Holocene relative shore‐level changes and development of the Ģipka lagoon in the western Gulf of Riga

Holocene relative shore‐level changes and development of the Ģipka palaeolagoon in the western Gulf of Riga are reconstructed using multiproxy analyses by combining litho‐, biostratigraphical and chronological data with remote sensing and geophysical data. The results show the development of the Ģipka basin from the Ancylus Lake/Initial Litorina Sea coastal zone (before c. 9.1 cal. ka BP) to coastal fen (c. 9.1 to 8.4 cal. ka BP) and gradual development of the Litorina Sea lagoon (c. 8.4 to 4.8 cal. ka BP) and its transition to a freshwater coastal lake (c. 4.8 to 4.6 cal. ka BP), fen (c. 4.6 to 4.2 cal. ka BP), and river floodplain (since c. 4.2 cal. ka BP). The highest shorelines of the Ancylus Lake and Litorina Sea were mapped at an elevation of 12–11 and 9 m a.s.l., respectively. A new relative shore level (RSL) curve for the western Gulf of Riga was constructed based on RSL data from the Ģipka area and from nearby Ruhnu Island studied earlier. The reconstruction shows that the beginning of the last marine transgression in the western Gulf of Riga started at c. 8.4 cal. ka BP, and concurred with the 1.9 m RSL rise event recorded from the North Sea basin. Diatom analysis results indicate the existence of the Ģipka lagoon between c. 7.7 and 4.8 cal. ka BP, with the highest salinity c. 6.1 cal. ka BP. During the existence of the brackish lagoon, settlement sites of the Neolithic hunter–gatherer groups existed on the shores of the lagoon in the period c. 6.0 to 5.0 cal. ka BP.

An interplay between glacial-isostatic adjustment (GIA) and postglacial sea-level rise in the slowly uplifting coastal areas of the Baltic Sea in SW Sweden (Berglund et al. 2005;Yu et al. 2007;Hansson et al. 2018), NW Russia (Sandgren et al. 2004;Miettinen et al. 2007), SE Finland (Miettinen 2004) and Estonia (Rosentau et al. 2013;Nirgi et al. 2020) has led to complex relative shore level (RSL) changes with multiple transgressionregression cycles (Fig. 1A).Because the differences in land uplift rates were quite small, in some of these areas the Litorina Sea highest shorelines (c.7.5 to 6.5 cal.ka BP) are today located at lower elevation (Nirgi et al. 2020) compared with the older Ancylus Lake shores (c.10.2 cal.ka BP), whereas in other locations the situation is reversed (Hansson et al. 2018).These sensitive areas are very suitable for the study of Holocene RSL changes and isostatic uplift in order to calibrate global GIA models (Tushingham & Peltier 1992;Lambeck et al. 1998) and to detect the melting episodes of large ice sheets during the Early to Middle Holocene.
After the damming and drainage of the Ancylus Lake, the last marine inundation of the Baltic Sea Basin (BSB) started at c. 9.8 cal.ka BP (Bennike & Jensen 2013), initiated by the inflow of saline waters from the North Atlantic.This transgression during the Early Holocene most likely followed the opening of the present-day Great Belt and Fehmarn Belt region (Feldens & Schwarzer 2012).Rising sea-level resulted in flooding of the southern Kattegat area c. 9.3 cal.ka BP (Bendixen et al. 2017).However, the timing of the transition from fresh to brackish water that marks the onset of the Litorina Sea is not yet clearly determined and might have started in the northern Great Belt region c. 9.0 cal.ka BP and east of the Darss Sill after 8.5 cal.ka BP (Andr en et al. 2011).The cause of the saline water inflow into the BSB is believed to be related to episodic melting of the Laurentide and Antarctic ice sheets and the resulting rapid sea-level rise events (Andr en et al. 2011).In the Blekinge area (SW Sweden) the initial transgression (L1) of c. 2.5 m took place between 8.5 and 8.2 cal.ka BP (Berglund et al. 2005), while the main transgression was centred around 7.6 cal.ka BP (Berglund et al. 2005;Yu et al. 2007).The latter is detected as a major RSL rise event also in the Gulf of Riga and Gulf of Finland areas, centred around 7.8 cal.ka BP (Rosentau et al. 2013;Nirgi et al. 2020).In the Rhine-Meuse Delta of the North Sea basin, a major RSL rise event of c. 1.9 m has been dated to 8.45 to 8.2 cal.ka BP, while a younger RSL rise event is not evident in the North Sea RSL record (Hijma & Cohen 2019).The end of the rapid melting of the Laurentide and Antarctic ice sheets resulted in a culmination of the Middle Holocene RSL high stand, which in the eastern Gulf of Riga area occurred at c. 7.3 cal.ka BP (Rosentau et al. 2011;Nirgi et al. 2020).
Although there are a few recent detailed RSL studies from the northern, eastern and southern coasts of the Gulf of Riga (Saarse et al. 2009;Grudzinska et al. 2017;Muru et al. 2018;Nirgi et al. 2020), there is a lack of comparable data from the western coast of the gulf with a similar uplift history.Our study area in G ß ipka, at the western coast of the Gulf of Riga (Fig. 1), is one of the coastal areas with complex systems of coastal landforms and a palaeolagoon where previous investigations (Grinbergs 1957;Eberhards 2000Eberhards , 2006) ) have shown potential for understanding Holocene RSL and GIA variability from the Ancylus Lake to the last marine phase.
In this study, we aimed to reconstruct RSL variability at the G ß ipka lagoon, with the main emphases on the Middle Holocene RSL rise, the duration of the highstand and following decline, as linked to the melting events of the large ice caps and related RSL rise events.We also aimed to provide palaeoenvironmental reconstructions of the G ß ipka lagoon system to reconstruct the site preferences of the hunter-gatherer communities that previous archaeological studies have shown were present in the area.The integration of archaeological and palaeoenvironmental records hopefully facilitates  Muru et al. 2017).C. Airborne Lidar topography of the study area with the location of the studied G ß ipka section (black triangle; 22°37 0 31 00 E; 57°34 0 36 00 N), the location of geological profile A-B and ground penetrating radar (GPR) profile C-D (Fig. 7), P urciems (A-F) and G ß ipka (A, B) Stone Age settlement sites (black diamonds, detail map with archaeological sites shown in Fig. 8).Locations of the Baltic Ice Lake coastal landforms at elevation 36-32 m a.s.l.(Rosentau et al. 2009) and Litorina Sea coastal landforms at elevation 10-9 m a.s.l.(Eberhards 2006) are also shown on the map.further testing of models of human behaviouralecological interactions (e.g. at lake or seashores), fundamental for understanding the hunter-gatherer lifeways (Gerasimov & Kriiska 2018;Groß et al. 2018).We present here the results of sedimentological, plant macrofossil and diatom analyses with new radiocarbon AMS and optically stimulated luminescence (OSL) ages together with newly available airborne Lidar elevation data, ground penetrating radar (GPR) survey results, and RSL data from the neighbouring areas (Kalnin ßa & Eberhards 2006;Muru et al. 2018).

