A synopsis of the African genus Whitfieldia (Acanthaceae: Whitfieldieae) and a key to the species

We present a synopsis and an identification key for the 13 species and one subspecies of the African genus Whitfieldia Hook. (Acanthaceae: Acanthoideae: Whitfieldieae). This work is based on the observation and analysis of morphological data from herbarium specimens and photographs of Whitfieldia species, and the study of all the relevant literature on the genus. Whitfieldia latiflos C.B.Clarke ex Stapf from Liberia is reinstated as an accepted species and a full description and illustration are provided. A new subspecies, W. colorata C.B.Clarke ex Stapf subsp. tigrina A.Grall & I.Darbysh., also from Liberia, is described. In addition, W. arnoldiana De Wild. & T.Durand and W. letestui Benoist are treated as synonyms of W. laurentii (Lindau) C.B.Clarke and W. preussii (Lindau) C.B.Clarke respectively for the first time. Eleven names in Whitfieldia are lectotypified. Reference specimens, relevant literature, habitat notes, distribution maps, taxonomic notes and a preliminary assessment of the conservation status and extinction risk are provided for each taxon. Four species are considered to be threatened of which two (W. latiflos and W. rutilans Heine) are assessed as Endangered (EN) while two others (W. purpurata (Benoist) Heine and W. thollonii (Baill.) Benoist) are placed in the Vulnerable (VU) category. In addition, two taxa are assessed as Near Threatened (NT) and one is currently considered to be Data Deficient (DD).


Introduction
Wild plants belonging to the genus Whitfieldia Hook. (Hooker 1845: t.4155) (Acanthaceae: Acanthoideae: Whitfieldieae) grow in the forests of tropical continental Africa from Guinea in the west to Ethiopia in the east, and south to Angola, Zambia and Mozambique, with highest diversity in the Guineo-Congolian phytogeographical region (Vollesen 2008). The genus is perhaps best known for the striking large whiteflowered species W. elongata (P.Beauv.) De Wild. & T.Durand ("white candles"), which is widely cultivated and has a range of uses from medicine to extraction of dye (Burkill 1985), but many of the species are highly attractive and have horticultural potential. Current literature and taxonomic databases record 12 (Vollesen 2008(Vollesen & 2013Burrows et al. 2018) to 14 accepted species of Whitfieldia (African Plants Database 2020; POWO 2020).

The tribe Whitfieldieae
Based on combined morphological analyses and DNA sequence data, Manktelow et al. (2001) recognised the tribe Whitfieldieae of the Acanthaceae to accommodate Whitfieldia together with two Afro-Malagasy genera, Chlamydacanthus Lindau and Lankesteria Lindl. Subsequent molecular phylogenetic studies have added the Malagasy endemic genera Camarotea Scott Elliot, Forcipella Baill., Leandriella Benoist and Zygoruellia Baill. to this tribe (McDade et al. 2008;Tripp et al. 2013). McDade et al. (2008) recognised a core Whitfieldieae clade, comprising Whitfieldia, Chlamydacanthus, Camarotea, Forcipella and Leandriella, which are monophyletic with strong support. They found Lankesteria to be only weakly supported as a basal member of Whitfieldieae s.l. Manktelow et al. (2001) proposed several synapomophies for Whitfieldieae s.l., notably (1) presence of concentric rings of ridges protruding from the testa of the seeds, and (2) pollen with a densely granular circular area surrounding the pores. Most members of core Whitfieldieae (where known) have biporate pollen grains (Manktelow et al. 2001;McDade et al. 2008), circular in apertural view and with a pronounced marginal girdle, which is very similar to the gürtelpollen of Justicieae: Isoglossinae (Kiel et al. 2006;Darbyshire 2009), although this similarity is homoplastic (McDade et al. 2008). Whitfieldia elongata is the only species of that genus with pantoporate pollen (Manktelow et al. 2001) which was otherwise reported in only one other species of Acanthaceae: Megaskepasma erythrochlamys by Prieu et al. (2017). However, Megaskepasma's pollen has equatorial apertures only (Kiel et al. 2018) and several other genera among Acanthaceae, such as Louteridium and Brunoniella, do also have pantoporate pollen as defined by Prieu et al. 2017 (Tripp et al. 2013). Other useful traits for the identification of core Whitfieldieae include left-contort corolla aestivation (Manktelow et al. 2001 also record imbricate aestivation in Whitfieldieae but this has not so far been observed by the current authors); an androecium of four stamens or very rarely two stamens and two staminodes with no filament curtain for all species (Manktelow et al. 2001;Tripp et al. 2013); and each floral or cymose unit being subtended by paired bracts (or bracteoles). These bracts are usually conspicuous and clasping or partially connate and often partially or entirely enclose the calyces. Based on these morphological traits, a further Malagasy genus -Vindasia Benoistis here included within Whitfieldieae for the first time. Core Whitfieldieae therefore contains 7 genera and ± 26 species, with a further 7 species in Lankesteria. Generic delimitation is problematic within the Whitfieldieae, the endemic Malagasy genera being particularly poorly differentiated morphologically (I. Darbyshire, pers. obs.).
