfolios in Europe designed to maximise the carbon sink, maximise the forest albedo, or reduce near-surface temperatures, Luyssaert et al. (2018) found that only the portfolio designed to reduce near-surface temperatures accomplished two of the objectives, i.e. to dampen the rise in atmospheric CO₂ and to reduce near-surface temperatures. This portfolio featured a decrease in the area of coniferous forest in favour of a considerable increase in the area of deciduous forest in northern Europe (from 130 000 to 480 000 km²).

3.3.4 Cryosphere

3.3.4(1) Snow

Generally, the relative decrease in snow amount is projected to be smaller in the colder northern and eastern parts of the Baltic Sea region than in the milder southern and western parts (e.g. Räisänen, 2021). There is agreement among the EURO-CORDEX RCMs that the average amount of snow accumulated in winter will decrease by around 50 % for land areas north of 60^∘ N for the RCP8.5 scenario by 2071–2100, relative to 1981–2010 (Christensen et al., 2022). South of 60^∘ N, the corresponding decrease is almost 80 %. The reduction in snow amount is slightly larger than in the figures presented by the BACC II Author Team (2015), which is consistent with larger average warming and, in the northern part of the area, a smaller precipitation increase projected in the RCP8.5 scenario compared to the SRES A1B scenario analysed by the BACC II Author Team (2015).

For Poland, two additional downscaling experiments were made to produce reliable high-resolution climate projections of precipitation and temperature using the RCP4.5 and RCP8.5 scenarios (Szwed et al., 2019). The results were used as input to a snow model (seNorge) to transform bias-adjusted daily temperature and precipitation into daily snow conditions. The snow model projected future snow depth to decrease in autumn, winter, and spring in both the near and far future. The maximum snow depth was projected to decrease 15 %–25 % by 2021–2050 in both scenarios. By 2071–2100, decreases under RCP4.5 and RCP8.5 were estimated to be 18 %–34 % and 44 %–60 %, respectively (Szwed et al., 2019).

3.3.4(2) Glaciers

The Special Report on the Ocean and Cryosphere in a Changing Climate (IPCC, 2019a) provided an assessment of future projected glacier mass reduction under various RCPs and treated Scandinavian glaciers separately (Hock et al., 2019). Previous projections, summarised in the Fifth Assessment Report of the IPCC (IPCC, 2014a; Vaughan et al., 2014), did not specifically focus on Scandinavian glaciers.

By 2100, likely (i.e. with a likelihood of 66 %–100 %) mass losses for high mountain glaciers are 22 %–44 % (RCP2.6) to 37 %–57 % (RCP8.5) of their mass in 2015. These losses exceed global projections for glacier mass loss of 18±7 % for RCP2.6 and 36±11 % for RCP8.5 (likely ranges; IPCC, 2019a). Glaciers in Scandinavia will lose more than 80 % of their current mass by 2100 under RCP8.5 (medium confidence), and many are projected to disappear, regardless of future emission scenarios (Fig. 28). Furthermore, river runoff from glaciers is projected to change regardless of emission scenario (high confidence) and to result in increased average winter runoff (high confidence) and in earlier spring peaks (high confidence; Hock et al., 2019).

Figure 28Mean projected glacier mass evolution between 2015 and 2100, relative to each region's glacier mass in 2015 (in percent) and ±1 standard deviation under RCP2.6, RCP4.5, and RCP8.5. Source: Hock et al., 2019; see their figure CB6.1.

Projections of future glacier mass loss depend crucially on climate projections providing surface air temperature and precipitation as forcing factors in process-based glacier models. For high mountain glaciers, such as those along the Scandinavian mountains that drain into the Baltic Sea, this is challenging, as the interplay of regional effects such as high mountain meteorology and elevation-dependent warming (Wang et al., 2016; Qixiang et al., 2018) with global climate is poorly understood (Hock et al., 2019). Surface air temperatures in mountain regions are projected to increase at an average rate of 0.3±0.2 ^∘C per decade until 2050 (very high confidence), i.e. faster than the present global average of 0.2±0.1 ^∘C per decade (Hock et al., 2019). Beyond 2050, surface air temperatures in high mountain regions are projected to increase under a high emission scenario (RCP8.5) and to stabilise at 2015–2050 levels under a low emission scenario (RCP2.6; IPCC, 2019a).

Projected changes in surface air temperature for the period 2071–2100 (compared to 1971–2000, under various emission scenarios) for the part of the Baltic Sea drainage basin that extends along the Scandinavian mountains (SMHI, 2020; Kjellström et al., 2016) will be of great importance for the assessment of future mass loss from glaciers draining into the Baltic Sea.

3.3.4(3) Permafrost

Due to recent warming, more than 20 % of the permafrost in the region was already lost in 1997–2018 (Fig. 13; Obu et al., 2020). As warming increases, so will the loss of permafrost (Sect. 1.5.2). Global projections show very limited permafrost in the region already at +2 ^∘C (Chadburn et al., 2017), but they (including Chadburn et al., 2017) do not account for peatland permafrost, which can persist for centuries outside of its climate equilibrium (Osterkamp and Romanovsky, 1999). Much of the permafrost in the Baltic Sea region was very close to its climatic boundary even before the recent acceleration of climate warming. Much of the lowland permafrost in palsas and peat plateaus in this region is very close to the 0 ^∘C thawing point and is likely relict permafrost persisting from the Little Ice Age (Sannel et al., 2016). Observations also show that lowland permafrost thaw has been going on for decades (Åkerman and Johansson, 2008). Preliminary analyses of permafrost loss in 1997–2018 suggest that this was roughly equally divided between alpine and lowland permafrost (22 % and 24 %, respectively; Fig. 13), which is in agreement with projections of loss of all types of Baltic permafrost in the future.

Permafrost thaw by climate warming is known to affect river runoff and its loads of carbon, nutrients, and contaminants such as mercury (Schuster et al., 2018; Vonk et al., 2015). The local effect of permafrost thaw in alpine headwaters can be significant (Lyon et al., 2009), but in the Baltic Sea basin, alpine permafrost thaw will likely have limited influence on the characteristics of river transport at their mouths on the Baltic Sea. This is because the alpine permafrost in the Baltic Sea drainage basin mainly affects solid bedrock or regolith, with almost no soil organic matter stored in permafrost (Fuchs et al., 2015). Thaw of permafrost in peatlands affects soils with very large stocks of organic material and has been suggested to cause large losses of peat carbon and nutrients into aquatic ecosystems (Hugelius et al., 2020). However, these projections are highly uncertain and based on studies of peatland thaw chronosequences in North America that may not be applicable to Fennoscandian permafrost peatlands (see Tang et al., 2018).

The extent of permafrost in the Baltic Sea drainage basin may decrease significantly in this century and, depending on which warming trajectory the Earth takes, may disappear altogether in the coming century. The thaw of alpine permafrost will have little effect on flows of water, carbon, and nutrients to the Baltic Sea. Thawing peatlands may increase the loads of carbon, nutrients, and mercury to the Baltic Sea, but these projections remain highly uncertain.

3.3.4(4) Sea ice

There are two new projections for sea ice in the Baltic Sea that have been produced after BACC II (BACC II Author Team, 2015). Luomaranta et al. (2014) used simplified regression and analytical models to estimate changes in sea ice extent (Fig. 29) and fast ice thickness. Due to their less demanding computational approach, they could base estimates on 28 CMIP5 models. As in the Arctic Ocean (Sect. 1.5.2), annual maximum ice extent and thickness were both estimated to decline in the future, but some sea ice will still form every year, even by the end of the century, which is in agreement with earlier studies (e.g. Haapala et al., 2001; Meier, 2002; Meier et al., 2004a). Under the RCP4.5 and RCP8.5 scenarios, the modelled mean maximum ice thicknesses in Kemi were projected to be 60 cm and nearly 40 cm, respectively, in 2081–2090. However, under the RCP8.5 scenario, two models projected Kemi to be ice free.