Geological and archaeological background
The former Litorina Sea lagoon in the G ß ipka area, nowadays completely filled in andwell drained, is located at the NW coast of the Gulf of Riga in the Irve Plain of the Coastal Lowland (Fig. 1).This slightly undulating plain stretching NW to SE, ~20 km long and up to 1 km wide, at an altitude of 6.4-9.4 m, is isolated from the Baltic Sea by the beach ridge-dune system of the Litorina Sea, reaching altitudes of 14-18 m.To the W and SW, the G ß ipka lagoon is bordered by the coastal plains, with Holocene accumulative coastal formations at altitude 12-13 m a.s.l. and coastal formations of the Baltic Ice Lake (32-36 m a.s.l.), located along the eastern slope of the Northern Kursa Upland (Fig. 1C).
The coastal plains fringing the NW coast of the Gulf of Riga are up to 6-7 km wide, and in our study area consist of three rather clear units.The Baltic Ice Lake plain, ~6 km wide, is the most distant from the sea and highest, with a surface altitude ranging from 36 m to 13 m a.s.l. in an EWdirection.An accumulative coastal ridge-dune complex above 32 m a.s.l.marks the Baltic Ice Lake shorelines (Fig. 1C).Closer to the sea is the Litorina Sea lagoonal plain, with surface elevations between 9.4 and 6.4 m a.s.l., and the highest point, at 10-9 m a.s.l., marks the highest Litorina Sea (Lit a) coastline along the western edge of the plain (Fig. 1C; Grinbergs 1957;Eberhards 2000Eberhards , 2006)).The lagoonal plain is drained by several SE-flowing meandering streams which have been partially straightened.The river Pilsupe (Fig. 1C), at the bank of which the studied G ß ipka outcrop is located, crosses the lagoon in a SW-NE direction and was originally a tributary of the larger river Vecupe and river Roja (Fig. 1C), the latter discharging into the Gulf of Riga in the southern part of the G ß ipka lagoon.In c. 1885, an artificial channel was dug through the Litorina Sea beach ridge-dune system allowing the river Pilsupe to discharge directly into the Gulf of Riga (Fig. 1C; Nomals 1930).After the channel was dredged in 1926/ 1927, fluvial erosion accelerated and led to the incision of Pilsupe, resulting in the exposure/outcropping of lagoonal gyttja layers along the banks (Galeniece 1928;Nomals 1930).The currently studied G ß ipka outcrop is located in the eastern part of the G ß ipka lagoon, at the left bank of the Pilsupe, close to the landward foot of the beach ridge-dune system of the Litorina Sea (Fig. 1C).This nearly 1-km-wide barrier spit separates the lagoon from the present coastline and represents a ridge-swale complex.According to Kalnin ßa & Eberhards (2006), this pattern is screened by elongated or parabolic dunes.
The altitude of the bedrock surface of Middle Devonian sandstones ranges from 0 m at the Ancylus Lake highest shoreline down to À15 m a.s.l. at the present coastline (Grinbergs et al. 1975;Veinbergs 1996).There is a pronounced scarp in the bedrock surface close to Ancylus Lake highest shoreline, where the bedrock surface altitude falls from 10 to 0 m a.s.l., which is amplified in the modern topography.The thickness of Quaternary cover is ~20 m at the coast, and less than 10 m along the Baltic Ice Lake abrasional plain.The Quaternary strata are composed of thin layers of Late Weichselian till and glaciolacustrine clay, covered mainly by marine and aeolian sandy sediments.Baltic Ice Lake and Ancylus Lake sediments occur only in places on the Irve Plain and reach a total thickness of up to 5 m.The Litorina Sea sandy sediments dominate the marine complex, covered by aeolian sand in the coastal ridge system of the Litorina Sea.An up to 6-mthick succession of gyttja, buried peat and organic-rich silty or clayey sands has been described in the G ß ipka area, associated with the former Litorina Sea lagoon, which is of major interest in the current study (Meirons 1993).
The first biostratigraphic investigation of the G ß ipka lagoon sediments was carried out by Galeniece (1928), who first interpreted two gyttja layers separated by fen peat as brackish-water Litorina Sea lagoonal deposits.Grinbergs (1957) and Veinbergs (1996) supported the conclusion about the brackish-water Litorina Sea lagoon by showing that the parallel dune ridge series on top of the barrier spit at the western edge of the G ß ipka lagoon represents the highest Litorina Sea coastline (Lit a) at 9.7 m a.s.l. and that two layers of gyttja represent two Litorina Sea transgressive phases with an intermediate lowstand marked by fen peat.It should be noted that neither of these investigations mention a lower fen peat horizon at an altitude c. 3 m a.s.l., which was first reported by Grinbergs et al. (1975).Grinbergs et al. (1975) describes an interval of wood peat containing a large amount of Betula pollen (58%), overlying a fine silt and sand horizon.They correlated the peat accumulation with the Yoldia Sea lowstand.Grinbergs et al. (1975) also presented the first chronological studies into the G ß ipka sediments, giving a conventional radiocarbon age of 8895AE85 years (Tln-10) for the lower fen peat and 4440AE75 years (Tln-13) for the lower part of the gyttja.
The first archaeological finds at P urciems were sherds of Neolithic pottery discovered in 1933 by Burchard and 1937 by Sturms along the actively eroding sand scarp on both sides of the Pilsupe river ( Sturms 1937).Six occupation units (dwellings) were discovered, four of them with in-situ stratigraphy (dwellings B-D and F), and another two (A and E) regarded as consisting of redeposited material ( Sturms 1937) (Fig. 1C).Later, Loze (2005) re-located dwelling F, where erosion was continuing, and discovered more Neolithic archaeological sites, G ß ipka A and G ß ipka B (Loze 2005(Loze , 2006)).

Sediment sampling
Sediment sampling of the 6-m-long G ß ipka outcrop on the left bank of the river Pilsupe (22°37 0 31 00 E, 57°34 0 36 00 N; 7.3 m a.s.l.) (top of the outcrop) was performed in 2017.A series of overlapping 1-m-long monoliths were extracted from the upper 4.5-m-long portion of the vertical succession containing organic matter.Ten-centimetre-diameter U-shaped plastic tubes were pressed into the sediment, cut from the back side and backed into plastic film for the transport.Monoliths were later used for subsampling for biostratigraphy and terrestrial plant macrofossils for further radiocarbon AMS dating.
Three samples for OSL dating were taken from three sand horizons described in the text by hammering 30-cmlong opaque plastic tubes without light exposure.Altogether six ~250 g sand sediment samples were collected from the succession for microtextural properties of mineral grains.

Luminescence dating
The time frame of sand sediment deposition was based on three OSL samples.The sampling tubes were opened under darkroom red subdued light conditions at the Lund Luminescence Laboratory, Sweden.The sediment at each end of the tube was retained for dose rate and water content determination.The sediment from the internal part of the tube was wet sieved, and the 180-250 lm fraction was treated with HCl, H 2 O 2 and density separation.In this way, the quartz extracts were obtained and further treated with HF and HCl.Finally, the quartz extracts were dried and re-sieved on the 180 and 250 lm set of sieves.
High-resolution gamma spectrometers at the Nordic Laboratory for Luminescence Dating, Aarhus University, Denmark were used for measurement (Murray et al. 1987), and the final environmental doses were calculated using the DRAC online calculator (Durcan et al. 2015).
Considering the proximity to the coast and several submergence-emergence events, average water content was assumed to be similar to the field water content for 50% of the depositional time and saturated for 50% of the time.
To obtain the equivalent dose, large (8 mm) single aliquots of quartz extracts were analysed in a Risø TL/ OSL reader DA-20 with a single aliquot regeneration protocol (Murray & Wintle 2000, 2003) preceded by IR/ blue ratio, preheating plateau and dose-recovery tests.Blue light sources (470AE30 nm; ~65 mW cm À2 ) and detection through 7 mm of U340 glass filter were used for OSL stimulation.For two samples, 17081 and À82, blue stimulation was used and post-infrared blue stimulation for the sample 17080 (Roberts & Wintle 2001).Equivalent doses (D e ) were calculated in the Risø Analyst 4.31 software.Aliquots were accepted if they revealed a test dose error of <10% (with the exception of sample 17082, where 15% was set), a recycling ratio within 10% of unity (15% for 17082), recuperation <5% of the natural, and for 17080, an IR/B ratio of ≤10%.Between 23 and 30 aliquots per sample were measured, but ~15% of these were rejected owing to poor recycling ratio or dim signals, thus resulting in 21-26 accepted aliquots (Table 1).The Central Age Model was used (Galbraith et al. 1999) for final age calculations.

