The middle Smithian (Early Triassic) ammonoid Arctoceras blomstrandi: conch morphology and ornamentation in relation to stratigraphy

The ammonoid genus Arctoceras (Hyatt) occurs across all palaeolatitudes, and is a key genus for middle Smithian biostratigraphical correlations globally. In this study, intraspecific variations in conch morphology, ornamentation and allometry are examined in relation to stratigraphic position. Arctoceras is the most abundant ammonoid genus in the middle Smithian of Svalbard. Originally, seven Arctoceras species were described from Svalbard. Later, as the importance of intraspecific variation was recognized, six of the Arctoceras species from Svalbard were treated as junior synonyms of Arctoceras blomstrandi (Lindström). Yet, the variations in A. blomstrandi conch morphology remain poorly quantified and the dependence on stratigraphic position, unknown. We quantify the intraspecific variation in conch morphology, ornamentation and allometry in relation to stratigraphy of the Svalbard Arctoceras. The results support the assignment of all Arctoceras morphotypes from Svalbard to a single species A. blomstrandi. The new data allow for an updated species description and open the way for the use in biostratigraphy of the endmember morphology A. blomstrandi var. costatus. We document consistent changes in conch morphology and ornamentation in the studied stratigraphic interval, with a distinct shift towards more evolute and ornate conchs in the top of the interval. The trends in the strength of ornamentation are partly explained by covariation with conch morphology (Buckman's law). The most marked shift in the conch morphology and allometric development of A. blomstrandi coincides with the onset of the positive carbon isotope excursion at the end of the middle Smithian, but pre‐dates the mid–late Smithian cooling of the sea surface.

E A R L Y Triassic ammonoids are of particular interest because they offer a detailed record of large-scale ecosystem recovery after the end-Permian mass extinction (Brayard et al. 2006;Ware et al. 2015). The magnitude and rate of this recovery are among the largest in the Phanerozoic. Understanding ammonoid evolution and dispersal in this critical phase requires detailed taxonomic work in order to provide accurate information on biodiversity, and on palaeobiogeographical, biostratigraphical and phylogenetic relationships. Within the Early Triassic, the Smithian-Spathian boundary is presently receiving a heightened interest. The boundary is marked by the final stage of the late Smithian extinction event of ammonoid and conodont species, which is connected to large-scale changes in climate and sealevel. A global upper Smithian hiatus has made the temporal extent of the event difficult to assess (Galfetti et al. 2007;Brayard & Bucher 2015;Widmann et al. 2020). The event signified the end of a diversification-extinction cycle that extended throughout the entire Smithian. The initial diversification phase took place during the early Smithian, with a high origination rate at the family level. The middle Smithian interval was marked by a thermal maximum (Romano et al. 2013;Goudemand et al. 2019), along with the appearance of short-lived ammonoid species in the Tethys Realm (Br€ uhwiler et al. 2010(Br€ uhwiler et al. , 2011. A widespread and stepwise extinction unfolded throughout the late Smithian (Br€ uhwiler et al. 2010;Brayard & Bucher 2015;Jattiot et al. 2016), which brought Ammonoidea close to extinction. Moreover, a worldwide glacio-eustatic hiatus spanned the Smithian-Spathian boundary (Hammer et al. 2019), the duration of which was quantified in South China by Widmann et al. (2020).
The Early Triassic ammonoids known from Svalbard were to a large extent described and named long ago (Lindstr€ om 1865;Mojsisovics 1886;Frebold 1930). A coherent regional stratigraphy was not developed until 1965developed until (Buchan et al. 1965, and prior to this, species were described from material collected with limited stratigraphic control, sometimes only referable to the formation level. Combined with a prevailing typological approach, the poor stratigraphic resolution obscured intraspecific and ontogenetic variation, which could consequently lead to taxonomic over-splitting. Over-splitting at the species level very likely leads to mistakes in biostratigraphic and palaeobiogeographic studies as well as to faulty interpretations of trends in phylogeny, evolution and diversity. Furthermore, palaeoecological analyses will be affected by the erroneous inflation of species richness. A typological approach to species definition evidently also leads to problems in the estimation of the effects of extinction and diversification events. This was, for instance, addressed by Jattiot et al. (2016) for the cosmopolitan genus Anasibirites at the late Smithian extinction, in which only two species out of 60 in the literature proved valid. Intraspecific variation in ammonoids can affect most morphological traits, but the most dramatic variations are found in the degree of compression/depression of the whorl section and in the degree of involution/ evolution of the conch. As embodied in the so-called Buckman law of covariation (Buckman 1892, p. 313;Westermann 1966, p. 305;Hammer & Bucher 2005;Monnet et al. 2015), these coiling parameters tend to correlate with the strength of the ornamentation, with wide and evolute conchs usually being the most ornate. In addition, allometry (change in shape through growth) and variations in adult size can complicate the circumscription of sound morphological species.
In Svalbard, the genus Arctoceras, along with associated taxa, indicates strata of middle Smithian age. In terms of relative abundance, Arctoceras largely dominates the middle Smithian ammonoid fauna of Svalbard. Representatives of Arctoceras collected in Svalbard were originally assigned to seven species (as detailed below), until Kummel (1961) reassessed the type specimens in light of intraspecific variation. Based on the original material, Kummel concluded that all previously described species had overlapping characteristics and he regarded them as ontogenetic or morphological variants of one species, Arctoceras blomstrandi (Lindstr€ om, 1865). However, prior to the present study, Arctoceras was never collected bed by bed in Svalbard, and Kummel (1961) had no information with which to assess the intraspecific variation in relation to stratigraphic position. Therefore, a thorough revision based on new bedrock-controlled material is necessary to confidently determine which species of Arctoceras are present in Svalbard. Moreover, a reassessment of Arctoceras is an ideal case study to understand how morphological variants and/or species might vary throughout a middle Smithian succession.
For this study, we collected extensive material of Arctoceras from four sections in central Spitsbergen (Figs 1,2) to document the morphological variation of the genus through geological time.
Lindstr€ om, € Oberg and Mojsisovics described their species under the genus Ceratites, which Diener (1915) later assigned to Arctoceras (Hyatt, 1900), with Ceratites polaris as the type species. Lindstr€ om worked on Svalbard ammonoids that had been collected by Nordenskj€ old in 1864. Lindstr€ om described the new species Ceratites? blomstrandi in Latin (he found the lobes in the sutures to be inconclusive with regards to assigning genus). Lind-str€ om's specimens had been collected from 'shale and sandstone' at 'Midterhuken' and, as seen from his illustrations, they were highly damaged (Lindstr€ om 1865). Lind-str€ om's initial species description (Lindstr€ om 1865, p. 4  Oberg had material of fair preservation and he confirmed the species Ceratites blomstrandi as suggested by Lindstr€ om (1865). € Oberg described the largest specimens in his collection as being largely like Lindstr€ om's C. blomstrandi but having true nodes/tubercles on the umbilical shoulder. The nodes would often turn into ribs on the lateral shell, he observed, and these did not cross the venter. The material came from calcareous concretions in 'the marly shale' at 'Svarta Klyfta' ( € Oberg 1877). According to Buchan et al. (1965), 'Svarta Klyften' (or 'Klyfta' in Swedish) is a section at Kapp Thordson (Kapp Thordsen on the southern tip of Dickson Land; Norwegian Polar Institute 2003), and likewise is the 'Isfjord-Kolonie' of Mojsisovics (1886) and 'Midterhuk' of Lindstr€ om (1865). This locates all the original material in the same area of central Spitsbergen, north of Isfjorden. € Oberg (1877) went on to describe the new species Ceratites costatus based on a single individual found in 1872 at Svarta Klyfta. C. costatus largely resembles C. blomstrandi but has notable differences. C. costatus is described as having a comparatively rounded umbilical shoulder with no tubercles, although € Oberg suggested that they might develop at an older stage. The ribs are strong and straighter than in C. blomstrandi, and on the outer whorls they cross the venter, forming convex arcs. Near the aperture, the venter appears slightly 'bumpy' between the ribs when viewed laterally. According to € Oberg (1877), the suture lines of C. blomstrandi and C. costatus are very similar, although C. costatus is supposed to have two umbilical lobes, as opposed to the one of C. blomstrandi.
In 1886, Mojsisovics reassessed the Arctoceras material of Lindstr€ om and € Oberg and established five new species: A. lindstr€ omi, A. € obergi, A. polare, A. simplex and A. whitei. The five species were based on few and generally poorly preserved specimens, making the suggested differences highly questionable. Furthermore, the splitting done by Mojsisovics is a typical example of the confusing of ontogenetic changes with species-specific traits.
In 1961, Kummel reviewed the earlier described material and provided new reliable illustrations of the holotypes and paratypes in order to compare actual differences and similarities of the fossils. Kummel (1961, fig. 1) included a diagram of the measured characteristics of more than 200 Spitsbergen Arctoceras specimens from the collection in the British Museum (Natural History), including the holotypes of the existing species. The diagram plots the chamber height (wh) and umbilical diameter (uw) against the conch diameter (d) of each specimen. All specimens fall within a general trend with no distinct groupings and, lacking the relevant stratigraphic data, Kummel concluded that only one Arctoceras species is  present in Svalbard, and that it has a randomly variable conch. After collecting Arctoceras in Svalbard across multiple sections, it became clear that the variation in the genus is not random with respect to stratigraphy. Individuals collected from the lower part of a section are always large and ribbed, individuals from around the middle always small, compressed and generally lacking ornamentation, and towards the top of a section the individuals again become slightly larger, less compressed and more strongly ornamented. In this work we re-evaluate Kummel's single Arctoceras species in Svalbard to test its robustness.

