Environmental changes around the Oligocene/Miocene boundary in the Limagne graben, Massif Central, France

– Continental environments are very sensitive to climatic variations. A unique opportunity to study the climate changes where this stage boundary is well constrained by fossils. Indeed, some localities of the Limagne Graben Basin are so rich in mammal remains that they have become a European reference for mammal biostratigraphy. The dominant sedimentary facies of the lacustrine deposits in the northern part of the Limagne Graben Basin are composed of poorly cemented marls and calcarenites containing various plants and animals remains ( e.g. algae, ﬁ sh bones and teeth, gastropods, ostracods, mammals, birds and reptiles remains) associated with stromatolites. Mammal remains, well described in this area from the literature, were used to constrain the chronostratigraphic context of this lacustrine basin, with re ﬁ nement thanks to new carbon and oxygen isotope measurements, palynological and sedimentological data. In this work, the available information obtained from a classical paleoecological study has been re ﬁ ned by new carbon and oxygen isotope analyses of carbonates and ﬁ sh teeth retrieved from the washed sediment residues. The results of this study con ﬁ rm that the Oligocene/Miocene boundary in Europe was a general period of aridity, associated with a cooler climate, which could be the cause of the decline in mammal paleodiversity observed during this period. The gradual evolution from brackish to fresh waters is indicated by the presence of ostracods and gastropods, and the major development of caddis ﬂ ies. This period is associated to humid climatic conditions while a gradual increase in temperature took place throughout the second part of the Aquitanian. These environmental changes were driven by strong variations of temperatures and a contrasted seasonality.


Introduction
Recent advances made in defining the evolution of global climate during the Cenozoic (Zachos et al., 2001) showed that a warming trend during the Late Oligocene reduced the extent of the Antarctic ice cap. Apart from a few events of glacial stages, the global volume of ice remained low until the Middle Miocene, along with a slight increase in bottom water temperature. One of these glacial events, referred as glacial maximum, occurred at the Oligocene/Miocene boundary as documented by Zachos et al. (2001). The present study focuses on this short-lived event and its consequences, in a region that was particularly sensitive to climate change: the continental environment of the Limagne Graben Basin located in the northern part of the Massif Central in France (Fig. 1). The lacustrine facies of this graben are particularly suitable for performing a detailed paleoclimatic study because of the exceptional density of paleontological remains (especially mammals) that enables a valuable dating of the sedimentary deposits. This manuscript reports on the main results of a multidisciplinary investigation of the "Limagne bourbonnaise" deposits at the Oligocene/Miocene boundary. To reconstruct past climate conditions, palynological, sedimentological and stable isotope analyses (d 13 C, d 18 O) of both carbonates (stromatolites and lacustrine limestone) and phosphates (fish teeth) have been performed.

Geological setting
The Limagne Graben Basin consists of several N-Strending grabens in the northern part of the Massif Central in France and forms a part of the European Cenozoic Rift System (Maury and Varet, 1980;Ziegler, 1990;Merle et al., 1998;Michon, 2000;Dèzes et al., 2004). The northern part of the basin is known as the "Limagne bourbonnaise", the central part as the "Limagne centrale" or "Grande Limagne", and the southern part as the "Limagne d'Issoire" and "Limagne de Brioude" (Fig. 1). Associated with rapid Oligocene subsidence, this structural heritage greatly influenced the sedimentary infill of the basin, which is as much as 2000 m thick in the "Limagne centrale" (Grolier and Tchimichkian, 1963;Morange et al., 1971;Gorin, 1975;Donsimoni and Giot, 1977). In the "Limagne bourbonnaise", the older sedimentary deposits are Late Lutetian in the eastern part of the basin and Bartonian in the western part (e.g. Late Eocene in age) (Riveline et al., 1988). The Rupelian (Early Oligocene) lacustrine sediments were preserved in the "Limagne centrale", "Limagne d'Issoire" and "Limagne bourbonnaise" (Guillot, 1991) and reflect a brackish environment characterized by the presence of gastropods (Potamides lamarckii) and coccoliths locally associated with evaporites.