AMS dating and age-depth modelling
Terrestrial plant macrofossils identified during macrofossil analysis were radiocarbon dated using the AMS 14 C dating method at Poznan Radiocarbon Laboratory, Poland.Altogether, seven AMS radiocarbon ages were obtained from the G ß ipka sediment succession, calibrated Wood charcoal samples preserved from the excavation of Neolithic dwellings at P urciems in the 1930s were AMS radiocarbon dated at the Laboratory of Mass Spectrometry, Centre for Physical Sciences and Technology, Lithuania, and calibrated as above.

Diatom analysis
In total, 32 samples for diatom analysis were prepared according to the procedure described by Battarbee et al. (2001).A drop of the enriched diatom sample was spread over a coverslip, dried overnight at room temperature, and mounted in Naphrax TM (refractive index À1.73) on a microscope slide.The quantitative analyses were performed with a Zeiss Axiophot light microscope under a 1009 oil immersion DIC objective.Three samples from the bottommost part of the sediment sequence at the depths 430, 448 and 450 cm were barren of diatoms.In 28 samples, ~230-300 valves (mean count 265 valves per sample) were counted and identified to species level to estimate the relative abundance of taxa.In one sample (at 355 cm) diatom preservation was very poor, and therefore the diatom count was below 200 valves.The diatom identification criteria and ecological information, including salinity preference, were obtained from the Baltic Sea intercalibration guides (Snoeijs 1993;Snoeijs & Vilbaste 1994;Snoeijs & Potapova 1995;Snoeijs & Kasperovi cien_ e 1996;Snoeijs & Balashova 1998) and other well-established diatom floras (Krammer & Lange-Bertalot 1986, 1988, 1991a, b;Witkowski et al. 2000;Lange-Bertalot et al. 2017) and literature (Denys 1991;van Dam et al. 1994), as well as from internet sources (World Register of Marine Species, https://www.marinespecies.org/,and AlgaeBase, Guiry & Guiry 2022).Diatoms were divided into groups according to their salinity tolerance: marine/brackish, halophilous, small fragilarioid taxa with brackish-water affinity, small fragilarioid taxa preferring fresh water, indifferent and freshwater taxa.Diatoms were also classified into groups by their preferred habitat: planktonic and periphytic taxa.Additionally, small fragilarioid taxa as a taxonomically complex and morphologically variable group that can live in benthos, as well as in the water column up to depths of 6-8 m (Brugam et al. 1998) as tychoplankton, were separated from periphytic taxa.This is a common practice in diatom studies from isolation basins, since this group is considered to represent pioneer taxa, which are able to adapt to rapid shifts in environmental conditions faster and more effectively than other diatoms (Lotter & Bigler 2000;Yu et al. 2004).The predominance of small fragilarioid taxa is related to environmental instability (Denys 1990) and in isolation basin studies has been observed just prior to, during or after the actual isolation event (Stabell 1985;Heinsalu et al. 2000;Risberg et al. 2005;Miettinen et al. 2007;Saarse et al. 2009;Grudzinska et al. 2017).
In order to facilitate interpretation of diatom data, stratigraphically constrained incremental sum of squares cluster analysis was performed on the full percentage diatom data to define diatom assemblage zones (DAZ) using Tilia CONISS (Grimm 2011).Detrended correspondence analysis (DCA; Hill & Gauch 1980) with detrending by segments was applied to detect major patterns of variation in the diatom assemblage data during the different development stages of the ancient G ß ipka lagoon.Detrended correspondence analysis was used as the diatom data had a gradient 2.7 standard deviation (SD) units long and to avoid an obvious arch effect resulting from the use of principal components analysis (Hill & Gauch 1980;Birks 1995).The DCA was performed using CANOCO 5.1 (after ter Braak & Smilauer 2018).As DCA is sensitive to rare species, the option in CANOCO 'downweight rare species' was applied.All taxa were square root transformed in order to stabilize the variances and reduce the effects of very abundant taxa (Birks 1995).Lithostratigraphy and diatom data were compiled using the program Tilia v.1.7.16 (Grimm 2011) and compiled in CorelDraw2019.

Plant macrofossil analysis
Altogether, 15 samples were analysed for plant macrofossils.One-centimetre samples with a volume ranging from 100 to 500 mL were collected at different depth intervals between 4.51 and 2.09 m to cover sediment units G2-G5.Preparation for plant macrofossil analysis followed conventional procedures (Birks 2001), using a sieve with 250-lm mesh size.A Zeiss Stemi 2000-C stereomicroscope was used to determine plant macroremains and select the remains of terrestrial plants for 14 C AMS dating.Relevant literature and atlases (Katz et al. 1965;Cappers et al. 2006;Velichkevich & Zastawniak 2006, 2008;Bojnansky & Farga sova 2007) along with a reference collection at the University of Latvia were used for plant macrofossil identification.

Textural properties of the sand
Scanning electron microscope analysis of the microtextural properties of mineral (quartz) grains permits detection of the type of sedimentary environment, namely aeolian, fluvial, glacial and coastal (Mahaney 2002;Smith et al. 2018).

GPR survey
Geophysical studies were carried out using ground penetrating radar (GPR), which offers good data coverage and penetration depth in sandy deposits.The GPR fieldwork was performed with a Zond 12-e doublechannel system (Radar System Inc.) using a commonoffset configuration with a co-polarized 300 MHz shielded antenna oriented perpendicular to the profile.Similar surveys in the coastal areas of Latvia indicate that 300 MHz antenna is the best compromise between the penetration depth and resolution (Karu ss et al. 2021;Lamsters et al. 2022).The time range for the measurements was 500 ns.A GPR survey was carried out in 2020 along a profile covering 2.2 km perpendicular to the coastline and zones of major coastal landforms (Fig. 1C).The equipment and settings used resulted in clear reflections down to 20 m depth in sandy sediments.
The GPR data were processed and interpreted with Prism 2.61 software.The processing routine included application of time-dependent signal gain function, Ormsby band-pass filter and background removal filter.Conversion from time to a depth scale was performed assuming that characteristic dielectric permittivity is 6, which corresponds to dry sand (Neal 2004).Three boreholes were drilled by hand auger down to a depth of 4.3 m along the GPR profile to verify the interpretation of the obtained reflections.For tracking, a GNSS RTK Emlid Reach RS2 receiver was used, connected to the radar.

Methods for RSL reconstruction
The RSL data from the G ß ipka section were used together with the RSL index points from the nearby Ruhnu Island (Muru et al. 2018) with similar isostatic uplift to compile a composite RSL curve for the western Gulf of Riga.Vertical and horizontal uncertainties of the RSL data were evaluated using the HOLSEA database format (Hijma et al. 2015) and are available in Table S3.