Figure 29Median, 5th, and 95th percentiles of the annual maximum sea ice extent of the Baltic Sea (in 10³ km²) estimated from 28 CMIP5 models. The right-hand side vertical axis shows the upper class limits for mild and average ice winters. (a) RCP4.5. (b) RCP8.5. Source: Luomaranta et al., 2014; their Fig. 5 is distributed under the terms of the Creative Commons CC-BY 4.0 License (http://creativecommons.org/licenses/by/4.0/, last access: 17 February 2022).

Höglund et al. (2017) used a more advanced approach to examine changes in sea ice conditions with a coupled ice–ocean model (Hordoir et al., 2019; Pemberton et al., 2017). They used downscaled atmospheric data from the EC-Earth and the Max Planck Institute ESMs and simulated the response of the ice for the RCP4.5 and RCP8.5 projections. Average annual maximum ice extent at the end of the century was projected to be 90–100 000 and 30–40 000 km² for the medium and high emission scenarios, respectively, and ice thickness was projected to decrease by 3–6 cm per decade. Höglund et al. (2017) also projected the mobility of the ice to increase but with little effect on future ridged ice production.

3.3.4(5) Lake ice

The latest model experiments demonstrate that the Baltic Sea catchment will experience a substantial reduction in lake ice cover in the future, with many lakes becoming ice covered only intermittently (Maberly et al., 2020; Sharma et al., 2019, 2021; Shatwell et al., 2019). This change will commence in the south and move northwards gradually. Lithuanian and Latvian lakes will lose their ice cover after +2 ^∘C warming, and further warming will gradually move winter ice loss northwards, so that, at +8 ^∘C warming, only lakes in northernmost Lapland will retain a winter ice cover (Maberly et al., 2020; Sharma et al., 2019, 2021; Shatwell et al., 2019).

3.3.5 Ocean and marine sediments

3.3.5(1) Water temperature

Ocean temperatures are rising at accelerating rates (IPCC, 2019a; Sect. 1.5.3(2)). For the end of this century, scenarios for the Baltic Sea projected a SST increase of 1.1 ^∘C (0.8–1.6 ^∘C, RCP2.6), 1.8 ^∘C (1.1–2.5 ^∘C, RCP4.5), and 3.2 ^∘C (2.5–4.1 ^∘C, RCP8.5) compared to 1976–2005 (Gröger et al., 2019, 2021b; see Table 14). In parentheses, the ensemble spreads indicated by the 5th and 95th percentiles are listed. These changes are slightly larger than the projected global SST changes (Sect. 1.5.3(2)). Ensembles other than the one by Gröger et al. (2019) gave similar results that vary between 1.9 ^∘C (RCP4.5) and 2.9 ^∘C (RCP8.5) for the ensemble mean temperature increase (Meier et al., 2022; see also Meier and Saraiva, 2020). By the end of the century, SST changes for the RCP8.5 scenarios significantly exceed natural variability. The largest open-sea warming is found in summer in the northern Baltic Sea, due to earlier melting of the sea ice (Figs. 30 and 31). Even higher warming of + 2–6 ^∘C (the range denotes RCP2.6 and RCP8.5 scenarios) is projected for the Curonian Lagoon by the year 2100 (Jakimavičius et al., 2018). The north–south gradient and pronounced seasonality of SST trends would be reduced in the future due to vanishing sea ice (Dutheil et al., 2021).

Figure 30(a) Ensemble mean sea surface temperature change (degrees Celsius) between 1970–1999 and 2070–2099 for winter (December through February; upper panels) and summer (June through August; lower panels) under RCP2.6, RCP4.5, and RCP8.5. Panel (b) is the same as panel (a) but for the standard deviation of the change, i.e. the ensemble spread (degrees Celsius). In total, eight different dynamically downscaled Earth system model (ESM) simulations are used. Data source: Gröger et al., 2019.

Figure 31Shown, from left to right, are the changes in summer (June to August) mean sea surface temperature (SST; degrees Celsius), annual mean sea surface salinity (SSS; grams per kilogram; hereafter g kg⁻¹), and annual mean bottom salinity (BS; g kg⁻¹) and winter (December to February) mean sea level (SL; centimetres) between 1978–2007 and 2069–2098. From top to bottom, the results of the ensembles by Meier et al. (2011a; ECOSUPPORT), under the A1B/A2 greenhouse gas emission scenario (white background), and by Saraiva et al. (2019b; BalticAPP), under RCP4.5 (grey background) and RCP8.5 (grey background), are depicted. Source: Meier et al., 2022.

The main driver of interannual variations in monthly mean SST is air temperature through the sensible heat fluxes (Meier et al., 2022). The second most important drivers are cloudiness over the open sea and latent heat flux and meridional and zonal wind velocities over coastal areas (the latter probably because of upwelling; Meier et al., 2022). In the vertical, the surface layer is warming more than the winter water, which is sandwiched between the surface layer and the halocline. Hence, the spring and summer thermoclines are becoming more intense (Gröger et al., 2019). Water temperature trends in the deepwater of those subbasins, such as the Bornholm Basin and Gotland Basin, that are sporadically ventilated by saltwater inflows originating from surface water were projected to be elevated as well (Meier et al., 2022). Projected changes in the vertical water temperature distribution are similar to those observed since 1850 (Kniebusch et al., 2019a).

For extreme events, projections suggest, inter alia, more tropical nights over the Baltic Sea, increasing the risk of record-breaking water temperatures (Meier et al., 2019a).

3.3.5(2) Salinity and saltwater inflows

Future changes in salinity will depend on changes in the wind fields over the Baltic Sea region (Lass and Matthäus, 1996), river runoff from the Baltic Sea catchment (Schinke and Matthäus, 1998), and mean sea level rise relative to the seabed of the sills in the entrance area (Meier et al., 2017, 2021a). A projected increase in river runoff will tend to decrease salinity, but sea level rise will have the opposite effect of tending to increase salinity because the water level above the sills at the Baltic Sea entrance would be higher, increasing the cross-sectional area of the Danish straits. As a result, saltwater imports from Kattegat would be larger. A 0.5 m higher sea level relative to the sill bottom at the end of the century would increase estimated Gotland Deep surface salinity by 0.7 g kg⁻¹ and bottom salinity by 0.9 g kg⁻¹ (Meier et al., 2017, 2021a). Furthermore, a hypothetically increasing westerly wind will block the freshwater flow out of the Baltic Sea. Consequently, the saltwater inflow is reduced due to mass conservation (Meier and Kauker, 2003).

Due to the large uncertainty in projected changes in wind fields over the Baltic Sea region (Sect. 3.3.1(5)) in changes in the freshwater supply from the catchment (Sect. 3.3.2(1)) and in global sea level rise (Sect. 3.3.5(4)), salinity projections showed a wide spread. No robust changes were identified because the two main drivers, river runoff and sea level rise, approximately compensate each other (Meier et al., 2021a). According to Saraiva et al. (2019b), river runoff would increase by about 1 % to 21 % at the end of the century, depending on the climate model under both RCP4.5 and RCP8.5. In the ensemble mean, such an increase would cause a decrease in surface and bottom salinity at the Gotland Deep of about 0.6–0.7 g kg⁻¹, with a large spread among the ensemble members. Assuming a negligible global sea level rise, the intensity and frequency of MBIs were projected to slightly increase due to changes in the wind fields (Schimanke et al., 2014). Hence, in ensemble studies that considered all potential drivers, no significant changes in salinity were projected as the ensemble mean (Meier et al., 2021a). In the case of salinity, global climate model uncertainty was identified to be the largest of all uncertainties (Meier et al., 2021a).

