Amplitude, frequency and drivers of Caspian Sea lake‐level variations during the Early Pleistocene and their impact on a protected wave‐dominated coastline

The Caspian Sea, the largest isolated lake in the world, witnessed drastic lake‐level variations during the Quaternary. This restricted basin appears very sensitive to lake‐level variations, due to important variations in regional evaporation, precipitation and runoff. The amplitude, frequency and drivers of these lake‐level changes are still poorly documented and understood. Studying geological records of the Caspian Sea might be the key to better comprehend the complexity of these oscillations. The Hajigabul section documents sediment deposited on the northern margin of the Kura Basin, a former embayment of the Caspian Sea. The 2035 m thick, well‐exposed section was previously dated by magneto‐biostratigraphic techniques and provides an excellent record of Early Pleistocene environmental, lake‐level and climate changes. Within this succession, the 1050 m thick Apsheronian regional stage, between ca 2·1 Ma and 0·85 Ma, represents a particular time interval with 20 regressive sequences documented by sedimentary and palaeontological changes. Sequences are regressing from offshore to coastal, lagoonal or terrestrial settings and are bounded by abrupt flooding events. Sediment reveals a low energy, wave‐dominated, reflective beach system. Wave baselines delimiting each facies association appear to be located at shallower bathymetries compared to the open ocean. Water depth estimations of the wave baselines allow reconstruction of a lake‐level curve, recording oscillations of ca 40 m amplitude. Cyclostratigraphic analyses display lake‐level frequency close to 41 kyr, pointing to allogenic forcing, dominated by obliquity cycles and suggesting a direct or indirect link with high‐latitude climates and environments. This study provides a detailed lake‐level curve for the Early Pleistocene Caspian Sea and constitutes a first step towards a better comprehension of the magnitude, occurrence and forcing mechanisms of Caspian Sea lake‐level changes. Facies models developed in this study regarding sedimentary architectures of palaeocoastlines affected by repeated lake‐level fluctuations may form good analogues for other (semi‐)isolated basins worldwide.


INTRODUCTION
Sediment deposited in the Caspian Sea forms excellent geological archives to assess the evolution of coastal sedimentary architecture under repeated lake-level oscillations. The Caspian Sea is the largest isolated lake in the world and forms, together with the Black Sea, the final remnants of the once vast Paratethyan domain (Fig. 1A). The Caspian Sea presents several characteristics that benefit preservation of coastal sedimentary successions, which are commonly limited in the open ocean due to strong coastal erosion during lowstands (Catuneanu et al., 2011). With limited tides, wave and wind activity (Hartgerink, 2005;Medvedev et al., 2016), the Caspian Sea offers enhanced preservation potential for coastal deposits. Furthermore, this basin forms an ideal sediment trap, since tectonic subsidence is extremely high in the south of the basin (Nadirov et al., 1997), as well as in the Kura Basin, forming an embayment of the Caspian Sea, corresponding to the foreland basin of both the Greater and Lesser Caucasus (Jackson et al., 2002a;Allen et al., 2003;Fig. 1B). Strong tectonic uplift of the surrounding mountain ranges, associated with dynamic regional climate (Morton et al., 2003), act as the main drivers of sediment supply into the basin (Abreu & Nummedal, 2007).
The water budget of the Caspian Sea shows strong variations through time, since it is highly dependent on regional changes in evaporation, precipitation and runoff (Degens & Paluska, 1979;Kroonenberg et al., 1997;Arpe et al., 2019). As a result, the Caspian Sea witnessed drastic and frequent lake-level changes up to 150 m amplitude over the last millions of years (Kroonenberg et al., 1997;Popov et al., 2006;Yanina, 2014;Krijgsman et al., 2019). The delicate balance between subsidence, sediment supply and water supply in enclosed basins enhances their sensitivity to lake-level and climate changes (Carroll & Bohacs, 1999). Lakelevel variations in isolated basins are therefore often of higher order of magnitude and frequency compared to the open ocean (Bohacs et al., 2003). These fluctuations are considered the main forcing mechanism acting on the evolution of coastal sedimentary architecture (Plint & Nummedal, 2000;Posamentier & Morris, 2000). Documenting coastal successions may provide valuable insights into drivers acting on coastal architecture, improving current  Popov et al., 2006, andAbreu &Nummedal, 2007). (B) Structural map of the Kura Basin in light grey, pinched between the Greater and Lesser Caucasus in dark grey (adapted from Jackson et al., 2002a,b). The location of the Hajigabul section is marked by a red star.
understanding of the range, timing and climate forcing of lake-level oscillations in the restricted Caspian Sea. Lake-level histories have been proposed previously (Jones & Simmons, 1996;Abreu & Nummedal, 2007), yet lack the resolution in the facies and age models to solve driving mechanisms of the Caspian Sea sedimentary evolution. Thick (ca 1 km) and continuous (ca 1 Myr) sedimentary successions, particularly in the former Kura Basin embayment (Fig. 1B), record frequent relative lake-level variations in the Pleistocene Caspian Sea (van Baak et al., 2013;Forte et al., 2015). Located on the northern margin of the Kura Basin, the Hajigabul section represents a 2 km long, well-dated sedimentary succession that covers much of the Pleistocene . The Early Pleistocene Apsheronian regional stage represents the most dynamic unit along this section, since the Caspian Sea repeatedly engulfed the Kura Basin during this time interval. Apsheronian deposits are marked by frequent alternations between mud-dominated offshore deposits regressing to sand-dominated coastal deposits, and occasionally to mud-dominated pedogenized lagoonal or terrestrial deposits. The remarkable quality of the section, its unique depositional setting and its robust age model are compelling attributes that allowed a detailed reconstruction of Caspian Sea lake-level variations in the Early Pleistocene. The coastal sedimentary architecture of this section is investigated with the intention to provide a detailed record of Caspian Sea lake-level changes. Thanks to the robust age and facies model, the magnitude and frequency of these lake-level variations may be documented throughout the Early Pleistocene. The overall aim is to better understand the drivers of environmental changes in the Caspian Sea, by comparing these oscillations to global climate cycles. This study may lay the foundations to appraise the mechanistic pathways of orbital cycles in restricted basins.