Methods for palaeogeographical modelling
The topographical data for the area around the G ß ipka site were provided in the form of a 1 9 1 m resolution Lidar digital terrain model (DTM).Palaeogeographical reconstructions were based on a GIS approach (Rosentau et al. 2011), where the palaeosea-level surfaces were subtracted from the DTM.
The palaeosea-level surface for the Litorina Sea maximum stand was interpolated using databases of coastal formations by Saarse et al. (2003Saarse et al. ( , 2007)).The surface had a vertical error of AE1 m.This was used as a reference surface, and the palaeosea-level surface for 5.5 cal.ka BP was calculated considering a linear decrease in shoreline tilting and the water level taken from the compiled RSL curve based on data from the Ruhnu Island (Muru et al. 2018) and the results of the present study.Palaeoshorelines were delineated based on analysis of the elevation and morphology of the coastal ridge-swale complex.Coastal landforms seawards of a palaeoshoreline, younger than the time modelled, were cut by the sea-level surface and then the topography of the whole palaeosea area was generalized to 50 9 50 m resolution and smoothed with a 350 9 350 m averaging filter.
In the Litorina Sea lagoon area, sediments younger than the time modelled were removed from the DTM.A lagoonal sediment body was modelled based on the shape and cross-section of the lagoon and the depth and ages of the sediments recovered from the outcrop at G ß ipka site (Fig. 1C).Separate sediment models were created for the time slices 7.3 and 5.5 cal.ka BP (maximum sediment thickness 3.5 and 2.65 m, respectively).In the lagoon area, the river valleys, cutting deep into the lagoonal sediments, were filled up to the level of the surrounding terrain (in general 6 m a.s.l. in the present DTM).Then the DTM was generalized to 10 9 10 m resolution and smoothed with a 50 9 50 m averaging filter to remove noise from modern landscape features inside the lagoon.The respective modelled lagoonal sediment body and the palaeosea-level surface were then subtracted from the modified DTM to create the palaeogeographical reconstruction of the time slice.
Altogether, seven AMS radiocarbon dates (Table 1) and three luminescence dates (Table 2) were obtained from this sediment succession.Two radiocarbon dates (Poz-107835 and Poz-107838) from the unit G3 were considered too old compared with the luminescence age from the same unit and compared with AMS dates (Poz-107868 and Poz-107869) from peat layer (G2) below (Fig. 2, Table 1).Dated plant material from the G2 peat succession (Table 1) is considered in situ origin as it represents typical fen (dominance of Carex spp.) to coastal wetland (a decrease in Carex spp.and an increase in aquatic plants -Cladium mariscus and Menyanthes trifoliata) vegetation, identified using plant macrofossil analyses (Table S2).In contrast, the redeposition of the terrestrial macrofossils (Poz-107835 and Poz-107838) from the former land surface is suspected at the time of the accumulation of G3 sand layer, being the likely reason for older ages of these samples.The final age-depth model is based on five radiocarbon dates and one luminescence date from the uppermost unit (G6) and covers the main litho-and biostratigraphic boundaries (Fig. 2).
The lowermost unit G1 (interval 600-450 cm) consists of several series of sands, where upward-coarsening alternates with fine to medium beach sand containing well-rounded gravel and pebbles (Fig. 2).Lateglacial clayey silt deposits underlie the G1 unit.Fine sand from the upper part of the unit was OSL dated to c. 9.3AE1.5 ka.Sediments of the unit G1 are covered by a 25-cm-thick layer (interval 450-425 cm) of welldecomposed dense fen peat (unit G2) with an average OM content of 85% (Fig. 2).According to the age-depth model, the accumulation of this lower fen peat unit in the G ß ipka sequence took place between c. 9.1 and 8.3 cal.ka BP (Figs 2, 3).The peat is covered by a 25-cm-thick layer of fine sand, containing OM interbeds a few centimetres thick (unit G3), typical for transgressive sands in the coastal area around the Gulf of Riga.Its average OM content is 4%, and the carbonate content is 1-7%, being higher (up to 7%) in the lowermost 15 cm compared with the upper portion (1-2%) of unit G3.According to the age-depth model the deposition of unit G3 took place between c. 8.3 and 7.7 cal.ka BP (Fig. 2).This age range is also confirmed by a single OSL age of c. 7.35AE0.44ka, obtained from the middle of unit G3.The sand layer is followed by a 63-cm-thick layer (interval 380-227 cm) of grey clayey gyttja (unit G4A) with an average OM content of 25%.In the lowermost 10 cm, fine sand lenses and peat fragments also occur.According to the agedepth model, deposition of this clayey gyttja took place between c. 7.7 and 4.85 cal.ka BP (Fig. 2).The clayey gyttja is overlain by a 17-cm-thick layer (interval 227-210 cm) of brown detritus gyttja (unit G4B) with a much higher OM content (average 38%) compared with unit G4A.According to the age-depth model, its deposition took place between c. 4.85 and 4.6 cal.ka BP (Fig. 2).This is followed by accumulation of a 45-cm-thick layer (interval 210-165 cm) of fen peat (unit G5) with an average OM content of 79%.Accumulation of the upper fen peat took place between c. 4.6 and 4.2 cal.ka BP and was followed by deposition of a 145-cm-thick layer (interval 165-20 cm) of partly cross-bedded alluvial silt to fine sand (unit G6) with an average OM content of 6%.Fine sand from the lowermost part of unit G6 was OSL dated to c. 3.96AE0.24ka (Fig. 2).

Diatom stratigraphy
A 2.73-m-long interval (155-428 cm) of the G ß ipka sequence was studied for diatoms.In total, 232 diatom taxa belonging to 70 genera were identified.The relative frequency of the most abundant diatom species is presented in Fig. 3.A table describing the diatom taxa in detail is given in Table S1.The diatom stratigraphy was divided into eight DAZ defined by cluster analysis (CONISS).There are clear changes in the diatom composition, showing a gradual transition from a freshwater to a brackish water environment, and finally isolation and the formation of a shallow lake (Fig. 3).

Statistical analyses of the diatom assemblages
The eigenvalues of the first two DCA axes are 0.43 and 0.13, respectively, and together these two axes explain 24% of the cumulative variation within the diatom data (Fig. S1).The first axis scores of DCA showed a gradient length of 2.3 SD units in the diatom assemblages and is plotted against sediment depth (Fig. 4).DCA 1 SD values >2 indicate freshwater conditions (DAZ-I and -II); SD 1.7-1.8,fresh water with slight brackish water influence (DAZ-VII and -VIII); SD 1.3-1.4,transition from fresh water to a brackish environment (DAZ-III); SD 0.6-0.9, a brackish environment with high turbidity (DAZ-IV to -VI); and SD <0.5, a brackish environment (DAZ-V to -VI).

Plant macrofossils
The plant macrofossil taxa from the G ß ipka section are given in Table S2.In the lower peat unit (G2) five samples were analysed (Table S2), three of which are from the base and two from the upper portion of the peat, showing a slightly different environment.In the lower part (449-450 cm) well-decomposed dense sandy peat contains detritus predominantly from small plants: rare Hypnales moss leaves and stems, mycorrhizal fungus Cenococcum geophilum sclerotia, rounded wood fragments and charcoal, hemp-agrimony (Eupatorium cannabinum), rushes (Juncus) seeds and Carex fruits.This interval is followed by hypnum-grass peat (448-449 cm) dominated by Hypnales leaves of herbaceous plants, Carex fruits, rare fruits of wet swampy meadows and fen plants: tormentil (Potentilla erecta) and purple marchlocks (Comarum palustre).The top part of the peat unit G2 at the transition from medium decomposed fen to silty gyttja (425-427 cm) is rich in Hypnales moss, vascular plant stems, leaf epidermis and aquatic animal chitin fragments.Together with the seeds of white water lily (Nymphaea alba), the triple-leaved bogbean Menyanthes trifoliata and fruits of great fen sedge (Cladium mariscus) occur.
In the overlying sand unit with OM interbeds (G3; 400-396 cm) small, partially rounded flat fragments of wood and charcoal are regularly present, along with mycorrhizal fungus Cenococcum geophilum sclerotia, megaspora of Selaginella selaginoides, needles of Pinus sylvestris and seeds of plants of different habitats (Table S2).
Three samples were analysed from the overlying grey clayey gyttja unit (G4a; 374-371 cm).The remains of aquatic plants dominate: seeds of Nuphar lutea, Nymphaea alba and Najas marina, and fruits of Ceratophyllum demersum and Zannichellia palustris (Table S2).In addition, the horned pondweed (Z.palustris) together with widgeon grass Ruppia maritima have previously been described from this sediment unit from the neighbouring borehole 16a (Kalnin ßa & Eberhards 2006).The horned pondweed prefers a stagnant or slowly flowing water environmentboth fresh and brackish water.Following Gessner (1957), Z. palustris belongs to a group of polyhaline and mesohaline plants found in waters with highly variable salinity ranging from 35 to 5&, while widgeon grass prefers a brackish water environment (Gorishina 1979).Two samples were analysed from the upper portion of the gyttja interval (unit G4B; Table S2).Among the representatives of the aquatic community, fragments of water chestnut Trapa natans fruit were found in the lower interval (225-227 cm).Remains of Stratoites aloides, Eleocharis palustris and Menyanthes trifoliata were found in the upper interval (214-215 cm).
Onlyone sample was analysed from the upper peat unit (G5; 210-209 cm).In this dense wood-grass peat, tree branches with bark, and bark and wood fragments dominate, along with Alnus glutinosa and Betula sect.Albae fruits and seeds of plants growing in shallow water, Eleocharis palustris and Menyanthes trifoliata, as well as sedge fruits were found.
Subangular (Fig. S2A) and rounded quartz grains (Fig. S2B) dominate among all investigated sand samples.Angular grains (Fig. S2C) occur only in unit G1 (up to 15%).Among mechanical microtextures, different-sized conchoidal features are common and abundant (>50% of investigated grains, Figs 5, S2D), and enriched with both straight and arcuate steps (Fig. S2E).Crescentic (Fig. S2F) and V-shaped percussion marks (Fig. S2G) occur commonly, except for the grains from G6 unit, where occurrence of these marks is the lowest among all investigated samples.Nearly all grains have precipitation, usually in depressions and holes, and show solution pits on their surface (Fig. S2H).The frequency of chemically induced, oriented etch pits is either sparse or moderate (Fig. S2I,J), but only 7% of grains from unit G3 have such pits on their surface.Finally, grains are mostly of medium surface relief (Fig. S2K).However, high-relief grains (Fig. S2L) occur moderately and sparsely in G1, and are absent in units G3 and G6.