3.3.5(3) Stratification and overturning circulation

Model-based estimations of future stratification are still rare and depend critically on how well the models project changes in the three-dimensional distributions of temperature and salinity. A first systematic attempt using a high-resolution coupled ocean–atmosphere model and five different global climate models explored future stratification under RCP8.5 (Gröger et al., 2019). They assumed a 10 % increase in river runoff (approximately the ensemble mean in Saraiva et al., 2019b) and an unchanged mean sea level in the North Sea at the end of the century. The ensemble consistently indicated a basin-wide intensification of the pycnocline by 9–35 % for nearly the whole Baltic Sea and a shallowing of the pycnocline depth in most regions, except for the Gulf of Bothnia (Gröger et al., 2019). The area with a pycnocline intensity >0.05 kg m⁻³ m⁻¹ increased by 23 %–100 %. The warm season thermocline likewise intensified in nearly the entire Baltic Sea (Gröger et al., 2019).

All ensemble members indicated a strengthening of the zonal, wind-driven, near-surface overturning circulation in the southwestern Baltic Sea towards the end of the 21st century, whereas the zonal overturning at depth was reduced by ∼25 % (Gröger et al., 2019). In the Baltic proper, the meridional overturning showed no clear climate change signal. However, three out of five ensemble members indicated at least a northward expansion of the main overturning cell. In the Bothnian Sea, all ensemble members showed a significant weakening of the meridional overturning.

As the study by Gröger et al. (2019) and previous projections (e.g. Meier et al., 2006) did not consider global sea level rise, these scenario simulations are no longer considered plausible (Meier et al., 2021a, 2022). Considering all drivers of changes in salinity in the Baltic Sea (wind, river runoff, and global sea level rise), neither the saline-induced stratification nor the overturning circulation was projected to change systematically among climate models (Meier et al., 2022). It was found that, under a RCP4.5 or RCP8.5 scenario, a linearly rising mean sea level by the figures suggested by IPCC (2019b) would approximately counteract the effects of projected river runoff increases and wind changes on salinity.

3.3.5(4) Sea level

Global mean and, thus, Baltic Sea level will continue to rise at an increasing rate. During this century, melting ice sheets in Antarctica and Greenland are expected to contribute more to the total sea level than in the past (e.g. Mitrovica et al., 2018). The fingerprints from melting ice sheets in Antarctica on sea level rise will be more pronounced in the Northern Hemisphere and introduce large uncertainties for the Baltic sea level rise. On the other hand, ice melt in Greenland has a relatively modest effect on sea level in the Baltic Sea. Furthermore, the sea level in shelf seas such as the Baltic Sea will rise more strongly than one would expect from the thermostatic expansion of the local water column only, due to spill-over effects from the open ocean (Landerer et al., 2007; Bingham and Hughes, 2012). In addition, the long-term rate of coastal land rise is not easy to estimate accurately, due to the limited length of global positioning system (GPS) measurements and frequently revised geological model values.

Estimates for the ensemble mean global sea level rise by 2100, relative to 1986–2005, ranged from 43 cm (RCP2.6) to 84 cm (RCP8.5), with likely ranges of 29–56 cm and 61–110 cm, respectively (IPCC, 2019a; see Sect. 1.5.3(1)). Note that the figures in Sect. 1.5.3(1) refer to the mean changes between 1986–2005 and 2081–2100 and the previous IPCC AR5 (2014b) with 10 cm lower changes in sea level for RCP8.5 compared to IPCC, (2019a). In particular for RCP8.5, sea level rise projections by the Fifth IPCC Assessment Report (IPCC, 2014a) were somewhat lower than those from the more recent Special Report on the Ocean and Cryosphere in a Changing Climate (IPCC, 2019a) because of the updated contribution from Antarctica based upon new ice sheet modelling.

For the period 2090–2099, relative to 1980–1999, and based on the SRES A1B scenario, the projected absolute sea level rise in the Baltic Sea was estimated to be about 80 % of the global increase (Grinsted et al., 2015). These results were confirmed by other studies for other scenarios and slightly different reference periods (e.g. Kopp et al., 2014; Grinsted, 2015) and summarised by Pellikka et al. (2020) who, for the period 2000–2100, documented an ensemble mean absolute sea level rise in the Baltic Sea of about 87 % of the global mean sea level rise.

Altogether, considering land uplift and eustatic sea level rise, very likely ranges (5 %–95 % probability) of relative sea level change between 2000 and 2100 under the most pessimistic IPCC emissions scenario (RCP8.5) were projected to be, for example, + 29 to + 162 cm in Copenhagen (median + 68 cm), − 13 to + 117 cm in Stockholm (median + 25 cm), and + 21 to + 151 cm in St. Petersburg (median + 59 cm; Grinsted et al., 2015). For coastal sites in the northern Baltic Sea, relative sea level changes in the Gulf of Finland in 2000–2100 were projected to be + 29 cm (− 22 to + 92 cm), − 5 cm (− 66 to + 65 cm) for the Bothnian Sea, and − 27 cm (− 72 to + 28 cm) for the Bothnian Bay, where the land uplift is larger (Johansson et al., 2014). The ranges in the latter study were estimated from the 5 % and 95 % cumulative probabilities, considering several published scenarios from the Third and Fourth IPCC Assessment Reports. In a recent study based upon IPCC (2019a), Hieronymus and Kalén (2020) also estimated a sea level fall in the northern Baltic Sea and a 70 cm rise in the south by 2100, based upon the RCP8.5 scenario. These upper bounds of the sea level rise projections imply a very strong future acceleration of the present rates. Current observations seem to show an acceleration, but the present magnitude is still small (Hünicke and Zorita, 2016; see Sect. 3.2.5(4)).

Recent efforts since the IPCC AR5 report (IPCC, 2014a) that focused on the contribution of Antarctic ice sheets to global mean sea level rise have shown that warming ocean water, melting the ice sheets from below, can lead to instabilities in the ice sheet dynamics. The ice sheets flowing from land into the ocean are in contact with the ocean floor out to the grounding line. From there on outward, the ocean is melting the ice from below and the ice sheets become thinner and lighter. If the weight of the ice sheet becomes less than the weight of the ocean water it replaces, then it floats up and away. The grounding line retreats inland where the ice sheet is thicker and the ice flow larger and reinforces the ice loss (Mercer, 1978). This and related feedback loops could lead to an extra metre of sea level rise until the end of the century (e.g. Sweet et al., 2017). The most recent estimates based on expert judgement (Bamber et al., 2019) for global mean sea level rise in 2100, relative to 2000, including these potential contributions (including land water storage), were 69 and 111 cm for low and high sea level scenarios, respectively. For the high sea level scenario, the likely range (5 % to 95 %) was between 62 and 238 cm.

Future changes in sea level extremes in the Baltic Sea depend on future changes in mean sea level and future developments in large-scale atmospheric conditions associated with changing wind patterns. Model projections disagree regarding atmospheric circulation changes, and therefore, their relevance for extreme future sea levels remains unclear (Räisänen, 2017). For the Baltic Sea, changing mean sea levels are expected to have larger effects on future extremes than changing atmospheric circulation (Gräwe and Burchard, 2012). Sea ice loss in the future will further directly expose the northern Baltic coastline to stronger storm surges.

Recent projections of extreme sea levels along Europe's coasts have considered all drivers by linear superposition, i.e. absolute mean sea level rise and land uplift, tides (small in the Baltic Sea), storm surges, and waves (Vousdoukas et al., 2016, 2017). The results suggested that extreme sea levels will increase more than the mean sea level due to small changes in the large-scale atmospheric circulation, such as a northward shift of the Northern Hemisphere storm tracks and westerlies and increases in the NAO/Arctic Oscillation (AO; IPCC, 2014b). These changes in the large-scale atmospheric circulation of the Baltic Sea region are, however, not robust among GCMs, giving the projections of extreme sea levels by Vousdoukas et al. (2016, 2017) low confidence.

3.3.5(5) Waves

The few existing wave climate projections for the Baltic Sea indicate an increase in the mean wave conditions, either in the whole area (Groll et al., 2017) or in its northern part (Bonaduce et al., 2019). This increase in the mean conditions has been linked to two main drivers, i.e. (1) increased wind speeds and (2) reduced seasonal ice cover.

