A Holocene relative sea-level database for the Baltic Sea

We present a compilation and analysis of 1099 Holocene relative shore-level (RSL) indicators located around the Baltic Sea including 867 relative sea-level data points and 232 data points from the Ancylus Lake and the following transitional phase. The spatial distribution covers the Baltic Sea and near-coastal areas fairly well, but some gaps remain mainly in Sweden. RSL data follow the standardized HOLSEA format and, thus, are ready for spatially comprehensive applications in, e.g., glacial isostatic adjustment (GIA) modelling. We apply a SQL database system to store the nationally provided data sets in their individual form and to map the different input into the HOLSEA format as the information content of the individual data sets from the Baltic Sea area differs. About 80% of the RSL data is related to the last marine stage in Baltic Sea history after 8.5 ka BP (thousand years before present). These samples are grouped according to their dominant RSL tendencies into three clusters: regions with negative, positive and complex (transitional) RSL tendencies. Overall, regions with isostatic uplift driven negative tendencies dominate and show regression in the Baltic Sea basin during the last marine stage. Shifts from positive to negative tendencies in RSL data from transitional regions show a mid-Holocene highstand around 7.5 r Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). A. Rosentau, V. Klemann, O. Bennike et al. Quaternary Science Reviews 266 (2021) 107071 Glacial isostatic adjustment


a b s t r a c t
We present a compilation and analysis of 1099 Holocene relative shore-level (RSL) indicators located around the Baltic Sea including 867 relative sea-level data points and 232 data points from the Ancylus Lake and the following transitional phase. The spatial distribution covers the Baltic Sea and near-coastal areas fairly well, but some gaps remain mainly in Sweden. RSL data follow the standardized HOLSEA format and, thus, are ready for spatially comprehensive applications in, e.g., glacial isostatic adjustment (GIA) modelling. We apply a SQL database system to store the nationally provided data sets in their individual form and to map the different input into the HOLSEA format as the information content of the individual data sets from the Baltic Sea area differs. About 80% of the RSL data is related to the last marine stage in Baltic Sea history after 8.5 ka BP (thousand years before present). These samples are grouped according to their dominant RSL tendencies into three clusters: regions with negative, positive and complex (transitional) RSL tendencies. Overall, regions with isostatic uplift driven negative tendencies dominate and show regression in the Baltic Sea basin during the last marine stage. Shifts from positive to negative tendencies in RSL data from transitional regions show a mid-Holocene highstand around 7.5

Introduction
Advance and retreat of the Scandinavian Ice Sheet caused significant mass redistributions in the Baltic Sea Basin (BSB) and surrounding areas. The retreat resulted in glacial isostatic adjustment (GIA) and variable relative shore-(sea-) level (RSL) changes in tens or even hundreds of metres after the last deglaciation. RSL records from the BSB area are therefore the key constraint to understand glacial isostasy and mantle viscosity (Lambeck et al., 1998). In both fields, GIA models are generally parametrised by a lithosphere of constant thickness and a stratified mantle with Maxwell rheology. The model's fit to shore-level data serves as the major constraint in such modelling as the data mirrors surface deformations manifested in RSL change during the last 20,000 years (Wu et al., 2013). Hence, the inference of the ice-load history (which to date always needs some GIA model information), the lithospheric thickness and the mantle viscosity structure depends strongly on the quality of the sea-level data used (Steffen et al., 2014).
The complex history of the BSB with up-dammed lake phases (Baltic Ice Lake and Ancylus Lake) and marine phases (Yoldia Sea and Littorina Sea) challenges the use of RSL records in GIA modelling. Standardised and uniform RSL databases covering the whole BSB were compiled by Lambeck et al. (1998, but data were not published) and Peltier (1992, 1993). The datasets included sea-and lake-level records and were used in global GIA modelling. Since Lambeck et al. (1998) several regional and local datasets with sea-level index points (SLIPs) or limiting data points have been published from areas where high-resolution RSL data were not previously available, especially from the eastern and southern BSB areas (Miettinen et al., 2007a,b;Rosentau et al., 2013;Grudzinska et al., 2017;Muru et al., 2017;Nirgi et al., 2020).
RSL highstands were well known already at the beginning of the last century, while high-quality data about sea-level lowstands have been available only in the recent past. Studies of underwater landscapes, especially recent research into rooted tree stumps Rosentau et al., 2017;, Mesolithic kitchen middens and refuse layers of the various coastal sites and shallow-water fishing constructions (Grob et al., 2018;Hansson et al., 2019;Astrup, 2019) provide new fresh means for understanding the interplay between sea-level rise and GIA during the mid-Holocene.
Here we compile and present a publicly available Holocene RSL database for the Baltic Sea basin and the Kattegat area. This includes SLIPs and upper and lower limiting data points from the different parts of the BSB, including well-known data points from the regions of the highest postglacial uplift in the Fennoscandian Shield area, but also filling some of the important data gaps related to offshore and eastern Baltic Sea areas. We discuss the indicative meaning of multiple RSL indicators including isolation basins, coastal peat layers, raised shorelines and archaeological coastal settlement layers. Lake-level indicators from the Ancylus Lake (10.7e9.8 ka BP) and the following transitional phase -Initial Littorina Sea (9.8e8.5 ka BP) -are also presented in the database. Finally, we analyse the spatial variability of the data and compare the observations with selected GIA model predictions. The database is based on the standardized HOLSEA format (Hijma et al., 2015) covering all Baltic Sea geographic regions with different postglacial uplift histories. It also provides possibilities for direct data comparisons with other regions within the Eurasian Ice Sheet complex, including datasets from Britain and Ireland (Shennan et al., 2018), the Atlantic coast from France to Portugal (Garcia-Artola et al., 2018), the Netherlands (Hijma and Cohen, 2019), the Russian Arctic (Baranskaya et al., 2018) and further regions on the globe (Khan et al., 2019).

