Timing of the last deglaciation phases in the southern Baltic area inferred from Bayesian age modeling

A new chronology of the last Scandinavian Ice Sheet retreat in the southern Baltic basin is proposed. Based on Bayesian age modeling, we show that the most likely ages of particular deglaciation phases are 16.5 ± 0.5 ka for the Gardno Phase, 15.6 ± 0.6 ka for the Słupsk Bank Phase, and 13.9 ± 0.5 ka for the Southern Middle Bank Phase. The Gardno moraines are correlated with the Halland Coastal moraines in southern Sweden and the Middle Lithuanian moraines in Lithuania and Latvia. Ice margin stillstands of the Słupsk Bank Phase and Souhern Middle Bank Phase are correlated with the Göteborg and Vimmerby moraines, and with the North Lithuanian (Haanja) and Otepää moraines. The average retreat rates of the ice margin of about 55 m/yr between the Gardno Phase and the Słupsk Bank Phase, and about 40 m/yr between the Słupsk Bank Phase and the Southern Middle Bank Phase suggest that the last deglaciation did not accelerate after the Gardno Phase when an extensive ice-dammed lake was formed in front of the retreating ice sheet. The ice margin was probably grounded rather than floating, which prevented its more rapid retreat. The timing of the two main ice margin stillstands at the Słupsk Bank and at the Southern Middle Bank corresponds to the cool periods around 15 and 14 ka interpreted from paleotemperatures of Greenland based on ice core GISP2. This suggests that the main phases of the last deglaciation in the southern Baltic region were at least partly triggered by climatic fluctuations in the Northern Hemisphere.


Introduction
A significant warming of climate in the Northern Hemisphere at the Last Glacial Termination triggered the retreat of the Scandinavian Ice Sheet (SIS) (e.g., Denton et al., 2010;Lambeck et al., 2010;Cuzzone et al., 2016;Hughes et al., 2016;Stroeven et al., 2016;Patton et al., 2017). After the Last Glacial Maximum (LGM), which in the northern continental Europe occurred between ~24 and ~18 ka (e.g., Ehlers et al., 2011;Hughes et al., 2016;Hughes et al., 2021), the southern margin of the last SIS receded gradually, leaving glacial landforms and sediments clearly visible in the landscape. The timing of the last SIS retreat has recently been reconstructed from extensive datasets of available geochronological data and geomorphological/geological evidence constraining the ice sheet advance and retreat compiled into a GIS system (e.g., Hughes et al., 2016;Stroeven et al., 2016). Moreover, the last deglaciation of the entire Eurasian Ice Sheet complex was also modeled by the thermomechanical approach validated against the available geochronological data on ice margin fluctuations (e.g., Patton et al., 2017). However, the geochronological constraints used in reconstructions and modeling come mainly from terrestrial settings, leaving marine areas much less recognized in terms of dating sites significant for the timing of the ice margin retreat.