GEOLOGICAL SETTING
Strata of Early Triassic age in Svalbard are represented by the Vardebukta and Tvillingodden formations on the west coast of Spitsbergen and by the Vikinghøgda Formation in the central and eastern parts of the archipelago. The Vikinghøgda Formation was deposited on an offshore shelf and represents the distal equivalent of the coastal to shallow marine Vardebukta and Tvillingodden formations (Mørk et al. 1999;Wignall et al. 2016). The Vikinghøgda Formation consists of three members: the Deltadalen Member was deposited during the Griesbachian and Dienerian, the Lusitaniadalen Member was deposited during the Smithian, and the Vendomdalen Member during the Spathian. The two latter members correspond to the Tvillingodden Formation in the west. The Lusitaniadalen and Vendomdalen members are relatively fossil-rich and fairly well-defined in terms of biostratigraphy (Buchan et al. 1965;Mørk et al. 1999;Hammer et al. 2019).
The lithological boundary between the Lusitaniadalen and Vendomdalen members more or less coincides with the boundary between the Smithian and Spathian substages. However, there is probably a hiatus spanning the latest Smithian, and the substage boundary is included in this gap (Hammer et al. 2019). In the Lusitaniadalen Member, the only preserved late Smithian zone is the early late Smithian Wasatchites tardus Zone (British Columbia, Tozer 1994; northern Siberia, Dagys 1994). In the low latitudes, the late Smithian ammonoid biostratigraphy includes at least two zones (North Indian Margin, Br€ uhwiler et al. 2010;Western USA Basin, Jattiot et al. 2017), only the oldest of which is correlated in the high latitudes (Hammer et al. 2019). Therefore, as stressed by Hammer et al. (2019), the youngest part of the late Smithian is typically missing in the Boreal Realm, indicating a hiatus. In the Lusitaniadalen Member, rocks yielding middle Smithian ammonoids are directly overlain by rocks representing the early late Smithian Wasatchites tardus Zone. The entire succession of strata with middle Smithian ammonoids (dominated by Arctoceras) has in Svalbard traditionally been assigned to a single zone, namely the Euflemingites romunderi Zone. However, the type species is in fact absent from the lower part of the section and only the upper part is assuredly within the E. romunderi Zone (Weitschat & Dagys 1989;Mørk et al. 1999). No older Smithian zones have been confirmed in the Lusitaniadalen Member.
Throughout the Boreal Realm, A. blomstrandi occurs in association with middle Smithian taxa (Korchinskaya 1972;Weitschat & Dagys 1989;Dagys & Ermakova 1990;Tozer 1994). Only in two cases are A. blomstrandi possibly associated with younger strata. From Ellesmere Island (Arctic Canada) in a single locality, Tozer (1994) described an association of A. blomstrandi with species of the late Smithian W. tardus Zone. Tozer himself questioned the authenticity of this assemblage of middle and late Smithian taxa that could have resulted from either condensation or reworking. The other case is the unconfirmed association of Arctoceras erebori (which we synonymize with A. blomstrandi) with xenoceltids and prionitids of the W. tardus Zone from Stensi€ ofjellet (Piazza et al. 2017). The ammonoids were in this case all extracted from a large concretion of which the internal structure and stratigraphy is unknown.

MATERIAL AND METHOD
We collected extensive ammonoid material along with sediment samples for organic carbon analysis in Sassendalen (central Spitsbergen, Svalbard) from two sections at Stensi€ ofjellet (STA and STC) and three sections at Wallenbergfjellet (WALA, WALB and WALD) (Fig. 1). In Stensi€ ofjellet, the two sections were reasonably close to each other and had similar developments, and they were measured and sampled as one. In Wallenbergfjellet, the sections were more distant and were logged separately. Fossils were not brought back from the distant WALD section, but it is included here because of its significance in biozone correlation. Lithological log, macrofossil contents and organic carbon isotopes from each section are shown in Figure 2.

