Ehrenberg originally described Navicula crux (Ehrenberg 1838: 184) in the raphid genus Navicula, because he observed a ‘sutura media interruptis [interrupted by a median groove]’ mistaking the broader, thickened sternum for a raphe structure. On his original drawings (Fig. 1A), however, he did not indicate a raphe structure, and in the original material only a valve lacking a raphe could be found (Fig. 1B–H). Although the largest valves in the Ehrenberg material seem to be somewhat more robust, most of the observed valves show a high similarity with the population in the type material of Staurosirella harrisonii (W.Sm.) E.Morales & C.E.Wetzel from Hull (United Kingdom) (Fig. 2). In addition, the populations observed in the Grunow and Kützing material collected in Moosach (Germany) (Supplementary materials 1–3) show that Navicula crux and Staurosirella harrisonii are conspecific. As Navicula crux is the oldest name in the same rank that was validly published, it has priority over S. harrisonii (Turland et al. 2018: Art. 11.4). Based on the observed morphological features, the latter species is transferred to the genus Staurosirella as Staurosirella crux (Ehrenb.) Van de Vijver & Kusber comb. nov.

Staurosirella crux (under the name Fragilaria harrisonii or Staurosirella harrisonii) has been considered by some authors as a synonym of the more commonly known Staurosirella leptostauron (Ehrenb.) D.M.Williams & Round (Hustedt 1931; Patrick and Reimer 1966; Krammer and Lange-Bertalot 1991). Staurosirella leptostauron was first published within the genus Biblarium Ehrenb. that was proposed for rejection by Williams (1986) in favour of the more commonly used genus Tetracyclus Ralfs (Wiersema et al. 2018). However, the name Biblarium leptostauron was introduced by Ehrenberg (1854: 8, plate XII, figs 35–36) who provided a name and an illustration for his new taxon. As he treated Bacillariophyta as animals (“Infusionstierchen”), the International Code for Zoological Nomenclature (ICZN) must be followed before treating a taxon name under the International Code for Botanical Nomenclature (ICN). The name was introduced as a species group name (ICZN 1999: Art. 11) without description or diagnosis (ICZN 1999: Art. 12.1) but with an indication because the name was assigned to an image of the taxon (ICZN 1999: Art. 12.2.7). Thus, the name is treated as valid under the ICN (Turland et al. 2018: Art. 38.7).

The search for S. leptostauron in Ehrenberg’s sample 2726 was not successful (contrary to S. crux). Due to the usage of Canada balm by Ehrenberg, the marked parts of the deposited mica were obscured and Ehrenberg’s specimens of this species could not be identified. Ehrenberg’s drawings, based on his own observations, show two different valves, one without striae and one with striae (Fig. 4A). According to Ehrenberg’s original drawings and measurements, Biblarium leptostauron might be 30–35 µm long with a width/length ratio between 0.52 and 0.55 (vs 0.67–0.78 in Biblarium crux as Ehrenberg called the latter taxon in 1845). The illustrated valve is more heteropolar than in B. crux and near the apices nearly parallel and much narrower than in B. crux (width-near-the-apex/width 0.23–0.30 vs 0.37–0.43 in B. crux). Biblarium leptostauron shares the main character (the “leptostauron”) with B. crux that was drawn with a raphe-like transapical structure and a central node due to misinterpretations by Ehrenberg of lengthwise broken valves. The striae of one valve (6/10 µm) of B. leptostauron are in the range of B. crux but drawn not as broad as in B. crux, the other valve lacks striae. It is questionable that both B. leptostauron valves are conspecific. It should be mentioned that Ehrenberg interpreted his Cassel material in different ways and his treatment in the Mikrogeologie (Ehrenberg 1854) was probably merely based on his old drawings instead of on new observations. It is also possible that the valve lacking striae is simply a representation of a girdle band. Thus, B. leptostauron as Ehrenberg conceived it, remains doubtful and in contrast to B. crux insufficiently described.

When Hustedt (1931) transferred Biblarium leptostauron to the genus Fragilaria as F. leptostauron (Ehrenb.) Hust., he listed many synonyms including Odontidium informe W.Sm. (Smith 1857: 10) and Staurosirella harrisonii (as Odontidium or Fragilaria). As these names are no longer considered to be synonyms of S. leptostauron (see the results in this paper), Hustedt’s transfer of Biblarium leptostauron to the genus Fragilaria in his treatment is not helpful to identify what should be comprised under that name.