During deglaciation, the last SIS formed distinct ice-marginal belts, traditionally ascribed to the three main phases of the last glaciation in Germany and Poland: Brandenburg, Frankfurt, and Pomeranian (Woldstedt, 1925(Woldstedt, , 1935. The local LGM occurred during the Brandenburg (Leszno) Phase (~24-23 ka) in Germany and western Poland, and later during the Frankfurt (Poznań) Phase (~19 ka) in eastern Poland and Belarus (Marks, 2012(Marks, , 2015. The subsequent ice-margin stillstand of the Pomeranian Phase is currently dated at ca. 17-16 ka (Marks, 2012) or 18-17 ka (Stroeven et al., 2016). This chronology has recently been established mainly by interpretation of calibrated radiocarbon ages (e.g., Marks, 2012), OSL dating (e.g., Wysota et al. 2009), and cosmogenic nuclide dating (e.g., Rinterknecht et al, 2006;Tylmann et al., 2019). However, much less is known about the timing of deglaciation phases along the present Baltic coast and within the southern Baltic basin. The chronology of the last deglaciation in the southern Baltic area was interpreted mostly from uncalibrated radiocarbon ages of organic sediments significant for constraining ice sheet advances and retreats, and the time-space correlation of relict ice-marginal landforms occurring on the southern Baltic floor with moraines in southern Sweden as well as in Lithuania, Latvia, and Estonia (e.g., Lundqvist, 1986;Rotnicki, Borówka, 1995;Uścinowicz, 1995Uścinowicz, , 1996Uścinowicz, , 1999Mojski, 1995Mojski, , 2000Marks 2002). As a consequence, the Gardno Phase was estimated at 14.5-13.8 ka BP, the Słupsk Bank Phase at 13.5-13.2 ka BP, and the Southern Middle Bank Phase at 13.0-12.8 ka BP. Later, those ages were calibrated to calendar years as 16.8-16.6 cal ka BP,.0 cal ka BP, respectively (e.g., Marks et al., 2016), although they seem too old and partly in conflict with the age of the Pomeranian Phase (17-16 ka) postulated by Marks (2012). The age of the Southern Middle Bank Phase was also corrected to 14.5 ka, according to newer, calibrated 14 C and 10 Be ages from southern Scandinavia, and the Słupsk Bank Phase was located between 15.5 and 14.5 ka (Uścinowicz, 2014). Recently, the age of the ice-margin stillstand of the Słupsk Bank Phase has been interpreted at ~15.2 ka on the basis of the OSL dating of glaciofluvial and glaciolacustrine deposits (Uścinowicz et al., 2019).
Despite the available geochronological data, the timing of the last deglaciation phases in this region and the correlation of subaqueous glacial landforms with terrestrial moraines are still problematic. The incomplete morphological record of particular ice margin positions and the inconsistency of the available dating constraints with morphostratigraphy make it difficult to establish any time-space correlations of the ice margin positions and to create time-slice reconstructions of the ice sheet (cf. Lüthgens, Böse, 2011;Marks, 2015;Hughes et al., 2016).
Our motivation for this contribution was to tackle these problems with a Bayesian approach to constraining the chronology of the last deglaciation in the southern Baltic area. The goal of the article is to propose a new chronology of the retreat of the last SIS in the southern Baltic basin based on the available geochronological data and morphostratigraphy. Here, we present a reconstruction of the ice margin retreat within a sector of the last SIS located in the southern Baltic basin.

Study area and geochronological constraints
The study area is located in the southern part of the Baltic basin and in the northern fringe of Poland ( Fig. 1A and B). It covers a part of the Polish middle-coast area with conspicuous moraines of the Gardno Phase and a part of the southern Baltic seafloor with icemarginal landforms of the Słupsk Bank Phase and the Southern Middle Bank Phase (Fig. 1B).
The morphology of the Baltic Sea floor north of the Gardno moraines is characterized by bulges oriented WSW-ENE (Słupsk Bank, Stilo Bank and Southern Middle Bank) separated by two linear depressions: an unnamed depression and the Słupsk Furrow. The Bornholm Basin limits the Słupsk Bank and the Southern Middle Banks from the west, whereas the Gdańsk Basin and the Eastern Gotland Basin limit them from the east (Fig. 1B).
Landforms and deposits formed in the Baltic basin during deglaciation were partly or completely destroyed by erosion due to southern Baltic transgression during the Holocene and/or became masked by marine sediments. However, some of them are still reflected in the seafloor morphology. Relicts of ice-marginal landforms occur in the area of the Słupsk Bank and the Southern Middle Bank at a depth of 16-30 m, recording the ice-margin stillstand during deglaciation after the Gardno Phase (Uścinowicz, 1995(Uścinowicz, , 1996(Uścinowicz, , 1999. They are correlated with the Słupsk Bank Phase and the Southern Middle Bank Phase of the last deglaciation in the southern Baltic area (Fig. 1B). Relicts of end moraines, ground moraines, boulder fields, glaciofluvial deltas, and ice-marginal lake plains were found on the Słupsk Bank (Pikies, 1995;Uścinowicz, 1999). Numerous remnants of end moraine ridges up to 10-14 m high with slope angles of 5-7° and a SW-NE orientation of the long axis occur in the western part of this ice-marginal zone (Kramarska, 1991a, b). There are also smaller ridges on the northern slope of the Słupsk Bank interpreted as De Geer moraines (Uścinowicz, 2010).
