Taxonomic novelties in African Dracaena ( Dracaenaceae )

In preparing the treatment of Dracaena for Flore du Gabon and Flore d’Afrique centrale, a relatively high number of taxonomic and nomenclatural novelties were discovered; these are presented here. Within Dracaena five species and one forma are described as new, D. bushii, D. haemanthoides, D. marina, D. wakaensis, D. waltersiae and D. laxissima forma aureilicia. Each new species is provided with a full description and taxonomic notes. Apart from that, five species are reinstated, D. braunii, D. nitens, D. perrottetii, D. tholloniana and D. usambarensis. A further 23 names are treated here as a synonym for the first time: D. bequaertii, D. buettneri, D. cuspidibracteata, D. densifolia, D. gabonica, D. gazensis, D. ledermannii, D. letestui, D. litoralis, D. longipetiolata, D. monostachya var. angolensis, D. oddonii, D. perrottetii var. minor, D. poggei, D. pseudoreflexa, D. reflexa var. buchneri, D. rubroaurantiaca, D. soyauxiana, D. talbotii, D. tessmannii, D. usambarensis var. longifolia, D. vanderystii and Pleomele heudelotii, while for four names a neotype and for 14 names a lectotype has been designated. Distribution maps are provided for a total of 23 species. An index of taxon names is included.


INTRODUCTION
The genus Dracaena L. has approximately 116 species, of which some 63 occur in Africa (incl.Madagascar) (Govaerts et al. 2017).The remainder is found in Asia, Australia and Central America.Most recent studies generally place the genus in the family Asparagaceae subfamily Nolinoideae (Stevens 2001onwards, Lu & Morden 2014, Angiosperm Phylogeny Group 2016), but others have placed it in Agavaceae (i.e., Hepper 1968b) or recognised it as a distinct family, Dracaenaceae, along with the genera Sansevieria Thunb.and Pleomele Salisb.(Bos 1992, 1998, Watson & Dallwitz 1992 onwards).
Various species of Dracaena have a high importance in horticulture (Singh & Dadlani 2000, Wijsman 2012), while others show medicinal (Staner & Boutique 1937) or social functions in marking graves, sacred sites and farm plots in many African societies (Sheridan 2008).Globally, Dracaena is in the top ten of most important crops in floriculture (Singh & Dadlani 2000), while in the Netherlands Dracaena is in the top five of the most exported pot plants with an annual turnover of c. 33 million euro (Wijsman 2012).It is remarkable that, even despite this importance, the taxonomy of various species is still unstable, while new species are being discovered on a regular basis (Mwachala et al. 2007, Mwachala & Fischer 2013), even in such well-studied places as the Canary Islands (Marrero et al. 1998).
Back in 2006, the first author started his investigations in Dracaena at Wageningen University, The Netherlands (WAG), by sorting out the enormous legacy left by the Dracaena specialist J.J. Bos (1939Bos ( -2003)).This legacy was composed of herbarium loans from all over the world, living plants in the greenhouse, pictures, literature and notes.The written data was digitized and a database compiled with all herbarium collections, before they were returned to their respective owners.The database, which is maintained by the first author, now contains data on more than 9 400 African specimens and on some 14 000 from elsewhere.Where possible, specimens were georeferenced and photographed.
During his Dracaena investigations, Bos had already recognised several new African species, but never came to publish these.He was also convinced that flower length was extremely variable and not a good character for species delimitation.In personal communication, Bos often referred to the type material of Dracaena mannii Baker which had led him to this conclusion: "The Berlin isotype of D. mannii for instance includes a considerably larger flower than any of its duplicates in other herbaria.Under these circumstances I fully endorse Hutchinsons view to treat them as one variable species" (Bos 1984).However, it turns out that the Berlin isotype (Mann 2329) consists of an inflorescence with smaller flowers while an envelope attached to the sheet contains a distinctly larger flower.With all other isotype sheets (at P, K and WAG) having flowers of the same much smaller size, we hypothesize that the single larger flower on the Berlin sheet does not belong to the same specimen/collection and was put there by accident.This unfortunate situation has thus led to a misconception of flower size as an unimportant character in Dracaena taxonomy!During our long experience with this genus, we have now learned that flower size is an important character, very useful for species recognition, and concluded that as a consequence Bos' species concept was often too wide.As a consequence, some names treated as synonym by Bos need to be reinstated as accepted species (De Roos 2014).For South Africa, Venter (1996) came to the same conclusion and reinstated D. transvaalensis Baker which Bos (1992) had treated as a synonym of D. aletriformis (Haw.)Bos.Such clarifications are also important for Dracaena plant breeders who use only a few clones (R. Scheffers, pers.comm.) and such reinstated 'new' species may point them to the potential of the introduction of a wider genetic variation to select from.
With Dracaena having been revised and treated for West Africa (Hepper 1968b, Bos 1984), southern Africa (Bos 1992), Ethiopia (Bos & Teketay 1997), tropical East Africa (Mwachala & Mbugua 2007) and the Zambesiaca region (La Croix 2010), the main gap in African Dracaena knowledge is in the Central African region.The first and second author are engaged in preparing the treatment for Flore du Gabon and Flore d'Afrique centrale.With Gabon harbouring most likely the most speciesrich rainforest in Africa (Sosef et al. 2017), it is not surprising that a relatively high number of taxonomic and nomenclatural novelties were discovered during the preparation of both Flora treatments.Most of these would seem out of place in such a treatment, and are thus published here.Since, apart from the species dealt with below, still others occur in the region covered by both Floras, no key is provided here.For identification, the interested user is referred to those Flora treatments which will appear soon.

