Early Eocene deep-sea benthic foraminiferal faunas: Recovery from the Paleocene Eocene Thermal Maximum extinction in a greenhouse world

The early Eocene greenhouse world was marked by multiple transient hyperthermal events. The most extreme was the Paleocene-Eocene Thermal Maximum (PETM, ~56 Ma), linked to the extinction of the globally recognised deep-sea benthic foraminiferal Velasco fauna, which led to the development of early Eocene assemblages. This turnover has been studied at high resolution, but faunal development into the later early Eocene is poorly documented. There is no widely accepted early Eocene equivalent of the Late Cretaceous-Paleocene Velasco fauna, mainly due to the use of different taxonomic concepts. We compiled Ypresian benthic foraminiferal data from 17 middle bathyal-lower abyssal ocean drilling sites in the Pacific, Atlantic and Indian Oceans, in order to characterise early Eocene deep-sea faunas by comparing assemblages across space, paleodepth and time. Nuttallides truempyi, Oridorsalis umbonatus, Bulimina trinitatensis, the Bulimina simplex group, the Anomalinoides spissiformis group, pleurostomellids, uniserial lagenids, stilostomellids and lenticulinids were ubiquitous during the early Eocene (lower-middle Ypresian). Aragonia aragonensis, the Globocassidulina subglobosa group, the Cibicidoides eocaenus group and polymorphinids became ubiquitous during the middle Ypresian. The most abundant early Ypresian taxa were tolerant to stressed or disturbed environments, either by opportunistic behavior (Quadrimorphina profunda, Tappanina selmensis, Siphogenerinoides brevispinosa) and/or the ability to calcify in carbonate-corrosive waters (N. truempyi). Nuttallides truempyi, T. selmensis and other buliminids (Bolivinoides cf. decoratus group, Bulimina virginiana) were markedly abundant during the middle Ypresian. Contrary to the long-lived, highly diverse and equitable Velasco fauna, common and abundant taxa reflect highly perturbed assemblages through the earliest Ypresian, with lower diversity and equitability following the PETM extinction. In contrast, the middle Ypresian assemblages may indicate a recovering fauna, though to some extent persistently disturbed by the lower-amplitude Eocene hyperthermals (e.g., Eocene Thermal Maximum 2 and 3). We propose the name ‘Walvis Ridge fauna’ for future reference to these Ypresian deep-sea benthic foraminiferal assemblages.


Introduction
Benthic foraminifera may constitute more than 50% of the total eukaryotic biomass at oceanic depths >1000 m [1][2][3], and are the biota most commonly represented in the microfossil record of the deep-sea floor, the largest habitat on Earth (e.g., [4]). Therefore, faunal and trace element/isotopic studies of deep-sea benthic foraminifera have been used extensively to infer past environmental conditions. These studies require detailed knowledge of benthic foraminiferal ecology and correct identification of morphological species to be chemically analysed, because many species show significant offsets in stable isotope and trace elemental signatures (e.g., [5,6]). However, the taxonomy of benthic foraminifera is problematic, since splitting of species is common, providing different names for morphologically indistinguishable species, especially for different geographic regions and for different age intervals. Compilations of Early Cretaceous-middle Miocene [7], Late Cretaceous-Cenozoic [7,8], Cenozoic [9], Late Cretaceous-Paleocene (K-Pg; [10]), and Paleocene-Eocene [11] faunas have been published, and benthic foraminiferal deep-sea biozones have been proposed for the Cenozoic [12][13][14][15][16]. In order to standardise descriptions of benthic foraminifera, Holbourn et al. [17] created a database of 300 deep-water species, including mostly taxa with stratigraphic and paleoecological significance, or those used in geochemical analyses. However, no detailed taxonomical global analysis similar to that for the Cretaceous-Paleocene Velasco fauna, as first described by Cushman [18,19] and White [20][21][22], updated by Schnitker [23], Tjalsma and Lohmann [11], and Alegret and Thomas [10], has been performed for the lower Eocene. Müller-Merz and Oberhänsli [24] studied Eocene deep-sea benthic foraminifera, but concentrated on the South Atlantic region only.
Despite the detailed faunal analyses across the Paleocene-Eocene transition, most studies do not follow the assemblages later into the early Eocene, and there is no consensus on the taxonomy of the most common early Eocene taxa. It has been suggested that early Eocene assemblages were less globally homogeneous than Paleocene assemblages [31], but this has not been well documented. We present a taxonomic compilation and partial revision of early Eocene deep-sea benthic foraminifera. This study covers almost the whole Ypresian (calcareous nannofossil and planktic foraminiferal biozones NP9b-NP12, CP8b-CP10, P5-P7, E1-E5 [64][65][66][67][68]), identifies the most common taxa and groups of taxa, and determines their paleogeographical and bathymetric distribution. The results of this compilation will improve current knowledge of early Eocene benthic foraminifera, as this is the first attempt to reconcile taxonomic assignments used by diferent authors and for different time intervals (e.g., Late Cretaceous, Early Cenozoic) globally. In addition, this study aims at evaluating variability in diversity, evolution of early Eocene faunas including extinctions and originations, and ecological preferences of Eocene species.

Data acquirement
We compared benthic foraminiferal assemblages at seventeen sites drilled by the Ocean Drilling Program (ODP) and the Deep Sea Drilling Project (DSDP). The sites have a broad global distribution, covering the Atlantic, Pacific and Indian Oceans (Fig 1; Table 1), allowing the assessment of geographic and bathymetric variations. We confined this study to open ocean drilling sites.
Most data have been previously published (Table 1), except those from ODP Sites 1258 and 1051 (S1 Table). PETM data from ODP Sites 1262 and 1263 have been reported partially in Foster et al. [88], data across ETM2 from those sites have been published by Jennions et al. [62], and across ETM3 preliminarily in Röhl et al. [87]. Complete data from ODP Sites 1262 and 1263 are included (S2 Table). D'haenens et al. [60] published benthic foraminiferal data across ETM2 from DSDP Site 401, but a taxonomically updated version of those data, including 18 additional samples from the most basal Eocene, is included here (S3 Table). All these samples were provided by the International Ocean Discovery Program (IODP) for the specific purpose of studying benthic foraminifera. No further permits were required for the described study. Sites where benthic foraminifera are recorded across both Ypresian intervals (lower and middle) are indicated with stars, those encompassing only the lower Ypresian are marked with triangles, and those covering exclusively the middle Ypresian are shown with squares. Color reference: green-Pacific sites, blue-Atlantic sites, red-Indian site. Paleogeographic reconstruction of~53 Ma from Hay et al., [69]. repository information (e.g., geographic location) and detailed list of the repository numbers are included in S1-S3 Tables. All these samples are accessible in the corresponding repositories.
The relative abundance of species (percentage) in the literature is typically presented by sample, but in order to estimate the abundance of each species across the two study intervals (lower and middle Ypresian) by site, the percentage of each species relative to the total number of specimens counted per site was calculated. When the number of specimens counted was not available from the published data (ODP Sites 1209, 1210, 1211 and 1212), we calculated the average percentage of every species by site, and then rescaled these averages to obtain an actual representative percentage (i.e., the fraction that such species represents within the total assemblage). We thus focused on the total relative abundance of each species across each study interval by site (S4-S6 Tables). These numbers depend upon the number of samples studied per time slice, but we consider them as representative of observations at each site.
In order to compare benthic foraminiferal assemblages from different sites as documented by different authors, it was necessary to standardise the data, because different authors use different taxonomic concepts, a major problem in compiling taxonomic data (e.g., [17]). After establishing informal morphological supra-generic groups to simplify the taxonomy (Table 2; S4 and S5 Tables), we performed a careful bibliographic revision and examined type specimens at the American Museum of Natural History, New York (AMNH) and at the Smithsonian National Museum of Natural History, Washington DC (NMNH). We then grouped species with similar morphologies and/or specimens that may belong to the same morphological species, but have been identified under different names (Table 3).
2.1.1 Sample preparation and picked specimens. We report new taxonomic data across the lower Eocene from ODP Sites 1051, 1258, 1262, 1263 and DSDP Site 401 (S1-S3 Tables). The very low numbers of benthic foraminifera at Sites 1051 and 1258 did not allow us to perform a detailed quantitative analysis of the assemblages, but we selected eight (Hole 1051A) and nine samples (Hole 1258B) from these sites to represent faunal composition in the central and tropical western Atlantic, respectively.
Samples were cleaned by removing the edges of the core in contact with the liner to avoid contamination, soaked in water with detergent, washed over a 63 μm sieve and the residue was dried at~50˚C. Benthic foraminifera were picked from the >63 μm fraction. An average of 386 specimens per sample were picked for DSDP Site 401, whereas an average of 120 specimens per sample was picked from ODP Site 1051, and 70 specimens per sample from ODP Site 1258 (S1 Fig). The low abundance numbers of the two latter sites are mostly due to strong dilution of benthic foraminifera by other components in the sediment. They are well below thẽ 300 specimens that are usually picked for quantitative analyses (e.g., [30,94]), and necessary to fully represent diversity. However, since in this case we are considering the relative abundance across the entire site, we considered that these numbers may reflect the overall diversity and abundance at Sites 1051 and 1258. For most samples at Sites 1262 and 1263, several hundreds of specimens were obtained, with the exception of an interval of strong dissolution at the base of the PETM that did not contain any benthic foraminifera, not even agglutinated forms.
2.1.2 Paleobathymetry. We use bathymetric divisions as defined in Van Morkhoven et al. [9]: upper bathyal (200-600 m), middle bathyal (600-1000 m), lower bathyal (1000-2000 m), upper abyssal (2000-3000 m) and lower abyssal (>3000 m). The early Eocene paleodepth of the studied sites (with the exception of ODP Site 1258) has been previously documented (Table 1; Fig 3) [31,95], mainly based on benthic foraminiferal assemblages. The original depth assignments of the assemblages are generally based on backtracking information (e.g., [10]) and/or comparison of assemblages as they occur along depth transects (e.g. [11]). We inferred the paleodepth at Site 1258 based on benthic foraminiferal evidence, as compared to asemblages at the other sites. Benthic foraminiferal assemblages at ODP Site 1258 are dominated by calcareous taxa, suggesting deposition well above the calcite compensation depth. Most of the abundant and common species from this site have been described from lower bathyal to abyssal depths, and include Quadrimorphina profunda, Clinapertina subplanispira, C. inflata, Anomalinoides spissiformis, Nonion havanense, Nuttallides umbonifera and stilostomellid species [11,96]. Other common species such as Nuttallides truempyi, Aragonia aragonensis, Oridorsalis umbonatus, pleurostomellids and Bulimina trinitatensis have been reported across wider depth ranges, varying from bathyal or middle bathyal to abyssal (e.g., [9,11,96]). Some common species such as Globocassidulina subglobosa and Bulimina tuxpamensis have a reported upper depth limit at upper bathyal depths [9]. However, some species such as A. aragonensis, G. subglobosa and occasionally even N. truempyi have been reported in outer neritic parts of the Tethyan shelves during the early Eocene (e.g., [97,98]), and Osangularia plummerae, another common species, typically inhabits a wide range of depths, from lower sublitoral to upper bathyal depths [99]. Thus we infer a lower bathyal to abyssal paleodepth for Site 1258 as most probable during the Ypresian.
Almost all sites for which data are presented here range from lower bathyal to upper abyssal paleodepths, except for the shallower Indian ODP Site 752, and the deepest Atlantic ODP Site 1262 and DSDP Site 550 (Fig 3). The vertical range of paleodepths of Pacific sites is narrower than that of the Atlantic Ocean sites, with paleodepths for Pacific sites ranging from lower bathyal to upper abyssal (~1300-3000 m), whereas paleodepth of Atlantic sites ranges from lower bathyal to lower abyssal (~1100-3900 m).