Within the wider classification of Acanthaceae, McDade et al. (2008) resolved Whitfieldieae as sister to (Andrographideae + Barlerieae) within their "BAWN" clade (i.e. the three tribes Barlerieae, Andrographideae and Whitfieldieae, plus the genus Neuracanthus Nees), but noted that this relationship is only weakly supported. This part of the Acanthaceae is in urgent need of phylogenetic studies to resolve relationships within and between the four major lineages (McDade et al. 2008;Darbyshire et al. 2019).
The genus Whitfieldia Hook. Whitfieldia itself is readily identified by having the combination of well-developed racemes, spikes or panicles; each cymose unit being pedunculate and single-flowered; and the calyces being large and showy, exceeding 9 mm in length. The calyces usually extend beyond or are subequal in length to the bracteoles and are often of a similar colour tone to the corolla, although they can also be green. Chlamydacanthus can also have racemose inflorescences but in that genus the cymose units are often 2 -3-flowered and the calyces are small (up to 8 mm long but usually shorter) and hidden within the bracteoles and are not similar in colour to the corolla. The other genera of core Whitfieldieae have fasciculate, glomerulate or shortly umbellate inflorescences.
On a molecular level, studies by Mankletow et al. (2001) and McDade et al. (2008) included only two species of Whitfieldia (W. elongata and W. "stuhlmannii", the latter almost certainly referring to W. orientalis Vollesen considering the geographic location [east of the Usambara Mountains, west of Kwani Forest reserve] of the specimen sampled, Iversen et al. 87520,n.v.). Both of the species sequenced were of East African origin. Further molecular work, with the addition of species from west and central Africa (where, based on the current work, c. 70% of the species are found) is needed to test the monophyly of the genus and biogeographical hypotheses regarding its diversification in Africa. This work is currently underway at Kew, using targeted sequencing of nuclear genes from herbarium specimens, as detailed in Brewer et al. (2019).
Aims of the present study Several regional taxonomic treatments of Whitfieldia are available, including for central Africa (Heine 1966;, eastern Africa (Vollesen 2004(Vollesen , 2008 and southern tropical Africa (Vollesen 2013). However, there is no modern global treatment of the genus. In particular, western African species occurring from Guinea to Cameroon are missing from all these recent regional studies and their identification remains problematic. Apart from W. elongata which can tolerate disturbed habitats, most species of Whitfieldia can be indicators of good quality forests (I. Darbyshire, pers. obs.), and several species are range-restricted, hence they are of significant conservation concern. To help identify and protect the species that are most at risk, a comprehensive taxonomic treatment, including identification tools and information on the distribution and ecology is needed.
So the present work aims to provide an identification key to species within the genus and a synoptic account of the recognised taxa, including distribution maps, notes on habitat requirements, citation of voucher specimens and, where available, illustrations to help in the identification of this fascinating group.

Material studied and cited
This study is based primarily on the examination of herbarium specimens held at BM, BR, K, P and PRE, together with additional material digitised at BRLU, G, LISC, MO and WAG, now transferred to L (herbarium codes follow Thiers, continuously updated). For the more common species (Whitfieldia elongata, W. lateritia Hook., W. laurentii (Lindau) C.B.Clarke), we cite a selection of specimens for each country where the species is found. For all other species, all specimens seen are cited. For most collections, at least one duplicate has been seen by the authors, indicated by an exclamation mark "!". When only a photograph has been seen, the herbarium code is followed by an asterisk "*". When only a field photograph of the collection has been seen by the authors, the collector and collection number are followed by two asterisks "**". Unseen specimens are only included where the authors have reasonable confidence that the identification is correct. An indication of the phenology ("fl." or "fr.", respectively for flowering or fruiting) has been added before the collection date to aid future phenological studies or targeted collection efforts.
Photographs of species of Whitfieldia in the field, most of them communicated by authors which are listed in the acknowledgements section, but also from  and Burrows et al. (2018), were studied where available to assist in the comparison of floral morphology and colour.

Measurements and terms employed
Measurements were made from dry material except for floral dissections, which were made following soaking in Aerosol OT 10% solution. The morphological terms used to describe the species follow Beentje (2016). The palynological terminology follows Manktelow et al. (2001) and McDade et al. (2008).

Typification
Wherever required and possible, accepted names and synonyms are lectotypified from amongst the original material.

Distribution Mapping
For most species, geographic coordinates were determined from specimen labels. Localities have been georeferenced wherever possible, primarily using online and published gazetteers (notably Fuzzy Gazetteer: http://dma.jrc.it/services/gazetteer/) and Google Earth™.
In some cases where species are more widely distributed in Africa and the number of specimens is too large to be manually georeferenced (Whitfieldia lateritia, W. elongata), data were first imported into GeoCAT (www.geocat.kew.org;Bachman et al. 2011) from GBIF (https://www.gbif.org) and herbarium specimens were then consulted to add missing points and allow a calculation of global extent of occurrence (EOO). In such cases, potentially anomalous points from GBIF were checked before acceptance. For most species, all coordinates were imported into SimpleMappr (Shorthouse 2010) to produce the distribution maps. For W. elongata, a polygon was drawn from the GBIF data and imported into QGIS version 2.14.4 (QGIS Team 2020).