During the Chattian period (Upper Oligocene), a fluvial supply from the north is observed (Wattinne, 2004). The lake at this time was well developed, as testified by the presence of stromatolites in contact with the basement around the basin edges. Nevertheless, a retreat of the shoreline is indicated at the end of the Chattian by the development of numerous paleosols (for more details see Roche et al., 2018).
During the Aquitanian (Early Miocene), the lacustrine domain was mainly restricted to the northern part of the basin (Limagne bourbonnaise; Fig. 2) with fluvial supplies coming from the east (Wattinne, 2004). The western part is characterized by a narrow basin (Fig. 2) filled with Miocene sediments. It is interpreted as an ancient stream that later migrated eastward (Hugueney et al., 1999) to become the present-day Allier River (Fig. 1). The Aquitanian, marking the last lacustrine episode recorded by the sedimentary deposits filling in the graben, was followed by the development of Burdigalian (Early Miocene) fluviatile systems. These deposits are characterized by marls and sandstones with feldspars and silex (Jung, 1946;Bout et al., 1966;Jeambrun et al., 1973;Giot et al., 1979) then are followed by the Pliocene "sand and clays of the Bourbonnais" formation (Tourenq, 1989).
The overall lacustrine sedimentary record of the Limagne Graben Basin is characterized by an alternation of marls and calcarenites associated with stromatolites showing various morphologies (Donsimoni, 1975;Freytet, 2000;Wattinne et al., 2003) and abundant mammal remains.
Some localities in the Limagne graben are so rich in mammal remains that they became reference levels for the European Standard Mammal Biozonation. For example, the faunal content documented in Coderet is used as the reference to define MP 30 (the last level of the Paleogene, referred to as MP) and in Montaigu-le-Blin to define MN2 (the Neogene referred to as MN; Fig. 3). Other mammal localities record the same associations at the Oligocene/Miocene boundary (Hugueney, 1997).
The general persistence of Late Oligocene lineages and the pattern of change being roughly the same throughout Europe suggests that the faunas of the earliest Miocene derived from an impoverished Oligocene assemblage. That is why the position of the Oligocene/Miocene boundary, in terms of mammalian faunas, was debated (Harland et al., 1982;Cavelier and Pomerol, 1983;Odin, 1994;Berggren et al., 1995;Odin et al., 1997;Ginsburg et al., 1998).
The base of the Neogene is fixed at 23.8 Ma (Hugueney et al., 2003) which is also the base of chron C6 Cn.2n (Fig. 3). The Créchy 1-2 level is well correlated with Brochene Fluch 53 (Switzerland) which is estimated as Oligocene by Picot (2002).
A preliminary overview of the mammal data highlights the location of the Oligocene/Miocene sedimentary boundary in the basin (Fig. 2).
The mammal sites were investigated more specifically in the eastern part of the basin (Créchy, Montaigu-Le-Blin, Gondailly, Chavroches), the Oligocene/Miocene boundary being particularly well exposed at the Créchy site (Hugueney et al., 2003).

Material and methods
Samples were collected from different quarries (172 samples from Montaigu-le-Blin; 204 samples from Créchy; 13 samples from Gannat). All layers of the studied sections were sampled and systematically washed to identify the coarse components (see Figs. 4, 5A and 5B, for an example of section sampling).
The samples were sieved through 1.0 mm, 0.5 mm, 0.250 mm, 0.125 mm, 63 mm, and 25 mm mesh sizes to remove the finer clay phase and to allow the coarser components to be identified (specifically floral and faunal contents), which were then sorted and counted in a semiquantitative way, with the quantitative abacus proposed by Terry and Chilingar (1955).
Nineteen selected samples were sent to the BRGM laboratory in Orléans, France, to extract the palynomorphs (see Wattinne (2004) for the protocol used). Only four of these samples yielded microflora.
