The primary sources for measuring mean sea level changes are tide gauges and radar altimetry from satellites. Both measure two different quantities, namely relative and absolute (geocentric) sea levels. Absolute sea level refers to the height of the sea surface relative to a geocentric reference, such as the reference ellipsoid, and is derived from satellite altimetry or GNSS-controlled tide gauges (where GNSS stands for global navigation satellite system; Schöne et al., 2009, 2011). Relative sea level refers to the height of the sea surface relative to the seafloor (or a local benchmark on land) and is derived from tide gauges or sea level reconstructions. Thus, contrary to the absolute sea level, the relative sea level is influenced both by variations in the height of the sea surface and by land uplift or subsidence.

The tide-gauge network of the Baltic Sea is one of the most densely spaced and longest-running networks in the world. Many stations have been in continuous operation since the early 19th century, and some stations have provided monthly averages for over 200 years (Ekman, 2009; Bogdanov et al., 2000; Kowalewska-Kalkowska and Marks, 2011). There are about 45 operational tide gauges with more than 60 years of data and with very good coverage of the western, southwestern, and northern Baltic Sea coasts (Hünicke et al., 2015). Eight additional gauges with data since 1961 in Latvian waters were recently added to this dataset (Männikus et al., 2019). Monthly averaged data are often provided to, and are open-access through, the Permanent Service for Mean Sea Level (PSMSL; e.g., Holgate et al., 2013). The Copernicus Marine Environment Monitoring Service collects data from a range of national stations at an hourly resolution with daily updates, although with varying quality control. The EMODnet Physics portal provides access to both data sources and, thus, combines a substantial number of long-term monthly and higher-frequency Baltic sea level measurements. The tide gauges measure relative to specific reference levels that differ between Baltic countries for historical reasons (Ekman, 2009). There are efforts to connect Baltic Sea gauges to a common datum, such as the Baltic Sea Chart Datum 2000 (BSCD2000; Schwabe et al., 2020). Additionally, uncertainties in local reference levels are introduced by factors including different measuring techniques and sampling frequencies (Ekman, 2009), changes in the reference points due to relocation of benchmarks, or anthropogenic changes in the tide-gauge surroundings, such as coastal or port development. Examples for the case of the Kronstadt gauge are described in Bogdanov et al. (2000). Phenomena occurring on timescales of years to decades such as subsidence (e.g., the sinking of piers due to unstable foundations or land sinking due to groundwater or gas extractions) can be other sources of uncertainties (Hünicke et al., 2015). Subsidence can further result from processes such as the compaction of unconsolidated alluvial soils, oxidation of organic materials, or increased surface load due to construction activities (e.g., Pelling and Blackburn, 2013).

Radar satellite altimetry from different satellite missions is available continuously from 1991 to the present (e.g., Schöne et al., 2010). Currently, the longest and most sound dataset is derived from a combination of the consecutive dual-frequency missions TOPEX/Poseidon, Jason-1, Jason-2, and Jason-3 (Legeais et al., 2018). The orbit repeats itself every 10 d, allowing for the construction of continuous time series since 1993. The missions form a pattern of ascending and descending orbits with an average cross-track distance of 100 km (at 60∘ N). Starting in 1991 with the European ERS-1 mission (followed by ERS-2, and the dual-frequency ENVISAT (European Space Agency – ESA) and AltiKa (India/France) mission), the Baltic Sea is also mapped every 35 d with an average cross-track distance of 55 km (at 60∘ N). With the launch of ESA's CryoSat-2 in 2010, a new altimetry concept was established using a synthetic aperture interferometric radar altimeter (SAR/SARin) to specifically map ice parameters but also sea surface heights closer to the coastlines. In 2016, the European Copernicus program launched a mission on Sentinel-3A (followed in 2018 by Sentinel-3B) carrying forward the uninterrupted altimetry program with 27 d repeat periods. The recent launch of Sentinel-6A Michael Freilich in November 2020 will continue the important missions with unprecedented accuracy.

Long-term changes in relative mean sea levels in the Baltic Sea are dominated by GIA, global sea level change, and other regional- to local-scale components and their interaction. This includes factors such as thermo- and halosteric contributions, or effects from long-term changes in wind and surface air pressure, currents in the North Atlantic that affect large-scale sea level, or variations in freshwater input in the Baltic Sea (Fig. 3).

Global mean sea level rise since the beginning of the last century was estimated from tide-gauge records with rates ranging between about 1 and 2 mm yr-1 (e.g., Oppenheimer et al., 2019). For the era of continuously operated satellite radar altimeters beginning in 1991, higher estimates ranging between about 3 and 4 mm yr-1 are reported (e.g., Nerem et al., 2018; Oppenheimer et al., 2019). These trends cannot be directly compared because of the strong decadal variability inherent in the records (e.g., Fig. 5b; Albrecht et al., 2011) and their different spatial representativeness. As the majority of tide gauges is located at the coast, their data are more representative of coastal changes. Satellite data typically reflect open-ocean changes and variability. Also, inhomogeneous data, comprising a few tide-gauge records in the 19th and early 20th centuries and satellite data with nearly global coverage in the 21st century, hamper quantification and comparability of trends (Jevrejeva et al., 2008).

For the Baltic Sea, an increase in absolute mean sea level of about 3.3 mm yr-1 was estimated based on available multi-mission 1992–2012 satellite altimetry data (Stramska and Chudziak, 2013). This figure is broadly consistent with the global average rate over that period. Current (1993–2015/2017) altimetry-derived Baltic Sea mean sea level trends are still comparable with the global average (Madsen et al., 2019a; Table 1). Across the Baltic Sea, the rates of absolute mean sea level rise vary between about 2 and 3 mm yr-1 in the southwestern parts and about 5 and 6 mm yr-1 in the northern parts for the period from 1995 to 2019 (Passaro et al., 2021; Fig. 5a).