Species Conservation (Extinction Risk) Assessments
The provisional conservation assessments presented here follow the categories and criteria of IUCN (2012). The term 'location' defines a geographically or ecologically distinct area in which a single threat-ening event can rapidly affect all individuals of the taxon present (IUCN 2012). Extent of Occurrence (EOO) was calculated using the GeoCAT tool. Area of Occupancy (AOO) is not used except in specific circumstances where we are confident that we have a good understanding of the number and distribution of localities for a species.
Type species: Stylarthropus brazzae Baill. Leiophaca Lindau (1911: 307 its local name in digo (Kenya, Tanzania) "Mkula usiku" translated as "grow by night" (Pakia 2006). The pollen of this species is spheroidal with 6 -9 scattered pores (Manktelow et al. 2001) and spinules on its surface. This is an unusual pollen type for this genus as all other species have a flattened, biporate, smooth pollen type Mankletow et al. 2001). While all other Whitfieldia species are said to be bee-pollinated (Mankletow et al. 2001;D. Notton, pers. comm.), W. elongata has a white long tubular corolla with exserted stamens typically associated with moth-pollination (sphingophily) (Faegri & van der Pijl 2013). The unusual pantoporate pollen structure could potentially be the result of the co-evolution of this species with a moth pollinator. However, this still needs to be tested.
(2) When Anderson (1863: 27)  (3) The protologue of Ruellia elongata P.Beauv. (De Beauvois 1804: 46) describes flowers of a 'beautiful blue sky colour' as depicted on the plate associated with the description of the species (De Beauvois 1804: tab. XXVI). However, not only should the flowers in this case have been described as and painted white, but they also should show a slight bend at the apex of the corolla tube and four longer and fully exserted stamens on the plate (instead of straight corollas with two short stamens). Such inaccuracies in the description and illustration could be justified by the poor state of the original collection (Palisot de Beauvois s.n.). However, faulty colouring of some of the plates in De Beauvois's Flore d'Oware et de Benin has first been reported in Marshall (1951). Later, Brenan (1952) described several of those plates as 'simply products of the imagination and sadly misleading to the botanist' (p. 194). (2004: 123; 2008: 267; 2013: 173); Burrows et al. (2018: 882, 889 can also be found in a more fragmented habitat. Detailed notes on the habitat, ecology and phytogeography of this species are recorded by Vollesen (2004). CONSERVATION STATUS. This species occurs in forest patches in coastal lowland East Africa. It has an EOO of 194,731 km 2 and an AOO of 116 km 2 , based upon using the standard recommended grid size of 2 × 2 km 2 . Whilst this calculated AOO is likely to be an underestimate of the true AOO because of incomplete sampling, it is probably not grossly so since the sizes of the extant forest patches within its range are small and declining due to forest clearing for conversion to agriculture, production of charcoal, timber logging, pole cutting and fuelwood gathering (Burgess & Clarke 2000). We have identified 17 to 20 locations for this species, and whilst this number is almost certainly declining, it is not close to meeting the threatened category thresholds under Criterion B. With further research, this taxon might qualify as threatened under Criterion A as some parts of its range have experienced over 30% forest loss of tree cover since 2000. However, in the absence of more information at present, this species is considered to be of Least Concern (LC). NOTES. Many East African herbarium collections previously identified as Whitfieldia stuhlmannii have been redetermined as W. orientalis. The latter is easily distinguished from the former by its tufts of floccose stellate hairs (see Fig. 1C) covering the young stems, its shorter basal, cylindrical portion of the corolla tube and its shorter inflorescences (1 -7 cm long vs 5 -10 cm). Moreover, the two white upper corolla lobes in W. orientalis are almost completely fused, whereas they are clearly divided in W. stuhlmannii. ( For a description of this species see Vollesen (2008: 268).

