Taxonomic novelties in Central African grasses (Poaceae), Paniceae 1

The Flore d’Afrique centrale initiative started back in 1942 aiming to treat all plant species of the Democratic Republic of the Congo (DRC), Rwanda and Burundi (Robyns 1946, 1949). The first series, of what was then called the Flore du Congo belge et du Ruanda-Urundi, deals with the Spermatophytes, the second one with the Pteridophytes. Series 3 and 4, on mosses and algae, were foreseen, but never realized. To date, about 60% of the estimated 11,000 species of vascular plants have been treated (Léonard 1994). Recently, a serious attempt has started to finish the remaining family treatments for the Flore d’Afrique centrale series 1 and 2 within a period of fifteen years (Sosef et al. 2014). More about the history of this Flora and its recent advancements have recently been published (Sosef 2016). One of the major families still to be dealt with are the Poaceae, with an estimated number of species occurring in that region of close to 750. Our preparation of a manuscript treating the grasses for the Flore d’Afrique centrale started end of 2013. The tribe Paniceae (sensu Clayton & Renvoize 1986 and Watson & Dallwitz 1992, not sensu Soreng et al. 2015) is the first to be dealt with. It is one of the largest in Central Africa in terms of the number of species (to date close to 200). Meanwhile, a fair number of taxonomic novelties have turned up, which seem out-of-place in a Flora treatment. Therefore, it was decided to publish these as a series of separate precursory publications, also so as to reach a wider audience. More information about the history of this Flora and its recent advancements is provided by Sosef (2016).


INTRODUCTION
The Flore d'Afrique centrale initiative started back in 1942 aiming to treat all plant species of the Democratic Republic of the Congo (DRC), Rwanda and Burundi (Robyns 1946(Robyns , 1949)).The first series, of what was then called the Flore du Congo belge et du Ruanda-Urundi, deals with the Spermatophytes, the second one with the Pteridophytes.Series 3 and 4, on mosses and algae, were foreseen, but never realized.To date, about 60% of the estimated 11,000 species of vascular plants have been treated (Léonard 1994).Recently, a serious attempt has started to finish the remaining family treatments for the Flore d'Afrique centrale series 1 and 2 within a period of fifteen years (Sosef et al. 2014).More about the history of this Flora and its recent advancements have recently been published (Sosef 2016).
One of the major families still to be dealt with are the Poaceae, with an estimated number of species occurring in that region of close to 750.Our preparation of a manuscript treating the grasses for the Flore d'Afrique centrale started end of 2013.The tribe Paniceae (sensu Clayton &Renvoize 1986 andWatson &Dallwitz 1992, not sensu Soreng et al. 2015) is the first to be dealt with.It is one of the largest in Central Africa in terms of the number of species (to date close to 200).Meanwhile, a fair number of taxonomic novelties have turned up, which seem out-of-place in a Flora treatment.Therefore, it was decided to publish these as a series of separate precursory publications, also so as to reach a wider audience.
More information about the history of this Flora and its recent advancements is provided by Sosef (2016).

Taxonomic studies on Central African grasses
The taxonomic study of Central African grasses basically started with the general work of Steudel (1854), but at his time only very few collections from DRC, Rwanda and Burundi were available.Only those collected in 1816 by the British Prof. Christian Smith (620 numbers in total, 45 of which are grasses) during his unfortunate journey under the command of Captain Tuckey (Brown 1818) were at Steudel's disposal.It was not until the 1870s that additional plant collections were made in DRC, mostly by several German collectors (Robyns 1949).Shortly after 1885, when the Belgium King Léopold II created the Independent State of the Congo, intensive botanical inventory work started.Franchet (1895) was the first to specifically study Central African grasses, and although he dealt with the French Congo, now the Republic of the Congo, and Gabon, he laid a firm foundation for further studies on grasses in the region.Shortly after that, De Wildeman studied the grass collections now coming into Belgium by the hundreds, or even thousands, due to an unprecedented collecting effort, followed by an equally impressive research effort, culminating in the first annotated checklist of Congolese plants, the Sylloge Florae Congolanae (Durand & Durand 1909).In the early 1900s, during the colonial period, several test gardens (Jardins Agrostologiques) were created by INEAC (Institut National pour l'Étude Agronomique du Congo Belge) to specifically study grasses for their agronomic characteristics.These were realized in various places of DRC (amongst others in Kisantu, Yangambi, Levervillenow Lusanga, Bandundu prov.