GEOLOGICAL BACKGROUND
Deep marine conditions existed in most of the Caspian Sea until its disconnection from the Black Sea at the end of the Miocene (Popov et al., 2006;Krijgsman et al., 2010). During the Pliocene, the basin was progressively filled in by deltaic sediment of the Productive Series (Reynolds et al., 1998;Hinds et al., 2004), major reservoir rock of the highly productive petroleum system present in the Caspian Sea (Vincent et al., 2010;Abdullayev et al., 2012). The basin underwent a major transgression following a likely Arctic flooding event during the Late Pliocene/Early Pleistocene Akchagylian regional stage (Richards et al., 2018;van Baak et al., 2019). The Caspian Sea became fully isolated from marine conditions around 2Á5 Ma (Richards et al., 2018). The Quaternary succession is consequently subdivided into regional chronostratigraphic units (see Krijgsman et al., 2019, for details) (Fig. 2A). During the Early, Middle and Late Pleistocene Apsheronian, Bakunian, Khazarian and Khvalynian regional stages, the Caspian Sea experienced strong lake-level oscillations and during extreme highstands episodic overflow events occurred towards the Black Sea (Yanina, 2014).
The Caspian Sea is subdivided into a North, Middle and South Basin (Fig. 1A). The South Caspian Basin presents a western extension, known as the Kura Basin, that corresponds to the foreland basin of both the Greater and Lesser Caucasus (Jackson et al., 2002a;Allen et al., 2003;Fig. 1B). Tectonic deformations affected the study area during the Quaternary, creating numerous strike slip faults and folds with mean directions between N130°and N160°. The Hajigabul Anticline, also known as the Hajigabul Mountain, the Kichik Kharami ridge or the Maliy Kharami fold, is located in the northern margin of the Kura Basin. This asymmetrical anticline is approximately 12 km long, 10 km wide and is oriented along a mean direction of ca N135° (Fig. 2B). The northern flank, dipping on average 20°to the north, is thrusted on top of the southern flank, dipping on average 65°to the south. The eroded southern flank of the anticline exposes the 2035 m thick Hajigabul section (N40Á13640556°, E48Á88694444°up to N40Á12366389°, E48Á86972222°) (Fig. 2C, Appendix S1). The anticline exposes clastic sedimentary deposits dated from the Pliocene Productive Series up to the Middle Pleistocene Khazarian regional stage (Bairamov et al., 2008;Lazarev et al., 2019).

MATERIAL AND METHODS
This study focuses on the repeated lake-level variations during the Early Pleistocene Apsheronian regional stage. Lake-level oscillations were reconstructed based on a detailed sedimentary log realized along the Hajigabul section at a decimetre-scale. Field observations focused on variations in lithologies, stacking patterns, colours, sedimentary structures and grain-size distributions, as well as the mineralogy, sorting, rounding, sphericity, texture and maturity of the sediment grains. Sediment colours were determined on fresh rocks using Munsell Soil Color Charts. Diverse sedimentary structures were documented, such as graded bedding, lamination, cross-stratification, pedogenic features, convolute bedding and erosional surface. Distinction was made between well-defined lamina, lamination and bed (Campbell, 1967). Petrographic optical microscopic descriptions were performed on 30 lm thick thin-sections, made perpendicular to the progradation direction of the observed sedimentary structures. Palaeocurrent directions were measured on three-dimensional cross-beds and corrected by unfolding the bedding planes. Measurements were subsequently statistically distributed over 12 intervals of 30°and plotted in a rose diagram with a maximum representability of 50%. Sedimentary facies were compared to well-established classifications (Postma, 1990;Miall, 2006), as well as several detailed referenced studies (see Depositional environments section). Taphonomic and mollusc  Bairamov et al., 2008). Red line indicates the studied section running along two strike-slip faults F1 and F2. (C) Panorama picture (looking north-west) of the Hajigabul section cropping out along the southern flank of the asymmetrical anticline with some anchor points along the stratigraphy.
observations provided additional information such as depositional energy, water depth and salinity, supporting environmental reconstructions. Molluscs were defined as in situ when presenting paired valves, or as reworked when displaying solitary, fragmented and/or abraded valves. Faunal assemblages are mostly composed of endemic species, identified following regional taxonomies (Andrusov, 1923;Kolesnikov, 1950;Wesselingh et al., 2019). Biotas displayed preferred salinity ranging from freshwater (0 to 0Á5 g l À1 ), oligohaline (0Á5 to 5 g l À1 ) to mesohaline (5 to 18 g l À1 ) conditions. Integration of data sets permitted to distinguish several facies fashioned by different depositional processes. Related facies were grouped into facies associations, representing distinct depositional environments.
Rhythmic sedimentary patterns between distal clay-dominated and proximal sand-dominated facies associations were assessed along the Hajigabul section. A facies depth rank curve was created ranging from 'I' for the deepest deposits, up to 'VII' for the shallowest deposits (Table 1). The facies depth rank shaped a relative lakelevel curve, which was used to identify stratigraphic sequences. Cyclostratigraphical analyses were subsequently realized on the facies depth rank curve. Using standard settings, Blackman-Tukey power spectra were generated with a Bartlett window in the Analyseries 2.0.4b program.
Data were normalized to the unit variance and equally-spaced through linear interpolation. In the same program, bandpass filters were applied at 90% confidence levels. The six anchor points of the magnetostratigraphic time frame previously established along the Hajigabul section Fig. 3A) allowed correlation to the geomagnetic polarity timescale (Gradstein et al., 2012). Based on this age model, cyclostratigraphical analyses allowed estimation of the frequency and timing of the rhythmic environmental changes and evaluation of the potential impact of Milankovitch eccentricity, obliquity and precession astronomical cycles (Laskar et al., 2011) and global sea-level variations (de Boer et al., 2014).

Depositional environments
A depositional model was created along the detailed lithological log of the Hajigabul section (Fig. 3B, Appendix S2). Eight facies associations were assembled representing distal shallow marine to proximal coastal, lagoonal or terrestrial depositional environments (Table 2). Facies associations were defined as sedimentary intervals developing between wave baselines, tide levels or storm levels. However, in the absence of major wave or tide activity in isolated basins, these boundaries were adjusted compared to those established in the open ocean. The storm-wave base (SWB) at the offshorelower shoreface transition, the fair-weather wave base (FWB) at the lowerupper shoreface transition and the highest storm level at the backshorecontinental transition can still be preserved. On the other hand, the mean low tide level at the upper shorefaceforeshore transition and the mean high tide level at the foreshorebackshore transition as found in the open ocean are absent in the Caspian Sea, recording a micro-tidal regime. In this peculiar setting, a classification based on the mean low water level and mean high water level was preferred for the most proximal facies associations, solely dependent on seasonal variations in evaporation, precipitation and runoff.