Grunow (1862) mentioned on page 368 Biblarium leptostauron as a synonym of Fragilaria harrisonii (W.Sm.) Grunow, clearly being unaware of Navicula/Biblarium crux. He also described at the same time two varieties: F. harrisonii var. rhomboides Grunow and F. harrisonii var. dubia Grunow. Analysis of the type material of F. harrisonii var. rhomboides showed that its morphology corresponds to that of Ehrenberg’s drawings (Ehrenberg 1854: plate XII, figs 35–36) (Fig. 4A) for Biblarium leptostauron, suggesting a likely conspecificity. However, Grunow (1862) linked B. leptostauron with F. harrisonii var. genuina, his name for what he considered the nominate variety of F. harrisonii, whereas for F. harrisonii var. rhomboides Grunow listed Odontidium harrisonii var. β and Staurosira pinnata Ehrenb. as synonyms. Unpublished data show that Staurosira pinnata in fact represents an entirely different species, often confused with Staurosira construens (Bart Van de Vijver unpubl. data).

Odontidium harrisonii var. β was first mentioned in Roper (1854: 77) as Odontidium tabellaria W.Sm., when Roper presented the observations he made on a handful of samples collected a few years earlier from the River Thames. Roper’s specimens were identified as Odontidium tabellaria by William Smith who “from a drawing, thought might be referred to his Odontidium tabellaria […]” (Roper 1854: 77; Roper’s published illustrations are labelled Odontidium harrisonii ?, Roper 1854: pl. VI, fig. 6a, b). To complicate matters, Odontidium tabellaria was not formally described until two years later by William Smith (1856: 17) and represents actually Staurosira tabellaria (W.Sm.) Leud.-Fortm. Roper’s illustrations (Roper 1854: plate VI, figs 6a, b) do indeed look similar to that of Smith (1856: 18, plate LX, fig. 374), who associated it with his newly described O. harrisonii W.Sm. But as the valves were smaller with more acute angles, Smith (1856) described it as O. harrisonii var. β (Supplementary materials 5–6; Gregory had noted this taxon earlier in Gregory 1854: 100, pl. IV, fig. 22, but the drawings are suggestive of something different altogether). In his comments, Smith (1856: 18–19) noted that this variety β resembled O. harrisonii so closely that he had no doubt as to include it as one of its varieties, although Smith never validly described it at the variety rank, despite the fact that he stated that O. harrisonii var. β was a widespread diatom taxon. Fragilaria harrisonii var. rhomboides was described six years later by Grunow (1862: 368) from Moosach (near Munich, Germany) (Fig. 4) without an illustration. The drawing in Smith (1856: plate LX, fig. 374) (Supplementary material 5A), shows a valve with a more rhombic, only weakly cruciform valve outline, very similar to Biblarium leptostauron. Fragilaria harrisonii var. rhomboides from Moosach should therefore be considered as a synonym for Biblarium leptostauron, and the material from the Grunow population offers a good opportunity to characterise the morphology of the latter.

Morales and Manoylov (2006b: 352, figs 26–33, 96–101) discussed the morphology of Fragilaria harrisonii var. rhomboides and this taxon was raised to species level in the genus Staurosirella as Staurosirella rhomboides (Grunow) E.Morales & Manoylov (Morales et al. 2010a: 43), however, basing their account on a North American population. Unfortunately, they did not study the original (type) material from Moosach, nor any other specimens mentioned in Smith (1856). However, after taking all these materials into account, there are clear differences between the North American population identified as S. rhomboides, and the original Moosach population considered to be the type. In Kützing’s sample from Moosach, a similar population was found (Supplementary material 2K–Z; also BM 18680, 18070, “Moosach”) compared to Grunow’s sample from the same locality. All valves in the Moosach population have a clear cruciform valve outline that is retained even in the smallest specimens (see Fig. 5L and Supplementary material 2Z), whereas the North American population has a more lanceolate to rhombic-lanceolate shape (Fig. 5). Additional differences in the structure of the apical pore field, the presence of short, parallel ridges on the virgae in line with the vimines (absent in the North American population) and the structure and more complex shape of the marginal striae, in our opinion exclude conspecificity (compare to Morales and Manoylov 2006b: figs 96–101). A comparison of the morphology of both populations made clear that the North American population represents a new species that will be described as Staurosirella moralesii Van de Vijver, Jüttner & D.M.Williams sp. nov.