Groll et al. (2017) projected wave climate at the end of 21st century, based on two different scenarios. They found a slight increase in the median wind speeds for most of the Baltic Sea area, which led to an increase of up to 15 % in median SWH. Using only one climate scenario, Bonaduce et al. (2019) found that decreased wind speed in the southern Baltic Sea led to a decrease in mean SWH, whereas increased wind speeds in the north, especially in winter, led to increased mean SWH. As neither study used multi-model ensembles of scenario simulations (an exception for the western Baltic Sea is the work by Dreier et al., 2021), and as there is large uncertainty in the projected wind speeds and directions, which is not attributed to the decline in ice cover, the results may not be representative. The projected changes in SWH estimates are therefore inconclusive.

Based on CMIP5 simulations, Ruosteenoja et al. (2019) estimated that, in future, mean and extreme scalar wind speeds are not likely to significantly change in the Baltic Sea area. Hence, mean wave conditions would not change. They also estimated that the frequency of strong westerly winds will increase, while strong easterly winds will become less common. These type of changes might have more significance on the frequency of extreme SWH values and their spatial patterns.

For extreme values, these studies give even less reliable results. The results of Groll et al. (2017), Suursaar et al. (2016), and Bonaduce et al. (2019) all indicated large spatial variability in how the projected extremes changed. In addition to the wind speed, extreme values are quite sensitive to wind direction, since fetch varies with direction due to the geometry of the Baltic Sea. Mäll et al. (2020) simulated how wave conditions during three historical Baltic Sea storms would change under climate change conditions. The results showed slight, but not significant, changes in extreme SWH values during the storms.

Future changes in seasonal sea ice conditions in the northern Baltic Sea are more reliable, and their effect on the wave climate is easier to estimate (Rutgersson et al., 2022). Mild ice winters have already become common, and new records of the lowest annual maximum ice extent have been recorded (see Sect. 3.2.4(4)). In the Baltic Sea, the ice season partly overlaps with the seasons of the strongest winds, namely autumn and winter. The mean and extreme values of SWH are therefore expected to increase in areas like the Bothnian Sea, which now typically has ice cover in winter, but will have lost it in the future Baltic Sea climate.

3.3.5(6) Sedimentation and coastal erosion

As a consequence of the probably accelerating sea level rise, coastal erosion will increase regionally to fill the increased underwater accommodation space. How much erosion will increase will depend not only on the rate of sea level rise but also on the intensity of storms (Zhang et al., 2017).

Coastal erosion, accretion, and alongshore sediment transport are primarily controlled by winds and wind-induced waves in the Baltic Sea. Projecting the future rate of coastal erosion or accretion in the Baltic Sea is highly uncertain because of a lack of consensus in the prediction of future storms. Neglecting potential change in future storms and assuming an intermediate sea level rise scenario (RCP4.5), an increment of 0.1–0.3 m yr⁻¹ in coastline erosion has been projected for some parts of the southern Baltic Sea coast (Zhang et al., 2017; Deng et al., 2015). Due to the prevailing westerly winds, the dominant sediment transport will continue to be eastwards along most of the southern Baltic Sea coast but with high variability along coastal sections, with a small incidence angle of incoming wind waves (Dudzińska-Nowak, 2017). It has been found that even a minor climate-change-driven rotation of the predominant wind directions over the Baltic Sea may substantially alter the structural patterns and pathways of wave-driven transport along large sections of the coastline (Viška and Soomere, 2013).

The presence of sea ice is an important factor moderating coastal erosion. Storm surges and wave run-up on the beach are much higher in ice-free periods than when there is even partial ice cover. The hydrodynamic forces are particularly effective in reshaping the shoreline when there is no ice and sediment is mobile (Ryabchuk et al., 2011). Due to global warming, both the area and duration of ice cover in the Baltic Sea will be reduced in future (Sect. 3.3.4(4)), thus increasing coastal erosion.

Foredunes will likely continue to form on prograding coasts but at rates influenced by the accelerating sea level rise (Zhang et al., 2017). Foredunes may tend to become higher, but with reduced prograding rate and wavelength, if sea level rise is accelerated or storm frequency increased. If the wind–wave climate is stable, the height of coastal foredunes on a prograding coast remains stable or increases linearly with a low to intermediate rate (<1.5 mm yr⁻¹) of sea level rise. An accelerating rate of sea level rise and/or changing storm frequency will lead to a nonlinear growth in height (following a quadratic or a higher power law; Zhang et al., 2017; Lampe and Lampe, 2018). The critical threshold that separates linear and nonlinear foredune growth in response to sea level rise is likely to be reached before 2050 in the RCP8.5 scenario (Zhang et al., 2017).

The yet-to-be-determined anthropogenic influence affects sediment transport and coastal erosion as well. Sediment transport and coastal erosion are relevant for coastal management, construction, and protection strategies. In general, two main types of management strategies exist for the Baltic Sea coast, i.e. (1) coastal protection by soft or hard measures and (2) adaptation to coastal change, accepting that in some places the coast would be left in its natural state (BACC II Author Team, 2015). However, administrative efforts for coastal protection differ among Baltic Sea countries, even between neighbouring states or nations. It has been found that engineering structures (e.g. piers and seawalls) may influence coastline change at a much larger spatial scale than the dimension of the structure itself.

3.3.5(7) Marine carbonate system and biogeochemistry

The BACC II Author Team (2015) concluded that model simulations indicated that climate change has a potential to intensify eutrophication in the Baltic Sea. However, they also showed that the implementation of nutrient load reductions according to the Baltic Sea Action Plan (BSAP; HELCOM, 2013a) may not only mitigate this effect but may even decrease the hypoxic and anoxic areas in the Baltic Sea. Note that the nutrient load abatement calculations of the BSAP did not take the effect of climate change into account. In contrast, the business-as-usual nutrients input scenario may increase the hypoxic area by about 30 % and even more than double the area affected by anoxia by 2100.

As atmospheric CO₂ rises, so will the concentration of CO₂ in the Baltic Sea surface water. This will influence the mean future pH, while eutrophication and enhanced organic matter production/remineralisation will increase the amplitude of daily and seasonal pH fluctuations without affecting the mean values much. Taking various nutrient load input scenarios into account, projections suggested that pH in the Baltic Sea surface water will decrease; in the worst-case emission scenario (atmospheric pCO₂ of 850 ppm – parts per million), the pH will drop by about 0.40 by 2100, while the decrease in a more optimistic emission scenario (550 ppm) will be smaller, i.e. about 0.26 (BACC II Author Team, 2015).

3.3.5(7.1) Oxygen and nutrients

Projected warming, increased precipitation, and global mean sea level rise may worsen eutrophication and oxygen depletion in the Baltic Sea by reducing air–sea fluxes and vertical transports of oxygen in the water column, intensifying internal nutrient cycling and increasing river-borne nutrient loads due to increased river runoff (Meier et al., 2011a, 2012b, c). However, the future response of deepwater oxygen conditions will depend mainly on nutrient loads from land (Saraiva et al., 2019a, b; Meier et al., 2021a; see Fig. 32). In contrast to the global ocean (see Sect. 1.5.4), future nutrient supplies will have a relatively larger effect on oxygen conditions and primary production than warming. With high nutrient loads, the changing climate will have a considerable negative effect, but if loads are kept low, climate effects can be small or negligible. Scenario simulations suggested that full implementation of the nutrient load reductions required by BSAP will significantly improve the eutrophication status of the Baltic Sea, irrespective of the driving global climate model (Saraiva et al., 2019b; Meier et al., 2021a) and regional coupled climate–environmental model (Meier et al., 2018a). Despite large inherent uncertainties of future projections, modelling studies suggested that the future Baltic Sea ecosystem may unprecedentedly change compared to the past 150 years (Meier et al., 2012a).