MATERIALS AND METHODS
The study was performed using herbarium material and applying standard herbarium techniques (De Vogel 1987).The vast majority of this material came from BR, BRLU, K, LBV, MO, P and WAG (now at L), with additions from A, B, BISH, BM, BRUNOY, C, E, EA, FHO, GOET, HUJ, IFAN, IG, L, MA, NY, O, POZG, PRE, SRGH, UC, US and Z. Acronyms of institutes holding herbarium collections follow Thiers (continuously updated).
Types not available in one of these were generally studied using the Global Plants facility (JSTOR 2000(JSTOR -2016)).Data and illustrations were stored in a local version of BRAHMS 7 (Filer 2001).Most of the living material was studied in the botanic garden Burgers' Zoo, Arnhem, The Netherlands, which houses the biggest collection of Dracaena in the world, thanks to efforts of Bos, Jongkind, Damen and several other WAG collectors.Furthermore, the first author visited the Botanic Garden Meise, the Royal Botanic Gardens Kew, the Muséum national d'Histoire naturelle Paris, the Poznań Palm House, the Botanical Garden of the Adam Mickiewicz University, Poznań, the Botanika Bremen, the Alter Botanischer Garten Göttingen, the Jardín Botánico Canario Viera y Clavijo Gran Canaria, the Arboretum de Sibang Libreville and the Dracaena nursery De Plaats at Honselersdijk, The Netherlands, to study living Dracaena material.Furthermore, in 2013 and 2015, the first author conducted two field trips (together with E.L.A.N.Simons) to Gabon to collect herbarium specimens, living material and to study the plants in situ.
The QGIS program (Quantum GIS Development Team 2017) was used to prepare the distribution maps, together with a layer of the UNESCO Vegetation map of Africa (White 1983).
In Africa, Dracaena is generally easy to propagate; only a small piece of the plant is required to form a new plant, and even when thrown away it may easily root and thus occur as an escape (Damen pers.obs.).Therefore, in the field it is sometimes impossible to tell whether an individual is wild or escaped from cultivation, or was even originally planted but then abandoned.Hence, it is not impossible that in the distribution maps presented here a number of collections from cultivated or escaped plants are incorrectly included.
IUCN conservation status assessments for all Central African species, also those not discussed below, are being prepared and will be published in a separate paper.

RESULTS
The results are presented below per species/taxon, in alphabetical sequence.Distribution -Cameroon, Gabon, Republic of the Congo, Democratic Republic of the Congo, Angola.