In their analysis of the Odontidium species described by Smith (1856), Morales et al. (2015) discussed Odontidium harrisonii and its variety β based on material from Burnham, Norfolk, United Kingdom, stating that this could be the type material for O. harrisonii. Most likely this is not the case, as the largest and most typical population of O. harrisonii is in the Hull material (see above), as also indicated by Smith (1856: 18). Morales et al. (2015: 458) wrote “We searched for the var. β in the same material we analysed for the nominate variety but could not find any organism resembling the drawing presented by Smith (1856: supplementary plate 60, fig. 374).” Our analysis showed that a small population of the variety β is indeed present in the Burnham sample (Supplementary material 5A–J). Moreover, in the Walker Arnott diatom collection (part of the Van Heurck collection in BR), several populations from the British Isles (Supplementary material 7) listed in Walker Arnott’s hand-written catalogue as O. harrisonii var. β, have a similar morphology compared to that in the Moosach population, confirming their conspecificity. Since Grunow used the Moosach population as material for the species description, we formally designate this material as lectotype for S. rhomboides.

In Europe, several populations were identified as Staurosirella rhomboides based on Morales and Manoylov (2006b) or Staurosira mutabilis (W.Sm.) Grunow as Werum and Lange-Bertalot (2004: plate 1) erroneously identified this taxon in 2004. Based on their morphological characteristics (valve outline, valve dimensions), the description of a new species, Staurosirella neorhomboides Van de Vijver, Kusber & Jüttner sp. nov. (Fig. 7), can be justified. Only a handful of published illustrations of these European populations could be found. Werum and Lange-Bertalot (2004) erroneously identified valves from Bavaria (Germany) as Staurosira mutabilis, discussing the rationale for using Staurosira instead of Staurosirella in the figure legend, an opinion now abandoned. In Morales et al. (2015), the type material of Staurosirella mutabilis (W.Sm.) E.Morales & Van de Vijver is illustrated, clearly contradicting the identification of the valves as S. mutabilis in Werum and Lange-Bertalot (2004). Peeters and Ector (2017: 267) reported this species from Burgundy (France) as Staurosirella sp4, illustrating a large population using LM and SEM. On the website ‘Diatom Flora of Britain & Ireland’, a population from Killen Burn (Scotland, United Kingdom) is shown as S. rhomboides (Jüttner et al. 2022). The newly described species S. neorhomboides, based on the Killen Burn population, differs from S. rhomboides (= S. leptostauron) in lacking the typical cruciform valve outline, and the largest valves have a more rhomboid outline instead (compare Figs 4 and 6). The smallest valves, however, have a more elliptic-lanceolate outline. The North American population, identified by Morales and Manoylov (2006b) as S. rhomboides, now described here as S. moralesii (Figs 5–6), is usually broader (valve width 5.5–9 µm vs 4–7 µm in S. neorhomboides), has a lower stria density (7–9 vs 10–11 in 10 µm) and a more elongated valve outline reaching almost 40 µm as maximum length (compared to maximum 18 µm in S. neorhomboides). Further differences are found in the ultrastructural details, such as spine shape (broad, singular in S. moralesii vs multiple, granulate in S. neorhomboides), the structure of the apical porefield (multiple rows of small pores in S. moralesii vs multiple slit-like rows in S. neorhomboides), the width of the striae (broader than the virgae in S. moralesii vs narrower in S. neorhomboides), and the depressed headpole in S. moralesii, a feature absent in S. neorhomboides. Therefore, both new species (Figs 5, 7) should be separated as different species. They also clearly differ from Staurosirella crux (Fig. 1), S. leptostauron (Fig. 4), and S. informis (Fig. 3).

Analysis of the type material of Odontidium informe (Fig. 3) showed that there are several morphological differences (more rhomboid instead of cruciform valve outline, less protracted apices) compared to the other species warranting its transfer to the genus Staurosirella as a separate species: Staurosirella informis (W.Sm.) Van de Vijver comb. nov. (Fig. 3). After it was validly described in 1857, this species has been rarely reported since, except in non-published collection catalogues and sample collections of the nineteenth century (Bart Van de Vijver pers. obs.). Another example of such a ‘forgotten taxon’ is Synedra fontinalis W.Sm., now Fragilaria fontinalis (W.Sm.) Van de Vijver et al., described from the same floristic excursion William Smith made to the Pyrenees in 1856, and recently restudied based on the type material at Edinburg Botanical Garden (Smith 1857; Van de Vijver et al. 2021c). The Scottish botanist George A. Walker Arnott (1799–1868) listed several samples in his collection from the British Isles as containing S. informis (as O. informe), one of which was also analysed in the current study to establish conspecificity (Supplementary material 4) with the original Pyrenees material.