Microhabitats.
The microhabitat distribution of benthic foraminifera was used to interpret the ecology of the assemblages following the TROX model (e.g., [100,101]), with the infaunal-epifaunal (or endobenthic-epibenthic) ratio used as a proxy for some combination of oxygen levels and trophic conditions at the seafloor. According to this model, a high relative abundance of infaunal taxa is indicative of a high food supply and/or low oxygen availability, and a low percentage of infaunal taxa is indicative of oligotrophic and well-oxygenated conditions at the seafloor. This model in its simplest form, however, may not be valid under all circumstances. For instance, calcareous infaunal taxa may be advantaged relative to epifaunal taxa during carbonate dissolution at the seafloor [88], or bottom currents may lead to trophic focusing, so that relatively high abundances of suspension feeding, infaunally living taxa occurred even at low primary productivity in the overlying waters [43].
Benthic foraminiferal taxa were assigned to microhabitats based on their test morphology: epifaunal species are characterised by milioline coiling, planoconvex and biconvex trochospiral tests, biconvex planispiral, rounded and flat tests, tubular, and leaf-like forms, and infaunal taxa mainly have cylindrical, spherical, ovoid, unilocular globose, rounded planispiral and streptospiral shapes, and also include taxa with a neck (e.g., [102][103][104]). However, the relation between morphology of the test and microhabitat is not straightforward. There are many exceptions among living taxa (e.g., [105]), so that assignments may be accurate only about 75% of the time [106]. Despite these complications, this index provides a basis for evaluating the relationship between benthic foraminifera and environmental conditions at the seafloor, as controled by primary productivity at the sea-surface through bentho-pelagic coupling [107], but influenced by variability in remineralization of organic matter during transport to the seafloor, as well as potential lateral influx of refractory organic matter (e.g., [108,109]).

Statistical analyses
Multivariate analyses, including cluster and non-metric multidimensional scaling (NMDS), have been performed to assess the early Eocene benthic foraminiferal assemblages, using PAST software v. 3.04 [110]. These techniques were applied in order to identify the main assemblages and the most typical benthic foraminifera across the Ypresian in distinct regions. Hierarchical cluster analysis is an explorative technique applied to identify groups of sites with similar benthic foraminiferal assemblages, using the unweighted pair group average algorithm (UPGMA) and the Pearson correlation as similarity coefficient. NMDS is an ordination method to visualise trends or groupings, and useful when the dataset shows null values, here represented by the absence of species. Considering that the values of the relative abundance (percentages) of the species may depend on the size and number of samples studied per site, we chose NMDS as an optimal technique, because it does not compare actual abundance percentages and transforms the distance measures into ranks, comparing these ranked distances with the rank of the Euclidean distances in the ordination plot [111]. The NMDS plots represent the data considering their rank order (based on abundances), but discarding absolute percentage values. The Euclidean distance was used as the similarity index because it is sensitive to high abundances.
Both analyses were carried out for the intervals corresponding to the lower and middle Ypresian. The lower Ypresian dataset includes 29 species groups from 14 sites (Table 4), and the middle Ypresian dataset consists of 42 species groups from 9 sites (Table 5). Tables 4 and 5 were constructed considering those taxa or groups of taxa that are common (!2 to <5%), abundant (!5 to <15%) or very abundant (!15%) in at least one site, but they do not include species identified in one publication only or by one author only, in order to avoid authorbiased results.

Taxonomic remarks
The taxonomy of some of the most common early Eocene species (Figs 4-6), which have mostly been included in distinct informal taxonomic groups, is shown below. Full names including authors are given in the taxonomic list (S1 Appendix).
3.1.1 Bolivinoides cf. decoratus group. This group includes species assigned to the genera Bolivinoides and Bolivina (Fig 5.4). Both genera are characterised by a biserial arrangement of   broad and low chambers, but they differ mostly by the outline of their test. Bolivina has an ovoid to triangular outline, whereas Bolivinoides is rhomboidal in outline. The Bolivinoides cf. decoratus group is based on the species originally identified as Bolivina decorata Jones, and includes specimens that were assigned to Bolivinoides decorata (at Site 865), Bolivinoides decoratus (at Sites 550, 690, 1051 and 1258) and Bolivinoides cf. decoratus (at Sites 689, 690, 1262 and 1263). According to the type description, B. decorata has an elongated test, broad at the apertural end and tapering to rounded point at the aboral extremity, and a surface ornamented with prominent, oblong tubercles arranged in oblique rows [112].

Bolivina huneri group.
As in the previous group, specimens assigned to the genera Bolivina and Bolivinoides are included in the B. huneri group (Fig 6.3). Its representative species, Bolivina huneri, was originally described by Howe [113], who described the small, elongated and only slightly compressed test, peripherically broadly rounded, with sides nearly parallel in the adult stage, and a wall ornamented with numerous, delicate and irregularly anastomosing costae, finely perforated.
D'haenens et al. [60] suggested the occurrence of intermediate forms between B. huneri and Bolivinoides crenulata. The latter species has a test about twice as long as broad, tapering with the greatest width near the apertural end, and is characterised by a wall ornamented by distinct, longitudinal ridges and a crenulated base of chambers, forming a pattern of irregularly rounded depressions. As far as we know, these transitional forms have been reported only at Site 401 during the early Eocene (biozones NP10-11) [60], and more studies are needed in order to determine whether they are rare and do not cover the full time range studied here, or if they have a more extended record. We have included in this group specimens named Bolivina huneri (at Site 550), Bolivinoides huneri (at Sites 1262 and 1263) and Bolivinoides crenulata/Bolivina huneri (at Site 401).
The species included in the B. huneri and B. cf. decoratus groups are somewhat similar, but can be distinguished based on the typical ornamentation of their tests. Although the ornamentation may be slightly blurred in some cases, B. decoratus does not show the pattern of irregularly rounded depressions as in B. crenulata, neither the irregularly anastomosing costae of B. huneri. For this reason, we allocated these species to different groups.

Fursenkoina fusiformis group.
This group includes species origninally recognised as Virgulina fusiformis by Cushman [114]. The type specimen is characterised by a small, somewhat compressed and fusiform test with the greatest width at about the middle portion of the test, and a bluntly pointed end. The arrangement of the inflated chambers in the type specimen is initially triserial, biserial in the later portion of the test. The sutures are very slightly depressed, and the aperture is elliptical, very small, at the upper end of the test.
The name of the genus Virgulina was deemed invalid because it was used to define a genus of a trematode worm, and Fursenkoina was proposed as the new name for this genus [115]. Species in this group have been assigned to the genus Stainforthia, possibly based on the arrangement of the earlier chambers. Fursenkoina is described as biserial throughout [116], but this is in disagreement with the original description of V. fusiformis. Stainforthia has Early Eocene deep-sea benthic foraminiferal faunas chambers that are triserially arranged in the early stage, at least in the microspheric generation, followed by a twisted biserial arrangement [116]. Specimens identified in the literature as Fursenkoina fusiformis and Stainforthia fusiformis were included in this group.