Dracaena acaulis
Notes -The protologue of D. acaulis cites no specimen, but mentions 'Golungo Alto, regio 2 (1000 -2400 pedes) in sylvaticis editioribus et ad rivulorum latera montium de, Alta Queta.Floret Julio.' as a reference.The specimen Welwitsch 3752 matches both the description and the reference to its locality in the protologue.There are several duplicates, but the LISU specimen is the only specimen with labels carrying all locality data of the protologue, a label in the handwriting of J.G. Baker and a plant description very similar to that published in the protologue.Therefore, this specimen has been chosen as the lectotype for D. acaulis.It is known that Welwitsch provided specimens collected on different dates and localities with the same number (Albuquerque et al. 2009).The specimens in BM and COI carry the same date (July 1855) as the specimen in LISU and thus probably belong to the same collection and can be regarded as isolectotypes.The other duplicates in K, LISU and P, without dates, as well as those of Welwitsch 3753 (with date April 1856), are disregarded.
Dracaena rubroaurantiaca was described from a plant with the same inflorescence type and leaf shape as D. acaulis.In the protologue, De Wildeman compared D. rubroaurantiaca with Dracaena fragrans (L.) Ker Gawl only; he probably had not seen any material of D. acaulis when describing his new species.
In his unpublished PhD thesis, Mwachala (2005) selected Plate 58 published along with the protologue (De Wildeman 1906: 228) as lectotype for D. rubroaurantiaca.In this protologue, however, De Wildemann cited: "dans les environs de Malella, 20-juillet 1895 (Alf.Dewèvre)".In BR, two sheets of the specimen Dewèvre 169 were found in a type folder together with Plate 58.One of the two sheets is annotated as 'Dracaena rubro-aurantiaca.De Wild.n.sp.'.Although there is no location written on this sheet, it is cited in Durand & Durand (1909) with the location 'île du Congo dans les env.de Malela', and the morphology fits the protologue.Thus, although the plate published in the protologue is part of the original material, it is preferred to select a herbarium sheet as the lectotype.
Map 1 Distribution of Dracaena acaulis Baker.
Sansevieria longifolia Welw. is an unpublished name written on labels mounted on several duplicates of Welwitsch 3752 and Welwitsch 3753.
Distribution -Gabon, Republic of the Congo, Democratic Republic of the Congo.
Notes -In his unpublished PhD thesis, Mwachala (2005) selected a neotype for D. acutissima.However, in the protologue Hua wrote: "Région de l'Alima (Thollon, n° 913), terrains sableux, sous-bois, fl.mars 1888".Material of Thollon 913 is available in several herbaria and hence the selection of a neotype is superfluous.The specimen P00442404 was chosen as the lectotype from among the three sheets available at P, since that is the only sheet carrying a label with the details mentioned in the protologue.Distribution -Nigeria, Cameroon, Equatorial Guinea, Gabon, Republic of the Congo, Democratic Republic of the Congo.

Dracaena bicolor
Notes -In the protologue, Hooker (1861) described the inflorescence of D. bicolor s.str.as a dense spicate-capitate compound raceme.Hepper (1968b) stated it is a shortly cylindrical, very dense spike, while according to Bos (1984) the inflorescence is subcapitate to thyrsoid.The inflorescence of D. bicolor s.lat.has several (4-10) dense clusters of c. 20 flowers forming a compact bracteate subcapitate inflorescence of up to 26 cm long.Further anatomic study might reveal that it has a thyrsoid structure, similar to that in Sansevieria (Budweg 2016).
Dracaena talbotii was described from a plant with the same inflorescence type as in D. bicolor, but with narrow leaves.We have observed leaf shape to be highly variable within D. bicolor (for example the collection Bos 4160 represents two distinct leaf forms collected from the same plant before and after it fell down and re-sprouted), reason why we do not recognise this form as a distinct taxon.Dracaena densifolia represents a form of D. bicolor with a more open inflorescence, but otherwise, including flower size, fits well within the range of D. bicolor.Distribution -Cameroon, Equatorial Guinea, Gabon, Republic of the Congo.

Dracaena braunii
Notes -Dracaena litoralis was recently described to accommodate a plant similar to the original material of D. braunii (type and plate in protologue of cultivated plant) but with smaller flowers and different leaf shape.However, we think the difference in the flower size is explained because D. braunii was based on cultivated material in which the flower size is often a little larger (phenotypic plasticity due to better growing conditions?).For example, a flower measured in Gabon (Damen et al. 499) was 18 mm long, but when this plant was transplanted and flowered in the greenhouse at WAG, the flower was 22 mm long.Flowers in D. braunii are nocturnal and only a few hours at full anthesis (Damen, pers.obs.).In the field, collections are usually made when the flowers are not yet at full anthesis and therefore still slightly smaller.In cultivation, where often the light levels are lower, full anthesis starts earlier during the evening and it is easier to collect the flowers at full anthesis.These phenomena could thus well explain the difference in flower size  observed by Mwachala & Fischer (2013).Moreover, we have been able to study D. braunii populations in the field, which occurred to be very polymorphic, possibly due to the extreme growing conditions, (salty places near the coast often grazed by buffalos).From those observations we conclude that leaf shape and phyllotaxis of D. litoralis fit comfortably within the range of D. braunii, as does the characteristic ecology.
In his unpublished PhD thesis, Mwachala (2005) selected Plate 20 in the protologue (Engler 1892: 479) as a lectotype for D. braunii.In that protologue, however, Engler cited a cultivated plant: "Die Pflanze wurde von J. Braun nach dem botan.Garten zu Berlin importiert und blühte daselbst im Aug. 1891." According to the label attached to the herbarium sheet in B with barcode B_10_0184054, it represents material from that plant and should thus be regarded as the original material to be preferred over a plate.This corrected typification was already communicated by Mwachala & Fischer (2013).Etymology.Dracaena bushii honours J.J. Bos (the Dutch word 'bos' translates to 'bush' in English) who has been a great inspiration of the first author, and at the same time credits Burgers Bush in Arnhem, The Netherlands, where the largest collection of living Dracaena in the world is being maintained, including a few specimens of D. bushii.

