Taxonomic novelties in Central African grasses (Poaceae), Paniceae 2

Background and aims – Within the framework of the renewed production of the Flore d’Afrique centrale, the grasses are being studied to accomplish their treatment. Taxonomic novelties, or other information not deemed appropriate in a Flora, are published in a series of separate papers of which this is the second. Methods – Standard herbarium techniques have been applied to material from BR, BRLU, GENT, P and WAG. Some types were studied through the JSTOR Global Plant facility. Key results – Novelties are presented for the genera Anthephora, Cenchrus (incl. Pennisetum) and Setaria. Three new combinations are made. Lectotypes are designated for five names. Ten names are treated as new synonyms of accepted species names, with explanations of the new taxonomic concepts applied.

Anthephora elegans Schreb. var. africana Pilg. (Pilger 1901: 119 There has been some doubt about the correct publication place and date of A. cristata. The name was published without reference to the basionym of Döll and without description by De Wildeman & Durand (1900: 60; although the title page of that journal issue gives "1901" as the year of publication, at the bottom of the first page of the article the publication date is stated to be December 29 th , 1900), and therefore has to be regarded as an invalidly published nomen nudum (Art. 38.1, Turland et al. 2018). The same name, validly published one year later by the same authors and with reference to the basionym, is therefore not to be regarded as a later homonym (Art. 53.1, Turland et al. 2018).
Antephora elegans var. africana was published citing four specimens, Buchholz 1875, Dinklage 464, Dewèvre 120 and Schlechter 12508 which are to be regarded as syntypes. Since the author worked at B, the lectotype should preferably be located there. All except the Schlechter specimen are not present at B and were presumably lost during the 1943 fire. At B, there are two sheets of Schlechter 12508, one of which has no spikelets left, the other with a few spikelets in an envelope glued onto the sheet. The latter is here selected as the lectotype, with several duplicates elsewhere.  (Donadio et al. 2009, Chemisquy et al. 2010, Veldkamp 2014. Cenchrus had about 22 species, Pennisetum some 80, while the other three genera were monotypic. The oldest name being Cenchrus, a fair number of names have already been recombined in that genus, notably by Morrone (in Chemisquy et al. 2010). However, some additional combinations related to central African taxa are useful (in the case of the widely cultivated crop C. americanum) or necessary (in the case of the re-instatement of P. nodiflorum as a distinct species). These are provided below. Lectotypes are assigned where appropriate.
Cenchrus americanus (L.) Morrone (in Chemisquy et al. 2010: 128). This is the well-known and widely cultivated Bulrush millet or Pearl millet. Its complicated nomenclature led to the acceptance of the name Pennisetum glaucum (L.) R.Br. for this species, with P. americanum (L.) Leeke treated as a synomym (see Clayton & Renvoize 1982 for a brief but clear review of the problem). However, in Cenchrus the epithet glaucus is already occupied (C. glaucus Mudaliar & Sundararaj = C. ciliaris L.) and hence the epithet americanum is the next available one and rightfully used by Morrone for the new combination.
This highly variable species is thought to belong to a complex of three taxa that frequently hybridize: C. americanus, C. sieberianus (Schltdl.) Verloove and C. violaceus (Lam.) Morrone; see Andrews & Kumar (2006) for an overview. Brunken (1977), who made a detailed study of the group, recognized these three taxa as subspecies of P. americanum. He considered the latter two taxa to be the wild forms and/or relatives of the cultivated C. americanus and his taxonomy is still widely used in Pearl millet crop science, where also their synonyms at subspecific level, respectively P. americanum subsp. stenostachyum (A.Braun & Bouché) Brunken and P. americanum subsp. monodii (Maire) Brunken, are regularly encountered in recent literature, for example in the authoritative Patil (2016) publication. However, while at species level the new combinations are available in Cenchrus, they are not for those who want to recognize the taxa at the level of subspecies. Regarding the existence of frequently occurring (partly) fertile hybrids between the three taxa, a recognition at subspecific level is probably more appropriate. Hence, to support further research related to the crop Pearl millet and for example gene banks holding accessions of C. americanus and its wild relatives, the correct names in Cenchrus for the two wild subspecies are provided below.
In several databases and literature sources, one may encounter the subspecific names Pennisetum glaucum (L.) R.Br. subsp. sieberianum (Schltdl.) Stapf & C.E.Hubb. and Pennisetum glaucum (L.) R.Br. subsp. violaceum (Lam.) A.Rich. Being older names at subspecies level, the epithets would have priority over the ones proposed below. However, both names were never validly published in the works referred to by the various sources and have thus not been taken into account. Cenchrus macrourus is a highly variable species characteristic of lake and river shores. Where many of its forms were previously recognized as distinct species, notably Clayton & Renvoize (1982) brought together the reticulum of local seg-regates into a single species, a view generally followed by others.