Oxygen isotope ratios of fish tooth phosphate are now considered as reliable low-temperature paleothermometers (Tudge, 1960;Kolodny et al., 1983;Shemesh et al., 1988, Lécuyer et al., 2013. It is widely considered that the oxygen isotope composition of fish teeth provides both water temperature and isotopic composition when the fish was alive and that this record is preserved after death. Oxygen isotope analyses were performed on fossil fish remains sampled from the Créchy and Montaigu-le-Blin quarries. The studied fossiliferous fish beds show a monospecific assemblage of Cyprinidae (Tarsichthys sp. TROSCHEL determined by J. Gaudant). The teeth were isolated through a simple washing of the sediment, then extracted and crushed for isotope analysis. About 30 teeth were pooled per sample, which represents a weight ranging from 15 to 30 mg. Sediment layers delivering more than 30 teeth (fossiliferous fish beds) were relatively scarce and were sufficient for only seven analyses (labelled here as layer a to g). Seven additional analyses were performed on teeth collected from contiguous layers (labelled here as    Becker et al. (2002): the base of the Aquitanian and the corresponding magnetostratigraphy) was erroneously placed in MP 30. It should be placed in level 58 of the section, see Picot (2002). layers A to G) that is to say a mixture with teeth coming from different layers but attributed thanks to mammal rests and sedimentological study at the same period of time. This mixture of teeth samples was necessary to obtain the minimum weight required for a high-precision oxygen isotope analysis of the tooth apatite phosphate (Tab. 1). The isolation of phosphate radicals from the fish teeth was performed by precipitating Ag 3 PO 4 crystals using a method developed by Crowson et al. (1991) and adapted by Lécuyer et al. (1993). The extraction and oxygen isotope analysis followed the method described by O'Neil et al. (1994) and modified by Lécuyer et al. (1998).
For the phosphates, we used the following isotopic fractionation equation (Lécuyer et al., 2013): Four samples of lacustrine limestones were selected from the fossiliferous fish beds (3 from Créchy, 1 from Montaigu-le-Blin) in order to compare their d 18 O values with those obtained from the fish teeth. The analyses were performed with a Finnigan Delta E isotopic ratio mass spectrometer at the Laboratoire de Géologie de Lyon.
Thin sections for petrographic analyses were made on 10 stromatolite samples and studied under optical transmitted light and cathodoluminescence; the latter technique revealed the internal structure of the stromatolite and locally the presence of diagenetic alteration. In addition, 42 laminations of a stromatolite from Créchy (10 cm high; CRE-1) were sampled with a microdrill core system in order to characterize the isotopic variations recorded during its growth.
Finally, five selected stromatolites (around 10 cm high) from Créchy, Gannat, Chavroches and Montaigu-le-Blin were powdered and analysed in bulk in order to characterize the global isotopic variations recorded in their structure through the Chattian/Aquitanian boundary.
For the carbonates we used (Anderson and Arthur, 1983) isotopic fractionation equation: The VSMOW/PDB transition can be calculated from the Coplen et al. (1983) The use of these equations requires an estimate of the water d 18 O value, which can be obtained using the Bowen and Wilkinson (2002)  The vegetal communities were mainly composed of coniferal forest elements (including Tsuga, Abies and Picea taxa) and subsequently of deciduous forest taxa represented by Betulaceae, Ulmaceae, Corylaceae, Fagaceae, Juglandaceae, Oleaeceae families. Mega-mesotherms specimens are poorly represented. Nevertheless, the discrepancies highlighted elsewhere between the Chattian and Aquitanian floras (Hochuli, 1979(Hochuli, , 1984Guinet, 1984;Chateauneuf and Nury 1995) also exist in the studied area. The disappearance of Tsuga pollen is observed in the upper part of the Chattian sequence and an increase of Euphorbiaceae-Juglandaceae is pinpointed at the base of Aquitanian. Correlations of corresponding associations with the Monbrison basin Sittler, 1969, 1976;Gorin, 1975) or the Provence basin (Bessedik, 1985;Chateauneuf and Nury, 1995) are difficult, due to scarcity and discontinuity of sampling in the Limagne basin area. However, the Créchy-Bas microflora, well dated by mammals (MP29), can be compared with the Provence basins. In the Provence basin, the microflora content at the Chattian-Aquitanian transition has been well documented: a cool dry phase took place during the upper part of the Chattian, but the following Aquitanian climatic changes still need to be better defined (Bessedik, 1985;Chateauneuf and Nury 1995;Nury, 1988;Nury et al., 2016), (Bessedik, 1985, Hugueney andTruc, 1976;Hugueney et al., 1999;Escarguel and Aguilar, 1997;Neubauer et al., 2015).

