Abstract
The middle Eocene interval at some Paris Basin localities was studied through high-resolution stratigraphy. Abundance data (499 species, 37 719 individuals) on the distribution of molluscs, collected at 12 shell beds of the middle Lutetian and lower Bartonian, formed the basis for a palaeoecological study. The middle Lutetian succession is subdivided into several elementary depositional sequences (EDS) interpreted as the product of relative sea-level change. Species-abundance distributions are better correlated with EDS than with geographical locality, suggesting that sea level played an important role in the distribution of palaeocommunities. Diversities were compared with analogous data from modern subtropical and warm-temperate intertidal and subtidal communities. We found that sea-level variation is responsible for a major change in the upper part of the middle Lutetian succession, leading from high- to low-diversity palaeocommunities. From base to top sampled palaeocommunities indicate a transition from high-energy and mesotrophic (EDS 2) to oligotrophic low-energy conditions of a sandy lower shoreface (EDS 4) to an upper shoreface (EDS 5 and lower Bartonian), the last with mangroves and a seagrass cover. Notwithstanding the Lutetian cooling, we found that subtropical conditions reached as far north as the Paris Basin. Our study suggests that climatic fluctuations might be obscured by facies control.
The middle Eocene of the Paris Basin belongs to one of the most intensely studied stratigraphic intervals of geology. Shell beds there have attracted scholars such as Guillaume Bruguière, Jean-Baptiste Lamarck, Gérard Deshayes and Alcide d'Orbigny, who contributed to establishing the principles of modern stratigraphy and the tenets of macroevolution (Rudwick 2005). As a corollary result of the many efforts, the lists of its macro- and microfossils are mostly completed (Merle 2008, and references therein; Courville et al. 2012; Lozouet 2014), but so far no species-abundance data of quantitative bulk samples have been used to perform palaeocommunity and diversity analyses, apart from a recent study devoted to one particular shell bed (Sanders et al. 2015). Such analyses, however, are of major interest because fossil assemblages of the Paris Basin show characters not known from benthic communities living today at analogous latitudes, where many families and genera have meanwhile undergone an almost complete turnover in importance (Lozouet 2014; Tomasových et al. 2014). For instance, the Paris Basin was located around 30–35°N within the warm-temperate belt at a time of general cooling with respect to the Early Eocene climatic optimum (EECO). Nonetheless, overall species richness calculated at the stage level suggests values similar to those of modern tropical hotspots (Huyghe et al. 2012a; Lozouet 2014; Sanders et al. 2015).
The thickest part of the middle Lutetian succession is composed of unlithified fine-grained sandstones, with a large component of biogenic carbonates. Original aragonitic shells and residual colour patterns are finely preserved and resemble much younger Neogene shells (Merle 2008; Caze et al. 2011). Other well-studied Palaeogene examples (e.g. the Pyrenees: Dominici & Kowalke 2014; SE Europe: Oppenheim 1901; the Middle East: İslamoğlu et al. 2011; Harzhauser et al. 2012), whether siliciclastic or carbonate-dominated, are poorly preserved. Similar good preservation is found only in some unlithified siliciclastic Bartonian successions of the US Gulf Coast, including the Gosport Formation (e.g. CoBabe & Allmon 1994) and the Stone City Formation (e.g. Zuschin & Stanton 2002). Sedimentary structures, mudstone partings and stratal surfaces are subtle, and massive bedding is frequent, so that field correlations between Paris Basin outcrops traditionally are not based on facies contrast but on macrofaunal content (Gély 1996). The main bioclastic components are mollusc shells, in concentrations of both sedimentological and biogenic origin. Large foraminifera are also abundant, although their accumulations do not take the form of the nummulitic banks that are common in most coastal and shelfal Eocene successions of nearby seas (Pyrenees: Dominici & Kowalke 2007; Huyghe et al. 2012b). If the Eocene is a ‘strange old world’ (Clyde 1999), its expression in the Paris Basin appears an utter mystery.
Paris Basin Palaeogene stratigraphy is relatively well understood, with isopachous relations across long distances (Gély & Lorenz 1991; Gély 1996). This allows a good correlation of local successions with regional subsidence patterns and global events. Available studies recognize tectonic stability from the Danian to late Bartonian (Robin et al. 2003). During this time, sedimentation was controlled by major sea-level fluctuations, typically in the range of third-order sea-level cycles (i.e. lasting 0.5–5 myr) and fourth-order sea-level cycles, which are well framed in a chronostratigraphic framework (Gély 1996; Huyghe et al. 2015). The third-order Lutetian cycle, for example, was formed by three fourth-order sea-level cycles. During the major transgression, which lasted more than 3 myr (Huyghe et al. 2015), the coastline moved in a southern direction in a stepwise fashion, with maximum flooding occurring in the middle Lutetian, apparently without superimposed high-frequency fluctuations (Gély 1996).
To explore the ecological structure of middle Eocene shell beds and shed further light on the history of sea-level changes, we have reconsidered three classical Lutetian localities, well known at least since the 18th century: Ferme-de-l'Orme and Grignon, both near Paris, and Fleury-la-Rivière, near Reims (Merle 2008; Courville et al. 2012). In addition to the Lutetian localities, we included the lower Bartonian succession at La Guepelle, to the NE of Paris, also one of the type areas for the definition of this stage (Pomerol 1981; Fig. 1). The significance of sedimentary facies was reconsidered and, to explore the nature of the Paris Basin biota, fossil assemblages were bulk-sampled in some key units. Abundance data on the distribution of molluscan species were evaluated and used for a multivariate analysis and to compute sample-level diversity. This allowed us to carry out for the first time a palaeocommunity analysis of the Paris Basin marine macrofauna, which provides information on the trophic structure of the benthic ecosystem. A key question is whether the diversity of the Paris Basin samples is comparable with that of modern tropical or warm-temperate environments. The sample diversity of the fossil assemblages, also studied for the first time in the Paris Basin, was therefore compared with that of modern assemblages from two well-studied shallow marine settings, the northern Adriatic Sea and the northern Red Sea, in warm-temperate and tropical climates, respectively (Zuschin & Hohenegger 1998; Sawyer & Zuschin 2010). Species-level abundance data and the sample-level quantitative comparison opened a new approach to interpreting Paris Basin shell beds. The influence of sequence architecture is also of major interest because most changes in first and last occurrences of species, and widespread changes in species abundance and biofacies, occur at sequence boundaries and at major transgressive surfaces (Holland 2000). This calls for evaluation of changes in taxonomic diversity in the context of environmental bias to ensure that these changes are not simply driven by sequence architecture (Smith 2001).