Palaeoenvironmental interpretation of the G ß ipka sediment succession
The palaeoenvironmental interpretation of the G ß ipka sediment succession is based on the above-described sediment and diatom stratigraphy, textural and structural properties of the sand units, plant macrofossil results, age-depth modelling based on AMS radiocarbon and luminescence chronology as well as GIS based palaeogeographical reconstruction.Sediment stratigraphy of the G ß ipka section is supplemented by newly obtained coring and GPR data along the west-eastoriented cross-section (Fig. 7), which is supplemented by earlier geological data by Eberhards (2003).All this shows a gradual transition from a freshwater coastal environment to a brackish-water environment, followed by isolation of a lagoon and formation of a shallow lake.Such a general trend in palaeoenvironmental change was interrupted by two low-water episodes, when the G ß ipka palaeolagoon turned into a wetland, reflected in two fen peat intervals in the studied G ß ipka sequence.
Ancylus Lake/Initial Litorina Sea coastal zone (before c. 9.1 cal.ka BP).-This period corresponds to the lowermost sand horizon (unit G1).Our primary expectation was that this sand complex probably reflects a high-energy coastal environment owing to the coastal site location (Fig. 7A).This assumption is supported by sedimentary structures and the texture of the sand of unit G1.For example, a thick-bedded sand of multiple reversed grading layers, a medium-bedded sand of multiple normal grading layers (Phantuwongraj et al. 2013) and mud rip-up clasts in a sand matrix (Kempf et al. 2017) are interpreted as indicating storm surge washovers.Also, upwards grading layers and mud rip-up clasts occur only in the lower sand (G1) at G ß ipka, meaning that this horizon has been affected by past storm action.This is additionally supported by quartz grain surface microtextures (Fig. 6), such as percussion marks, fresh surfaces and others (Costa et al. 2012, b).The uppermost part of the coastal sand complex, at elevation 1.3-2.8m a.s.l.(unit G1), was deposited at c. 9.3AE1.5 ka, during the Initial Litorina Sea, suggesting that the whole regressive complex occurs at the time when the retreating shoreline of the Ancylus Lake/Initial Litorina Sea passed the G ß ipka area.New Lidar elevation data and GPR results show that this nearshore sand complex extended up to 2 km landward from the G ß ipka section, forming a well-developed belt of coastal landforms with thick aeolian capping, most probably marking the highest shoreline position of the Ancylus Lake (Fig. 7).Fine grained clayey-silty deposits were described in core B3 (Fig. 8), behind these coastal landforms, suggesting the existence of an Ancylus Lake lagoon.We interpret GPR reflector 1 as the surface of the glacial till rising below the Ancylus Lake coastal landform up to the elevation 7 m a.s.l. and reflector 2 up to the elevation 13 m a.s.l. as the contact between Ancylus Lake coastal sands and aeolian capping.
Coastal fen (c.9.1 to 8.4 cal.ka BP).-This period corresponds to the major part of the lower fen peat layer (unit G2).The onset of the lower peat accumulation at c. 9.1 cal.ka BP shows that RSL dropped below 3 m a.s.l.during the regression and that fen peat accumulation in the coastal depression at G ß ipka started soon after the water level decrease in the study area.The occurrence of plant macrofossils of coastal wetland habitats, such as Menyanthes trifoliata, Cladium mariscus, Juncus, Carex and Eupatorium cannabinum (Table S2), suggests fen formation in a coastal area that was probably also seasonally flooded, as suggested by the presence of mineral matter in the peat layer (Fig. 2).
Transition from fen to coastal wetland (c.8.4 to 8.1 cal.ka BP).-This period corresponds to the uppermost part of the lower peat layer (unit G2) and the lowermost part of the sand layer (unit G3) of DAZ-I (Fig. 3).In the uppermost 3 cm of the peat, the first diatoms appear, showing a transition from fen to coastal wetland owing to the rising water table at c. 8.4 cal.ka BP, as derived from our age-depth modelling.The diatom composition suggests a freshwater environment, but also includes small fragilarioid taxa preferring a slightly brackish environment, which might be an indication of seawater intrusion.Also, the abundance of macrofossils of various aquatic plants, such as Nymphaea alba, Cladium mariscus, Potamogeton, Scirpus lacustris, Menyanthes trifoliata, Batrachium and Myriophyllum verticillatum along with Carex (Table S2), is characteristic for this environment.Changes in lithology from sandy peat accumulation to fine sand deposition with redeposited OM interbeds are interpreted as marking the beginning of the marine transgression in the G ß ipka area.
Shallow-water coastal estuary (c.8.1 to 7.7 cal.ka BP).-This period corresponds to the major part of the sand layer with OM interbeds (unit G3) in DAZ-II (Fig. 3).Rapid changes in diatom assemblages along with alternating mineral and organic-rich sediment layers suggest a highly variable environment, as in an estuary with a strong interplay between sea-level rise and freshwater discharge from a river.Also, the similar composition of the plant taxa compared with the underlying unit G2 indicates probable redeposition owing to activation of erosional processes along the G ß ipka basin.
The fine sand of unit G3 lacks angular quartz grains, and this combines with the rare occurrence of high-relief grains.These features distinguish this horizon from the sands of unit G1 and, combined with rather stable grainsize properties (mean, sorting, skewness) and (semi-) horizontal/cross-bedding, argue for an environment of lower energy, rather than intense storm action.In coastal environments subangular to rounded grains with silica precipitation and enriched with oriented etch pits and V-shaped percussion cracks are likely (Krinsley & Doornkamp 1973;Mahaney 2002;Kali nska-Narti sa et al. 2017;Itamiya et al. 2019;Martewicz et al. 2022), and this description largely coincides with the outline of the investigated grains, except for the subangular shapes.The sand horizon yielded a number of subrounded and rounded grains almost without high relief, but instead with dulled surfaces covered with V-shaped marks.These latter marks develop in fluvial transport (Lind e & Mycielska-Dowgiałło 1980), probably suggesting mixed conditions during the deposition of G3 sand, with both fluvial and wave influence (moderate oriented etch pitting).The presence of asymmetric current ripples in G3 sands also supports this conclusion (Fig. 5F).
Coastal lagoon (c.7.7 to 4.8 cal.ka BP).-This period corresponds to the uppermost part of the sand layer with OM interbeds (unit G3) and the clayey gyttja layer (unit G4A) in DAZ-III to -VI (Fig. 3).A high degree of environmental instability with mass occurrence of fragilarioid taxa coupled with the occurrence of some broken parts of the strongly brackish-tolerant Campylodiscus clypeus (Ehrenberg) Ehrenberg ex K€ utzing and C. echeneis Ehrenberg ex K€ utzing at c. 7.7 to 7.3 cal.ka BP (DAZ-III) is recorded.This, along with plant macrofossils typical of the transgressive phase of the Litorina Sea, such as seeds of Najas marina, Zannichellia palustris and Ceratophyllum demersum (Yu et al. 2004), is interpreted as a result of the quickly rising sea-level and onset of lagoon formation owing to development of the coastal barrier in the area NE of the G ß ipka section.Nearshore sands together with subfossils of Cerastoderma edule, Macoma baltica, Mytilus edulis (Meirons 1993), Peringia ulvae and Ecrobia ventrosa (pers.comm.M. Rudz ıte) are seen up to an elevation of Fig. 7. A. Geological cross-section through the G ß ipka basin, based on coring data by Eberhards (2003), some new studied sediment sections (G ß ipka outcrop; cores B1-B3) and interpretation of Lidar and GPR data.B. GPR images of the section crossing the G ß ipkabasin.Numbers show the most prominent GPR reflectors described in the text.We interpret reflector 1 as the surface of the glacial till rising below the Ancylus Lake coastal landform up to the elevation 7 m a.s.l. and reflector 2 as the contact between Ancylus Lake coastal sands and aeolian capping.Internal reflector 3 is interpreted as a layering of a dune.The locations of the geological cross-section (A-B) and GPR image (C-D) are given in Fig. 1C.
This quickly rising RSL period in the G ß ipka record is followed by a gradual increase in brackish and halophilous diatom taxa and the establishment phase of the brackish lagoon at c. 7.3 to 7.1 cal.ka BP (DAZ-IV).The brackish periphytic diatoms represented by Karayevia submarina, Planothidium delicatlum and P. engelbrechtii, including Achnanthes fogedii, recorded as a typical fossil taxon in the Litorina Sea sediments (Snoeijs & Kasperovi cien_ e 1996;Witkowski et al. 2000;Grudzinska et al. 2014), suggest the existence of the G ß ipka brackish lagoon, with the strongest influence of the Litorina Sea between c. 7.1 and 5.8 cal.ka BP (DAZ-V) and the highest salinities c. 6.1 cal.ka BP (Fig. 4).At the same time the strongest marine water intrusions were observed in Lake Lilaste in the southern part of the Gulf of Riga (Grudzinska et al. 2017).
In previous studies (Sepp€ a et al. 2000;Risberg et al. 2005;Grudzinska et al. 2012Grudzinska et al. , 2013)), mass abundance of small fragilarioid taxa, which are considered as pioneer species with the ability to cope successfully with rapid changes in salinity and turbidity (Yu et al. 2004), was interpreted as a sign of the isolation.
Hence, the dominance of Pseudostaurosiropsis geocollegarum, Pseudostaurosiropsis punctiformis (Witkowski, Metzeltin & Lange-Bertalot) Witkowski, Seddon & Pli nski and Nanofrustulum sopotensis, coupled with a noticeable decline in brackish periphytic diatoms, for a millennium-long period between c. 5.8 and 4.8 cal.ka BP (DAZ-VI), following our age-depth modelling, indicates slow and gradual isolation of the lagoon, most probably owing to the gradual lowering of the RSL.
Freshwater coastal lake (c.4.8 to 4.6 cal.ka BP) with possible intermittent brackish water inflows.-This period corresponds to the detritus gyttja layer (unit G4B) and DAZ-VII (Fig. 3).The diatom composition shows a distinct freshening of the environment, characterized by the replacement of small fragilarioid taxa that prefer brackish water by species preferring freshwater conditions.The presence of Nanofrustulum sopotensis and N. krumbeinii in the lake can be explained by the higher electrical conductivity that remained after isolation, although it cannot be excluded that the lake had a small connection to the Gulf of Riga into which brackish water may have occasionally inflowed owing to stronger easterly winds.This, together with the abundance of the macrofossils of various aquatic and coastal plants, suggests the existence of a shallow eutrophic coastal lake (Table S2).The main water supply during this period was most probably from the rivers and streams entering the lake.The water level in the coastal lake was regressive and most probably controlled by falling RSL, finally leading to overgrowing of the lake, which started at c. 4.6 cal.ka BP.
Coastal fen (c.4.6 to 4.2 cal.ka BP).-This period corresponds to the upper fen peat (unit G5), which is barren of diatoms.The plant macrofossil assemblage (Table S2) suggests a coastal peatland which was probably also seasonally flooded, as indicated by the presence of mineral matter in the peat.Gradually increasing OM content upwards in the peat and disappearance of aquatic plants indicate drier conditions in the later phase of fen development.
River floodplain with coastal lakes (since c. 4.2 cal.ka BP).-This period corresponds to the upper sand and silt interval (unit G6) in DAZ-VIII (Fig. 3).In the geological cross-section this period is characterized by various subaqueous and terrestrial sediment layers, interpreted as a floodplain deposit, including small local lakes, swamps and channel deposits (Fig. 7).The lowermost part of the floodplain deposit is represented in the G ß ipka section by laminated fine sand with OM containing freshwater periphytic diatoms and small fragilarioid taxa, suggesting at least periodic existence of isolated freshwater lakes in the floodplain.Subrounded and rounded sand grains almost without high relief but insteadwith dulled surfaces and coveredwith Vshaped marks suggest fluvial transport of grains and therefore strong river input into the lake.