Sedimentology and associated faunal content
Sections in the Créchy, Montaigu-le-Blin, Gannat and Gondailly quarries were studied in detail for facies analysis. Four major depositional environments have been proposed (Wattinne, 2004): a fluvial domain, characterized by channels infilled by quartz and calcarenite pebbles, conglomerates and sands composed of quartz and locally heavy minerals; a proximal lacustrine domain characterized by noncemented calcarenites composed of oncolites, oolites, ostracods, gastropods, fish remains, algal debris and caddisflies, all of them being deposited in a shallow-water environment favourable to stromatolite growth; a deeper eutrophic lacustrine zone characterized by wackestones, micropackstones, green marls, and bioaccumulations of ostracods or gastropods; a palustrine zone, leading sometimes to protosols with root traces and mudcracks, associated with terrestrial fauna (birds bones, eggshell fragments, mammal remains).
Several repetitive centimetric sequences characterized by alternations of stromatolite-bearing calcarenites and clayey layers have been observed. Most of the clayey layers were halting the growth of the underlying stromatolites. The clayey layers deposits are interpreted as changes in the climatic conditions, with a more arid climate leading to eutrophic conditions.
The Oligocene and Miocene sedimentary deposits contain lacustrine faunal and floral associations which vary in terms of relative abundance around the Oligocene/Miocene boundary. The main faunal and floral assemblages have already been described in the Créchy quarry (Hugueney et al., 2003).
The Oligocene deposits (calcarenites and marls) are dominated by ostracods (Créchy, Gannat) with rare gastropods   (Bithinella arvernensis BOUILLET = Pseudamnicola arvernensis REY) indicative of a fresh-water environment. The abundance of fish remains at the end of Oligocene is interpreted as successive thanatocoenoses formed during dry periods (Hugueney et al., 2003). Moreover, at the end of the Oligocene, a major development of palustrine facies leading to protosols has been observed. Root traces, mudcracks and fossil bird remains indicate a retreat of the shoreline. Associated faunal content show specific gastropods (Hydrobia sp.), which are indicative of a brackish water environment at Créchy (P. Lozouet, pers.com.), (Fig. 6). The Miocene deposits (MN1, MN2) are mostly characterized by calcarenites associated with stromatolites, caddisflies, gastropods and ostracods (Fig. 6). The most abundant lacustrine gastropods are Valvata sp. and Pseudamnicola arvernensis, which are usually found in a fresh-water environment (Créchy, Montaigu-le-Blin, Gondailly). All these variations in fauna content are interpreted as resulting from salinity variations in response to a drier event at the Oligocene/Miocene boundary.
The shallow-water lacustrine environment was very sensitive to climate changes and the alternance of palustrine and lacustrine facies through time is interpreted as reflecting changes in the paleo-lake levels that may be linked to rainfall fluctuations associated with a marked seasonality.

Stable oxygen isotope analysis 7.1 Stable isotope analysis of bulk stromatolites
The stromatolites located in the calcarenite levels grew preferentially during wet periods . The data (Tab. 3) show that Aquitanian stromatolites have higher d 18 O values (À0.4; À0.1) than Chattian stromatolites (À2.9 to À1.6). Mammal remains (Hugueney, pers com) indicate that Chavroches deposits are older than the Montaigu-le-Blin ones despite their belonging to the same biozones. At the same time, the Chattian stromatolites have negative to slightly positive d 13 C values (À3.2 to 1.5) whereas Aquitanian stromatolites  Hillaire-Marcel, 1992, 1993;Andrews et al., 1993Andrews et al., , 1997Arenas et al., 1993Arenas et al., , 1997Zamarreño et al., 1997). As these stromatolites are well dated by sedimentary deposits, we can observe a general trend in the lake with a gradual increase of water evaporation associated with higher photosynthetic activity (Fig. 7).