It is possible that Fragilaria leptostauron var. woerthensis A.Mayer (Mayer 1937: 73) might also represent S. informis, but analysis of Mayer’s type material from Bavaria (Germany), which was not available for this study, will be necessary to confirm this hypothesis. It is unclear whether Mayer’s material still exists (Rolf Klee pers. comm.). The original drawings in Mayer (1937: plate IV, figs 14–16) show a lanceolate-rhomboid, isopolar valve, most likely identical to S. informis, but more detailed morphological observations are necessary to confirm this. Bey and Ector (2013: 268) illustrate a taxon provisionally named ‘Staurosirella chavauxii nom. nud.’, which, according to Bey and Ector (2013) “will be described by E. Morales”, but that has never been done (Guiry and Guiry 2023). It appears as ‘S. chavauxii Morales et al. in prep.’ in a study on the biodiversity of calcareous springs in Poland (Okoń et al. 2020: 656). The absence of published records of S. informis in the twentieth century is most likely a consequence of the fact that the species has been confused with or has been considered a synonym of S. rhomboides or S. leptostauron. The results of the present study contradict now this presumed synonymy and reinstall the taxon as an independent species.

One taxon, often considered morphologically closely related to S. leptostauron, but with a heteropolar, ovoid valve outline, is Staurosirella dubia, first described as Fragilaria harrisonii var. γ dubiaGrunow (1862: 368, plate IV(7), fig. 8a–d) from the Stienitzsee (Lake Stienitz) near Berlin (Germany). The original material of the latter, catalogued by Grunow as his sample 552, shows a rather large, robust, solitary species of Staurosirella. This is in contrast to the population indicated by Morales and Manoylov (2006b: figs 13–25) showing isopolar, linear to linear-lanceolate valves with weakly protracted (in longer specimens), acutely rounded apices. The original (unpublished) drawings Grunow made of the Stienitzsee population (Fig. 10A) confirm the heteropolarity of the species, although these drawings clearly differ from some of the valves used to illustrate Fragilaria harrisonii var. γ dubia in Grunow (1862: plate IV(7), fig. 8a–d), the basionym for S. dubia, showing mostly isopolar valves with rostrate apices. The origin of the latter drawings could not be determined as no sample number was written next to these original Grunow drawings in the Grunow drawing collection held at W.

Heteropolar valves similar to S. dubia are often reported as Staurosirella martyi in the literature. For instance, LM images in Lange-Bertalot et al. (2017: plate 11, figs 47–51), identified as S. leptostauron var. dubia (Grunow) Edlund, a synonym of S. dubia, present less similarity with the type of S. dubia in the Grunow material, than the valves illustrated on the same plate as S. martyi (Lange-Bertalot et al. 2017: plate 11, figs 52–56). It is therefore very likely that valves identified as S. martyi in fact represent S. dubia. Conspecificity of the type of S. dubia and the presumed S. dubia population in Morales and Manoylov (2006b) has to be excluded, not only due to the type specimens of S. dubia being heteropolar and its valve dimensions, but also based on the absence of spines (and spine vestiges) in the type population, which is contrary to the North American population where all valves possess double marginal spines on the virgae between striae (see Morales and Manoylov 2006b: figs 90–92, 95). The confusion probably occurred because the type material of both the original material of Fragilaria harrisonii var. dubia and Opephora martyi Hérib., the latter being the basionym for S. martyi, was never studied before. Witkowski et al. (1996) discussed the biogeography and morphology of S. martyi (as Fragilaria martyi (Hérib.) Lange-Bert.). However, none of the illustrated valves were taken from the original material from Neussargues (France). The valves from Joursac (France) (Witkowski et al. 1996: figs 1–6) in fact represent Opephora cantalense Hérib., a species also described in 1903 by Héribaud and now considered a synonym of S. martyi (Krammer and Lange-Bertalot 1991; Witkowski et al. 1996). The other valves illustrated in Witkowski et al. (1996: figs 7–52) originate from various populations worldwide (e.g. Austria, Chile, Germany, Iceland). Based on our analysis of the S. martyi type material, it is clear that most likely these valves do not belong to S. martyi and represent several (probably new) taxa.