Dracaena bushii
Shrub to 2.5 m tall, with a single stem arising from a woody rhizome; stem to c. 5 cm diam, up to 50 cm long, yellowish grey, marked by dark green annular scars of fallen leaves, occasionally forming aerial roots.Leaves alternate, equitant; pseudopetiole bluish dark green, longer than the lamina, to 80 cm long, sulcate along its upper surface, gradually extending into a sheathing base, clasping the stem for distinctly more than its circumference; lamina ovate, up to 65 by 20 cm, corrugated lengthwise, shiny dark green above, dull pale green beneath, midrib paler, variegated with scattered transversely orientated oval pale green dots, each dot with a yellow centre, coriaceous, leaf base rounded, leaf tip acuminate, mucro to 2 cm long.Inflorescence terminal, usually erect; peduncle up to 8 cm long, smooth, terminating in a bracteate capitulum of up to 13 cm diam; bracts enveloping the capitulum, broadly ovate, c. 5 by 3 cm, scarious, dark brown; flowers in fascicles of 2 or 3, each fascicle closely enveloped by 3.5-4 cm long light brown bracts.Flower: pedicel 0 -1 mm long; perianth white, 45 -50 mm long, lobes 15 by 2 mm, spreading, each with a single central distinct costa; stamens inserted near the throat, up to 2 mm shorter than the lobes, filaments white, terete, inflated, with subulate tips, anthers pale yellow, 3 by 1 mm; ovary oblong, up to 3 by 2 mm; style filiform, to 1 mm exserted, white, stigma capitate, shallowly trilobed, c. 1 mm diam.Fruits depressed globose, shallowly 1-3-lobed, c. 14 by 9 mm, bright orange to bright red, with 2 or 3 seeds, receptacle of up to 4 mm long.Seeds smooth, ellipsoid, up to c. 10 by 7 mm, flattened where pressed against adjacent seeds.
Ecology -Understory of secondary forest, forest edges, humid places, on sandy soil and clayish river banks; at 100-1500 m altitude.Flowering in June (fide Letouzey 13823).Note -The description of Dracaena sp.aff.phrynoides Hook in Cheek et al. (2004) matches the new species D. bushii.Unfortunately, the herbarium at Kew could not locate the three specimens cited in this publication during a recent visit of the first author, and so this could not be checked.Distribution -Guinea, Sierra Leone, Liberia, Ivory Coast, Ghana, Nigeria, Cameroon, Equatorial Guinea, Gabon, Central African Republic, Republic of the Congo, Democratic Republic of the Congo, Tanzania, Angola, Zambia.

Dracaena camerooniana
Notes -Dracaena oddonii was described from a plant with larger leaves and a longer petiole than the typical D. camerooniana.We do not recognise this form as a distinct taxon, because leaf shape within D. camerooniana is highly variable and both leaf forms can occur even on the same plant (Damen, pers.obs.).
Dracaena silvatica was described from a plant with much smaller flowers than typical D. camerooniana.This ecotype has been observed several times in the wild and grows in deep shade and inundated forest where the flowers and fruits are often poorly developed or even malformed (Damen, pers.obs.).It is regarded as a phenotype not worthy of a distinct taxonomic rank.

Dracaena cerasifera Hua -Map 7
Dracaena cerasifera Hua ( 1897 Distribution -Liberia, Ivory Coast, Ghana, Cameroon, Equatorial Guinea, Gabon, Central African Republic, Republic of the Congo. Note -In the protologue, Hua gives 'Ogooué (Mgr Leroy)' to indicate the material seen.This material is represented at P by three sheets, two with flowers and one with fruits, which match the description provided in the protologue.The sheet chosen here to serve as lectotype is richest in flowers.Distribution -Nigeria, Cameroon, Equatorial Guinea, Gabon, Democratic Republic of the Congo.

Dracaena congoensis
Note -The protologue describes a plant exhibited in 1872 by Robert Bullen of the Glasgow Botanic Gardens.No herbarium material of the original plant seems to have been preserved.Although we could not find direct proof of the statement by Bos (1984) that William Bull received plants of D. goldieana from Edinburg Botanic Garden, it is a highly likely assumption that William Bull received material from this same plant and published the drawing carrying that name in his Retail List (Bull 1877).This drawing thus represents the material closest to the original cultivated plant and we decided to select that as the neotype.Etymology.Bos did already recognise this species as distinct and provisionally named it D. haemanthoides on his identification slips (fide Bos 10828), probably because it resembles the growth habit and capitate flowers of some species of Haemanthus.