The area of the Southern Middle Bank also consists of relicts of glaciofluvial deltas and ground moraine (Pikies, 1995;Uścinowicz, 1999).
Luminescence dates of mineral sediments are available for the area between the Gardno and Słupsk Bank moraines and for the Southern Middle Bank (Fig. 1B). They come from glaciofluvial deltas and ice-marginal lake deposits or coastal ridges, and may be used to constrain deglaciation chronology, since glaciofluvial deltas and ice-marginal lakes are correlated with particular phases of the ice sheet retreat (Uścinowicz et al., 2019). OSL dates of glaciofluvial deltas on the Słupsk Bank range from 9.77 ± 0.83 to 21.30 ± 2.00 ka, whereas OSL dates of ice-marginal lake sediments deposited in front of the Słupsk Bank cover an extremely large spectrum of ages: from 11.09 ± 0.79 to 135.0 ± 12.0 ka. Sediments on coastal ridges in the Gardno-Łeba Lowland north east of the Gardo moraines and interpreted as morphological evidence of the southern extent of an ice-marginal lake formed south of the Słupsk Bank range from 11.03 ± 0.73 to 16.13 ± 0.94 ka (Uścinowicz et al., 2019). One TL date is available for glaciofluvial sand overlying till on the southern Middle Bank, and it shows the age of this sand to be 13.20 ± 2.00 ka (Kramarska et al., 1990). The conventional 14 C ages were calibrated according to the IntCal20 calibration curve (Reimer et al., 2020). For details see Tab. S1 in Supplementary Materials.

Materials and methods
The sites of sediment dating that were used in the Bayesian analysis are located in the Gardno-Łeba Lowland close to the ice marginal zone of the Gardno Phase and on the southern Baltic seafloor close to the ice marginal zones of the Słupsk Bank Phase and the Southern Middle Bank Phase (Fig. 2). Radiocarbon dates were obtained from 2 sites in the Gardno-Łeba Lowland, whereas luminescence dates come from sediments cored within the Southern Middle Bank, the Słupsk Bank, and seafloor depressions south of the Słupsk Bank ( Fig. 2A), and from sediments on coastal ridges in the Gardno-Łeba Lowland ( Fig. 2B and C). The chronology of deglaciation was modeled on the basis of the two oldest radiocarbon ages of sediments overlying the boulder pavement interpreted as remnants of till in the Gardno-Łeba Lowland (Rotnicki, Borówka, 1995), 26 OSL ages of glaciofluvial, glaciolacustrine, and coastal ridge deposits (Uścinowicz et al., 2019), and one TL age of glaciofluvial sand (Kramarska et al., 1990) (Tab. 1). Radiocarbon ages are based on conventional Liquid Scintillation Counter (LSC) measurements of bulk organic samples taken at the Gliwice Radiocarbon Laboratory (Rotnicki, Borówka, 1995). The conventional 14 C ages were calibrated with OxCal 4.4 according to the IntCal20 calibration curve (Reimer et al., 2020). OSL ages are based on opto-luminescence measurements of 90-125 µm quartz grains taken at the Gliwice Luminescence Laboratory using the single-aliquot regenerative-dose (SAR) protocol for determination of equivalent doses for individual aliquots (Uścinowicz et al., 2019). The Central Age Model (CAM) was used to calculate equivalent doses for particular samples on the basis of aliquot distribution (Galbraith et al., 1999). TL age is based on thermoluminescence measurements of 80-110 µm quartz grains taken at the University of Gdańsk Luminescence Laboratory using the regenerative-dose protocol (Kramarska et al., 1990). Moreover, to additionally constrain our chronological model, we used the available literature data related to the timing of (1) the Pomeranian Phase in Poland (Marks, 2012;Stroeven et al., 2016) and (2) the deglaciation of Gotland (Anjar et al., 2014).
