Sightings of the Rare Globorotalia Cavernula in the Subantarctic South of Africa: Biogeochemical and Ecological Insights

In their extensive documentation of living planktic foraminifers, Bé (1967) writes, “we encountered one species that was unlike any other in our world-wide plankton collection.” Globorotalia cavernulaBé, 1967, is unique in its extremely deep, open umbilicus and overlapping final chambers that form a twisted test with crater- or cavern-like depression for which it is named (Fig. 1). Another unusual aspect of this species was the apparently restricted geographic distribution of its living populations to the Subantarctic Pacific (Bé, 1967; Brummer & Kučera, 2022). Most planktic foraminifer species (i.e., morphotypes, not necessarily genotypes) in the Southern Ocean have a circumpolar distribution and a Northern Hemisphere (Arctic or Subarctic) counterpart population (Bé, 1967; Darling & Wade, 2008; Schiebel & Hemleben, 2017; Lam et al., 2022). Yet, Bé (1967) found G. cavernula to be concentrated west of the Drake Passage, with a few sightings south of Australia and none from the Atlantic sector. Unknown to Bé (1967), this new species had already been documented by Boltovskoy (1966) in the high-latitude westernmost South Atlantic but was not named due to the limited number of specimens (Boltovskoy, 1969; see  Appendix A for translation of Boltovskoy, 1966). In Bé's net tows, G. cavernula was sufficiently abundant to support the new species designation: occurring at 29 sites and contributing >5% of the total planktic foraminifer assemblage at six of them (Bé, 1967). Since its discovery, G. cavernula has been found at only five sites in the modern ocean (Table A1): two net tows in the Indian Subantarctic (Boltovskoy, 1969), two net tows in upwelling regions outside of the Southern Ocean (one in the northeast Atlantic off Senegal, Cifelli & Benier, 1976; and one in the eastern equatorial Pacific off Panama, Fairbanks et al., 1982), and a single sediment trap south of New Zealand (Northcote & Neil, 2005).

Globorotalia cavernula is thought to have evolved during the Pleistocene from Globorotalia truncatulinoides d’Orbigny, 1839 (Kennett & Srinivasan, 1983; Aze et al., 2011). Kennett & Srinivasan (1983) noted that Globorotalia petaliformisBoltovskoy, 1974 (found in Middle-Late Miocene sediments of the southeast Indian Ocean), and Globorotalia crozetensisThompson, 1973 (from Pleistocene sediments of the Crozet Basin), are closely related forms to G. cavernula. The latter has since been reclassified as a variant of G. cavernula (unaccepted, subjective junior synonym; Brummer & Kučera, 2022). It has further been suggested by Brummer & Kučera (2022) that G. cavernula is not its own species but instead one of the known cryptic species of G. truncatulinoides (de Vargas et al., 2001; Quillévéré et al., 2013), or possibly aberrant juvenile G. truncatulinoides (addendum to Hornibrook, 1982).

Here, we report new sightings of live-collected G. cavernula from net tows in the African sector of the Southern Ocean and combine these with existing records to investigate its modern and past distributions. We take advantage of developments in high-resolution microscopy and chemical microanalysis since the original discovery of G. cavernula to characterize its shell ultrastructure and trace-element composition. Femtosecond laser ablation–inductively coupled plasma–mass spectrometry (fs-LA-ICP-MS) allows us to quantify the trace element ratios of individual chambers and, thus, investigate changes in physiology, depth habitat, and biomineralization during adult ontogeny. Through comparison of these inter-chamber trends and whole-shell averages with fellow Globorotalia (including G. truncatulinoides) and other species, we seek to glean new insights into the ecology and habitat of this rare species.

Living planktic foraminifers were collected from a total of 14 tows during two voyages of the R/V S.A. Agulhas II in the Southern Ocean south of Africa (Smart et al., 2020). The first voyage was in winter in the Atlantic sector (July–Aug 2015; VOY016), and the second was in late summer in the Indian sector (April–May 2016; VOY019). Plankton were sampled from the upper water column (25–90 m) using a double 1-m²-opening, 200-μm-mesh bongo net, towing at 1–1.5 knots for ∼90 min. Before each tow, hydrographic profile data were collected using Sea-Bird conductivity-temperature-depth (CTD), florescence, and oxygen sensors mounted to a rosette water sampler. Tows targeted the deep chlorophyll maximum (DCM) or the middle of the mixed layer if no DCM was present. Bulk tow material was preserved in a 5–10% formalin‐seawater solution (pH-buffered with sodium borate) and refrigerated at 4°C until processing (following Ren et al., 2012). Foraminifers were separated from the bulk material by wet sieving and density separation (addition of a 200 g/L NaCl solution), then rinsed with deionized water and dried in a fume hood (for more details see Smart et al., 2020).

Globorotalia cavernula specimens, identified by their distinctive morphology (Fig. 1), were separated under a microscope using a wet-picking brush. Six of these shells were mounted at various orientations on carbon tape and gold coated for SEM imaging using a Zeiss LEO 1530 Field Emission SEM at the Max Planck Institute for Chemistry (MPIC) in Mainz, Germany. Backscattered electron images were taken using an accelerating voltage of 15 kV at a working distance of 11.2–11.3 mm, and secondary electron images were taken using 10 kV at 9.5 mm. Pore density was estimated from the spiral side to minimize curvature and using the penultimate chamber to avoid potential irregularities seen in final chambers (Constandanche et al., 2013; Burke et al., 2018).