-and Eala).It was Father H. Vanderyst who published several articles on these and other grasses in the region (Vanderyst 1918(Vanderyst , 1919(Vanderyst , 1925)), although he was much hindered by the limitations World War I implied on the access to collections and libraries.This made him decide to publish his new names as being provisional and therefore invalid in the light of our present-day Code (McNeill et al. 2012, Art. 36.1b).In that same period, Stapf produced his Magnum Opus on the grasses in the Flora of Tropical Africa series (Stapf 1919(Stapf , 1920)), describing a large number of new species based on collections from DRC. Vanderyst continued his studies on grasses but with more focus on grass vegetation types and their agronomic values (for example Vanderyst 1932).By that time, Walter Robyns, who later became Director of the National Botanic Garden of Belgium (then Jardin botanique de l'État) and would be the main driving force behind the start of the Flore d'Afrique centrale series in 1942, had taken a keen interest in grasses and published two volumes of a Flore agrostologique du Congo belge et du Ruanda-Urundi (Robyns 1929(Robyns , 1934) ) dealing with the tribes Paniceae, Maydeae and Andropogoneae (with Robyns 1932 being a precursor for his 1934 publication).Unfortunately however, he was never able to complete this work and no major studies on the grasses of Central Africa were produced for over fifty years.Then, the grasses of Rwanda were dealt with by Troupin (1988), while Ndabaneze (1989) provided a most useful checklist of the grasses of Burundi.
To date, the grasses of most of the countries neighbouring DRC, Rwanda and Burundi have been treated, either as an account in a Flora series (Clayton 1970, 1989, Clayton et al. 1974, Clayton & Renvoize 1982, Cope 1999, Koechlin 1962, Launert 1971, Sosef 1999) or as a separate publication (Harker 1960, Kami 1997, Napper 1965, van der Zon 1992).Apart from the DRC, notably the Central African Republic and Angola still remain without a specific treatment of their grass flora.

MATERIALS AND METHODS
All studies were performed using herbarium material and applying standard herbarium techniques (de Vogel 1987).Acronyms of institutes holding herbarium collections follow Thiers (continuously updated).The vast majority of material came from BR, with additions from BRLU, GENT, LG and WAG.Types not available in one of these, were generally studied using the Global Plants facility (JSTOR 2000(JSTOR -2016)).
Brachiaria glycerioides Chiov.has been regarded as a synonym of Urochloa deflexa (Schumach.)H. Scholz (Clayton et al. 2006 and various other databases), but the original description clearly mentioned the characteristic flattened tip of the lower glume of Acroceras, and the auriculate leaf base of A. amplectens.Moreover, I was able to study a photocopy of the type specimen, kept at Modena, Italy (MOD), which confirmed the above.Stapf (Stapf 1920: 624).-Type: Angola, S. Angola, Cunene marshes, 21 May 1909, H.H.W. Pearson 2024 (lecto-: K, designated here).Stapf (1920: 624) mentioned four specimens which are to be regarded as syntypes.Up to now, no author seems to have chosen a lectotype from amongst these, although Pearson 2024 at Kew bears a label indicating it is the 'lectotype'.As there is not much difference between the four specimens, which all carry an identification tag from O. Stapf, it seems best to follow Kew's 'suggestion' and to formally designated Pearson 2024 as the lectotype.
Panicum guluense is an obscure name, invalidly published by Vanderyst.In GrassBase (Clayton et al. 2006) is was previously treated as a synonym of Panicum mueense Vanderyst, but it seems unclear from where this information originated.The type specimen was located at BR and turned out to represent Acroceras zizanioides.
Although widely introduced in the tropics as a forage grass and a regular escape, this species was never reported from the wild in the DRC.Various specimens prove it was introduced in several experimental gardens from the early 20 th century onward.Now, two specimens, A.Schmitz 489 and Lisowski 53819, collected in 1947 at Keyberg and in 1970 at Lubumbashi respectively, indicate it seems to have escaped into the wild at least in the South-East of the country (Haut-Katanga).

Paspalum distichum L.
Although widely distributed, this species is rare in Central Africa.Four specimens, Lejoly 5162, Lisowski 49764 and 52484 and Nyakabwa 1, collected in the years 1978, 1979 and 1981 probably from the same locality along a road in Kisangani, Province Orientale, DRC, represent the first records of this species in this country.Their locality suggests an introduction but of unknown origin.