We used 26 out of 33 available OSL ages of genetically diversified sand, gravel, and silt in the southern Baltic area correlated with the Słupsk Bank Phase (Uścinowicz et al., 2019; Tab. S2), which resulted from a preliminary selection based on their distribution and identification of 7 outliers (Fig. 3). The latter refer to glaciolacustrine sediments from an icemarginal lake and are clearly overestimated, probably due to an incomplete bleaching of sediment grains during their transport and deposition (Uścinowicz et al., 2019). The age range of the 26 OSL dates used in the modeling of the chronology is from 9.77 ± 0.83 to 21.30 ± 2.00 ka, whereas the range of dates identified as outliers and not included in the modeling is from 37.10 ± 2.40 to 135.0 ± 12.0 ka (Fig. 3).

Construction of the chronological model
The deglaciation chronology was inferred from a hypothetical "relative-order" model of the expected chronological order with events arranged into a pseudo-stratigraphical order reasoned solely from landform-sediment relationships independent of the numerical dating controls (Fig. 4). The Sequence model in OxCal was divided into a series of Phases, each representing the stages of deglaciation in the analyzed region which may be correlated with particular dating controls. Thus, each Phase consists of a group of dating controls and is separated by Boundary commands, which delimit the duration of each Phase and generate an age posterior density estimate. Moreover, we used After ("terminus post quem") and Before ("terminus ante quem") commands to constrain the chronology when stages evidently postdate or pre-date particular events. The whole Sequence is constrained by Boundary commands, which delimit the start and the end of the model (Fig. 4A).
The start of the Sequence was defined by the age of the Pomeranian Phase, which based on recent estimates for northern Poland (Marks, 2012;Stroeven et al., 2016) may be roughly constrained to 18-16 ka. Then, the "Gardno Phase" was introduced (Boundary) with the constraint that it must post-date (After) the Pomeranian Phase. The Phase "post-Gardno" consists of radiocarbon ages (Gd-4776 and Gd-6117) from the oldest organic deposits overlying the boulder pavement being a remnant of till correlated to the last ice advance in the Gardno-Łeba Lowland (Figs. 2A, and 4B). Subsequently, the "Słupsk Bank Phase" was introduced with Boundary and Phase consisting of a group of OSL dating controls (GdTL samples) from glaciofluvial deltas, ice-marginal lake deposits, and coastal ridge sediments (Uścinowicz et al., 2019). Then, the later stage of deglaciation, the "Southern Middle Bank Phase" was defined with a Boundary command as well as Phase consisting of one TL date (UG-793) of glaciofluvial sediments occurring on the Southern Middle Bank (Fig. 4A and B).
The constraint was that it had to pre-date (Before) the most likely timing of the deglaciation of Gotland (13.7-12.3 ka) based on recalculated surface exposure 10 Be ages of Anjar et al. (2014). The Sequence is closed with the Boundary "End" command. The notation of commands used to process the algorithms is available in the Supplementary Materials of the article (Tab. S3). The run of the Sequence model was conducted in the Outlier mode, which assumes that outliers are distributed according to a student T distribution with 5 degrees of freedom; the scale is allowed to lie anywhere between 10 0 to 10 4 years (Bronk Ramsey, 2009b). In the initial model, the dating controls were all entered with a prior probability of 0.05 of being an outlier. Ages that clearly do not fit the model (characterized by the agreement index with the model A < 10%) were treated as outliers and removed. We discuss ages identified as outliers and possible reasons of their incompatibility with the model below. Ages having a much higher agreement index with the initial model, but not higher than 60%, and exceeding the 0.05 threshold of probability of being outliers in the initial model results, were downweighted by being assigned a prior probability of 0.75 of being outliers. Then, a re-run of the same Sequence model was conducted for the chronological sequence without outliers and with down-weighted ages. Finally, the agreement index for the re-run model (Amodel) was used to evaluate the reliability of the chronological sequences obtained (Chiverrell et al., 2013).