One of the six uncleaned specimens (which was from tow M6; Table 1) was selected for high-resolution, single-shot fs-LA-ICP-MS analysis of the spiral side at the MPIC, following the method of Jochum et al. (2019). Spot sizes of 30 μm were analyzed at low fluence of 0.7 J/cm² and low pulse repetition rate (1 Hz and 100% energy for Mg/Ca, 5 Hz and 30% energy for the rest), using a 200-nm wavelength ESI NWRFemto laser system coupled with a SF Thermo Element2 ICP-MS (Jochum et al., 2014). Pre-ablation was done for surface cleaning before each measurement and, where possible, each chamber was measured more than once. A representative, stable portion of each ablation profile was selected in order to exclude any surface contamination and artifacts of beam penetration (Fig. A1). International reference materials from the USGS (MACS-3 synthetic calcium carbonate) and NIST (SRM 610 and 612 silicate glasses), were used for calibration and quality control. Data were corrected as described in Jochum et al. (2019). Repeated measurements of MACS-3 yielded relative standard deviations of 14% for Mg, 11% for Sr, 12% for Na, 11% for Mn, and 18% for Ba (n = 9 measurements in each case), but larger for B at 54% (n = 7). A seventh specimen (tow W3; Table 1) underwent MilliQ water and ethanol rinses, oxidative cleaning using a buffered H₂O₂ solution and a brief (30-second) weak HNO₃ leach, following the protocol of Repschläger et al. (unpublished data, in review) modified from Barker et al. (2003). This cleaned specimen was measured by fs-LA-ICP-MS in the same way as described above, yielding two laser spots per chamber. In both cases, the final chambers (F-0) did not yield useable results. Uncleaned F-0 was too thin to yield reliable data, and cleaned F-0 broke off during cleaning. Due to the use of NaCl in the density separation step, the Na/Ca data are not interpreted.

For geographic and stratigraphic context, previously reported sightings of G. cavernula were compiled (Table A1), including occurrences in the water column (i.e., from net tows and a single sediment trap) and seafloor sediments (from recent to Eocene age). Existing databases provided a foundation for this compilation, particularly ForCenS for surface sediments (Siccha & Kučera, 2017) and FORCIS for planktonic foraminifers from the water-column (Chaabane et al., 2022, 2023; de Garidel-Thoron et al., 2022). These records were supplemented with occurrences from the literature, the PANGAEA online repository, and ODP/DSDP cruise reports. Searches were conducted in October 2022 using Google Scholar, GBIF, EOL (which includes records from the Smithsonian NMNH), and one record from the Neptune database (based on Lazarus, 1994; Spencer-Cervato, 1999). Where available, Table A1 includes information on the abundance of G. cavernula in each collection and indicates whether or not photographic evidence (or illustrations) could be found to support the identification. Globorotalia crozetensis (Thompson, 1973; reclassified as G. cavernula) and closely related G. petaliformis (Boltovskoy, 1974) were also included in the compilation.

Globorotalia cavernula specimens were found in four of the 14 tows (two in winter, two in late summer), constituting ≤0.5% of total foraminifers in all cases (Table 1). These four tows, with target depths of 25–80 m, span a latitude range of 42–46°S (red diamonds, Fig. 2), straddling the Subantarctic and Polar Frontal Zones (the latter zone bounded to the south by the Polar Front at ∼51°S). On average, at the depth of collection (51±24 m), water temperature was 7.6±1.8°C and salinity was 34.0±0.2 (n = 4). The tow with the most G. cavernula specimens (i.e., ten at site W3) came from the wintertime Atlantic sector at 46.0°S, 5.5°E at a depth of 80 m, where in situ temperature was 5.4°C, and salinity was 33.9 (Table 1).

The latitude range of our G. cavernula-containing tows and the associated seawater properties are largely consistent with the observations of Bé (1967) in the Pacific sector of the Southern Ocean (green diamonds, Fig. 2). There, G. cavernula were found in a ∼6° wide belt just north of the Polar Front. While present in tows as deep as 500–1000 m, this species was most commonly found in their 0–250 m tows, where temperatures ranged from 3.3 to 12.5°C. Their highest concentrations of G. cavernula (>50 individuals/1000 m³ water) were at temperatures of 4–6°C and salinities of 34.2–34.3 (Bé, 1967). The slightly warmer and fresher values accompanying our Indo-Atlantic G. cavernula may be due to the shallower focus of our sampling (≤90 m).