Paspalum glumaceum Clayton
This species was published in 1975, as a segregate from the P. scrobiculatum complex (Clayton 1975, Koning & Sosef 1985) and said to occur in Sudan, Uganda, Kenya, Tanzania, Malawi, Zambia, Zimbabwe and Madagascar.Several specimens were discovered from savanna regions in the North and South-East of the DRC (Garamba National Park, Upemba National Park and Katanga province), thus extending its known range.

Urochloa
The genus Urochloa P.Beauv.has often been recognized as being a close relative of the younger but much more speciose Brachiaria (Trin.)Griseb.However, although seemingly important characters such as the presence of a mucro on top of the upper lemma, the position of the lower glume (adjacent to or facing away from the rachis) or the apex of the spikelet being rounded to acute, or rather long acuminate, may provide diagnostic features for a distinction at regional level, these characters do not hold at the global scale.Moreover, over the past decades, morphological as well as molecular evidence has accumulated showing that in fact the majority of the species are better accommodated within Urochloa (see notably Webster 1987, 1988, Morrone & Zuloaga 1992, and Torres González & Morton 2005 and references therein).An exception could be made for three species: the widely distributed Brachiaria eruciformis (Sm.)Griseb., B. malacodes (Mez & K.Schum.)Scholz and B. schoenfelderi C.E.Hubb.& Schweick from Zimbabwe, Angola and Namibia, respectively.Unfortunately, because of a typification issue, the name Brachiaria could not be maintained for this small group of species.A proposal to conserve the name by Veldkamp (1996) was rejected (Brummitt 1998), reason for Veldkamp (2004) to, reluctantly, erect the new genus Moorochloa Veldk. to accommodate these species.
Although the merger of the majority of Brachiaria with Urochloa has found support from experts working on South American, Asian and Australian grasses, African grass specialists seem to have been reluctant to accept it.As a result, many of the African Brachiaria species have no name available in Urochloa yet.Below, these are provided for those occurring in the Central African region, along with several taxonomic remarks.Although above I argue the necessary name changes of almost all African Brachiaria species, it seemed unwise to provide new combinations in Urochloa for the ones not occurring in Central Africa of which I did not study the taxonomy in detail.The two Uitenhage specimens have not been located, but K holds a specimen of each of the Drège collections.The various annotations, amongst which the name Panicum subquadriparum in Nees' hand, are more or less identical, but the sheet chosen to serve as the lectotype carries an analysis though of unknown origin.Both sheets clearly represent the species now known as Urochloa arrecta.
Specimens of the closely related U. mutica (Forssk.)T.Q.Nguyen may occasionally have only few paired spikelets in each raceme and can thus easily be confused with U. arrecta.The two seem sufficiently distinct based on additional characters of the rachis (0.5-1 mm wide in U. mutica and 1-1.5 in U. arrecta) and the culm nodes (densely bearded in U. mutica and glabrous to pubescent in U. arrecta).
Although Rendle provides two syntypes, Welwitsch 2726 and 2829, Stapf (1919) distinctly chose the latter arguing that the first belonged to a different taxon.As such, Stapf effectively lecotypified the name with Welwitsch 2829.
In the protologue of Panicum comatum, Richard (1850: 376) cited two collections, Quartin Dillon s.n. and Schimper 1196, both from Ethiopia.The latter has been chosen here as lectotype, because it has many more duplicates and is therefore much wider distributed and more easy to consult in the future.It concerns a form with upper glumes and lower lemmas carrying a horizontal line of long white hairs at about ¾ from the base.This form seems to be much more common in eastern Africa.
The name Panicum scalare Mez is to be regarded as a later homonym of Panicum scalarum Schweinf., and as such is illegitimate according to the ICBN (McNeill et al. 2012).Scholz (1978) argued that the name was therefore not avail-able for a combination within Brachiaria, and hence that B. scalaris Pilg.was illegitimate, and proposed the new name B. pilgerana.However, Art. 58.1 of the ICBN (McNeill et al. 2012) is clear in stating that a later homonym is available for re-use in a new combination, but that the author of the later homonym is not referred to in the new combination and that the name has priority only from the publication date of the new combination.This renders B. pilgerana superfluous.