Relative mean sea level changes and their secular trends are strongly affected by vertical land movements and vary considerably across the Baltic Sea. Geologically, the Baltic Sea region is divided into the uplifting Fennoscandian Shield in the north and the subsiding lowland parts in the south (Harff et al., 2007). Land uplift rates in the northern parts are in the order of several millimeters per year and are comparable to the climatically induced rates of sea level rise in the 21st century. Relative mean sea level trends in the Baltic Sea show a corresponding north–south gradient, reflecting these crustal deformation rates due to GIA. In the northern parts, the relative mean sea level decreases with a maximum rate of about 8.2 mm yr-1 in the Gulf of Bothnia (Hünicke et al., 2015). This corresponds to the area with maximum GIA-induced crustal uplift (Peltier, 2004; Lidberg et al., 2010). In the southern Baltic Sea, the relative mean sea level increases at a rate of about 1 mm yr-1 with a gradient in a northeasterly direction (Richter et al., 2012; Groh et al., 2017). These findings are supported by numerous studies analyzing Baltic relative mean sea level trends on a national basis (Suursaar et al., 2006a; Dailidienė et al., 2012; Männikus et al., 2019). Different observation periods and analysis techniques hamper comparison to some extent. A more comprehensive figure on Baltic sea level trends and variability was provided by Madsen et al. (2019a), who used a statistical approach to combine data from century-long tide-gauge records with results from hydrodynamic modeling based on atmospheric reanalysis data and then compared the results with those from satellite altimetry records (Table 1). The analysis again emphasizes that the rate of sea level rise strongly depends on the period considered but that present-day rates are higher than those previously observed. This is also supported by an analysis of trends over successive 40-year periods at the Warnemünde tide gauge that shows pronounced variability in decadal sea level trends with present trends in the upper range of the observed values (Fig. 5b).

Atmospheric forcing in the form of wind and precipitation can affect the basin-average sea level (through changes in the total volume of water) as well as the internal distribution of water volume within the Baltic Sea basin. Winds, more particularly the strength of the westerly winds, modulate the exchange of water masses with the North Sea (Gräwe et al., 2019). During periods of weak or moderate wind conditions, a two-layer system of in- and outflow prevails in the Danish straits, whereas during periods of strong winds, in- or outflow occurs across the entire water column, the strength of which is controlled by the sea level gradient between the Kattegat and the western Baltic Sea (Sayin and Krauss, 1996; Stigebrandt, 1983; Mattsson, 1996). The alignment of the prevailing wind directions in this region with the geographical orientation of the connecting straits between the Baltic and the North seas favor conditions in which stronger than normal westerly winds push water masses into the Baltic Sea raising the overall sea level (with the inverse being true for weaker than normal westerly winds).

On average, there is a strong salinity increase from the northeast to the southwest of the Baltic Sea. This gradient results from the along-basin separation between the main saltwater (Danish straits) and freshwater (rivers discharging into the Gulf of Finland and the Bothnian Sea) sources. It leads, through density, to a corresponding sea level gradient which, in combination with the volume of the freshwater input and the effects of the prevailing westerly winds, gives rise to a corresponding sea level variation of about 35–50 cm across the Baltic Sea (Ekman and Mäkinen, 1996). Kniebusch et al. (2019) showed that there is pronounced decadal variability in the freshwater forcing, leading to a corresponding variability in the along-basin salinity gradient which, in turn, affects the sea level gradient. Over the period from 1900 to 2008, Kniebusch et al. (2019) also detected a trend in the salinity gradient but found the trend to be sensitive to the exact period chosen and small compared with the decadal variability. Likewise, precipitation anomalies affect the distribution of the Baltic mean sea level through corresponding changes in salinity and water density (Hünicke and Zorita, 2006). Currently, no robust long-term trends in the freshwater budget are observed (Johansson, 2016; Rutgersson et al., 2014) that would add to presently observed rates of Baltic sea level change.

The relation between Baltic sea level variability and the state of the large-scale atmospheric circulation has been the subject of numerous studies (Kahma, 1999; Johansson et al., 2001; Lehmann et al., 2002; Dailidienė et al., 2006; Hünicke and Zorita, 2006; Suursaar et al., 2006a; Johansson and Kahma, 2016; Chen and Omstedt, 2005; Omstedt et al., 2004). Mostly, these studies have focused on relations between the NAO

The North Atlantic Oscillation (NAO) basically describes a meridional pattern in sea level pressure (SLP) with higher than normal SLP around the Azores and lower than normal SLP over Iceland and vice versa. Variability in this pattern is physically linked to the intensity of the westerlies in the European region.

The contribution of the different mechanisms behind current sea level trends can be also estimated from the analysis of simulations with regional ocean models driven by observed atmospheric and global mean sea level forcing (Gräwe et al., 2019). This approach has the advantage that the effects of land movement on the Baltic sea level are explicitly neglected so that the contribution from other factors can be disentangled. In the simulation of Gräwe et al. (2019), the Baltic sea level rose at a rate of about 2 mm yr-1 over the past 50 years, a rate that is slightly higher than the global average. Most of this sea level rise in the Baltic Sea was caused by a corresponding increase in the sea level in the North Atlantic Ocean. Model results along with data from Latvian waters (Männikus et al., 2020) further suggest a heterogeneous pattern of sea level rise, with larger rates in the northern Baltic Sea and smaller rates in the southeastern Baltic Sea, that appears to be the result of a poleward shift of atmospheric pressure systems (Gräwe et al., 2019).

For the future, global mean sea level rise is expected to have the largest impact on future Baltic sea level changes (Grinsted, 2015; Hieronymus and Kalén, 2020). It is anticipated that most of the future Baltic absolute sea level rise will be strongly linked with corresponding large-scale changes in the North Atlantic and the factors modulating these changes. These factors are mainly the thermal expansion of the water column, contributions from melting of the Antarctic ice sheet (Grinsted, 2015), and imprints from the variability and change in the Atlantic Meridional Overturning Circulation (Börgel et al., 2018).