Whitfieldia stuhlmannii
DISTRIBUTION. This species is known from the forests of central and western D. R. Congo and in eastern Tanzania. In 2018, it was also found for the first time in Rwanda (Fischer 577/2018, see below). It may also occur in western Uganda in the forests of the Semliki National Park, close to the type locality, and might also be found in Burundi, but no collections of Whitfieldia stuhlmannii are known from those countries to date HABITAT AND ECOLOGY. Whitfieldia stuhlmannii grows in wet, undisturbed or somewhat disturbed forest. It is commonly found by streams in riverine forests, and occurs at 750 -1300 m a.s.l. CONSERVATION STATUS. This species has been collected in Tanzania and D. R. Congo, with a large EOO of 735,508 km 2 . However, except for three collections, most of the Congolese collections are old with little or outdated locality information, hence not all are mapped. In addition, eastern Congo has been under-collected and so we cannot use the calculated AOO (68 km 2 ) as a suitable measure for this assessment. A decline in the forest cover is ongoing within the range of Whitfieldia stuhlmannii due to conflict and mining operations in D. R. Congo (Butsic et al. 2015). Of the two known localities in Tanzania, Ntakatta Forest is protected as a Forest Reserve whilst Mount Livandabe is not protected. However, there is still considerable intact forest cover within the EOO of W. stuhlmannii at present. Moreover, even though most collections are from primary forests, a few specimen labels mention that this species can tolerate some disturbance ("forêt secondaire" or "Hevea"). Based upon those observations, this species is assessed as of Least Concern (LC). ( and so they should be treated as a single species under the earlier name, W. laurentii. Whitfieldia laurentii is also very similar to W. stuhlmannii in many respects: inflorescence size, corolla shape and pubescence of the different elements of the flower. Lindau (1894), when describing W. laurentii, also noted that these species were very similar but that W. laurentii has hairy bracteoles, a character that we have not observed on the dry material of W. laurentii. Lindau also added that the inflorescences of W. laurentii are pauciflorous and axillary, but this distinction is not obvious from our own observation of the two taxa. However, the flowers of W. stuhlmannii always show a very clear colour pattern on the corolla with its distinctive purple lobes contrasting sharply with the pure white tube and throat, and the corollas of W. stuhlmannii are generally larger than those of W. laurentii (c. 24 -59 mm long, vs 20 -25 mm long). Widespread logging for timber, construction, firewood and charcoal, and for clearing areas for subsistence agriculture is a serious threat to the biodiversity of the Mayombe Forest. In addition, the impact of slash and burn cultivation practice is a major concern in this region (Ron 2013). Transfrontier Conservation Areas in this region have been designated and could secure the future of Mayombe species, an area known to be of high botanical importance (Romeiras et al. 2014). No collections have been made in the area since 1989 and recent observations or collections are needed to confirm its current status in the Mayombe region today. As for the Aubeville area in Congo Republic where the species was collected in 1951, the forest has since suffered from industrial wood exploitation (1953 -1988), exploitation of goods and services by local communities, and finally, since 1988, strong pressure from slash and burn agriculture (Kimpouni 2008). Currently, W. liebrechtsiana is not known from any protected areas (Protected Planet 2020) and considering the small number of locations and the continuing decline in extent and quality of habitat, it is assessed as Vulnerable under criterion B: VU B1ab(iii,v). NOTES. With its long subulate bracteoles (see Fig. 1F), long slender inflorescences and its corolla which is purple in its upper half, becoming striped with purple lines in the middle and then transitioning to cream/ white, Whitfieldia liebrechtsiana is unlikely to be confused with any other species. Shrub, up to 150 cm tall; leafy stems stout, 4 -5 mm in diam., somewhat 4-angular or becoming terete, swollen above nodes (constricted in dry state), shortly pubescent when young with multicellular spreading hairs, later glabrescent. Leaves petiolate, petiole 1.5 -4 cm long, shortly pubescent; blade oblong-elliptic, c. 16.5 -19 × 6.5 -8.2 cm, base shortly attenuate then abruptly widened, margin entire, apex shortly acuminate; lateral veins c. 8 -9 pairs, these and midrib pale and prominent beneath; shortly pubescent on main veins beneath, elsewhere glabrous; cystoliths minute, dot-or shortly rod-shaped, surface beneath also with minute sunken dark glands. Inflorescences axillary in distal leaf axils, racemiform, 3.7 -13 cm long, secund; primary peduncle 9 -30 mm long, this and spike rachis pubescent with short multicellular hairs, these appearing deflated in dry state; cymes at each inflorescence axil single-flowered, opposite, becoming ± widely spaced at maturity; bracts along rachis green, linear or narrowly lanceolate, (4.5 -) 7 -13 × 0.6 -3 mm, shortly pubescent beneath, glabrous above; pedicels of individual cymes 1.5 -3.5 mm long, densely short-pubescent; bracteole pairs fused towards base, clasping the calyx, each pale yellowish-green, broadly ovate, 7 -11 × 7 -8.5 mm, base rounded or shallowly cordate, apex obtuse or shortly and bluntly acuminate, external surface shortly pubescent, internal surface glabrous; flowers sessile above bracteoles.

Whitfieldia liebrechtsiana
Calyx divided to the base into five oblong-oblanceolate lobes, pale yellowish-green but darker towards apices, lobes somewhat unequal in size, posterior lobe largest, 9.5 -11.7 × 3.6 -4.5 mm, other four lobes marginally smaller and narrower, each with apex rounded and surface with 3 or 5 subparallel veins, both surfaces glandular-pubescent. Corolla reddish-purple, lobes and distal portion of tube speckled purple and with darker purple veins, ± 21.5 mm long when dissected and flattened, glabrous externally; basal portion of tube somewhat bulbous, 3.7 -4.5 long and 4.3 -5 mm in diam., above which tube is slightly constricted and bent through c. 90°, and abruptly expanded into markedly saccate throat 11.5 -13 mm long and 10 -13 mm in diam., with few long multicellular hairs on floor within and with a raised ring of dense silky, blunt unicellular hairs at insertion point of stamens; limb of five ovate-triangular lobes in a weak 2+3 configuration, left-contorted in bud, becoming somewhat reflexed at anthesis; adaxial pair of lobes 5.5 -6.5 × 5.7 -6.7 mm, slightly asymmetric, fused for 1 -2.5 mm at base; 3 abaxial lobes 6 -7.5 × 5.5 -6.5 mm, apices slightly emarginate. Stamens 4, didynamous, held against roof of corolla; filaments sigmoid-shaped, longer pair 10 -15 mm long, glabrous; anthers included within corolla tube or those of longer pair partially exserted, thecae narrowly oblong, 2.9 -3.3 mm long, connective palepuberulous and with few short-stalked glands. Ovary 2 -2.5 mm long, glabrous, within cupular disk 0.8 -1 mm tall; style markedly curved, held against corolla roof, ± 19 mm long, shortly pubescent in distal half; stigma capitate, 0.5 mm in diam. Capsule only seen in immature state, 17 mm long including stipe 6.5 mm long, fertile portion broadly expanded, rounded with obtuse apex, glabrous; seeds not seen. Fig. 2.