Figure 32Ensemble mean summer (June to August) bottom dissolved oxygen concentration changes (millilitres per litre; hereafter mL L⁻¹) between 1978–2007 and 2069–2098. From left to right, the results of the nutrient load scenarios Baltic Sea Action Plan (BSAP), reference (REF), and business as usual (BAU) are shown. From top to bottom, the results of the ensembles by Meier et al. (2011a; ECOSUPPORT), under the A1B/A2 greenhouse gas emission scenario (white background), and by Saraiva et al. (2019b; BalticAPP), under RCP4.5 (grey background) and RCP8.5 (grey background), are depicted. Source: Meier et al., 2022.

By the end of the century (2069–2098), the ensemble mean hypoxic area is projected to change only slightly under reference (− 14 % for RCP4.5 and − 5 % for RCP8.5) and high (− 2 % for RCP4.5 and + 5 % for RCP8.5) nutrient load scenarios, compared to 1976–2005 (Saraiva et al., 2019b). Nutrient loads in the reference scenario are the average loads in 2010–2012. The high, or worst, scenario assumes changes caused by a fossil-fuelled development, coupled to increasing river runoff (Saraiva et al., 2019a). Changes in nitrogen and phosphorus loads were estimated from assumptions on regional population growth, changes in agricultural practices, such as land and fertiliser use, and developments in sewage treatment (Zandersen et al., 2019; Pihlainen et al., 2020). Under the BSAP scenario, the ensemble mean hypoxic area will be reduced by 50 %–60 % at the end of the century, in comparison with 1976–2005 (Saraiva et al., 2019a). The relative reductions in hypoxic area may decrease with increasing sea level (Meier et al., 2021a).

In the same model ensemble (Saraiva et al., 2019a), BSAP implementation is projected to reduce the water column phosphate pool in the Baltic Sea by 59 % (RCP4.5) and 56 % (RCP8.5) by the end of the century, and even the reference loads would lead to a decline by 24 % (RCP4.5) and 18 % (RCP8.5). Also, a larger ensemble (Meier et al., 2018a) of eight biogeochemical models forced by outputs of seven ESMs downscaled by four different RCMs projected that the BSAP reduced phosphate concentrations in the Baltic proper, Gulf of Finland, and Bothnian Sea despite climate change, with the largest reductions in surface concentrations by approximately 3 mmol m⁻³ in the Gulf of Finland. Present-day nutrient loads led to small increases in surface phosphate concentration in the Baltic proper, a small decline in the Gulf of Finland, and little change in Bothnian Sea and Bothnian Bay. Little change was predicted for DIN concentrations in the Baltic proper, whereas simulations showed an increase in the Gulf of Finland and the Bothnian Sea, regardless of whether nutrient loads were kept at the present level or whether loads were reduced.

Furthermore, future projections suggested that the sea ice decline in the northern Baltic Sea may have considerable consequences for the marine biogeochemistry because of changing underwater light conditions and wave climate (Eilola et al., 2013). Eilola et al. (2013) found that, by the end of the century, the spring bloom would start by up to 1 month earlier and winds and wave-induced resuspension would increase, causing an increased transport of nutrients from the productive coastal zone into the deeper areas.

For the Baltic proper, the internal nutrient cycling and exchanges between shallow and deeper waters were projected to be intensified, and the internal removal of phosphorus may become weaker in future climate (Eilola et al., 2012). These effects may counteract the efforts of planned nutrient input reductions.

Uncertainties in projections from Baltic Sea ecosystem models have recently been systematically assessed for the first time (Meier et al., 2018a, 2019b, 2021a; Meier and Saraiva, 2020). Larger sources of uncertainty are global and regional climate model uncertainties, in particular concerning global mean sea level rise and regional water cycling (Meier et al., 2019b). The mechanism behind the correlation between large-scale meteorological conditions in the different climate periods and oxygen conditions in the Baltic Sea is not well understood and subject to ongoing research. With respect to nutrient concentrations, uncertainties in conditions at the North Sea boundary and difficulties in simulating the long-term response of the Baltic Sea biogeochemical system to changes in nutrient inputs also play a role.

Under the BSAP scenario, mean nitrogen fixation would decrease (Meier et al., 2021a), and record-breaking cyanobacteria blooms may no longer occur in the future, but record-breaking events may reappear at the end of the century in a business-as-usual nutrient load scenario (Meier et al., 2019a).

3.3.5(7.2) Marine CO₂ system

The rising atmospheric pCO₂ due to anthropogenic emissions will increase the mean pCO₂ of surface seawater and, thus, has the potential to lower the pH. However, the magnitude of pH changes will also depend on the development of total alkalinity concentrations (A_T; Omstedt et al., 2012) Future A_T changes in the Baltic Sea will be shaped by both external inputs (riverine runoff and inflows from the North Sea) and internal generation. The latter is due to the biogeochemical processes of organic matter production and remineralisation, especially under euxinic conditions. Kuznetsov and Neumann (2013), who used A_T as a tracer in a model (no internal processes included) calculated that, on average, A_T in surface Baltic Sea waters would decrease by about 150 µmol kg⁻¹ by 2100, a change corresponding to an assumed decrease in salinity. Simulations by Gustafsson et al. (2019) that included most of the biogeochemical processes affecting A_T (except S burial and Fe oxide availability) showed that A_T in the central Baltic Sea in the business-as-usual scenario will first increase, by about 100 µmol kg⁻¹ by 2050, and then revert to present levels by 2100. If BSAP is implemented, A_T will decrease by about 150 µmol kg⁻¹ in 2100 from present levels. Irrespective of the nutrient load scenario, pH is eventually expected to decrease in the central Baltic Sea due to anthropogenic CO₂ emissions. Assuming the A1B CO₂ emission scenario (pCO₂ increase to 700 µ atm by 2100), pH will drop to about 7.9 and 7.8 under business-as-usual and BSAP scenarios, respectively.

All available future projections of changes in alkalinity are highly uncertain, as they address only a few out of multiple factors determining the alkalinity pool in the Baltic Sea, i.e. changes in salinity, river runoff, weathering in the catchment, organic matter production (eutrophication), remineralisation (especially at low redox conditions), and processes in sediments (including pyrite and vivianite formation; Kuliński et al., 2021).

3.3.6 Marine biosphere

3.3.6(1) Pelagic habitats

3.3.6(1.1) Microbial communities

The effects of climate change on microbes and the functioning of the microbial loop have been studied by experiments in which temperature, salinity, dissolved organic matter (DOM), and OA were manipulated. In general, microbial activity and biomass increased with increasing DOM and temperature (Ducklow et al., 2009), but effects can be mixed. For instance, an increase in DOM in the northern Gulf of Bothnia enhanced the abundance of bacteria, whereas a temperature increase (from 12 to 15 ^∘C) decreased their abundance, probably due to a simultaneous increase in bacterivorous flagellates (Nydahl et al., 2013).

In the southern Baltic Sea, the impact of OA was limited, and the bacterial community responded primarily to temperature and phytoplankton succession (Bergen et al., 2016). In experiments where CO₂ was increased and salinity decreased (from 6 to 3 g kg⁻¹), heterotrophic bacteria declined (Wulff et al., 2018). In experiments with increasing temperature (from 16 to 18–20 ^∘C) and reduced salinity (from 6.9 to 5.9 g kg⁻¹), the Baltic proper microbial community also showed mixed responses, probably due to indirect food web effects (Berner et al., 2018).

3.3.6(1.2) Phytoplankton and cyanobacteria

The projected increase in precipitation is expected to increase nutrient loads, especially into the northern Baltic Sea (Huttunen et al., 2015), and together with increased internal loading of nutrients, several modelling studies project an increased phytoplankton biomass by the end of the century (Meier et al., 2012b, c; Skogen et al., 2014; Ryabchenko et al., 2016).