Lithology, fossil occurrences and biozonation
The level datum (0 m) used for the logs (Fig. 2), corresponds to the boundary between the Lusitaniadalen and Vendomdalen members. In Wallenbergfjellet, the boundary is consistent with the definition by Mørk et al. (1999) and is found on the upper surface of a yellow-weathering, silty bed. The bed is part of the grey shales of the Lusitaniadalen Member and the top of the bed corresponds to  . At Stensi€ ofjellet, the yellow-weathering silty bed appeared to be missing and an abrupt change from grey to black shales is used as the member boundary. Further field work by one of us (HB) in 2020 confirmed that the boundary bed is actually present at Stensi€ ofjellet, although it was not exposed in 2017. All the studied sections can be expected to have highly similar lithological developments. Selective preservation is obvious in the Lusitaniadalen Member in all sections, given that ammonoids are found only within early diagenetic carbonaceous lenticular horizons and nodules. The surrounding shales almost exclusively yield flattened bivalves, if anything at all. The lowermost Arctoceras specimens were sampled at À28 m (section WALB) in relation to the Lusitaniadalen-Vendomdalen boundary, while the highest sample was located at À2.1 m (section WALA) (Fig. 2). We confirmed Wasatchites tardus, demonstrating the early late Smithian W. tardus Zone at different heights in the separate sections. The upper boundary of the W. tardus Zone is uncertain, but it is probably located slightly above the lithological boundary (Hammer et al. 2019). The uppermost W. tardus specimen was sampled in section WALA at c. À0.1 m, while the lowermost was sampled in section WALD at À2.7 m. In all sections there is a minor gap between in situ fossils of the E. romunderi Zone and the W. tardus Zone, but in section WALD that gap is as little as 0.3 m, with Arctoceras found at À3 m.
The vertical extent of the W. tardus Zone is difficult to define precisely because it is relatively thin, and may be demonstrated from as little as only one nodular horizon in a section. An attempt to use the organic carbon isotope curves to aid correlation between sections is shown in Figure 2. The WALD section has the longest proven extent of the W. tardus Zone, with three observations spanning 1.6 m. The gap between the lowest W. tardus and the highest Arctoceras specimen (demonstrating the E. romunderi Zone) is also thinner here than in the other sections. The gap has a d 13 C org value of À33& VPDB (Vienna Pee Dee Belemnite standard). This value is traced across all studied sections as a tentative correlation of the interval separating the E. romunderi and W. tardus Zones (Fig. 2). The suggested d 13 C org correlation line falls close to the uppermost Arctoceras samples in each section, and might indicate that these are in fact of similar age, despite differences in lithological development between sections. However, care is taken in ordering samples from separate sections in relation to each other, and for the following analyses, the sampled Arctoceras horizons are treated as belonging to one of four stratigraphic groups: Group I, II, III or IV (Figs 2-6). The number of stratigraphic groups was set to minimize the risk of incorrect correlations and to obtain sufficient sample sizes for statistical analysis.
Arctoceras were collected in association with the following ammonoid species (Fig. 2).
Tellerites furcatus ( € Oberg, 1877): several specimens of T. furcatus (Fig. 5H, I) were found at À5 m (horizon STA8). The species was placed exclusively in the Wasatchites tardus Zone by Weitschat & Dagys (1989), but we have been able to confirm it only in a single horizon in the Euflemingites romunderi Zone.
Pseudosageceras sp. indet. (Diener, 1895): a single Pseudosageceras specimen was found in association with Arctoceras at À11.1 m (horizon STA7). The genus has earlier been observed only in the W. tardus Zone (Weitschat & Dagys 1989;Mørk et al. 1999). Given that Pseudosageceras is a long-raging genus, the variable age occurrence is not unexpected.

Fossil data
The material collected for this work consists of 135 Arctoceras specimens (of which seven specimens are measured twice at different growth stages) from 18 horizons (Figs 2, 7). Levels where Arctoceras specimens were present but not collected are shown in parentheses in Figure 2. Sample sizes from the different horizons are highly dissimilar, and 12 horizons are represented by five or fewer sets of measurements. The material is curated at the Pal€ aontologisches Institut und Museum, Universit€ at Z€ urich, Switzerland, and figured specimens have received catalogue numbers of this institution (PIMUZ). Arctoceras specimens were measured and described following Klug et al. (2015). For each specimen, measurements were taken for the largest preserved diameter d and corresponding whorl width ww, whorl height wh and umbilical width uw. When the conch was not complete on both sides, the half width was measured on the intact side and the full whorl width inferred. All measurements are available in the Dryad Digital Repository (Hansen et al. 2020). In some damaged specimens the best measurements were obtained not on the outermost whorl, but on a younger part of the conch (measurements noted with a '*' in the online dataset; Hansen et al. 2020). Specimens with diameters below c. 10 mm were not included in analyses because of the low measuring accuracy. Qualitative characterization of shell ornamentation (i.e. ribs) is also considered for each specimen (weak/strong/absent). When a specimen was too damaged, the ornamentation was characterized as 'unknown' (Hansen et al. 2020). Data processing was done using the free software PAST, ver-  The d 13 C org curves from the Stensi€ ofjellet, WALA and WALD sections were prepared at the Pal€ aontologisches Institut und Museum, Universit€ at Z€ urich, Switzerland and measured at the University of Lausanne, Switzerland. Fifty-five samples were selected for bulk organic carbon isotope composition (d 13 C org ); all pulverized using a ceramic mortar.
About 5 g of powdered samples were dissolved in 6 M hydrochloric acid to remove all carbonates. After centrifuging, residues were rinsed several times with deionized water and centrifuged until neutrality was reached. The residues were dried overnight at 45°C, and d 13 C org values of the homogenized residues were measured using a Carlo Erba 1500 elemental analyser connected to a ThermoFisher Delta V Plus mass spectrometer at the University of Lausanne. The samples were individually wrapped in tin foil cups and sequentially allowed to react  with an injected quantity of oxygen while continuously flushed with He carrier gas. The sample was oxidized in the reactor at c. 1050°C using cobalt(II) oxide as the catalyst. Excess oxygen in the He-stream was adsorbed in a reactor column filled with metallic Cu held at 500°C. The CO 2 produced was passed over a magnesium perchlorate (Mg[ClO 4 ] 2 ) trap to remove H 2 O, and a gas chromatograph to separate the N 2 from the CO 2 , before the CO 2 was carried by the He-stream into the mass spectrometer for isotopic analysis. The reproducibility of several in-house standards used is better than 0.1&, and they are calibrated against USGS-24 graphite (À16.0& VPDB).