Description
Facies Association I is composed of 1 to 22 m thick dark bluish-grey (GLEY2-4/5B) to dark greenish-grey (GLEY1-4/5G) mudstone. The base of the succession consists of 1 to 10 m thick homogeneous intervals of massive mudstone (Facies Fm). Massive mudstone is overlain by 3 to 15 m thick intervals of mudstone with horizontal planar silt lamination (Facies Fl - Fig. 4A). Lamina sets are 0Á5 to 1Á0 cm thick and comprise laminae 1 to 3 mm thick. Throughout FA.I, sediment is commonly affected by dispersed low intensity bioturbation, made of 1 to 3 cm long and 0Á3 to 0Á5 cm wide vertical burrows. Sediment contains some dispersed in situ mesohaline to oligohaline mollusc species of Dreissena, Apsheronia, Monodacna and Hyrcania (Appendix S3).

Interpretation
Facies Association I documents alternations between very low and low depositional energies. Massive mudstone (Facies Fm) is deposited out of suspension in low energy open waters. Such deposits have previously been identified as hemipelagic deposits (Stow & Tabrez, 1998). Laminated mudstone (Facies Fl) represents intermittent low energy distal bottom-current  Table 1). activities. Similar bottom-currents have been documented as storm-generated currents transporting coastal sediment towards more distal settings (H equette & Hill, 1993). FA.I represents interchanges between very low and low depositional energies, which are increasing towards the top as lamina sets become thicker and more frequent. Together with the vertical burrows, they point towards depositional environments located in the sublittoral zone. FA.I represents relatively shallow offshore settings, with sediment deposited below the SWB.

Description
Facies Association II consists of 2 to 42 m thick dark greenish-grey (GLEY1-4/5G) mudstone. At the base, mudstone contains thicker and coarsergrained horizontal planar lamination of silts to fine sands (Facies Fl). Lamina sets are 1 to 5 cm thick and contain 3 to 10 mm thick laminae. Upward, mudstone displays lenticular bedding made of 1 to 5 cm thick and 3 to 10 cm long asymmetrical lenses composed of silt to very fine sands (Facies Fs - Fig. 4B). Mudstone sporadically alternates with 0Á5 to 3Á0 m thick beds of greyishbrown (2Á5Y-5/2) very fine to medium-grained sandstone. Sandstone is commonly deposited in normal graded beds and comprises three different types of sedimentary structures. Sandstone contains trough cross-stratification filled in with mud, with preserved cross-set thicknesses between 5 cm and 50 cm (Facies St) and presents tabular and sigmoidal cross-stratification, with preserved cross-set thicknesses between 5 cm and 50 cm (Facies Sc). Finally, sandstone comprises horizontal planar lamination (Facies Sl), made of 1 to 10 mm thick lamina sets containing 1 to 3 mm thick laminae. Throughout FA.II, sandstone beds commonly display sediment loading. Sediment documents dispersed low intensity bioturbation made of 1 to 3 cm long and 0Á3 to 0Á5 cm wide vertical burrows. Sediment also contains rare dispersed in situ molluscs, as well as some 1 to 5 mm thick shell strings with transported shells and shell debris, ranging from mesohaline to oligohaline salinities. Observed fauna were Dreissena, Apsheronia, Monodacna, Hyrcania and Clessiniola species (Appendix S3).

Interpretation
Facies Association II records interchanges between low and moderate depositional energies. Laminated mudstone (Facies Fl) represent  fluctuations between low energy, open-water conditions and moderate energy bottom-current activities, referenced in the literature as distal storm-generated currents (H equette & Hill, 1993). Asymmetrical lenticular bedding (Facies Fs) occasionally occurs upward within this facies association, attributed in enclosed environments to intermittent wave action (Reineck & Wunderlich, 1968;de Raaf et al., 1977). Sandstone with trough cross-stratification (Facies St) is known to result from migrations of 2D and 3D sinuous crested dunes (Hamblin, 1961). Sandstone with tabular or sigmoidal cross-stratification (Facies Sc) is interpreted as resulting from migrations of 2D straight crested dunes (Allen, 1963). Sandstone with horizontal planar lamination (facies Sl) is known to be created by upper regime plane-bed flow, under high energy bottom-currents (Cheel, 2006), during redistributing of proximal sediment by storm activity towards more distal depositional settings (Reineck & Singh, 1972). Sediment loading affecting these deposits typically occurs when sandstone is rapidly deposited on top of water-saturated mudstone, leading to fluid expulsion and in situ sediment deformations (Oliveira et al., 2009). Reworked shell strings indicate sporadic energy increases, possibly formed as a result of storm activities (Starek et al., 2010;Hampson et al., 2011). Sediment recorded in FA.II highlights variations between low and moderate depositional energies, with energies increasing towards the top as lamina sets and lenses become thicker and more frequent. Deposits are disturbed more repeatedly by moderate wave activity. FA.II represents lower shoreface settings, with sediment deposited between the SWB and the FWB.

Description
Facies Association III is composed of 4 to 46 m thick greyish-brown (2Á5Y-5/2) fine to coarsegrained sandstone. Sandstone is deposited in 0Á5 to 8Á0 m thick normal graded beds, with sharp horizontal bases. Sandstone displays four lithofacies types. The first lithofacies consists of sandstone with trough cross-stratification with preserved cross-set thicknesses between 5 cm and 50 cm (Facies St). The second lithofacies is made of sandstone with tabular and sigmoidal cross-stratification with preserved cross-set thicknesses between 5 cm and 50 cm (Facies Sc - Fig. 4C). The third lithofacies is characterized by sandstone with horizontal planar lamination with lamina sets between 1 to 10 mm thick and 1 to 3 mm thick laminae (Facies Sl

Interpretation
Facies Association III records moderate to high energy depositional energies. Sandstone with trough cross-stratification (Facies St) is known to form by migrations of 2D and 3D sinuous crested dunes (Hamblin, 1961). Sandstone with tabular and sigmoidal cross-stratification (Facies Sc) is interpreted as the result of migrations of 2D straight crested dunes (Allen, 1963). Sandstone with horizontal planar lamination (Facies Sl) is known to result from lower stage planebed flow, under high energy bottom-currents (Cheel, 2006), during redistributing of proximal sediment in distal settings during storm activity (Reineck & Singh, 1972). Sandstone with swaley cross-stratification (Facies Ss) is thought to be created under oscillatory waves of high velocity, in shallow waters with low aggradation rates (Dumas & Arnott, 2006). Sediment loading affecting the sandstone indicates its rapid deposition on top of water-saturated mudstone, creating fluid expulsion and in situ sediment deformations (Oliveira et al., 2009). Deposition of reworked shell strings indicates relatively high energy depositional environments, attributed in the literature to storm activity (Starek et al., 2010;Hampson et al., 2011). FA.III represents upper shoreface settings, extending from the FWB up to the mean low water level.