The absence of G. cavernula from their Atlantic sector tows as far east as 10°E led Bé (1967) and subsequent authors (e.g., Brummer & Kučera, 2022) to believe that this species is limited to the region west of Drake Passage. However, this was likely due to their sampling latitude (50–70°S) being largely south of the Polar Front in the Atlantic sector. Indeed, the lesser known Boltovskoy (1966) sightings of G. cavernula (that predated its naming) were in Subantarctic waters of the westernmost South Atlantic (off Argentina, September 1963). However, given their proximity to Drake Passage and the prevailing flow of the Malvinas Current, the specimens found off the coast of Argentina may have been carried there from the Pacific. Our observations from the African sector (covering 41–54°S, 0–38°E) more definitively expand the known modern range of the species and advocate for a circumpolar distribution. In the Pacific sector, G. cavernula have most frequently been found in October and November (Bé, 1967, tows; Northcote & Neil, 2005, sediment trap). Future sampling in austral spring should clarify whether this seasonality holds for the African sector as well.

Compiled data from seafloor sediments show that G. cavernula has been found at a broad range of low- to mid-latitude sites but rarely in the Southern Ocean and never in the (sub) Arctic (Table A1 and references therein). In almost all cases (where abundance is known), this species makes up a very small percentage (<0.5%) of the total population (i.e., reported as ‘rare’ to ‘isolated’; Table A1). Most fossil occurrences are either adjacent to the Southern Ocean (off New Zealand, Walvis Ridge etc.) or in tropical-subtropical upwelling zones (Eastern Equatorial Pacific and Canary Currents). Yet several occurrences are outside of these more temperate waters, seemingly at odds with their modern (water-column) and recent (sediment) distributions in cooler waters. While some of these sightings could be explained by seasonal upwelling (e.g., Holocene sediments from the Cariaco Basin, recent sediments from the East China Sea), others cannot. At these warmer, non-upwelling sites (e.g., Pleistocene sediments from the North Brazil Current and Gulf Stream), G. cavernula may have been transported from cooler habitats by these strong currents and their eddies. Other possibilities include a deeper primary depth habitat for G. cavernula in these environments, akin to G. truncatulinoides (e.g., Healy-Williams et al., 1985), and an expanded or shifted geographic range during the Pleistocene associated with shifting hydrography (equatorward migration of the Southern Ocean fronts and contracting subtropical gyres). Broader implications for G. cavernula are discussed under ‘Ecological and Evolutionary Implications’.

Globorotalia cavernula tests are low trochospiral coiling, keeled, and asymmetrically biconvex (more convex on the umbilical side, flatter on the spiral side). Sutures are curved and somewhat step-like on the spiral side (resembling a gentle spiral staircase; Fig. 3.1b), and more radial on the umbilical side. The umbilicus is deep and wide with an angular shoulder (rim of the “crater”). In some specimens, the final chambers of the adult test increasingly protrude out of the coiling plane (Fig. 3.3). The keel also becomes less distinct to absent in the final two chambers. The interiomarginal aperture is arched and bordered by a clearly defined lip. On average, the maximum diameter of the specimens examined under SEM is 313±44 μm (n = 6; excluding the cleaned shell where chamber F-0 broke off). They have three whorls, with 4.5–5.5 chambers in the final whorl. The coiling direction for all specimens found (n = 14) is sinistral.

The walls of G. cavernula tests are thin (∼2–7 μm thick in chamber F-0; Fig. 3.4), finely macro-perforate and relatively smooth, particularly on the spiral side where pustules are fairly small and concentrated near the keel. There, the largest pores (but also the least dense) occur on the innermost (earliest) whorl, where pustules appear to have merged. On the umbilical side, pustules increase in size and number from the earliest to the most recent chambers and are particularly noticeable at the base of the aperture (Fig. 3.3) and along the umbilical shoulder. The largest pores of the umbilical side seem to be along the sutures (Fig. 3.3). Pore density (estimated from the spiral side, penultimate chamber; Fig. 3.1a) is approximately 0.021 pores/μm². All specimens contained cytoplasm at the time of collection, and none have a secondary calcite crust. The square, flat crystals visible on some SEM images (Fig. 3.1b) are residual NaCl from the density separation step.

In general, SEM imaging of our specimens confirms the characteristic morphology and surface features of G. cavernula described by Bé (1967). The degree of twisting and chamber overlapping (i.e., imbrication) varies among specimens, in both our collection and that of Bé (1967), making some specimens more difficult to identify than others. In the case of the G. cavernula holotype, the strongly twisted appearance is likely enhanced by its kummerform final chamber (as depicted in Fig. 1). However, a kummerform chamber is not required to produce this coiling trait, as demonstrated by Fig. 3.3 (and Fig. A2) where the final chambers of a non-kummerform specimen “overshoot” a regular trochoidal coiling mode. Our specimens are slightly smaller than those measured by Bé (1967), having 4.5–5.5 chambers in the final whorl instead of 5–6, and maximum diameters of 313±44 μm (n = 6) compared to their 326±93 μm (n = 6). In sediment samples, G. cavernula is typically smaller than (adult) G. truncatulinoides (Hornibrook, 1982; Hayward, 1983). The same is true in our net tow samples, where G. cavernula is, on average, ∼20 μm smaller in diameter than G. truncatulinoides in the same collection (based on the three tows where data are available, containing 196 measurements of G. truncatulinoides; Smart et al., 2020).