Regional setting
The Baltic Sea is a semi-enclosed intra-continental and almost tide-less sea with a total area of about 392,978 km 2 (without the Kattegat), and with a catchment area about four times larger than the area of the Baltic Sea itself. The transition zone between the Baltic Sea and the North Sea, and thus between brackish and oceanic water masses, is the Kattegat (22,287 km 2 ), which is sometimes included as part of the Baltic Sea (Lepp€ aranta and Myrberg, 2009). The present-day tidal range is 15e30 cm in the Danish straits, 2e5 cm in most of the Baltic Sea, up to 10 cm in the northern Gulf of Finland (Lepp€ aranta and Myrberg, 2009) and 17e19 cm in the easternmost part of the Gulf of Finland (Medvedev et al., 2013).
The BSB area experienced several glaciation events related to the major climatic shifts during the Last Glacial Period (Marine Isotope Stages (MIS) 4 to 2 (Batchelor et al., 2019). The Last Glacial Maximum (LGM) occurred during MIS 2, at around 20 ka BP and represented the coldest phase of the last glacialeinterglacial cycle with the largest ice volume and the most extensive areal coverage. The maximum ice thickness was around 3000 m in the area of the fastest uplift along the western coast of the Bothnian Sea (Kierulf et al., 2021).
Since the LGM, the BSB has undergone a number of lake and sea stages (for details see Andr en et al., 2011), starting with the Baltic Ice Lake (BIL). This freshwater body lasted from about 16 ka BP (Houmark- Nielsen and Kjaer, 2003) to 11.7 ka BP and was formed by meltwater inflow from the retreating ice sheet in the north and by river discharge from the south and east (Fig. 1). It is assumed that during the initial stage of the BIL the lake level was approximately at ocean level and, then, increased relative to the latter. According to a study by Muschitiello et al. (2016), the first drainage of the BIL at the outlet of Mt. Billingen took place at 12.87 ka BP, lowering the BIL by ca 5e10 m in a few years (Bj€ orck, 1995). Evidence from the Arkona Basin in the southern Baltic indicates a more pronounced lowering of the shore level of ca 20 m . During the cold Younger Dryas the ice margin re-advanced and the BIL was again dammed by the ice sheet (Bj€ orck, 2008). Soon after, the ice sheet started to retreat from the Younger Dryas position (marked by distinctive ice-marginal formations, such as Salpausselk€ as in Finland), and the water level of the BIL suddenly dropped within 1e2 years by ca 25 m just prior to 11.7 ka BP (Bj€ orck, 1995;Walker et al., 2009). During that stage the BIL has been connected to the sea via south-central Sweden, while southern Sweden has formed a peninsula connected to Denmark and Germany. The drop of water level marks the beginning of the next stage in the Baltic Sea history, the Yoldia Sea (YS) from 11.7 to 10.7 ka BP ( Fig. 2A). It took about 400 years before saline water penetrated into the Baltic Basin, initiating the brackish phase of the YS which lasted from 11.3 to 11.1 ka BP (Andr en et al., 2002;Obrochta et al., 2017). Land uplift in south-central Sweden stopped the water inflow from the sea and the YS gradually turned into a freshwater lake. A rapid Ancylus Lake (AL) transgression has been documented in the southern Baltic. It started about 10.7 ka BP, reached a highstand ca 10.3 ka BP (Fig. 2B) and lasted until 9.8 ka BP (Bj€ orck, 2008;Hansson et al., 2018). The Ancylus Lake stage is characterised by a 'tipping bathtub effect' resulting in regionally varying shore-level signatures (Andr en et al., 2011). In the northern parts of the Baltic regression was seen. The highest lake level is marked by the AL beach in south-east Sweden and Gotland (Svensson, 1989(Svensson, , 1991, Latvia  and Estonia . As land uplift was larger in the north than in the south and the lake level rose at the same time in general up to 10 m mean sea level, the southern coasts of the AL became inundated. At 10.2 ka BP, a new outlet formed via Mecklenburg Bay, Fehmarn Belt, and the Great Belt to the Kattegat , which likely led to an initial lowering of the AL of about 5 m followed by a fluvial phase, the Dana River along the new outlet. At ca 9.8 ka BP, the lake level and the sea level balanced again, and saline water began to enter the lake (Andr en et al., 2000;Berglund et al., 2005). The river developed as the drainage pathway of the AL and several smaller lakes formed in the western Baltic Sea area towards the Kattegat and North Atlantic. But the magnitude of the AL level drop is controversial. Thus Jensen et al. (1999) and Lemke et al. (1999) saw little evidence for erosion by the Dana River. The Littorina Sea (LS) transgression marks the most important stage of inundation associated with inflow of saline waters from the North Atlantic and finally shaping the present Baltic Sea south of the Scandinavian uplift region. This early Holocene transgression most likely progressed initially through the morphological depression of the Dana valley in the present-day Great Belt and Fehmarn Belt region (Bj€ orck, 2008;Feldens and Schwarzer, 2012). The Early Littorina Sea, Mastogloia Sea or Initial Littorina Sea (ILS), is a transitional phase from the AL to the LS from 9.8 to ca 8.5 ka BP with an almost freshwater character ( Fig. 2C; Berglund et al., 2005). Rising sea level resulted in flooding of the southern Kattegat area around 9.3 ka BP  and adjusted with the lowering lake levels around 9.0 ka BP at approximately 30 m below sea level (b.s.l.). As a consequence of a further rising sea level during the final phase of deglaciation the riverine AL outflow from the Baltic Sea basin turned into a brackish-marine inflow from the North Atlantic passing the Great Belt and Fehmarn Belt depressions and flooding successively the areas above the 30 m b.s.l. (Ernst, 1974).
The timing of the transition from fresh to brackish water that marks the onset of the LS is not yet clearly determined and might have started in the northern Great Belt region around 9.0 ka BP and east of the Darb Sill after 8.5 ka BP (Andr en et al., 2011). The cause for the saline water inflow into the BSB is believed to be related to episodic melting of the Laurentide and Antarctic ice sheets (Andr en et al., 2011). However, since 8.5 ka BP a LS transgression has been recorded in many different locations around the BSB suggesting the onset of the last marine stage Lampe et al., 2011;Rosentau et al., 2013;Nirgi et al., 2020).
The end of the rapid melting of the Laurentide and Antarctic ice sheets resulted in a slow-down of RSL rise in the areas of near-zero uplift  and the culmination of the mid-Holocene RSL highstand in the areas of the slow postglacial uplift around 7.5e6 ka BP ( Fig. 2B; Yu et al., 2007;Rosentau et al., 2013). Brackishmarine diatom assemblages from the sediments of the Landsort deep suggest that the highest surface water salinities in the BSB occurred between 7.1e5.4 ka BP (van Wirdum et al., 2019), thus during the highstand or slightly later. The beginning of the late LS stage (ca 3.0 cal ka BP; Berglund et al., 2005) is identified by microfossils (Bj€ orck, 1995(Bj€ orck, , 2008(Bj€ orck, , 2008, but is not yet clearly defined based on results of many independent studies. According to Andr en et al. (2000) the late LS begins "where the siliceous microfossils assemblage that requires a more marine environment decreases".
RSL changes and uplift patterns for the recent past have been recorded by tide gauge measurements complemented by repeated levelling (Ekman, 1996(Ekman, , 2009(Ekman, , 2009Kakkuri, 1997;Douglas and Peltier, 2002). During the last decades continuous point positioning (time series of the coordinates) from permanent GNSS (Global Navigation Satellite System) networks has also become available for determinations of crustal movements with respect to the Earth centre of mass, e.g., within the BIFROST (Baseline Inferences from Fennoscandian Rebound Observations, Sea-level and Tectonics) project (Vestøl et al., 2019;Kierulf et al., 2021, Fig. 3).