Description
Facies Association IV consists in 1 to 23 m thick intervals dominated by light yellowish-brown (2Á5Y-6/3) fine to very coarse-grained sandstone. Sandstone is deposited in 0Á5 to 8Á0 m thick graded beds, which display sharp horizontal contacts with underlying sandstone. Sandstone contains three different types of sedimentary structures. The first one consists of horizontal planar lamination made of lamina sets of 1 to 10 mm thick and comprising 1 to 3 mm thick laminae (Facies Sl). Sandstone also contains wavy bedding made of symmetrical ripples with amplitudes of 3 to 5 cm, wavelengths of 7 to 10 cm and preserved cross-set thicknesses between 5 cm and 50 cm (Facies Sw -Fig. 4D). Furthermore, sandstone displays low-angle cross-lamination with preserved cross-set thicknesses between 5 cm and 50 cm (Facies Sa). Throughout FA.IV, laminae are commonly draped by finer-grained sediment or millimetrescale shell fragments. Sandstone contains abundant continuous 1 to 10 cm thick erosive layers of transported shell debris and shells. Layers display mixed assemblages of mostly oligohaline genera, with Dreissena, Apsheronia, Monodacna, Hyrcania, Didacna, Corbicula, Clessiniola, Theodoxus and Laevicaspia (Appendix S3).

Interpretation
Facies Association IV represents high energy depositional energies. Sandstone with horizontal planar lamination (Facies Sl) is documented to be created by lower stage plane-bed flow regime, under high energy bottom-currents (Cheel, 2006) and found in the attenuated wave zone along the coastline (Martel & Gibling, 1991). Sandstone with wavy bedding (Facies Sw) is recognized as being formed under stationary wave oscillations, affecting the sediment floor (de Raaf et al., 1977). Sandstone with low-angle crosslamination (Facies Sa) is linked to migration of low relief dunes, under swash and backwash currents, during half-stationary wave oscillations impacting the sediment floor (de Raaf et al., 1977). FA.IV represents foreshore settings, extending from the mean low up to the mean high water level, subjected to daily wave swash.

Description
Facies Association V is composed of 1 to 7 m thick intervals dominated by light yellowishbrown (2Á5Y-6/3) medium to very coarse-grained sandstone. Sandstone is deposited in 0Á2 to 1Á0 m thick graded beds, with sharp horizontal lower contacts. Sandstone contains two lithofacies types. The first lithofacies is dominated by sandstone with horizontal planar lamination made of lamina sets 1 to 10 mm thick and comprising 1 to 3 mm thick laminae (Facies Sl). The second lithofacies consists of conglomeratic sandstone (Facies Gm -Fig. 4E). Sandstone is matrix-supported and contains common millimetre-scale lithic pebbles, glauconite grains and mud clasts, as well as abundant reworked oligohaline molluscs and rare reworked Neogene planktonic foraminifera (Fig. 5A and B). These two lithofacies display sandstone with low compositional maturity, because it is mostly composed of quartz and feldspar grains, and highlights high textural maturity, as it comprises well-sorted and rounded grains presenting moderate-sphericity. These deposits occasionally contain shell pavements, displaying partial or total cementation of the grains in a post-diagenetic carbonate matrix ( Fig. 5C and D). Shell pavements contain mixed assemblages with oligohaline species of Dreissena, Apsheronia, Monodacna, Hyrcania, Didacna, Corbicula, Clessiniola, Theodoxus, Laevicaspia and Turricaspia (Appendix S3).

Interpretation
Facies Association V represents very high depositional energies. Sandstone with horizontal planar lamination (Facies Sl) is considered to represent lower stage plane-bed flow regime, under high energy bottom-currents (Cheel, 2006) and is found in the attenuated wave zone along coastlines (Martel & Gibling, 1991). Shelly conglomeratic sandstone (Facies Gm) is recognized as a lag deposit, formed under very high energy wave reworking in the swash zone (Bourgeois & Leithold, 1984). It may form shell pavements when undergoing post diagenetic dissolution, identified as the main mechanism of formation for beach cement (Kneale & Viles, 2000). Such beach rocks typically represent very high energy sediment reworking and winnowing (Kidwell & Aigner, 1985). The lateral continuity of indurated sandstone layers, parallel to the palaeocoastline, points towards the absence of mouth bars or tidal inlets, and the presence of a continuous low-relief coastline. FA.V represents backshore settings, extending from the mean high water level up to the highest storm level, where reworking and winnowing are affecting sediment deposition.

Description
Facies Association VI is composed of 1 to 22 m thick alternations between olive brown (5Y-4/3) sandstone and olive-grey (5Y-4/2) or reddishbrown (5YR-4/3) mudstone. Massive mudstone (Facies Cm - Fig. 4F) is present in 1 to 5 m thick intervals. It randomly alternates with 1 to 15 m thick mudstone with horizontal planar silt lamination, made of lamina sets 0Á5 to 3Á0 cm thick and comprising 1 to 3 mm thick laminae (Facies Cl). Sediment also contains 0Á5 to 5Á0 m thick light yellowish-brown (2Á5Y-6/3), very fine to medium-grained sandstone beds. Sandstone presents horizontal planar silt lamination with lamina sets 1 to 10 mm thick and 1 to 3 mm thick laminae (Facies Cs). Massive and laminated mudstones are occasionally affected by centimetre-scale depositional faults. Throughout FA.VI, sediment contains abundant millimetre to centimetre-scale terrestrial organic material fragments. It also displays common vertical roots 1 to 5 cm long and 0Á1 to 0Á3 cm wide and rare vertical burrows 1 to 3 cm long and 0Á3 to 0Á5 cm wide. Sediment is occasionally affected by reddish oxidation. Furthermore, it contains sporadic 1 to 5 cm thick white chalk layers, enriched in charophytes and fish bones. Mudstone commonly contains dispersed in situ freshwater to oligohaline molluscs, including Dreissena, Apsheronia, Monodacna, Corbicula, Laevicaspia, Esperiana, Lymnaea, Valvata, Theodoxus, Planorbis, Unio and Bythinia (Appendix S3). The same genera are only rarely present in sandstone, where they are reworked in centimetre-thick layers.

