Palaeoenvironmental evolution of Lake Gacko (Southern Bosnia and Herzegovina): Impact of the Middle Miocene Climatic Optimum on the Dinaride Lake System.

In the Early to Middle Miocene, a series of lakes, collectively termed the Dinaride Lake System (DLS), spread out across the north-western part of the Dinaride-Anatolian continental block. Its deposits, preserved in numerous intra-montane basins, allow a glimpse into the palaeoenvironmental, palaeobiogeographic and geodynamic evolution of the region. Lake Gacko, situated in southern Bosnia and Herzegovina, is one of the constituent lakes of the DLS, and its deposits are excellently exposed in the Gračanica open-cast coal-mine. A detailed study of the sedimentary succession that addresses facies, sediment petrography, geophysical properties, and fossil mollusc palaeoecology reveals repetitive changes in lake level. These are interpreted to reflect changes in the regional water budget. First-order chronologic constraints arise from the integration of radio-isotopic and palaeomagnetic data. (40)Ar/(39)Ar measurements on feldspar crystals from a tephra bed in the upper part of the sedimentary succession indicate a 15.31 ± 0.16 Ma age for this level. The reversed magnetic polarity signal that characterises the larger part of the investigated section correlates to chron C5Br of the Astronomically Tuned Neogene Timescale. Guided by these chronologic data and a detailed cyclostratigraphic analysis, the observed variations in lake-level, evident as two ~ 40-m and seven ~ 10-m scale transgression-regression cycles, are tuned to ~ 400-kyr and ~ 100-kyr eccentricity cycles. From the tuning, it can be inferred that the sediments in the Gacko Basin accumulated between ~ 15.8 and ~ 15.2 Ma. The economically valuable lignite accumulations in the lower part of the succession are interpreted to indicate the development of swamp forests in conjunction with lake-level falls corresponding to ~ 100-kyr eccentricity minima. Pedogenesis, rhizoliths and palustrine carbonate breccias in the upper part of the section reveal long-term aridity coinciding with a ~ 400-kyr eccentricity minimum. Eccentricity maxima are interpreted to trigger lake-level high-stands. These are accompanied by eutrophication events caused by enhanced denudation of the surrounding basement and increased detrital input into the basin. The presented age model proves that Lake Gacko arose during the Middle Miocene Climatic Optimum and that the optimum climatic conditions triggered the formation of this long-lived lake.

Introduction

The Dinaride–Anatolian land formed a major barrier between Paratethys and the proto-Mediterranean Sea during the Early and Middle Miocene (Fig. 1A). It harboured a widespread and long-lived array of lakes, collectively known as the Dinaride Lake System (DLS). Deposits of the individual lakes have been preserved in numerous intra-montane basins (Fig. 1B) throughout the Dinaric Alps from Slovenia to Serbia and Montenegro (Harzhauser and Mandic, 2008; Krstić et al., 2003; Rasser et al., 2008). They form an extraordinary archive for the palaeoenvironmental, palaeobiogeographic and geodynamic evolution of the region.

Fig. 1

The studied area. (A) Palaeogeographic position between the Paratethys and the Mediterranean Tethys during the Langhian (after Rögl, 1998). (B) Geographic position showing distribution of other major DLS basins (after Mandic et al., 2008). (C) Geological setting of the Gacko Basin with indicated position of two studied localities (after Mojićević and Laušević, 1965; Mirković et al., 1974). The numbering and colour code of Miocene lithostratigraphic units coincide with illustration D. (D) Detailed geological map of the western part of the Gacko Basin (after Thomas and Frankland, 2004) with indicated position of the Gračanica section. The relative position of represented lithostratigraphic units is indicated by the column on the left. For colour codes see Fig. 2.

Lake Gacko was the southernmost constituent of the DLS (Fig. 1B). Excellent outcrop conditions in the Gračanica open-cast coal-mine (Thomas and Frankland, 2004) provide a superb opportunity to study the development of this lake. Clear vertical changes in facies as well as faunal assemblages characterise the exposed sedimentary succession. The archives of this highly sensitive lake registered subtle environmental variations as alternations of coal, marl, lacustrine- and palustrine limestone. The presence of one thick and several thinner ash layers testifies to concurrent volcanic activity in Southeastern Europe and facilitates the construction of a time-frame through 40Ar/39Ar dating. The simple synclinal structure of the basin and the readily available basic geologic, lithostratigraphic and palaeontologic data provided by the Gračanica coal-mine Exploration and Production Authority and a number of past scientific investigations (Brusina, 1897; Katzer, 1921; Milojević, 1966, 1976; Mirković, 1980; Neumayr, 1880; Thomas and Frankland, 2004) provides a suitable framework for this high-resolution palaeoenvironmental study.

Kochansky-Devidé and Slišković (1978), subdivide the evolution of the DLS in two subsequent stages, each characterised by different molluscs. Based on its phylogenetically advanced mollusc assemblage, Lake Gacko pertains to the younger stage. Its lifetime should therefore coincide with the late stage evolution of Lake Sinj in Croatia, another constituent lake of the DLS. For the latter, a high-resolution description of the depositional history, a reconstruction of the vegetation and climate dynamics based on pollen records, and a solid chronostratigraphic framework were recently published (De Leeuw et al., 2010; Jiménez-Moreno et al., 2008; Mandic et al., 2009). Newly obtained results from Lake Gacko are thus readily comparable with data from Lake Sinj and provide a better insight into the dynamic faunal as well as palaeoenvironmental evolution of the DLS as a whole.

Geological setting

The Gacko Basin (Fig. 1C) is a typical intra-montane karst polje (large flat-floored valley) situated at approximately 930 m above sea-level. The 40 km², strongly elongated tectonic depression is oriented in a NW-SE direction, parallel to the strike of the Dinarides (Milojević, 1966, 1976; Mirković, 1980; Mirković et al., 1974; Mojićević and Laušević, 1965, 1973; Muftić, 1964). Its margins are defined by normal faults (Thomas and Frankland, 2004). The karstic basement consists of predominantly shallow-water deposits that accumulated on the Dinaric/Adriatic carbonate platform in the Mesozoic. In the Late Cretaceous, the platform disintegrated and flysch deposition started in the north-eastern tectonic unit (Fig. 1C). Intensive thrusting subsequently affected the south-eastern tectonic unit and, during the Palaeocene and Eocene, flysch accumulated there as well. Late Eocene clastic molasse-type sediments, termed the Promina Formation, are the last marine deposits in the region. The Miocene lacustrine sediments form a simple but slightly asymmetric syncline structure, with its NW-SE oriented axis slightly offset to the southwest. Post-depositional tectonics shifted the north-western part of the syncline southward and divided it from the rest of the basin by S–N striking anticlinal folds (Thomas and Frankland, 2004) (Fig. 1C and D).

The Miocene sedimentary infill of the Gacko Basin comprises about 360 m exclusively lacustrine sediments (Milojević, 1976; Mirković, 1980; unpublished exploration and drilling data by Gacko Mine and Thermal Power Plant company; Figs. 1D and 2). The architecture of the sedimentary succession can be interpreted as a single transgression–regression cycle of the lake. The initial flooding of the basin resulted in an about 20-m-thick basal conglomerate with a matrix of sand and clay (Fig. 1D — Unit 1). Three marl-to-coal sequences follow on top of this conglomerate, each about 50 m thick in the central part of the basin (Fig. 1D — Units 2 to 4). The coals bear taxodiacean stems and trunks that are indicative of swamp conditions at the lake margin. Slightly upward in the section, the predominance of marls indicates drowning of the swamp environment and installation of a perennial lake (Fig. 1D — Unit 5). A thick volcanic tuff marks the top of this marl unit. Above the tuff, a second coal interval points to returning swamp and mire conditions due to a drop in water level. This is the terminal phase of Lake Gacko (Fig. 1D — Unit 6). The volcanic tuff and second coal interval are restricted to the central part of the basin and are outside of the scope of the present study.