SEM imaging also adds new details on the ultrastructure of G. cavernula. For example, the pore density of G. cavernula (0.021 pores/μm²) is fairly typical for Globorotalia (all of which are non-spinose and macroperforate), which normally range from 0.01 to 0.03 pores/μm² (averaging 0.016±0.007 pores/μm², n = 142, based on six Globorotalia species from a range of tropical to mid-latitude core-top sites; Burke et al., 2018). For comparison, G. truncatulinoides has an average pore density of 0.017±0.006 pores/μm² (n = 45) in that same study. With a larger dataset, it might be possible to distinguish between G. cavernula, G. truncatulinoides, and its pseudo-cryptic species (i.e., genotypes) on the basis of pore density, in the same way that porosity has been used to distinguish between the two genotypes of Globigerinella siphonifera d’Orbigny, 1839 (Huber et al., 1997), and among the three genotypes of Orbulina universa d’Orbigny, 1839 (Morard et al., 2009; Marshall et al., 2015). More likely to be diagnostic is whole-test outline analysis (based on edge and umbilical views), which has been shown to correctly assign G. truncatulinoides specimens to one of the two main genotype clusters 75% of the time (Quillévéré et al., 2013). For example, in a lateral-view comparison, even the most axially compressed (i.e., least conical) forms of G. truncatulinoides (Types III and IV) have a much less pronounced umbilical shoulder (“crater rim”) than G. cavernula. Other features that typically set G. cavernula apart from G. truncatulinoides are its more strongly arched aperture in the final chamber(s) (in lateral view), and its more gradual chamber-size increase (as seen in lateral and spiral views).

A pervasive feature of G. truncatulinoides in sediments is a secondary calcite crust that forms deeper in the water column before gametogenesis (e.g., Duckworth, 1977; Hemleben et al., 1985; Reynolds et al., 2018). While it is not surprising that live-caught G. cavernula lack a gametogenic crust, it is notable that such a crust, so far, appears to be absent from fossil (downcore) specimens too (Hornibrook, 1982; Hayward, 1983). Future collections and/or trace element analysis of fossil G. cavernula could verify if this reflects a difference in mean depth habitat, life cycle (e.g., depth migration for reproduction), or biomineralization between the two species.

In the uncleaned specimen, most trace element-to-calcium ratios (TE/Ca) are unrealistically high, likely due to the presence of cytoplasm and other non-bound organics on the live-caught specimen, so we do not attempt to interpret these data (Table 2, top). The cleaned specimen yielded TE/Ca within literature ranges for all elements (Table 2, bottom), although Na/Ca is higher than expected (8.2±1.1 vs. ∼4–7 μmol/mol in culture (Allen et al., 2016; Bertlich et al., 2018) and core-tops (Zhou et al., 2021)) for shells outside of the highly saline Red Sea (>7 μmol/mol; Mezger et al., 2016). Our elevated Na/Ca is likely due to the concentrated NaCl solution used in the density separation step and is not discussed further.

The Mg/Ca ratio is a well-established paleothermometer, where species-specific relationships are used to infer the temperature at the time and depth of shell calcification (Erez & Luz, 1983; Nürnberg et al., 2000; Jentzen et al., 2018). Globorotalia cavernula has a relatively low Mg/Ca (2.1±0.7 mmol/mol), consistent with a cool-water environment, but not as low as predicted based on some other Globorotalia (G. truncatulinoides; Reynolds et al., 2018; Globorotalia inflata d’Orbigny, 1839; Cléroux et al., 2008; grey and purple, respectively, in Fig. 4). Our measured Mg/Ca is within the range of surface tow-caught Globigerina bulloides d’Orbigny, 1826 from the same area as ours (and thus temperature conditions; ∼1.5–3 mmol/mol at 5–13°C; Martínez-Botí et al., 2011; green diamonds in Fig. 4), as well as North Atlantic core-top G. bulloides (∼1.8–3.5 mmol/mol; green line in Fig. 4) and deeper-dwelling Globorotalia (∼1.1–2.1 mmol/mol), but at a higher temperature range (10–18°C) than our site (Cléroux et al., 2008). The Mg/Ca of G. cavernula is also higher than Neogloboquadrina pachyderma Ehrenberg, 1861 (core tops) and Neogloboquadrina incompta Cifelli, 1961 (tows and core tops) under similar temperature conditions to ours (Martínez-Botí et al., 2011; Tierney et al., 2019; data not shown). Part of the Mg/Ca elevation in G. cavernula may be due to a methodological difference. The literature values are mostly from solution-based analyses (combining whole shells), which can differ from laser-derived (single shell, multi-chamber) averages. For Mg/Ca in G. truncatulinoides, Reynolds et al. (2018) find the last three chambers to be representative of whole-shell Mg/Ca. This would yield a lower Mg/Ca of 1.7±0.1 mmol/mol (n = 2, as F-0 is missing) for G. cavernula, more consistent with the cold waters where it was collected (light blue circle in Fig. 4; values in brackets in Table 2). But even with this correction, G. cavernula remains above the typical Mg/Ca-to-temperature relationship for G. truncatulinoides (which predicts ∼1.1 mmol/mol for our 5.4°C waters; Reynolds et al., 2018; grey line in Fig. 4) and also above Neogloboquadrina, as seen previously for G. bulloides (Cléroux et al., 2008; Tierney et al., 2019) and to some extent, Globorotalia hirsuta d’Orbigny 1839 (Elderfield & Ganssen, 2000; not shown here). An important caveat is that very little Mg/Ca data exist for G. truncatulinoides in high-latitudes environments, such that the cold-water Mg/Ca data in temperature calibrations (grey triangles in Fig. 4) derive solely from deep-water calcification (i.e., gametogenic crust formation). Thus, it is quite possible that pre-gametogenic G. truncatulinoides living and calcifying within the Subantarctic mixed layer have a higher Mg/Ca, more similar to G. cavernula.