Interpretation
Facies Association VI records alternations between very low and moderate depositional energies in organic-rich environments. Organicrich massive mudstone (Facies Cm) is recognized to be deposited from suspension in very low energy standing waters (Link & Osborne, 1978;. Mudstone with horizontal planar silt lamination (Facies Cl) is typically documented as overbank deposition during washover events (Schwartz, 1982). Sandstone with horizontal planar silt lamination (Facies Cs) is known to be deposited during intermittent increases of energy disrupting very low energy standing waters and is linked to occasional overbank deposition during washover storm events of higher energy (Schwartz, 1982). FA.VI records organic-rich and well-oxygenated environments, interpreted as lagoonal settings.

Description
Facies Association VII consists of 1 to 10 m thick reddish-brown (5YR-4/3) or greenish-grey (GLEY1-6/10GY) mudstone. The succession contains 1 to 5 m thick massive pedogenized mudstone intervals (Facies Pm - Fig. 4G). They are irregularly intercalated by 1 to 10 m thick pedogenized mudstone with horizontal planar lamination of silts to very fine sands (Facies Pl). Lamina sets are 0Á5 to 1Á0 cm thick and comprise 1 to 3 mm thick laminae. Mudstone contains rare in situ freshwater molluscs, such as Corbicula, Melanoides, Unio and Planorbis (Appendix S3). Mudstone contains various scales of slickenslides, reddish and greenish colour mottling, and development of blocky peds. It commonly contains millimetre-scale organic material fragments, vertical roots 1 to 5 mm long and 0Á1 to 0Á3 cm wide, as well as vertical burrows 1 to 3 cm long and 0Á3 to 0Á5 cm wide.

Interpretation
Facies Association VII represents alternations between very low and low depositional energies.
Pedogenized mudstone (Facies Pm) is suggested to be deposited from suspension in very low energy standing waters (Link & Osborne, 1978). Pedogenized laminated mudstone with horizontal planar lamination (Facies Pl) is typically documented as overbank deposition during washover events (Schwartz, 1982). The presence of mottling and slickenslides represents pedogenesis (Retallack, 2001;Kraus & Hasiotis, 2006). FA.VII displays organic-rich environments, subject to subaerial exposure and pedogenesis, and is interpreted as terrestrial floodplain settings.

Description
Facies Association VIII consists of a 0Á1 to 0Á5 m thick shelly interval, expressed in three different manners according to its stratigraphic relation with neighbouring facies associations. Facies Association VIIIa occurs on top of backshore deposits (FA.Vsee Backshore facies association section). It consists of erosive shelly intervals made of light yellowish-brown (2Á5Y-6/3), medium to very coarse-grained sandstone (Facies TS1 -Fig. 4H). Shelly sandy deposits form condensed layers that contain abundant millimetre-scale, well-sorted, abraded shell fragments. Faunal assemblages record mixed salinities from freshwater to oligohaline. These layers include millimetre-scale glauconite grains and centimetre-scale lithic pebbles, moderately sorted, well-rounded and with moderate-sphericity.

Interpretation
Facies Association VIII represents transitional intervals from backshore, lagoonal or terrestrial settings to offshore settings (FA.Isee Offshore facies association section). Facies Association VIIIa highlights an energetic and rapid deepening on top of backshore deposits, generating sediment erosion, reworking and winnowing. Similar sandy shelly lag deposits have previously been interpreted as erosional ravinement surfaces (Kidwell & Aigner, 1985;Scarponi et al., 2013). Facies Association VIIIb displays a less energetic and more gradual deepening on top of lagoonal deposits, leading to increasing salinity associated with a reduction in the amount of shells. Even if no ravinement surface has been observed in the field, these gradual changes in faunal assemblages are interpreted as transgressive events, also referred to as progressive and transgressive overlap (Grabau, 1906). Facies Association VIIIc represents an energetic and rapid deepening on top of terrestrial deposits, producing sediment erosion and reworking. Similar muddy shelly lag deposits have previously been interpreted as erosional transgressive surfaces (Driese & Foreman, 1991). In brief, FA.VIII represents sudden transgressive events of offshore settings on top of backshore, lagoonal or terrestrial settings.

Stratigraphic architecture
Within the Apsheronian interval, facies associations are repeatedly deposited following the same stratigraphic order, forming sedimentary successions recurrently regressing from distal to proximal depositional environments. Sedimentary successions start with ca 9 m of bluish-grey to greenish-grey mudstone deposited in offshore settings (FA.I). Upward, they record ca 15 m of alternating greenish-grey mudstone and greyishbrown very fine to medium-grained sandstone representing lower shoreface settings (FA.II). Fine to coarse-grained greyish-brown sandstone with small-scale sedimentary structures becomes increasingly dominant over mudstone illustrating ca 14 m of upper shoreface settings (FA.III). Sedimentary successions continue with ca 7 m of fine to very coarse-grained greyish-brown sandstone with larger-scale sedimentary structures representing foreshore settings (FA.IV). Upward, ca 2 m of medium to very coarsegrained yellowish-brown sandstone with lowangle sedimentary structures is found illustrating backshore settings (FA.V). Palaeocurrent directions measured on cross-sets in foreshore and backshore depositional settings and corrected for the bedding planes highlight mean unidirectional dip directions between 180°and 210°, with 65% of all palaeocurrent dip directions recorded within a range from 150°to 240° (  Fig. 6). Sedimentary successions are sometimes overlain by ca 6 m of brownish lagoonal mudstone (FA.VI), occasionally overlain by ca 5 m of reddish or greenish terrestrial mudstone (FA.VII). Each sedimentary succession corresponds to a regressive sequence, prograding southward from offshore up to coastal, lagoonal or continental environments. Regressive sequences are overlain by conglomeratic shell lags on top of backshore deposits (FA.VIIIa), transitional shelly mudstone on top of lagoon deposits (FA.VIIIb) or muddy shell lags on top of terrestrial deposits (FA.VIIIc) that represent flooding events. Repeated transgressive events on top of regressive sequences highlight recurrent relative lake-level variations whose amplitudes are discussed below.
Relative lake-level variations display repetitive patterns throughout the section. In total, 20 regressive cycles, separated by 20 highstands, were found approximately every 60 AE 50 m along the facies depth rank (Fig. 3C). As the amplitude of regressive sequences is rather constant, this suggests no major environmental change throughout the entire sedimentary record, such as the substantial deepening recorded in the Plio-Pleistocene Akchagylian regional stage . The very good preservation of sequences, aggrading under unidirectional and non-erosive depositional processes, attests to their deposition during normal regressions and in sufficient accommodation space. Regression events are overlain by major erosive flooding surfaces, formed under rapid lake-level increases of high amplitude. The amplitude of these sudden transgressive events probably prevented the formation of transgressive system tracts on top of the flooding surfaces. The regularity of these repeated lake-level fluctuations, shaping 20 sequences formed under normal regression and sediment aggradation, demonstrate the lack of significant subsidence rise, loss of sediment supply or lake-level drop (Posamentier et al., 1998). Instead, it points towards high sediment supply continuously balanced by high subsidence rates, overruled by repeated transgressive events along the coastline, whose recurrence is driven by external forcing mechanisms.