Stromatolites from Créchy (CRE-1) show a positive correlation between d 18 O and d 13 C values (r = 0.75; Fig. 8), except for samples 24 and 27. The covariance between the d 18 O and d 13 C values is ascribable to the isotopic evolution of the water mass and indicates rather stable hydrological conditions (Talbot and Kelts, 1990), i.e. a closed lake where primary production is influenced by periods of either water dilution or concentration. This suggests that during the growth of stromatolites, the water of the lake had probably led to establishing an isotopic equilibrium between atmospheric CO 2 and the dissolved inorganic carbon (DIC) enhanced by photosynthetic activity. This process involves a 13 C enrichment of the water DIC due to the preferential use of the 12 C by the microbial community during photosynthetic activity (Rau, 1978;Herczeg and Faifbanks, 1987;Merz, 1992;Casanova and Hillaire-Marcel, 1993). Two periods are characterized by uncorrelated d 18 O (more negative) and d 13 C values (less negative) as revealed by samples 24 and 27 (Fig. 8). They indicate disturbances that affected the lake hydrology during the stromatolite growth. These disturbances could have various origins, such as a northward opening of the basin or a local  river input. Lastly, it can be seen that the least negative d 18 O values are invariably located at the "surface" levels present in the stromatolite. Studying these surface levels in thin section and by cathodoluminescence shows that they are affected by dissolution gulfs associated with early diagenesis. They are related to halts in the algal growth. The isotopic record is thus most likely related to an alternation of humid and dry periods changes recorded in the stromatolite structure. Observed isotopic patterns also highlight the discontinuous growth of the stromatolite as well as the gradual increasing rate of evaporation of the lake. The general trend of the isotopic signal is interpreted as a global drying of the area through time Hillaire-Marcel, 1992, 1993;Arenas et al., 1993Arenas et al., , 1997Andrews et al., 1993Andrews et al., , 1997Zamarreño et al., 1997). As all the studied stromatolites recorded these periods of early diagenesis only at their surface levels, this observation allows the comparison of the average d 13 C and d 18 O values between the various stromatolites.

Results for the Limagne fish teeth
If it is postulated that the elevation of the Chattian and Aquitanian lakes were close to sea level (Briot et al., 2001), and applying the Bowen and Wilkinson (2002) equation to the isotopic results (Tab. 5), we obtain a d 18 O value of the lake water ranging from À5 to 0‰ (VSMOW). We emphasize that this value is only an approximation and does not take into account the altitudinal variations of the paleo-landscape. Those equations applied to a fossil lake cannot provide water temperatures but rather their potential variations through time. Accordingly, the isotopic record is interpreted in terms of positive or negative offsets relative to an arbitrary and constant water d 18 O value (À3‰) during the periods of aridity. The d 18 O value of the lake water was controlled by its hydrological balance (drainage, precipitation, evaporation), although variations through time most likely resulted from changes in either air relative humidity, air temperature and wind velocity. Sedimentological analysis contributed to put limits on these uncertainties; for example, tenches live in an eutrophic medium characterized by clayey deposits and lacustrine limestones, which is an environment associated with arid climates, whereas stromatolites of the Limagne graben grew under rather wet conditions. Lacustrine limestones could have thus precipitated in the higher part of the water column due to the intense phytoplanktonic activity during periods of aridity. Samples of those limestones (Tab. 5) were selected in order to compare their d 18 O values with those obtained from the fish teeth supposedly related to the same period of aridity (Tab. 6). The results depend on whether the sample corresponds to a single or a mixed layer, with the calculated temperature offsets varying between À4°C and þ2°C (Tab. 6). These results show that the temperature variations between the Chattian and the Aquitanian recorded by the fish teeth are related to a weak climatic warming during the Aquitanian period.

