Vegetation , floristic composition and structure of a tropical montane forest in Cameroon

Floristic composition, forest structure and vegetation characterisation have been treated widely in tropical forest ecosystems (Campbell et al. 2006; Lutz et al. 2012; Neelo et al. 2015; Noumi 2013; Zhong et al. 2015), yet some tropical forest zones remain less studied. One such area is the tropical montane forest zone along the continental part of the Cameroon Volcanic Line (Ayonghe et al. 1999; Marzoli et al. 2000; Sainge et al. 2017), where few studies have assessed forest structure and composition along elevational gradients using permanent sampling plots (Gonmadje et al. 2011; Sainge 2016; Sunderland et al. 2003; Tchouto 2004). Rather, most work in this region has been based on more common general plant collection methods (e.g. Cheek, Onana & Pollard 2000; Thomas 1996, 1997).


Introduction
Floristic composition, forest structure and vegetation characterisation have been treated widely in tropical forest ecosystems (Campbell et al. 2006;Lutz et al. 2012;Neelo et al. 2015;Noumi 2013;Zhong et al. 2015), yet some tropical forest zones remain less studied. One such area is the tropical montane forest zone along the continental part of the Cameroon Volcanic Line (Ayonghe et al. 1999;Marzoli et al. 2000;Sainge et al. 2017), where few studies have assessed forest structure and composition along elevational gradients using permanent sampling plots (Gonmadje et al. 2011;Sainge 2016;Sunderland et al. 2003;Tchouto 2004). Rather, most work in this region has been based on more common general plant collection methods (e.g. Cheek, Onana & Pollard 2000;Thomas 1996Thomas , 1997. Some studies have examined changes in species composition and diversity across environmental and geographic gradients (Gentry 1988;Imani et al. 2016), but vegetation structure and composition are also influenced strongly by elevation (Imani et al. 2016;Richter 2008). Vegetation systems at different elevations on different substrates in montane ecosystems differ in biomass production, carbon storage and biodiversity conservation value (Richter 2008). Globally, although the biodiversity of elevational gradients in the tropics have seen much attention (e.g. Desalegn & Beierkuhnlein 2003;Fischer, Blaschke & Bassler 2011;Lovett, Marshall & Carr 2006;Rutten et al. 2015), this subject remains little studied in the Cameroon Mountains.
The study of elevational gradients in different mountain systems in the tropics began as far back as 1800, with the work of Alexander von Humboldt and Charles Darwin. This early work led many scientists to centre research on these questions (Fischer et al. 2011), aiming to understand the distributional changes of species along different mountain gradients (Hemp 2002;Lovett et al. 2006;Rutten et al. 2015). In East Africa, the work of Lovett et al. (2006), Rutten et al. (2015) and Desalegn and Beierkuhnlein (2003) classified and delimited montane forest as starting at elevations from 700 to 1000 m above sea level (MASL).
The Cameroon Mountains have been termed by different authors as the Mountains of the Cameroons (Durrell 1954), Biafran Forests and Highlands (Burgess et al. 2007;Cronin et al. 2014), Cameroon Line (Nono et al. 2004) or Cameroon Volcanic Line (Ayonghe et al., 1999;Marzoli et al., 2000). This chain of volcanic and plutonic isolated mountains covers 40 877 km 2 (Sainge 2016) and is isolated by ~2000 km from the Albertine Rift and by 1500 km-2000 km from the highlands of Liberia and Guinea (Alweny, Nsengiyumva & Gatarabirwa 2014;Burgess et al. 2007). This region ranks high among the biodiversity hotspots in the upper and lower Guinea forest and is thus considered to be of global conservation importance (Barthlott et al. 2005;Marchese 2015;Myers et al. 2000;Onana 2013;White 1983).
A few brief botanical collections were made in the Rumpi Hills area between 1976 and 1984, with a total of 57 botanical specimens collected by D. Dang, R. Letouzey, S. Polhill, B. Satabie and D. Thomas (National Herbarium of Cameroon database). It was not until 1996 that Thomas (1996) did a more intensive, rapid botanical survey of the area. Later, during 2000-2004, brief reconnaissance trips were made by D. Thomas, D. Kenfack and M. Sainge, during which 68 specimens were collected and deposited at the Missouri Botanical Garden Herbarium and the National Herbarium of Cameroon. The RHFR and surrounding areas are threatened by encroachment from both small farm estates and large-scale agro-industrial companies (Kupsch, Bobo & Waltert 2014).
In this study, we aimed to characterise vegetation patterns across the RHFR and to present a first detailed assessment of its vegetation structure and composition. This study is one element in a longer-term effort to assemble an understanding of the flora of the region. The specific objectives of this study were thus to understand the composition, structure and patterns in the RHFR along an elevational gradient.