Timing of environmental changes
Frequency of relative lake-level variations may be assessed along the Hajigabul section thanks to its rather well-constrained age model. By correlating the six existing magnetostratigraphic age tie points with the geomagnetic polarity timescale (Gradstein et al., 2012), the exposed sedimentary record, deposited between 550 m and 1630 m, is dated from ca 2Á1 to 0Á85 Ma . Based on this time frame, sedimentation rates were extrapolated and the interval of interest, recording repeated lake-level fluctuations between 400 m and 1615 m, is dated from ca 2Á3 to 1Á2 Ma (Fig. 7A). Considering this entire interval, the 20 lake-level highstands display an average frequency of ca 53Á3 kyr. However, the age model is not as equally robust along this interval. The lower part of the section (150 to 510 m) and the upper part of the section (1100 to 1700 m) only document two age tie points each. With the presence of four age tie points, the stratigraphically continuous middle part of the section (510 to 1100 m) displays a very robust time frame and records a sedimentation rate of ca 1Á54 m kyr À1 . Based on this very well-constrained interval, the nine lake-level highstands occur with an average frequency of ca 41Á1 kyr, in the range of the 41 kyr obliquity cycles.
A Blackman-Tukey transformation of the lakelevel curve realized along this interval to the time domain reveals spectral power at frequency 0Á0195069 and bandwidth 0Á00827091 with 90% confidence. Spectral analyses result in a large number of peaks with a dominant one between 36 kyr and 89 kyr. The applied 36 to 89 kyr bandpass filter depicts the lake-level cycle well, apart from two cases (Fig. 7B). In the first case, the filter shows peaks that are not present in the lake-level curve. One extra cycle occurs in the filter between highstand 1/2, and one after highstand 20. This may be due to the poorly constrained age model in the lowermost and uppermost parts of the section. In the second case, the filter groups cycles 7/8, and cycles 11/12 into one. The frequency of these highstands is too high to be picked up by the filter. The age model is not able to resolve whether these highstands relate to different forcing or whether these were potentially deposited under lower sedimentation rates. Except for these two cases, the representability of lake-level variations by the bandpass filters is very high.
When plotted against one another, lake-level highstands match relatively well to obliquity maxima within the freedom of the age model (Fig. 7C). The additional cycles created by the filters in time and cycle domains between highstands 1/2, and above highstand 20, are similarly present in the obliquity curve. The two cycles banded together by the filters between highstands 7/8 and 11/12 are ungrouped due to the lower sedimentation rates. Two additional cycles are created between highstands 13/14 and 18/19 due to the higher sedimentation rates. Even if some dissimilarities are documented, the correlation between lake-level variations, bandpass filters and obliquity cycles is rather high. The relative lake-level curve also appears in the range of global sea-level changes (Fig. 7D). Consequently, Early Pleistocene lake-level in the Kura Basin appears to have been sensitive to 41 kyr obliquity cycles, driving ice-volume and global sea-level.