For absolute sea level trends, the relative contribution from future melting of the Greenland and Antarctic ice sheets is the major source of spatial variation across the Baltic Sea. Melting of ice from the large ice sheets leads to changes in the Earth's gravitational field, rotation, and crustal deformation – the patterns of which are called sea level fingerprints (Plag and Jüttner, 2001). For any given place on Earth, these fingerprints are specific to the location of the ice melt. The Baltic sea level is substantially more sensitive to melting from Antarctica than from Greenland. For example, if an ice volume melts from Antarctica that is equivalent to an increase in the global mean sea level of 10 cm, the increase in the absolute mean sea level in the Baltic Sea would be about 11 cm (+10 %). If the same volume melts from the Greenland ice sheet, the response of the Baltic Sea level would be smaller and spatially more heterogeneous. Here, the values would vary around zero such that the southern Baltic Sea would experience an increase in absolute sea level of about 2 cm (20 % of the globally averaged sea level rise from Greenland melting), while the northern Baltic Sea would expect a corresponding sea level fall (Hieronymus and Kalén, 2020).

For the Swedish coast, Hieronymus and Kalén (2020) provided sea level projections accounting for such variations. When compared to earlier estimates, this resulted in lower projections in 2100. However, as more recent estimates of potential future contributions from Antarctic melting increased (Oppenheimer et al., 2019), Hieronymus and Kalén (2020) still found their projections to be broadly consistent with earlier estimates, because their more careful treatment of spatial inhomogeneities was balanced by the larger estimates for Antarctic melting.

Stronger winds and increased runoff can further contribute to future Baltic sea level rise in the order of some centimeters (Meier, 2006; Hünicke, 2010; Johansson, 2014; Karabil, 2017; Pellikka et al., 2018). Future changes in the Baltic Sea freshwater budget may also affect the regional distribution of the Baltic mean sea level (Hünicke et al., 2017). If present links between atmospheric forcing and the Baltic mean sea level are extrapolated into the future, their contribution to the future Baltic mean sea level can be estimated from climate projections. Presently, such contributions are thought to be minor (Weisse and Hünicke, 2019). As GIA continues and will persist to dominate relative mean sea level changes in the northern parts of the Baltic Sea, these areas are expected to see a continued, although decelerated, decrease in relative mean sea level in the future, while the southern parts are expected to experience a relative mean sea level rise slightly exceeding the global average (Räisänen, 2017).

One of the expected implications of anthropogenic climate change would be an acceleration in sea level rise over time. So far, global mean sea level rise seems to have accelerated over the 20th century (Nerem et al., 2018; Oppenheimer et al., 2019; Dangendorf et al., 2019), although this is still a topic of ongoing debate (Kleinherenbrink et al., 2019; Veng and Andersen, 2020). Robust detection of acceleration is hampered to some extent by changes in the measurement system. While most of the data over much of the 20th century originate from coastal tide gauges mostly in the Northern Hemisphere, satellite altimetry has provided a nearly global figure mostly from the open ocean since 1991. Differences in estimates of satellite intra-mission biases also have substantial effects on estimated acceleration rates (Kleinherenbrink et al., 2019). As tide gauges in the Baltic Sea provide some of the longest, best quality controlled, and more homogeneous records, they may provide some support in assessing acceleration. From a statistical analysis of long Baltic sea level records, Hünicke and Zorita (2016) could indeed detect an acceleration in sea level rise. While the tests were not powerful enough to detect a significant acceleration in most of the records taken individually, results are robust for the set of gauges against different definitions of acceleration. The overall magnitude of the acceleration is small and, if continued unchanged over the whole 21st century, would add just a few centimeters to the sea level rise. The latter is, however, unlikely, as an increase in the acceleration of sea level rise is expected for the future – in particular when the large ice sheets melt more rapidly.

Acceleration of Baltic sea level rise also displays a spatial pattern with accelerations in the northeastern parts 3 times as large as those in the southwest. The spatial structure is compatible with what would be expected from an expected deceleration of GIA. However, applying the theoretical deceleration calculated with models of the dynamics of the Earth's crust (Spada et al., 2014), the resulting order of magnitude is still too small to explain the apparent acceleration of relative coastal sea level. It has to be kept in mind that most tide gauges are affected by the rather strong noise of local vertical movements; thus, the detection of a second-order change, like acceleration, is challenging.

Sea level extremes are measured using the same primary devices as for mean sea level: tide gauges and radar altimetry. For wind waves, data from visual observations, wave buoys, or radars mounted on platforms are also common. Spatial coverage of instrumental wave measurements in the Baltic Sea is scarce and limited to a few regions. So far, no long-term data exist from the central and eastern Baltic Proper and the Gulf of Riga (Suursaar et al., 2012). In the northern parts, measurements are often limited to ice-free periods, as buoys are usually removed during winter to prevent damage from ice. The homogeneity of data can be problematic due to effects from the relocation of measurement sites or the replacement of instruments (Hünicke et al., 2015). Wave data from measurements are often used in combination with reconstructions from numerical wave hindcasts (Cieślikiewicz and Paplińska-Swerpel, 2008; Nikolkina et al., 2014; Björkqvist et al., 2018; Soomere et al., 2012), where they provide a reasonable characterization of open-sea wave fields.

Baltic sea level extremes occur over a wide range of spatial and temporal scales. Contributions can arise from phenomena acting at small scales, such as wind-wave run-ups occurring at local scales with variations in the range of seconds, up to phenomena such as variations in the volume of the entire Baltic Sea characterized by formation timescales of up to a few weeks and even longer persistence (Soomere and Pindsoo, 2016). In between these ranges, other phenomena such as storm surges contribute to sea level extremes. All of these phenomena are generated mostly by meteorological and to some extent by astronomical factors (Weisse and Hünicke, 2019). From a climate perspective, this indicates that any relevant change in meteorological forcing may be associated with corresponding changes in Baltic sea level extremes.