Cenchrus mildbraedii, endemic to Uganda and adjacent Rwanda, is highly similar in habit (coarse rhizomatous perennial with erect culms, sheaths coriaceous and glabrous, the lower ones flabellate), inflorescence (peduncle scabrid or puberulous below the spike-like terminal inflorescence), involucrum with some 10 to 20 bristles of which only the longest is overtopping the single spikelet, and the strongly reduced lower and upper glumes. Cenchrus mildbraedii would be distinct from the widespread C. macrourus due to its much shorter lower lemma, ½ to ⅔ of the length of the spikelet, where that of C. macrourus is between ¾ of the length of the spikelet and full length of the spikelet. The type material of C. mildbraedii was studied in detail, as well as highly similar material collected in 1929 on the type locality, the saddle between the Sabyinyo and Gahinga volcanoes (Humbert 8636, 2 sheets at BR, [BR0000005866847, BR0000005866908]). It turned out that although both glumes are highly reduced, the lower lemma is in fact almost as long as the spikelet. This error probably has its origin in the protologue itself, where Mez published an otherwise highly detailed description in which he stated "gluma I. minutissima vix reperienda, ……; gluma II. item quam maxime diminuta sed quam praecedens paullo longiore, ……; gluma III. item valde diminuta, 1-3 mm longa, squamiformi, apice rotundata, 0-1-nervia" [freely translated: lowest glume very minute, hardly visible, …..; upper glume also reduced though a little longer than the one preceding, …..; glume 3 also very much diminished, 1-3 mm long, scale-like, apex rounded, 0-1-veined"]. In Paniceae, glume 3 is generally seen as the lemma of the lower floret. However, it seems that Mez either miscalculated the number of elements in the spikelet or misinterpreted the spikelet structure, because when comparing with the type material, the description of glume III clearly relates to the upper glume in this case. This observational error has been copied by Stapf & Hubbard (1934), Robyns (1934), Clayton & Renvoize (1982) and finally even in the otherwise high quality Grassbase webservice (Clayton et al. 2006-) and thus remained uncorrected for almost a century.
Pennisetum kisantuense was published rather obscurely, since it was only mentioned in the key of Vanderyst (1925: 685). No specimens were cited, but several collected by Vanderyst carry the name of the species in his handwriting. Clayton & Renvoize (1982: 690) indicated Vanderyst 5030 at K as one of several syntypes, and hence this cannot be seen as a formal lectotypification. Unfortunately, no duplicates of Vanderyst 5030 are present at BR. Hence, I deemed it better to select a syntype which I have been able to study in detail (Vanderyst 4650, Kisantu (Inkisi), Jul. 1914 implying it was already regarded as a synonym by another authoritative work. According to me, and my colleague Ton van der Zon (van der Zon 2019), that is an error. The latter recently published the correct name for the species in the genus Cenchrus, but it seems appropriate to add some explanation on the confusion of names. The species was treated as a synonym of P. divisum (J.F.Gmelin) Henrard (= C. divisus (J.F.Gmelin) Verloove, Govaerts & Buttler; Verloove et al. 2014) by the African Plant Database (2019), which is based on Lebrun & Storck (1995). That database correctly regards P. dichotomum (Forssk.) Delile as a synonym of P. divisum. However, Robyns (1934) (2019), below P. dichotomum (Forssk.) Delile. Possibly, this situation has caused confusion and the error to occur. Otherwise, C. nodiflorus has sufficient diagnostic morphological features, and is characteristic of rapids in the Congo River, from Kinshasa to Kisangani (D.R. Congo), see Léonard (1994).
Cenchrus nodiflorus has a superficial resemblance (many swollen nodes, relatively short inflorescences, upper glume with 5 to 7 nerves, at least ⅔ as long as the spikelet) to C. divisus but is a much coarser species. The plants are up to 2.5 m high, with leaves up to 40 cm long, all bristles of the involucre glabrous (scabrous) and spikelets 3.5-4.5 mm long, while C. divisus reaches up to 1.5 m, with leaves up to 15 cm long, internal bristles plumose, and spikelets 6.5-8.5 mm long.
The epithet would translate to "with knotty flowers" and is often used for species having their flowers in tight glomerules. Since this does not really apply to this species, but rather its culms with swollen nodes are a fairly striking feature, one wonders if Franchet may have made a mistake and wanted to name it 'nodiferum'? This is not overly clear from the protologue though, and it is certainly not a correctable error in the sense of the International Code of Nomenclature for algae, fungi, and plants (Art. 60, Turland et al. 2018). So, this rather strange epithet is the correct one to be used for this species.