Archaeological chronology and location of Neolithic sites
The earliest evidence of human settlement is from G ß ipka B, located on the landward side of the Litorina Sea dune belt at 9 m a.s.l.(Fig. 8; Loze 2006).The assemblage of Comb Ware pottery indicates occupation c. 6.0 to 5.5 cal.ka BP; organic-tempered pottery has been recovered too, showing that the site was also used during the time interval c. 5.5 to 5.0 cal.ka BP.This is typical for western Latvia: there is very little evidence of coastal settlement before c. 6 cal.ka BP, with abundant remains from c. 6-5 cal.ka BP, and once again a paucity of evidence for coastal occupation after c. 5 cal.ka BP.
All of the other Neolithic sites/dwellings (P urciems A-F and G ß ipka A) are located within the dune belt at elevations of 9-13 m a.s.l., archaeological stratigraphy taking the form of thin occupation layers in the dune sands (Figs 8,9).These occupations have likewise yielded finished and semi-manufactured amber ornaments, and are known in particular for the finds of clay figurines and associated evidence of 'ritual' activities involving ochre ( Sturms 1937;Loze 2005Loze , 2006)).The five new AMS radiocarbon dates obtained for P urciems dwellings C, D and F span the period c. 5.6 to 5.0 cal.BP (Table S4), and the characteristics of the pottery from the other P urciems dwellings and the G ß ipka A site indicate occupation approximately during the same period (B erzin ß s 2008; Lõugas & B erzin ß s 2023).

Discussion
Ancylus Lake development in the western Gulf of Riga An unresolved issue in the postglacial development of the west coast of the Gulf of Riga is the maximum water level and extent of the Ancylus Lake.Our luminescence date from the uppermost part of the high energy nearshore sand complex (unit G1) suggests that at c. 9.3AE1.5 ka the retreating coastline of the Ancylus Lake/Initial Litorina Sea passed the G ß ipka area, as visible from the storm footprints recorded in the lower sand unit (G1).GPR results combinedwith Lidar elevation data show that this sand complex extends up to 2 km landward from the studied G ß ipka section, forming at its landward limit a belt of coastal landforms at an elevation of 11-13 m a.s.l., covered by aeolian sands (Fig. 7).This NW-SE oriented concave belt of coastal formations can be observed for at least 30 km and probably marks the maximum extent of the Ancylus Lake (Fig. 1C), thus contradicting the earlier interpretation by Grinbergs (1957), who considered these to be Litorina Sea formations.The elevation of the Ancylus Lake highest shoreline at G ß ipka (ridge foot at 11-12 m a.s.l.; Fig. 7) is also similar to the elevation of the Ancylus Lake maximum (13 m a.s.l.) mapped in the P€ arnu area (Saarse et al. 2003), eastern Gulf of Riga, with a similar uplift rate, where it is dated to c. 10.2 cal.ka BP (Nirgi et al. 2020).Deposition at similar time frame between c. 10.2 and 10.82 ka is also known from western Latvian coast in the ancient Ventspils lagoon (Kali nska et al. 2022).Later, during the lake regression, a series of shore-parallel sandy coastal ridges (Fig. 1C) were formed seaward of the Ancylus Lake highest shoreline, as seen in the geological profile, partly buried under younger lagoonal and alluvial deposits (Fig. 7).The formation of a series of storm-induced ridges at gradually lower altitude supports the continuous outflow scenario of the Ancylus Lake after its highstand, convincingly documented in the southern Kattegat (Bendixen et al. 2017).