The most prominent and most relevant phenomena contributing to sea level extremes in the Baltic Sea are storm surges, wind waves, and a preconditioning that leads to increased water volumes and, thus, sea levels in the entire Baltic Sea before the onset of a storm (Suursaar et al., 2006b; Madsen et al., 2015). Preconditioning is associated with periods of prevailing westerly winds that increase the sea level gradient across the Danish straits. In turn, the increased sea level gradient leads to higher inflow and higher Baltic Sea water volumes (Samuelsson and Stigebrandt, 1996). While the long-term average barotropic outflow from the Baltic Sea is about 7 km3 d-1 or 80 000 m3 s-1 (Winsor et al., 2001), transports across the Danish straits can reach values of up to about 45 km3 d-1 in both directions, which corresponds to a sea level change of about 12 cm d-1 over the entire Baltic Sea (Mohrholz, 2018). Major inflow events are associated with typical volumes in the order of about 100 km3, corresponding to a Baltic sea level increase of about 24 cm (Matthäus and Franck, 1992). Typically, such variations have timescales of about 10 d or longer (Soomere and Pindsoo, 2016) while atmospheric variability on shorter timescales primarily leads to a redistribution of water masses within the Baltic Sea basin (Kulikov et al., 2015) or between the Baltic Proper and the Gulf of Riga (Männikus et al., 2019).

Storm surges refer to changes in the sea level in coastal waters that are primarily caused by the action of spatially varying atmospheric pressure on the sea surface (Pugh and Woodworth, 2014) and by wave setup in specific locations (Soomere et al., 2013). They represent a substantial threat to the low-lying coastal areas of the Baltic Sea, in particular in the southwestern regions (Wolski et al., 2014), the Gulf of Finland (Suursaar and Sooäär, 2016; Averkiev and Klevannyy, 2010), the Gulf of Riga (Suursaar and Sooäär, 2016; Männikus et al., 2019; Suursaar et al., 2006b), and the Gulf of Bothnia (Averkiev and Klevannyy, 2010). Comparing 50 years of hourly sea level data from gauges across the entire Baltic Sea, Wolski and Wiśniewski (2020) showed that Pärnu in the Gulf of Riga and Kemi in the Gulf of Bothnia are particularly prone to storm surges. In their analysis, these two gauges were characterized by both the highest average number of surges per year and the highest average number of hours per year with extreme sea levels. Regarding height, a surge in November 1824 in the Gulf of Finland was responsible for the highest recorded water level in the Baltic Sea (4.21 m above tide-gauge zero in St. Petersburg; Wolski and Wiśniewski, 2020). In the western Baltic Sea, a storm in November 1872 caused water levels to exceed 3 m at many gauges (Feuchter et al., 2013; Wolski and Wiśniewski, 2020). For the Gulf of Riga and the western Baltic Sea, respective storm surge heights of around 2 m and 1–1.5 m are typical (Wolski and Wiśniewski, 2020).

Storm surges in the Baltic Sea can last from several hours to almost a day (Wolski and Wiśniewski, 2020). Because of the seasonal cycle in wind speed, storm surges are highest and most frequent during fall and winter (Weidemann, 2014). The presence of sea ice may substantially reduce the effectiveness of wind in generating storm surges. In winter in the Gulf of Bothnia, the piling up of water is strongly suppressed by the existence of sea ice (Zhang and Leppäranta, 1995). When preconditioning exists – that is, high Baltic Sea water volumes – even moderate wind and wind surges may lead to coastal sea level extremes (e.g., Weisse and Weidemann, 2017).

Wind waves in the Baltic Sea show a pronounced seasonal cycle, with higher values in winter and lower values in summer. This behavior is associated with a corresponding seasonal cycle in wind speed (Björkqvist et al., 2018) and leads to extensive variations in the wave energy flux (Soomere and Eelsalu, 2014). Generally, waves are higher in open waters than in coastal waters. To date, the highest measured waves in the Baltic Sea have been reported from a wave buoy in the northern Baltic Proper. During a storm in December 2004, this buoy recorded a significant wave height of 8.2 m (Björkqvist et al., 2018; Tuomi et al., 2011). During windstorm Gudrun in 2005, measured significant wave heights reached 7.2 m in the Baltic Proper and 4.5 m in the Gulf of Finland. For this storm and away from sensors, model simulations suggested the existence of extreme significant wave heights of up to 9.5 m off the northwestern coast of Latvia (Soomere et al., 2008). More recently, very high waves with significant wave heights of up to 8.1 m have also been recorded in the Bothnian Sea (Björkqvist et al., 2020). However, in the Bothnian and the northern Baltic seas, the seasonal presence of sea ice typically modifies the wave climate and limits wave heights during the freezing season (Tuomi et al., 2011).

In the coastal zone, the height of the wave extremes is considerably smaller, but other wave-related processes such as wave setup or run-up (swash) can make substantial contributions to short-term sea level extremes. Wave setup refers to a wave-induced increase in the mean water level caused by the release of momentum from the dissipation of a long sequence of breaking waves, and swash is caused by the run-up of single waves on the beach (Melet et al., 2018). For example, for some coasts in the Gulf of Finland (Soomere et al., 2013, 2020) and along the shores of the West Estonian archipelago (Eelsalu et al., 2014), potential wave setup may reach values of up to 70–80 cm. Moreover, the shape of the shoreline and irregular bottom topography can strongly modify local wave conditions (Tuomi et al., 2012, 2014), and wave setup and run-up may both have a strong influence on the level of flooding and erosion at the shore (Dean and Bender, 2006). Thus, coastal wave climate, its impacts, and its long-term changes strongly depend on location.