Material and methods

The stratigraphic succession of Lake Gacko was logged and sampled in the abandoned SW part of open-pit B of the Gračanica mine in the NW part of the basin (Thomas and Frankland, 2004) (Figs. 1C, D and 2). It is located between GPS WGS1984 datum points N 43° 10′ 37.5″, E 18° 29′ 17.7″ and N 43° 10′ 33.0″, E 18° 29′ 7.4″. The section has a stratigraphic thickness of 76.8 m. It starts with a 2.8-m-thick coal seam which is the lowermost lignite layer of the basin infill. This coal seam, buried under mine debris during the field study, overlies the clayey marls of the “footwall deposits of coal II” unit of Milojević (1966). The top of the section corresponds to the level of the Gacko Polje plane (Fig. 2).

Fig. 2

The results of deposition and facies analysis and lake-level-change. Positions of lithological units defined for the studied section and their correlation with the general Gacko Basin lithostratigraphic division are shown (Milojević, 1976; Mirković, 1980; Thomas and Frankland, 2004 and unpublished exploration and drilling data by Gacko Mine and Thermal Power Plant company). The colour code of the lithological log reflects the natural sediment colour; for detailed lithological description see text. Position of discussed thin-sections and mollusc samples is indicated, and gamma ray and magnetic susceptibility logs are reproduced. Results of mollusc palaeoecological analysis and positions of inferred transgression–regression cycles and possible intensities of relative lake-level rises are indicated as well.

Detailed sedimentological logging was carried out in the field and allowed first-hand analysis of the represented facies types and their architecture. Thin sections of selected carbonate facies types were prepared for detailed microfacies analysis in the laboratory. The fossil record was studied both in the field and in the lab.

Mollusc palaeontology

Well-preserved molluscs are rare in the section and are concentrated in its lower and middle part. Aragonite leaching in the upper, carbonate-dominated part is demonstrated by rare gastropod moulds. Only three samples (mollusc samples 1, 3 and 4, Fig. 2) were suitable for quantification because, for other samples, shell decalcification or a strongly cemented rock matrix prevented washing and sieving. The quantified samples were sieved at fraction > 250 μm. One additional mollusc sample (2), taken from the abandoned Vrbica open-cast mine, was investigated (GPS WGS1984 Position: N43° 08′ 42.7″ E18° 33′ 35.8″; Figs. 1C and 2). This sample is situated in the adjoining hanging wall of the main coal seam, above a single 10-cm-thick coal layer intercalated in dark brown clayey and silty marls bearing Mytilopsis frici. The presence of the latter marker fossil in combination with the lithology and lithostratigraphic position of the site allows straightforward correlation to the lower part of the Mytilopsis-marl unit of the Gračanica section (Fig. 2). The abandoned Vrbica opencast mine is located beside the road, halfway between Avtovac and Gacko (= Metohija) and therefore corresponds to the classical fossil mollusc site of Brusina (1897) and Neumayr (1880).

Geophysical logging and spectral analysis

To better understand the fine-scale vertical variations in the detrital rock component, gamma ray and magnetic susceptibility measurements were taken with a hand-held “Compact Gamma Surveyor” scintillation gamma radiometer and “SM-20” magnetic susceptibility metre with a sensitivity of 10− 6 SI units (GF Instruments, Brno, Czech Republic). The distances between the geophysical point measurements were 10 cm (GR) and 5 cm (MS), respectively (Fig. 2).

Power spectra are calculated for the magnetic susceptibility and natural gamma radioactivity in the depth domain using the PAST programme (Hammer et al., 2001) and the Lomb periodogram algorithm for unevenly sampled data. The data were detrended prior to spectral analysis by subtraction of the linear regression line to improve resolution at low frequencies. Gaussian band-pass filtering was conducted using the AnalySeries time series analysis tool developed by Didier Paillard (Paillard et al., 1996).

Magnetostratigraphy

In order to construct a magnetostratigraphy for the Gacko Basin, 40 sites with two standard palaeomagnetic cores each were drilled (Fig. 2). The resolution achieved in the section stretching from the coal beds at the base to the prominent volcanic ash at the top was around 1 site per 2 m stratigraphically. Samples were collected with a gasoline-powered hand-held drill. The orientation of all samples was measured using a magnetic compass. Bedding planes were similarly determined at regular intervals. Subsequently, both bedding planes and sample orientations were corrected for the local magnetic declination, adding 4° east.

Palaeomagnetic analysis was carried out at the Palaeomagnetic Laboratory of Utrecht University. The cores were sliced into multiple specimens of which one per site was subsequently subjected to Alternating Field (AF) demagnetization. After each demagnetization step, the natural remanent magnetization (NRM) of the samples was measured on a 2G Enterprises DC Squid cryogenic magnetometer (noise-level 3 × 10−12 Am2). AF demagnetization was performed by a laboratory-built automated measuring device applying 5–20 mT increments up to 100 mT by means of an AF-coil interfaced with the magnetometer. The characteristic remanent magnetization (ChRM) was identified by inspection of decay-curves and vector end-point diagrammes (Zijderveld, 1967). ChRM directions were calculated by principal component analysis (Kirschvink, 1980).

Geochronology

An approximately 1-m-thick greenish and clayey volcanic ash in the uppermost part of the section was sampled twice for 40Ar/39Ar dating (Fig. 2). Both samples were taken from the same part of the ash-layer but with a lateral distance of ~ 10 m. The samples were processed in the Department of Isotope Geochemistry (VU Amsterdam). The bulk samples were crushed, disintegrated in a calgon solution, washed and sieved over a set of sieves between 63 and 250 μm. The residue was subjected to standard heavy liquid as well as magnetic mineral separation techniques. The fraction of grains larger than 150 μm of both samples contained abundant feldspar crystals. These were handpicked and leached with a 1:5 HF solution in an ultrasonic bath during 5 min. The mineral separates were then loaded in a 10 mm ID quartz vial together with Fish Canyon Tuff (FC-2), Drachenfels one (f250–500) and Drachenfels two (f > 500) sanidine. The vial was irradiated at the Oregon State University TRIGA reactor in the cadmium-shielded CLICIT facility for 10 h.

After return to the VU Amsterdam laboratory, ten splits of both samples were loaded into a copper sample-tray together with Drachenfels as well as Fish-Canyon sanidine standards. The tray was pre-heated under vacuum using a heating stage to remove undesirable atmospheric argon. Thereafter, samples were placed in the UHV sample chamber and degassed overnight. The samples were then fused using a Synrad CO2 laser in combination with a Raylase scan-head as a beam delivery and beam diffuser system. After purification, the resulting gas was measured with a Mass Analyzer Products LTD 215-50 noble gas mass spectrometer. Beam intensities were measured in peak-jumping mode in 0.5 mass intervals over the mass range 40–35.5 on a Balzers 217 secondary electron multiplier. System blanks were measured every three to four steps. Mass discrimination was monitored by frequent analysis of aliquots of air. The irradiation parameter J for each unknown was determined by interpolation using a linear fit between the individually measured standards.

All 40Ar/39Ar ages were calculated with the in-house developed ArArCalc software (Koppers, 2002), applying the decay constants of Steiger and Jäger (1977). The age for Fish Canyon Tuff sanidine flux monitor used in age calculations is 28.201 ± 0.046 Ma (Kuiper et al., 2008). The age for the Drachenfels sanidine flux monitor is 25.42 ± 0.05 Ma (Kuiper et al., in prep). Correction factors for neutron interference reactions are 2.64 ± 0.017 × 10− 4 for (36Ar/37Ar)Ca, 6.73 ± 0.037 × 10− 4 for (39Ar/37Ar)Ca, 1.211 ± 0.003 × 10− 2 for (38Ar/39Ar)K and 8.6 ± 0.7 × 10− 4 for (40Ar/39Ar)K. Errors are quoted at the 1σ level.