DISTRIBUTION. Whitfieldia latiflos is only known from two collections in the northwest of Liberia: the type was collected 20 miles from Kakata, and the second    (Darbyshire 2014). The species then had only eight locations and its threat status was justified by the decline in the extent and quality of its habitat due to agricultural expansion. Whitfieldia letestui was assessed as Endangered (EN), being considered endemic to southwest Gabon, with only nine specimens known from eight subpopulations. Of these, four subpopulations occur in a mining concession and two in an oil concession, both of which are experiencing some lowlevel human pressure including forest clearance, whilst a seventh subpopulation occurs in a selectively logged forest. This led to the conclusion that the quality of habitat of this species is being reduced, with only 4 locations left in total (Stévart et al. 2016). However, these populations are treated here as a single species, and a new location, that of Ebo in Cameroon, has been added. The new EOO is now calculated as 137,816 km 2 and 13 locations have been identified. Within the EOO, the species still appears to be scarce, with few suitable forests which could host this species. Moreover, as far as Cameroon is concerned, all specimens collected fall outside the range of formally protected areas that are managed for their biodiversity, such as National Parks, although Ebo was formerly proposed as such. Similarly on Bioko, the single known specimen of this species comes from a non-protected area. It is possible that this species is under-collected due to a short flowering period (Lachenaud et al. 2018). However, with few collections known in total and most specimens coming from non-protected areas, the species is now assessed as Near Threatened: NT B2ab(iii). NOTES. Lachenaud et al. (2018) considered Whitfieldia letestui to be endemic to southwestern Gabon. However, Heine (1966: 44) had previously noted that this species was very close to W. preussii, but the scarcity of the collections prevented him from making taxonomic changes. The characters Heine cited for potentially distinguishing the two species (including the density of the inflorescence, the pilosity of the bracteoles and the length of the flower peduncle) do not clearly discriminate two taxa in the light of the collections available for study today. Consequently, following a thorough review of material identified as W. letestui and W. preussii, we conclude that these species cannot be distinguished and W. letestui is synonymised here. For a description of this species see Heine (1966: 36). HABITAT AND ECOLOGY. Neither the labels for the only two collections known nor the protologue of Whitfieldia purpurata give us any indication of its ecology. However, both specimens come from the Lastoursville area of Gabon, which supports evergreen humid forest (Escourrou 1988). From Google earth, (www.google.com/earth/index.html), we estimated that W. purpurata grows between 260 and 440 m a.s.l. CONSERVATION STATUS. This species is only known from the Lastoursville area, which is currently unprotected. A good forest cover in the area still exists away from the main roads but slash and burn and mining activities are present in the Lastourville area. Much collecting has been conducted in this region, where this species has not been recollected for almost 90 years. However, in the Ogooué-Lolo province more generally, a large part of the forest remains unexplored (Jongkind 2018), and Whitfieldia purpurata could well grow in those unexplored areas. As there is currently too little knowledge on the populations of this species, it is assessed as Data Deficient (DD). NOTES.
(1) In Flore du Gabon, Heine (1966: 37) noted that the type collection (the only specimen known for  Whitfieldia purpurata at the time, see Note 2) could be an extreme form of W. brazzae (Baill.) C.B.Clarke with bracteoles and flowers particularly large and a glabrous inner corolla surface. However, the inflorescences of W. brazzae are much more lax. As for the bracteoles, with more collections of W. brazzae being available to us today, we observe that even though its bracteoles are wide, they are elliptic to ovate and never fully rounded such as in the type of W. purpurata (see Fig. 3A). Moreover, we have not seen any specimen of W. brazzae with a purple corolla (the corolla of W. brazzae varies from pale yellow to salmonpink or orange). As for the corolla inner surface, it is often also glabrous in W. brazzae, so this character does not help to separate the two species.