Changes in the Baltic Sea volume, storm surges, and extreme sea levels are tightly coupled. As increased volumes provide preconditioning for extremes in the entire sea (Pindsoo and Soomere, 2020), higher Baltic Sea volumes generally lead to higher extremes under otherwise similar storm conditions (Weisse and Weidemann, 2017). More specifically, westerly storms not only cause surges on the eastern Baltic Sea coast but also increase the volume of the sea. During subsequent storms, extreme sea levels are then higher than without preconditioning.

Two other noticeable effects may further contribute to Baltic sea level extremes: seiches and meteotsunamis. In the Baltic Sea, seiches with periods of up to tens of hours and e-folding times of up to 2 d may develop under certain atmospheric conditions (Leppäranta and Myrberg, 2009). Details of these oscillations are debated and are still not fully understood. From numerical studies, Wübber and Krauss (1979) proposed a series of basin-wide seiches with periods of up to 31 h. Other authors have argued that the existence of such basin-wide oscillations is not entirely supported by data. Thus, they have suggested that such sea level oscillations in the Baltic Sea could alternatively be considered as an ensemble of weakly coupled local seiches with periods between 17 and 27 h in the Gulf of Riga, the Gulf of Finland, and the Belt Sea (Jönsson et al., 2008). Oscillations with similar periods also occur in the West Estonian archipelago (Otsmann et al., 2001). For the Gulf of Riga, Suursaar et al. (2002) also note the existence of a 5 h barotropic seiche. When favorably coupled with storm surges or in resonance with atmospheric forcing, such oscillations may contribute to very high sea level extremes at the coast (Suursaar et al., 2006a; Weisse and Weidemann, 2017; Wolski and Wiśniewski, 2020). Meteotsunamis are generated by moving atmospheric disturbances that trigger resonant sea level fluctuations. While not extremely frequent in the Baltic Sea, cases were described by Pellikka et al. (2014, 2020) for the Gulf of Finland or by Holfort et al. (2016) for the western Baltic Sea. Amplitudes may be in the order of 1 m (Pellikka et al., 2014, 2020).

The contribution from tides to the height of Baltic sea level extremes is mostly small, as the connection to the open ocean and, thus, co-oscillation is limited. Overall, tidal ranges are mostly between about 2 and 5 cm (Witting, 1911, as cited in Leppäranta and Myrberg, 2009). In the western sea areas, tidal ranges of up to 10–30 cm are observed (Defant, 1961, as cited in Leppäranta and Myrberg, 2009). Due to resonance, tidal ranges of up to 20 cm are also found near St. Petersburg at the end of the Gulf of Finland (Medvedev et al., 2013). While co-oscillation is limited, locally generated tides, albeit small, may also contribute in the same order of magnitude at some places (Schmager et al., 2008). In the western Baltic Sea, tidal conditions are predominantly semidiurnal, while in the Gulf of Finland and the Gulf of Riga, diurnal tides prevail (Medvedev et al., 2013; Schmager et al., 2008). While tidal ranges are generally small, they still have noticeable effects on the return periods of sea level extremes (Särkkä et al., 2017).

Variability and long-term changes in Baltic extreme sea levels can occur for various reasons but are primarily linked to changes in the relative mean sea level and atmospheric conditions. Relative mean sea level changes will modify the base upon which other atmospheric drivers of extremes will act, although the response is not necessarily linear (e.g., Arns et al., 2015). For example, for the same wind field, and hence surge levels, higher extremes are expected under higher relative mean sea levels. Also, changes in the driving atmospheric conditions will lead to changes in the statistics of waves, surges, and so on, which in turn will affect the sea level extremes. Moreover, nonlinear interaction between the contributions, potentially nonstationary behavior of the population of extremes (Kudryavtseva et al., 2018), and local effects such as from bathymetry or the shape of the coastline can make overall effects on sea level extremes highly nonadditive (e.g., Arns et al., 2015).

Based on tide-gauge data for different periods, several studies have revealed trends in Baltic sea level extremes. These trends were found to originate mainly from a corresponding change in mean sea level (Marcos and Woodworth, 2017; Ribeiro et al., 2014; Barbosa, 2008) or an increase in the magnitude of the preconditioning (Soomere and Pindsoo, 2016; Pindsoo and Soomere, 2020). For the periods from 1960 onwards, Marcos and Woodworth (2017) showed that the height and duration of sea level extremes increased or decreased at sites with a respective increasing or decreasing relative mean sea level and that these trends mostly disappear when the mean sea level signal is removed. Only for some of the eastern and northernmost stations do the trends indicate some contribution from corresponding changes in the large-scale atmospheric circulation and regional wind patterns (Barbosa, 2008; Ribeiro et al., 2014). Using data from a hindcast simulation ignoring GIA, Pindsoo and Soomere (2020) found increases in the height of Baltic sea level extremes for the period from 1961 to 2004/2005. These increases were strongest in the Gulf of Finland, the Gulf of Riga, and the eastern Baltic proper (Fig. 7). Pindsoo and Soomere (2020) concluded that the increases along the Swedish coast and the Gulf of Bothnia were almost entirely a result of global mean sea level rise and an increase in the maximum water volume of the entire sea. Conversely, they noted substantial contributions from stronger local storm surges to explain the higher values along the eastern shorelines.

Long-term changes in the wave climate may further contribute to changing extremes through corresponding adjustments of wave transformation in the surf zone (e.g., wave setup and swash). So far, there is no conclusive large-scale figure, and results vary strongly depending on the period and region. For the Arkona Basin, Soomere et al. (2012) analyzed wind-wave variability and trends based on 20 years of observation and a 45-year wave hindcast. They concluded that the wave height in this area exhibits no long-term trend but reveals modest inter-annual and substantial seasonal variations. For shorter periods (1993–2015), estimates from satellite altimetry data suggest a slight increase in annual mean significant wave height in the order of 10 % (about 12 cm or 0.005 m yr-1). Spatially, wave height increased in the central and western parts of the sea and decreased in the eastern parts (Kudryavtseva and Soomere, 2017). Nikolkina et al. (2014) analyzed a multi-ensemble wind-wave hindcast covering the entire Baltic Sea using different atmospheric forcing and periods (1970–2007 and 1957–2008). While these authors found the hindcasts consistently describing the known spatial patterns with relatively severe wave climate in the eastern parts of the Baltic proper and its subbasins, they could not infer consistent conclusions on long-term changes mainly due to differences in the atmospheric forcing used in the model simulations.