Urochloa comata has generally been keyed out by the fact that its spikelets are born in fascicles or even on short sidebranches along the raceme.However, various grass publications (for example Clayton 1989, Clayton & Renvoize 1982, Phillips 1995) state that occasionally such structures are also present in the closely related Brachiaria scalaris, though only at the base of the lower racemes.After careful examination, I have observed the spikelet structure of the two taxa to be quite the same, and reached the conclusion that the latter is to be regarded as a poor form of U. comata.Similarly, Brachiaria coronifera, of which the distinction from B. scalaris was already doubted by for example Clayton & Renvoize (1982: 596), turned out to represent similar poor forms, but with a horizontal line of hairs below the top of the spikelet (like in the type of U. comata, see above).This rather wide variation in spikelet indumentum is known to occur in several species of this group though.A paratype specimen of B. coronifera at BR (Schlieben 768) even represents a mixture of both forms.For further information, see below in the discussion following U. villosa.Many authors have treated U. humidicola as a species distinct from U. dictyoneura, based on the fact that the first would be a stoloniferous (not tufted) plant, with only 2 or 3 racemes, smaller spikelets (3.8-5(-5.2) mm long) and leaves lacking a contra-ligule.After a careful study of the Central African specimens, I concluded that this distinction does not hold.Several specimens in fact show that stolons grow out of a fairly dense tuft, while the other characters show a continuum and hence overlap between the two taxa.I therefore follow Veldkamp (1996) who already united the two.Chromosome counts for U. dictyoneura s.s. of 2n = 42 and for U. humidicola of 2n = 72 have been published (Schultze-Kraft 1992, Schultze-Kraft & Teitzel 1992) suggesting that a polyploid complex may be at hand.
Most recent authors have distinguished three closely related species: U. eminii, U. decumbens and U. ruziziensis.All possess a winged rachis with ciliate margins, a clasping lower glume concealing a short internode between the lower and upper glume, and more or less pubescent (rarely even glabrous) spikelets.U. ruziziensis was described (as Brachiaria ruziziensis) indicating that the main difference between this species and U. eminii would be the presence of a short internode between the lower and upper glume which U. eminii would lack.This clearly is an erroneous observation, since U. eminii also possesses such an internode (also verified on the isotype at BR).Then, the taxon U. decumbens was already treated as being conspecific with U. eminii by Robyns (1934), but subsequent authors (notably the authoritative Clayton & Renvoize 1982) kept it distinct, based on the fact that U. eminii would be a tufted annual with lower glumes more than ½ the length of the spikelet while U. decumbens would be a stoloniferous perennial with lower glumes less than ½ as long as the spikelet.Kami (1997), in his treatment of the grasses of the Republic of the Congo, and Morrone & Zuloaga (1992) in their overview of South-American species of Brachiaria and Urochloa, distinguished U. decumbens and U. ruziziensis based on their resp.perennial and annual habit and the presence/absence of a well-developed terminal spikelet.Furthermore, these authors indicate that the rachis width is distinctive: in U. decumbens it is 1-1.7 mm, in U. eminii 2-3 mm and in U. ruziziensis 3-5 mm.In his Catalogue of Burundi grasses, Ndabaneze (1989) reduced U. ruziziensis to a variety of U. decumbens, unfortunately without any argumentation.Troupin (1988) stated that Ndabaneze's decision was based on the presence of intermediate specimens.Finally, Clayton & Renvoize (1982) noted that U. ruziziensis could well be nothing more than a local segregate of U. decumbens.
Upon closer study of the fairly abundant material available at BR, I came to the conclusion that none of the characters above hold.An apparently annual habit can be observed, but this is not correlated with any of the other characters, and could simply relate to the specific plant being young.Stolons (basically prostrate culms rooting at the nodes) are very frequently occurring, also in seemingly annual plant material.The length of the lower glume is variable, but again, does not seem to have any correlation with for example rachis width.In conclusion, it is best to regard these names as referring to a single, though variable, species.
Occasionally, some specimens have glabrous spikelets (a form described as B. bequaertii) although in that case sparsely pubescent spikelets were sometimes observed on the same plant, confirming that this falls within the variability of this species.The distinction with the related U. rugulosa (see be-low) becomes difficult, but can be established based on the much more rugose lemma of the upper floret, the generally shorter lower glume with only 5 or 7 veins (not 9 to 13), and the upright or shortly ascending culms of the latter.Forms with glabrous spikelets may then also be difficult to distinguish from U. brizantha (Hochst.ex A.Rich.) R.D.Webster, especially since in such specimens the spikelets tend to be paler in colour and often tinged with purple, typical for U. brizantha.The latter can be distinguished based on its more plump spikelets in a single row, with blunt and incurved apices of both lemmas.It is not impossible that such 'intermediate' specimens in fact represent hydrids between the two.