Middle Lutetian at Ferme-de-l'Orme, Grignon and Fleury-la-Rivière, and lower Bartonian at La Guepelle, in the Paris Basin, with position of bulk samples. Small numbers and arrows at Grignon represent the extent of sedimentary units as defined by Guernet et al. (2012). Large numbers in circles represent fifth-order depositional sequences as recognized in this paper, with main transgressive surfaces correlated according to the literature (Gély 1996; Merle 2008). Description of the succession at La Guepelle is from B. Pattedoie (pers. comm., 2007). A major unconformity separates the middle Lutetian from the lower Bartonian (Huyghe et al. 2015; not all intervening strata are shown here). Insets show explanation of symbols and location map of the study sites. Labels in rectangles refer to bulk samples collected in main shell beds.
Specifically, this approach allowed us to reconstruct Paris Basin benthic palaeocommunities at a time of gradual change during the middle Eocene and through climatic thresholds. The joint input from sedimentary facies analysis and palaeoecology sheds light on external factors controlling the record. This study allows a better understanding to be obtained of the effects of global sea-level fluctuations in driving both the structure and the composition of benthic communities.
Sedimentary facies
Paris Basin strata are roughly horizontal and isopachous (Gély 1996; Merle 2008), connecting outcrops as far as 160 km from one another. Sections are oriented along depositional strike, which is the reason for little lateral facies change within the studied succession (Fig. 1). The most detailed sedimentological description of the ‘Falunière de Grignon’, where the Middle Lutetian is thickest and most complete, has been given by Jean-Pierre Gély and Didier Merle (cited by Guernet et al. 2012; Sanders et al. 2015). Those researchers described and numbered the sedimentary units from 1 to 8, starting from the top (i.e. lower numbers represent younger units), a basic framework adopted here. In this phase of our research, middle Lutetian palaeoenvironments are interpreted based on recent palaeontological studies carried out at Grignon (Guernet et al. 2012; Huyghe et al. 2012a) and Fleury-la-Rivière (Courville et al. 2012; see also Tomasových et al. 2014) and from our own fieldwork. Starting from the oldest, the sedimentary units of Gely & Merle (cited by Guernet et al. 2012) are summarized as follows.
Units 7 and 8. Bioturbated calcareous sandstone with abundant glauconite and quartz and sparse shell material. Thickness 3–4 m. Based on field evidence, the association is characterized by Cardium (Orthocardium) subporulosum d'Orbigny 1850. A tabular shell bed in the upper part is marked by sparse gravels and Turritella shell debris. Fine gravels in the lower part of the Fleury section, within sedimentological shell concentrations (Fig. 2a) associated with bones of marine vertebrates, are here correlated with the preceding shell bed. Units 7 and 8 have a complex stratigraphy, at least in their lower part at Grignon. Samples PB27 and PB16 were collected in the lower part of the Fleury-la-Rivière section, in deposits correlative with the upper part of unit 7 at Grignon (Merle 2008; Courville et al. 2012).
Middle Eocene shell beds of the Paris Basin. (a) Sedimentological shell concetration at Fleury-la-Rivière, near an unconformity separating the middle Lutetian from the Ypresian (EDS 2). (b) Campanilopa giganteum biogenic shell bed in plan view (Fleury-la-Rivière, EDS 2). (c) Loosely packed bioclastic fabric at Grignon (EDS 4): disarticulated bivalves interspersed with shell hash. (d) Densely packed bioclastic fabric (Grignon, EDS 4; height of outcrop about 20 cm): disarticulated and nested bivalve shells concentrated in pockets. (e) Potamides lapidorum shell bed at Ferme-de-l'Orme (EDS 5): laterally continuous shell bed within shelly sands. (f) Shell bed in plan view, with colonial corals and bivalves (upper part of EDS 4, Grignon). For scale: pencil, 16 cm; coin, 1.5 cm.
Unit 6. Cross-bedded, locally bioturbated calcareous sandstones, rich in quartz and glauconite, about 2 m thick. No samples were evaluated from this unit.
Unit 5. Highly bioturbated massive calcareous sandstone with abundant quartz and glauconite, about 1.5 m thick. In the lower part of the unit, loosely packed shell beds are characterized by Campanilopa giganteum (Lamarck 1804), which dominates the local biogenic shell concentration with its high biomass (Fig. 2b). The upper shell beds, with a species-rich mollusc association, are instead dominated by Sigmesalia intermedia (Deshayes 1832) at Fleury-la-Rivière (Courville et al. 2012). Sample PB9 was collected in unit 5 at Grignon.
Unit 4. Massive calcareous sandstone with shells in loosely (Fig. 2c) or densely packed shell beds (Fig. 2d), or also resting on irregular erosional surfaces in the basal part of the unit, in event beds of sedimentological origin. Glauconite is rare. Thickness is about 2 m. The association is characterized in the field by Chama lamellosa Lamarck 1806 and abundant Orbitolites complanatus Lamarck 1801. Sample PB6 was collected at Grignon.