Several mesocosm studies have investigated the effects of warming on southern Baltic Sea phytoplankton communities. Warming accelerated the phytoplankton spring bloom and increased primary productivity (Sommer and Lewandowska, 2011; Lewandowska et al., 2012; Paul et al., 2015). The total phytoplankton biomass still decreased due to the negative effects of warming on nutrient flux (Lewandowska et al., 2012, 2014).

OA may enhance phytoplankton productivity by increasing the CO₂ concentration in the water. The biomass of southern Baltic Sea autumn phytoplankton increased in mesocosms simulating OA (Sommer et al., 2015). In many experiments, OA had, however, little effect on phytoplankton community composition, fatty acid composition, or biovolume in spring or autumn (Paul et al., 2015; Bermúdez et al., 2016; Olofsson et al., 2019).

It has been suggested that climate change may increase the blooming of toxic species, such as the dinoflagellate Alexandrium ostenfeldii (Kremp et al., 2012, 2016) and the cyanobacterium Dolichospermum sp. (Brutemark et al., 2015; Wulff et al., 2018). There are also contradictory results indicating that OA and warming may decrease the biomass of Nodularia sp. and Dolichospermum sp. (Eichner et al., 2014; Berner et al., 2018). Several modelling studies projected increases in cyanobacteria in the warmer and more stratified future Baltic Sea (Meier et al., 2011b; Andersson et al., 2015; Neumann et al., 2012; Chust et al., 2014; Hense et al., 2013), but other modelling studies projected that the environmental state of the Baltic Sea will be significantly improved, and extreme cyanobacteria blooms will no longer occur if BSAP is fully implemented (Meier et al., 2018a, 2019a; Saraiva et al., 2019a; see Fig. 33).

Figure 33The same as Fig. 32 but for annual mean Secchi depth changes (metres). Secchi depth changes indicate changes in water transparency caused by phytoplankton and detritus concentration changes. Source: Meier et al., 2022.

3.3.6(1.3) Zooplankton

The effects of increasing temperature and OA on zooplankton have been studied experimentally. In Acartia sp., a dominant copepod in the northern Baltic Sea, warming decreased egg viability, nauplii development, and adult survival, and both warming and OA had negative effects on adult female size (Garzke et al., 2015; Vehmaa et al., 2013, 2016).

In contrast, the effects of climate change on microzooplankton (MZP) seem to be mostly beneficial. Warming improved the growth rate of southern Baltic Sea MZP, which led to a reduced time lag between phytoplankton and MZP maxima, improving the food supply to microzooplankton in warm conditions (Horn et al., 2015). Aberle et al. (2015) showed that, while protozooplankton escaped predation by slower growing copepods at low temperatures, at warmer temperatures, small ciliates in particular became more strongly controlled by copepod predation.

3.3.6(2) Benthic habitats

3.3.6(2.1) Macroalgae and vascular plants

The effects of climate change on bladderwrack, Fucus vesiculosus, have been studied in a number of experiments. OA appears to have a relatively small effect on macroalgae (Al-Janabi et al., 2016; Wahl et al., 2020), while temperature effects can be significant. The effects of increasing temperature are not linear, however. Growth or photosynthesis is not impaired under projected temperature increase (from 15 to 17.5 ^∘C), but at extreme temperatures (27 to 29 ^∘C), photosynthesis declines, growth ceases, and necrosis starts (Graiff et al., 2015; Takolander et al., 2017). In very low salinity (2.5 g kg⁻¹), sexual reproduction of F. vesiculosus ceases (Rothäusler et al., 2018).

The direct and indirect effects of changes in temperature, salinity, and pH may alter the geographic distribution of many species in the Baltic Sea. Assuming a decline in salinity, a reduced penetration of marine species, such as bladderwrack, eelgrass, and blue mussel, into the Baltic Sea has been predicted (Vuorinen et al., 2015). A large number of other species is affiliated with such keystone species, and species distribution modelling has indicated that, e.g., a decrease in bladderwrack will have large effects on the biodiversity and functioning of the shallow water communities of the northern Baltic Sea (Jonsson et al., 2018; Kotta et al., 2019). The responses of eelgrass, Zostera marina, to climate change and eutrophication mitigation have recently been modelled by Bobsien et al. (2021).

Experiments on climate change effects have also been made with other macroalgae and vascular plants. Thus, OA increased the growth of the opportunistic green alga, Ulva intestinalis, in the Gulf of Riga (Pajusalu et al., 2013, 2016). Other studies showed that charophyte photosynthesis increased under high pCO₂, whereas eelgrass did not respond to the elevated pCO₂ alone (Pajusalu et al., 2015). Salinity decline is projected to decrease the distribution of Z. marina and the red alga Furcellaria lumbricalis, whereas warming will probably favour charophytes (Torn et al., 2020).

3.3.6(2.2) Zoobenthos

The effects of warming on invertebrates are nonlinear. The respiration and growth of the isopod Idotea balthica increased up to 20 ^∘C and then decreased at 25 ^∘C (Ito et al., 2019). Many marine invertebrates, including isopods, will also directly and indirectly suffer from decreasing salinity (Kotta et al., 2019; Rugiu et al., 2017) and OA. The size and time to settlement of the pelagic larvae of the Baltic clam Limecola balthica (Macoma balthica) increased with OA, suggesting a developmental delay (Jansson et al., 2016), whereas OA had no effect on the isopod Saduria entomon (Jakubowska et al., 2013) or the larvae of the barnacle Amphibalanus improvisus (Pansch et al., 2012).

Several modelling studies have estimated the relative effects of hydrodynamics, oxygen, and food availability on Baltic Sea zoobenthos under various scenarios. As the assumptions about climate change and nutrient inputs differ, a comparison between studies is impossible, and final conclusions cannot been drawn. In previously hypoxic areas, benthic biomass was projected to increase (until 2100) by up to 200 % after re-oxygenating bottom waters, whereas in permanently oxygenated areas, macrofauna may decrease by 35 % due to lowered food supply to the benthic ecosystem (Timmermann et al., 2012). It has, however, been concluded that nutrient reductions will be a stronger driver for Baltic Sea ecosystem than climate change (Friedland et al., 2012; Niiranen et al., 2013; Ehrnsten et al., 2019). These studies suggested that benthic–pelagic coupling will weaken in a warmer and less eutrophic Baltic Sea, resulting in gradually decreasing benthic biomass (Ehrnsten et al., 2020).

3.3.6(3) Non-indigenous species

It is often suggested that climate change will favour invasions by non-indigenous species worldwide (Jones and Cheung, 2014). It has been shown that non-native benthic species typically occur in areas with reduced salinity, high temperatures, a high proportion of soft seabed, and low wave exposure, whereas most native species show an opposite pattern (Jänes et al., 2017). Modelled temperature and salinity scenarios suggested an increase in Ponto–Caspian cladocerans in the pelagic community and an increase in dreissenid bivalves, amphipods, and mysids in the benthos of coastal areas in the northern Baltic Sea by 2100 (Holopainen et al., 2016). Disentangling factors facilitating the establishment of non-native species demands long-term surveys and data from multiple environments in order to distinguish climate-related effects from other ecosystem-level drivers (Bailey et al., 2020). In addition, studies on changing connectivity are needed (e.g. Jonsson et al., 2020).

3.3.6(4) Fish

Climate change may affect Baltic Sea fish through changes in water temperature, salinity, oxygen, and pH, as well as nutrient loads, which indirectly affect food availability for fish. The responses of cod larvae to OA and warming have been studied experimentally. Some studies found no effect on the hatching, survival, or development rates of cod larvae (Frommel et al., 2013), while, in others, the mortality of cod larvae doubled when exposed to high-end OA projections (RCP8.5). Several modelling studies, however, project low abundances of cod towards the end of the century, due to continued poor oxygen conditions (Niiranen et al., 2013; Wåhlström et al., 2020).