Bulimina elongata group.
This group includes species with a triserial, long and slender, smooth test, including Bulimina elongata and B. thanetensis (Fig 5.5). These two species mainly differ in the longitudinal twist of the test in B. thanetensis. Close examination of several paratypes of B. thanetensis (CC 58889, CC 58890 and CC 58891, NMNH), however, showed that this twist is not clearly present in some specimens, which thus can be easily confused with B. elongata. Bulimina thanetensis may have become extinct during the early Ypresian, slightly after the start of the PETM (e.g., [14,117]), whereas B. elongata continued into the early Eocene. Holbourn et al. [17], however, reported an extended stratigraphic range of B. thanetensis further into the Eocene.
The holotype of B. thanetensis (CC 35855, NMNH) is long and slender, with a markedly twisted test. The first chambers are very small, increasing slowly in size towards the middle part of the test, where they start to increase in size more rapidly, and are more inflated, thus the sutures are also more marked. The holotype slightly differs from deep-sea forms in its more rounded, squat chambers, and less-prominent sutures [17]. Typical long and slender tests of B. thanetensis have been found at Maud Rise (Site 690, e.g., [14,17]), whereas the specimens called B. cf. thanetensis in Egypt are stouter (e.g., [117,118]), and resemble those depicted by Tjalsma and Lohmann [11].
Neither the holotypes nor other primary types of B. elongata are deposited at the AMNH or NMNH, but we observed diverse Eocene specimens identified as B. elongata (CC 59037, CC 59030, CC 59031, CC 59026, CC 59036, CC 59028, CC 9545 and CC 59025, NMNH). We found a large variety in the width of the test, and in the length-width ratio. Only five specimens from slides CC 59031 and CC 59026 show truly long and slender tests with nearly parallel lateral sides, slightly inflated chambers and slightly depressed sutures. We also checked some hypotypes and plesiotypes from younger time intervals, but these specimens have a more ovate form and flush sutures, in contrast with the Eocene specimens. Nørvang [119] suggested that a trochospiral initial end of B. elongata occurs in the microspheric form, and thus these specimens do not correspond to the genera Caucasina as suggested by others [120].
Based on the specimen figured in D'haenens et al. [60], specimens here called Bulimina cf. simplex (originally named Bulimina thanetensis) at Site 401 (S3 Table) were included into this group because they have slightly parallel lateral sides (as in B. elongata), but do not have a truly long test. However, they do not have the bulging and protruding chambers which are typical for B. simplex (see B. simplex group).
3.1.5 Bulimina simplex group. This group includes buliminids characterised by a very simple, smooth and straight test, rounded in cross section (e.g., Bulimina simplex, B. bradburyi and B. tuxpamensis), and without ornamentation or other specific features (Fig 4.14 and 4.15). Early Eocene deep-sea benthic foraminiferal faunas These species are commonly confused in the literature due to the lack of clearly distinguishable features. Holbourn et al. [17] suggested that B. bradburyi and B. tuxpamensis are morphologically close, and that they may be conspecific.
According to the original description of B. simplex, the chambers from the first three spires are flat, and the last ones bulging and protruding [121]. On the contrary, B. bradburyi has inflated chambers [122], and they are broader than in B. simplex. In contrast to the sutures in B. tuxpamensis, in B. bradburyi they are distinct and depressed. Bulimina tuxpamensis has a very stout test, regular in outline, with wide and limbate, not depressed sutures [123]. This author pointed out that some specimens of B. tuxpamensis tend to become slightly three sided, but in our observations this species can be easily distinguished from buliminids with truly triangular cross-section, such as Bulimina prolixa and Pyramidina rudita (see below).
We noted the above differences between B. bradburyi and B. tuxpamensis during the examination of the paratypes of the first species (CC 38962, CC 559449, NMNH) and the plesiotypes of the second one (CC 23852, CC 62507 and USNM 623763, NMNH). However, the plesiotypes of B. simplex (CC 56845, CC 42645 and CC 24625, NMNH) are in general more slender than B. bradburyi and B. tuxpamensis, and show a broad variety of test-shapes. Some have a narrow test at the initial end, with greatest width at the apertural end, whereas others have a more fusiform test. The shape of the chambers varies, from square-rounded and inflated, to longer and more inflated.
Based on the original description of Cushman and Parker [124], B. prolixa is longer and narrower than P. rudita, and its sutures are slightly depressed. We examined several paratypes of B. prolixa (CC 32707 and CC 32708, NMNH). The specimens have a slender test with triangular cross-section, although they may vary from a typical buliminid shape (largest width close to last chamber) to somewhat fusiform (largest width closer to mid-test).
In contrast, P. rudita has a rougher surface but smaller test, with distinct, depressed and sigmoidal sutures [124]. This species is characterised by distinctly concave sides, as observed in several specimens from two slides at the NMNH (CC 32719 and CC 32723). Georgescu et al. [125] emended this species based on the pore mounds of the test and assigned it to the genus Pseudouvigerina. According to these authors, this genus includes turrilinids with pore mound ornamentation, but their illustrated specimens have a more elevated and pyramidal apertural face than the specimens in Ypresian samples. Moreover, P. cristata [126], the type species of Pseudouvigerina, has an aperture surrounded by a neck (e.g., [127]), in contrast with the loopshaped aperture of the genera Bulimina and Pyramidina. The original description by Cushman [128] points out that Pseudouvigerina is characterised by a biserial early stage, becoming triserial later in its development, whereas Loeblich and Tappan [116] mention that the test is triserial throughout, as also seen in the micro-CT scan of Pseudouvigerina sp. in Speijer et al. [129]. Early Eocene deep-sea benthic foraminiferal faunas We did not observe an early biserial stage or an aperture surrounded by a neck in our specimens, neither in the SEM images of Pyramidina rudita hypotypes in Georgescu et al. [125], and therefore disagree with Georgescu et al. [125], and retain P. rudita in Pyramidina.
3.1.7 Costate buliminid group. This group includes Bulimina semicostata and B. jarvisi, which have a test with coarse perforations and irregularly-shaped, in some cases anastomosing costae (Fig 5.6). Van Morkhoven et al. [9] pointed out that these species are difficult to distinguish, and Tjalsma and Lohmann [11] suggested that transitional forms may occur in the middle Eocene (biozone P14).
The typical B. jarvisi is longer, and its greatest diameter occurs approximately at two thirds of the distance from the proloculus, whereas B. semicostata has a blunter triangular cross-section, a greatest diameter at the penultimate chamber, and an imperforate area extending from the aperture to the final chambers [9]. Based on our observations of the holotype of B. jarvisi (CC 23128, NMNH) and cotypes of B. semicostata (CC 51871 and CC 59482, NMNH), B. jarvisi is narrower and longer than B. semicostata, and has more inflated chambers and more depressed sutures, in agreement with the original description of Cushman and Parker [130]. The costae in B. jarvisi proportionally cover a larger area of the test than in B. semicostata. This ornamentation is finer in B. jarvisi than in B. semicostata [17].
Additional species that may be related to this group include B. asperoaculeata and B. glomarchallengeri. Bulimina asperoaculeata (originally spelled as aspero-aculeata) shows less distinctive costae and deeper sutures than B. jarvisi [131]. The microspheric form of B. semicostata may be confused with B. glomarchallengeri, which also slightly resembles the juvenile specimens of B. callahani, but B. glomarchallengeri has more inflated chambers. According to Olshanetskiy [132], B. glomarchallengeri had its first appearance during the late Eocene, later than the range of B. semicostata.
Other Ypresian buliminids with costate tests (e.g., B. virginiana, B. farafraensis and B. alazanensis) are not included in this group because they can be easily distinguished from B. semicostata or B. jarvisi based on the general morphology of their test. For example, the hypotype of B. virginiana (USNM 139271, NMNH) has a small, narrow test, clearly longer than wide, triangular in cross-section with concave sides and fine costae over the whole test. Some specimens of B. virginiana from ODP Site 1051 and DSDP Site 550 have flat rather than concave sides. Both B. alazanensis and B. farafraensis have a somewhat larger width/length ratio, sometimes with a small spinose projection at the initial end, but the former species has prominent, continuous and more pronounced longitudinal costae, whereas B. farafraensis (junior synonym of B. aksuatica according to Deprez et al., [98]) shows numerous long, well-defined costae. Alegret and Ortiz [117] pointed out that B. farafraensis may show more irregular costae than the holotype, ocassionally branching. According to LeRoy [133], B. farafraensis can be distinguished by its rounded triangular cross-section and uninterrupted longitudinal costae, which we clearly observed on the holotype (CC 58013, NMNH).