Dracaena haemanthoides
Shrub, single-stemmed or occasionally acaulescent, up to c. 50 cm high; stem unbranched, erect, up to c. 20 cm long, densely leafy towards the top, greenish.Leaves alternate, equitant; pseudopetiole broadly winged, gradually extending into a sheathing base, clasping the stem for slightly more than its circumference; lamina oblanceolate or narrowly elliptic in smaller specimens, up to 80 by 12 cm, thickly coriaceous, glossy dark green above, pale green or occasionally dark purple with midrib prominent for c. 3/4 of its length beneath, base gradually tapering, apex acuminate, subulate, mucro up to 3 mm long.Inflorescence terminal, erect, up to 18 cm long; peduncle smooth, purplish green, base with cuspidate transitional leaves, merging into purplish green cymbiform bracts of up to 16 by 10 mm which subtend the closely packed glomerules containing well over 10 flowers each, forming a subcapitate inflorescence.Flower: pedicel 1-2 mm long, persistent basal part to 1.5 mm long; perianth creamy white, 51-59 mm long, tube longitudinally striped, pink-purple on the outside, lobes reflexed, c. 11 by 3 mm, apex obtuse with minute mucro, outside greenish white, inside white; stamens inserted near the throat, up to 1 mm shorter than the lobes, filaments filiform, white, anthers c. 1.5 mm long, pale yellow; ovary ovoid, 3 by   Notes -Dracaena bequaertii was described from a lianescent plant with a shorter inflorescence than the typical D. laxissima, but otherwise fits that species' morphology.From studying a large number of herbarium collections, it appears that the inflorescence length in D. laxissima can vary between 5 and 50 cm, and hence we do not recognise D. bequaertii as a distinct taxon.
Dracaena thomensis was encountered as a name written on the identification slip mounted on the specimen Quintas 1, deposited in K (K000255924).The name does not seem to have been formally published.Dracaena tessmannii was described from a plant with the same inflorescence and leaf shape as D. mannii.In the protologue, Engler & Krause (1910) compared D. tessmannii with D. fragrans only.We assume they probably did not see any material of D. mannii when describing this new species.

Dracaena laxissima
All four syntypes of D. usambarensis var.longifolia at BR are ± similar in the sense that they are composed only of leaves mounted on a sheet.They were collected at different locations on different dates.The lectotype chosen represents the only specimen with a plant description written on the label.This description is most likely used as the basis for the description in the protologue.Dracaena marina is similar to the East African Dracaena usambarensis Engl.; it is easily distinguished from that species by the oblanceolate to narrowly oblanceolate leaves, its bright red to yellow inflorescence axes, and the perianth lobes that are longer than the tube.In D. usambarensis the leaves are narrow elliptic, the inflorescence axes are green to orange and the perianth lobes are shorter than the tube.-Type: Breteler 14640 (holo WAG (WAG0047304); iso LBV (LBV0008874), WAG (WAG0047305), WAG-spirit (WAG0029799)), Gabon, Gamba region.c.S2°40' E10°00', 11 Nov. 1998.
Etymology.Bos already recognised this species as distinct and had provisionally named it D. marina, because it is often found close to the ocean.With this name, Bos also wanted to honour Ms Marina Wassink, the former secretary of the Plant Taxonomy Department at Wageningen University, The Netherlands, who had been a great help to him.
Shrub or large tree of up to 40 m high; trunk up to 60 cm diam, forming stilt roots when growing in inundated areas or mangroves; bark smooth to shallowly longitudinally fissured, grey, slash creamy white, not exuding resin; leaves densely set along the branches, tufted at the apex in older specimens, stem visible between the leaves.Leaves spirally arranged, sessile; lamina variable, oblanceolate and up to 30 by 9 cm, but on fast growing shoots often narrowly oblanceolate and up to 60 by 8 cm, coriaceous, dark dull green above, pale green with prominent midrib below, secondary veins inconspicuous, base gradually tapering, then abruptly expanding and clasping the stem for 1/2 to 3/4 of its circumference, apex acute to acuminate, mucro c. 3 mm long.Inflorescence terminal, erect, a much branched panicle, up to 75 by 55 cm; peduncle smooth, flexible (like rubber), bright yellow to bright red; bracts supporting the branches triangular, up to 10 by 7 mm (distally decreasing in size), early caducous, coriaceous; flowers solitary or arranged in fascicles of up to c. 5, each solitary flower or fascicle subtended by a small triangular, membranous, caducous bract with scarious margin, up to 2 mm long.Flower: pedicel bright orange, to 15 mm long, persistent basal part to 13 mm long, surrounded by early caducous scarious floral bracts of c. 1 mm long; perianth creamy white, 35-55 mm long, lobes 24-33 mm long; stamens inserted near the throat, up to 3 mm shorter than the lobes, filaments filiform, white, anthers c. 2.5 mm long, pale yellow; ovary ovoid, 3 by 2.5 mm; style filiform, exserted for up to 3 mm, white, stigma c. 0.5 mm diam.Fruits globose to depressed globose, up to 23 by 28 mm, glossy, orange-brown to red, usually with 1 seed, receptacle swollen, up to 6 by 3  Notes -Although we have not been able to trace Stanfield FHI47067, Keay's description of an unidentified arborescent Dracaena based on this specimen leaves little doubt that it is conspecific with D. marina.It also matches the location of other specimens of D. marina.Dracaena marina can be confused with D. mannii since it keys out as such in Bos (1984Bos ( , 1992)), Mwachala & Mbugua (2007) and La Croix (2010).It is easily distinguished from D. mannii by its flowers being less than half as long as those of D. mannii.Distribution -Cameroon, Equatorial Guinea, Gabon, Central African Republic, Republic of the Congo, Democratic Republic of the Congo, Angola.