Depositional facies analysis and lake-level change

Lithology

A detailed description of the lithological record is essential in order to document the character of fine facies shifts observed in the section. It forms the backbone of the palaeoenvironmental interpretation, guides the reconstruction of lake-level changes, and provides a solid framework for the cyclostratigraphic model put forward.

In the investigated section, 6 lithostratigraphic units designated with the letters A to F are recognised. The lower three units are dominated by coal, the upper three by carbonate deposits. Coal is represented by a soft-brown coal (lignite) with the following average quality values at the studied site (after Thomas and Frankland, 2004): total moisture 37.4% a.r., ash 15.1% a.r., total sulphur 1.22% a.r. and net caloric value 9.623 kJ/kg.

The present lithostratigraphic classification follows, with two exceptions, the classification established by Milojević (1966, 1976) (Fig. 2). Unit B includes, however, besides “footwall coal I” also two adjoining units, namely Fossarulus- and Melanopsis-marl. The latter two are not well developed at studied locality and cannot be clearly differentiated from the “footwall coal I”. This coincides with the observation by Milojević (1966) that the two marl units, attaining up to 50 m in the basin centre, reduce toward the basin margin to metre scale. Furthermore, “hanging wall marls” in the studied section comprise two distinct lithological units termed Units E and F. The top of the “hanging wall marls”, together with overlying “tephra” and “coal units”, could not be investigated in the studied section because they were present exclusively in the central part of the basin (Figs. 1D and 2).

Unit A (“footwall coal II”)

Unit A (9.8 m) is three-folded, comprising a 3-m-thick coal seam in the lower part, a 2.8-m-thick marl in the middle part and a 4-m-thick coal seam on top. It overlies the whitish-grey clayey marl. The lower coal seam is intercalated by marl lenses and inter-layers and incorporates weakly coalified remains of tree stumps and branches. The marl interval bears common mollusc remains and is portioned by a 10 cm coal seam into lower, laminated, wood fragments-bearing, beige part and an upper, coal laminae-bearing, greyish part.

Unit B (“footwall coal I”, Fossarulus- and Melanopsis-marl)

Unit B (13 m) consists of alternating coal seams and dark greyish limestone. The coal layers show rough cyclic architecture by a thickening upward trend up to the 3-m-thick coal seam in the middle part of the unit, followed by a thinning upward series. The lower boundary is transitional, characterised by coal-limestone alternation at a 10 cm scale. In the lowermost 2.7 m the limestone is dark greenish-greyish with brownish dots and abundant impurities (coalfield wood fragments) and common mollusc remains. White carbonate crust accumulations can also occur in transitional zones between coal and limestone. The bed transitions in this part of the unit are still gradual. At the next coal seam, the lower coal boundaries become sharp and planed, resulting in lithologically very clear limestone differentiation (Fig. 3.1). The upper coal boundaries are sharp as well, but commonly undulated due to original surface relief made by taxodiacean tree trunks and branches deposited on top of coal seams and covered subsequently by limy deposits. Five laterally traceable grey limestone beds are present. The coal seams are characterised by larger and smaller limestone lenses and interlayers of restricted lateral extension. They include common, horizontally oriented, weakly coalfield remains of tree stumps, trunks and branches at the metre-scale.

Fig. 3

Field views of the most typical depositional facies. (1) Alternation of coal and grey limestone around the boundary of Units B and C. (2) Section interval including Units D2 and E1 with grey limestone, thin, coal seam-related dreissenid shell accumulations, tufa-lithoclast breccia and rhizolith limestone. (3) View of the brecciation surface (arrow) at base of Unit D2. (4–5) Grey limestone of Unit D2 with accumulated shells of Mytilopsis frici. (6) Tufa-lithoclast breccia at base of Unit E. (7) View of the succession of the lower part of Unit F with indicated position of 0.5-m-thick, greenish tephra layer. (8) The very top of the section (Unit F) with organic-rich clay and marl-bearing, thin coal seam intercalations.

Unit C (“main coal”)

This unit (< 10.2 m) comprises the topmost coal seam interval. It includes the best-quality coal because of the lowest relative volume of intercalated limestone beds. That volume, however, as already observed by Thomas and Frankland (2004), increases westwards. Also, the thickness increases in that direction. The unit starts with a 1.4-m-thick homogeneous coal seam with sharp and plane lower boundary (Fig. 3.1). The overlying, 3.4-m-thick interval shows intensive alternation of coal and limestone beds. At the studied site, coal predominates in the interval (Fig. 3.1). To the west, however, limestone replaces it progressively, becoming there thicker than the coal. On top, a homogeneous, up to 3.4 m thick, high-quality coal seam occurs with only sporadic limestone lenses. Finally, the intensive alternation of coal and limestone beds (maximum thickness 1.6 m) marks the very top of the Unit C. Eastwards, both latter subunits decrease in thickness by up to 50%.

Unit D (“Mytilopsis-marl”)

Unit D (14.3 m) is characterised by thick intervals of grey, fossil-poor limestone. Still including coal seam intercalations, it represents a transitional unit between the coal- and the limestone-dominated section parts. It comprises two subunits bounded by an erosional hiatus. The lower subunit (8.2 m) comprises, in the lower part, dark grey limestone (equal to limestones of previous units) passing upward into dark grey limestone. Therein, intercalations of a few up to 10-cm-thick coal seams and dark greyish clayey marl beds occur. Overlying this is a 1.2-m-thick dark greyish clay bed, topped by beige limestone bearing coalified wood fragments; it passes upwards into light greyish, intensively fractured limestone. The upper subunit overlies an erosional surface (Figs. 2, 3.2 and 3.3). The low-grade relief is planed by up to 30-cm-thick mollusc-bearing sandstone filling up the 30-cm-deep fissures in the underlying limestone; above it is rich in dark, coalified organic matter. This is followed by an about 4-m-thick prominent, light grey, initially banded then homogeneous limestone interval, which lacks fossils. Its upper half bears dark nodules introduced after a thin (1 cm) coal intercalation persisting laterally for more than 100 m. A dark limestone (2.2 m) follows upward. Its lower part is composed of sand-infilled bioturbations, the middle part of an enhanced sandy component, and the upper part of thin coal seams alternating with bivalve coquinas (Figs. 3.2, 3.4 and 3.5). The unit ends with beige limestone comprising still thin coal and coquina inter-layers.

Unit E (“hanging wall marls” — breccia unit)

The lower boundary of this unit, introducing an abrupt change of depositional style in the succession, is marked by an erosional hiatus (Fig. 2). Unit E is 16.2 m thick, but its thickness enlarges significantly in westward direction, where an intercalation of up to 2-m-thick cross-bedded carbonate sands is present. In the studied section the unit starts with beige breccia (0.8 m) composed of intraclasts and palustrine limestone fragments overlying the erosional surface (Figs. 3.2 and 3.6). The next 0.9 m interval first shows a dark brown laminated sandy limestone bed, then a characean oogoniae-bearing, beige sandy limestone bed, and finally a dark greenish carbonate breccia layer. This is followed above by a 3.5-m-thick, dark grey to dark brown, microtube-bearing limestone. Gastropods, still rare in the lower part, become common in the upper part, accumulated to coquinas. This interval bears grey laminated limestone in its lower part and breccia intercalations in its upper part. The characean oogoniae that are already present in the previous interval become a common constituent of limestones within the following 1-m-thick, light greyish bed. The thickest breccia horizon follows; it is beige, attaining 3.5 m. A single 20-cm-thick dark limestone bed is intercalated in its upper part. The upper 7.2 m of the unit is represented through alteration of characean limestone with maximally 0.9-m-thick breccia intervals. Rare gastropod remains can be present in the characean limestones. The top of the unit is marked by the section's uppermost breccia bed (Fig. 2).