Fusiform buliminid group.
This group has a complex record in the taxonomic literature, and includes buliminids with a typical fusiform shape, smooth test, inflated chambers and depressed sutures, resembling Praebulimina reussi (more inflated form) and Bulimina kugleri (slender form; Fig 5.10). According to Hofker [134], the genera Praebulimina and Bulimina are distinguished by the complexity of the inner tooth-plate structure linked to the aperture in Bulimina, whereas the toothplate in Praebulimina is rather simple. This author thought that Praebulimina might be considered ancestral to Bulimina, if the toothplate is considered the most typical characteristic of these genera. It has been suggested that P. reussi (with its typical subterminal, comma-shaped aperture) was common in the Maastrichtian and became extinct at or close to the Cretaceous/Paleogene boundary globally [135,136], although Sprong et al. [137] reported it from Danian strata. The subtle differences between the genera Praebulimina and Bulimina, however, have not been clearly described in the literature, and not been observed by all authors.
The species B. kugleri is characterised by a marked fusiform shape, with long inflated chambers, and the greatest width in the middle part of the test, with an aperture characterised by a high, arched, curved opening at the base of the inner margin of the last chamber [138]. The holotype of this species (CC 38199, NMNH) clearly shows these features, but we found considerable variability in the fusiform shape in the nineteen paratypes observed (CC 38257, NMNH), as also observed by Thomas [139]. Some of these specimens have the greatest width in the middle part of the test, but with a broader end part, so that the general form is closer to that of a typical buliminid. Additionally, some paratypes have a smaller width-to-length ratio than the holotype and specimens from the Southern Ocean Sites 689 and 690 [139]. This author pointed out that the specimens from the Southern Ocean closely resemble B. kugleri, but that they are smaller than the type specimens.
Chambers of P. reussi rapidly flare from the pointed initial part, and the test has its broadest part in the last whorl [140]. Praebulimina reussi has a subterminal, comma-shaped aperture. We observed three plesiotypes, one of which (CC 39602, NMNH) has a markedly pointed initial part with small chambers that increase gradually in size, but the chambers rapidly increase in size at approximately one third of the test, resulting in very inflated chambers. The initial part of the other two plesiotypes (CC 22433 and CC 51863, NMNH) is less pointed, the chambers do not abruptly increase in size, and the greatest width occurs in the middle part of the test. We observed that the last chambers of P. reussi are occasionally elongate and inflated, resembling B. kugleri. Similarly, Alegret and Thomas [10] suggested that P. reussi may vary in elongation of the test, ranging from low and globular to elongate and fusiform. In the latter case, these two species may be distinguished because B. kugleri does not show the small, low and compressed early chambers of P. reussi.
The identification of these species is problematic, and they have been confused with each other, or misidentified in the literature. For example, Pardo et al. [141] may have misidentified P. reussi as the species Praebulimina carseyae in their Maastrichtian material [142]. The specimens determined by White [21] as B. ovula Reuss were later renamed P. reussi (Morrow) (e.g., [16,30]) because Bulimina ovula Reuss was an invalid name, being a homonym of B. ovula d'Orbigny [140]. There thus has been considerable confusion between specimens named B. ovula Reuss and B. ovula d'Orbigny, the latter being an extant form with a much more inflated test. Another homonym is Bulimina ovula Terquem, which is now placed in the genus Buliminella because it has more than 3 chambers per whorl. Some authors consider Bulimina ventricosa [143] as synonym with P. reussi (e.g., [144]).
In addition to P. reussi and B. kugleri, the fusiform buliminid group includes similar Ypresian specimens that have been assigned different names, such as Bulimina sp. 1 (Fig 5.8  3.1.9 Globocassidulina subglobosa group. This group is integrated by the homonym species Globocassidulina subglobosa (Fig 5.16), which is recognised by its large, thick and subglobular test, with few chambers and an obliquely-set loop-like aperture on the ventral face of the terminal segment [145].
The group includes specimens identified as Globocassidulina globosa (Hanken) at Site 752 by Nomura [93], which according to the SEM image in that publication ( [93]: Plate 5, Figure 16) correspond to G. subglobosa.
3.1.10 Pullenia jarvisi group. We included into this group the species Pullenia jarvisi, P. subcarinata and P. americana, although their original descriptions indicate they are different species. Pullenia jarvisi has a planispiral test, which is completely involute, biumbilicate and somewhat depressed, with a lobulate outline and about five slightly inflated chambers, separated by depressed sutures. The aperture consists of a low slit extending from one umbilicus to the other, with a strongly convex apertural face [146]. Pullenia subcarinata has a planispiral convex test with a non-lobate outline, slightly keeled, and convex umbilicus [147]. This species shows six chambers, its sutures are not depressed, and the upper part of the last chamber is not convex. The type specimen of P. americana, which is actually a junior synonym of Pullenia quinqueloba Cushman (not Reuss), has a planispiral, completely involute, much compressed and slightly umbilicate test, peripherally rounded. Pullenia americana has 5 to 6 somewhat inflated chambers, separated by slightly depressed and curved sutures, and an elongate aperture at the base of the apertural face, low at the sides and considerably higher in the middle [146]. We think that P. subcarinata and P. americana have commonly been confused and misidentified in the literature, despite these differences. For example, the SEM images of P. subcarinata in Nomura and Takata [37] resemble P. jarvisi because of their slightly lobate outline, somewhat depressed sutures and a last whorl with five chambers. Additionally, the upper part of the last chamber is not acute, as in the type figure of P. subcarinata. On the other hand, we expected to observe a higher apertural face than depicted in the SEM image of the specimens named P. americana by Nomura [93], and we argue that those specimens are more similar to P. jarvisi. Due to inconsistencies in the identification of these specimens, we decided to include them into a single group, but other flat Pullenia species such as P. quinqueloba (Reuss) and P. salisburyi are excluded as they may be distinguished from this group mostly by their higher apertural face. Likewise, synonyms of the species that we considered misidentified should be detached from this group (e.g. P. quinqueloba).
In addition, we found that many specimens assigned to P. jarvisi in Arreguín-Rodríguez and Alegret [63], and Pullenia sp. 1 in Alegret and Thomas [148] do not belong in the genus Pullenia, because they are not truly planispiral, but should be assigned to Clinapertina subplanispira. Our data from DSDP Site 550 have been modified accordingly in the S5 Table, and the percentages have been recalculated. (Fig 5.14 and 5.15). Both species are distinctly trochospiral, with the last chamber extending over one side. They closely resemble each other, as both species have been described as having four chambers in the last whorl and a lateral chamber expansion on the ventral side [149]. Tjalsma and Lohmann [11] figured specimens of Abyssamina poagi with three chambers.

Abyssamina incisa-poagi group. This group includes Abyssamina incisa and A. poagi
According to Schnitker and Tjalsma [149], A. incisa has an accentuated surface relief, and much greater inflation and length of the lobate extension of the last chamber, although very rare morphologically intermediate specimens may occur. We observed some of these intermediate or transitional specimens (e.g., in samples from DSDP Site 550), with the last chamber more pronounced than in the holotype of A. poagi (USNM 305101, NMNH) but less than in the holotype of A. incisa (USNM 305100, NMNH), and with the degree of depression of the sutures intermediate between both holotypes.
Abyssamina quadrata is somewhat similar to the species of this group, but it can be easily distinguished by its almost planispiral test. Abyssamina quadrata has slightly depressed sutures, as does A. poagi [149], and has four chambers in the last whorl, but is characterised by a quadrate test that is spheroidal in cross-section. Comparing the holotypes, we observed that the last chamber is more inflated in A. quadrata (USNM 305103, NMNH) than in A. poagi, but it does not look as pronounced as in A. incisa, because all chambers in A. quadrata are inflated. The globular outline of A. quadrata is clearly observed in the images shown by Holbourn et al. [17]. Abyssamina poagi differs from A. quadrata in its longer and narrower chambers, which tend to overlap in the umbilical region, by its more strongly curved umbilical sutures and by its shorter, wider aperture [9]. Schnitker and Tjalsma [149] pointed out that A. quadrata slightly resembles small and rounded specimens of Pullenia jarvisi, but the aperture in A. quadrata is asymmetrical, whereas it runs from umbo to umbo in P. jarvisi.
Quadrimorphina profunda also resembles the abyssaminid species, although this species has a low trochospiral test, compressed laterally, with 3 1 /2-4 1 / 2 chambers in the last whorl. Its involute side greatly resembles a four-chambered Abyssamina. The main differences with the Abyssamina species are that Q. profunda is trochospiral, with a strongly evolute side (absent in the abyssaminids) that exposes earlier whorls, it has a peripheral aperture with a tendency for extending somewhat onto the evolute side, and it does not show an extension through the umbilical side, as in the abyssaminids. Thus, we did not include it into this group. The figured A. poagi specimen in Holbourn et al. [17] might, in fact, be a misidentified Q. profunda, in view of the shape of its last chamber extending over the evolute side.
3.1.12 Eilohedra weddellensis group. Eilohedra weddellensis and Alabaminella weddellensis are generic assignments used for the species originally described as Eponides weddellensis (Fig 6.8). According to Earland [150], Ep. weddellensis has a minute biconvex test, with a rounded peripheral edge and an aperture characterised by a minute slit on the inner edge of the final chamber on the ventral side. This species has five chambers in the last whorl, separated by distinct, flush sutures on the dorsal side, and depressed on the ventral side.
The taxonomy of this species is not fully resolved, thus we are not able to designate the best fitting generic assignment, but here we point out the main differences among these genera. Specimens of the genus Eponides have a biconvex test with angular to carinate periphery. Neither Eilohedra nor Alabaminella show such a feature, instead they have a rounded periphery. Moreover, Alabaminella has a biconvex test, whereas Eilohedra has a flattened umbilical side and a convex spiral side [116], and they differ in the presence of an apertural plate-like lip in Alabaminella [151], and the thightly coiled test of Eilohedra [152].

Anomalinoides spissiformis group.
This group includes morphologically similar species that have been commonly confused in the literature, such as Anomalinoides spissiformis, A. praespissiformis and A. praeacutus (also called A. praeacuta) (Fig 4.7). The original description of A. praespissiformis states that it closely resembles A. spissiformis: the former is characterised by a smaller size, fewer chambers and more strongly curved sutures [153]. Our observations suggest that the size and number of chambers may vary within a species, thus these criteria do not appear to be sufficient to distinguish between species. For example, Tjalsma and Lohmann [11] mention that the specimens identified as A. spissiformis are smaller and more tightly coiled than the holotype. The curvature of the sutures may be a better criterion to distinguish these species, and we leave this possibility open for future studies.
According to the type description of A. praeacutus, its periphery is rounded, but the figure of the apertural view in the original description of Vasilenko [154] shows a more acute periphery on the first part of the last whorl. This feature is used by some to distinguish this species from the other two, although there is no agreement in the literature. For example, the specimens identified as A. praeacutus and A. praespissiformis by Nomura and Takata [37] and D'haenens et al. [60], respectively, resemble A. spissiformis. Due to the lack of consistent identification of these species, we included them in one group.
There is also confusion with the species A. ammonoides, which has been commonly identified in the Cretaceous and Paleocene (e.g., [10]), whereas morphologically similar forms have been identified as A. spissiformis in Eocene material (e.g., [155]). Due to the strong similarity between these two species, Arreguín-Rodríguez and Alegret [63] called their specimens from Site 550 Anomalinoides ammonoides/spissiformis (Fig 4.8 and 4.9). We suggest that these species may be synonyms, with the name A. ammonoides commonly being used for Cretaceous and Paleocene specimens, and A. spissiformis for Eocene specimens.

Anomalinoides capitatus group. Specimens called Anomalinoides capitatus, A.
capitatus/danicus, and A. rubiginosus are included into this group. The species commonly named Anomalinoides danicus (originally described as Cibicides danica) is a junior synonym of Anomalinoides rubiginosus [9,10], characterised by a trochospiral, closely coiled test, rounded to somewhat lobulate on the periphery, chambers and sutures rather indistinct, 9 to 10 chambers in the last whorl and a very coarsely perforated wall on both sides. Van Morkhoven et al. [9] and Bolli et al. [7] pointed out that A. rubiginosus ranges from the Upper Cretaceous to the upper Paleocene, and Alegret et al. [38,39] suggested that this species became extinct at the PETM. Speijer [99] and Holbourn et al. [17], however, mention a range for this species up into the lower Eocene.
Anomalinoides capitatus differs from A. rubiginosus in the imperforate periphery of the early chambers, fewer chambers (5-6) in the final whorl, which increase gradually in size, and in the marked sutures varying from thin and depressed in the final chambers to thick and raised in the earlier ones. Van Morkhoven et al. [9] considered that A. rubiginosus, A. capitatus, A. semicribratus, and A. globosus represent a chronocline from the Late Cretaceous to the present. Thus, the specimens called A. capitatus/danicus by Nomura [93] may correspond to transitional forms. Tjalsma and Lohmann [11] recognised some of these transitional forms in the upper Paleocene.