RESULTS
Size and maturity Figure 7 shows a box and dot plot of the measured conch diameters within each horizon. The median value (shown as a vertical line) is given in each box. An obvious change in specimen size is observed from the lower to the middle part of the section between the samples of horizon WALB2 (À21.9 m) and of WALB3 (À15.5 m). Although the basal horizons (Group I, À28 m to À21.9 m) contain no specimens smaller than 64 mm in diameter, the rest of the section (À15.5 to À2.1 m) is dominated by small specimens <100 mm. Only a single complete specimen from horizon STC2 reaches a diameter of 125 mm. In some horizons (STC12, WALB4, STC2 and WALA3), scattered fragments of large individuals (with an inferred diameter of more than 100 mm; Hansen et al. 2020) indicate that the observed size trend is in part due to preservation bias. The preserved conch sizes in the lower (Group I, À28 to À21.9 m) and upper (Group IV, À3.6 to À2.1 m) horizons appear fairly symmetrically distributed. Conversely, the size distribution of the horizons in between (Group II, À15.6 to À6.4 m; Group III, À5 m) is markedly skewed, with many relatively small specimens and a right tail of larger specimens. This distribution pattern could be interpreted as a higher juvenile mortality in the palaeoenvironment of these horizons. Another interpretation could be that the individuals of these horizons have mechanically lost most of their flattened body chambers during extraction in field and therefore appear smaller, overall. Unfortunately, the material is rarely preserved with sutures and there is not much direct evidence for or against the two interpretations. The general absence of sutures likewise makes it difficult to assess at what diameter Arctoceras reached maturity. Generally, in specimens in which the sutures are visible, the body chamber is at least partially preserved, showing that the entire phragmocone at the time of death can be expected to be preserved across the material (Figs 3-6). Septal crowding at the end of the phragmocone, which is diagnostic of maturity, is rarely observed in the material. From horizon WALB2 at À21.9 m a poorly preserved F I G . 7 . Box and dot plot showing the size (diameter) distribution of specimens within the sampled horizons (sorted according to stratigraphic group). Boxes extend from the 25th to the 75th percentiles, with the median shown as a vertical line within the box. Solid triangular data points signify internal data from a larger specimen. The number of data points in each horizon is noted at the left of the boxes. Grey shading signifies the smallest known specimen with strong ribs within a horizon.
specimen (PIMUZ37612; Fig. 3C) shows crowding of both ribs and sutures. The specimen was at least 170 mm in total diameter and the last suture sits at a diameter of c. 125 mm. From the upper part of the section at À3.1 m (STC2), a very poorly preserved fragmental specimen (PIMUZ37605) appears to have approximated ribs towards the venter. The fragment originates from an individual with a diameter of at least 150 mm. The sutures are, however, not preserved in PIMUZ37605. These two specimens might indicate that the maximum size of Arctoceras is constant or near-constant across the stratigraphy, but to thoroughly assess the question of size and maturity, more and better material will be needed.

Ornamentation
Ornamentation, mainly in the form of relatively distant lateral ribs, as a rule stronger on outer whorls, is unevenly distributed through stratigraphy. The notion of whether ribs are strong or weak is qualitative and somewhat arbitrary. However, distinctly strong ribs are dominant in the lower horizons (Group I, À28 to À21.9 m), where specimens are large. In the middle part of the section (Group II, À15.6 to À6.4 m), all specimens lack strong ribs except for a single fragmental large individual (inferred original diameter of 100-125 mm) at À13.1 m (WALB4).
In the upper part of the section (Groups III and IV, À5 to À2.1 m), strong ribs are present even though specimens are not generally larger than in the middle part.
Ribbing in Arctoceras has a highly flexible nature and usually appears very gradually through growth. Figures 3-6 show selected specimens from the four stratigraphic groups.
Conch morphology in relation to stratigraphy Figure 8 shows scatterplots of the individual measured coiling parameters of all specimens and demonstrate how the four stratigraphic groups cluster in morphospace. Figure 8A shows the umbilical width index (uw/d) versus the diameter of all specimens. Large specimens of Group I (À28 to À21.9 m) appear to form a linear trend, consistent with the smaller specimens from Groups II (À15.6 to À6.4 m) and III (À5 m), and describe a widening umbilicus in relation to diameter with age/size. Specimens of Group IV (À3.6 to À2.1 m) have overlap with the other groups, but many individuals plot in the far upper left part of the morphospace (i.e. they are more evolute than any other specimens of similar size). The specimens from Group III (À5 m) have a large spread in umbilical width indexes, but appear to form an intermediate cluster between Groups II and IV, with which they have a large overlap. Figure 8B shows whorl height normalized to diameter (wh/d) versus diameter. As could be expected, the distribution is similar to that of Figure 8A, but vertically inverted, and whorl height decreases in relation to diameter with age/size (involute specimens have relatively large values for wh/d). Figure 8C-D demonstrates two different, but closely related, expressions of conch shape (Fig. 8C: ww/d, conch width index; Fig. 8D: ww/wh, whorl width index). A similar trend is seen in the two plots: specimens from Group I form a cluster mostly separate from the rest of the material. Groups II and III are unable to be separated and also overlap largely with Group IV. Group IV does, however, spread far into the upper part of the morphospace, having wider conchs and whorl sections than other specimens at similar diameters.
The morphological parameters of the stratigraphic groups can also be expressed in two dimensions, as in the box plots in Figure 9. Figure 9A (uw/d) shows a peculiar signal (reversed in Fig. 9B for wh/d): Group I has large values, but the immediately overlaying Group II has comparably small values. Through Groups II to IV, the values increase again to a level comparable to that of Group I. Using the descriptive terms from Klug et al. (2015), individuals of Groups I and IV are nearly all subinvolute (uw/d, 0.15-0.30), while Groups II and III tend heavily towards involute conchs (uw/d, <0.15).
When considering the conch width (ww/d) and whorl width (ww/wh) indices (Fig. 9C, D), there is a slight, but unmistakable, increase in values from Groups I to IV. However, the size bias between the stratigraphic groups is severe (Fig. 7), and the relationship between size and conch morphology (ontogentic development) is investigated further below.