As for mean sea level, extreme sea levels are linked and correlated with the large-scale atmospheric circulation and its variability. The NAO is often used to characterize the state and the variability of the large-scale atmospheric circulation (Johansson, 2014; Marcos and Woodworth, 2018). There are several mechanisms that contribute to the link between Baltic sea level extremes and the NAO. The positive correlation between the phase of the NAO and mean sea level in the northwestern European shelf seas (Woolf et al., 2003) suggests that higher than normal mean sea levels occur during positive phases of the NAO. This would lead to an increase in the baseline sea level upon which wind- and pressure-induced extremes will act. In addition, there is also a positive relationship between the phase of the NAO and the frequency of westerly winds. Increased frequencies of westerly winds may lead to a higher than normal water volume in the Baltic Sea on average, which again would increase the baseline. Eventually, potential relationships between changes in the NAO and regional wind patterns would contribute to corresponding changes in wind surges and waves.

Using tide-gauge data, Johansson (2014) and Marcos and Woodworth (2018) showed that the positive correlation between the NAO and Baltic sea level extremes persisted even when long-term mean sea level changes were removed. This indicates that the NAO influences on Baltic sea level extremes are not only limited to the effects of changes in the mean but have contributions from NAO effects on Baltic Sea volume and/or locally generated wind surges and waves. This conclusion was further supported by a model study in which a coupled North and Baltic Sea model was forced solely by wind and sea level pressure from 1948 to 2011, thereby explicitly excluding effects from global mean sea level rise and rising temperatures (Weidemann, 2014). In this experiment, periods of high water volume in the Baltic Sea occurred more often during positive NAO phases and vice versa, and lower wind speeds were generally needed to sustain higher sea level extremes when the volume was above normal (Weisse and Weidemann, 2017).

Future changes in sea level extremes in the Baltic Sea crucially depend on two factors: future changes in relative mean sea level and future developments in large-scale atmospheric conditions associated with changing wind patterns. For some regions, changes in the frequency or thickness of sea ice may also have an impact. Relative sea level changes will strongly vary across the Baltic Sea because of the existing spatial gradient in GIA and the spatial inhomogeneity associated with the uncertain relative contributions of melting from Antarctica and Greenland (see Sect. 2.1.2). 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). GIA is expected to continue at rates similar to those observed over the last century. Absolute mean sea levels are expected to rise in the entire Baltic Sea, but exact rates are uncertain and depend on the models, scenarios, and periods considered (see discussion in Sect. 2.1.2; Grinsted, 2015; Hieronymus and Kalén, 2020).

Potential future changes in long-term mean and extreme wind speeds are highly uncertain (Räisänen, 2017). In the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC), a poleward shift of Northern Hemisphere midlatitude storm tracks and an ensemble average increase in the NAO index are reported as possible future developments (Kirtman et al., 2013). Because of the large internal variability and the large variability among models and scenarios, there was, however, only medium confidence in these projected changes (Kirtman et al., 2013). Recent studies indicate that the poleward shift of the westerlies is most pronounced in summer and less obvious in other seasons (Zappa and Shepherd, 2017). Because storm surges in the Baltic Sea are primarily a winter and fall phenomenon, this implies that contributions from changes in atmospheric circulation to future Baltic Sea storm surge climate remain uncertain. This is in agreement with climate model simulations investigated for the Second Assessment of Climate Change for the Baltic Sea Basin (BACC II Author Team, 2015) that were highly inconsistent for projected changes in wind speeds at the end of the 21st century (Christensen et al., 2015; BACC II Author Team, 2015).

Despite these uncertainties, some storm surge and wind-wave projections seem to report relatively robust changes towards 2100. Using data from a multi-model ensemble, driven with atmospheric data from eight different climate models and the two Representative Concentration Pathway (RCP) emission scenarios RCP4.5 and RCP8.5, Vousdoukas et al. (2016) projected an increase in storm surge heights in the order of about 5 %–10 % across the entire Baltic Sea towards the end of the 21st century. For wind waves, Groll et al. (2017) reported an increase of about 5 %–10 % in mean and extreme wave heights using data from two realizations of two emission scenarios. In contrast, using a substantially larger ensemble consisting of data from six climate models and three emission scenarios, Dreier et al. (2021) found no consistent long-term changes in wave climate along the German Baltic Sea coast, and their projected changes in extreme wave height varied between about -10 % and +6 %. This again suggests that projections of future wind waves and storm surges in the Baltic Sea are still highly dependent on the atmospheric scenario, the climate model, and the realization used for the projection. Nevertheless, because of the larger effect of changing mean sea levels on the extremes, regions with expected increases in relative mean sea level are still highly likely to experience an increase in sea level extremes.