Small Cibicidoides/Anomalinoides group.
This group includes small trochospiral, somewhat compressed specimens assigned to the genera Cibicidoides and Anomalinoides, typically with an unequally biconvex to nearly planoconvex test with numerous (usually 12-15) chambers in the last whorl. After careful examination of the data from the study sites, we included Cibicidoides micrus, Anomalinoides cf. acutus, Cibicidoides subcarinatus and C. ungerianus into this group (Fig 6.10-6.14).
According to the description of Bermúdez [156], C. micrus has some protruding shell growth, usually a continuous, circular protuberance, at the center of the dorsal surface. We checked the holotype (CC 62431, NMNH) and two paratypes (CC 62432, NMNH) of C. micrus, and observed a depressed spiral suture that delimits a circular protuberance formed by a small edge-like extension at the base of the chambers. This feature is more clearly seen in the holotype, but difficult to observe in one of the paratypes, and it was not observed in the specimens identified by Arreguín-Rodríguez and Alegret [63] at Site 550.
After reviewing our material, we recognise that specimens called C. micrus at DSDP Site 550 and ODP Site 1051 closely resemble those called A. cf. acutus at ODP Sites 1262, 1263 and 690 [30,62] and Anomalinoides sp. B in Cretaceous-Paleogene sediments at ODP Site 1262 [157], but these do not fully agree with the type description of A. acutus [158]. Furthermore, in view of the images of D'haenens et al. [60] and Nomura and Takata [37], we suggest that the specimens named C. ungerianus and C. subcarinatus respectively, are more similar to C. micrus or A. cf. acutus than to the type figures and descriptions of these taxa. Cibicidoides ungerianus has an acute and carinate periphery, but the specimens illustrated in D'haenens et al. [60] do not show a carinate periphery. Cibicidoides subcarinatus has an equally biconvex test with a subcarinate periphery and limbate, comma-shaped sutures, whereas the specimens figured by Nomura and Takata [37] show a more planoconvex test with comma-shaped but slightly depressed sutures.

3.1.16
Cibicidoides with an umbo group. Species of this group are distinguished from other biconvex Cibicidoides by the elevations of shell material on the dorsal side, and the occurrence of an umbo on the ventral side. Cibicidoides alleni and C. dayi are included in this group.
The general morphology of both species is similar, but according to Alegret and Thomas [10] C. dayi has a strongly depressed spiral suture around a prominent umbo instead of a true elevation of shell material at the base of the chambers on the dorsal side; in addition, its umbo on the ventral side is rather indistinct. In the photographs of type specimens of C. dayi (AMNH-F1-19908 and AMNH-F1-19909, available at http://foraminifera.eu/amnh.php), an umbo on the ventral side may be observed only on specimens with a marked spiral suture, whereas the umbo is indistinct in other specimens.
Specimens identified as Anomalinoides trinitatensis by Takeda and Kaiho [36] are included in this group because they closely resemble the species included in this group (C. alleni and C. dayi). In the original description of Nuttall [159], A. trinitatensis also shows a growth of shell material on the dorsal side and a boss on the ventral side, but it has a more planoconvex test. However, specimens in Takeda and Kaiho [36] differ from A. trinitatensis, in having fewer but more inflated chambers, than those in the type figure, and there is no umbo on the ventral side.
Cibicidoides micrus also has a depressed spiral suture on the dorsal side, but it is not included into this group because of its smaller size and flatter test, and it was included in the group of small Cibicidoides/Anomalinoides.

Cibicidoides mundulus group.
This group consists of trochospiral species with a morphology similar to that of species such as Cibicidoides mundulus, C. praemundulus, C. pseudoperlucidus and C. proprius (Fig 5.24-5.26). All have a planoconvex or unequally biconvex to lenticular test, with a plug on the umbilical side and a coarsely perforate wall.
Cibicidoides mundulus and C. praemundulus have lenticular tests with an imperforate band along the periphery, 3 whorls with 10-12 chambers in the last one, and a plug or umbo. According to Van Morkhoven et al. [9], they differ because C. praemundulus is smaller and more compressed, and does not have pores concentrated along the spiral sutures as in C. mundulus. Van Morkhoven et al. [9] and Holbourn et al. [17] showed that the degree of curvature in the sutures, the pattern of pores and the test size of C. mundulus are variable, and they considered C. kullenbergi a junior synonym of C. mundulus.
The species C. proprius is more planoconvex, and C. pseudoperlucidus is biconvex but with a somewhat flattened ventral side. Brotzen [131] described that C. proprius may vary in convexity, from planoconvex to biconvex. These two species have a slightly lobate, acute periphery, and a finely perforate wall. Cibicidoides mundulus and C. praemundulus and these two species have a plug or boss, and have 3 whorls, but C. pseudoperlucidus and C. proprius usually have fewer chambers in the last whorl than the other two species. Berggren and Aubert [160] suggested that C. proprius and C. alleni might be considered as synonyms, arguing that the variation in convexity of the test (C. proprius-planoconvex, C. alleni-biconvex) is not a differentiating feature of these two species. However, Alegret and Thomas [10] documented that C. alleni also differs from C. proprius in the calcite ridge along the spiral side, and in the lack of an acute periphery as in the latter species. For the purpose of this study, we consider these two species (C. alleni and C. proprius) sufficiently distinct to be separated into different groups (C. alleni is included in the Cibicidoides with an umbo group, and C. proprius in the C. mundulus group).
Specimens called Cibicidoides eocaenus in D'haenens et al. [60] were re-named as Cibicidoides howelli (S3-S5 Tables). Cibicidoides howelli may have become extinct during the PETM (e.g., [38]). Alegret and Thomas [10] suggested that 2 paratypes of C. howelli (CC 38529), deposited at the NMNH, should be included in C. proprius, because they lack the rounded last portion of the test and have an acute periphery. Based on the SEM images in D'haenens et al. [60], we suggest that these specimens (originally called C. eocaenus) should be included in C. proprius, due to their resemblance to the paratypes of C. howelli described by Alegret and Thomas [10], and thus are incorporated in the C. mundulus group.

3.1.18
Cibicidoides eocaenus group. The species Cibicidoides eocaenus, C. tuxpamensis and C. perlucidus are included in this group. The test of C. eocaenus is small, nearly circular and unequally biconvex (may vary from planoconvex to biconvex), its umbilical side is somewhat cone-shaped, and the spiral side is somewhat less convex. The chambers (12)(13)(14)(15) are inflated, separated by curved and limbate sutures. Additionally, C. eocaenus has a distinct spiral suture line and frequently a prominent umbilical umbo [9]. In spite of the latter feature, C. eocaenus is not included in the Cibicidoides with an umbo group because of its characteristic cone-shaped umbilical side, which makes them easy to distinguish.
Van Morkhoven et al. [9] compared topotype specimens and suggested that the species C. eocaenus and C. tuxpamensis may be conspecific. The latter species is very similar to C. perlucidus, characterised by its umbilical plug and acute periphery, but they differ because C. tuxpamensis has a more rounded periphery. According to these authors, there are gradational variations between these two species, which may be considered as ecophenotypic variants and the names perlucidus and tuxpamensis may be used to refer to the periphery of each morphotype. Following this criterion, Arreguín-Rodríguez and Alegret [63] used the names C. eocaenus perlucidus and C. eocaenus tuxpamensis to designate specimens with an acute and rounded periphery respectively at Site 550 (Fig 6.15 and 6.16). These names have also been applied in the classification of benthic foraminifera at Sites 1051 and 1258 (this study), and these species have been included here into the C. eocaenus group.

Flat Gyroidinoides group. This group includes
Gyroidinoides species characterised by compressed, flat to slightly biconvex tests and a rounded periphery such as Gyroidinoides depressus, G. planulatus (both alternatively assigned to Valvalabamina, e.g., [98,99,118]) and G. complanata (Fig 5.27-5.29). The small differences in e.g., the shape of the sutures, the degree of inflation of the chambers, and the number of chambers in the last whorl make these species difficult to distinguish.
Gyroidinoides depressus may have somewhat inflated later chambers and depressed sutures, whereas chambers in G. planulatus and G. complanata are not inflated, thus their sutures are not depressed. The latter two species may differ in the shape of the sutures on the ventral side, which are slightly curved in G. planulatus [161] and straight and nearly radial in G. complanata [162].
3.1.20 Nuttallides umbonifera group. The original description of Nuttallides umbonifera defines a biconvex test with a boss in the umbo on the ventral side [163] (Fig 5.18-5.20), but specimens with a small boss, or even without a boss, have been commonly assigned to this species [120]. We reviewed the holotype of N. umbonifera (USNM 26162, NMNH), which has an unequally biconvex test, which is more convex on the ventral side, with an acute and slightly lobulate periphery [17]. On the dorsal side, the holotype is evolute, showing oblique to slightly curved sutures. On its ventral side, it shows an umbo, the chambers are trapezoidal in shape, and the sutures are oblique to curved, and depressed. The aperture is a slit that extends from the periphery to the umbo, with a fold perpendicular to the base of the last chamber.
Additionally, we observed specimens that resemble N. umbonifera (with an umbo), but in which the first chambers of the last whorl do not have a clearly trapezoidal shape on the ventral side, as in the type specimens, and with more chambers. These specimens have been called Osangularia sp. 1 at DSDP Site 550 [63] and at ODP Site 1051 (this study), although they are difficult to distinguish from the 'typical' N. umbonifera (Fig 5.21-5.23). According to Loeblich and Tappan [116], the aperture of Osangularia is T-shaped [164], but Nuttallides shows a ventral aperture at the base of the chamber, with a lip, extending from the umbilical boss to the periphery, formed parallel to the plane of coiling and bordered marginally by a small fold of the apertural face [165]. The apertural features, however, are not clearly seen in many specimens, especially small and/or poorly preserved ones, so that identification may be difficult. The variations in number and shape of chambers of N. umbonifera and the so-called Osangularia sp. 1 may represent interspecific variability of the same species (i.e., N. umbonifera).
The genus Nuttallinella may be distinguished from Nuttallides based on the presence of an umbilicus-rather than a boss-on the ventral side [166]. Cretaceous specimens similar to N. umbonifera, however, have been identified as Nuttallinella ripleyensis or Nuttallinella sp. at ODP Site 1262 [157]. These species have slightly less lobulate tests than N. umbonifera, and their sutures on the dorsal side are less oblique. Neither of these two species are included in this group because they are not present in the early Eocene.
Considering the general morphology of the test, and based on our observations as well as on the images of the Ypresian specimens in the literature, we suggest that specimens identified as Nuttallides umbonifera? (e.g., [14]), Nuttallides sp. 2 [37], Osangularia sp. 1 [63], Osangularia sp. [36] and Eponides elevatus [36] should be included in this group. They may be related species, represent intraspecific variability, or include several, morphologically not well-defined species.