Sedimentary architecture of the Kura Basin palaeocoastline
The architecture of sediment deposited along the northern margin of the Kura Basin during the Early Pleistocene Apsheronian regional stage records repeated Caspian Sea incursions within this embayment (Fig. 1A). The facies model of the coastal sedimentary successions deposited in this bay displays several features typically generated under wave activity. Sediment regularly documents planar-lamination formed by storm-generated currents, transporting proximal sediment towards more distal depositional settings (H equette & Hill, 1993). Deposits commonly display wavy bedding, low-angle crosslamination and swaley cross-stratification, generated under oscillatory wave action (de Raaf et al., 1977;Dumas & Arnott, 2006;Ainsworth et al., 2011). Sediment frequently contains shelly beach rock, formed by wave reworking and winnowing in the swash zone (Bourgeois & Leithold, 1984;Davis & Hayes, 1984;Kidwell & Aigner, 1985). Upward, washover deposits occur, formed under wave reworking (Davis & Hayes, 1984).
Some characteristic features of wave-dominated coastlines nevertheless have not been found. Wave-dominated systems commonly contain sedimentary structures generated under longshore currents (Davis & Hayes, 1984). Longshore sediment drifting typically leads to the formation of barrier bars and spits (Bhattacharya & Giosan, 2003), commonly observed along the restricted Mediterranean, Black Sea and Caspian Sea (van der Meulen & Salman, 1996;Lahijani et al., 2009Lahijani et al., , 2019Matishov et al., 2013;Vespremeanu-Stroe et al., 2017). These structures usually include varied current directions (Nielsen & Johannessen, 2009). Apsheronian sediment is still deposited under quite unidirectional palaeocurrent directions, flowing to the southwest (Fig. 6), perpendicular to the Kura Basin palaeocoastline oriented north-west/south-east (Fig. 1B). Moreover, sediment does not record any aeolian influx, typically found along barrier bars and spits (Davidson-Arnott & van Heyningen, 2003). Such structures document limited lateral continuity (Davidson-Arnott & van Heyningen, 2003). In the Kura Basin, however, sediment bodies may be traced over at least 4 km laterally on aerial photographs of the Hajigabul anticline (Fig. 8). Although most features  (Laskar et al., 2011). (D) Global sea-level curve (de Boer et al., 2014). In order to discuss the simultaneous evolution of these four curves, the time frame of the section  is plotted by dotted black lines against the geomagnetic polarity timescale (Gradstein et al., 2012). The yellow lines highlight the intervals between magnetostratigraphic sample levels and the orange striped interval displays the hiatus present at the top of the section. Fluvial influence along the northern coastline of the Kura Basin appears to have been rather limited. Sandstone layers do not display major fluvial features such as strong basal erosive or channelization typical of fluvial channels. Sandstone bodies are laterally continuous and are conformably deposited on top of underlying sediment. Additionally, sediment lacks fluvial evidence such as high-angle trough crossstratification (Ainsworth et al., 2011), hyperpycnal flows or enrichment in terrestrial organic material (Jorissen et al., 2018). Sediment input originates from small northern sources with limited impact on the coastal morphology. Deposits yield evidence for a northern sediment source with palaeocurrents flowing southward, offshore from the northern margin of the Kura Basin.
Due to the isolation of the Kura Basin from the open ocean, its coastline did not experience any tide activity. Sediment does not contain characteristic tidal deposits, such as swash sedimentary structures, intertidal clay-drape couplets (de Boer et al., 1988), intensive bioturbation in foreshore sediment, or extensive development of current ripples below the fair-weather wave base (FWB) (Frey & Dashtgard, 2011). These features are typically recorded within foreshore sediment, deposited between the mean low tide and mean high tide in open marine environments. However, isolated basins usually display extremely low tidal ranges, and foreshore deposits represent in this case a transitional interval between mud-dominated shoreface and sand-dominated backshore settings, lacking indications for tide activity.
Wind action along the coastline also appears to have been limited, with a lack of major storm winds. The documented succession lacks any evidence of strong wind-induced features such as aeolian beach deposits (Otvos, 2000) or sediment resuspension (Booth et al., 2000). The northern coastline of the Kura Basin appears to have been protected by the Greater Caucasus from the strong northern winds recorded nowadays around the Caspian Sea (Kosarev, 2005). In brief, the northern margin of the Early Pleistocene Kura Basin shaped a straight wave-dominated coastline, rather dissimilar to those in the open ocean due to the lack of tide activity, or any major fluvial or aeolian interference.
Along this protected coastline, wave energy was less significant than in typical wave-dominated systems. With average wave heights of Fig. 8. Aerial photograph of the Hajigabul section from Google Earth â with mapping of the regional stages boundaries, the two strike-slip faults F1 and F2 and some of the most characteristic layers, emphasizing their lateral continuity over distances of at least 4 km along the anticline. 0Á25 m during calm seas and 1 to 4 m during storm events (Terziev et al., 1992;Amirinia et al., 2017), this coastline is in fact classified as a low energy wave-dominated beach (Jackson et al., 2002b). Following recognized beach classifications (Short, 2006), the sedimentary architecture of the Kura Basin palaeocoastline is ranked as a low energy wave-dominated reflective beach. The low reflectivity of such coastlines enhances offshore sediment transport and increases sediment accretion. With a sedimentation rate around 1Á5 m kyr À1 , the Kura Basin coastline recorded rapid aggradation. Reflective beaches are generally composed of relatively coarse sandstone due to waves breaking closer to the shoreline (Short, 2006). Similarly, the studied sedimentary succession is composed of relatively coarse grains. Moreover, such coastlines are characterized by very narrow or even completely absent surf zones (Aagaard et al., 2013). The Apsheronian coastline probably formed such a narrow and steep beach, since it was situated along the northern margin of the foreland basin of the Greater Caucasus. However, if proximal backshore settings were relatively narrow and steep, distal foreshore, shoreface and offshore settings were rather wide and shallow. The steepness of the distal palaeocoastline was undoubtedly scaled down compared to the present-day southern Iranian coast of the Caspian Sea (Kosarev, 2005) or compared to beaches along the open ocean (Short, 2006). The succession misses the deposition of major mass-transport deposits, typical for deep and steep shelves (Postma, 1995), steeply dipping sedimentary structures or erosional chutes (Orton & Reading, 1993). The Kura Basin shaped in fact a shallow, protected and restricted embayment of the Caspian Sea. The basin was filled in by sediment originating from the northern Greater Caucasus and from the Kura Delta confined further to the west (Mamedov, 1997;Hoogendoorn et al., 2005;Abreu & Nummedal, 2007). Sediment was probably transported, washed and deposited under weak river and wave action along the northern margin of the Kura Basin. While this study provides a good analogue for a protected wave-dominated coast, additional research is still needed to fully assess sedimentary architecture within a basin devoid of major tide and wave activity.

Amplitude of Early Pleistocene Caspian Sea lake-level variations
In enclosed basins, with limited tide and wave activity, facies association boundaries are delimited between wave baselines and/or water levels that are expected to be shallower than in the open ocean. Waves in the present-day offshore Caspian Sea typically display wavelengths around 50 m during storm events (Hartgerink, 2005). The storm-wave base (SWB) (offshorelower shoreface transition), situated at a depth of half the wavelength, would consequently be located at maximum 25 m water depth. Following the same approach, with average wavelengths estimated around 32 m (Hartgerink, 2005), the FWB (lowerupper shoreface transition) would be situated at maximum 16 m water depth. Within oceanic settings the lowest SWB lies between 20 m and 200 m water depth, and the FWB between 5 m and 20 m water depth (Nichols, 2009). Furthermore, in the Caspian Sea, the mean low and high water levels are mostly driven by seasonal variations in evaporation, precipitation and runoff, acting on the water budget of this endoreic basin (Degens & Paluska, 1979;Kroonenberg et al., 1997). Such seasonal fluctuations generate lake-level variations up to AE0Á2 m amplitude in the present-day Caspian Sea (Chen et al., 2017). The mean low water level (upper shorefaceforeshore transition) would consequently be situated around À0Á2 m and the mean high water level (foreshorebackshore transition) around +0Á2 m. Such levels are very dissimilar compared to oceanic systems, where tidal ranges display variations between AE1 m and AE7 m amplitude (Nichols, 2009). Finally, waves are between 1 to 4 m high during storm events in the present-day Caspian Sea (Terziev et al., 1992;Amirinia et al., 2017). The highest storm level (backshorecontinental transition) would therefore be situated at maximum +4 m. In oceanic settings, waves may reach up to 12 m height during storm events (Bouws et al., 1996).
Consequently, offshore sediment in the Kura Basin would be found between approximately À40 m and À25 m, lower shoreface sediment between À25 m and À15 m, upper shoreface sediment between À15 m and À0Á2 m, foreshore sediment between À0Á2 m and +0Á2 m, and backshore sediment above +0Á2 m. Lagoonal and continental sediment would be deposited around 0 m, behind the beach barriers (Fig. 9). However, elongated bays like the Kura Basin usually attenuate the energy of wave influx (Caliskan & Valle-Levinson, 2008). The facies boundaries may have been shallower in the Kura Basin, shielded from major storm, wave and wind activity by the eastern Caucasus extension. Nevertheless, without any additional information like characteristic ichnofossils or robust subsidence rate studies, it remains extremely challenging to provide absolute water depths for these facies boundaries. For now, relative lake-level variations are approximated to oscillate between À40 m and 0 m relative water depth. Water levels in the Early Pleistocene Kura Basin may be estimated based on the assumption that surface levels were not tectonically elevated or depressed compared to the present-day topography. During lowstands, the western and middle part of the Kura Basin recorded terrestrial environments (Agust ı et al., 2009;Krijgsman et al., 2019), the eastern part documented offshore environments (van Baak et al., 2013) and the northern margin of the basin recorded coastal, lagoonal and terrestrial environments (Fig. 10A). At À20 m, the water level would have been 8 m above today's absolute water level. This 8 m water increase compared to the present-day would have been enough to flood the easternmost part of the Kura Basin, keep the westernmost part emerged and only have limited impact on its northern margin, where littoral environments could have developed. During highstands, the western part of the Kura Basin still recorded terrestrial environments  and the eastern part still documented offshore environments (van Baak et al., 2013). However, repeated alternations between terrestrial and nearshore environments were recorded in the middle part of the basin (Nevesskaja, 2007;Krijgsman et al., 2019) and along its northern margin (Fig. 10A). At +20 m, the water level would have been 48 m above today's absolute water level. Following such a water level increase compared to present-day, highstands would not have reached the far western end of the basin still recording terrestrial settings, while the middle part of the basin and its northern margin would have been episodically flooded, and the eastern part of the basin would have been constantly flooded. Early Pleistocene lake-level variations in the Kura Basin consequently oscillated with an amplitude of ca 40 m, between water levels À20 m and +20 m (Fig. 10B). Fig. 9. Summary of facies associations, depositional processes and faunal assemblages, illustration of lateral relationships in between the different depositional environments and the relative water depth of facies association boundaries.