(2) The specimen Le Testu 7255, determined by Heine as Whitfieldia thollonii (Heine 1966: 40) was also collected in Lastoursville in Gabon. Its corolla (25 × 7.5 mm) is, as in W. purpurata, more campanulate than funnel-shaped. The original label describes its colour as "jaune abondamment rayée de brun [yellow heavily streaked with brown]", confirmed by the observation of the dry specimens, although the stripes are not as neat as in W. thollonii. Instead, they resemble the flowers of W. purpurata. Indeed, the flowers of W. purpurata have been described as purple on the type label but a yellowgreen corolla heavily striped with brown (or purplishbrown) lines could also be perceived as purple. Indeed, a similar case has indeed been observed in the description of corollas of W. colorata subsp. tigrina, described both as "dingy purple" and "yellow with darker red lines". However, in a dry state, the corollas of that taxon all look similar with thick blurred dark purple stripes on a cream or yellow background. This is also the case in W. purpurata. As for the bracteoles in Le Testu 7255, they are not so clearly rounded as in the type of W. purpurata; instead, they have a rounded to ovate base with a distinct triangular acumen. However, measurements of the bracteoles at anthesis also exceed the range of bracteoles measured in W. thollonii (11 -11.5 × 7.5 -8.5 mm vs 9 × 7 mm). Since this collection is the only one so far determined as W. thollonii from Gabon, disjunct from the rest of its distribution in the Congo Republic and D. R. Congo, we believe it should be best placed under W. purpurata pending further observation of populations in the Lastoursville region. For a description of this species see Heine (1966: 37). HABITAT AND ECOLOGY. Whitfieldia rutilans is a rare species which grows in evergreen forests including along riverbanks, at 300 -500 m a.s.l. CONSERVATION STATUS. This species was first collected by Le Testu in Ndanda and Sacamicanda in the Lastoursville area which, even though unprotected, still has a good forest cover away from major roads. More recently, Whitfieldia rutilans was rediscovered by Texier and Bidault in the same area in eastern Gabon, adding three collections from two subpopulations to the ones already known and increasing the EOO of the species to 3,046 km 2 . Those recent c o l l e c t i o ns we re m a d e i n f o r e s t l o g g i n g concessions. Texier 349 and Bidault 1991 come from the SEEF concession, where the timber activity was halted in 2017 (Jongkind & Texier 2019) and Texier 840 comes from the CEB concession where the exploitation is recent but ongoing and threatening the fourth location for this species (N. Texier, pers. comm.). More generally in this region, even though the forest loss seems low beyond the main roads, timber exploitation in this non-protected area is continuing and is expected to further affect the quality of habitat for this species. It is therefore considered here to be Endangered under criterion B: EN B1ab(iii)+2ab(iii). A preliminary assessment was already available online (Stévart et al. 2016) with the same conclusion. NOTES. The flower colour of Whitfieldia rutilans was described by Heine (1966) as a bright ochraceous red or deep orange-red, based on the label data of the two Le Testu collections. Since then, the species has been re-collected by Texier and his photos show a pale orange corolla similar to that in some populations of W. brazzae and W. lateritia. However, W. rutilans differs from the first species by its short pedicels, dense inflorescence and glandular calyx lobes, and from the latter by its narrower corolla (4.2 -5.5 mm wide vs 8 -11 mm). More generally, W. rutilans is easily distinguished by its leaves reaching 28 cm in length (see Fig.  3B), which is unusually long for the genus.  Lindau (1895: 306). Type: Congo Republic, bois de Linzolo, fl., s.d., Thollon 210 (lectotype P! P00439334, selected here; isolectotype P! P00439335). Fig. 3D. Whitfieldia gilletii De Wild. (De Wildeman 1905: 316); Durand & Durand (1909: 421)  For a description of this species see Evrard & Demillecamps (1992: 96) while considering note (2) below. BR0000013109240). HABITAT AND ECOLOGY. Whitfieldia thollonii is recorded from forests, often alongside rivers, but also in palm plantations. The altitude range was not recorded on specimen labels and our knowledge on the ecology of this species is limited. CONSERVATION STATUS. With all the populations considered, we calculate an EOO of 44,635 km 2 and between 12 and 15 locations known historically. However, there is a distinct smaller core range (EOO = 7,553 km 2 ) in the vicinity of the two capital cities, Brazzaville and Kinshasa. The collections from there are rather old with the last one from 1975. It is possible that due to deforestation and urban expansion, those locations do not exist anymore today, at least halving the number of locations. The AOO in this core range is only 48 km 2 , but total AOO is potentially a lot larger for this species as we also need to consider the two outlier locations in Kikwik and Goa in the east. No recent collections of this species have been made so our knowledge on the current population is only estimated. To apply a precautionary approach, we assess it as Vulnerable VU B2 ab (i,ii,iii,v) due to the high level of threat in the core of the range and the assumption that fewer than 10 locations are extant. However, the extinction risk would be reduced if further subpopulations are found within its EOO and away from the large conurbations of Brazzaville and Kinshasa. NOTES. (1) Previously, this species was considered to be also found in Gabon (Heine 1966), based solely on the collection Le Testu 7255 from Lastoursville. We believe that this collection is actually referable to Whitfieldia purpurata (see Note 2 under W. purpurata above).