Isometry/allometry of umbilical width in relation to stratigraphy
The degree of conch evolution appears to be ontogenetically controlled, given that the umbilical width index (uw/d) generally increases with size across the dataset (Fig. 8A). We study this ontogenetic pattern using nonlinear regression of umbilical width against conch diameter for the allometric equation y = ax b (i.e. forcing a zero intercept), with bootstrapped confidence intervals. We chose this method over the classical log-log transformation (which would yield a linear regression) because it gives a more direct and easily interpretable picture of the ontogenetic development (de Queiroz 1998). The regression coefficients for the allometric equation are given in Table 1 and the resulting allometric curves are shown in Figure 10A.
There is a considerable reduction in the allometric coefficient b from Group I at the bottom of the section up to Group III at À5 m, describing a shift from allometry to sub-isometry in the ontogenetic development of uw. Although the 95% confidence intervals are overlapping between adjacent groups, there is no overlap between Groups I and III, and the trend is reliable (Fig. 10A, inset). Sub-isometry is also seen for Group IV at the top. An immediate problem with the data is the severe size bias between groups. In Group III the growth pattern beyond d =~70 mm is uncertain, and uw might actually have a more allometric development than can be proven from the recovered material. The sub-isometric growth in Group IV is more well-founded and reliable. At the same time as the allometric coefficient is shown to decrease upsection, the scaling coefficient a increases from 0.03 at the bottom of the section up to 0.17 at the top. The trends in coefficients a and b lead to a complex relationship between size, degree of evolution (uw/d) and stratigraphy. At small sizes (d < 60 mm), the overall conch evolution undisputedly increases from the lower levels to the top of the section (Fig. 10A). At intermediate sizes (d = 60-130 mm), the trend becomes obscured, but the most evolute conchs are still found at the top of the section (Group IV). At large conch sizes (d > 130 mm), Group I from the bottom of the section overtakes the other groups due to the strong positive allometry of this group. We are now able to explain the peculiar trend seen in Figure 9A: given that only intermediate and large specimens were recovered in stratigraphic Group I, this group is represented solely by comparably evolute forms (relative large uw/d ratio). Groups II and III appear less evolute than Group I, mainly because only smaller conchs were recovered. Conversely, the relatively more evolute signal in Group IV is not an artefact of size, but an actual morphological development towards larger umbilical widths in the juvenile and sub-adult conchs.

Isometry of whorl width in relation to stratigraphy
We investigate the trends of whorl width (ww) against conch diameter (d) as a function of stratigraphic position, as was done for umbilical width above. The regression coefficients for the allometric equation are given in Table 2 and the resulting allometric curves are shown in Figure 10B. The allometric coefficients (Fig. 10B) are not significantly different from isometry (b = 1) except in Group III at À5 m, where a slight negative allometry is indicated. However, there is a stratigraphic trend, given that Group I from the bottom of the section exhibits the smallest whorl widths across all conch sizes. Meanwhile, Group IV from the top has the largest whorl widths for all conchs larger than 50-60 mm in diameter. Figure 9C depicted very much the same trend, which is now proven largely independent of the size bias. In Figure 9D, Groups I-III all have similar median values of ww/wh, in spite of the large differences in conch sizes within these levels. Group IV is the only stratigraphic group with a markedly different median value, indicating that a shift in conch morphology is taking place at the top of the section between Groups III and IV.

Ornamentation and conch morphology: Buckman's law of covariation
Buckman's first law of covariation (Buckman 1892, p. 313;Westermann 1966, p. 305;Monnet et al. 2015;Moulton et al. 2015) states that more evolute and depressed individuals within a species tend to have coarser ornamentation. A simple explanation for this common phenomenon in terms of isometric scaling was given by Hammer & Bucher (2005). For testing Buckman's law of covariation in our material, we correlated strength of ribbing with uw/d (describing the degree of evolution) and ww/wh (describing the degree of compression) across the complete dataset. Strength of ribbing in each specimen was assessed qualitatively and assigned to three classes: absent, weak, and strong. Specimens with unknown ornamentation (due to surface damage) were not included and the total number of data points in the analysis is N = 126. A specimen can easily have weak or absent ribbing on the inner whorls followed by strong ribbing on the outer whorls (but not the other way around). Only the strongest degree of ribbing is noted. The strength of ribbing is, in the analysis, considered as an ordinal value reflecting an underlying continuous distribution, while uw/d and ww/wh are continuous variables. When testing the degree of ribbing versus ww/wh, we obtained r = 0.18, p = 0.036 (Spearman's rank-order correlation). This

Group
Level ( The relatively strong correlation between the degree of evolution and ornamentation (r = 0.48) is in accordance with Buckman's law, and this can be at least partly invoked as responsible for the strong ornamentation of the stratigraphic groups I and IV, in which the observed degree of evolution is strongest (Fig. 9A).
Trends in bulk organic carbon isotope values from STA, WALA, and WALD are closely comparable. In all three sections, d 13 C org from the lower part of the studied middle Smithian Lusitaniadalen Member ranges between c. À34& and c. À33& (Fig. 2). The onset of a slight positive trend is observed at STA, c. 5 m below the Lusitaniadalen-Vendomdalen boundary; the positive shift is distinctly expressed from 2 m below the boundary and continues until c. À29& is reached. At WALA, d 13 C org shows a more gradual positive shift from À33.34& at the base to À28.79& at the top of the studied section (Fig. 2). At WALD, d 13 C org in the lowest part of the studied succession is c. À33.5&. It starts to increase 3.5 m below the Lusitaniadalen-Vendomdalen boundary and reaches c. À30& in the lowermost part of the Vendomdalen Member. In all three successions, preliminary observations (ESH) indicate that particulate organic matter does not significantly change in composition throughout the studied intervals. Thus, a compositional change in organic matter is unlikely to be the driving mechanism for the observed changes in bulk organic carbon isotopes values (Tyson 1995).

DISCUSSION
We have investigated the morphology, ornamentation and allometry of Arctoceras in Svalbard across stratigraphy. The taxon shows a large variation in conch parameters within each stratigraphic group, along with notable differences between groups. Even so, we consider the material as belonging to one species, A. blomstrandi. The overlap between adjacent stratigraphic groups is generally large and none of the groups can be completely separated from the bulk of the material, therefore the erection of two or more distinct species is not justified. Buckman's law of covariation can be shown to be partly responsible for the variation seen in the Svalbard arctoceratids, and no mode of ornamentation is decoupled from the main trend of the material. We still document a clear stratigraphic trend in A. blomstrandi, which is discussed and quantified further here.

Conch morphology, ornamentation and allometry in relation to stratigraphy
We have shown above how the bias in preserved conch size across stratigraphy is not the only controlling factor on the recovered conch morphologies in A. blomstrandi. There are undisputable differences in the ontogenetic development between the bottom and top of the section. The degree of conch evolution is the trait creating the most prominent variation in A. blomstrandi because this is also linked to the strength of ribbing (Buckman's law of covariation). When considering the ontogenetic development in umbilical width, a shift happens between Groups II and III from allometry to isometry (Fig. 10A). This means that in most of the section (À28 to À6.4 m), uw/d can be expected to increase during ontogeny, giving a slight un-coiling effect of the conch, a trait common in arctoceratids. Conversely, in the upper part of the section (À5 to À2.1 m), the umbilical width will tend to remain near constant in relation to the diameter throughout growth. From Group III to IV, further changes take place as the initial width of the umbilicus in relation to diameter increases. This is also seen from Figure 8A. When considering the ontogenetic development of the whorl width no shift is seen in the type of growth, given that all groups grow isometrically. However, the trend of larger relative sizes in the top of the section is seen here as well. Considering the trend in whorl section shape (ww/wh, Fig. 9D), the morphological shift is restricted to Group IV. When it comes to the development of ornamentation in Arctoceras blomstrandi, there are some uncertainties due to poor preservation and size biases. We have indications that in Groups I and II, strong ribs do not appear in small conchs (d < 100 mm) (Fig. 7). Conversely, in Groups III and IV, examples of very small individuals with prominent ribbing are known (e.g. PIMUZ37594; Fig. 5E). We showed that Buckman's law of covariation is in effect, and that the emergence of strong ribbing in small conchs is related especially to the widening of the umbilicus. The morphological changes in the conch of Arctoceras blomstrandi are strongest in Group IV, but appear to be initiated within Group III. Group III is in fact a single horizon, and the onset of morphological change is very well-defined at c. À5 m below the lithostratigraphic boundary.