Climate change may also be positive for fish stocks. Warmer spring and summer temperatures have been projected to increase productivity of sprat (Voss et al., 2011; MacKenzie et al., 2012; Niiranen et al., 2013). For herring, results are more varied as both an increase (Bartolino et al., 2014) and a short-term decrease (Niiranen et al., 2013) have been projected.

Multi-species modelling has also emphasised the role of climate for cod stocks. If fishing is intense but climate remains unchanged, cod declines, but not very dramatically, while if climate change proceeds as projected, cod disappeared in two models out of seven, even with the current low fishing effort (Gårdmark et al., 2013). Different scenarios yield very different outcomes, however. A medium CO₂ concentration scenario (RCP4.5), low nutrients, and sustainable fisheries resulted in high numbers of cod and flounder, while high emissions (RCP8.5) and high nutrient loads resulted in high abundance of sprat (Bauer et al., 2018, 2019). All these studies assumed a more or less pronounced decrease in salinity.

3.3.6(5) Marine mammals

Ringed seal and grey seal – sea ice

Climate change is projected to drastically reduce the extent of seasonal sea ice in the Baltic Sea (Luomaranta et al., 2014; Meier, 2006, 2015; Meier et al., 2004b). At the end of the 21st century, ice will probably, in most years, be confined to the Bothnian Bay, the eastern Gulf of Finland, the Archipelago Sea, and the Moonsund (sound between the Estonian mainland and the offshore western islands of Saaremaa, Hiiumaa, Muhu, and Vormsi) and the eastern parts of the Gulf of Riga, such as Pärnu Bay (Meier et al., 2004b), with corresponding changes in the breeding and moulting distribution of ringed seals. Aside from these projections, ice cover has been even more limited in all the southern areas in recent years. Extirpation of one or more of the three southern breeding ringed seal populations is possible (Sundqvist et al., 2012; Meier et al., 2004b).

The ringed seal is an obligatory ice breeder that digs lairs in the snowdrifts on offshore ice for protection of the pup (e.g. Smith and Stirling, 1975). The Baltic grey seal prefers loose floes of drift ice (Hook et al., 1972) but can also breed on land (Jüssi et al., 2008). Overall pup survival in land-breeding grey seals is probably lower than for ice breeders (Jüssi et al., 2008). Absence or low quality of sea ice will adversely affect pup survival and quality in ice-breeding seals. The effects can be seen by the end of the breeding season and beyond (Jüssi et al., 2008). Grey and ringed seals are capital breeders, i.e. their pup quality depends on effective transfer of maternal energy (fatty milk) during a short, intensive lactation period. Timing of birth for both species is strongly adapted to the availability of the optimal breeding platform of sea ice. The height of the pupping season is around February to early March when the extent and strength of sea ice is usually greatest. The immediate breeding success can be defined as the survival and quality of the offspring at the end of the breeding season, but breeding conditions may have population consequences by affecting the survival and fitness of the pups throughout their lives (McNamara and Houston, 1996; Kauhala and Kurkilahti, 2020). A warming climate with higher air and water temperatures will decrease the extent of ice cover, the ice thickness, and the overlaying snow cover, as well as the stability and duration of the ice.

Loss of habitat is critical for the reproductive success of the ice-associated seals, especially the ringed seal, and can eventually lead to local population decreases and changes in breeding distribution, starting in the southernmost parts of its range. The ringed seal populations breeding in the Gulf of Finland, Gulf of Riga, and Archipelago Sea (SW Finland) are already small and vulnerable to any negative changes in habitat quality.

Harbour seal and grey seal – flooding of haul-outs

Harbour seals and grey seals rely on undisturbed haul-out areas for key life cycle events such as breeding, moulting, and resting (Allen et al., 1984; Thompson, 1989; Watts, 1996; Reder et al., 2003). In the southern Baltic Sea, relative sea levels have risen by 1 to 3 mm yr⁻¹ over the interval 1970–2016 (Sect. 3.2.5(4)), and increased rates of sea level rise are expected in the future (EEA, 2019c; Grinsted, 2015). A low emission scenario for the 21st century projects an additional sea level rise of 0.29–0.59 m and a high emissions scenario an extra 0.61–1.10 m, but substantially higher values cannot be ruled out (Grinsted, 2015; IPCC, 2019a). A high emission scenario is, thus, likely to flood all current seal haul-outs in the southern Baltic Sea and many important localities in Kattegat, while under a low emission scenario, most haul-outs in the southern Baltic Sea will be flooded, while others will be reduced to small fractions of their current area. In the northern and central Baltic Sea and northeastern Kattegat archipelago areas, seals will have alternative islets and skerries and are not likely to be affected to the same degree as in the southern and in western Kattegat. In parts of the Gulf of Bothnia, relative sea levels may even fall due to postglacial rebound (EEA, 2019c).

Harbour porpoise

There are no direct studies of the effects of climate change on harbour porpoises in the Baltic Sea. Hence, the following is based mostly on informed guesswork and on a few studies in other areas. There are a multitude of ways that changes in one parameter can affect others, and we do not currently have the knowledge to predict the cumulative effects this might have on the Baltic Sea harbour porpoise population.

Harbour porpoises are present from Greenland to the African coast and the Black Sea and seem to have a rather wide thermal tolerance. Therefore, even though it is projected that we will see a 1.1–3.2^∘C increase in SST in the Baltic Sea (Sect. 3.3.5(1)), it seems unlikely that this will directly affect harbour porpoise distribution, unless the Baltic Sea harbour porpoise is specifically adapted to colder temperatures. If this is the case, a northwards range shift might occur. With the expected future decrease in sea ice extent, the winter habitat available for the harbour porpoise in the northern Baltic Sea would increase.

Harbour porpoises are small cetaceans, with limited capacity to store energy, that mostly live in cold environments. Hence, they need to eat almost constantly (Read and Hohn, 1995; Wisniewska et al., 2016) and are therefore expected to be tightly dependent on their prey (Sveegaard et al., 2012). Their main prey species in the Baltic proper are cod (at least before the recent cod stock collapse), sprat, herring, gobies, and sand eel (where present). Climate-induced changes in, for example, SST, fronts, stratification, and, to some degree, currents will affect the distribution, abundance, and possibly the quality of prey species and, in turn, the harbour porpoise population. Their distribution may shift as they follow their prey, and potential food shortages might lead to starvation, with possible population effects.

It has been hypothesised that the susceptibility of marine mammals to disease may increase as temperature increases. Higher temperatures can increase pathogen development and survival rates, facilitate transmission among individuals, and increase individual susceptibility to disease. The negative effects of disease and environmental contaminants on individual fitness will obviously worsen if the animal is also under nutritional stress.

Seals and changes in the distribution of prey species

Any large alteration of the ecosystem can affect the distribution of seals if there are climate-related changes in the abundance and distribution of their main prey species, such as herring, sprat, and cod, as is possible with climate change. Such changes in top consumer distribution have been modelled in other sea areas, such as the UK continental shelf, where the current distribution of harbour seals did not match the projected future distribution of their prey well (Sadykova et al., 2020). There are large differences between Baltic Sea climate projections of future salinity (Saraiva et al., 2019b), and other factors such as temperature, eutrophication, predation, and competition also affect fish distributions. Thus, future changes in abundance and distribution of seal prey species, such as herring and cod, are difficult to project (Dippner et al., 2008, 2019; Lindegren et al., 2010; Vuorinen et al., 2015).

3.3.6(6) Waterbirds

Climate change scenarios agree in projecting a strong temperature increase in the Arctic and sub-Arctic. This will likely cause a northward expansion of species ranges, with colonisation by new breeding and wintering species, and local species declines following migration of populations to ice-free northern waters (Pavón-Jordán et al., 2019; Fox et al., 2019).