Units 2 and 3. Lithified massive calcareous sandstone about 2.5 m thick with very rare glauconite, sparse shell beds and abundant Orbitolites in the lower part. In the upper part, the association is dominated by Seraphs sopitus (Solander cited by Brander 1776) and Avicularium lithocardium (Linné 1771), and some colonial corals occur (Fig. 2f). Samples PB1 and PB3 were collected at Grignon, and PB10 at Ferme-de-l'Orme.
Unit 1. Massive calcareous sandstone, about 0.5 m thick, with thin laterally continuous shell beds (Fig. 2e, sample PB14). The mollusc association is dominated in the field by Potamides lapidorum (Lamarck 1804), Vicinocerithium echidnoides (Lamarck 1804) and Saxolucina saxorum (Lamarck 1806), and miliolids dominate among the benthic foraminifera.
The higher content of glauconite in older units may reflect lower sedimentation rates, but the decreasing quartz content in the up-section direction gives evidence of the contrary.
Our investigations of the lower Bartonian succession cropping out at La Guepelle, also known as Guepelle Sandstone Fm and Beauchamp Sandstone Fm (biozone NP16: Huyghe et al. 2015), are based on an unpublished description by B. Pattedoie (pers. comm., 2007). The third-order sequence boundary at the base of the Bartonian succession (Fig. 1; see Huyghe et al. 2015) was not observed in the field. Sedimentary facies at La Guepelle include a number of massive sandstone units a few metres thick. They are characterized by a sparse or loosely packed bioclastic fabric and thinner calcareous intervals in the middle part of the succession, topped by an important transgressive surface marked by boreholes of bivalves (Fig. 1). These characters led previous researchers to the interpretation of a lower shoreface environment, comparable with that sampled in the middle Lutetian (e.g. Huyghe et al. 2015). A coarsening-upward facies sequence above the transgressive surface, topped by coarse-grained sandstones with cross-bedding, indicates the passage to a high-energy upper shoreface seafloor. Otherwise, as for the middle Lutetian, a better palaeoenvironmental reconstruction is expected from a palaeoecological analysis of sampled shell beds.
Sequence stratigraphy
The Palaeogene Paris Basin infilling is subdivided into five large-scale depositional sequences of million-year duration, bounded by major unconformities (third-order sequences in the sense of Haq et al. 1987) and formed by the stacking of smaller units (Gély & Lorenz 1991; Huyghe et al. 2015). The Lutetian third-order sequence, of about 8 myr duration, has been divided into three fourth-order sequences: sequence V (comprising the Glauconie grossiére Fm, the Calcaire à Nummulites laevigatus Fm and the Calcaire grossier Fm), sequence VI (Marnes et caillasses inférieures Fm and Falun de Foulangues Fm) and sequence VII (Marnes et caillasses supérieures Fm). These units approximately correspond to sequences A, B and C of Gély (1996). Sequence V is of long duration (about 5 myr) and comprises one sea-level fall dated around the middle of biozone NP15 of calcareous nannoplankton biostratigraphy, within Polarity Chron C20, followed by a new, rapid transgressive event and a long, stepwise fall, until reaching a major lowstand at the boundary C20–C19 (Kominz et al. 2008). Evidence for one episode of these global fluctuations is provided by the fourth-order sequence boundary at the top of NP15 (V–VI unconformity of Huyghe et al. 2015).
The middle Lutetian localities under study here belong to the highstand systems tract (HST) of sequence V. Around Paris, these sediments sharply overlie glauconitic strata of the lower Lutetian, whereas furthest west, near Reims, they rest on lower Eocene deposits (Gély 1996), indicating an important erosion preceding the major middle Lutetian transgression (Chron C20). The middle Lutetian is formed by elementary units A6–A10, interpreted as parasequences (Gély 1996; Huyghe et al. 2012a). As an alternative hypothesis, we propose that beds and bedsets are instead stacked to form small-scale sequences bounded by subtle unconformities, expressing low-amplitude fifth-order cycles and forming the building blocks of the whole succession (elementary depositional sequences: EDSs, numbered 1–6) (Fig. 1). Analogous high-frequency fluctuations, similarly expressed by metres-thick units, are recognized in Lutetian carbonates of the Spanish Pyrenees, increasing in amplitude and thickness in the Bartonian (Huyghe et al. 2012b).
Facies changes justifying the grouping of bedsets into EDSs include the following: (1) the presence of fine-grained gravels (lower part of EDS 2 at Grignon and Fleury: Fig. 2a); (2) sharp-based coarse-grained strata with sedimentological shell concentrations (lower part of EDS 3 at Grignon and Fleury; lower part of EDS 4 at Grignon); (3) enrichment in carbonates in the middle part of EDS 4; (4) occurrence of a shallow subtidal fauna in EDS 5. EDSs 2–4 are stacked to form a retrogradational pattern, with calcareous sands of EDS 4 possibly representing a relatively deeper setting, although not exceeding the lower limit of influence of wave action during storms (i.e. lower shoreface). All 10 middle Lutetian bulk samples belong to biozone NP15 of calcareous nannoplankton stratigraphy (Huyghe et al. 2015). Samples were collected in the transgressive tract of EDSs 2–5 (EDS 1 and EDS 6 were not sampled), in biogenic shell beds from upper or lower shoreface deposits (corresponding to ‘nearshore’ and ‘inner shelf’ samples, respectively, of Tomasových et al. 2014).
A superficial comparison suggests that also the Bartonian at La Guepelle may be subdivided in a number of EDSs. Two sequence boundaries are therefore proposed, with some transgressive surfaces intervening in the succession (Fig. 1). Two bulk samples were collected in the lower EDSs, which were possibly deposited in an upper shoreface environment.
Most of the 12 samples were collected in a shell-rich and bioturbated, massive sandstone lithofacies, indicative of low-energy subtidal conditions below fair-weather wave base. Shallower conditions were inferred from the presence of dispersed fine gravels (PB27), abundance of intertidal gastropods (Potamides: PB14), and position in a coarsening-upward succession (PBN7, PBG). High glauconite content indicates low sedimentation rates. Samples PB1 and PB3 were collected in a calcareous lithofacies with denser accumulation of biogenic shell beds (Fig. 2d), suggesting a further lowering of sedimentary input with respect to underlying and overlying strata.