Unit F (“hanging wall marls” — tephra-clay unit)

Unit F (11.5 m) is lithologically very variable (Fig. 2). The base is marked by a ripple-bedded characean limestone followed by a series of light-coloured, beige and grey characean and ostracod massive limestones. In the central part of the unit the limestone becomes intercalated by up to 0.9-m-thick, greenish sandy and clayey tephra layers (Fig. 3.7). Deposition of organic-rich brownish clays is restricted to the uppermost 3.9 m of the section. Initially, clay alternates with limestone. In the middle part of the interval, brownish clay bears gastropod fragments. Within the final 1.1 m the gastropod-bearing clay becomes dark brown, finally bearing thin coal inter-layers at the very top of the section (Fig. 3.8).

Carbonate depositional facies

Four main limestone facies found in the studied section were investigated additionally in thin sections to more precisely interpret the depositional facies. Special emphasis was given to the microtubuli-bearing limestone and the intraclast breccia in respect to the character of included components. For sample positions see Fig. 2.

Greyish limestone

This limestone type dominates the lower part of the section. It is present in the Units B, C and D and is associated there with coal deposits. These limestones are always in direct stratigraphic contact with latter, representing the coal seam intercalations or alternating with thicker coal layers (Fig. 2). Such packstones commonly bear molluscs and coalified plant remains (Figs. 4.1 and 4.2). Gacko Basin corresponds, by its geological setting, to the synchronous Sinj Basin. The latter likely accommodates DLS palaeo-lake deposits and comprises, within its topmost infill part, similar coal-limestone alternations (Mandic et al., 2009). There, the limestone intervals are regarded to represent the lake-level rises, whereas the coal phases are correlated with the lake-level falls. The same depositional model applies for the Gacko Basin.

Fig. 4

Photomicrographs (optical microscope) of selected lacustrine carbonate facies. (1–2) Organic matter- and mollusc-bearing grey limestone, sample 1, Unit B. (3–4) Rhizolith limestone, sample 2, Unit E1. (5–6). Tufa-lithoclast from tufa-lithoclast breccia, sample 3, Unit E1. (7–8) Characean limestone, sample 4, Unit F.

Rhizolith limestone

This limestone bears no fossil remains except for common, up to 3-mm-long, black, calcitic microtubuli. It is dark greyish and dark brownish, restricted to the lower part of Unit E. There, it dominates one 3.7-m-thick interval (Fig. 2). The thin sections (Figs. 4.3 and 4.4) show packstone with size-sorted, non-oriented microtubuli floating in a micritic matrix. Tubuli attain maximally 1 mm in radius and are commonly fragmented. Their walls are thin and composed of dark, globular-micrite. Subordinated are microtubuli with white, calcititic, non-micritised walls. These walls are multi-tubular in cross-section, supporting the classification into calcified root cortices. Following examples by Froede (2002) and Košir (2004), the dark microtubuli are furthermore interpreted as fossilised root casts. Consequently, all microtubuli can be termed rhizoliths. Fragmentation and size-sorting of the rhizoliths point to their transport from the place of origin, usually positioned at a pedogenic, vegetated lake fringe. Rhizolith production is bound to the type of pedogenesis that is characteristic for arid and semi-arid climate conditions (Flügel, 2004).

Tufa-lithoclast breccia

The breccia intercalations are restricted to Unit E, occurring from its base to the top, with the thickest bed (2.5 m) in the middle part of the unit (Fig. 2). Breccia beds usually have sharp lower boundaries that can be clearly erosional; the upper boundaries are transitional. Layers are dominantly beige, but thin inter-beds of dark brownish or greenish matrix are present as well. They are associated with rhizolith and characean limestone. The lithoclast composition is dominated by up to 10-cm-large calcareous tufa, although coalified wood fragments and lacustrine limestone clasts are also present. The latter show, in thin section, fringe cement crusts of plant stems (Figs. 4.5 and 4.6). The stems usually attain about 4 mm in radius and belonged to aquatic macrophytes. The limestone can therefore be classified, after Ford and Pedley (1996), as paludal tufa. Its origin was bound to ephemeral ponds developed along the lake fringe. The breccia is therefore related to erosional processes of marginal environments that could develop during the lake-level low stands as marked by an erosional hiatus or be transported into the lake during storm events.

Characean limestone

This limestone facies is restricted to Units E and F, alternating there with tufa-lithoclast breccia, rhizolith limestone and organic-rich clay (Fig. 2). Its colour is dominantly beige, sometimes, light or dark grey. Except for a single ripple-bedded interlayer, it is massive. Sporadically occurring massive micritic limestone with absent or rare characean oogonia or the ostracod-bearing limestone of the Unit F is closely associated with this facies type. In thin section, whole and fragmented oogonia are visible, floating in the micritic matrix (Figs. 4.7 and 4.8). Based on density variation, it represents wackestone to packstone. The presence of characean limestones points to the presence of open lacustrine conditions (Platt and Wright, 1991).

Fossil molluscs are very good indicators of palaeoenvironmental changes in lacustrine settings (Cohen, 2003). They are highly sensitive to hydro-climate fluctuations and record abiotic factors such as water chemistry, depth, and turbulence. Hence, the evaluation of their taxonomy, species-richness, and proportional abundances helps reconstruct changes in depositional environment and is an essential element in assessing lake-level variations.

The investigated samples comprise 18 species-level taxa of exclusively lacustrine molluscs (Figs. 5 and 6). Their taxonomic identifications are based on Brusina (1874, 1897, 1902), Kochansky-Devidé and Slišković (1978), Neumayr (1869, 1880) and Olujić (1999). All but Mytilopsis frici (Figs. 3.4 and 3.5) are represented in the quantified samples. M. frici is an endemic dreissenid bivalve that belongs to the M. drvarensis clade. It is rounded and flattened, attaining a size of 30 to 50 mm. The flattened morphology is partly secondarily induced by sediment compaction and therefore the specimens could originally have had somewhat more inflated morphologies. The specimens show a dorso-ventral sinus, which distinguishes them from their predecessor — the very similar M. drvarensis. Their identification as M. frici is in agreement with previous results of Kochansky-Devidé and Slišković (1978).

Distribution

On top of the lower coal seam of Unit A, shell accumulations with the pulmonate snails Gyraulus and Lymnaea are present. In the overlying beige limestone, rissooidean gastropods have been additionally observed. A quantified sample from that layer (mollusc sample 1 in Fig. 2) comprises 8 species (Fig. 6). The pulmonates are still frequent therein, represented by Gyraulus pulici (23.6%) and Radix hyaloleuca (17.9%), but rissooideans are slightly more commonly represented with the stenothyrid Bania prototypica (19.8%) and the hydrobiid Fossarulus with two species, F. fuchsi (13.2%) and F. buzolici (13.2%) (Fig. 5).

Fig. 6

Diagramme showing mollusc distribution in four quantified samples. The major supra-generic taxonomic categories mentioned in the text are indicated.

Fig. 5

Mollusc taxa identified in quantified samples. (1–4) Melanopsis lyrata Neumayr, Mollusc sample (MS) 3. (5) Radix hyaloleuca (Brusina), MS 1. (6) Fossarulus bulici Brusina, MS 1. (7) Fossarulus buzolici Brusina, MS 1. (8) Fossarulus fuchsi Brusina, MS 2. (9–10) Prososthenia neutra Brusina, MS 3. (11) Gyraulus pulici (Brusina), MS 1. (12) Planorbarius sp., MS 2. (13) Pseudamnicola stosiciana Brusina, MS 1. (14) Bania prototypica (Brusina), MS 1. (15) Gyraulus pulici (Brusina), MS 3. (16) Ferrissia illyrica (Neumayr), MS 2. (17) Orygoceras dentaliforme Brusina, MS 2.

The pulmonate snail accumulations are still present in Unit B. Its lowermost 4 m bear abundant mollusc remains concentrated in transitional zones between coal and limestone beds. Present therein are Gyraulus pulici, Ferrissia illyrica and Lymnaea klaici (Fig. 5). Additionally present but rare are the rissooidean snails Fossarulus cf. buzolici and Stalioa cf. parvula. In the upper part of the unit, molluscs are absent except for the 1.3-m-thick, grey limestone interval, which bears rare Radix cf. hyaloleuca. Upward, in Unit C, no molluscs have been observed (Fig. 2).