Both input ages and modeled ages were reported with 1 σ uncertainty (68.2% probability).

Results
The model based on the assumed sequence of events and all dating controls (Tab. 1, Fig. 4A) showed a very poor agreement index Amodel = 0.5%. This suggested that the results of the initial Sequence were not reliable and some outliers and problematic ages must occur among the dating controls. We identified 10 outliers with the individual agreement index A < 10% (Tab. 1). One radiocarbon date belonging to the Phase "post-Gardno" (sample Gd-4776) showed a low agreement index of 5.3%. The calibrated age of this sample is 17.50 ± 0.24 ka b2k, and it is most probably too old an age for the post-Gardno phase, which may result from the redeposition of organic matter and the contamination of this sample with older carbon. It is highly probable because a bulk sample of organic material dispersed in ice-dammed lake clay and silt was dated (Rotnicki, Borówka, 1995). Moreover, no additional data supporting the apparent age of these samples (e.g. palynology) are available. Nine out of twenty-six OSL ages belonging to the Phase "Słupsk Bank" (samples GdTL) also had very low agreement indexes ranging between 3.9 and 9.2%. Most of these ages are too young (between 9.77 ± 0.83 ka and 11.45 ± 0.84 ka), which may result from a partial or total bleaching of grains after their deposition in the period of dead-ice blocks melting, just before and during marine transgression (Uścinowicz et al., 2019). The age of one sample is too old (21.30 ± 2.00 ka), which is most probably the result of an incomplete bleaching of grains during glaciofluvial transport. We also identified 3 OSL ages belonging to the Phase "Słupsk Bank" (samples GdTL) that had an agreement index with the initial model in the range of 33.6 to 47.6%.
These are OSL ages 12.35 ± 0.80, 12.16 ± 0.94, and 12.10 ± 0.83 ka, which are slightly too young within the Sequence (Tab. 1). The causes of the insufficient compatibility of these dates with the model could be similar to the ones mentioned above when analyzing outliers with very low agreement indexes (partial or total bleaching of grains after their deposition).
Thus, they were down-weighted by being assigned a prior probability of 0.75 of being outliers. Assuming this time frame for the last deglaciation of the southern Baltic basin and for the spatial distribution of the main ice-marginal landforms, a rough estimation of the average retreat rate of the ice sheet front was proposed (Fig. 5)  as earlier suggested by Raukas et al. (1995). Timing of deglaciation in the region of the Middle Lithuanian moraines was previously estimated at about 16.5-16.0 ka (Lasberg, Kalm, 2013) what is also similar to our modeling constraint for the Gardno Phase.  (Tylmann, et al. 2019) and for the Pomeranian Phase ice margin stillstand ~18-17 ka (Stroeven et al., 2016).
The retreat rate between Pomeranian Phase and Gardno Phase ice margin stillstands (~17.5 ka to ~16.5 ka) was slightly higher, and it may have amounted to at least ~70 m/yr. However, these results suggest that the retreat rate of the ice margin was relatively stable from the early stages of deglaciation after the local LGM, to the later stages of the ice sheet recession in the Baltic basin. The lack of a clear acceleration of the ice margin retreat rate in the Baltic basin compared to that in northern Poland indicates that, despite the formation of an extensive icedammed lake in front of the retreating ice sheet after the Gardno Phase (Uścinowicz, 1995(Uścinowicz, , 1999, the ice sheet margin was still grounded rather than floating. This may have prevented a more rapid recession, characteristic of a floating ice front, which usually amounts to hundreds or even thousands of meters per year (c.f., Dowdeswell et al., 2020). This situation probably persisted until the Southern Middle Bank Phase. Moreover, the retreat did not accelerate within the southern Baltic, as it was the case in the southern Sweden, probably also due to much thicker ice infilling southern Baltic basin in comparison to the southern Scandinavia.