Our analyses of the type material of both S. dubia and S. martyi revealed, however, that both species share a large number of common morphological features, such as the absence of marginal spines, a heteropolar valve shape, the shape of the apical pore field, and an apical depression at the headpole. Larger valves in the fossil type material of Staurosirella martyi (Figs 8–9) from Neussargues (France) have a typical clavate outline with a clear constriction between the valve centre and the headpole, a feature often lacking in the populations of S. dubia (Fig. 10). Valves in the type population of S. martyi tend to be longer with a length of up to 40 µm whereas S. dubia valves in the type material never exceeded 25 µm. However, in another (historical) population, found in a sample collected near Ormesby (Norfolk, United Kingdom) and present in the William Smith collection in BR (Supplementary material 8), longer valves were observed (up to 35 µm). The only difference between the type of S. dubia and the type of S. martyi seems to be the stria density with a slightly lower stria density in S. martyi (6–7 in 10 µm) compared to 7–8 in 10 µm in S. dubia, but this difference seems so minor that conspecificity between the two species is very likely. Although Grunow described Fragilaria harrisonii var. γ dubia in 1862 and Opephora martyi was described 40 years later, the latter will still have priority in the species rank following ICN Article 11.2 (Turland et al. 2018).

We also investigated the type population of Opephora cantalense (Fig. 11) from nearby Joursac (France), but could not find any morphological differences between these taxa. The main difference Héribaud (1903) indicated was the broader sternum in O. cantalense compared to S. martyi (“Se distingue de 1’Opephora martyi par la forme de l’aréa, par le nombre et la disposition des côtes, ainsi que par la striation de la face connective.”), and a different stria number in 10 µm (6 for S. martyi, 5.5–6 for O. cantalense). These differences are insufficient to separate these taxa. Héribaud also described several varieties in either taxon as var. capitata, but analysis of the original material showed that they simply belong to the normal cell diminution series, and therefore should not be considered as separate taxa. Héribaud (1908: 10) also described the heteropolar Opephora glangeaudi Hérib., illustrated by two line drawings (Héribaud 1908: plate XIV, figs 17–18), which is possibly another species of Staurosirella. Unfortunately, the original material from the Dépôt de la Garde (Cantal, France) could not be accessed. Based on the two small drawings of Héribaud, it is impossible to accept or reject conspecificity with either species of Staurosirella.

Round proposed the new genus Martyana Round for Opephora martyi, based on the absence of an apical pore field at the headpole, the presence of a step at the headpole, the absence of spines, and the structure of the areolae and striae (Round et al. 1990: 673). Morales and Manoylov (2006b) already discussed the genus Martyana in detail and concluded that there were insufficient morphological characters to erect a new genus and included Opephora martyi in Staurosirella, making the genus Martyana a younger synonym of the genus Staurosirella.

An analysis of the type material of S. martyi (and the populations identified as S. dubia) also indicated that the populations used to illustrate these species in Morales and Manoylov (2006b) do not belong to either, but represent new species. The population of S. martyi from Willow Creek (USA) was re-investigated here (Figs 12–13). The observed population shows some similarity with the type of S. martyi, but all valves are much wider (7–10 µm with only 5–7.5 µm width for S. martyi) and larger with a length in several specimens reaching 70 µm, given the valves a more robust outlook. Moreover, the areolae in the Willow Creek sample are usually clearly visible in LM contrary to S. dubia where the areolae are almost never or very difficult discernible. We therefore exclude conspecificity and following a comparison with all previously discussed taxa, the population is described as a new species: Staurosirella manoyloviana Van de Vijver, Jüttner & D.M.Williams sp. nov. (Fig. 11). Another morphological closely related species, Staurosirella subrobusta E.Morales & Manoylov, has comparably large valves but has an almost isopolar valve outline, a broader sternum, and a dense series of granulate spines on the valve margins (Morales and Manoylov 2006b), the latter absent in S. manoyloviana.

The results of our study highlight the importance and even necessity of including historical material in taxonomic studies of diatoms. Although phycologists in the nineteenth century, such as Ehrenberg and Grunow, usually made very accurate drawings, comparison between these drawings and LM observations is not always easy and misinterpretations can occur, leading to incorrect identifications, taxonomic drift, and/or shoehorning new populations into historical names. It is therefore crucial that taxonomic analyses, especially those involving taxonomic changes (new combinations, transfers, and synonymies), include the revision of original material, when it is available. A better analysis and documentation of historical material is the basis for avoiding making incorrect or incomplete taxonomic interpretations. Historical diatom collections, such as the Grunow collection in Vienna (W), have become more accessible, facilitating the search for these materials (Schuster et al. 2023).