Database compilation
The data sets compiled in this study cover all the geographic coastal regions of the Baltic Sea including the Kattegat area (Fig. 3) and extend over the entire Holocene period after the drainage of the Baltic Ice Lake at about 11.7 ka BP (Bj€ orck, 1995;Andr en et al., 2011). Altogether, 77 attributes of the HOLSEA database format (version 2019 at https://www.holsea.org/) were provided by different data providers of all circum-Baltic countries (Supplement 1). Country reports addressing specific issues related to the data  (Peltier, 2004) with location of the study area and the ice-dammed Baltic Ice Lake in the Baltic Sea Basin area. Red contours marks ice sheet extents during the last glacial maximum around 20e21 ka BP. Palaeotopography is reconstructed with the same ice-sheet model and the VM2 earth model associated with ICE-5G model. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) availability, collection, quality and sample's indicative meaning are provided in Supplement 2.
A few data sets were not provided in the HOLSEA data format of Hijma et al. (2015), see Supplement 2. The question arose how to deal with these data formats especially when it was not possible to transfer the provided data one-to-one into the HOLSEA workbook format. Accordingly, we decided to transfer all data sets into a Structured Query Language (SQL) database system (we chose PostgreSQL) where the original content of each table, usually a spread sheet, was imported and stored in separate relations or tables (Fig. 4). Relation-specific rules were defined in order to convert the data into the HOLSEA format, defined as views in the database system. The contributors were consulted about the rules. Finally, the union of all views was exported back into the HOLSEA workbook excel format. Advantage of this procedure is that we keep the original content and the transparency of the data conversion.
Radiocarbon data were partly re-calibrated after delivery. For this, we used the software OxCal, Version 4 developed by the Oxford Radiocarbon Accelerator Unit (www.c14arch.ox.ac.uk/oxcal. html), which was installed locally and, so, could be implemented into the processing chain. Ahead of the data transfer into the HOLSEA worksheet, other information like 1-sigma ranges or probability density functions determined by the software are provided in the database system. Luminescence ages were standardized to 1950 CE to be consistent with the radiocarbon ages (cal BP). This involves 38 samples in the database with age adjustments between 48 and 64 years.