An independent proxy for the ice margin retreat rate in the southern Baltic basin may be a geomorphological record available on the NW slope of the Słupsk Bank. A DTM obtained by the multibeam echosounding of this area reveals numerous moraine ridges with a distinctive spatial distribution and morphology (Fig. 6). These are relicts of closely spaced sub-parallel ridges orientated mainly NE-SW (Fig. 6A) and usually characterized by asymmetric cross profiles with steeper SE slopes (Fig. 6B). This system of ridges was interpreted as a remnant of De Geer moraines, which indicate a subaqueous type of deglaciation after the ice margin stillstand of the Słupsk Bank Phase (Uścinowicz 2010). The main arguments supporting the De Geer moraine interpretation of these ridges are as follows: (1) the regular spatial arrangement of closely spaced asymmetric ridges; (2) their morphometric parameters similar to those of typical De Geer moraines (up to hundreds of meters long, tens of meters wide, up to 10 m high, and spaced from tens to hundreds of meters apart); (3) their geological structure with boulders/cobbles on the top and till inside (Uścinowicz, 2010). We used them to estimate the possible retreat rate of the ice margin, assuming that this kind of moraines may have been formed annually (e.g., Bouvier et al, 2015;Sinclair et al., 2018). The distance between particular ridges varies from tens to hundreds of meters, and smaller ridges are usually spaced more closely (~60-90 m apart) than bigger ridges (mostly ~100-300 m apart). Moraines located to the NW of the main moraine belt correlated with the Słupsk Bank Phase show a minimum distance of ~65-100 m between particular ridges and a distance of ~100-220 m between the main ridges (Fig. 6B). It suggests more rapid recession than our estimation of the average retreat rate of the ice margin after the Słupsk Bank Phase, that is ~40 m/yr (Fig. 5). The estimation based on the modeled ages of ice margin stillstands is only an approximation of the ice margin retreat rate, which in fact varied over time and space, as perhaps reflected by the diversified spacing of De Geer moraine ridges. However, the minimum distance between closely spaced ridges (~65 m) is a similar order of magnitude as the ice margin retreat rate constrained by our modeled chronology (40 m/yr), so we assume that the moraines may have been formed annually. Perhaps, not all ridges were preserved at the seafloor since deglaciation (they are relicts of glacial landforms), what may explain their wider spacing of hundreds of meters. So, the estimated average retreat rate of the ice sheet after the Słupsk Bank Phase is at least partly confirmed by the seafloor geomorphology preserved on the NW slope of the Słupsk Bank.

Time-slice reconstruction
Our modeling included calibrated radiocarbon ages, luminescence ages, and other chronological constraints (relative sequence, age of Pomeranian Phase, and deglaciation of Gotland). Therefore, the chronology obtained is based on more solid lines of evidence than previous interpretations of uncalibrated radiocarbon ages (e.g., Rotnicki, Borówka, 1995) and time-space correlations of ice margin positions (e.g., Lundquist, 1994;Lagerlund et al., 1995;Raukas et al., 1995;Uścinowicz, 1996Uścinowicz, , 1999. The proposed chronology of the SIS retreat and the space-time correlation of ice margin positions during particular phases of deglaciation in the southern Baltic region make it possible to propose a time-slice reconstruction of the ice sheet. We reconstructed the most probable configuration of the southern front of the SIS during the time period between 18-17 ka and 14 ka (Fig. 7).