(2) Previous authors have noted that it is difficult to separate Whitfieldia thollonii from W. brazzae (Heine 1966: 42;Evrard & Demillecamps 1992: 96). However, Evrard & Demillecamps distinguished these species on the basis of the 10 following characters: a) corolla length; b) corolla length/width ratio; c) bracteole shape; d) density of glands on the calyx; e) colour of the corolla; f) leaf shape; g) position of the calyx in the bracteoles; h) pilosity of the inner corolla; i) pedicel length; and j) pollen morphology. Below, we consider these characters in turn.
(a) According to , the corolla length of Whitfieldia brazzae can reach 33 mm while W. thollonii can reach 34 mm but no range of measurements were given. From our observations, there is no significant difference in corolla length between the two species. (b) We were equally unconvinced by the argument that the length/width ratio (l/w) of the corollas differs in these taxa, as several specimens identified as W. thollonii by Evrard & Demillecamps have a l/w ratio which is higher than the supposedly maximum of 2.6 (e.g. Oddon 3754, Achten 482A), and so within the ratio range that they give for W. brazzae. (c) De Wildeman (1905: 317) and Heine (1966) agreed that W. thollonii and its synonym W. gilletii have elliptic bracteoles while W. brazzae and its synonym W. sylvatica have ovate bracteoles, whilst  argue it is the exact opposite. This disagreement could be explained by different definitions of those shapes by the authors. However, we looked at all specimens cited by those authors, as well as additional material, and we have seen significant variation in the shape of bracteoles ranging from elliptic to ovate with several intermediates, hence not resolving into two well-defined taxa. These differences in bracteole shape are also sometimes visible on a single collection, including on the type specimen for W. brazzae, De Brazza 165, which has both ovate and elliptic bracteoles. (d) In Evrard & Demillecamps (1992: 96), both species are described as having "a calyx covered with very small white glands" and no difference in the density of glands has been observed by us. (e) The colour of the corolla for W. brazzae is described as brown-yellow by  but a specimen they cited under that species (Compère 1919) was described on the specimen label as pink when fresh and shows clear purple lines when dry. Several specimens from P identified as W. brazzae have flowers which are orange (Farron 5113) or salmon red (Pobéguin 239). For W. thollonii, many labels do not indicate the colour of the fresh corolla. However, purple-brown streaks are present on corollas of all dried specimens of W. thollonii. Moreover, flower buds for this species tend to show dark purple lobes. Hence, even though the corolla colour for both species is more variable than has previously been described, we agree that dark stripes and dark lobes on the corollas are a good indicator of W. thollonii; W. brazzae can show stripes but they are faint and regular. (f) For these two species, as in most Whitfieldia species, the leaves are variable in size and shape within a species and consequently their measurements cannot be used for identification (the exceptions being W. rutilans, where the leaves are exceptionally large [see Fig. 3B], and W. purpurata where the leaves are distinctly cuspidate). (g) We did not find a significant difference in the position of the calyx: both taxa can have calyces which are either exserted from or hidden within the bracteoles. This is due partly to variation in the shape of the bracteoles, which can be markedly acuminate hence covering the calyx lobes, and partly to the maturity of the calyx as it continues to grow post-anthesis. (h) Even though the inner corolla of W. thollonii is said by  to be pilose, we found several specimens (e.g. Gillet s.n. 1903Gillet s.n. & 1904 for which it is glabrous as in W. brazzae. (i) There is a clear difference in pedicel length when the types of W. brazzae (18 -25 mm long) and W. thollonii (4 -13 mm long) are compared. However, when all specimens are considered together, this character also loses its strength and appears to be clinal. This is illustrated by the specimens Oddon 3554, Achten 482A and Achten 482B from D. R. Congo, which have unusually long pedicels for W. thollonii (respectively up to 20 mm, 17.5 mm and 25 mm long).(j) Only minute differences were found in the pollen of W. thollonii and W. brazzae as published by . The addition of the photograph of pollen from another specimen of W. brazzei (Fidao s.n.) during this study confirmed the similarities.
In conclusion, then, many of the characters previously used to separate Whitfieldia thollonii from W. brazzae do not hold up when the full range of material is examined and only corolla colour appears to differ consistently between the two. However, whereas W. brazzae has been collected multiple times both at the border between southwestern Gabon and Congo Republic and in the Brazzaville-Kinshasa area, W. thollonii seems much more confined to the second location only. More field observations and photographs are needed to understand at what rank these taxa should be separated. Only photographs of W. brazzae have been seen by the authors (see Fig.  3E), and photographs of W. thollonii would enable us to verify if other characters, such as the orientation of the corolla lobes, help distinguish these two taxa better. Additional studies might reveal these species to be better treated as only infraspecifically distinct.
the Batéké Plateaux (Protected Planet 2020), where it has been collected several times, including in the last fifteen years. It is therefore considered to be of Least Concern (LC). However, it would be useful to gather more details on current population size, trends and threats. It would be interesting, for example, to assess this species under the criteria A once we have more information on the current rate of decline. NOTES. (1) Baillon (1890b: 823) and Clarke (1899: 67) have spelt the epithet "brazzei", the genitive form of "brazzeus". However, this does not follow the established rules of Latin grammar for botanical nomenclature. Since the species has been named after Pierre Savorgnan de Brazza, the correct epithet should be "brazzae" (Heine 1966;Turland et al. 2018, Articles 60.8 and 60.9).