Dracaena nitens
Notes -Dracaena mannii, D. perrottetii and D. usambarensis are very similar to D. nitens and according to Bos (1984) even represent synonyms of the same species.However, D. nitens differs from the other species in having flowers 23-37 mm long and linear to narrowly elliptic leaves of up to only 1.5 cm wide, and is reinstated here.
The protologue of D. nitens does not cite any specimens, but it should be a plant collected by Welwitsch in Golungo Alto or Pungo Andongo, bearing flowers in September.The Welwitsch 3743 specimen deposited in BM carries a plant description in the hand of Baker.This description is obviously used as the basis for the description in the protologue of D. nitens.Also, on this sheet we find the annotation 'Lectotype' written by Bos.Therefore, we here follow Bos's suggestion and have chosen this specimen as the lectotype for D. nitens.Although there are several other Welwitsch collections (at COI, G, K, LISU, P and WAG) bearing the number 3743, it is known that Welwitsch regularly gave plants collected on different dates and localities the same number (Albuquerque et al. 2009).Since there is no indication that the other sheets represent the same collection (actually, some bear a different date) these are disregarded as isolectotype material.
Both syntypes of D. reflexa var.buchneri could not be located at B, and are likely to have been destroyed.No duplicates could be found elsewhere, but it is still possible one may show up in, for example, LE or UC.Also, in the absence of any suitable material to serve as a neotype, we have for the moment refrained from designating one.The analysis in the protologue and the location fit within the range of D. nitens, which is why we consider this variety to belong to that species.
Notes -Dracaena nitens, D. mannii, and D. usambarensis are very similar to D. perrottetii and according to Bos (1984) are even synonyms of the same species.However, D. perrottetii differs from the other species in having subfalcate leaves usually wider than 1.5 cm, and is reinstated here.Dracaena perrottetii var.minor was described from a plant with smaller leaves and a simple inflorescence.Studied material shows that leaves in D. perrottetii are polymorphic and that both branched and unbranched inflorescences occur even on the same plant.Therefore we do not recognise such a form as a distinct taxon.
Pleomele heudelotii was described based on the same material as D. perrottetii var.minor.Notes -Dracaena sanderiana was first exhibited by Hort.Sander in Earl's Court (1892), and published the same year with a description, but without illustration.Original material of the plant exhibited has not been traced and has probably not been conserved.One year later, D. sanderiana was exhibited by Hort.Sander in Ghent and an illustration was published in Gard.Chron., ser. 3, vol. 13 (1893).This illustration most likely represents the same plant as originally exhibited in 1892 and is hence chosen here as the neotype.
Dracaena sanderiana is an important indoor ornamental plant (Morsy & Elshahawy 2016) also sold under the name D. braunii and 'Lucky Bamboo' (Aslam et al. 2013).According to The World Checklist of Asparagaceae (Govaerts et al. 2017), D. sanderiana is a synonym of D. braunii, but unfortunately no reference is given to the origin of this information.We have not found any revision or flora that underlines this statement, although Baker (1898) stated that D. sanderiana is similar to D. braunii, he surely did not regard them as conspecific.
The flowers of D. sanderiana are 5 times longer than those of D. braunii, while the leaf base is not congested as in D. braunii.
Here we treat them as distinct taxa.
The inflorescence of D. poggei fits well within the range of D. sanderiana and the type material represents nothing more than a non-variegated form of D. sanderiana.Reason why we do not recognise this form as a distinct taxon.Luckily, the publication of D. sanderiana predates that of D. poggei only by a few days, so that the first can be maintained for this important ornamental.
Dracaena vanderystii was described from a plant with comparatively narrow leaves.The inflorescence fits well within the range of D. sanderiana and observations in herbaria, in the field and on cultivated material all show that leaf shape within Dracaena is highly variable.Therefore, we do not recognise this form as a distinct taxon.In his unpublished PhD thesis, Mwachala (2005) designated the same lectotype for D. vanderystii, but stated it was to be found in P.This is probably a typo which is corrected here.Distribution -Democratic Republic of the Congo, Rwanda, Burundi, Sudan, Ethiopia, Uganda, Kenya, Tanzania, Zambia, Malawi, Mozambique, Zimbabwe.
Note -The only material cited in the protologue is Steudner 477 from Dschibba, Gondar.Both Bos & Teketay (1997) and Mwachala & Mbugua (2007) cited the holotype as being destroyed in B. However, we found this specimen filed under Dracaena deisteliana Engl. in the herbarium in Berlin.The protologue of D. tholloniana cites Thollon 31 as the type and states 'Ndjolé, sur le haut Ogooué, sous-bois, fl.Janvier 1895'.However, at P, Thollon 31 represents an Asteraceae (Eclipta prostrata (L.) L.) with a different date and location.Both Bos (1992) and Mwachala & Mbugua (2007) cite Thollon 91 as the holotype for D. tholloniana but give no reason for this change in number.They probably, as we, assume it was a printing error.
There are two duplicates of Thollon 91 at P (P00442286, P00442285), the specimen with barcode P00442286 matches the locality data and collecting date mentioned in the protologue and is annotated 'Dracaena tholloniana Hua' in Hua's handwriting.
Dracaena longipetiolata was described from a fruiting specimen with a long petiole.However, such long petioles are also found in specimens otherwise perfectly fitting within D. tholloniana, while intermediates also exist.Furthermore, the infructescence also fits within the range of D. tholloniana.Therefore, we do not recognise this form as a distinct taxon.
According to Bos (1984), Dracaena monostachya var.angolensis is synonym of D. aubryana s.lat.It was described from a plant with a short peduncle and leaves with a rounded base and long petiole.These characters fit those of D. tholloniana, as does the lectotype chosen earlier by Bos.Distribution -Democratic Republic of the Congo, Burundi, Kenya, Tanzania, Zambia, Malawi, Mozambique, Zimbabwe, South Africa.