In the lower part of Unit D, molluscs are absent in the studied section. In contrast, in the Vrbica open-cast mine (Fig. 1C) molluscs are very common at this particular stratigraphic position (mollusc sample 2 in Fig. 2). The quantified sample from there comprised the highest recorded species richness of 12 species (Fig. 6). Its composition is dominated by the rissooidean Prososthenia neutra (30.4%) and Fossarulus bulici (23.5%) followed by the previously completely absent Melanopsis lyrata (15.7%). A small content of pulmonate snails known already from the underlying coal-bearing succession is additionally present. The latter decrease even further in the middle part of Unit D, which is characterised by the maximum dominance of Melanopsis lyrata (32.2%) (mollusc sample 3, Figs. 2, 5 and 6). Taxonomic richness remains high with 9 identified species. Prososthenia neutra (25.6%) and Fossarulus fuchsi (15.6%) are additionally frequent therein. In the topmost part of the unit, monospecific shell accumulations of Mytilopsis frici are strikingly associated with thin coal seams.

Molluscs are still present throughout Unit E, but in its upper part commonly leached, preserved only as moulds. Mollusc diversity decreases in the lower, organic-matter-richer interval. Hence, mollusc sample 4 (Fig. 2) from the base of the brownish rhizolith limestone bears only 3 species; the dominant species here are Fossarulus bulici (67.9%) followed by Radix hyaloleuca (17.9%) and Gyraulus pulici (14.3%) (Figs. 5 and 6). Most of the mollusc remains present up to the top of the section belong to Fossarulus. With the reappearance of organic-matter-rich sediments at the top of Unit F, large lymnaeid snails known from the base of the Unit B occur again, together with F. bulici and F. buzolici.

Palaeoecology

The recorded succession of mollusc assemblages supports the interpretation of the studied interval for a large-scale transgression–regression series. Whereas its lower and upper parts are characterised by ephemeral pond assemblages dominated by pulmonate snails such as Radix and Gyraulus, its middle part shows a completely different, highly diversified fauna dominated by Melanopsis and Prososthenia.

The dominance of Melanopsis and Prososthenia is reminiscent of the Sinj Basin, where both genera are abundant especially in the upper part of the basinal infill (Mandic et al., 2009; Olujić, 1999). The very diverse and fairly well-preserved mollusc assemblage suggests a deposition in a long-lived palaeo-lake environment. The Melanopsis–Prososthenia assemblage indicates agitated shallow-water, open-lake conditions. The assemblage with Radix and Gyraulus is, in contrast, characteristic for wetland and marsh faunas inhabiting small temporal lakes and ponds. It marks the marginal position of the site during low-stand conditions.

The additional species present in the lower part of the succession, such as Unio or Pisidium, on the other hand fit well into the picture of a rather marginal and temporal facies. Nevertheless, the extremely low species richness in Units E and F points to the introduction of strong environmental stress, coinciding with the occurrence of lowermost paludal tufa breccia. The long-lived lake fauna probably became extinct in the lake and did not recover until the end of the studied section. In contrast, Fossarulus, common throughout the section, is highly abundant particularly in Units E and F. This demonstrates its presence in a pioneer guild of that palaeo-lake.

The dominance of Fossarulus and Melanopsis in two separate lithostratigraphic units of the lower part of the basinal infill was already recognised by Milojević (1966) and since then was used for its ecostratigraphic division (Fig. 2). The present study refines this division in recognising the reoccurrence of Fossarulus and the decease of Melanopsis domination in the upper part of the succession, starting with the Unit E. As discussed above, that division must reflect the change of general palaeoenvironmental conditions from a dominantly low lake-level and arid climate during the Fossarulus-assemblage phase in the lower and upper parts of the succession to a dominantly high lake-level humid climate during the Melanopsis-assemblage phase in the middle part of the section (Fig. 2).

Geophysical logging

Natural gamma radioactivity as well as magnetic susceptibility is correlated to the amount of detrital material such as clay minerals (Emery and Myers, 1996) in the rock. It allows the evaluation of changes in detrital sediment input into the basin as a function of land denudation. However, volcanic matter (Hardardottir et al., 2001) and radioactive elements potentially bound to organic matter (Jiménez-Moreno et al., 2009; Saracevic et al., 2009) may influence the signal. Usually, however, these disturbances of the detrital signal can easily be recognised due to their point-wise distribution and distinctly higher values than those available from the terrigenous clay component.

Magnetic susceptibility

The change in magnetic susceptibility reflects well the lithological change along the section (Fig. 2). The periods of high magnetic susceptibility (MS) are bound preferably to grey limestone intervals. In contrast, coal, tufa-lithoclast breccia and characean limestone (but also volcanic ash intercalations) at the top of the section generally show low MS values. Furthermore, the MS log shows striking correlation with the lithological units described in Section 4.1. The Units A, C, E and F are characterised by decreased, B and D by increased MS values. In Unit B, the increased values are restricted to the lowermost and topmost part. Unit D, with generally increased MS, has minimum values concentrated at its top, its base and the erosional boundary between the two subunits. Unit F also shows a slight MS increase in the middle part, interrupted by the ash intercalation, and decrease in marginal parts.

The MS increase can be interpreted as reflecting the increased detritus input in the basin during dominantly wetter climate conditions and during lake-level high-stands. Such an interpretation fits well with the MS pattern recorded in the section. The three dominantly coal-producing periods (Unit A, middle part of Unit B and Unit C), representing a marginal lake environment, show decreased MS. In contrast, their boundary intervals (lower and upper parts of Unit B), represented by grey limestone intercalations and marking the lake-level rise, show increased MS. As pointed out in Section 4.1, these intervals represent condensed marginal facies of transgressional limestone units reaching up to 50 m thickness in the basinal centre. Also, in Unit D, the decreased MS is bound to coal intercalations and to erosional surfaces, proving the lake-level fall and emersion there. The lake-level low-stand at the base of Unit F, suggested by the decreased MS, is demonstrated there by the presence of ripple-bedded limestone overlying the tufa-lithoclast breccia at the top of Unit E.

Natural gamma radioactivity

The gamma ray (GR) log shows generally higher values in the coal-bearing units (A, B and C), the lower part of Unit D and in Unit F (Fig. 2). The strong increase in the latter unit is clearly bounded to ash intercalation, the slight increase above it, in contrast, to organic matter-rich clay. In the lower part, coal shows generally somewhat lower values than the limestone-dominated intervals (lower and upper parts of Unit B and lower part of Unit D). The exception is the peak in the coal seam of the middle part of Unit B. The carbonate-dominated interval in the upper part of the section (upper part of Unit D and Unit E) shows generally decreased GR with a minimum in the thickest tufa-intraclast breccia.

The exceptionally increased GR in the coal seam in mid Unit B most probably reflects radioactive matter input to the basin due to a single volcanic event. The capability of organic matter to bind the radioactive isotopes, and occasional tephra fall events, distort the GR in a manner that is difficult to extract from the signal provided by the siliciclastic input to the basin and the lake-level rise events. Nevertheless, the pattern observed in the MS log for the lower part of the section is also present in the GR log. There, the base and the top of Unit B and the lower part of Unit D show an increased GR, suggesting lake-level rise events there. The generally low GR throughout Unit E suggests a relative lake-level low-stand with maximum reached by the thick tufa-lithoclast breccia.

Transgression–regression cycles

The studied succession shows a number of larger and smaller-scale lake-level oscillations represented by striking depositional facies shifts along the section. In a lacustrine setting such oscillations reflect changes of the regional water budget in response to variations in precipitation, evaporation and groundwater level (Cohen, 2003). Consequently, the depositional history of lakes is highly sensitive to the interplay between humid (high-stand) and arid (low-stand) climate conditions.