The Sr/Ca ratio is not strongly affected by temperature or salinity but may respond to large changes in dissolved inorganic carbon or the Sr/Ca of seawater (Lea et al., 1999; Martin et al., 2000; Allen et al., 2016). The Sr/Ca of G. cavernula (1.36±0.07 mmol/mol; Table 2) is most similar to core tops from the North Atlantic (G. bulloides, G. truncatulinoides, and G. inflata; Cléroux et al., 2008) and (sub)tropical western Pacific (Trilobatus sacculifer Brady, 1877; Zhou et al., 2021). In the context of the Subantarctic, our Atlantic sector G. cavernula is intermediate in Sr/Ca between core-tops from the Indian and Pacific sectors (1.27 mmol/mol for G. bulloides and 1.42 mmol/mol for N. pachyderma, respectively; Martin et al., 2000). Compared with net tows from the Cariaco Basin, G. cavernula is at the lower end of the Sr/Ca ranges for both deep and shallow dwellers (Davis et al., 2020), perhaps related to the higher Sr/Ca of upper-ocean seawater in that region (Lebrato et al., 2020).

The remaining trace element ratios (B/Ca, Mn/Ca, and Ba/Ca) are insensitive to temperature and unaffected or weakly affected by salinity (Hönisch et al., 2011; Allen et al., 2012, 2016). The B/Ca ratio (68±11 μmol/mol), a potential proxy for pH or phosphate (Henehan et al., 2015), is most similar to G. inflata in core tops from the Southern Ocean (near our tows), North Atlantic and South Pacific (∼50–80 μmol/mol; Yu et al., 2007), as well as to O. universa in culture (56–92 μmol/mol for pH range 7.6–8.7) and in core-tops from the Gulf of Mexico (71–76 μmol/mol; Allen et al., 2011, 2012).

In the shells of living foraminifers, Mn/Ca has been found to vary with the [Mn] of seawater (Munsel et al., 2010), which is higher near terrestrial sources and in low-oxygen, remineralization zones (Klinkhammer et al., 2009; Steinhardt et al., 2014). With a Mn/Ca of 8.5±7.3 μmol/mol, G. cavernula falls within the broad range of field measurements (∼0.3 to >100 μmol/mol; Davis et al., 2020, tows; Reynolds et al., 2018, traps; Jonkers et al., 2012, core tops) but at the lower end, more similar to shallow- than to deep-dwellers in low-latitude settings (e.g., ∼1–15 μmol/mol in Mozambique Channel traps; Steinhardt et al., 2014). This is consistent with the well-oxygenated (6.3 ml/l at W3; Table 1), low-[Mn] (∼0.2–0.5 nM; Boye et al., 2012; van Hulten et al., 2017), open-ocean waters where G. cavernula was collected.

Similarly, Ba/Ca has been used as a tracer of deep water masses and freshwater runoff in which Ba is enriched (Hall & Chan, 2004; Weldeab et al., 2007; Hönisch et al., 2011). At the same time, the Ba/Ca of non-spinose species is typically higher than predicted based on ambient seawater (Lea & Boyle, 1991), leading to suggestions that they consume Ba-rich prey (Bahr et al., 2013) or inhabit a Ba-enriched marine snow environment (Fehrenbacher et al., 2018; Richey et al., 2022). The [Ba] of seawater decreases northwards across the Southern Ocean surface, spanning ∼40–60 nM in Subantarctic waters (Jeandel et al., 1996; Hsieh & Henderson, 2017). The Ba/Ca of G. cavernula (2.3±0.5 μmol/mol) is lower than both spinose and non-spinose species from tows in the upwelling- (Ba remineralization-) influenced Cariaco Basin (Davis et al., 2020), where seawater [Ba] is higher (45–85 nM; Falkner et al., 1993). The Ba/Ca of G. cavernula is more similar to but at the lower end of Globorotalia (including G. truncatulinoides) from tows and traps in the open North Atlantic (∼2–13 μmol/mol), where seawater [Ba] is lower (∼40 nM; Lea & Boyle, 1991). The lower Ba/Ca of G. cavernula in the Subantarctic, despite the relatively high ambient [Ba] and being a non-spinose species, might indicate that it does not inhabit marine snow or have a particularly Ba-enriched diet in this environment.