Hanzawaia mantaensis.
In the studied material, this species has been found to be common only at ODP Site 1051. We examined several type specimens of Hanzawaia mantaensis at the AMNH (No. 19954), and were able to identify the holotype shown in the type figure, which is stored in the same slide as several syntypes (as Anomalinoides mantaensis at http:// foraminifera.eu/amnh.php).
The holotype has a planoconvex test with an acute periphery. The dorsal side is flat and evolute. It shows numerous (usually 10) narrow chambers in the last whorl, separated by curved, limbate sutures. The ventral side is convex and involute, and the sutures are curved and depressed. According to the original description [167] and to Holbourn et al. [17], this species shows a boss on the ventral side (clearly seen in the holotype), but the syntypes do not have a boss. This feature seems to be a pronounced, large proloculus (possibly with a slightly thickened wall) rather than a true boss of solid calcite.
The specimens identified at ODP Site 1051 do not exactly agree in all features with the holotype, but are more similar to the smaller syntypes in the slide, with a somewhat more convex ventral side, somewhat wider chambers, less raised sutures and a less pronounced umbilicus on the dorsal side, although the proloculus is pronounced (Fig 6.18-6.20). Considering these characteristics, we suggest that this syntype is a juvenile form of H. mantaensis, and we consider the specimens from Site 1051 also as juvenile forms.

The lower Ypresian.
Data on the lower Ypresian include 29 species or species groups with a relative abundance !2%, of which 24 (>82%) are abundant or very abundant, and only 5 (~17%) show a common relative abundance. These 29 taxa are represented in 14 ocean drilling sites from the Pacific, Atlantic and Indian Oceans (Figs 1 and 7). The species N. truempyi and Oridorsalis umbonatus, and the B. simplex group, pleurostomellids and uniserial lagenids are ubiquitous (i.e., taxa with a global distribution), whereas Bulimina trinitatensis and the A. spissiformis group, the P. jarvisi group, lenticulinids and stilostomellids are present in all regions, although not at all sites (Table 4).
Three clusters (based on faunal content) are recognised in two separate branches in the dendrogram (Fig 7). One cluster (green in Fig 7) groups together all Pacific sites (except for Site 865), another cluster (blue) includes the South Atlantic sites, and the third cluster (purple) groups sites which have a seamount setting, including Pacific Site 865 and Southern Ocean Sites 689 and 690. Separate branches hold Indian Site 752 (red) and North Atlantic Site 401 (grey). The groups of sites and individual branches are characterised by the abundance of species or groups of taxa (Fig 7). For instance, taxa 1-5 (Fig 7) seem to mainly represent Pacific assemblages, taxa 6-9 (Fig 7) are more common at Indian Site 752, whereas taxa 10-14 (Fig 7)  characterise the assemblage at North Atlantic Site 401. The seamount assemblages are mostly represented by taxa 16-29 (Fig 7), somewhat similar to the SE Atlantic assemblages where these taxa are present but generally less abundant. The difference between those assemblages lies in the higher abundance of A. quadrata, Nonionella robusta, C. subplanispira or A. aragonensis at the SE Atlantic sites (taxa 23-25, and 29; Fig 7). In addition, the species N. truempyi (taxon 4) and Q. profunda (taxon 1) are abundant at the SE Atlantic sites (especially deepest Site 1262), but common or present at the seamount sites.
The NMDS plots (coordinates 1-2 in Fig 8A, and coordinates 2-3 in Fig 8B) show the species distributed according to their abundance rank, thus species closer to each other indicate similar abundances. The taxa Q. profunda and N. truempyi are markedly more abundant than all other species included in the 95% confidence ellipse (Fig 8A). Quadrimorphina profunda is very abundant in almost all Pacific sites and abundant in the deepest SE Atlantic site (Site 1262). Nuttallides truempyi is very abundant in the East Pacific Rise sites, and abundant at Shatsky Rise (Pacific), SE Atlantic and Indian sites. The species distribution across coordinates 2 and 3 ( Fig 8B) shows that Tappanina selmensis and Siphogenerinoides brevispinosa are highly significant in the lower Ypresian. Tappanina selmensis is very abundant at Southern Ocean Site 689, and abundant at the other seamount sites, Atlantic sites (NA and SE) and at the East Pacific Rise sites. On the other hand, S. brevispinosa is very abundant at Site 690 (Southern Ocean) and abundant at the other Southern Ocean site, as well as at most Shatsky Rise sites (except for the shallowest Site 1209). Inside the confidence ellipse the distribution of the species is mostly associated to the geographic regions where they are abundant, at least considering those species abundant in one region only (i.e., marked with one color only in Fig 8).

The middle Ypresian.
The middle Ypresian, for which we have data for fewer sites than for the lower Ypresian and no data for the East Pacific and Shatsky Rise sites, still shows a higher overall species richness (number of species with abundance !2%) than the lower Ypresian, with 42 species or taxa groups. The proportion between abundant (or very abundant) and common species is more balanced than in the lower Ypresian, with 25 taxa (>59%) abundant or very abundant, and 17 (~40%) common. The species N. truempyi, O. umbonatus and B. trinitatensis, and those included in the B. simplex group, G. subglobosa group, lenticulinids, pleurostomellids, polymorphinids, stilostomellids and uniserial lagenids are ubiquitous, whereas A. aragonensis, the A. spissiformis group and the C. eocaenus group are present in all regions, but not at all sites ( Table 5).
The dendrogram shows three clusters and two separate branches (Fig 9). One branch corresponds to Pacific seamount Site 865 (purple in Fig 9), and one to Indian Ocean Site 752 (red), whereas all three clusters contain Atlantic sites. Sites 401 and 1051 form the North Atlantic cluster (grey), which is markedly distinctive. North Atlantic Site 550 (also in the N Atlantic, though at a greater paleodepth) is more similar to Southern Ocean Site 690, and falls in the cluster 'high latitude' (dark blue), which resembles the Central-SE Atlantic cluster (light blue) and Site 865 (purple) in faunal content.
Taxa 1-8 as well as taxa 39-42 (Fig 9) mostly define the North Atlantic assemblages, being the first group of taxa more characteristic at Site 401 and the second group more distinctive at Site 1051. Taxa 33-38 (Fig 9) represent the Indian Ocean assemblages, whereas taxa 16-21 (Fig 9) are typical of the seamount Pacific assemblages. The high latitude Atlantic cluster is characterised by taxa 27-32, as well as by taxa 11 and 12 (Fig 9). The Central-SE Atlantic sites are distinguished by taxa 22-26 (Fig 9) as well as by taxon 10, whereas other species such as Q. profunda, pleurostomellids and N. truempyi (taxa 9 and 14-15; Fig 9) seem to be equally representative of various regions.
Nuttallides truempyi and B. virginiana are the most significant species across the middle Ypresian, based on their higher abundance ranks compared to all other species located within the 95% confidence ellipse, which includes taxa with similar abundances (Fig 10A). Nuttallides truempyi is abundant at all studied sites except for North Atlantic Sites 1051 and 401, whereas B. virginiana is typical of these two sites, being very abundant at the shallower Site 1051 and abundant at Site 401. Tappanina selmensis and the B. cf. decoratus group also may be important species in the middle Ypresian interval (Fig 10B). The first species is very abundant at Site 690  Table 4. https://doi.org/10.1371/journal.pone.0193167.g008 Early Eocene deep-sea benthic foraminiferal faunas (Southern Ocean, higher latitude cluster), and the latter is abundant at both Atlantic high latitude sites (550 and 690) and at the Pacific seamount Site 865. Abundant species from North Atlantic, Indian and Central-SE regions are easily recognised as groups (considering those species abundant in one region only, i.e., marked with one color in Fig 10), whereas abundant species from high latitude sites (550 and 690) are more spread out inside the confidence ellipse. . The abundance of the species located outside the 95% confidence ellipse is significantly distinct (higher) in comparison with the species inside the ellipse. Common species are marked with black dots, and abundant/very abundant species are indicated with colored circles. The color of the circle represents the sites where such species are abundant. Color reference: light blue-SE Atlantic sites, dark blue-high latitude Atlantic sites, red-Indian site, grey-NA sites, and purpleseamount site. Abbreviations found in Table 5. https://doi.org/10.1371/journal.pone.0193167.g010 Early Eocene deep-sea benthic foraminiferal faunas