Integrated data on lithology, carbonate depositional facies, mollusc palaeoecology and geophysical logging support the recognition of seven transgression–regression cycles (TRC) for the investigated succession at the ~ 10-m scale (Fig. 2). Especially the relative lake-level low-stands can be pointed out with accuracy. In the lower part of the section they are bounded to coal-dominated intervals (Unit A, middle part of B and C), representing the basin-ward progression of the vegetated lake-fringe. In Unit D they are represented by emersion horizons and erosional boundaries. In the upper part of the section the shallowest depositional environments are indicated by a thick breccia interval in mid Unit E, by ripple-bedded limestone overlying breccia at the base of Unit E and by the coal intercalations at its very top.

Superposed to that basic TRC pattern, a larger, ~ 40-m scale lake-level fluctuation is present. It shows minimum lake-levels during TRC 1 and TRC 5 and maximum values during TRC 3 and TRC 7 (Fig. 2). The minimum lake level in TRC 5 starts with the base of Unit E and is marked by an abrupt change of lithology and mollusc content. All point to a severe lake-level fall at its lower boundary. The massive occurrence of rhizoliths is in accordance with that interpretation, indicating semi-arid climate conditions during TRC 5 (Flügel, 2004). Their abrupt disappearance with the onset of TRC 6 is followed upsection by the gradual disappearance of tufa-lithoclast breccia and ever more dominant characean limestones. This points to increasingly open-lake conditions. The reoccurrence of lignite deposition in TRC 7 also points to a renewed increase in regional humidity.

TRC 1 (10 m)

Its lower boundary is marked by the beige limestone in mid Unit A between two thick coal seams. Its ephemeral-pool mollusc assemblage suggests the lake retreat as a cause for the short-term restriction of coal deposition. Minimum values in both GR and MS logs agree with that interpretation.

The maximum flooding is correlated with the interval dominated by grey limestone located in the lower parts of Unit B. Deposition of grey limestone marks the drowning of the vegetated fringe due to the lake-level rise. The increased GR and MS values at this interval agree with this interpretation of carbonate depositional facies. The prolonged open lake deposition is furthermore indicated by thick synchronous limestone deposited in the basin centre.

TRC 2 (15 m)

Its lower boundary is marked by the thickest coal seam in the middle of Unit B. This seam represents the longest persistence of the organic matter-producing depositional environment at the lake margin. The maximal flooding is marked by the grey limestone at the top of Unit B. Based on its position immediately below the Main Coal Unit of the lithostratigraphic Gacko Basin division by Milojević (1976) (= Unit C), it correlates with the lake deepening event, which produced thick limestone deposits in the basinal centre. The increased MS and GR also reflect that correlation. The introduction of a Melanopsis-assemblage demonstrates the onset of perennial-lake conditions.

TRC 3 (9 m)

The lower boundary coincides with the thickest coal seam of Unit C. This is the most massive and most homogeneous coal layer of the whole investigated section, reflecting possibly a mire or peat depositional environment.

The maximum flooding and installation of high-stand conditions are indicated by the change of the limestone colour from grey to brown. This marks the start of the deposition under suboxic bottom conditions in the period of highest humidity, enhanced detritus input and deepest depositional conditions. The preservation of organic matter at the lake bottom can be explained by a thermocline and/or enhanced primary production in the lake. The intercalated brownish clay layers possibly mark the concentrated detrital material typical for deep lake deposition.

This particular TRC probably marks the deepest depositional conditions in the whole section. This conclusion is strongly supported, in addition to by the previous observations, by highest MS and GR, pointing to highest detrital input into the basin. Furthermore, the strongly diversified mollusc assemblage at this particular interval proves the establishment of a long-lived palaeo-lake in the basin.

TRC 4 (8.1 m)

The lower boundary is marked by the erosional surface developed on brecciated grey marls (Fig. 3.6). This surface marks the emersion through a lake-level fall. The internal brecciation of the surface was probably enhanced by root action. The maximal flooding surface (MFS) is correlated similarly to TRC 3 with the colour change from light to dark grey limestone. The strong lake deepening, resulting in partial suboxic bottom conditions, is indicated already by the dark nodules below that surface. The placement of the MFS is additionally supported by the increased values of MS and GR.

TRC 4 is characterised by the presence of an extraordinarily thick, monotonous grey limestone interval. Its occurrence up to the basinal margin reflects the strong lake-level rise combined with persistent and stable open-lake conditions within that interval. The persistence of the highly diversified mollusc assemblage also reflects the long-lasting, stable lake conditions.

TRC 5 (8.2 m)

The lower boundary coincides with the boundary between Units D and E. It is marked by the erosional surface developed above the coal seam-bearing series. The surface is covered by the tufa-intraclast breccia. This marks the retreat of the open-lake conditions and installation of marginal depositional settings comprising a palustrine environment of temporary, macrophyte-vegetated and carbonate-enriched ephemeral ponds. This coincides with the renewed development of an ephemeral pool mollusc fauna (Fossarrulus-assemblage), which replaces completely the diversified fauna of TRC 3 and 4.

The MFS is again correlated with the change from dark grey into dark brown colour now occurring within the rhizolith limestone. As pointed out in Section 4.2.2 the limestone represents a shallow open-lake setting, with rhizoliths redeposited from the pedogenic, vegetated lake-fringe during the lake-level rise. The increase in organic matter within the limestone presumably corresponds with lake eutrophication as a result of enhanced input of denudation material. The rhizolith abundance in the sediment points to a short transport. Their massive production, restricted moreover to this particular interval, can be interpreted as a maximal aridification event within the studied section.

TRC 6 (9.1 m)

The thick tufa-lithoclast breccia marks the lower boundary of this TRC. It represents a long-lasting lake-level low-stand that produced the thickest lithoclast accumulation recorded in the section. The MFS is correlated with the thickest characean limestone-bed, indicating the long-term persistence of open-lake conditions. Absent hypereutrophication of the lake during this cycle reflects the very low primary production in the lake, possibly due to relatively dry climate conditions lasting throughout the interval. Nevertheless, the absence of rhizoliths indicates generally wetter conditions than those characterising TRC 5.

TRC 6 (11.5 m)

This TRC has its lower boundary in a ripple-bedded characean limestone. Its upper boundary is settled with the intercalated lignite laminae at the top of the section (Fig. 3.4), marking the reoccurrence of marginal lake depositional conditions. The MFS is correlated with the thickest limestone interval. It bears characean oogonia and ostracods, indicating a long-term, open lake depositional phase for this interval. The dark brownish, organic matter-rich clays on top mark the termination of the carbonate production and the renewed general humidity increase in this interval. The depositional cycle is disturbed by massive volcanic tuff intercalations in its middle part.

Chronology

The intensities of this characteristic component range between 1 × 101 and 1.5 × 103 mAm − 1. No gyroremanence is observed and, after application of a 100 mT field, the remaining NRM of most samples is negligible. In the upper half of the section a normal low-field overprint is removed between 0 and 32 mT, and a reversed high-field component is demagnetized at higher fields. In the lower, coal-dominated part of the section, all samples reveal normal polarities. The inclinations of the normal polarity component above 32 mT correspond to the inclination of the present-day geomagnetic field at the Gacko site. These directions are therefore interpreted as a secondary overprint. The inclinations of the reversed components are strikingly lower, but in very good agreement with the results from the Sinj Basin (De Leeuw et al., 2010). These reversed directions are interpreted to be of primary origin, which implies that the upper part of the Gacko succession was deposited during a reversed polarity chron (Fig. 7).