RSL indicators
The Holocene RSL database for the Baltic Sea holds 1099 accepted data points ( Fig. 5; Appendix A) of which 867 are sea-level data points and 232 are data points from the AL phase and the following transitional ILS phase from 10.7 to 8.5 ka BP (Fig. 2). From these, about one-third are SLIPs, represented by isolated lake basins, salt marsh deposits, basal peats, raised beach ridges and other indicators (Fig. 5). About one-third are terrestrial limiting points represented by freshwater peat deposits and Mesolithic and Neolithic cultural layers from the coastal zone. Marine limiting points include often near-shore marine and lagoon deposits from the isolation basin studies and other indicators. Their indicative meanings are summarized in Table 1 and the water levels are . As the tidal range in the BSB is very small and clearly below the mean significant wave heights (Tuomi et al., 2011) the SLIPs in the database are not corrected for tidal effects. AL and ILS index points are represented by isolated lake basins, submerged tree stumps and other indicators (Table 1). According to different authors the water level in the AL was up to 10e20 m above the Atlantic sea-level Rosentau et al., 2013;Hansson et al., 2018) while the ILS water level was close to the Atlantic level (Andr en et al., 2011).
The largest number of RSL data including SLIPs are from the time period between 8.0 and 6.0 ka BP, and are associated with the research interest to study the changes during the mid-Holocene RSL rise (Fig. 5). The Late Holocene is rather poorly represented in the dataset and only few data points are available for the last 2 ka BP. The rejected 204 data points, not used in further analyses, are also included in the database, thus the total number of the points in the Baltic data set is 1303 (Appendix A). The largest number of rejected data is related to raised beach ridges, which can be used as SLIPs if their chronologies can be established. AL and ILS lower (marine) limiting points were also rejected as they can form either  Highest shorelines, beach and storm ridges Foot of the beach ridges and other coastal landforms with luminescence or radiocarbon ages from the same or neighbouring sites AL level ± 1 m Submerged tree stumps In situ tree stumps that died during marine transgression  up to 2.5 m above AL level ± 0.5 m Initial Littorina Sea index points (9.8e8.5 ka BP)

Isolation basins
Stage of isolation determined from lithology, geochemistry, loss-on-ignation and diatom assemblages and sometimes by other microfossil evidences ILS level ±(0.2 me0.5 m) Submerged tree stumps In situ tree stumps that died during marine transgression  up to 2.5 m above ILS level ± 0.5 m Fen peat to brackish water sediment transition (transgressive contact) Intercalated fen peat with brackish-water sediments (gyttja) on top ILS level ± 0.5m above or below MSL. Other rejected data include terrestrial and marine limiting data points, which form a cluster with very limited number of data points where the shore-level trends cannot be independently estimated with satisfactory confidence. Overall, the accepted RSL indicators show the regional-scale variation and are further discussed following the developmental stages of the BSB (Fig. 2): 11.7e10.7 ka BP (YS); 10.7e9.8 ka BP (AL); 9.8e8.5 (ILS) and since 8.5 ka BP (LS). RSL data for the last marine stage (LS) were grouped according to their dominant RSL tendencies into three clusters: regions with negative, positive and complex (transitional) RSL tendencies (Fig. 6).
4.2. RSL data and tendencies during the 11.7e10.7 ka BP (YS stage) This group includes RSL data points from 16 original regions (SLIPs n ¼ 13; limiting points n ¼ 84) from the BSB and the Kattegat. No RSL data points are available from the Bothnian Bay and most of the Bothnian Sea regions because these areas were still covered by the Scandinavian Ice Sheet during the YS stage ( Fig. 2A). About 95% of the RSL data are terrestrial and marine limiting points. Terrestrial limiting points include intercalated peat and gyttja deposits. Marine limiting points include sedimentary indicators with dates from marine shells or wood from marine or coastal sediments. A few SLIPs are coming from the isolated lake basins studies, from the sedimentary indicators and submerged rooted tree stumps.
The highest SLIPs are at an elevation of 118 m above the present day sea level (a.s.l.) in SW Finland (Fig. 7) while in the southern Baltic rooted pine stumps at H€ av€ ang in Sweden indicate an RSL of 22 m b.s.l. (Fig. 8) and rooted pine stumps at Lithuanian offshore even lower the RSL to around 34 m b.s.l. (Fig. 9). Coastal and marine deposits with marine shells from the Belt Sea region, which was separated from the BSB by a land bridge, suggest RSL levels of over 35 m b.s.l. (Fig. 9). RSL data points from SW Finland and € Osterg€ otland regions in Sweden suggest negative tendencies (Fig. 7) and RSL data points from Hav€ ang and offshore Lithuania suggest positive RSL tendencies (Figs. 8 and 9) during the YS.
4.3. RSL data and tendencies during the 10.7e9.8 ka BP (AL stage) This group includes sea-level data points from the Kattegat, Halland (Sweden) and G€ oteborg (Gothenburg, Sweden) regions (SLIPs n ¼ 3; limiting points n ¼ 7) and lake-level data points of the up-dammed AL from the eight different regions (SLIPs n ¼ 30; limiting points n ¼ 75).
Sea-level data points are represented by marine limiting points from the Kattegat region, mostly by dated marine shells and molluscs from coastal deposits and suggest RSL levels over 25 m b.s.l. (Fig. 9). In the slowly uplifting Halland region in Sweden the RSL rose from 15 b s.l. to 14 m a.s.l. during this period (Fig. 8).
Lake-level data points of the AL include SLIPs derived from the isolated lake basins at different elevations and from rooted tree stumps. Isolation events have been detected by using diatom stratigraphy, threshold measurement and radiocarbon dating supported by counting of clay varves from site-specific sequences. Large lake diatoms like Ellerbeckia arenaria and Aulacoseira islandica have been typically used to distinguish the AL stage from an isolated lake stage . AL data comprise also a large number of terrestrial limiting points including peat deposits and archaeological sites especially from the southern and eastern Baltic. Lake limiting points include mostly sedimentary indicators from AL deposits in the southern Baltic.
In the Bothnian region only negative lake level tendencies are available indicating that postglacial land uplift was faster than the effect of the up-damming of the AL. In the Ångermanland region in Sweden, the highest isolation basin was found at an elevation of 268 m a.s.l. and was dated to 10.6e10.2 ka BP (Fig. 7). Regions with moderate to slow uplift in Blekinge and Hav€ ang (both Sweden) as in Tallinn and P€ arnu (both Estonia) show positive lake-level tendencies from 11.0 to~10.2 ka BP and negative lake-level tendencies from 10.2 to~9.8 ka BP (Fig. 8). The magnitude of up-damming varies from 20 to 10 m depending onthe intensity of the uplift and the availability of lake-level data points. Submerged rooted tree stumps from Blekinge and Hav€ ang at different elevations suggest a rapid rise of the AL level reaching about 20 m with a rate of~40 mm per year . The oldest Mesolithic archaeological RSL indicators from the H€ av€ ang and P€ arnu regions are also related to the AL transgression.
4.4. RSL data and tendencies during the 9.8e8.5 ka BP (ILS stage) This group includes sea-level data points from the Kattegat, Samsø Belt and Vendsyssel Thy (all Denmark), as from G€ oteborg to Halland in Sweden (SLIPs n ¼ 5; limiting points n ¼ 13) and ILS data from the eight regions of the BSB (SLIPs n ¼ 30; limiting points n ¼ 97).
Sea-level data points are represented by marine limiting points from the Kattegat region, mostly by dated marine shells and molluscs from coastal deposits. They suggest RSL levels above 23 to 22 m b.s.l. (Fig. 8). ILS data include SLIPs derived from isolated lake basins at different elevations. Isolation events have been detected by using diatom stratigraphy, threshold measurement and radiocarbon dating. During this stage, the distribution of brackish water diatoms like Mastogloia smithii is common to separate the ILS stage from an isolated lake stage . RSL data points from the ILS show negative tendencies in the northern BSB (Fig. 7) and positive tendencies in the southern BSB (Fig. 9). SLIPs from the Ångermanland and G€ astrikland regions in Sweden show a fast drop in RSL at that time. About 9.5e9.1 ka BP, the RSL at the Lomtj€ arn site in Ångermanland was about 180 m a.s.l. (Fig. 7). In the Usedom/Rügen region in Germany, in the southern BSB, the RSL was 16 m b.s.l. around 8.9 ka BP (Fig. 9). SLIPs and terrestrial limiting points from Hav€ ang and P€ arnu show rather stable and low RSLs around 5 m b.s.l. to 1 m a.s.l. during the ILS. Despite the first evidences of marine water inflows in the southern BSB during the ILS (Andr en et al., 2011), stable and low RSL levels in Hav€ ang and P€ arnu suggest that RSL changes in the BSB were not fully controlled by sea-level rise in the world's oceans (Lambeck et al., 2014). Thus, the Early Holocene rapid sea-level rise, well