The Sr/Ca ratio is the most stable of the ratios measured, varying by <5% between chambers (Fig. 5; Table 2). This is in line with previous work, showing Sr/Ca to be fairly invariant within individual shells (Anand & Elderfield, 2005; Hathorne et al., 2009; Dueñas-Bohórquez et al., 2011; Jonkers et al., 2012). In contrast, Mg/Ca decreases substantially from peak values in chamber F-4 until F-2, with a slight increase into the penultimate chamber, F-1 (Fig. 5; Table A2). A similar pattern has been observed in G. truncatulinoides, but with a continued increase into the final chamber, F-0 (not measured here; Anand & Elderfield, 2005). In most other non-spinose species (e.g., Globorotalia menardii Parker, Jones & Brady, 1865; Neogloboquadrina dutertrei d’Orbigny, 1839), Mg/Ca has been found to increase through the last several chambers and attributed to the outward thinning of their low-Mg/Ca crusts (Steinhardt et al., 2015; Fehrenbacher et al., 2017; Jochum et al., 2019; Davis et al., 2020). Such a crust is absent from our specimens, so the Mg/Ca trends through the final whorl of G. cavernula may reflect temperature changes during the adult phase (e.g., downward migration in the water column or even summer-to-winter cooling; from F-4 to F-2). However, chamber-to-chamber trends in Mg/Ca (of both crust and ontogenetic calcite, i.e., inner layer) are surprisingly consistent across non-spinose species from vastly different environments, with different mean depth habitats and expected behaviors; including deep-dwelling G. truncatulinoides from the mid-latitude North Atlantic (Anand & Elderfield, 2005; Fig. 6A) and thermocline-dwelling N. dutertrei from the tropical Mozambique Channel (Steinhardt et al., 2015; Fig. 6B). This consistency suggests that a common physiological mechanism or biomineralization strategy among non-spinose species (including G. cavernula, at least before crust formation if it occurs), is more likely than a seasonal trend specific to the timing of our net tows.

For Ba/Ca, the relative chamber-to-chamber trends in G. cavernula (Fig. 5) closely match those seen in non-encrusted G. truncatulinoides from the Gulf of Mexico (not shown; Richey et al., 2022), although the absolute values are much lower in our specimen. In the ontogenetic calcite of G. truncatulinoides, Ba/Ca generally decreases through the outer whorl, hypothesized to reflect adult specimens being less deeply buried in marine snow than juveniles (Richey et al., 2022). Another possible explanation, that does not require marine snow, is the decreasing number of Ba-enriched bands in more recently formed chambers (with layers being sequentially added to the whole shell during new chamber formation). The Mn/Ca of G. cavernula is particularly variable (RSD = 87%; n = 3), but similar in absolute values to the ontogenetic-specific Mn/Ca in N. dutertrei from core-tops in the Mozambique Channel, which also peak in chamber F-3 (Jonkers et al., 2012). Lastly, B/Ca shows no coherent trends between chambers.

While it is possible that the intra-shell patterns in TE/Ca discussed here reflect systematic effects of biomineralization (e.g., calcite layering, chamber formation), environmental conditions, or depth migration during life, many more shells would be needed to assess these effects. Indeed, inter- and intra-chamber trends in the ontogenetic calcite of much more intensively studied foraminifera have been documented but have yet to be decoded (e.g., Jonkers et al., 2012; Steinhardt et al., 2015; Reynolds et al., 2018).

By inhabiting both (sub)polar and low-latitude upwelling zones, the modern ocean distribution of G. cavernula is not unlike other cool–cold water species (e.g., N. pachyderma and N. incompta, G. bulloides Type II). However, its absence from the northern high latitudes (Subarctic) in both modern and paleo-distributions (Fig. 2) is more reminiscent of G. truncatulinoides (Bé & Tolderlund, 1971). Globorotalia truncatulinoides emerged around 2.8–2.3 Ma (Lazarus et al., 1995; Spencer-Cervato & Thierstein, 1997) but only diversified into cool-water genotypes (III and IV) that colonized the Southern Ocean fronts in the Late Pleistocene, around 300 ka (de Vargas et al., 2001). As explained by Darling & Wade (2008), the asymmetric distribution of G. truncatulinoides between the two hemispheres might be related to its relatively recent evolution but is likely also affected by differences in water-column structure and deep/intermediate water-mass properties between north and south (given the deep habitat of this species). However, the low abundances of G. cavernula make it inherently difficult to distinguish chance expatriation by currents or eddies from its actual adaptive range. Thus, the comparisons and interpretations made here should be considered hypotheses to be tested by future work.

If representative, our compilation suggests that the geographic range of G. cavernula is more restricted today than it was in the recent past. The colder climate of the Pleistocene likely made larger expanses (or different areas, e.g., subtropical gyre margins) of the low-latitude surface ocean hospitable to cool-water species like G. cavernula, as seen for N. pachyderma, G. bulloides, and Turborotalita quinqueloba Natland, 1938 (Yamasaki et al., 2021). Similarly, the equatorward shifts of the Southern Ocean fronts during the ice ages that characterize the Pleistocene might contribute to the paucity of occurrences (i.e., only one site) in Pleistocene-aged sediments south of today’s subtropical front. However, this does not explain the absence of G. cavernula from recent/Holocene sediments in the Southern Ocean. One possibility is that G. cavernula only very recently colonized the Subantarctic (i.e., since the last glacial period) and has not accumulated in sufficient numbers to stand out in core-tops or surface sediments. Perhaps it is more likely that G. cavernula continues to be mistaken for (or considered a variant of) G. truncatulinoides or a benthic species (e.g., of the genus Gavelinopsis Hofker, 1951). The two planktic species are probably more difficult to tell apart in Southern Ocean sediments due to the smaller and more axially compressed forms of G. truncatulinoides (Types III and IV) that occur there, i.e., compared to the larger, more conical subtropical tests. Furthermore, with the standard protocol of picking 300 specimens for census counts (CLIMAP, 2009), the typical abundance of ≤0.5% would amount to <1–2 G. cavernula specimens in a sample and could be easily overlooked.