Dracaena usambarensis
Notes -Dracaena nitens, D. mannii, and D. perrottetii are very similar to D. usambarensis and according to Bos (1984) are synonyms of the same species.However, D. usambarensis is easily distinguished from the others by its large flowers of 40-55 mm long and is reinstated here.
Three syntypes of D. usambarensis are mentioned in the protologue: one collected at Quilimane (probably a Stuhlmann collection), Volkens 65 and Volkens 1938.We could not find this material in B and assume it was destroyed during the WWII bombing of the Berlin-Dahlem herbarium.The only material we could trace was found in K and BR and represent photos of the herbarium sheets Volkens 65 and Volkens 1938.We agree with Mildbraed (Mildbraed & Perkins 1910) that Volkens 1938 belongs to D. afromontana Mildbr.and should be excluded.The photos of Volkens 65 match the description in the protologue and has been of great help in fixing the application of the name D. usambarensis.The photo in K is bigger and clearer than that in BR.Therefore, we have chosen the photo of Volkens 65 in K as the neotype of D. usambarensis.
Dracaena brachythyrsa is an unpublished name written on the label mounted on Peter 13906 deposited in B (B_10_0184055) and WAG (WAG.1154506).
The protologue of D. gazensis cites two syntypes with different locality data and collecting date, but with one collection number Swynnerton 80 and so a lectotype has to be assigned.Bos recognised this and annotated the BM specimen BM000911614 as lectotype but never formalized his choice.We agree with Bos that the BM specimen with both leaves and an inflorescence is most suited to serve as the lectotype.The other duplicate in BM, with different date, and the specimens in Z and K that carry a different number, Swynnerton 80a, are disregarded.We also disregard the specimen Swynnerton 80b from SRGH mentioned in Flora Zambesiaca as the holotype for D. gazensis (La Croix 2010), although we did not see material of this specimen.
Dracaena gazensis was described from a plant with the same inflorescence type, flower length and leaf shape as D. usambarensis.In the protologue, Rendle only compared D. gazensis with D. deremensis Engl.; he probably had not seen any material of D. usambarensis when describing his new species.
The protologue of D. pseudoreflexa cites two syntypes, Mildbraed 2813 and Mildbraed 2175.Both specimens are probably lost at B during WWII and we could only trace photos of Mildbraed 2813 in K and BR.The drawing in the protologue is the only original material left and has enough detail to fix the application of this name; therefore we have chosen this drawing as a lectotype for D. pseudoreflexa.Mwachala & Mbugua (2007) cites a specimen Mildbraed 2813 deposited in B as the holotype of D. pseudoreflexa, hence ignoring the syntype Mildbraed 2175.We have not been able to trace Mildbraed 2813 in B.
Dracaena pseudoreflexa was described from a plant with smaller leaves than the typical D. usambarensis.Leaf shape within D. usambarensis is highly variable and intermediates also exist (Damen, pers.obs.).We do not recognise this form as a distinct taxon.Mildbraed also mentioned that the tube length equals the lobes in the flower of D. usambarensis compared to the lobes being shorter than the tube in D. pseudoreflexa.The protologue of D. usambarensis has no information about the length of the lobes and we only found specimens with lobes shorter than the tube.Distribution -Ghana, Nigeria, Cameroon, Equatorial Guinea, Gabon, Republic of the Congo, Democratic Republic of the Congo, Angola (Cabinda).