RSL data and tendencies since 8.5 ka BP (LS stage)
RSL data points from the BSB and Kattegat since 8.5 ka BP include 274 SLIPs and 468 limiting data points and are further grouped into three clusters according to their dominant RSL tendencies: regions with negative, positive and complex (transitional) RSL tendencies. Overall, regions with isostatic uplift-driven negative tendencies dominate and show regression and decreasing volume of the BSB during this last marine stage (Fig. 6).

Regions with negative RSL tendencies
This cluster of RSL data points comprises 12 original regions from the northern BSB including the Ångermanland region with evidence of the highest postglacial uplift in northern Europe. Thus, between 9.1 and 7.8 ka BP RSL dropped in Ångermanland from 152 to 128 m a.s.l. (Fig. 7). SLIPs in this cluster contain data collected from isolated lake basins at different elevations, which were studied using diatom stratigraphy, threshold measurements and radiocarbon dating supported by counting of clay varves from site-specific sequences. Thresholds have been identified in the field and their elevations were determined typically by using benchmarks, but also by applying airborne LiDAR (Light detection and ranging) elevation data (Supplement 1). Accelerator mass spectrometry (AMS) and conventional radiocarbon dating of terrestrial plant material supported by age-depth modelling is the most typical method to establish chronologies of an isolation event. Data points from raised beaches are also available for this region, however, due to a poor age control some of these data points are rejected for further analyses. SLIPs from Ångermanland, G€ astrikland, S€ odermanland (all Sweden) and Finland SW cover rather well the entire last marine stage while for most of the other regions the last 6000 years are quite poorly represented by RSL data points (Fig. 7).

Transitional regions
This is the largest cluster of RSL data points and it includes 14 original regions from the BSB and Kattegat (Fig. 8). These RSL data comprise a large number of terrestrial limiting points including peat deposits and Mesolithic and Neolithic archaeological sites but include also SLIPs derived from intercalated peat layers, submerged rooted tree trunks, isolated lake basins using diatom stratigraphy and radiocarbon dating, or luminescence dated coastal landforms. In general, the transitional regions in the BSB and the Kattegat show positive RSL tendencies from 8.5 to~7 ka BP and negative RSL tendencies afterwards (Fig. 8). Shifts from positive to negative tendencies are dated in Finland S to~7.3 ka BP (7.4e7.1 ka BP), in the Karelian Isthmus (Russia) to~7.1 ka BP (7.3e6.8 ka BP), in Narva-Luga (Estonia) to~7.3 (7.7e6.9 ka BP) and in the Tallinn region tõ 7.5 (8.1e6.8 ka BP) ( Fig. 8; Supplement 1). RSL data points from the P€ arnu and Blekinge regions combine SLIPs collected from onshore and offshore areas and show shifts from positive to negative tendencies at~7.3 ka BP and at~6.5 ka BP, respectively (Fig. 8). Thus, the age of the turning point from the positive to negative tendency in the transitional regions of the BSB vary between 7.5 and 7.1 (6.5) ka BP, depending somewhat on the intensity of the uplift. This age is consistent with the end of the final melting of the Laurentide Ice Sheet and a remarkable slow-down in global sea-level rise (Lambeck et al., 2014).