The lower Ypresian fauna
The lower Ypresian assemblages have been described as typical post-extinction faunas (e.g., [31,36,37,139]), including fewer species (in this case, species with a relative abundance !2%), and higher dominance of abundant species (>5%). This is as expected, because this interval is directly above the Paleocene-Eocene boundary (covering up to ∽0.8 m.y. after the P-E boundary), thus includes the peak PETM environmental disturbance directly postdating the severe extinction of deep-sea benthic foraminifera. The environmental disturbance includes increased temperature, low oxygen levels at some locations, and ocean acidification. Its effects on benthic foraminiferal fauna persisted for at least 260 kyr after the extinction, as suggested by low diversity assemblages at Maud Rise [30]. The decreased diversity probably was the combined result of the global extinction and local or temporary disappearance of species (e.g., [8]). The concomitant dominance of abundant species likely reflects the proliferation of opportunistic species in perturbed environments (e.g., [168]).
The relative abundance of the taxa in the lower Ypresian assemblages (data compilation from 14 sites; Fig 7) appears to be linked mainly to geographic regions (Pacific, Atlantic and Indian Oceans), and to a lesser extent to specific environmental settings (seamounts). Nonseamount Pacific sites were located in truly pelagic, open ocean settings far away from coastal influence (Fig 1), and contain more abundant taxa that are generally considered oligotrophic (such as Q. profunda and N. truempyi; e.g., [31]) than the Walvis Ridge Atlantic sites, with the latter closer to continental margins. Proximity to continental margins may be related to a higher amount of food reaching the seafloor, due to coastal upwelling along continental margins and higher input of nutrients by rivers (e.g., [169]), and also due to lateral transport of refractory organic matter (e.g., [170]).
The fact that the 'seamount cluster' includes sites at great distance from each other (Atlantic and Pacific Oceans) suggests that these assemblages may have been controlled by parameters specific to the seamount setting (e.g., active current patterns), independent of their geographic location. Such assemblages are strongly dominated by cylindrical infaunal taxa (such as stilostomellids), which seem to prefer relatively high organic carbon flux regularly delivered to the seafloor in non-seamount settings [171]. However, at seamount ecosystems, these taxa may have developed a suspension feeding strategy as they adapted to a specific food supply brought in or concentrated by current systems around the seamounts (e.g., [43]). Such currents are enhanced around the steep topography of the seamounts, causing removal of fine-grained food particles from benthic communities in some places, and concentration in other parts by trophic focusing (e.g., [172,173]).
North Atlantic Site 401 (paleodepth~1900 m) does not cluster with other Atlantic sites, and has an about equal abundance of infaunal and epifaunal taxa (Table 4), in contrast to most other sites. This site is characterised by a high abundance of epifaunal taxa (in comparison to other sites) such as Epistominella exigua (indicative of a pulsed food supply [107,174,175]), the C. mundulus group, and the N. umbonifera group, which are less abundant at other sites. This assemblage might reflect a more variable food supply at these fairly high northern latitudes, possibly with high seasonality, or higher levels of oxygenation than at other sites (see below, 4.3). For this time interval, we have no data from other North Atlantic sites to test hypotheses about the parameters controlling the benthic assemblages.
Quadrimorphina profunda, N. truempyi, T. selmensis and S. brevispinosa are the most notable species, as inferred from the NMDS plots (Fig 8). Nuttallides truempyi and the deep-water species Q. profunda may indicate lower food supply (e.g., [15,31,36]). Inside the 95% confidence ellipse, abundant species characterising the Pacific assemblages are located relatively closer to the oligotrophic species N. truempyi and Q. profunda than the rest of species typical from other regions (Fig 8A), which agrees with the fact that Pacific sites are located in open ocean settings (far from continental margins), with lower food supply to the seafloor. Moreover, N. truempyi may also reflect a potential resilience to surviving in carbonate-corrosive waters [31,38,39].
On the other hand, T. selmensis and S. brevispinosa may have behaved as opportunistic taxa, blooming in response to environmental perturbations associated to the PETM. Both species tolerate stressed/disturbed environments (e.g., [38,39,139,176]) and possibly indicate a high food supply (e.g. [177]), although this is debated for T. selmensis (see e.g., [31,178]). Other buliminids such as the B. simplex group and the fusiform buliminid group are located closer to T. selmensis and S. brevispinosa in the NMDS plot, but inside the confidence ellipse (Fig 8B). Sprong et al. [137] suggested that B. kugleri (included in the fusiform buliminid group) may have been an opportunistic species, reported from a wide paleodepth range and dependent on food supply, and commonly abundant just after the Paleocene-Eocene extinction [14,139]. Additionally, Takeda and Kaiho [36] suggested that Q. profunda is an opportunistic species that may have become more tolerant to stressed conditions after the PETM, thus accounting for its location relatively close to the other opportunistic species (Fig 8B). Following this line of reasoning, we suggest that the N. umbonifera group may also have had an opportunistic behavior, considering that N. umbonifera and Osangularia sp. 1 (both included in the N. umbonifera group) peaked in abundance together with other opportunistic taxa across the H2 event at DSDP Site 550 [63].

The middle Ypresian fauna
Assemblages in the middle Ypresian interval have a higher diversity (i.e., higher number of species) than those from the lower Ypresian, and higher equitability, i.e., a similar number of abundant and common taxa despite the smaller number of studied sites (9), and the occurrence of several lesser hyperthermal events (see e.g., [179]). The higher diversity of these assemblages may reflect long-term recovery after the PETM extinction through evolution, migration, and return of 'Lazarus' taxa (e.g., Siphonodosaria subspinosa, Strictocostella pseudoscripta [8]). This sustained recovery suggests that the effects of the Eocene hyperthermals on the benthic fauna on a global scale were considerably less than the effect of the PETM, despite local short-term, reversible declines in diversity and changes in assemblage composition during the early Eocene hyperthermals [43,60,62,63].
As in the lower Ypresian, the sites generally cluster largely by geographical location rather than bathymetry, with a North Atlantic and an Atlantic cluster, and separate branches for Indian and Pacific Oceans. The exception appears to be the clustering of North Atlantic Site 550 with Southern Ocean Site 690 (high latitude cluster), whereas Sites 550 and 401 are clustering far apart, though they are in relatively close geographic proximity (with Site 550 at greater depth; Figs 1 and 3). Sites 401 and 550 share high abundances of the N. umbonifera group and the G. subglobosa group, but Site 550 does not have abundant B. virginiana, and neither other species such as E. exigua, the B. elongata group or the C. mundulus group, probably due to its greater depth (see below). Despite their seamount setting location, Site 690 is separated from Site 865 in the middle Ypresian (in contrast to the lower Ypresian interval) (Fig 9). Site 690 is characterised by higher abundances of the opportunistic T. selmensis and the S. brevispinosa group than Site 865 (although these sites share high abundances of stilostomellids), possibly indicating that Site 690 and Site 550 (which with it clusters) were more susceptible to environmental perturbations during hyperthermal events, with more pronounced effects at great depths (Site 550) and higher latitudes (Site 690). This relation between fluctuations in abundance of opportunistic taxa and hyperthermals is inferred, for example, from major assemblage changes documented during the ETM2 and H2 events (including variations in relative abundances and decreased number of specimens) at Site 550 [63]. Furthermore, the high susceptibility of these two sites may be related to the abyssal, oligotrophic setting at Site 550, which makes it very sensitive to trophic changes, and by the seamount setting at Site 690, making the site sensitive to hydrographic changes probably enhanced by its higher latitude location.
The middle Ypresian assemblage in the Indian Ocean differs from the other assemblages, as in the lower Ypresian, possibly because the site is much shallower (~1000 m paleodepth), with a higher food supply and more stable seawater carbonate saturation.
According to the NMDS plot (Fig 10), the most distinguished species during the middle Ypresian interval are B. virginiana, N. truempyi, T. selmensis and the B. cf. decoratus group, with the first species showing palaeogeographic and paleobathymetric restrictions. Bulimina virginiana has been documented from shallower depths along the Northeast coast of the United States (e.g., [176,180,181]), and its abundant occurrence at North Atlantic Sites 1051 (~1500 m paleodepth) and 401 (~1900 m paleodepth) may indicate that it migrated from shallower depths after the PETM extinction. At NW Atlantic Site 550 (~3900 m paleodepth), B. virginiana is present but not abundant, suggesting that it may have not been able to thrive at sites deeper than~2000 m. Furthermore, B. virginiana was not recognised at any of our other study sites, suggesting that it may have been restricted geographically, occurring in the North Atlantic only.
Abundant B. virginiana might indicate higher food supply at North Atlantic Sites 401 and 1051, considering that it is a buliminid taxon that lived at shallower, more eutrophic depths. This interpretation is supported by the opposite location of N. truempyi and B. virginiana in the NMDS plot (Fig 10), which may indicate distinct ecological preferences with N. truempyi reflecting oligotrophic conditions [31,182]. This hypothesis is supported by the location of uniserial taxa relatively closer to N. truempyi than to other species (Fig 10B), because stilostomellids may indicate overall oligotrophic conditions but fairly active current systems (as at Site 865), leading to trophic focusing and decoupling between productivity and arrival of food to the seafloor [43]. Nuttallides truempyi may also indicate carbonate corrosive waters (e.g., [31]), which could be supported by the location of lenticulinids, a group that is resistant to dissolution [183], and other infaunal taxa such as O. umbonatus, the B. simplex group, uniserial lagenids, pleurostomellids, Q. profunda, or the fusiform buliminid group, somewhat close to N. truempyi. However, we did not find evidence for CaCO 3 -corrosiveness at the studied sites during the middle Ypresian interval (see section 4.3).
The B. cf. decoratus group might be interpreted as reflecting an abundant food supply, since it is a buliminid taxon (sensu lato), but its abundance in assemblages with abundant oligotrophic taxa at Site 550 [63] suggests that this group is tolerant to meso-oligotrophic conditions. Likewise, Boscolo Galazzo et al. [184,185] argued that the species B. crenulata and B. huneri (which resemble B. decoratus) might indicate a lower food flux than other buliminids. The proximity with the opportunistic T. selmensis [178] in the NMDS plot (Fig 10B) suggests that the B. cf. decoratus group might have had an opportunistic behavior. These taxa, like T. selmensis, are biserial with an ornamented wall.