RSL changes in the western Gulf of Riga during the Litorina Sea
The lower fen peat (unit G2) accumulation on top of the coastal sands suggests low RSL (below 3 m a.s.l.) during the Initial Litorina Sea from c. 9.1 cal.ka BP (Fig. 9).Data from the G ß ipka succession do not provide information about the lowest RSL, but it is likely that BOREAS Holocene shore-level changes, western Gulf of Riga RSL dropped close to or even below the present sea-level, as there is evidence of RSL below the present level at that time in the P€ arnu area in the eastern Gulf of Riga owing to a similar uplift history (Fig. 1; Nirgi et al. 2020).Based on AMS radiocarbon-dated terrestrial macrofossils in tandem with the onset of unit G3 accumulation, the subsequent arrival of fresh to slightly brackish transgressional waters at G ß ipka is dated to c. 8.4 cal.ka BP (Table S3).After initial flooding of the area, wave and fluvial-influenced shallow-water estuary sediments with interbeds of redeposited terrestrial organic matter accumulated around 8.1 to 7.7 cal.ka BP with an estimated RSL at G ß ipka c. 4-6 m a.s.l.(Fig. 9).Lithological changes from the accumulation of shallow water sand (G3) towards lagoonal gyttja clay (G4a) accumulation and changes in diatom composition at c. 7.7 cal.ka BP (Fig. 4; Table S3) mark the ongoing rise in RSL and the formation and growth of the barrier in front of the lagoon.The rising RSL period in the G ß ipka record is followed by the establishment phase of the brackish lagoon at c. 7.3 to 7.1 cal.ka BP and probably marks the culmination of the Litorina Sea transgression at the mapped marine limit of the coastal barrier at elevation c. 9 m a.s.l.(Fig. 7; Table S3).Following our RSL reconstruction, during the initial period of transgression, between c. 8.4 and 7.7 cal.ka BP, the water level rise was slower, around 2 m, compared with a subsequent sea-level rise of 4 m from c. 7.7 cal.ka BP until the culmination of the transgression at c. 7.3 cal.ka BP (Fig. 9).

Possible causes of the RSL change and comparison with other RSL reconstructions
The onset of the transgression at G ß ipka concurs with the age of the major RSL rise event detected in different locations around the globe (Li et al. 2012;Amorosi et al. 2013;Wang et al. 2013;Lawrence et al. 2016) and dated in high precision in the Rhine-Meuse Delta (North Sea) to between 8.45 and 8.2 cal.ka BP (Hijma & Cohen 2019).This about 1.9 m RSL rise event (Hijma & Cohen 2019) is related to the rapid melting of the Laurentide Ice Sheet and drainage of the glacial lake Agassiz-Ojibway c. 8.5 to 8.2 cal.ka BP (Lajeunesse & St-Onge 2008).In the BSB, this event probably initiated the onset of a brackish-water phase in Femern Belt (Bennike et al. 2022), a 1-2 m transgression (L1) accompanied by relatively low salinity, in the Blekinge area (Berglund et al. 2005) and the onset of the Litorina Sea transgression in the slowly uplifting V€ astervik-Gamlebyviken (Katrantsiotis et al. 2022), Narva-Luga (Rosentau et al. 2013), P€ arnu (Nirgi et al. 2020) and western Gulf of Riga areas around 8.5 to 8.4 cal.ka BP.Despite somewhat different uplift rates (Fig. 1A), these areas experienced almost simultaneous RSL rise, probably owing to the rapid nature of sea-level rise.
The second transgression (L2) in Blekinge (Berglund et al. 2005) seems to be the most prominent period of RSL rise, centred around 7.8 to 7.6 cal.ka BP and documented in the western Gulf of Riga, as well as in many areas around the BSB (Yu et al. 2007;Rosentau et al. 2013;Sander et al. 2015;Nirgi et al. 2020).It is interesting that the rapid sea-level rise at 7.8 to 7.6 cal.ka BP is not detected in the high-resolution RSL data from the Rhine-Meuse Delta.On the one hand, it may suggest regional RSL phenomena related to GIA processes (collapsing forebulge; Garc ıa-Artola et al. 2018); on the other hand, it may relate to the interpretation of the Rhine-Meuse data.
The culmination of the Litorina Sea transgression at elevation 9 m a.s.l. in the western Gulf of Riga around c. 7.3-7.1 cal.ka BP (Fig. 9) is consistent with the end of the final melting of the Laurentide Ice Sheet c. 7.0 cal.ka BP  S3.Clay figurine from the P urciems site associated with 'ritual' activities involving ochre ( Sturms 1937) marks the habitation period between 6.0 and 5.0 cal.ka BP.
and a significant slow-down in global sea-level rise starting from c. 6.7 cal.ka BP (Lambeck et al. 2014).Thus, after c. 7 cal.ka BP the rate of postglacial uplift in our study area exceeded the rate of global sea-level rise (Fig. 9).

RSL and salinity changes in the lagoon
The percentage of brackish and marine diatoms in the G ß ipka lagoon shows the strongest influence of the Litorina Sea at about 1000 years after the RSL highstand (Fig. 9), between c. 7.1 and 5.8 cal.ka BP, with the highest seawater salinities around 6.1 cal.ka BP (Figs 4,S1).This delay probably reflects the regional salinity changes in BSB related to the ongoing widening of the straits at € Oresund and Great Belt in combination with decreased freshwater supply to the BSB owing to the climate shift (Gustafsson & Westman 2002;Andr en et al. 2011;Ning et al. 2017).The study by Gustafsson & Westman (2002) also shows that freshwater input into the BSB was 15-60% lower than present during the maximal salinity phase around 6 cal.ka BP and showed a salinity decrease during the subsequent millennia, agreeing with our results.

Development of the G ß ipka lagoon and Neolithic huntergatherers
Neolithic hunter-gatherer groups in G ß ipka and P urciems settled in the sandy grounds on top of the Litorina Sea barrier system, preferring the landward side of the system (G ß ipka A and P urciems A-D) and the lagoon shores (G ß ipka B and P urciems E) (Fig. 8).The valley-like depression in historical topographic maps from the beginning of the twentieth century (Karte des westlichen Russlands 1914Russlands -1919) ) may indicate that somewhere between sites P urciems D and F a passage to the lagoon also may have existed (Fig. 8), although owing to later human-induced disturbances (dredging of the channel) during the twentieth century (Nomals 1930), this morphological evidence is inconclusive.
Settlements between seashore and lagoon may be regarded as a typical locus of Neolithic settlement in the sandy coastal belt along the southern and western shores of the Gulf of Riga.Thus, sites from the same period have been discovered in the sandy belts separating the former lagoons and coastal lakes in the region (B erzin ß s et al. 2016, 2022a).The G ß ipka-P urciems sites, occupied c. 6-5 cal.ka BP, are concentrated at the deepest, seaward side of the lagoon with water depths up to 3-4 m (Fig. 8C).Salinity reconstructions show that during the occupation the lagoon was a brackish water body with high productivity and calm waters, and therefore probably well suited for the installation of fishing constructions, as at the Priedaine site (Fig. 1B) at the shore of a concurrent palaeolake at the head of the Gulf of Riga.
The faunal remains from neighbouring Silin ßupe and Priedaine (Fig. 1B; B erzin ß s et al. 2016, 2022b) indicate that sites at the shore of lagoons and coastal lakes could also serve as bases for hunting trips to catch terrestrial mammals, probably because the waters offered good transport connections along the coastal belt and with the hinterland.It is evidently not fortuitous that the G ß ipka-P urciems concentration of sites is located directly opposite the mouth of the Pilsupe river, entering the lagoon from the landward side (Fig. 8C): this waterway would have connected the coastal community with inland areas, important for access to resources and for contact with inland groups.Hence, the P urciems-G ß ipka locality may be regarded not only as a very attractive location for hunting and fishing in the Stone Age, but also as a node of transport and communications.
The G ß ipka B site, by the shore of the lagoon, has produced the earliest evidence of occupation.Following our palaeogeographical reconstructions, the open sea was less than 0.5 km distant, and it may be that conditions on the narrow dune belt were too inhospitable for settlement.Subsequently, occupation did spread to the dune belt itself, and this area also had a particular social and religious significanceas seen from the evidence of activities involving the application of powdered ochre and the fragmentation of clay figurines (Fig. 9; Loze 2005).The P urciems-G ß ipka area evidently lost its particular attractiveness as a settlement locus along with terrestrialisation of the former lagoon and lake from c. 4.6 cal.ka BP.