Drivers of Early Pleistocene Caspian Sea lake-level changes
Lake-level fluctuations, as recorded laterally over at least 4 km in the Kura Basin, are similarly observed along nearby sections further to the north-east along the Caspian Sea coast (Richardson, 2012). In view of the fact that lakelevel oscillations are documented consistently in the basin, the observed sedimentary sequences are bounded by third-order transgressive events as characterized in the Embry model (Embry, 1993). As such, the studied lake-level variations do not represent depositional sequences as mentioned in the Vail & Mitchum (1977) model, nor genetic stratigraphic sequences as defined by the Galloway (1989) model. Therefore, the local lake-level curve in the present study may be used to analyse the evolution of the Caspian Sea lake-level during the Early Pleistocene.
Caspian Sea lake-level variations seem to have been strongly driven by Early Pleistocene climate conditions. The robust lake-level curve reconstructed in this study includes 20 highstands with an amplitude of ca 40 m and a frequency of ca 41 kyr. Highstands are in range with the 41 kyr obliquity cycle and correlate in cycle and time domain with obliquity maxima (Fig. 7C). Obliquity minima are known to have strengthened Northern Hemisphere glaciations on the Eurasian continent until the mid-Pleistocene transition (Ruddiman et al., 1989;Maslin et al., 1998). Obliquity minima may have driven increases in ice-sheets during colder and drier climates in the Caspian catchment areas, which may have caused lake-level lowstands by decreased runoff. Lagoonal and terrestrial settings established around the Kura Basin embayment. On the other hand, obliquity maxima may have generated warmer and wetter climates that drove lake-level highstands by increased northern river runoff. Following lake-level transgressions, the Kura Basin recorded more distal settings mostly driven by regional climate changes.
After the mid-Pleistocene transition, it appears that a strong climatic switch occurred in the Caspian Sea system. Extensive research has shown a correlation between lake-level highstand and glacial conditions (Kroonenberg et al., 1997;Yanina, 2014). Extreme lake-level lowstands likely resulted from water capture in ice-sheets and permafrost during dry periods, whereas lakelevel highstands may have resulted from melting of the ice-sheets under humid conditions. Transgression events reached over 150 m amplitude since the Middle Pleistocene, compared to the maximum 40 m amplitude lake-level variations during the Early Pleistocene. The lake-level regime thus experienced a major change during the mid-Pleistocene transition. Overall, Caspian Sea lake-level variations seem to be strongly driven by allogenic forcing mechanisms. Although the present study offers a rare view inside the Early Pleistocene Caspian Sea driven by obliquity climate cycles, further research is still required to fully understand the precise impact and mechanistic pathways of orbital forcing and of (inter)glacial periods on lake-level oscillations.

CONCLUSIONS
During the Early Pleistocene, the Kura Basin formed an embayment of the isolated Caspian Sea. Thick sedimentary successions were deposited on its northern margin, related to high subsidence and aggradation rates. Water and sediment supply originated from northern mountainous sources and from reworked sediment of the Kura Delta confined in the western part of the basin.
The Early Pleistocene Apsheronian regional stage highlights a 1050 m thick interval along the Hajigabul section with eight facies associations representing distal shallow marine to proximal coastal, lagoonal or terrestrial depositional environments. Sediment is deposited in 20 regressive sequences about 60 m thick, bounded by sudden flooding surfaces. The section highlights repeated lake-level changes affecting the Kura Basin embayment.
This geological archive enables reconstruction of a detailed coastal facies model for an isolated sedimentary basin subject to water-level variations. The protected palaeocoastline of the Kura Basin documented limited tide, wind and wave interference, due to reduced storm events compared to the open ocean. This elongated embayment was shielded by the eastern extension of the Greater Caucasus. As a result, the palaeocoastline shaped a low energy, wave-dominated, reflective beach profile. Under these circumstances, wave baselines delimiting the different facies associations (storm-wave base, fair-weather wave base, mean low water level, mean high water level and highest storm level) are assumed to be shallower than in the open ocean.
Frequent lake-level variations occurring along the palaeocoastline of the Kura Basin represent lake-level for the entire Caspian Sea. These are estimated with amplitude of ca 40 m water depth and a frequency of ca 41 kyr. Lake-level variations in the Early Pleistocene Caspian Sea were possibly driven by the 41 kyr astronomical cycles. This study delivers new insights in terms of coastal morphology along the Caspian Sea, as well as estimations regarding the amplitude, frequency and drivers of relative lake-level changes.