Climate evolution around the Oligocene/ Miocene boundary
The Créchy quarry shows a great variety of rodents preserved in the Late Oligocene beds, but also registers the Oligocene/Miocene transition with a drastic reduction of the fauna content and diversity at the beginning of the Neogene (Hugueney et al., 2003). The basal Miocene impoverished faunas are dominated by eomyids and glirids and furthermore locally, zapodids. Later, the richness in species is greater as observed in Montaigu-le-Blin, which yielded an Early Miocene (MN2a) fauna characterized by 18 families and 33 species, but only 11 rodents. The faunal community indicates a more open tropical landscape. No younger fauna have been observed in the Limagne graben. Although episodes of appearance and disappearance took place in the Limagne area during the Late Oligocene (MP29-30), the most drastic event took place at the MP30/MN1 boundary Among the persistent lineages, new and more derived species developed at the Miocene base (Hugueney, 1997): at the family level, the entire Theridomyidae family, endemic to the European Late Eocene and Oligocene, disappeared; at the genus level, Cainotherium became the unique representative of the Cainotheriidae family, and Amphiperatherium of the Didelphidae family and Creodont Hyaenodon while the Artiodactyls Microbunodon and Anthracotherium disappeared. Neither the Eomyid Eomys nor the Cricetid Adelomyarion, Glirid Gliravus, and Castorid Rhizospalax crossed this boundary; at the species level, all the Oligocene species of Plesiosminthus disappeared and a single species, P. myarion appeared. Legendre (1987Legendre ( , 1989 demonstrated that local extinctions in mammal faunas can be correlated with successive temperature decreases and that the more noticeable effect of a temperature drop is a decrease in the number of species weighing less than 500 g, which is the case for most of the Oligocene rodents. In this case, the change between the Late Oligocene and earliest Miocene faunas could be related to one or more successive climatic changes associated to enhanced seasonality (Hugueney, 1984;Comte, 2000;Becker et al., 2002).
The most comprehensive studies of the Chattian/Aquitanian floras and microfloras and their relationships to paleoclimates were mostly carried out on the western border of the Mediterranean Sea (Rhone Valley, Provence, Languedoc Roussillon, Catalonia, Portugal). The occurrence of some species (Prosopis) in the Provence area, the Swiss Molasse (Berger, 1990) and the Northern German Tertiary (Utescher, 2000) may indicate increasing climatic aridity during the Late Oligocene and Early Miocene (Guinet, 1984). Regression of angiosperms and the correlative development of Pinaceae and herbaceous plants reflect a dry period before the return of the Aquitanian-Burdigalian thermophitic floras (Chateauneuf, 1972;Chateauneuf and Nury, 1995;Bessedik, 1985;Pais, 1986).
The results obtained from the present study enable to outline the modality of the climatic evolution through the Oligocene/Miocene boundary in the Limagne graben. The sedimentary analysis of the late Oligocene deposits records an increase in palustrine indicators (mudcracks, root traces), reflecting a retreat of the shoreline. The lake water changed from fresh to brackish during the Chattian (Hugueney et al., 2003) as testified by the presence of specific gastropods (Hydrobia sp.) and flamingo nests (Paleolodus ambiguus). This period is also characterized by the development of thanatocoenoses of tenches preserved in the argillaceous beds of the end of the Oligocene.
Throughout the Oligo-Miocene boundary, the lacustrine faunal associations show a general evolution from brackish to fresh water (Tab. 7), as indicated by the appearance of freshwater ostracods and gastropods, a major development of caddisflies, and the disappearance of brackish-water gastropods (Hugueney et al., 1999;Hugueney et al., 2003).
Furthermore, the stable isotope compositions of stromatolites indicate that evaporation increased gradually toward the Oligocene/Miocene boundary.
The pollen data confirm this trend of increasing aridity furthermore revealing a cooling stage at the Oligocene/ Miocene transition. These cool conditions were first highlighted by Gorin (1975), based on the presence of Pinaceae-rich pollen assemblages. The increasing aridity associated with a cooler climate may have led to the well-documented mammal impoverishment (Hugueney, 1984;Legendre, 1987Legendre, , 1989Comte, 2000;Becker et al., 2002) A wetter climate prevailed during the lower part of the Aquitanian (MN1) followed by a gradual increase in temperature toward the upper part of the Aquitanian (MN2).  (Gorin, 1975) and by the high fish tooth d 18 O values measured throughout the Chattian/Aquitanian boundary.
Our study confirms that the global climate evolution around the Oligocene/Miocene boundary involved several successive temperature changes as already suggested by Hugueney (1984), Comte (2000) and Becker et al. (2002), along with a general trend of increasing aridity that settled during the Aquitanian (Tab. 7).