Fig. 7

Palaeomagnetic results for the Gacko section plotted in equal area diagrammes as well as in stratigraphic order. Between 40 and 67 m the magnetic signal is characterised by reversed directions, interpreted to be of primary origin. The tilt corrected (tc) and non tilt corrected (ntc) Zijderveld diagrammes for Ga18 are typical demagnetization diagrammes for samples from this part of the section. The lower, coal-dominated part of the section generally shows normal directions. The Zijderveld diagrammes of sample Ga3 are typical for this interval. The tilt corrected (tc) and non tilt corrected (ntc) display both the normal and reversed directions. The red circle represents the present-day field direction at the location of the section. The normal directions clearly have a higher inclination than the reversed directions and are moreover statistically indistinguishable from the present-day field. Therefore, the normal directions are interpreted to be an overprint. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Isotopic dating

The results of the 40Ar/39Ar total fusion experiments of the Gacko samples 1 and 2 are given in the Supplementary information. The weighted mean plateau age of 15.36 ± 0.03 Ma for Gacko 1 is equivalent to the weighted mean plateau age of 15.37 ± 0.02 Ma for Gacko 2 (analytical errors only). However, the inverse isochron intercepts deviate from the atmospheric argon ratio of 295.5 and are indicative for excess argon. Therefore, the inverse isochron age is regarded as the best age estimate of the ash layer. A combined inverse isochron (Fig. 8) yields an age of 15.31 ± 0.03 Ma and trapped 40Ar/36Ar component of 338 ± 12. The uncertainty increases to ± 0.16 Ma when uncertainties in J, the age of the primary standard and decay constants, as reported in Kuiper et al. (2008) and Steiger and Jäger (1977) respectively, are included.

Fig. 8

Inverse isochron diagramme of multiple grain 40Ar/39Ar fusion experiments on Gacko 1 and 2 feldspar. The inset shows the data in more detail. The solid line is the isochron line, the dashed line is the reference line representing the plateau age (15.37 ± 0.16 Ma) and an 40Ar/36Ar intercept value equal to modern atmosphere (295.5), and the ellipses represent the 1σ analytical error.

Correlation to the GPTS

Potential options to correlate the observed interval of reversed polarity in the middle to upper part of the section are C5Bn.1r (14.877–15.032 Ma), C5Br (15.160–15.970 Ma) or C5Cn.1r (16.268–16.303 Ma: Lourens et al., 2004). When taking the 40Ar/39Ar age of the exposed volcanic ash layer into account, it is clear that it can be exclusively correlated to chron C5Br. This means that the deposits between 40 m and 67 m of the section must have accumulated between 15.974 and 15.160 Ma (Lourens et al., 2004), although the exact start and end of sedimentation cannot be determined based on that result. This implies that Lake Gacko developed during the Early Langhian, corresponding to the final stages of the Miocene Climatic Optimum. In the stratigraphic terminology of the Central Paratethys, it correlates to the Early Badenian (Piller et al., 2007).

Astronomical tuning

The inferred superposition of a ~ 10-m scale on a ~ 40-m scale transgression-regression depositional cycle suggests the changes in lake-level might be orbitally paced and are potentially controlled by ~ 100-kyr and ~ 400-kyr eccentricity cycles. To test if this assumption is robust, the MS and GS records were subjected to a spectral analysis in the depth domain, aiming to better visualise their rhythmicity and amplitude modulations (Fig. 9).

Fig. 9

Spectral analysis in depth domain of magnetic susceptibility (MS) and gamma ray (GR) logs, calculated with PAST programme (Hammer et al., 2001). Vertical dashed line marks the non-significant part of the diagramme with frequencies beyond the four-cycle bandwidth. Horizontal dashed lines mark the 0.01 and 0.05 significance levels.

Spectral analysis

The Lomb periodogram for the MS data series shows a single significant power interval with two peaks corresponding to cycle thicknesses of 11.3 m and 8.6 m (Fig. 9). The significant power interval is even broader in the GR data and covers the 16.8 m to 5.6 m range. Within this range, a single high power-peak is detected at a cycle periodicity of 8.4 m. These spectral power distributions imply the presence of a ~ 10-m-scale sedimentary cyclicity in the section and corroborate the inferences based on the depositional facies analysis.

Gaussian band-pass filtering was then used to extract the frequency component of the most significant power-peaks in the Lomb periodogram. Band-pass filters were centred on the peaks and their band-width was defined according to the width of the intervals retrieved in the Lomb periodogram. Filtering of the cycle periodicities of the 8.4 m GR peak with a band-width of 16.8–5.6 m, and filtering of the 11.3 m MS peak with a band-width of 16.5–8.6 m, reveals a close correlation with the inferred TRC cycles (Fig. 10).

Fig. 10

Astronomical tuning of the depositional transgression–regression cycles inferred for the Gračanica section to the ~ 100-kyr and ~ 400-kyr eccentricity curves of La2004 (Laskar et al., 2004). The GR and MS records are additionally shown with their filtered components in depth domain (see text for discussion). Band-widths of the filtered records are according to their power spectrum as illustrated in Fig. 9.

The filtered GR log shows two intervals in which it deviates from the TRC cyclicity (dashed line in Fig. 10), both related to extraordinarily strong natural gamma radioactivity input. As pointed out in Section 4.4.1 these GR peaks are bounded to volcanic tuff fall into the basin. Since they coincide with minima in MS values they cannot be related to increased detrital input. Very low intensities in the MS signal cause a resolution problem in TRC cycles 5 and 6 (dashed line in Fig. 10). However, this low-intensity interval reflects a phase of minimum detrital input into the basin, which suggests a relatively low lake-level.

Periodic changes in eccentricity as forcing factor of the observed lake-level variations

Prominent ~ 100-kyr and ~ 400-kyr eccentricity-forced climate variability characterises the Miocene Climatic Optimum between 16.9 and 14.7 Ma (Holbourn et al., 2007). Continental lacustrine records in central Spain for the corresponding time interval provide clear evidence of eccentricity control on the depositional environment (Abels et al., 2010; Krijgsman et al., 1994), and in the central Mediterranean, climate variability recorded in marine records reflects the eccentricity cycle (Abels et al., 2005).

Adoption of eccentricity (~ 100-kyr and ~ 400-kyr) as the main forcing factor for the transgressive-regressive cycles in Lake Gacko implies a mean sedimentation rate of ~ 0.1 m/kyr for the studied section. Based on lithostratigraphic correlation with the units established by Milojević (1966), the studied succession can be correlated to a minimum interval of about 210 m in the centre of the basin. The minimum sediment accumulation rate for the central part of the Gacko Basin would thus amount to ~ 0.3 m/kyr. This is in good agreement with accumulation rates obtained in the Miocene Lake Sinj of the DLS (de Leeuw et al., 2010).

Several other lignite-bearing lacustrine successions in south-eastern Europe also express a sedimentary cyclicity indicative of orbitally controlled lake-level fluctuations (Van Vugt et al., 1998, 2001). Lignite-marl alternations in the Late Miocene to Early Pliocene Ptolemais coal mine in northern Greece were for example shown to predominantly reflect precessional forcing (Van Vugt et al., 1998). A ~ 21-kyr precessional forcing for the ~ 10 m thick cyclicity of the Gacko Basin would, however, imply an accumulation rate of ~ 1.3 m/kyr. Such a high accumulation rate is only probable in basins with a strong fluvial input, which Gacko is not.

Since strongly eccentricity-forced climate variability characterises the MCO and has already been shown to exert a strong influence on contemporary continental lacustrine environments in Spain, it is the most likely forcing factor for the transgression–regression cycles of Lake Gacko. The resulting sedimentation rates are in good agreement with previously published sedimentation rates elsewhere in the DLS.

Correlation to the astronomical curves

Detailed microfacies analyses of the Spanish Late Miocene successions showed that minima in the ~ 100-kyr and ~ 400-kyr eccentricity amplitudes correspond to prolonged dry climate periods and lake-level falls (Abels et al., 2008). This supports the correlation of humid regional climate phases to intervals with maximum eccentricity values in the astronomical target curve.

Despite its substantial uncertainty, the 40Ar/39Ar isotope age of 15.31 ± 0.16 Ma provides a first-order correlation point that ties the upper part of the section to the increase of the contemporary ~ 400-kyr cycle (Fig. 10), in agreement with the inferred increase of humidity from TRC 5 to TRC 7. The ~ 10-m depositional cycles of presumed ~ 100-kyr eccentricity origin can consequently be correlated to the astronomical curve.