Coastal erosion and sedimentation as well as longer-term shoreline changes tend to be researched at local, sediment compartment, or regional scales, where particular combinations of wind, wave, sediment characteristics and availability, and sea level variations and extremes at a range of scales can be considered. Generally, remote sensing data have been widely applied to investigate coastline change at various spatial and temporal scales. Different sources of remote sensing data include satellite images (e.g., Tiepold and Schuhmacher, 1999), aerial photographs (e.g., Furmanczyk et al., 2011; Dudzinska-Nowak, 2017), terrestrial (Eelsalu et al., 2015) and airborne laser scanning data (e.g., Hartleib and Bobertz, 2017), and orthophoto maps (e.g., Zhang et al., 2017; Dudzinska-Nowak, 2017). Such data were used, for example, to quantify the rates of coastline change over the past decades along the sandy southern Baltic Sea coast and more recently for the Russian waters (Ryabchuk et al., 2020). Also, historical maps with scales between 1:250000 and 1:5000 dating back to the 1820s have provided useful information on the coastline change in the Baltic Sea at a longer timescale (Deng et al., 2017a; Hartleib and Bobertz, 2017). As tides are small in the Baltic Sea, the baseline of the seaward-most foredune or cliff, which represents the cumulative effect of deposition or erosion, has been used as an indicator to quantify coastline change (e.g., Dudzinska-Nowak, 2017; Zhang et al., 2015).

The Baltic Sea region can be geologically divided into the uplifting Fennoscandian Shield in the north and the subsiding lowlands in the south (for details, see Fig. 2.8 in Harff et al., 2007). Coastline change is characterized by a corresponding north–south gradient (Fig. 8) along which the main driving force gradually shifts from GIA to atmospheric and hydrodynamic forcing. Vertical land movement ranges from uplifts of up to almost 9 mm yr-1 in the north to subsidence of up to 2 mm yr-1 in the south. Since the onset of the Holocene, these vertical movements have caused a persistent marine regression at the northern coasts and a marine transgression along the southern coasts of the Baltic Sea (Harff et al., 2007, 2011). The transitional area between the northern uplift and the southern subsidence is located at the southern coast of the Gulf of Finland (Rosentau et al., 2017).

About half of the shores of the Baltic Sea are comprised of either extremely resistant bedrock or relatively slowly changing cliffs of limestone or morainic materials. Due to this geology and the ongoing GIA, coastal erosion is not regarded as an issue of concern in Finland or for the majority of the Swedish coast (Pranzini and Williams, 2013). The other half of the Baltic Sea shores located primarily in the south and the east are sedimentary and are susceptible to coastline change due to erosion and accretion. This part of the coast is characterized by a series of barrier islands and sandy dunes connected with soft moraine cliffs. A typical cross-shore dune profile at the southern Baltic coast features an established or a series of foredune ridges with typical heights between 3 and 12 m a.m.s.l. (meters above mean sea level) (Łabuz et al., 2018). At the backshore behind the established foredune ridges, drifting or stabilized dunes in transgressive forms 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).

As with other (semi-)enclosed seas, dominant processes relevant for erosion and accretion on Baltic Sea shores differ from those described in the classic coastal process literature. As on other shorelines, erosion is generally largest during storms when nearshore waves are longest, highest, and thus most energetic. Differences arise on Baltic Sea shorelines, however, from the lack of substantial tides, the lack of long wind-generated swell, the frequently occurring high angles of storm waves approaching the shore, and partially the presence of a seasonal ice cover. Wave energy and direction are the dominant drivers of sediment transport, erosion, and accretion in the Baltic Sea. Regional properties of the energy supply and the wave-driven transport are described in Soomere and Viška (2014), Kovaleva et al. (2017), or Björkqvist et al. (2018). A broader contextualization for the entire Baltic Sea can be found in Hünicke et al. (2015) and Harff et al. (2017).

On most shorelines, erosion is concentrated over relatively short periods around high tide, but as tides are very small in the Baltic Sea, elevated average water levels, which provide conditions conducive to erosion, can be maintained for extended periods (Johansson and Kahma, 2016; Soomere and Pindsoo, 2016). Therefore, Baltic Sea volume, storm surges, wave setup, the presence or absence of sea ice, and long-period wave energy from infragravity or edge waves are the main factors influencing the duration and location on the beach profile where sediment can be mobilized and erosion may occur. Systematic synchronization of water level and wave intensity may considerably modify the width of the affected nearshore strip and the depth to which profile changes may occur (Soomere et al., 2017a).

The classic cut-and-fill concept of beach erosion and recovery (e.g., Brenninkmeyer, 1984) assumes that the most energetic steep waves induce beach erosion and mostly cross-shore transport of sediment to the deeper part of the shore, while sandbar formation, transport of sediment onshore, and the accretion of the same beach appear during calmer wave conditions with less energy and longer wave periods (constructive swell). Thus, changes on the beach primarily follow the incident wave energy level (Masselink and Pattiaratchi, 2001). This cycle is less significant in the Baltic Sea where the wave regime is highly intermittent and contains very small proportions of low-intensity constructive long swell waves (Broman et al., 2006; Soomere et al., 2012). If waves approach the nearshore area at large angles with respect to shore-normal, the classic cut-and-fill cycle of beach change is modulated by a relatively intense alongshore movement of sediment when compared with open ocean shores with a similar wave intensity (Soomere and Viška, 2014).

While the strongest alongshore and cross-shore sediment transport in the nearshore area (surf and swash zones) usually takes place during extreme wave events, the most rapid shoreline changes (both erosion and accretion) occur when high waves attack the shore at relatively large angles (Ashton et al., 2001), in particular when the angle of wave approach is unusual for the specific location. The latter indicates that specific shore segments may be sensitive to erosion for a particular wave direction only. Several small pocket or headland-confined beaches with very small amounts of sand are in a fragile, yet almost equilibrium, state as they are geometrically protected against winds from many directions (Caliskan and Valle-Levinson, 2008).

Compared to average conditions in the world oceans, storm waves in the Baltic Sea often approach the shore at relatively large angles (Soomere and Viška, 2014; Pindsoo and Soomere, 2015). On open shores, such waves drive much more intense alongshore transport than waves of comparable height that approach the shore almost perpendicularly. When the approach angle exceeds a threshold of about 45∘ (Ashton et al., 2001), the predominance of high-angle waves can lead to the explosive development of large spits and sand ridges. The growth of such structures has been observed in the eastern part of the Gulf of Finland (Ryabchuk et al., 2011b, 2020).