Notes -Both D. letestui and D. mildbraedii were described from plants with long linear leaves.From observations in the herbarium and during fieldwork, we have observed leaves of D. viridiflora to be extremely variable even on the same plant.Therefore, we do not recognise these forms as distinct taxa.
The type specimen of D. ledermannii, Ledermann 1483, the only specimen cited in the protologue, was probably lost at B during WWII.However, B also holds the specimen Ledermann 6119, which is annotated by Krause (S.Bollendorff & P. Hiepko, pers.comm.) as D. ledermannii and fits the description provided in the protologue.We have identified it as belonging to D. viridiflora and chosen that specimen as the neotype for D. ledermannii.
Subshrub of up to c. 70 cm high, single-stemmed; stem usually unbranched, prostrate, to c. 5 cm high.Leaves up to 4, in a rosette; pseudopetiole erect, slender, up to 52 cm long, terete, dark green, margin involute, abruptly extended into a short c. 1 cm long dark green to violet sheathing base, clasping the stem for more than its circumference; lamina ovate to broadly ovate, to 23 by 11 cm, coriaceous, plicate, above glossy and uniformly dark green, below paler green, both surfaces variegated with transversely orientated greenish yellow to white round to oval dots, midrib inconspicuous, with up to c. 8 parallel nerves impressed above and prominent below, remaining nerves and venation inconspicuous, but conspicuous in herbarium specimens, base rounded, leaf tip declinate, broadly cuspidate, mucro c. 4 mm long.Inflorescence terminal, erect; peduncle smooth, c. 7 cm long; raceme c. 3 cm long; bracts up to 6, early caducous, broadly triangular to cymbiform, 10 by 8 mm, with aristate tip of up to 6 mm long, scarious, distally decreasing in size, sheathing at the base of the inflorescence to amplexicaul at its apex.Flower: pedicel up to 5.5 mm long, articulated near the apex; old dried flower 20 mm long, colour unknown, tube 6 mm long; filaments and anthers not seen; ovary ovoid, up to 4 by 2 mm, style not seen.Fruits depressed globose, up to 15 by 20 mm, shallowly 1-3-lobed, bright orange-red.Seeds ellipsoid, c. 7.5 by 5 mm, with a smooth testa.
Distribution -Endemic to Gabon, only known from the Ngounié province.
Ecology -Growing in old secondary forest and on relatively dry forested slope near waterfall; at 700 -950 m altitude.Occasionally forming a population covering several square meters.Etymology.Dracaena waltersiae honours Mrs Gretchen Walters who has collected the type and brought the plant to my attention.With her work, notably that on the Plateaux Batéké, she has made a substantial contribution to our knowledge of the flora of Gabon.

Map 2
Distribution of Dracaena acutissima Hua.Map 3 Distribution of Dracaena bicolor Hook.Map 4 Distribution of Dracaena braunii Engl.

Map 17
Distribution of Dracaena sanderiana Sander ex Mast.Map 18 Distribution of Dracaena steudneri Engl.Map 19 Distribution of Dracaena tholloniana Hua.
Dracaena bushii is similar to Dracaena phrynioides Hook., both having capitate inflorescences and leaves with long pseudopetioles.Dracaena bushii is easily distinguished from that species by its flowers which are twice as long and the presence of a distinct stem up to 50 cm long, and an equitant (with overlapping leaf bases) leaf arrangement, where D. phrynioides is subacaulous with its leaves arranged in a rosette.