Definition of a 'new' morphological endmember
It is the variations in conch coiling and ribbing that create the largest differences in appearance in A. blomstrandi. The earlier distinction between seven species in Svalbard was beyond doubt a gross exaggeration of the importance of each trait. However, given that we have shown that A. blomstrandi undergoes directional changes to its morphology through time, the question arises whether one species is in fact sufficient to cover the variation.
Kummel (1961, fig. 1) presented a plot of whorl height (wh) and umbilical diameter (= uw) against the conch diameter of Spitsbergen A. blomstrandi specimens in the collection at the British Museum (Natural History). As Kummel pointed out, all specimens fall within the same general trend and do not form obvious groupings. However, it is interesting to note how the holotype of A. costatus falls slightly outside the main plot with a relatively large umbilical width, as is the trend in A. blomstrandi towards the top of the stratigraphy. The A. costatus description was also the most distinct of the species descriptions following the initial description of A. blomstrandi by Lindstr€ om (Kummel 1961). In the present material we observe a number of costatus-like specimens that are in fact all found towards the top of the sections. However, although a specimen such as PIMUZ37616 (WALB8, d = 62.28 mm; Fig. 6C) is practically identical to the description by € Oberg (1877) (see 'History of Arctoceras blomstrandi' above), there are some immediate problems with the costatus species description. In PIMUZ37592 (STA3, d = 98.23 mm; Fig. 6E), the ribs are distant and distinct but do not cross the venter, and the umbilical shoulder is not rounded. Rather, the umbilical shoulder is raised and it could even be argued that it is developing very faint tubercles towards the aperture. In PIMUZ37609 (STC3, d = 63.39 mm; Fig. 6D) the ribs cross the venter but the umbilical shoulder is again distinct and slightly raised, not rounded. The overall impression of costatus sets it apart from the bulk of Arctoceras specimens but the distinction between morphotypes is a difficult one to draw. This also means that the exact stratigraphic distribution of the costatus morphotype is uncertain. All the distinct specimens in our material are found in stratigraphic Group IV between À2.1 and À3.6 m. In Figure 11, we have plotted measurements of the holotype, paratype and lectotype specimens of the original Arctoceras species alongside the present material. Conch measurements are taken from Kummel (1961), Lindstr€ om (1865) and Mojsisovics (1886). Due to the poor preservation of the type material and differences in measuring standards, the given values are not entirely trustworthy; also, not all characteristics could be measured (hence, the uneven number of measurements included in the plots in Fig. 11). Still, all data points fit reasonably within the same trends as are described in this study, supporting the conclusion of a single species as drawn by Kummel (1961). In Figure 11A-B, the holotype of costatus plots neatly in the same areas as specimens sampled from the upper levels of the stratigraphy.
According to € Oberg (1877), the suture line of costatus bears two umbilical lobes, as opposed to one in A. blomstrandi. Kummel (1961) was not able to confirm any consistent differences between suture lines of the various Arctoceras morphotypes in Svalbard, and even points to the fact that the suture line might vary on opposing sides of the same individual. In the present material, suture lines are not generally preserved, and the question has been difficult to study in full detail. In Figure 4H the suture line of the costatus-like specimen PIMUZ37616 is shown, and the two umbilical lobes are indeed present. However, the same appears to be the case in, for example, PIMUZ37596, a slim un-ornamented specimen, although the suture line proved difficult to draw at the umbilicus in all specimens (compare Figs 4J and 5D).
Here, we refrain from elevating the costatus morphotype to species level based on the generally high flexibility of the Arctoceras conch morphologies and the large overlap of morphologies across stratigraphic groups. Still, the costatus morphotype is a distinct endmember restricted to the top of the section, which makes it potentially useful in biostratigraphy. More costatus material from Svalbard and other Boreal localities might push our understanding of this particular morphotype, but currently there is no strong support for reinstating costatus as a separate species. For future biochronological descriptions and discussions, the endmember could be referred to as A. blomstrandi var. costatus.

Conch size in relation to stratigraphy
A substantial reduction in average conch size occurs in Arctoceras blomstrandi between Groups I and II from the sample at À21.9 m to the sample at À15.6 m. The associated changes in morphology (reduced degree of evolution and strength of ribbing) between these horizons are explained by the allometric effects of the size reduction. The reduction in observed sizes could be a preservation artefact, but may alternatively reflect changes happening in the palaeoenvironment. Perhaps an unstable environment led to increased juvenile mortality, seen as the deposition of an unproportionate number of many small conchs. Alternatively, the habitats of age groups switched, and juveniles became the dominant inhabitants of the area.
In Groups III and IV, the overall recovered conch size does not change substantially from Group II, but 'mature' morphology and ornamentation are developed at smaller sizes than previously, and the endmember morphology A. blomstrandi var. costatus may be observed. The lack of evidence of full maturity (in the form of approximating sutures and/or ribs) would suggest that A. blomstrandi at the top of the section did in fact grow to sizes similar to those seen in the bottom of the section. Only one fragmental specimen from À3.1 m might be used in direct support of this assumption: specimen PIMUZ37607 (Fig. 6F), which is the largest known individual from Group IV, with a diameter of 124.7 mm. It has costatus-like straight ribbing on an inner part of the conch and more 'classic' sigmoidal ribbing closer to the aperture. This could be interpreted as the retaining of the fully mature ornamentation at large sizes across the stratigraphy. However, the general absence of very large individuals (d > 130 mm) in most of the stratigraphy is difficult to explain if a constant mature size and ornamentation was maintained at all times. Preservation size biases would hardly be that consequent throughout that many metres of sediment. A possibility is, as mentioned above, that mature or near-mature individuals migrated to other areas. Finally, it is also a possibility that A. blomstrandi did actually evolve towards smaller mature sizes towards the top of the stratigraphy. It would be very interesting to see these possible answers studied in similar ways in other regions in the Boreal Realm.
places them in the peak phase of the middle Smithian negative carbon isotope excursion (Hammer et al. 2019) that is associated with a spore spike and an ecological crisis of land plants (Hochuli et al. 2016) during the middle Smithian thermal maximum (Romano et al. 2013). The morphological changes in A. blomstrandi that give rise to A. blomstrandi var. costatus are seen to be initiated at c. À5 m in relation to the top of the Lusitaniadalen Member (and c. 2.6 m below our inferred base of the W. tardus Zone; Fig. 2). Thus, the initial change in Arctoceras morphology coincides with the initial increase in d 13 C in the uppermost middle Smithian, leading up to the major positive shift across the Smithian-Spathian boundary (Hammer et al. 2019). As mentioned above, A. blomstrandi might in fact be subject to miniaturization in the upper 5 m of the stratigraphy, which is not an uncommon response to environmental stresses (Urbanek 1993), and could very well be a likely explanation for the stratigraphic evolution observed in A. blomstrandi. However, we have not been able to assess the maturity of the present material, and it is not currently possible to test for an evolutionary change towards smaller conchs (miniaturization).