The amplitude modulation pattern of the band-pass filtered component of the GR curve for the interval between TRC 3 to TRC 6 shows a striking correlation to the modulation pattern of the ~ 100-kyr eccentricity curve between − 15.261 and − 15.648 Ma. Thereby, the aridification event (with extended caliche building and thick rhizolith accumulations constricted to TRC 5) corresponds precisely with the minimum of ~ 400-kyr eccentricity values. Consequently, the best fit correlation to the astronomical target curve indicates that the studied section has been deposited between 15.8 and 15.2 Ma.

Discussion

Dinaride Lake System mollusc phylostratigraphy revised

Based on the presence of a phylogenetically progressive endemic mollusc fauna, the Gacko Basin was regarded as belonging to the younger DLS stage by Kochansky-Devidé and Slišković (1978) (Fig. 11). The presented biostratigraphic and cyclostratigraphic data confirm this hypothesis. Correlation of the studied section to the upper part of the Lučane section in the Sinj Basin (De Leeuw et al., 2010) indicates that the Mytilopsis frici-bearing horizon in the Gacko Basin postdates the common last occurrence of M. drvarensis in the Sinj Basin and predates the first occurrence of the still more advanced Mytilopsis aletici. Although the timing of the appearance of these three Mytilopsis species is now clear, the evolutionary mode in the clade remains uncertain because M. frici has not been found in the Sinj Basin yet. Both options – the iterative evolution of two species in different basins, as well as the gradual evolution of M. drvarensis into M. aletici through M. frici – are still possible.

Fig. 11

Stratigraphic position of the Lake Gacko within the Dinaride Lake System geochronology (see text for description) and its correlation with the global climate change (Zachos et al., 2001).

Dinaride Lake System palaeo(bio)geography revised

As indicated above, the evolutionary history of the DLS is generally considered to comprise two successive stages (Kochansky-Devidé and Slišković, 1978). The older phase is characterised by the occurrence of phylogenetically primitive dreissenids, while during the younger phase more progressive dreissenids developed. Strata with primitive dreissenids are present in nearly all lacustrine basins of the Dinarides. It is thus thought that the DLS extended over the entire Dinarides and even into the southern part of the Pannonian Basin during its older phase. Evolutionarily progressive dreissenids are exclusively found in the south-western basins, whereas marine strata cover lacustrine sediments of the first DLS phase in the north-eastern basins. This led to the hypothesis that during the younger phase, named Lake Herzegovina by Krstić et al. (2003), the DLS retreated south-westwards in response to flooding of the northern-eastern lakes by the Paratethys Sea.

The currently available chronologic data, however, do not support this hypothesis. Flooding of the north-eastern lakes by the Paratethys occurs at, or slightly after, 14.9 Ma because the first marine sediments contain calcareous nannoplankton indicative of the NN5 zone (Ćorić et al., 2009). The new chronostratigraphic and cyclostratigraphic data for Lake Gacko and Lake Sinj, two of the main constituents of Lake Herzegovina, indicate that both had already disappeared by the time of marine flooding.

Therefore, the absence of evolutionarily progressive dreissenids such as Mytilopsis frici and Mytilopsis aletici north of the Gacko–Kupres–Šipovo line, i.e. from the part of the DLS stretching from the Sarajevo Basin to the Pannonian Basin, must be explained by other palaeogeographic processes. This might include an early presence of a watershed or the presence of a north–south climatic gradient. Such a gradient was already shown to exist in the Middle-Miocene Paratethys (Harzhauser et al., 2003). Future efforts, including stable isotope research, should point out the exact cause for the lack of an advanced assemblage in the northern DLS basins in the period between 16 and 15 Ma.

Impact of the MCO on the Dinaride Lake System

After the Mi1 Glaciation that struck the Earth in the earliest Miocene, global temperatures started to rise. They peaked during the Middle Miocene Climatic Optimum (MCO), which lasted from 16.9 to 14.7 Ma (Zachos et al., 2001) (Fig. 11). The rising temperatures had a large impact on European ecosystems. Coral reefs, crocodiles, taxodiacean and mangrove vegetation spread northwards into the Central European epicontinental Paratethys Sea (Harzhauser et al., 2003), and ectothermic vertebrates thrived in the Alpine Foreland Basin (Böhme, 2003). Lake Gacko arose during the high times of the MCO (Fig. 11), characterised by minimum ice cap volume and increased greenhouse gas levels (Holbourn et al., 2007; Zachos et al., 2001). This suggests that the optimum climatic conditions stimulated lake formation in the Dinarides. Apparently, the global rise in temperature induced a critical change in the regional evaporation–precipitation balance. The strong expression of eccentricity forcing that accompanied the MCO (Holbourn et al., 2007) expressed itself through major fluctuations in the water budget of Lake Gacko and resulted in cyclic changes in lake-level.

The presented cyclostratigraphic results demonstrate that the palaeoenvironment of Lake Gacko was very sensitive to climatic fluctuations. The coincidence of the rise and disappearance of Lake Gacko with the MCO provides additional evidence of the climate dependence of the DLS. In contrast, the preservation of lacustrine deposits in the geological record depends on the creation of sufficient accommodation space. A severe extensional regime reigned in the Pannonian Basin during the Early and Middle Miocene and might have triggered the formation of basins in the Dinarides (Ilić and Neubauer, 2005). This beneficial geodynamic regime provided the depressions in which the DLS settled and provided sufficient accommodation space to preserve its deposits.

Conclusions

The optimum climatic conditions of the Middle Miocene stimulated formation of a perennial lake in the Gacko Basin. This palaeo-lake mirrored cyclic variations in the ~ 100-kyr and ~ 400-kyr eccentricity of the Earth's orbit, known to have profoundly influenced the Middle Miocene climate. The resulting regional variation in hydro-climate induced recurrent changes in its sensitive palaeoenvironment. A detailed study of the lithological, palaeontological and geophysical information locked up in its lacustrine record documents these cyclic variations.

The lignites in the lower part of the basin infill indicate a vast swamp environment dominated by taxodiacean forests that extended across the whole basin. In response to the changing climate, lake-level rose, resulting in periodic swamp disintegration and marl deposition. Eventually, lignite deposition ceased and a perennial lake developed. This resulted in thick carbonate deposits in the middle part of the succession. Well-diversified endemic mollusc assemblages characterise this long-lived lacustrine environment.

The subsequent appearance of palustrine carbonate breccias and rhizolith limestones indicates the onset of a striking aridification phase with largely decreased lake-levels. Swashes caused reworking of the lake's vegetated rim and temporary marginal ponds. Going upwards in the section, these facies types gradually vanish, and a thickening upward sequence of characean limestone beds characterises the uppermost part of the section. This indicates a lake-level rise inducing longer periods of open-lake conditions. The low-stand period with organic-rich, swamp-related sediments at the very top of the section suggests the return of a predominantly arid climate.

Seven ~ 10-m-scale transgression–regression cycles characterise the studied section. Using the 15.31 ± 0.16 Ma 40Ar/39Ar age for a volcanic ash layer in the upper part of the section as a tie-point, these small-scale cycles are tuned to the ~ 100-kyr cycles of the eccentricity curve. This correlation indicates the investigated sediments accumulated between ~ 15.8 and ~ 15.2 Ma, i.e. during the Early Langhian. The reversed palaeomagnetic polarity of the carbonate-dominated part of the section is correlated to chron C5Br.

The combined chronostratigraphic and cyclostratigraphic data of the present study indicate that the deposits of Lake Gacko are time-equivalent to the upper part of the sedimentary sequence in the Sinj Basin. This confirms previous biostratigraphic correlations based on the similarity of their highly evolved mollusc assemblages. This suggests that the evolutionary stage of the autochthonous molluscs is a powerful tool for regional biostratigraphic correlation.