Regions with positive RSL tendencies
This cluster of RSL data points includes 13 original regions from the BSB (Fig. 9). In the Poel, Rügen/Hiddensee, Usedom/Rügen, Fishland/Zingst and Salt Meadows regions in Germany from salt meadows transgressed basal peat samples have been used mainly, which was supported by diatoms, pollen and plant macro remains, forming about 80% of all SLIPs in this cluster (Fig. 9). Terrestrial limiting points include in situ tree stumps and Mesolithic material covering the period ca 9e6 ka BP. AMS or conventional radiocarbon datings of terrestrial macrofossils from the peat layers have been used for the chronology. Data points from the Fehmarn Belt and Kieler Bucht regions in Germany include various terrestrial and marine limiting data points derived from offshore sediment cores drilled during the 1970s and 1980s. Their chronology covers the whole last marine stage and is based on conventional radiocarbon dating of lake, marine and peat deposits. Newer ages from the Fehmarn Belt from the 2000s are also based on AMS dating. Data points of the Arkona Basin are from the 1990s and include only marine limiting points (Fig. 9). Data points from the Vistula Lagoon in Poland include mostly terrestrial limiting points containing submerged peat layers supported by pollen analyses and in situ tree stumps. Their chronology covers the last 7000 years and is based on AMS radiocarbon dated terrestrial macrofossils, wood and bulk peat samples dated during the 21st century. Sample elevation uncertainties of this off-and onshore region is around ±0.1 m.
Data points from the Gulf of Riga in Latvia comprise terrestrial and marine limiting points including peat and gyttja deposits supported by diatom analyses. Their chronology covers the period from 9 to 2 ka BP and is based on AMS radiocarbon dated terrestrial macrofossils and gyttja samples as well as conventional radiocarbon dates of peat, wood and gyttja samples. Sample elevations of the onshore samples were determined using topographic maps with uncertainties of ±0.5 m.
Data points from the Belt Sea in Denmark include various terrestrial and marine limiting data points derived from offshore sediment cores suggesting RSL below 20 m b.s.l. around 8.5e8.0 ka BP (Fig. 9). However, SLIPs from Rügen/Hiddensee and Usedom/ Rügen show up to 5 m higher RSL levels for the same period. SLIPs and terrestrial limiting points from Poel, Rügen/Hiddensee, Usedom/Rügen, and Fishland/Zingst also indicate the slowdown in Holocene RSL rise between 8 and 7 ka BP (Fig. 9).

Vertical uncertainties in RSL data
Various methods were used to reconstruct SLIPs and limiting points in the BSB, which can cause different chronological and vertical uncertainties in the dataset. The main uncertainties are discussed in this chapter and specific details are provided in the country reports in Supplement 2. Converting original data to HOLSEA format provided a good template to systematically analyse the uncertainties, which has been done in most cases by original data producers and has been further discussed during the set up of the database. The RSL data from nine original data sets was presented.
A large number of SLIPs from the regions with negative and complex RSL tendencies comes from isolated lake basins. The detailed threshold identification, coring and elevation measurement is a standard procedure in sea-level studies. Still, for several sites original elevations have been taken from topographic maps. For these sites, the threshold elevations have been corrected by the data providers during the database compilation by using newly available LiDAR elevation data, which provide the possibility to diminish the vertical accuracy to ±0.1e0.2 m (Supplement 2). For underwater sites, the vertical accuracy is estimated to be ±0.5 m.
In offshore sea-level studies in regions with negative sea-level tendencies, the elevations are estimated based on the measured water depth and the shore-level height with vertical accuracy around ±1 m. Sampling uncertainty depends on used coring equipment and methodology. In isolated basin studies, typically Russian corers have been used with uncertainties around ±0.05 m while in vibra-coring and offshore gravity coring the uncertainties are higher being around ±0.15 m.
Sediment compaction for basal peat samples is not significant, but for intercalated peat and gyttja samples, individual compaction factors have been considered. These errors have been evaluated site-by-site also using compaction models. The highest compaction values are related to the offshore and coastal samples from the southern BSB with values up to 5.9 m .
We note that the height or depth of the data refers to a certain height system in place in the respective country. Most national height systems, for example RH2000 in Sweden, are national realisations of the European Vertical Reference System (EVRS) which in the countries surrounding the Baltic Sea refer to either NAP or the Kronstadt tide gauge. The latter is the zero level for the Baltic Height System (BHS77) in Russia and previously in many countries in Eastern Europe, and this zero level differs up to 0.2 m from the NAP. Hence, we applied a correction for such data points.
If known, the national height system is indicated in the database and, if needed, a correction is applied to transfer the height of an old height system to the current EVRS-related one. Such a correction is generally smaller than a few decimetres. If the height system is unknown, the year of sample discovery is listed and an uncertainty of 0.5 m is added. This concerns 539 samples, mostly related to the older offshore and some onshore data of the southern Baltic and Kattegat, and amounts to 40% in the database. We suggest a conservative uncertainty of 0.5 m based on findings by Nordman et al. (2015), who reanalysed RSL data, mainly based on varves, from Ångermanland, Sweden, as well as corrections applied to some Swedish data after transformation of data from the former national height system RH70 to the current RH2000. Nordman et al. (2015) identified different height systems for different portions of the Ångermanland record, which is close to today's land uplift maximum. The reference levels of two of the systems differ by 68 cm and have reference epochs being 70 years apart. We note though that the RSL data discussed in Nordman et al. (2015) were found in different decades since the beginning of the last century, while our database mainly contains data since the 1950s and data far away from the uplift maximum. Hence, 68 cm is considered as likely too large to be set as uncertainty. Our corrections to some Swedish data points, also near the uplift maximum, are at a level of 0.2e0.3 m and concern two height system epochs being 30 years apart. We thus think that an uncertainty of 0.5 m includes any uncertainties due to an unknown height system. The user is of course free to either revise this uncertainty value or update the database with the correct height value in the national height system related to the EVRS.
Besides the observational uncertainties, various climatological and geological factors may affect the accuracy of the RSL data. However, these are very difficult to assess. In the tide-less BSB, wind may cause high sea levels at the narrow ends of bays and gulfs, like the Gulfs of Finland, Bothnia and Riga. As most cyclones travel over the BSB from the SW or W to E, storm surges are usually generated in the E or NE sections of the Baltic Sea (Hünicke et al., 2015) and may affect Holocene RSL.
Local crustal (block) movements along existing fault lines have been detected in western Sweden (Risberg et al., 1996) and SE Sweden (Risberg et al., 2005) by investigating Holocene isolated lake basins. We have not identified any local crustal instabilities in the RSL data nor did we apply any correction for such. Kierulf et al. (2021) recently analysed the vertical velocity field derived from GNSS around the Baltic Sea and could show local motions that cannot be explained by GIA or other known processes. Hence, the user should be aware that such local instabilities may sporadically be present and further corrections may be necessary.