In the protologue of B. decumbens, Stapf (1919) mentioned two specimens: Dummer 1070 and Speke & Grant 488.Both are present at K, but the first has several duplicates elsewhere and is therefore selected as the lectotype.
The protologue cited two collections, Schimper 289 from Aduoa and Schimper 1833 from Chiré.Duplicates of both collections are available in many herbaria and represent ample material of the same species.Thus, without any clear reason for a preference, the specimen cited first in the protologue was chosen as the lectotype.
This species is difficult to distinguish from poor forms of U. comata (formerly accommodated in Brachiaria scalaris) and from U. villosa.It differs from both in the shape of its spikelets, which is broadly obovate, its culm nodes which carry a dense tuft of spreading grey hairs and its lower glume being not clasping and surrounding the spikelet base for only about ⅔.In U. comata the lower glume is also not clasping, but its line of insertion surrounds the entire spikelet.In U. villosa the lower glume is clasping, hence with the lower margins running vertical for a short distance after spreading out and there is a (very) short rachilla between lower and upper glume.
In the past, the majority of authors made a distinction at species level between the perennial and annual form, naming them U. mosambicensis and U. trichopus respectively.However, I could not trace any additional morphological feature to support such a taxonomic distinction.In addition, the similar and large variation in pilosity of the spikelets being present in both forms corroborates the idea that a single species is at hand.Life form, although important from the viewpoint of cultivation (the species is traded and is a palatable pasture grass, see Brink 2006aBrink , 2006b)), can in itself not serve as the only diagnostic feature at species level.Reason why these taxa are synonymized here.
Meanwhile, although U. trichopus is widespread in tropical Africa and east to Yemen and Birma, it was never reported from Central Africa.Recently, a wrongly identified specimen collected in 1954 in Haut-Katanga (south-eastern DRC) at BR turned out to represent the first record from this region: van Oosten 240.
This narrowly endemic species occurs only in the Haut-Katanga region of the DRC.It is closely related to U. serrata, but characterized by its annual habit, lanceolate to narrowly elliptic leaf blades, turbinate spikelets, distinctly asymmetric and ligulate pedicel tip and long hairs on the pedicels extending well beyond the spikelet.Several collections from near Kolwezi represent a perennial form with narrower leaf blades.Others from the same region show less turbinate spikelets and less characteristic long hairs on the pedicels.Further study is needed to assess the taxonomic status of these forms and, eventually, this species.
Above, in the discussion following U. comata, it has been explained that B. scalaris is to be regarded as a poor form of U. comata.Indeed, the distinction between such poor forms and U. villosa is quite difficult.After careful examination of all material available at BR (several hundred sheets), including type material of B. scalaris, as well as that available through the JSTOR Global Plants website (http://plants.jstor.org), I have observed that the only difference between the two species seems to be the clasping lower glume in U. villosa (with its lower margins running vertical and touching each other or even slightly overlapping), where it is not clasping in U. comata (margins diverging right from the base).In U. comata the lower floret can be male or sterile, while in U. vil-losa it is always sterile.Spikelets from both species can have a rounded or acute apex.
The species Brachiaria leucacrantha was already suspected to represent not more than a form of Urochloa (Brachiaria) xantholeuca by Clayton & Renvoize (1982: 597) and Clayton (1989: 78).After studying the material at BR, I reach the same conclusion.The spikelets of B. leucacrantha are more strongly acuminate than those of U. xantholeuca and carry two tufts of long hairs on the upper glume and lower lemma.However, slightly less acuminate spikelets sometimes already show this tendency to have longer hairs towards the tip as well, and one may even argue to synonymize both taxa.Because the leucacrantha-form seems to be restricted to eastern Africa, while U. xantholeuca occurs throughout tropical Africa, it is deemed best to assign it some taxonomic status, for which the variety level seems most appropriate.
The protologue of the name Panicum leucacrantum provides three syntypes: Holst 2077 (which could be an error for 2097), 2805 and 4163.In B, Holst 2097 holds a copy of the protologue (and drawing of spikelet details of later date), but that cites #2077, giving rise to some confusion or doubt.The collection numbered 4163 does not have an original label from Holst.Holst 2805 being by far the richest material at B, with an original label, and the only one with known duplicates elsewhere (at least K and M), then seems to be the obvious choice for the lectotype.