Analytical methods
Bulk samples (0.5 l) were washed through a series of sieves (1.0, 2.0 and 4 mm mesh-sizes). Species were separated and identified using the monograph of Cossmann & Pissarro (1904–1913) and by comparison with museum collections hosted at the University of Florence, following modern taxonomic updates (Merle 2008). All recognizable mollusc specimens in the fractions were counted and combined for the purpose of this study. Raw abundances per sample, for a total of 37 719 specimens and 499 species distributed in 12 samples, and data on locality and stratigraphic interval served as the basis for a Q-mode multivariate analysis. Cluster analyses (Ward's method) and ordination techniques (non-metric dimensional scaling, NMDS) and analysis of similarity (ANOSIM) were applied to a matrix of square-root transformed per cent abundance data. The similarity matrix for NMDS and ANOSIM is based on the Bray–Curtis Index. NMDS results are based on a 2D plot (stress = 0.1553), with Axis 1 explaining 47.42% and Axis 2 20.37% of the variance. In the 3D plot, stress lowers to 0.115, with Axis 3 explaining only 7.52% of variance. In particular, ANOSIM provides a way to test statistically whether there is a significant difference between two or more groups of sampling units, based on stratigraphy or geography. The R-values of the ANOSIM, which can vary between zero (no difference between samples) and unity, were used to evaluate relations between samples. Once groupings among samples, recognized by cluster analysis and NMDS, were established, the fossil fauna was used to explore the nature of co-occurrences by way of an R-mode analysis, carried out on a simplified version of the dataset. This was obtained by including only those species that contributed at least 2% of the total abundance of at least one sample, and the species that co-occurred in at least four samples, no matter what their frequency. By doing so, the species list was downgraded to less than one-fifth of the original list, and the total abundance decreased to just two-thirds (S = 91, n = 23 592). Results were expressed as frequency of a given species in the original dataset. Per-sample, species-level diversity was measured on the original dataset (S = 499, n = 37 719) through rarefied species richness and evenness. Richness was calculated with 95% confidence intervals at two values: n = 79, corresponding to the minimum number of individuals (in sample PB3), and n = 623, corresponding to the second lowest number of individuals (sample 94-6, from the Red Sea). Evenness was measured through the Shannon–Wiener index. These values were compared with those of modern analogues from the warm-temperate Adriatic Sea (Sawyer & Zuschin 2010) and the subtropical Red Sea (Zuschin & Hohenegger 1998), which are taxonomically and taphonomically comparable (i.e. the fauna consists mostly of molluscs) and were sieved on 1 mm mesh-size. Mean values for the Red Sea, Adriatic Sea and Paris Basin were measured for both values of richness and evenness. This allowed us to evaluate the climate affiliation of Paris Basin shallow marine biota, which during the middle Eocene inhabited latitudes today associated with cool-temperate waters. All statistical analyses were carried out with the software package Primer 6.0 (Clarke & Warwick 2001) and PAST 3.10 (Hammer et al. 2001)
Results
Multivariate statistics
In the Q-mode cluster analysis, two groups are recognized at a similarity level of 15% (Fig. 3a). One larger group comprises all middle Lutetian samples (cluster b–e), excluding PB14, the youngest Lutetian sample in our analysis, which grouped with the lower Bartonian sample PBG (cluster a). The second sample from the lower Bartonian, PBN7, stands as an outlier. Within the middle Lutetian cluster, three samples from EDS 4 at Grignon form a smaller and tighter subset (cluster c: similarity above 40%). The other three clusters within the large Lutetian group consist of two samples each, which are always from different EDSs and different localities (clusters c–e).
Multivariate analysis. (a) Cluster analysis. (b) Plot of NMDS; samples grouped by clusters shown in (a). (c) Plot of NMDS; samples grouped by locality. (d) Plot of NMDS; samples grouped by chronostratigraphic unit (EDS).
The dataset includes four localities and five EDSs of chronostratigraphic significance, four from the middle Lutetian, at c. 42 – 48 Ma, and one from the lower Bartonian, at 39–40 Ma. The three samples from Fleury-la-Rivière, all collected within a 5 m thick sandstone succession and located as much as 160 km from the other section, are scattered in three groups. This suggests that geographical position played a minor role in structuring the distribution of the Paris Basin mollusc biota. Similarly, the two samples collected at Ferme-de-l'Orme, a section also only a few metres thick, are also widely separated in the sample ordination plot (Fig. 3c). The Q-mode clusters make more sense when interpreted in terms of chronostratigraphic units. In the NMDS ordination, all EDS 4 samples, although from two different localities, plot close together (Fig. 3d), with no overlap with those from other stratigraphic units. EDS 2 and EDS 3 can be similarly separated. Displaced from these are the sample from EDS 5 and the lower Bartonian samples. PB14, from the middle Lutetian of Ferme-de-l'Orme, plots relatively far from the other middle Lutetian samples, and has affinities with the lower Bartonian sample PBG as found in the cluster analysis (Fig. 3a). This result is mirrored by the position of PB14 and PBG at high values along Axis 1 of the NMDS plot (Fig. 3d). The ANOSIM shows a lower Global R when the samples are grouped by locality (Global R = 0.626; Table 1), than by EDS (Global R = 0.733; Table 2).