Conclusions
• The Ancylus Lake highest shoreline was formed in the G ß ipka area at an elevation of c. 11-12 m a.s.l., reaching an elevation ~2 m higher and located ~2 km landward of the highest shoreline of the Litorina Sea.During the regression of the Ancylus Lake/Initial Litorina Sea, a series of shore-parallel coastal landforms, today buried under younger sediments, were formed before c. 9.3AE1.5 ka, when the regressive coastline passed the studyarea at G ß ipka.• During the existence of the lagoon, settlement sites of Neolithic hunter-gatherer groups were established on the shores of the lagoon in the period c. 6.0-5.0 cal.ka BP.Salinity reconstructions based on diatom assemblages show that during the occupation the lagoon was a brackish waterbody with high productivity and calm waters.The G ß ipka-P urciems Stone Age settlement concentration occupied a particularly advantageous location: directly opposite the mouth of the Pilsupe River, which would have served as a pathway connecting the coastal communities with territories further inland, thus facilitating access to resources and contact with inland groups.The P urciems-G ß ipka area evidently lost its particular attractiveness as a settlement locus along with terrestrialisation of the lake from c. 4.6 cal.ka BP.

Fig. 1 .
Fig. 1. A. Overview map of the Baltic Seawith the location of the study area, sites discussed in the text and present-day land uplift isobases (mm a À1 , Vestøl et al. 2019).B. Palaeogeographical reconstruction of the Litorina Sea maximum in the Gulfof Rigawith relative shore-level (RSL) isobases at c. 7.3 cal.ka BP and with the most important Stone Age settlement sites around the gulf (after Muru et al. 2017).C.Airborne Lidar topography of the study area with the location of the studied G ß ipka section (black triangle; 22°37 0 31 00 E; 57°34 0 36 00 N), the location of geological profile A-B and ground penetrating radar (GPR) profile C-D (Fig.7), P urciems (A-F) and G ß ipka (A, B) Stone Age settlement sites (black diamonds, detail map with archaeological sites shown in Fig.8).Locations of the Baltic Ice Lake coastal landforms at elevation 36-32 m a.s.l.(Rosentau et al. 2009) and Litorina Sea coastal landforms at elevation 10-9 m a.s.l.(Eberhards 2006) are also shown on the map.

Fig. 2 .
Fig. 2. Stratigraphy and age-depth model of the sediments of the G ß ipka section with identified units (G1-G6), loss on ignition and luminescence and radiocarbon ages discussed in the text.For the location of the section see Fig. 1C.

Fig. 3 .
Fig. 3. Percentage diagram of selected (≥2% in two or more samples) diatom taxa from the G ß ipka section (arranged according to salinity preference).Diatom assemblage zones (DAZ) are defined by the results of cluster analysis (CONISS).

Fig. 4 .
Fig. 4. Summary diagram of lithology, modelled ages, loss-on-ignition (%), diatom distribution based on salinity tolerance and habitat classification (%), DCA 1 showing changes in diatom composition (standard deviation units), and development stages, interpreted based on the description of the lithology, the results of loss on ignition and the changes in diatom composition.

Fig. 5 .
Fig. 5. A. General sediment architecture of the lower sand of the unit G1 (the white square shows deformations in sand as seen in Fig. 6B).B. Deformations (arrows) in the upper part of the lower sand of unit G1 at the contact with the lower fen peat (unit G2). C. Details of alternation of coarser (gravel/pebble) and finer (silt) interlayers.D. Upwards coarsening successions (arrows) in the lower horizon.E. Details of rip-up silt clasts enriched with micaceous minerals.F. Semi-horizontal sand with OM interbeds of the unit G3, with two sets of asymmetric current ripples in the middle of the sand sheet.G. Semi-horizontal sand of the upper horizon (unit G6).

Fig. 8 .
Fig. 8. Palaeogeographical reconstructions of the development of the G ß ipka area with RSL isobases.A. Litorina Sea lagoon during the RSL highstand at c. 7.3 cal.ka BP.B. Litorina Sea lagoon at c. 5.5 cal.ka BP and during the Neolithic occupation.C. Detail of the lagoon and its possible passage at 5.5 cal.ka BP with the location of the P urciems (A-F) and G ß ipka (A, B) settlement sites and studied G ß ipka section.

Fig. 9 .
Fig. 9. Composite RSL curve for the western Gulf of Riga combining RSL index points from Ruhnu Island (Muru et al. 2018) and RSL data from the G ß ipka area.Environmental changes in the G ß ipka area are presented at the bottom and Baltic development stages (after Andr en et al. 2011) at the top of the graph.For the locations of the sites see Fig. 1.The RSL data in HOLSEA database format are presented in TableS3.Clay figurine from the P urciems site associated with 'ritual' activities involving ochre( Sturms 1937) marks the habitation period between 6.0 and 5.0 cal.ka BP.

Table 1 .
List of radiocarbon dates and dated material from the G ß ipka section.fr = fragment.Menyanthes trifoliate seeds and Carex fruits, Betula and Alnus fruit fr.,

Table 2 .
Summary of radionuclide concentrations, dose rates, water content, equivalent doses and Central Age Model (CAM) ages.s.e.= standard error; n = number of quartz aliquots.
• A new RSL curve for the western Gulf of Riga has been constructed based on the RSL data from the G ß ipka area and from the nearby Ruhnu Island.The reconstruction shows that the last marine transgression in the western Gulf of Riga started at c. 8.4 cal.kaBP and concurred with the 1.9 m RSL rise event between 8.45 and 8.2 cal.kaBP recorded from the North Sea basin.Owing to the rapid nature of the sealevel rise, it caused almost simultaneous transgression in BSB areas with different postglacial land uplift rates.Most of the RSL rise in the western Gulf of Riga occurred later, between c. 8.2 and 7.3 cal.kaBP, followed by RSL fall owing to the significant slow-down in global sea-level rise and ongoing postglacial land uplift.•New diatom, plant macrofossil, textural and structural sediment properties and chronological data from the G ß ipka section show the development of the G ß ipka area in connection with the RSL changes, with the following stages: open coast with regressive Ancylus Lake/Initial Litorina Sea coastline before c. 9.1 cal.ka BP, coastal fen (c.9.1-8.4cal.ka BP) and wetland (c.8.4-8.1 cal.ka BP), estuary (c.8.1-7.7 cal.ka BP), Litorina Sea lagoon (c.7.7-4.8cal.ka BP), freshwater coastal lake (c.4.8-4.6 cal.ka BP), fen (c.4.6-4.2cal.ka BP) and river floodplain (since c. 4.2 cal.ka BP).