If salinity in the Baltic Sea decreases, invertebrate species serving as prey for waterbirds (e.g. blue mussels for common eiders) are likely to change in distribution, body size, and quality as food, with consequences for the distribution, reproduction, and survival of the waterbirds that eat them (Fox et al., 2015). Predicting the consequences of climate change for piscivorous seabirds is complex because the effects are not uniform among Baltic Sea fish species. For example, an expected increase in recruitment and abundance in an important prey species (sprat; MacKenzie et al., 2012; Lindegren et al., 2012) and declining numbers of large piscivorous fish (cod) may favour fish-eating birds, although management efforts to improve cod stocks may counteract the expected increase in sprat and lead to population declines of their main bird predator, the common guillemot (Kadin et al., 2019). Herring, another important prey species, is reported to be negatively affected by decreasing salinity (declining energy content; Rajasilta et al., 2018).

A rising sea level will reduce the area of salt marshes available for the breeding of waders and foraging by geese (Clausen et al., 2013), and other coastal habitats would likewise be affected (Clausen and Clausen, 2014). Sea level rise in combination with storms may cause loss by erosion of current coastal breeding habitats and flood breeding sites, thus affecting the breeding success of coastal waterbirds. Climate change can also be expected to affect waterbirds in the Baltic Sea by changing the incidence of diseases and parasites (Fox et al., 2015).

Marine food webs

Climate change and other anthropogenic environmental drivers are expected to change entire marine food webs, from coastal to offshore, from shallow to deep, and from pelagic to benthic (sedimentary) as species distributions are impacted and key nodes and linkages in the food webs are altered or lost (Lindegren et al., 2010; Niiranen et al., 2013; Leidenberger et al., 2015; Griffiths et al., 2017; Kotta et al., 2019; Gårdmark and Huss, 2020). These climate-driven changes will also, when combined with societal changes, affect aquatic ecosystem services, for instance the future primary production (a supportive ecosystem service) and fish catches (a provisioning ecosystem service; Hyytiäinen et al., 2021).

Certain marine species, e.g. cod and bladderwrack, may decline in both distribution and abundance (Gårdmark et al., 2013; Takolander et al., 2017), whereas others, e.g. sprat and certain mainly coastal freshwater fish, may increase (MacKenzie et al., 2012; Bergström et al., 2016). An increase in cyanobacteria blooms has also been projected, especially for the central Baltic Sea (Meier et al., 2011b; Funkey et al., 2014), while increased flow of DOC may reduce both primary and secondary production in the northernmost low saline areas, with pronounced brown water runoff (Wikner and Andersson, 2012; Figueroa et al., 2021). The responses also depend on human intervention, i.e. the success of nutrient reduction schemes, and are most probably nonlinear (Hyytiäinen et al., 2021; Ehrnsten et al., 2020). However, it can be summarised that – if only climate change is accounted for – most studies tend to project a decline in the overall state of the ecosystem, and a long-term decline in the provision of associated ecosystem services to humans is likely if the climate change is not significantly mitigated.

6.23.1 Lower trophic levels

The summer cyanobacteria bloom in the Baltic proper, and increasingly in recent years in the Bothnian Sea, is considered as one of the main problems of Baltic Sea eutrophication, and the nitrogen fixation it carries out is an important process in Baltic ecosystem models (Munkes et al., 2021). It has long been considered limited by the availability of phosphorus (Larsson et al., 1985; Granéli et al., 1990). It is, therefore, remarkable that it is not possible to predict interannual variations in cyanobacteria blooms observed by satellites from water chemistry (Kahru et al., 2020; Hieronymus et al., 2021).

There are significant knowledge gaps related to the quantification of nitrogen fixation and the fate of the fixed nitrogen in the Baltic Sea pelagic zone. Direct nitrogen fixation measurements were, until recently, dogged by method problems, and even if these are now hopefully largely resolved (Klawonn et al., 2015), the enormous patchiness of cyanobacteria blooms remains a huge problem. The alternative approach of directly measuring the increase in total combined nitrogen during the bloom (Larsson et al., 2001) requires very high precision, also suffers from patchiness problems (Rolff et al., 2007), and has not been used much. Finally, the amount of nitrogen fixed can be estimated by modelling, based on uptake of CO₂ or phosphorus and assuming a Redfield N : P or C : N ratio. This theoretically highly attractive approach (Eggert and Schneider, 2015) is hampered by the possibility of non-Redfieldian ratios and has made some biologists sceptical by predicting high nitrogen fixation in spring when there are not sufficient known nitrogen-fixing autotrophs in the water to carry out this nitrogen fixation.

While total nitrogen in the water column clearly increases during the summer cyanobacterial bloom, just a couple of months later this increase seems largely to have disappeared, even though sediment traps find little evidence that nitrogen has settled out of the upper mixed layer. Sediment trap measurements might be gross underestimates, or there are unidentified sites of denitrification or other overlooked nitrogen sinks in the water column. Nitrogen fixation is a central process in Baltic ecosystem models, and better observationally based estimates of processes in the nitrogen cycle are required for assessing their credibility (Munkes et al., 2021).

6.23.2 Marine mammals

There is a great need for further research on the effects of climate change on marine mammals, especially on the critically endangered Baltic proper harbour porpoise and the Baltic ringed seal, given the multiple threats and cumulative impacts on these populations. Seal and porpoise foraging distribution and the relation of seals to haul-out sites is not well known. The requirement of sea ice for the successful breeding of ringed seals has not been sufficiently assessed. Land breeding of grey seals is not monitored regularly in most Baltic states. The effects of interspecific competition on distributions are not known. Range contraction can be conceptualised as having three stages (Bates et al., 2014), i.e. performance decline, population decrease, and local extinction, all of which should be studied. For example, studies on performance decline, such as physiological conditions that reduce reproductive potential (Helle, 1980; Jüssi et al., 2008; Kauhala et al., 2017, 2019) are important (Bates et al., 2014). Breeding success of Baltic Sea ringed seals in normal winters is poorly known, as the lairs in pack ice snowdrifts are rarely found. Likewise, observations of the effects of poor ice conditions on the breeding success of ringed seals in mild winters are very limited, but the lack of protection from breeding lairs against harsh weather and predators is assumed to be highly negative.

6.23.3 Waterbirds

The complex interaction between many primary parameters affected by climate change makes it hard to identify which environmental changes are actually causing changes in waterbird populations. It is currently not known in detail how shifts in distribution and timing of migration match the availability and quality of food, and thus, the importance of potential temporal mismatches between food availability and requirements is unknown. In addition, changes in waterbird distribution are likely to alter inter- and intraspecific competition. Resolving these issues requires investigation of effects at other levels of the food web (e.g. loss of bivalves from areas of reduced salinity, species, and size–class composition of fish communities) and their consequences for waterbirds. So far, knowledge of climate change effects on waterbirds in the Baltic Sea are mostly restricted to ducks (including diving and dabbling ducks), with much less known for other quantitatively important components of the waterbird community, i.e. divers, grebes, waders, gulls, and auks. Interactions between fish and piscivorous waterbirds in particular need more attention. Responses to climate change are likely to vary between waterbird species and groups. There is still little information on which species are mostly affected (negatively or positively) by changes in climatic conditions, and the uncertainty is, therefore, large in terms of how species in future waterbird assemblages will interact and the consequences for the functioning of the Baltic Sea. To gain a better understanding on how single species (or groups with similar ecology that are often closely related) will respond to climate change is critical for projecting effects of climate change on waterbirds around the Baltic Sea.

6.23.4 Marine food webs

Some changes observed in marine food webs have been partly attributed to warming, brightening, and sea ice decline on long timescales. Other drivers, such as eutrophication or fisheries, may, however, predominate, and many records are too short to allow the attribution of the observed changes to climate change. Although effects of warming, OA, and dissolved organic matter on some ecosystem functions have been identified in mesocosm experiments, changing food web interactions are still impossible to project. It is, however, important to include the marine biosphere in management strategies for tackling the complex interactive aspects of climate-change-related effects on the marine ecosystem and human adaptations to them (Andersson et al., 2015; Stenseth et al., 2020).