Comparisons with Arctoceras observations in Arctic Canada and Siberia
In 1994, Tozer described A. blomstrandi from several Smithian localities in Canada (Ellesmere and Axel Heiberg Islands, British Columbia and Alberta) and noted how they often have prominent umbilical tubercles. Rare specimens were, however, strikingly different. Tozer illustrated a specimen (GSC28154; Geological Survey of Canada) from Svartefjeld Peninsula that he assigned to A. blomstrandi (following Kummel 1961), but which he described as looking like A. lindstroemi (Mojsisovics) or A. costatum [A. costatus] ( € Oberg). Where other specimens from the same section bear weak ribs and distinct tubercles, GSC28154 bears strong ribs and only barely perceptible tubercles. Furthermore, the specimen was found in possible association with late Smithian taxa, and Tozer tentatively placed it in the W. tardus Zone (Tozer 1994, p. 74, pl. 27, fig. 1). Based on the given illustration (diameter appears to be no more than 90 mm), GSC28154 does indeed resemble A. blomstrandi var. costatus. The placement of GSC28154 in the W. tardus Zone is questionable and is probably due to severe condensation. In that case, GSC28154 can be expected to originate from the very top of the Euflemingites romunderi Zone, which gives it the same approximate age as A. blomstrandi var. costatus in Svalbard. The morphological evolution observed in A. blomstrandi in Svalbard would appear to be present also in Arctic Canada, and is hence useful in biostratigraphy in the Boreal Realm.
From Arctic Canada, Tozer (1994) also described the possibly late Smithian species Arctoceras gigas from the Blind Fiord Formation (W. tardus Zone?) at Lindstr€ om Creek, Ellesmere Island. Arctoceras gigas is defined as having sometimes dense tubercles, an inclined umbilical wall and a wider umbilicus than A. blomstrandi (Fig. 11A). Where A. blomstrandi has an umbilical width of less than 30% of the diameter (often c. 25%) A. gigas has an umbilical width of more than 30% (Tozer 1994). We find this to be in general accordance with our data, in which only three specimens have umbilical width indexes of more than 0.29. It is interesting to note that although two of these are large individuals from the bottom of the section, one is from the upper part at À3.1 m (horizon STA3) and is certainly a juvenile (d = 14.34 mm). The taxonomic relationship between A. gigas and A. blomstrandi would be highly interesting to see examined, given that A. blomstrandi is now shown to evolve towards a wider relative umbilicus. Tozer (1994) did not illustrate the suture line of A. gigas for our direct comparison to other Arctoceras species. Dagys & Ermakova (1990) reported observations (in Russian) of Arctoceras blomstrandi from the Kolymensis Zone (= middle Smithian) of Siberia. The illustrated Arctoceras specimens with corresponding description and measurements fit well within the trends seen in Svalbard. The one suture line (Dagys & Ermakova 1990, fig. 18A) is of the more denticulate type, but still within the variation expected in A. blomstrandi (Fig. 4H-K). According to Dagys & Ermakova (1990), A. blomstrandi has a limited conch variability in Siberia (uw/d, 0.19-0.28; wh/d, 0.41-0.52; ww/d, 0.25-0.36; ww/wh, 0.55-0.79), and the major intraspecific variation arises from differences in the strength of ornamentation. Measurements of only five Russian specimens have been reported, and it is very possible that the authors did not sample the full range of conch variation in Siberian A. blomstrandi. The five Siberian specimens are fairly large and plot with Group I (and partly with Group IV) from Svalbard (Fig. 11). One Siberian specimen (372/37, d = 91 mm) was measured at three growth stages. The youngest stage (d = 17.4 mm) plots in the areas where Groups III and IV overlap, while the middle growth stage (d = 42 mm) plots with Groups II and III (Fig. 11). The three consecutive measurements of specimen 372/37 point to a very plastic development of the relative width of the umbilicus and the whorl. However, the specimen appears highly weathered from the illustration (Dagys & Ermakova 1990, pl. 9 fig. 2), and the measurements are considered with some caution here. Another specimen, 370/37 (Dagys & Ermakova 1990, pl. 9 fig. 2), bears straight ribs reminiscent of those of A. blomstrandi var. costatus, but also bears umbilical tubercles. The specimen has a much slimmer conch than specimens of Group IV (370/37 is marked with a dot in Fig. 11), and is morphologically related to the large specimens of Group I. As at the time of writing, the A. blomstrandi var. costatus morphotype is unconfirmed in Siberia, and it is not currently possible to say if the morphological trends from Svalbard may be observed in Siberia.

CONCLUSION
It is a general issue in ammonoid palaeontology that longand wide-ranging genera need improved definitions of conch morphology variability between, and within, species. A well-understood specific variation across stratigraphy would be beneficial to the biostratigraphical resolution and is something worth attempting within each basin. The middle Smithian Arctoceras blomstrandi was initially split into seven species (A. blomstrandi, A. costatus, A. lindstr€ omi, A. € obergi, A. polare, A. simplex and A. whitei) in the late 1800s, only to be reformed as a single species by Kummel in 1961. However, the variation in conch morphology and ornamentation observed in the field appears to fall in stratigraphically distinct groups, and a revision of Svalbard Arctoceras was needed. For this study, we have collected new and extensive Arctoceras specimens from Svalbard within the Boreal Realm. Sections on Stensi€ ofjellet and Wallenbergfjellet in Sassendalen were logged and sampled to obtain material with a high degree of stratigraphic control. Several stratigraphic horizons yielded five or fewer specimens, and the horizons were sorted into four stratigraphic groups to enable meaningful analysis. We find that the variability in the investigated specimens does not exceed normal intraspecific variation. The variation is, however, not random, as proposed by Kummel (1961), but depends on stratigraphic position. A bias in recovered conch sizes between different parts of the section complicated the morphological analysis, but the ontogenetic development was shown to vary between the stratigraphic groups. In the major part of the A. blomstrandi species range, conch morphology and ornamentation follow the same ontogenetic trends. The observed variation is to a large degree controlled by the different developmental stages of the individual specimens. Towards the top of the species range, a shift in the conch morphology takes place, and the relative expansion of the umbilicus changes from allometric to isometric. At the same time, the whorl width is seen to increase slightly, and the whorl width index reaches larger values in the top than in any other part of the section. Ornamentation in the form of ribbing (coupled to conch morphology through Buckman's law of covariation) becomes stronger in the top, while umbilical tubercles are apparently not more common than earlier. The morphological shift in A. blomstrandi is located c. 5 m below the Smithian-Spathian boundary, and possibly 2.6 m below the middlelate Smithian boundary. A morphotype similar to the synonymized species A. costatus of € Oberg (1877) is hence observed only in the uppermost part of the stratigraphy (3.6-2.1 m below the Smithian-Spathian boundary), and is considered as a morphological and stratigraphic endmember within the A. blomstrandi complex. The documented morphological trends in Spitsbergen provide a first reference study for forthcoming comparisons with occurrences from other latitudes. Such studies will ultimately contribute to disentangling evolutionary from ecological responses of ammonoids during the middle Smithian thermal maximum, i.e. the warmest climatic episode throughout the entire Early Triassic. Remarks. Seven species of Arctoceras described from Svalbard in the nineteenth century were lumped into the senior species A. blomstrandi (Lindstr€ om, 1865) by Kummel (1961). The one species of Arctoceras in Svalbard remains valid, but rare material assigned to the synonym species C. costatus ( € Oberg, 1877) is acknowledged as a distinct endmember morphology with a limited stratigraphic range, and is termed A. blomstrandi var. costatus ( € Oberg 1877, pl. 4, fig. 3).