Results of analysis of similarity (ANOSIM), factor ‘locality’
Results of analysis of similarity (ANOSIM), factor ‘EDS’
The R-mode cluster analysis highlighted eight main groups of species contributing to the similarities of samples within each EDS (Fig. 4, clusters labelled A–H). Group A is the largest and comprises 41 species, which preferentially occur in samples collected from EDS 4 at Grignon and Ferme-de-l'Orme. The bivalve Paraglans calcitrapoides is abundant and, together with the gastropods Vermicularia conica and Collonia striatus, ubiquitous and exclusive of this interval. Cluster D also mainly contributed to the grouping of samples from EDS 4. Some of the species of clusters A and D, although not ubiquitous, were found only in samples from this chronostratigraphic interval, including Striarca decipiens, S. quadrilatera, Barbatia barbatula, Cyclocardia squamosa, Plagiocardium granulosum, Emarginula costata, Lapparentia fischeri and, in cluster D, Hipponix cornucopiae, Sigmesalia trochoides and Siphonaliopsis minuta. Species groups A and D, even when broken down to smaller clusters, are composed of a mixture of trophic groups. These comprise many suspension-feeding, shallow infaunal bivalves (Carditidae, Cardiidae, Veneridae, etc.), many infaunal and epifaunal carnivore gastropods (Natica, Vexillum, Crassispira, etc.), some microherbivores (Tricolia, Pusillina) and parasites (Tenuiscala). Cemented forms are rarer than in other clusters. Among the species of cluster D, Eopleurotoma plicaria, Natica epiglottina and Haustator mitis are particularly characteristic of sample PB3 and also occur in sample PB16, from EDS 2; this explains in part the grouping of these two samples in Q-mode analyses (Fig. 3).
R-mode cluster dendrogram (analysis conducted on a subset of the original dataset) compared with sample composition, in stratigraphic order, with frequencies expressed semiquantitatively.
Cluster B comprises species occurring mostly in a sample from EDS 3 (PB9, from Fleury-la-Rivière), such as suspension-feeding (Haustator imbricatarius) and carnivorous gastropods (Olivancillaria mitreola). Glycymeris pulvinata (a suspension feeder) and Fustiaria circinata (a deposit feeder) are ubiquitous in older Lutetian samples, disappearing up-section (EDS 5, Bartonian). Overall, the suspension-feeding bivalves form the best represented trophic group.
Clusters C and F contain species particularly abundant in EDS 2, none of them being exclusive to this interval. Species belong to a variety of trophic niches, with abundant microherbivorous gastropods Bittium semigranosum and B. transenna and many suspension-feeding bivalves and gastropods (e.g. Cardiidae, Veneridae, Corbulidae, Turritellidae). Carnivores are less diversified or abundant than elsewhere.
Clusters E and H contain species abundant in lower Bartonian shell beds, the first with species of sample PBN7, the second with species of PBG. Cluster H, characterized by the herbivorous gastropods Potamides lapidorum and an unidentified rissoid, is exclusive of younger samples (comprising EDS 5). Cluster E, in turn, contains shallow infaunal suspension-feeding bivalves such as Callocardia nitidula and Bicorbula gallica, and chemosymbiotic lucinids (Parvilucina pusilla) that range from EDS 2 to the lower Bartonian. Cubitostrea cubitus (cluster E) is found in both lower Bartonian samples, whereas it is absent in the Lutetian. Finally, cluster G groups species characteristic of sample PB14 from EDS 5 at Ferme. Here, characteristic species point to a vegetated bottom, with abundant herbivorous gastropods Bittium duchasteli and Tympanotonos semicoronatus. The chemosymbiotic lucinid bivalve Parvilucina turgidula from cluster G points to organic-rich bottoms with local sulphuric conditions, possibly associated with the presence of seagrass.
A few species were found throughout the study interval, but the majority are either middle Lutetian or lower Bartonian. Potamides lapidorum is abundant and exclusive of EDS 5 and lower Bartonian samples, whereas no species in the restricted dataset (S = 91) is exclusive of the lower part of the middle Lutetian (EDS 2 or EDS 3).
Richness and evenness comparisons
Owing to differences in taxonomic and taphonomic conditions and sampling design, there are not many modern examples that can be compared with our fossil dataset. The two case histories that we have selected are taxonomically and taphonomically comparable with the Paris Basin study because they both consider the whole benthic molluscan shelly fauna, including aragonitic shells, which is well preserved in the Paris Basin collections. Sampling designs are also comparable, with bulk samples sieved through the same sieve sizes across the three studies that are being compared. The two modern settings examined, the Red Sea and the Adriatic, like the Paris Basin are from vast shallow seas enclosed by land masses. The wide range of habitats included in both modern settings allows for a meaningful comparison with the middle Eocene subtidal fossil assemblage. In the end, we expect differences in composition, evenness and abundance of single samples to be under the control of climate, with the Red Sea and the Adriatic serving as templates for subtropical and temperate settings, respectively. To facilitate comparison, the information on bottom facies and water depth for all samples is provided on a per-sample basis (Table 3).
Comparison of rarefied species richness and evenness of molluscan faunas from samples collected in the modern Red Sea (Zuschin & Hohenegger 1998), the modern Adriatic Sea (Sawyer & Zuschin 2010) and in the middle Eocene of the Paris Basin (this study)
The analysis confirms the uniqueness of samples from EDS 4: this unit has the highest values of rarified species richness and evenness among all samples, most notably PB1 and PB10, collected in calcareous sandstones at Grignon and Ferme-de-l'Orme, respectively. The lowest evenness is present in lower Bartonian samples at La Guepelle (PBN7 and PBG) and the upper part of the middle Lutetian at Ferme-de-l'Orme (PB14). Another sample with low evenness is PB16, from the upper part of EDS 2 at Ferme. Our diversity study therefore confirms a stratigraphic trend already indicated by multivariate analysis: low or intermediate evenness in samples from EDS 2 and EDS 3, very high values in EDS 4, and a drop of evenness in EDS 5 and in the lower Bartonian. When Paris Basin rarefied data are compared with their modern analogues, we find that values encountered in EDS 4 are similar to those measured in assemblages from modern coral sand and reef slope habitats; that is, possibly the highest diversities encountered in the tropical Red Sea (Table 3; Zuschin & Hohenegger 1998). In contrast, in EDS 5 and the lower Bartonian shell beds the palaeocommunities show the high dominance typical of intertidal assemblages from the modern warm-temperate Adriatic Sea (Table 3; Sawyer & Zuschin 2010).