The single sighting of G. cavernula from the Eocene (square in Fig. 2; simply noted as ‘present’ by Toumarkine et al., 2005) would imply that G. cavernula pre-dated G. truncatulinoides. However, this record is not accompanied by a photograph or illustration, so we cannot verify the identification. The Miocene occurrences of G. petaliformis (x’s in Fig. 2) are photo-documented and were even classified as ‘common’ to ‘abundant’ in Middle Miocene samples (Boltovskoy, 1974; Table A1). While G. petaliformis and G. cavernula have some morphological features in common (e.g., the “spiral staircase” chamber arrangement, generally stronger in G. petaliformis), their peripheral outlines are quite different (more lobate in G. petaliformis). The lack of G. petaliformis records since its naming makes it difficult to evaluate the relationship between the two species. We found no Pliocene records of either species that could bridge the temporal gap between the last (most recent) G. petaliformis and the first (oldest) photo-documented G. cavernula (Pleistocene; Krasheninnikov et al., 2005).

Previous authors have hypothesized that G. cavernula is a cryptic species or aberrant juvenile form of G. truncatulinoides (addendum to Hornibrook, 1982; de Vargas et al., 2001; Quillévéré et al., 2013). Based on our initial pore density measurements, G. cavernula is not statistically different from G. truncatulinoides. However, it is not uncommon for different Globorotalia species’ pore density ranges to overlap (Burke et al., 2018), so this finding is not conclusive. In terms of gross morphology, G. cavernula has several features that distinguish it from all known morphotypes of G. truncatulinoides, particularly its deep, crater-like umbilicus and twisted, overlapping arrangement of chambers. In smaller specimens with more subtle imbrication (i.e., less step-like; more trochoidal coiling), the more strongly arched aperture of G. cavernula can be a useful diagnostic.

Ultimately, molecular genetics is needed to confirm the relationship between G. cavernula and G. truncatulinoides. Whether or not the two are separate biological species, comparison with the more abundant and better-studied G. truncatulinoides could provide insight into the ecology or behavior of G. cavernula. For example, the similarity in inter-chamber TE/Ca trends between G. cavernula and G. truncatulinoides, for both encrusted (as for Mg/Ca and Sr/Ca in Anand & Elderfield, 2005) and non-encrusted shells (as for Mg/Ca in Reynolds et al., 2018; and Ba/Ca in Richey et al., 2022), may hint at a similar progression (e.g., changing ecological demands) during the recorded phase of development (i.e., adult ontogeny; Fig. 6). At the same time, the substantially lower Ba/Ca for G. cavernula than in other non-spinose species (including G. truncatulinoides) from a range of environments cautions against assuming analogous behavior (e.g., marine snow habitat) for G. cavernula.

Based on the absolute values, it is tempting to speculate that the higher-than-predicted Mg/Ca of G. cavernula (for its low-temperature setting) suggests a shallower habitat in the water column than G. truncatulinoides, or at least that G. cavernula does not follow the species-specific temperature relationship of G. truncatulinoides. If robust (i.e., confirmed by additional G. cavernula measurements), it could argue for a separate species designation for G. cavernula. However, the lack of Mg/Ca data on non-encrusted G. truncatulinoides (grey circles in Fig. 4) from the high-latitude (i.e., cold) mixed layer and on (sub)fossil G. cavernula that have completed their life cycle prevent a fair comparison. Perhaps more compelling is the lack of crust in G. cavernula seen in sediments so far, which could suggest that, unlike G. truncatulinoides, its life cycle is confined to the mixed layer.

The modern distribution of G. cavernula in the Southern Ocean appears to be circumpolar, not limited to the region west of Drake Passage as commonly thought (Bé, 1967; Brummer & Kučera, 2022). The fact that both of our relatively limited sampling campaigns yielded G. cavernula (albeit in low numbers), and in seasons so far thought to be outside of its prime, suggests that this species may not be as rare as it was perceived to be. Instead, the apparent rarity of G. cavernula in the modern ocean (i.e., in the water column) may be due to the paucity of net tow collections in the Southern Ocean. The absence of G. cavernula records from recent/Holocene sediments across the Southern Ocean might be a result of it being grouped with G. truncatulinoides, which are smaller and less distinctively conical at these latitudes (G. truncatulinoides Types III and IV). Under-reporting of G. cavernula occurrences could be exacerbated by it being an impractical target for most geochemical and paleoceanographic analyses that require large numbers and favor species with continuous representation downcore, or simply a lack of awareness of its existence when picking. However, such biases would be expected to apply to downcore picking of older sediments as well. Thus, we propose that the larger number and more widespread distribution of G. cavernula records in Pleistocene-aged sediments reflects a true range expansion or geographic shift of the species under a colder Pleistocene climate.