The presence of sea ice during a storm can modify this general pattern (Omstedt and Nyberg, 1991). The hydrodynamic forces are particularly effective in reshaping the shore when no sea ice is present and when the sediment is mobile (Orviku et al., 2003; Ryabchuk et al., 2011a). Storm surges are generally higher in the absence of sea ice (Omstedt and Nyberg, 1991). During extreme storm surges, strong waves may reach unprotected and unfrozen mobile sediment on higher sections of the shore that are out of reach for the waves during times with average water levels (Orviku et al., 2003). Land–ice interaction in the coastal zone is also crucial during ice winters. The principal erosion mechanisms are the wind-driven shore “ride-up” and pileup of ice and the ice growth down to the sea bottom with resulting transport of bottom sediment when the ice drifts out to sea (Girjatowicz, 2004; Leppäranta, 2013; Orviku et al., 2011). The former mechanism is also a risk to structures close to the shoreline.

Because of the dominant role of westerly winds that take up more than 60 % of the year over the southern Baltic Sea (Zhang et al., 2011a) and the sheltering effect of the land in the west, wind waves are typically more developed in the eastern Baltic Sea than in the western part. As a result, sediment transport and dune development are generally more facilitated along the eastern and southeastern coast. For example, the coastal dunes with the largest size and height are developed along the Polish and the Lithuanian coast. Being open to large hydrodynamic loads over short periods, although with almost no low and long-period swell that is typically associated with beach accretion, Baltic Sea shorelines are conceptually vulnerable to erosion. Waves in the Baltic Sea are also relatively short, implying a narrow surf zone and less powerful wave run-up than on the open ocean shores.

The Baltic Sea wave climate is highly intermittent and anisotropic, following the intermittency and anisotropy of the wind climate; that is, there are only short periods and few directions from which the higher waves approach the shorelines. Only about 1 % of the total onshore annual energy flux arrives within the calmest 170–200 d, about 60 % arrives within 20 d, and as much as about 30 % arrives during the stormiest 3–4 d (Soomere and Eelsalu, 2014). A simple consequence of the intermittency and anisotropy of wave fields and the complicated geometry of the coast is that the evolution of Baltic Sea shores is a steplike process (Soomere and Healy, 2011). They remain unchanged and seemingly stable for long periods until a storm from an unusual direction causes massive change. The most extreme erosion events will occur when such a combination comes along with high, normally storm-surge-related, water levels (Tõnisson et al., 2013b). However, most of the time, changes are very slow and require high-resolution measurements, such as laser scanning techniques (Eelsalu et al., 2015; Sergeev et al., 2018), to be detected. The anisotropy of the wave climate combined with the orientation of large sections of the shoreline leads to a large-scale asymmetry of the sediment flux (which is mostly counterclockwise on the shore of the Baltic Sea proper; Figs. 8, 9; Soomere and Viška, 2014). Owing principally to the overall sediment deficit on the Baltic Sea sedimentary shores (Pranzini and Williams, 2013), actual sediment transport is frequently limited by the availability of suitably sized sediment. Thus, the real sediment flux is only a few percent of the potential flux (Soomere and Viška, 2014). Consequently, there is only slow sediment accumulation in flux convergence areas (Fig. 9), and accumulation features may be destroyed easily as sediment transport during a single storm from an unusual direction may be substantially higher. This emphasizes the role of storms and storm waves in the Baltic Sea as substantial factors controlling sediment transport and coastal change.

Hydrodynamic conditions strongly affect coastal morphogenesis. Along the southern and eastern Baltic Sea coast, coastal morphogenesis has been extensively studied for more than a century (Keilhack, 1912; Kolp, 1978; Kliewe, 1995; Lampe et al., 2007; Zhang et al., 2010, 2014; Tõnisson et al., 2013a; Furmanczyk and Musielak, 2015; Harff et al., 2017; Deng et al., 2019). Barrier coasts are generally resilient to changing climate and can maintain their morphology provided that there is a neutral or positive sediment budget (Zhang et al., 2014) and beach migration is unimpeded (Cooper et al., 2020). The foredunes form a natural barrier for coastal protection along a major part of the southern Baltic coast. The part of the Baltic coastline that is protected by engineering structures or newly formed foredune ridges has been able to sustain its general shape and function (both ecological and economical) in the past decades, while most of the remaining parts, including the soft cliffs and old dune sections, have been subject to continuous and increased erosion (Łabuz, 2015). Most coastline erosion along the southern Baltic Sea is caused either by storms or by human-induced depletion of sediment supply (e.g., as a side effect of engineering structures). Existing studies reveal a highly nonlinear relationship between storm intensity and the rate of coastline erosion along the southern Baltic Sea coast (Zhang et al., 2011b, 2015). Additionally, a strong relationship has been found between rates of coastline change and the relative level of human development. Even a local and modest level of development (e.g., beach nourishment and pier construction) was found to influence the long-term coastline change at larger spatial scales (up to 100 km) (Deng et al., 2014; Dudzinska-Nowak, 2017). Because of this and the strong dependence on the angle at which the waves approach the shore, even smaller human interventions or a climate-related change in the predominant wind and wave directions in the Baltic Sea may substantially alter the structural patterns and pathways of wave-driven transport and functioning of large sections of the coastline (Viška and Soomere, 2012).

Human activities have become significant drivers of coastal change. With sediment transport being largely confined to shallow water, even smaller coastal construction works such as small boat harbors can significantly disrupt natural sediment transport pathways, leading to significant local (and frequently undesirable) coastal changes. Ship traffic, with the advent of strongly powered and fast large vessels, has become a significant driver of coastal processes and has caused erosion locally (Soomere et al., 2009). As is the case worldwide (UNEP, 2019), the demand for coastal sand for construction, industry, and beach nourishment will become increasingly significant for sediment supply.