Middle Eocene molluscan palaeoecology
The quantitative analysis of middle Eocene Paris Basin shell beds has highlighted a change in the composition of mollusc palaeocommunities that parallels and expands knowledge based on lithology and stratigraphy. Evenness and complexity in trophic relationships generally increase from sandstones of EDS 2–3 to calcareous sandstones of EDS 4, dropping to a minimum in calcareous sandstones of EDS 5 and lower Bartonian sandstones. This change is also reflected in the relative position of samples in ordination space (Fig. 3), although relationships here are more complex to interpret and a simple gradient cannot be extrapolated. The following discussion highlights three possible end-members of a continuum, where factors such as nutrient level, water depth and trophic interaction with plants contribute to shaping macrofaunal benthic communities.
Mesotrophic sandy bottoms
The lowermost sample PB16, from the middle Lutetian (EDS 2), and PBN7, from the lower Bartonian, show low evenness (Table 3), but they occupy different positions in the NMDS plot (Fig. 3). Dominant species of either assemblage, however, belong to the suspension feeders. These include a turritelline gastropod, Haustator mitis (32% of frequency in PB16), and two infaunal venerid bivalves, Calloocardia nitidula and Callista elegans (respectively 32% and 23% in PBN7). Suspension feeders can act as opportunist species within shallow marine benthic communities, taking advantage of high quantities of seston and capable of facing rapid changes of environmental conditions. Accordingly, high dominance of turritelline gastropods is frequently associated with deltaic and upwelling zones, where seasonally driven nutrients abound (Allmon 1976). As upwelling is unlikely, given the palaeogeographical setting of the Paris Basin, some form of deltaic influence can be hypothesized for the PB16 palaeocommunity. In contrast, rapid burrowers such as the dominant venerid bivalves in PBN7 are abundant in high-energy shallow subtidal bottoms, where they can withstand the vagaries of a rapidly shifting sandy substrate and take advantage of the abundant suspended organic matter (Stanley 1977). The separation between PB16 and PBN7 in the ordination plot can thus be explained by two different, but similarly unpredictable subtidal environments. An important species-level turnover has, however, taken place during the several million years that separate the two. This calls for also considering evolutionary change in explaining the distance of PBN7 from all other samples. This is not the case for PB3, also from the middle Lutetian (EDS 4), which shares with PB16 abundant Haustator mitis (eight individuals, 32% in PB16), explaining their proximity in ordination space. Rarefied species richness and evenness of this sample are relatively high, but have to be interpreted with caution because of the small size of the sample (n = 79).
Oligotrophic sandy bottoms
All other middle Lutetian samples are plotted in the middle part of the NMDS. Samples from EDS4, including PB1, PB4 and PB6 from Grignon and PB10 from Ferme-de-l'Orme, have the highest evenness, comparable with that of open marine coral sands of subtropical settings such as the Red Sea (Table 3). Similarly, we can assume that the upper half of EDS 4 was deposited under oligotrophic conditions in an ecosystem characterized by narrow niches. Accordingly, the palaeocommunity is evenly composed of herbivores (Emarginula costata, Collonia stratus, Ptychocerithium inabsolutum, Vicinocerithium sp., Tricolia sp., Pusillina nana, Lepidochitona grignonensis), carnivores (Tenuiscala cloezi, Payraudeautia perforata, Lapparentia fischeri, Volvarinella spp., Eopleurotoma plicaria, Cryptoconus spp.), some deposit feeders (Rimella fissurella), and suspension feeders. The last are in turn evenly represented by byssate epifauna (Barbatia barbatula, B. angusta, Striarca decipiens), cemented epifauna (Vermicularia conica, Vermetia sp.) and infauna (Paraglans calcitrapoides, Crassatina triangularis, Varicorbula minuta, Plagiocardium granulosum, etc.). These characters suggest a subtidal setting deeper and furthest from terrestrial inputs with respect to all other palaeoenvironments considered.
Among samples from EDS 4, PB10 shows many similarities to PB22 from EDS 3, including the abundance of the herbivore Bittium transenna, the suspension-feeding Turritellidae (Sigmesalia intermedia and Haustator funiculosus) and the Veneridae Callista elegans and Cyclocardia serrulata. Because the last two are abundant also in PBN7 from high-energy subtidal bottoms, and because the Turritellidae more often prefer high-nutrient settings, we infer that Ferme-de-l'Orme and Fleury-la-Rivière, where PB10 and PB22 were collected, respectively, were shallower and closer to terrestrial inputs with respect to Grignon. The last two remaining assemblages to consider, PB27 and PB9 from EDS 2 and 3, respectively, can also be interpreted as representing more unpredictable palaeoenvironments than those encountered at Grignon within EDS 4. Evenness is not as high as in the latter, and opportunistic suspension feeders often abound at the expense of other trophic groups; this is the case with Haustator funiculosus (23% in PB27), Cubitostrea plicata (21% in PB9, 17% in PB27) and Glycymeris pulvinatus (20% in PB9, 2% in PB27).
Analogies of Paris Basin rarefied diversities with those measured in modern shallow marine mollusc communities in the Red Sea strengthen the above palaeoecological interpretation. Modern coral sands are good analogues of samples PB1, PB4, PB6 and PB10 (EDS 4, excluding PB3) because they have even proportions of the main trophic groups (herbivores, suspension feeders and carnivores), and because opportunistic suspension feeders such as Turritella or Corbula are missing (Zuschin & Hohenegger 1998). The lack of chemosymbiotic bivalves in EDS 4 palaeocommunities, although they are diversified and abundant in Red Sea subtidal coral sands, can be explained by the lack of a seagrass cover in Paris Basin bottoms during this part of the middle Lutetian.