(2) Whitfieldia brazzae is very likely present in the Central African Republic and some specimens (e.g. Fidao s.n.) have been collected very close to the Congolese border with that country. However, no specimen from that country has been seen by the authors so far.
(3) For separation from the closely allied Whitfieldia thollonii, see the Notes section under that species. HABITAT AND ECOLOGY. Whitfieldia lateritia is found in disturbed and undisturbed forest, especially near streams, in gullies or canyons, between 190 and 1010 m a.s.l. CONSERVATION STATUS. This species has a large EOO of 174,821 km 2 and is known from many collecting sites. It has been collected frequently and it is still today fairly common in suitable forest habitats. According to the notes of past collectors, it also tolerates some disturbance. Nevertheless, it is likely to be declining due to forest loss, although not to such an extent that it would qualify under any of the IUCN threat criteria. This species is therefore assessed as of Least Concern (LC). NOTE. Whitfieldia lateritia is a rather frequently encountered species in surviving forest of Upper Guinea which shows a continuous variation of colour tones in the bracteoles, calyx (green to red) and corolla (from pale orange, brick-coloured, to pink or red, with or without stripes, sometimes dots, see Fig. 3F). A wide and continuous range of bracteole form has also been observed across its range, varying from broadly ovate to lanceolate or suborbicular, with or without an acumen. Moreover, their width ranges from 5 -12 mm, a rather wide range compared to other species in the genus. The indumentum on the floral parts also varies, with the bracteoles ranging from almost glabrous to pubescent. However, the stem and the pedicels are always villose with multicellular hairs, which facilitates its distinction from W. colorata (glabrous to puberulous) within its distribution range, and from the similar W. brazzae (glabrous) in Gabon and the Congos, where W. lateritia does not occur.
years, the Bomi Hills are also targets for open cast iron ore mining (Gunn et al. 2018). However, the subspecies is also present in areas of good intact forests within the Gola National Parks in both Sierra Leone and Liberia. Consequently, it is assessed as of Least Concern (LC).
et al. 2014). Three specimens were collected in a classified Forest (Forêt de la Niégré: Jongkind 4665; Leeuwenberg 2475 & 2757), but that area has not been exempt from deforestation since the 1970s and much land in this supposedly protected area has been sold to migrants who converted the forest into cocoa plantations (Eblin et al. 2018). Two more Ivorian collections (Leeuwenberg 2223; Breteler 6063) were made east of the Forêt de la Niégré in the 1950s and the 1960s respectively, and there is now very little forest cover left at that location. Another Ivorian collection was made in Tabou in 1973(Konning 2434 and the area is now heavily urbanised with little forest remaining, so there is little chance that the location still exists. This subspecies was collected in the Taï Forest in 1999 (Menzies CC1546). However, the Taï National Park, despite its protected status, has suffered from deforestation due to the expansion of cocoa plantation, illegal timber production, slash and burn agriculture and bush fires (Boissieu et al. 2007; Barima et al. 2016).
One collection was made between Bacanda and Tiepo [Chiepo] in 1965 (Aké Assi 8442) in the Go Bodiénou Classified Forest. Little is known on the protection of that Forest in the last 50 years, and even though some forest is still visible in the area on Google Earth, with the absence of protection (Protected Planet 2020), one can doubt of the presence of W. colorata subsp. tigrina there today. The locations in the east of Liberia, however, seem much more intact with good forest cover still present. Moreover, several collections were made in the Sapo and Cestos-Senkwhen National Parks where the forest is well protected. This portion of Liberia is also largely under-collected and has good suitable habitat for this taxon, hence it is rather difficult to calculate a precise AOO. To conclude, even though the situation for this species is perilous in Ivory Coast, the east of Liberia is much more promising with national parks and a good forest cover in most of the area where the subspecies can be found. We therefore assess this subspecies as Near Threatened: NT B1ab(iii,iv,v). NOTES. The authors contacted Carel Jongkind in 2018 to ask him if he had seen Whitfieldia latiflos while conducting fieldwork in Liberia. While he had not, he described interesting plants that he had collected from the Ivory Coast/Liberia border area and which did not key out as either W. lateritia or W. colorata. Those plants were new to us, and thanks to Carel's observations, a careful combination of morphological observations and mapping led us to conclude that three taxa were indeed present in the area. These are here treated as W. lateritia, W. colorata subsp. colorata and W. colorata subsp. tigrina. The two latter taxa appear much closer to each other (narrow calyx lobes, dense inflorescence) than to W. lateritia, particularly in a dry state, hence they were treated as subspecies in this synopsis. However, additional future collections, photographs, and observations at the fruiting stage might lead to a re-evaluation of their rank.

COLORATA
The two subspecies cannot be separated at the vegetative or fruiting stage. As a consequence, a few collections with fruits or fallen flowers cannot be assigned to one or the other subspecies with confidence. One flowering specimen (Voorhoeven 134, WAG) is unusual in its dense inflorescence with rather large bracteoles and could not be identified at the subspecies level; it needs further study. licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/ licenses/by/4.0/.