We thank Dee Breger, who made the original line drawings of G. cavernula. If not for her detailed and memorable illustration of the holotype (Fig. 1), G. cavernula would likely have escaped our notice as well. Also, thanks to the Journal of Foraminiferal Research for permission to re-print these drawings, originally published in Bé (1967). We are indebted to Raphaël Morard who attempted molecular genetic analysis on our serendipitously collected G. cavernula. We thank Catherine V. Davis for insightful discussions on the trace element data and George Scott for sharing his expert eye for morphology and ideas on speciation. This manuscript was also improved by inputs from an anonymous reviewer. Thank you to Bruce Hayward for pointing us in the right direction on our search for previous sightings of G. cavernula, and to Chiara Cappelli for assisting with the translation of Boltovskoy (1966;  Appendix A). This work was supported by the South African National Research Foundation (Grants 111090 and 120714) and the Max Planck Institute for Chemistry (MPIC) in Mainz, Germany (S.M.S.). Thank you to Jan Leitner at the MPIC who performed the SEM analyses. We are grateful to the Department of Environmental Affairs, South Africa, for providing equipment and technical support for net tow deployments, as well as the to the captain and crew of the R/V S.A. Agulhas II in winter 2015 and late summer 2016 for safe and productive voyages. Data are available online on the PANGAEA data repository (https://doi.org/10.1594/PANGAEA.962773) and in Appendix Tables A1 and A2, which can be found linked to the online version of this article.

Table A1. Compiled record of previous occurrences of G. cavernula (and closely related forms). Data available online at https://doi.org/10.1594/PANGAEA.962773 and linked to the online version of this article.

Table A2. Individual-chamber trace element/calcium data for the cleaned G. cavernula specimen from net tow W3. Table A2 can be found linked to the online version of this article.

Here we provide the relevant excerpt (p. 21–22) of Boltovskoy (1966), “La zona de convergencia subtropical/subantártica en el Océano Atlántico: (parte occidental); un estudio en base a la investigación de foraminíferos-indicadores” published by Argentina's Secretaría de Marina, Servicio de Hidrografía Naval. Note that the text in square brackets in the translation are our own additions for clarity.

20. Globorotalia sp. “A”

This species has only been found in the material of expedition “Productivity I”. In all the consulted literature, I could not find any species to which the found specimens could be assigned. It is morphologically highly variable. It consists of approximately 2½–3 whorls, the last one with 5–6 chambers, which are sometimes not well-aligned. The dorsal sutures are commonly (though not always) limbate. Thin walls, but clearly perforated and in some specimens covered with pustules similar to those of G. hirsuta. The most characteristic feature of this species is the large, deep, and clearly visible umbilicus.

Due to the shell architecture, the umbilicus, and other morphological features, the described specimens are very similar to Globigerina (Globotruncana) marginata turona Olbertz. However, the much more convex ventral side and the greater number of whorls in the latter do not allow for a corresponding identification. Additionally, it should be noted that Olbertz’s subspecies was found in the upper Cretaceous, while our material is recent.

The possibility that [the shape and size of] the umbilicus in the different species of Globorotalia is generated due to some ecological factor is not excluded, and in this case, the separation of these specimens as a [different] species is not justified, of course. Until this circumstance is further clarified, I prefer to leave the species in question in open nomenclatura.

Due to the fact that it is very rare and its [(i.e., this morphotype's)] taxonomic position is problematic, this species cannot be used as an [environmental/water-mass] indicator in the studied region.

20. Globorotalia sp. «A»

Esta especie ha sido hallada solamente en el material de la operación “Productividad I”. En toda la bibliografía consultada no pude encontrar ninguna especie a la cual podría adscribir los ejemplares encontrados. Es muy variable morfológicamente. Está compuesta de alrededor de 2½–3 vueltas, la última con 5–6 cámaras, las cuales a veces no están bien ordenadas. Las suturas del lado dorsal comúnmente (aunque no siempre) son limbadas. Paredes finas, pero nítidamente perforadas y en algunos ejemplares cubiertas con papilas del tipo de las de G. hirsuta. El rasgo más característico de esta especie es la gran concavidad umbilical, profunda y bien visible.

Por la construcción del caparazón, la concavidad umbilical y otros rasgos morfológicos, los ejemplares descriptos son muy parecidos a Globigerina (Globotruncana) marginata turona Olbertz. Sin embargo, el lado ventral mucho más convexo y el mayor número de vueltas en esta última, no permite la identificación correspondiente. Además hay que tomar en cuenta que la subespecie de Olbertz fue encontrada en el Cretácico superior, mientras que nuestro material es reciente.

No está excluida tampoco la posibilidad de que la concavidad umbilical se origina en las diferentes especies de Globorotalia debido a algún factor ecológico, y en este caso la separación de estos ejemplares como una especia, por supuesto, no es justificable. Hasta que esta circunstancia no se aclare más, prefiero dejar la especie en cuestión en la nomenclatura aperta.

Debido a la circunstancia de que es muy rara cuantitativamente y a que su posición taxonómica es problemática, esta especie no puede ser usada en la región estudiada como indicador.