Cenchrus trachyphyllus (Pilg.) Morrone (in Chemisquy et al. 2010: 129). -Pennisetum trachyphyllum Pilg. (Pilger 1901: 122 Panicum bongaense Pilg. (Pilger 1902: 44). -Setaria bongaensis (Pilg.) Mez in Mildbraed (1922: 20) The name Panicum thollonii, later correctly transferred to Setaria, has been treated as a distinct species by for example Stapf & Hubbard (1930) and Robyns (1934). Later, it was regarded as a synonym of S. homonyma (Steud.) Chiov. by Clayton (1989), followed by many others such as van der Zon (1992) and Morrone et al. (2014), although Webster (1993) regarded it as a name with unknown status. Morrone et al. (2014) designated the lectotype from amongst two syntypes (Thollon 826 and Hens 62). However, after studying this type material at BR and P, it showed that it deviates in important characters from true S. homonyma. The most striking feature is that often the lower leaf blades sit on a slender pseudopetiole up to 5 cm long. Additional diagnostic characters distinguishing it from S. homonyma are: leaf blades flat, only slightly plicate towards the base, inflorescence axis and rachis scabrous and upper glume covering about ⅘ of the upper lemma which is finely transversely rugulose. S. homonyma has leaf blades without pseudopetioles, that are plicate throughout their length, its inflorescence axis and often also its rachis are pilose (rarely glabrous), while the upper glume completely covers the upper lemma which is more prominently transversely rugose. The characteristics fit within the variation of S. barbata, where the distinction of folds in the leaf blade is highly variable, and where the leaf blade is sometimes contracted into a long pseudopetiole. I could not find any reasons to separate this widespread form of S. barbata into a distinct taxon, since the variation in the above-mentioned characters appears to be continuous.
Two other names, previously treated as synonyms of S. seriata Stapf (S. gracilipes C.E.Hubb., see also the discussion on S. kagerensis below) and of S. homonyma (Steud.) Chiov. (Panicum bongaensis Pilg.), the types of which show high similarity with the type of S. thollonii, are also to be regarded as synonyms of S. barbata rather than of the other two species.
Setaria kwamouthensis was treated as a synonym of S. thollonii by Robyns (1934) but was placed amongst the doubtful and excluded names by Morrone et al. (2014). Two syntypes (Vanderyst 4607 & 4619) were located at BR. Since the first was collected at Lekanu, and the second at Kwamouth, the latter was chosen as the lectotype. Mez (Mez 1917: 58 According to the revision of the Old World species of Setaria by Morrone et al. (2014), the presumed annual species S. seriata appears to be a rare but relatively widespread species, occurring from Ivory Coast to Zambia, with S. gracilipes C.E.Hubb. as a synonym. Upon closer examination of the holotype specimen of S. seriata, the leaf morphology and spikelet structure proved highly similar to that of the perennial species S. kagerensis. Both have plicate but comparatively narrow leaf blades, a paniculate inflorescence with spikelets in ± unilateral racemes, a lower glume of up to ⅓ of the length of the spikelet while the upper glume is ⅘ or more of this length, and a smooth to papillose upper lemma and palea. In the protologue, Stapf remarked that the species was "probably annual", because the only material available consisted of an inflorescence and a single leaf. Later, Clayton (1989) erroneously treated the annual S. gracilipes C.E.Hubb. as a synonym of S. seriata, which was probably the start of a complicated confusion of names and mixing of species descriptions (see also above, the discussion below S. barbata). This situation is now clarified, at least for Central Africa. Setaria microprolepis Stapf has long been treated as a synonym of S. homonyma (Steud.) Chiov., probably because of its presumed transversely rugose upper lemma and an upper glume ± equalling the spikelet. However, S. homonyma is an annual species with narrowly elliptic leaf blades, while the type material of S. microprolepis shows a perennial plant with linear leaves. Upon closer inspection the upper lemma was not rugose, as mentioned in the protologue, but only papillose. Both features then referred the material to S. kagerensis.

Setaria kagerensis
In its present circumscription, S. kagerensis has a distribution centred in East Africa, occurring from Sudan and For several decades, authors have been more or less copying the same remark about the weakness of the distinction between S. acromelaena and S. nigrirostris (incl. S. incrassata (Hochst.) Hack.; see for example Clayton & Renvoize 1982: 527;Cope 1995: 234;Phillips 1995: 238). Most authors stated the first is the annual counterpart of the second. The recent revision by Morrone et al. (2014) stated the same, but gave an additional difference, namely the fact that S. acromelaena would have a sparsely to densely pilose inflorescence axis mixed with the dense short-hispid indumentum, while that of S. nigrirostris only bears a short-hispid indumentum. The latter difference does, unfortunately, not hold, since for example the specimens Bos 9081 from Ethiopia and Comité Spécial du Katanga 10 from south-eastern D.R. Congo clearly represent annual plants but without any pilose hairs on the axis. I have argued before (Sosef 2016b, on the distinction between Urochloa mosambicensis (Hack.) Dandy and U. trichopus (Hochst.) Stapf) that in grasses annual habit alone cannot be used as a taxonomic distinctive character at species level, and hence the two taxa are to be united and their names to be regarded as synonyms.
Setaria abyssinica Hack. var. annua Chiov. has been rightfully regarded as a synonym of S. acromelaena and is hence now also synonymous with S. nigrirostris. The type information has been updated.

ACKNOWLEDGMENTS
I am grateful to the curators of the following herbaria, for providing access to their collections and/or assisting in tracing specific materials: BR, BRLU, GENT, P, WAG. Constructive remarks to an earlier manuscript version were received from three reviewers and the Editor, for which I want to thank them.