Figures 3-6
Emended diagnosis. Discoidal conch of moderate to sometimes strong compression. Conch is most commonly subinvolute, but small individuals (d <~70 mm) may be involute, and very rare individuals of various sizes can be subevolute. Flanks are subparallel to slightly convex, grading imperceptibly onto the venter. The whorl section is oval and widest at about mid-flank. The venter is narrowly to broadly rounded. Umbilical shoulder is angular and the umbilical wall vertical. Ornamentation consists most often of simple ribs, sometimes weak umbilical tubercles and rarely ventral strigation. The ribs are sinuous to straight, arranged distantly with a radial to prorsiradiate orientation. The ornamentation is variable across the species with some small to medium individuals lacking all sculpture, but ribs typically grow stronger through ontogeny. As a rule ribs are strong on large individuals (d >~100 mm), and umbilical tubercles are not seen on small individuals. Suture line is ceratitic with broad saddles and lobes bearing few lobules (2-4 in immature specimens and max. 6-7 in large mature specimens) (Fig. 4H-K).
In the top few (~2.6) metres of the Arctoceras blomstrandi stratigraphic range, the species tends towards relatively wider whorl sections and more evolute conchs compared with the rest of the range. Within this interval, the rare A. blomstrandi var. costatus may be observed. The morphotype bears prominent ribs that tend to be straight and may cross the venter. The umbilical shoulder can be somewhat rounded and there are no true umbilical tubercules, although the umbilical shoulder can be slightly raised. Whorl section becomes almost rectangular.
Remarks. Arctoceras erebori from possibly the upper Smithian of Svalbard (Piazza et al. 2017) is indistinguishable from juveniles of A. blomstrandi both in ornamentation and in conch morphology (Fig. 11). Arctoceras erebori is described as having ventral strigation and sometimes a keel. Ventral strigation is not unknown in A. blomstrandi, and Lindstr€ om (1865) included the feature in the original description of C. blomstrandi. Examples of (very weak) ventral strigation are seen in PIMUZ37616 (Fig. 6C) and PIMUZ37592 (Fig. 6E). The proposed keel in A. erebori is present in only one specimen (PMO210.489) on a part of the venter that is collapsed, and the feature is here considered a preservation artefact. The suture line of A. erebori is of the more denticulate type, but not exceeding the variation otherwise seen in A. blomstrandi. The precise stratigraphic level of A. erebori is unknown, but according to museum notes, the material was found within a concretion in association with taxa typical of the W. tardus Zone. This association could not be confirmed from any of our bedrock-controlled collections. The association of Arctoceras with the W. tardus Zone is likely to be the result of local compaction or even reworking, and A. erebori is here considered a junior synonym of A. blomstrandi.
Arctoceras tuberculatum (Smith, 1932) from the early and middle Smithian of USA (Nevada, Kummel & Steele 1962;Jattiot et al. 2017;Jenks & Brayard 2018;California, Kummel & Steele 1962;Utah, Brayard et al. 2013) has a conch morphology very similar to that of A. blomstrandi as demonstrated in Figure 11. However, the umbilical shoulder of A. tuberculatum tends to be raised and to bear stronger tuberculation than is seen in any A. blomstrandi specimens. The suture line may be helpful in distinguishing between the two species, given that the lateral lobe is more dentate in A. tuberculatum than in A. blomstrandi in specimens of similar size.
Arctoceras rubyae (Jenks & Brayard, 2018) from the late early Smithian and early middle Smithian of Nevada has ornamentation and an umbilical width index comparable to that of A. blomstrandi, but the species is recognized from its extremely compressed conch (ww/d, 0.20-0.22).
Arctoceras gigas (Tozer, 1994) from the late middle Smithian or possibly the early late Smithian of Ellesmere Island (Arctic Canada) has a very similar appearance to A. blomstrandi. Arctoceras gigas is represented only by few and poorly preserved specimens, but is defined as being larger than A. blomstrandi, as well as having a wider umbilicus (uw/d <0.30) with denser umbilical tubercles. Arctoceras gigas is indeed very large, but the specimens measured by Tozer (1994) seem to share ontogentic trends with both A. blomstrandi and A. tuberculatum (as suggested by Brayard et al. 2013) (Fig. 11). More material with suture lines preserved would be needed to either confirm the species A. gigas or to treat it as a synonym of another Arctoceras species.
Arctoceras strigatus from Guangxi (South China) (Brayard & Bucher 2008) has a very weak radial ornamentation and profound strigation, setting it apart from A. blomstrandi. The suture line is also more dentate in A. strigatus.
Arctoceras septentrionale (Diener, 1895) and A. subhydapsis (Kiparisova, 1961) from the Smithian of South Primorye (Russia) as described by Shigeta & Zakharov (2009), are difficult to confirm as separate species, because they might as well represent different maturity stages. Arctoceras septentrionale is not truly distinguishable from juvenile A. blomstrandi in either ornamentation or conch morphology (Fig. 11). Arctoceras subhydapsis is more evolute and has a more dentate suture line than A. blomstrandi, but more material of the two Russian species would be useful before a confident conclusion can be made on the taxonomy.
Arctoceras schalteggeri from the early Smithian of Pakistan (Br€ uhwiler et al. 2011) has a conch morphology similar to that of A. blomstrandi, but ornamentation is simpler and weaker, and the suture line appears to be more denticulate.