Common features in the Ypresian time intervals
A common feature of benthic foraminiferal assemblages across both Ypresian intervals is the marked and widespread dominance of infaunal taxa [14,29], including taxa placed in the buliminid group sensu lato (triserial and biserial forms), in the uniserial groups (stilostomellids, unserial lagenids), or the pleurostomellids. The dominance of infaunal groups, which has been linked to some combination of high food-low oxygen conditions (TROX model, [100,101]) in the present oceans, is not easy to explain during the warm early Eocene. In the modern oceans, food supply to the seafloor is generally linked to high primary productivity in surface waters. However, the transfer of food from the surface to the seafloor is not reflected by a simple logarithmic function, as traditionally thought (e.g., [186,187]), but it varies strongly by location and ecosystem structure (e.g., [188,189]). Over longer time periods and across climate change events, one might expect that the link between surface primary productivity and food supply to the benthos (bentho-pelagic coupling) has not necessarily always been strong and may have been broken, for instance through processes of trophic focusing on seamounts [43], carbonate dissolution in bottom waters [88], or changes in the amount of remineralization of organic matter, with higher rates expected at higher temperatures (e.g., [108,109,185]).
Despite these complexities, food supply to the deep sea overall is expected to have been lower in warmer than in cooler periods of Earth history (e.g., [15]), due to a combination of factors including increased stratification in warmer oceans, which limits upwelling of nutrient-enriched deeper waters, thus primary productivity (e.g., [95]), and the increased remineralization of primary-produced material due to higher metabolic rates or remineralizers, which limits export productivity (e.g., [108,109,185]). This hypothesis is supported by the observation that delivery of food to the seafloor may have declined strongly during hyperthermal events such as ETM2 [60,62].
The high abundance of infaunal taxa in the warm Paleogene oceans has been linked to lower oxygen conditions rather than to a high food supply [190,191], due to the lower solubility of oxygen at higher temperatures. However, moderately-low oxygen conditions (rather than anoxia) do not appear to strongly affect living deep-sea benthic foraminifera (e.g., [101,107]. In fact, some foraminifera have been observed to be able to calcify under anoxic conditions [192]. The lower oxygen hypothesis is, however, supported by geochemical evidence that the deep sea may have been less oxygenated in the early Paleogene in general [193][194][195]. Oxygen minimum zones in the Paleogene may have been more extensive, thus covering more of the seafloor, with more pronounced deoxygenation during hyperthermal events [196][197][198][199], at least at Atlantic sites [197]. Despite the observation that low oxygen conditions do not seem to affect modern living benthic foraminifera in overall well oxygenated cool oceans, the overall, long-term lowered oxygen conditions of the Paleogene may have affected the composition of benthic assemblages. This may not have been a direct effect, but could have operated through giving a competitive advantage to infaunal taxa using nitrate as electron acceptor through denitrification, as seen in many extant species, especially deep-infaunal taxa (e.g., [200][201][202][203]. This denitrification process could have been more pronounced at high nitrate levels in the deep waters, e.g. as produced by high rates of remineralization. In addition, the occurrence of multiple short-lived hyperthermal events during the early Eocene (e.g., [179]), associated with widespread carbonate dissolution in the deep sea, may have resulted in pore waters being more carbonate-saturated than bottom waters, at least in some areas in the Atlantic Ocean [88]. These conditions would give competitive advantage to calcareous infaunal taxa that lived and calcified buried in the sediment, in contact with the more saturated pore waters rather than bottom waters. However, we do not see evidence for the long-term occurrence of widespread, highly corrosive bottom waters at the depths of our sites, because non-calcareous agglutinated benthic foraminifera are not common in any of the studied samples (lower and middle Ypresian), or in the South Atlantic samples studied by Müller-Merz and Oberhänsli [24], and because the carbonate content is generally high at the studied sites (except during the hyperthermals at some locations, e.g., [37,82]). In addition, reconstructions of the position of the Calcium Carbonate Compensation Depth (CCD) show that it was shallower than today, but not by more than several hundred meters, leaving most of our sites above the CCD (with the exception of the PETM and possibly ETM2) [204].
Another common feature in the assemblages across both time intervals is the fact that neither the species (and/or groups of species) nor the number of taxa with a global distribution (i.e., ubiquitous species) differ significantly: only 4 taxa that were not ubiquitous in the lower Ypresian became ubiquitous in the middle Ypresian, and 1 group from the lower Ypresian was no longer widely distributed in the upper interval. We suggest that the palaeogeographic distribution of benthic assemblages did not change significantly within the early Eocene (lower and middle Ypresian), although we have data on fewer sites for the middle Ypresian interval (specifically lacking data on the equatorial Pacific and Shatsky Rise, with the exception of seamount Site 865). Our present dataset indicates that the relative abundance of benthic foraminifera in the Ypresian was linked mainly to geographic regions (Pacific, Atlantic and Indian Oceans), and to a lesser extent to specific environmental settings (seamounts). Sites on depth transects (East Pacific Rise, Shatsky Rise, Walvis Ridge) generally group closely together, whereas sites at similar depths in different regions (e.g., Sites 401-North Atlantic, 1209-Pacific and 690-Southern Ocean, all at~1800-2000 m) do not cluster together (Fig 7). This observation suggests that paleobathymetry (within the studied range) was a less important determining factor than geography for this time interval, which is in contrast with the cosmopolitan Velasco fauna. Note that we are not certain why the one assemblage from the Indian Ocean is clearly distinct from the Atlantic and Pacific clusters, as it differs in both geography and a shallower paleodepth (1000 m) (Fig 3). We argue that the overall similarity between assemblages along a depth transect, studied by the same set of authors (Shatsky Rise: Takeda and Kaiho [36]; equatorial Pacific: Nomura and Takata [37]; Walvis Ridge: Thomas and Shackleton [34], Jennions et al. [62]; Maud Rise: Thomas [14]) is probably not due to different approaches to taxonomy. These geographic differences persist in our taxonomic compilation (morphological groups rather than specific taxa), which aimed at correcting for such effects and making this global compilation possible.
Our assessment of benthic foraminifera across the two Ypresian intervals provides new insights into the main structure and composition of early Eocene global assemblages. Most published compilations have mainly focused on the major extinction across the Paleocene-Eocene boundary (e.g., [11,15,26,27,31]), documenting the transition from the long-lived Cretaceous-Paleocene (Velasco) fauna, which showed transient assemblage changes across the Cretaceous/Paleogene impact event (e.g., Alegret et al. [205], and references therein), to the Paleogene fauna, which persisted from the PETM to the Eocene/Oligocene turnover. According to those studies, the Eocene post-extinction fauna was characterised by low-diversity and high dominance of some species like N. truempyi, abyssaminids and/or buliminids (e.g., [31,36,139]). Our analysis of the lower Ypresian interval confirms this basic pattern, i.e., we confirm that the lower Ypresian globally contains fewer common species, and recognise N. truempyi and Q. profunda (usually included in the abyssaminid group) as some of the most important species. We also confirm the abundance of buliminid taxa (Figs 7 and 8). In addition, we note that both Ypresian assemblages contain abundant pleurostomellids and stilostomellids (at least in some regions), which are quantitatively rare and unimportant within the preceding Velasco fauna [160]. In contrast to our Ypresian records, these authors noted that several agglutinated taxa are typical of the Velasco fauna. On the other hand, taxa such as N. truempyi and buliminids remain equally abundant in both faunas (Velasco and Ypresian). Furthermore, the abyssal Barbados fauna (2500-3000 m paleodepth) from the middle Eocene-Oligocene contrasts to both Ypresian assemblages by its decreasing abundance of N. truempyi (which disappeared at the end of the Eocene together with Abyssamina species), lower abundance of buliminids, and dominance of G. subglobosa [206]. Therefore, a sucession of faunas such as Velasco-Ypresian (also called Walvis Ridge fauna, see next section)-Barbados may cover the transition of deep-sea benthic foraminiferal assemblages from the Cretaceous through the Oligocene.

Walvis Ridge fauna
Benthic foraminiferal assemblages from Walvis Ridge across the lower and middle Ypresian intervals may be considered as representative deep-sea fauna (Figs 7 and 9), recovering from a major perturbation (PETM) during a warm period with superimposed hyperthermal events (early Eocene). Therefore, we propose the name 'Walvis Ridge fauna' for future reference to the early Eocene (Ypresian) deep-sea benthic foraminiferal assemblages.
Not all species included into this fauna shown a global distribution at lower bathyal-abyssal paleodepths (1500-3500 m), but similar assemblages can be recognised in other Atlantic regions and in the equatorial Pacific during the lower-middle Ypresian (∽56-50.7 Ma [25]).

Conclusions
Our study synthesizes and evaluates, for the first time, early Eocene (Ypresian) benthic foraminiferal assemblages living in a greenhouse world with superimposed, short-lived warming events, after a significant extinction. Thus, this taxonomic and paleoecologic compilation constitutes a tool for reconciling different taxonomic concepts and provides an analysis of benthic foraminiferal diversity variability in space and time, and recovery and diversification from PETM extinction.
Variations in early Eocene benthic foraminiferal assemblages at the studied sites in the Pacific, Atlantic and Indian Oceans are mainly correlated to differences in palaeogeography (proximity to continental margins, seamount settings, or open-ocean pelagic settings), and to a lesser degree to paleobathymetry.
The lower Ypresian assemblages are characterised by lower diversity and equitability, i.e., high dominance of abundant species (or groups of taxa) that reflect a perturbed fauna, which resulted from the global extinction associated to the PETM. The most abundant species in this interval may have had an opportunistic behavior and were tolerant to disturbed environments. In the middle Ypresian interval, benthic foraminiferal assemblages reflect a recovering fauna, as expressed in higher diversity and increasing numbers of common taxa, and less affected by environmental perturbations despite the repeated occurrence of hyperthermal events of lesser magnitude than the PETM. The increased diversity and equitability resulted from migration from shallower waters, return of Lazarus taxa from refugia, and evolution of new taxa. We propose the name 'Walvis Ridge fauna' for future reference to the early Eocene (Ypresian) deep-sea benthic foraminiferal assemblages, which lived in a greenhouse world with superimposed hyperthermal events, during a phase of recovery from the PETM extinction.
Supporting information S1 Appendix. Taxonomic list. Most common benthic foraminifera during the early Eocene and other species mentioned in the text.