About 18-17 ka, the SIS margin was located in northern Germany, northern Poland, and southern Lithuania along the Pomeranian moraines. The modeled age of 16.5 ± 0.5 ka for the ice margin re-advance of the Gardno Phase indicates a rather rapid re-advance after the deglaciation of northern Poland during the Pomeranian Phase. However, during the ice-sheet retreat from the Pomeranian moraines to the Gardno moraines there was at least one or two stillstands known as the Mecklenburg Phase (e.g., Lagerlund et al., 1995, Rinterknecht et al., 2014 or the Rosenthal Phase and the Velgast Phase (e.g., Hoffmann, 2002). A few minor local glacial marginal zones located between the Pomeranian and the Gardno moraines are also shown by Marks (2005) in Pomerania in Poland. This means that up to the Gardno Phase the southern Baltic basin was still infilled with ice ( Fig. 7A). An important question arise, however, with regard to deglaciation of Bornholm. 10 Be surface exposure ages reported by Houmark-Nielsen et al. (2012) indicate the most likely timing of ice sheet retreat there between 17 and 16 ka. Mean 10 Be age calculated based on eight the most reliable ages from Bornholm was reported as 16.6 ± 0.9 ka (Anjar et al., 2014). However, based on the most recent global 10 Be production rate (Borchers et al., 2016) these ages are significantly older and mostly fall into the range 19-18 ka (Fig. 5). Using the recent Scandinavian reference 10 Be production rate (Stroeven et al., 2015) makes them ~2.8% older than ages calculated with the global 10 Be production rate. It is in conflict with age estimation for ice margin positions south and west of the Baltic basin (cf. Houmark-Nielsen, 2011;Marks, 2012;Stroeven et al., 2016;Tylmann et al. 2019), and also with chronology proposed in this research (Fig. 7). The reason could be that surface exposure ages from Bornholm are too old due to significant amount of inherited 10 Be occurring within particular ages, so that they do not indicate the actual timing of deglaciation there (cf. Houmark-Nielsen et al., 2012). Other scenario could be, if we assume the apparent 10 Be ages as good indicators of deglaciation chronology, that Bornholm massive remained ice free as a nunatak despite ice sheet advances between 18-17 ka and 15.5 ka. This scenario is, however, highly speculative as no other evidences about "nunatak history" of Bornholm was found so far.  The age of the two main ice margin stillstands included in our reconstruction is slightly younger that in the model proposed by Stroeven et al. (2016), who calculated deglaciation isochrones that may corresponds to these stillstands at 16 and 15 ka, respectively. Hughes et al. (2016), in their reconstruction of the Eurasian Ice Sheet evolution, also showed that the most probable limit of the southern front of the SIS falls within the time-slices of 16 ka at the Słupsk Bank and 15 ka at the Southern Middle Bank. Our modeling, however, is based on new available geochronological data from the southern Baltic seafloor, which were not included in either of the abovementioned reconstructions of the entire ice sheet. Moreover, the timing of the two main ice margin stillstands during the deglaciation of the southern Baltic basin roughly corresponds to the timing of two cool periods around 15 ka and 14 ka interpreted from paleotemperatures of Greenland based on ice core GISP2 (Alley, 2004). This suggests that the main stillstand phases of the SIS during the last deglaciation of the southern Baltic region were at least partly triggered by climatic fluctuations in the Northern Hemisphere.

Conclusions
The The average retreat rate of the ice margin within the southern Baltic basin was approximately 40-55 m/yr, which is comparable with the retreat rate estimated for deglaciation after the local LGM in northern Poland (~50 m/yr). This suggests that the retreat rate of the ice margin was relatively stable from the early stages of deglaciation after the local LGM to the later stages of the ice sheet recession within the Baltic basin. The margin of the ice sheet there was probably still grounded rather than floating, which may have prevented a more rapid recession and acceleration of the retreat rate in the Baltic basin. The two main ice margin stillstands during deglaciation roughly correspond to the two cool periods around 15 ka and 14 ka interpreted from paleotemperatures of Greenland based on ice core GISP2, which suggests that the last SIS in the southern Baltic responded to climatic fluctuations in the Northern Hemisphere.
The contribution of co-authors to the manuscript is the following: Declaration of competing interest: The authors declare that they have no known competing financial interests or personal relationships that could have influenced the work reported in this paper.