Age and chronological uncertainties in RSL data
About 97% of the SLIPs and limiting dates are radiocarbon dated (AMS or conventional dating) and have been calibrated using the OxCal program (Version 4.3). Terrestrial samples were calibrated using the IntCal13 calibration curve and marine samples using the Marine 13 calibration curve (Reimer et al., 2013) and reported with 2s confidence interval. Most of the radiocarbon ages were corrected for isotopic fractionation. About 75% of the radiocarbon dates originate from terrestrial samples such as seeds and fruits or wood samples from sediments or freshwater peat. Such material is usually considered reliable material for chronological studies. About 30% of the radiocarbon measurements are AMS ages. Bulk sediment samples such as gyttja or marine mud have also been widely used for radiocarbon dating. Bulk samples may contain a mixture of carbon of different ages and are therefore less reliable. Age-depth modelling was also used to date the SLIPs in the sediment sequences with multiple radiocarbon dates. Thus, the modelled age represents SLIP age and the actual radiocarbon dates represent terrestrial and or marine limiting points.
In the Ångermanland region, radiocarbon dating of the SLIPs is supported by counting of clay varves from site-specific sequences and, for three regions (Samsø, Lithuania, Hiiumaa) luminescence dating has also been used for chronology. Luminescence dating provides the possibility to date coastal deposits without carbon content, but suffers from larger chronological uncertainty than present in radiocarbon dating.
We advise the user interested in GIA modelling that the data   (Vestøl et al., 2019) or other national transformations in place. The uncertainty for such corrections is usually negligible if the height system is known, while we suggest setting 10% of the height correction as uncertainty if the height system is unknown. The latter is based on tests we made with Swedish data near today's land uplift maximum. The difference of a geodetically correct transformation from one height system to another compared to one with simply using a land uplift model was 1e2 cm for total height corrections of 20e30 cm.

RSL data and GIA model predictions
RSL data points from the last marine stage (since 8.5 ka BP) were compared with the publicly available GIA predictions including the global ICE-5G ice history (Peltier, 2004; with a lithospheric thickness of the Earth model: 80, 100, 120 km) and the ICE-6G_C ice history together with its corresponding Earth model VM5a (Argus et al., 2014;Peltier et al., 2015).
Comparisons of RSL data sets with GIA predictions show relatively good fit with RSL data from the regions with negative tendencies including the regions with highest uplift in Ångermanland, G€ astrikland and Uppland ( Fig. 7 and 10). This was expected because some of these data were used as major constraints in ice model developments and used in the generation of both ICE-5G and ICE-6G_C models. Comparison of SLIPs from these regions shows that standard deviations values remain typically below 10 m, with the exception of the Gunnarsbyn data which shows systematically a lower RSL compared to the model predictions and a standard deviation of 10e14 m (Fig. 10).
Regions with positive RSL tendencies also show relatively good fit with GIA predictions with exception of the Fehmarn dataset which contains only one SLIP ( Fig. 9 and 10). Standard deviation values are also around 5 m, however for the ICE-6G_C model the differences are somewhat higher compared to other predictions (Fig. 10). Terrestrial limiting points from Vistula, Lithuania and the Gulf of Riga also show several meters lower levels compared to model predictions during the mid-Holocene.
The RSL data from the transitional regions shows a rather poor fit with GIA predictions, especially in the eastern BSB (Fig. 8). The mid-Holocene RSL highstand was reached in transitional regions of the BSB around 7.5e7.1 (6.5) ka BP indicating a slow-down of sealevel rise while ICE-5G and ICE-6G_C models predict this highstand ca 500e700 years earlier (Fig. 8). For the Karelian Isthmus, Narva-Luga and P€ arnu regions, ICE-5G and ICE-6G_C model predictions fail to predict a marine transgression at 8.5e7 ka BP that is clearly documented in RSL records. For these three areas, standard deviation values are also highest for the ICE-6G_C model being at about 20e25 m and for ICE-5G at about 10e24 m (Fig. 10). In the Finland S, Tallinn and Blekinge regions the differences are somewhat smaller, being at about 5e10 m. For the transgression period at about 8.5e7 ka BP, proxy reconstructions suggest clearly a lower RSL, which is further confirmed by terrestrial limiting points from Hav€ ang (Figs. 8 and 10).
Geographically the highstand area is crossing the Jylland and Blekinge areas in the western BSB and Latvia, Estonia, and SE Finland. Compared to ICE-6G_C model predictions the highstand area fits well with the RSL records in the western BSB but locating it ca 200e300 km northward in the eastern BSB (Fig. 11). Differences in RSL elevations and location of the highstand zones between model predictions and proxy reconstructions in the eastern BSB may suggest that the contribution of ice loading is overestimated in the ICE-5G and especially in the ICE-6G_C models as the eastern BSB region has been previously rather poorly covered with the RSL data.