Mangroves and seagrasses
Younger samples PBG and PB14 occupy the lower right quarter of the ordination, forming a loose cluster (Fig. 3a). Their trophic composition justifies, for both, the interpretation of palaeocommunities from a very shallow subtidal sandy bottom, with very high organic content from the decomposition of vegetal matter. The simplest palaeocommunity is represented by lower Bartonian sample PBG with a high dominance of herbivorous gastropods of the families Potamididae (Potamides lapidorum, Tympanotonos semicoronatus), Cerithiidae (Ptychocerithium lamellosum) and Rissoidae (Pseudotaphrus buccinalis and an undetermined species), followed by the carnivores (Amauropsina canaliculata, a naticid) and some suspension feeders, either byssate- (Trinacria media) or cemented-epifaunal (Cubitostrea cubitus), with only rare infauna (Cyclocardia pulchra).
Next in composition comes sample PB14 from the upper part of the middle Lutetian (EDS 5). It shares with PBG the dominance of cerithioidean herbivorous gastropods, mostly belonging to the same species (Potamides lapidorum, Tympanotonos semicoronatus, Ptychocerithium lamellosum), but also the cerithiids Bittium duchasteli and B. semigranosum. Rissoidae, however, are absent; their ecological role is played in part by a Pseudomelaniidae (Bayania sp.), and trophic complexity and evenness are considerably higher than in PBG. In particular, the suspension feeders are much more diverse and equally distributed within the epifauna (e.g. Trinacria deltoidea, Striarca quadrilatera, Barbatia angusta) and the infauna (e.g. Cyclocardia serrulata, Gari dutemplei, Venerella secunda, Katelysia texta). Alongside the suspension feeders, the infauna is occupied by abundant chemyosymbiotic lucinid bivalves (Parvilucina turgidula), indicative of a substratum rich in organic content. Carnivores are represented by rare Oliviidae (Olivancillaria mitreola) and Marginellidae (Volvarinella eburnea).
Modern Potamididae show parallel distribution with mangrove trees along the tropical intertidal belt (for the relations between Eocene and modern potamidids, see Reid et al. 2008; Dominici & Kowalke 2014), but the presence of Rissoidae (PBG) and miliolid foraminifera (PB14), together with lucinid bivalves, points also to the presence of seagrass. The comparison of rarefied richness and evenness for these two samples with those from the Red Sea points to a species-abundance distribution similar to that of a modern mangrove environment (Table 3). We therefore suggest a very shallow subtidal environment, closely associated with a mangrove ecosystem, possibly covered with sparse seagrass.
Discussion
Several aspects complicate a sequence stratigraphic interpretation: the thinness of the sections (5–15 m) and the slow net sedimentation rates they imply, the pervasive bioturbation that hampers the development of a facies model, the paucity of outcrops and the orientation of the section along depositional strike. However, available data on the sequence architecture of low-accommodation Cenozoic siliciclastic successions allow our sequence stratigraphic interpretation to be considered as a viable working hypothesis. In fact, data-rich studies from similar closed marine embayments developed along passive continental margins allow for the recognition of similarities. A very good analogue is the middle Eocene Clairborne Group, in the onshore US part of the Gulf of Mexico Basin (Hackley 2012). Claiborne formations contemporaneous with those here under study are the Welch Formation (middle Lutetian), the Sparta Sand and the Cook Mountain Formation (lower Bartonian). Similarly to their Paris Basin analogues, they represent shallow marine sediments and are only a very few metres thick, glauconitic, richly fossiliferous and laterally persistent for tens of kilometres (Stenzel 1940; Gimbrede 1962; Yancey & Davidoff 1994), implying low sedimentation rates and an even topography. A younger analogue is found in the Miocene of Maryland, further north along the US east coast. This is characterized by laterally continuous and rich shell beds cropping out for tens of kilometres in a strike direction, with unconformity-bounded regressive–transgressive cycles stacked to form a retrogradation succession lasting more than 10 myr, but only about 30 m thick (Kidwell et al. 2015).
The sedimentary facies analysis and palaeoecology of middle Eocene Paris Basin shell beds allow an interpretation of the nature of the shallow marine benthic ecosystem and its abiotic factors during a peculiar time in the Cenozoic. The ‘doubthouse’ comprised a period between the Early Eocene ‘greenhouse’ climate, with the highest temperatures of the Cenozoic, and the Oligocene ‘icehouse’, featuring a sharp decrease in global temperatures and the formation of the Antarctic ice sheet. This period was characterized by a global temperature decrease interrupted by a temporary amelioration during the lower Bartonian (the Middle Eocene Climatic Optimum; MECO) lasting about 500 kyr (Bohaty & Zachos 2003). Palaeoenvironmental change during the middle Lutetian has been recently evaluated by geochemical and palaeontological studies conducted on the Grignon succession, which are directly comparable with our approach (Fig. 5). The isotopic study on mollusc shells (Huyghe et al. 2012a, 2015) and the study on ostracod assemblages (Guernet et al. 2012) suggest that maximum depths were reached in the middle part of their parasequence A6, corresponding to the lower part of EDS 2, in our study represented by sample PB27 from Fleury-la-Rivière (Figs 1 and 5). In our interpretation, however, PB27 is representative of a shallowing with respect to underlying sediments (Glauconie grossière Fm: Huyghe et al. 2015; EDS 1) because of its palaeoecological signal (i.e. relatively low evenness and high abundance of opportunistic suspension feeders). Sedimentary facies analysis and the palaeoecology of mollusc shell beds suggest instead that the middle Lutetian (fourth-order) maximum flooding surface (MFS) coincided with the lower part of shelly calcareous sandstones in the upper part of the section, at the MFS (fifth-order) of EDS 4 (Fig. 5). This interval is also marked by the richest ostracod assemblage (Guernet et al. 2012). This fourth-order MFS would thus pinpoint the beginning of the sea-level highstand recognized in global sea-level curves at around 45.0 Ma (Kominz et al. 2008). The sequence boundary separating EDS 4 and EDS 5 would