Conclusion
We provide a standardized and publicly available Holocene RSL database for the Baltic Sea and the Kattegat Sea with 867 sea-level data points. The database also includes 232 data points from the Ancylus Lake phase (10.8e9.8 ka BP) and the following transitional phase (9.8e8.5 ka BP) in the Baltic Sea history, distinguished from the marine data. About 80% of RSL data is related to the last marine stage in the Baltic Sea history since 8.5 ka BP. This part contains 274 SLIPs and 468 marine and terrestrial limiting points which are grouped according to their dominant RSL tendencies into three clusters: regions with negative, positive and complex (transitional) RSL tendencies. Overall, regions with negative tendencies, associated with intense and still ongoing postglacial uplift, dominate and show falling RSL in the BSB. Shifts from positive to negative tendencies around 7.5e6.5 ka BP in transitional regions are consistent with the end of the final melting of the Laurentide Ice Sheet. Comparisons of RSL data with GIA predictions including global ICE-5G and ICE-6G_C ice histories show good fit with RSL data from regions with negative tendencies, whereas in the transitional areas in the eastern BSB the predictions overestimate the RSL and fail to predict a mid-Holocene RSL highstand derived from the proxy Fig. 11. Distribution of the mid-Holocene RSL highstand area for the considered GIA models together with RSL data points. The area is determined as the region where the tendency changes between 10 and 5 ka BP. The colours indicate the age of the highstand determined from ICE-6G_C(VM5a). Inside the ring, the model predicts a continuous RSL fall, and outside a continuous rise. RSL data points with a negative tendency are shown as blue dots and points with a positive tendencyare shown in red. Note that the mid-Holocene highstand timing is shifted from mid-to early Holocene due to interference with the uplift history. It is detected in the zone where negative and positive RSL tendencies co-occur and shows, the GIA model fails to predict the highstand zone in the eastern BSB. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) reconstructions. Differences in RSL elevations and the locations of the highstand zones between model predictions and proxy reconstructions in the eastern BSB may suggest that the contribution of ice loading is overestimated in the ICE-5G and especially in the ICE-6G_C models. This example thus shows, among others, the potential of the database to serve in and revise ice history reconstructions.
Finally, we note that there is also a large number of late-glacial RSL data available related to the Baltic Ice Lake stage in the BSB history, which are useful in GIA modelling extending the time span beyond the Holocene (Lambeck et al., 1998(Lambeck et al., , 2010. As more adequate and precise observations of isolations become available for the Baltic Ice Lake, the extension of the open-access Baltic database towards the late-glacial period would be a well-justified task for the future.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements
We thank Kurt Lambeck and an anonymous reviewer for their critical reviews to improve the paper. We also thank Svante Bj€ orck for helpful discussion on RSL data. Jukka-Pekka Palmu, Annika Åberg and Susanne Åberg are acknowledged for compiling the Ancient Shoreline Database at the Geological Survey of Finland. Tina Kempe is thanked for help with data transformation to the current Swedish national height system RH2000. The research of Alar Rosentau was supported by Estonian Research Council grant PUT 456. The contribution of Volker Klemann, Milena Latinovi c and Meike Bagge was supported by German Federal Ministry of Education and Research (BMBF) as Research for Sustainability initiative (FONA); www.fona.de through Palmod projects (FKZ: 01LP1503A, 01LP1502E, 01LP1918A). The GPR-based studies in the Danish area received funding from the Danish Research Council for Independent Research through a grant to Lars Nielsen. Sediment samples were dated with the luminescence method at the Nordic Laboratory for Luminescence Dating. The contribution of Yuriy Kublitskiy and Dmitry Subetto was funded by Russian Foundation for Basic Research (18-05-80087) and was supported by the Ministry of Education of the Russian Federation (FSZN-2020-0016). The research of Ieva Grudzinska was supported by Estonian Science Foundation grant 9031, Estonian Research Council grant IUT 1-8 and the Doctoral Studies and Internationalisation Programme DoRa. Most of the figures were created by the Generic Mapping Tools GMT4 (Wessel and Smith, 1998) and GMT6 (Wessel et al., 2019). This paper is a contribution to IGCP project 639 'Sea Level Change from Minutes to Millennia', supported by UNESCO and IUGS.

Appendix A. Supplementary material 1
Excel file containing SLIPs and limiting points included in the Holocene relative sea-level database of the Baltic Sea. The database is provided at the GFZ data services (https://doi.org/10.5880/GFZ.1. 3.2020.003).

Appendix B. Supplementary material 2
Document file with country-wise description of the Holocene relative sea-level data. The document is provided at the GFZ data services (https://doi.org/10.5880/GFZ.1.3.2020.003).

Appendix C. References to original publications of the RSL data
A large number of papers are only referenced in the database. In order to tribute the work of these primary authors, we decided to add also those references to the main document, following the discussion in Düsterhus et al. (2016): Finland: Alhonen et al. (1978); Donner and Eronen (1981);