therefore mark the long and stepwise sea-level drop occurring between about 44 and 43 Ma. This resulted in a major change in shell bed composition recorded during the ensuing transgression, which is correlated with the global rise occurring at about 42 Ma (Kominz et al. 2008), when only very shallow subtidal depths were attained (sample PB14, collected at Ferme-de-l'Orme). A short-lived sea-level fluctuation is recorded at Ferme-de-l'Orme at the base of a thin muddy interval (possibly deposited in an intertidal environment), overlain by upper shoreface deposits dominated in the field by the turritellid Sigmesalia (nutrient-rich shoreface environment; Fig. 1). The sea level then fell again, in coincidence with the pre-MECO climatic cooling. During this time, the third-order sequence boundary was formed, separating the Lutetian from the Bartonian (Huyghe et al. 2015). The La Guepelle succession can be correlated with the general transgression occurring during the MECO, centred at around 40.0 Ma (base of Chron C18n.2n: Huyghe et al. 2015). Similarly to the transgressive systems tract of EDS 5 and EDS 6, depths did not attain the same magnitude as in the middle Lutetian. Cooling during the later part of the middle Lutetian, and the further temperature drop preceding the MECO (Bohaty & Zachos 2003), may explain the species-level turnover registered at the stage level (Cossmann & Pissarro 1904–1913). Our sample-level data, however, point to a facies shift at the sequence boundary between EDS 4 and EDS 5, which controls mollusc diversity. We therefore suggest that sequence stratigraphical architecture controls patterns of faunal change at this regional scale, as has been shown for several other regions and time intervals (e.g. Brett 1998; Patzkowsky & Holland 1999; Smith et al. 2001; Zuschin et al. 2011).
Middle Lutetian stratigraphic units recognized in this study, compared with units described by previous researchers. All middle Lutetian samples used in our palaeoecological analysis were projected onto the Grignon succession, taken as a template of this time interval.
The middle Eocene history of mollusc benthic communities recorded in the Paris Basin is just one step in the long and complex march leading from early Eocene greenhouse shallow marine ecosystems to the modern fauna (Lozouet 2014). In this time interval, the macroevolutionary turnover was particularly severe at intertidal and shallow subtidal depths (Tomasových et al. 2014). Middle Lutetian communities, however, are similar to modern tropical assemblages such as those living in coral sands and seagrass bottoms of the Red Sea, sharing high evenness and complex trophic relationships (Table 3; Zuschin & Hohenegger 1998), and unlike warm-temperate faunas of the Adriatic Sea, which are dominated by fewer species at all depths (Table 3; Sawyer & Zuschin 2010). This evidence confirms previous suggestions (e.g. Huyghe et al. 2012a) that, even during the middle Eocene cooling, subtropical conditions of the Tethyan realm (Lozouet 2014) reached as far north as the Paris Basin. The same situation was not found in the later part of the middle Lutetian, nor during the MECO, notwithstanding an important global climatic amelioration. Before concluding that shell beds of EDS 4 record the last evidence of tropical affinity among mid-latitude shallow marine communities in Europe, younger assemblages from a comparable facies must be sought.
Conclusions
Important middle Eocene successions of the Paris Basin were studied, starting from a re-evaluation of sedimentary facies and through new bulk-sampling in mollusc shells beds. The study interval comprises the middle Lutetian and the lower Bartonian, which are separated by a major depositional sequence boundary on a regional scale. Our study reached the following conclusions.
Both intervals can be subdivided into small-scale unconformity-bounded units (i.e. metre-scale elementary depositional sequences; EDSs), interpreted as the product of high-frequency (fifth-order) cycles of sea-level change. We identified five EDSs in the middle Lutetian and two in the lower Bartonian. Sequence boundaries were recognized at the base of sandstones with fine gravels or cross-bedding, within successions dominated by bioturbated fine-grained shelly sandstone and shelly calcareous sandstone.
The species abundance distribution of molluscan assemblages from 12 bulk samples of major shell beds, used for multivariate analysis and diversity analysis, helped reconstruct the composition and structure of the benthic communities living at this peculiar time of Palaeogene climate transition. Palaeocommunities from middle Lutetian EDS 1–4 are marked by high evenness and trophic complexity, which peaked in EDS 4. These indicators dropped in EDS 5, possibly in EDS 6 (not sampled), and in lower Bartonian shell beds.
When compared with modern assemblages, middle Lutetian communities strongly resemble mollusc species from coral sands at subtropical latitudes, but less so warm-temperate shallow marine communities. This is counterintuitive if Paris Basin palaeolatitudes are considered. It does, however, agree with previous knowledge on stage-level species richness, calculated after more than two centuries of research on these successions. We therefore confirm that the middle Lutetian Paris Basin constitutes the remnant of a former early Eocene ‘greenhouse’ world, which disappeared as the icehouse world of the Eocene–Oligocene boundary was approached.
The steps with which the climatic transition occurred cannot be understood unless the appropriate facies is sampled in younger sediments. This was not possible at the study sites because EDS 5, EDS 6 and Bartonian assemblages are from shallower water depths. The vertical stacking of Paris Basin palaeocommunities suggests that the peak of tropical diversity of EDS 4 coincides with a fourth-order sea-level highstand recorded on a worldwide basis.
Acknowledgements and Funding
The following people ensured the completion of this paper: D. Merle, B. Pattedoie, P. Legrand, A. Tomasových and M. Stachowitsch. We are grateful to S. Holland for a thorough review of the paper. This work was supported by the Austrian Science Fund (FWF) project P19013-Bio and MIUR/PRIN 2010–2011 (‘Past Excess CO2 worlds: biota responses to extreme warmth and ocean acidification’).
- © 2016 The Author(s)
References
Palaeocommunities, diversity and sea-level change from middle Eocene shell beds of the Paris Basin