Study area
The Rumpi Hills Forest Reserve lies near the south-western extreme of the Cameroon Mountain range, in Ndian Division, South-West Region, Cameroon. It stretches across latitudes 4.6°N -5.0°N and longitudes 8.8°E -9.4°E, with an elevational range of 50 m -1778 m. It covers an area of 458 km 2 (Forestry Ordinance 51 1941). Data were collected in clusters of 1-ha sampling plots: southern plots (numbers 1-8) were located on level terrain at elevations of 50 m -200 m; northern plots (numbers 18-25) on fairly level terrain at 400 m -600 m; four plots (numbers 14-17) on basaltic rock at 250 m -300 m; and eastern plots (numbers 9-13) on undulating terrain at 1200 m -1778 m ( Figure 1).
The climate of the RHFR is typical of equatorial Cameroon, being hot and humid, with two distinct seasons: dry (December to March) and wet (April to November). An annual total rainfall of 5000 mm has been reported for the reserve (Nembot & Tchanou 1998). Temperature fluctuates with elevation, with the coldest temperatures at the top of Mount Rata; although no climate station is located in this area, Nembot and Tchanou (1998) reported a mean temperature of 22°C for the reserve. This reserve forms a topographic platform for different river sources that supply the Chad, Benue, Sanaga, Congo and Manyu rivers (Ngwa 1978). Rivers originating in this reserve flow in five directions: north into Lake Chad via the Logone River; northwest via the Benue River, the Kimbi River and the Katsina Ala River; southwest into the Gulf of Guinea via the Ndian (Moriba), Moko, Meme, Mungo and Wouri rivers; southeast via the Kadei River, a tributary of the Congo River, and west into Nigeria via the Munaya and Mbo rivers.
The reserve per se is free from human settlements, as no villages are located within its core area. Twelve villages are within 1 km -3 km of the reserve margins: Matamani in the northwest; Mata in the north; Madie and Dikome Balue in the east; Munyange and Nalende in the south; Mbange, Bossunga, Motindi and Lipenja Mukete in the west; and Meka and Besingi in the northwest.

Field sampling
A reconnaissance survey based on topographic and vegetation maps of the reserve (Letouzey 1985) was carried out to identify homogeneous areas of putatively different vegetation types (Sainge & Cooper 2014). Data collection was done from February to June 2015, using 25 1-ha plots, which were placed based on accessibility to sample different vegetation types and elevations ( Figure 1). Each plot measured 500 m long × 20 m wide and was subdivided into 25 quadrats of 20 m × 20 m. For each plot, global positioning system (GPS) coordinates were recorded for the four corners, including start and end points (Appendix 1), via careful, repeated measures to assure accuracy (coordinates of these permanent plots are available at http://hdl.handle. net/1808/25180). In each plot, all trees and lianas with diameter at breast height (dbh, 1.3 m above ground) of ≥ 10 cm were identified, measured with a diameter tape, tagged, recorded and mapped using their GPS coordinates. Smaller trees, shrubs and lianas (dbh < 10 cm) were sampled and measured with calipers in 10 m × 10 m quadrats located in every fifth 20 m × 20 m quadrat in each 1-ha plot. The forest was divided into four vertical strata: trees < 10 cm dbh as understory, 10 cm -30 cm dbh as mid-canopy, 30 cm -60 cm dbh as canopy and ≥ 60 cm dbh as emergent species. Finally, we recorded non-plot-based observational data (i.e. general plant collections) to detect and include species not present on the standardised plots.

Taxonomy and plant identification
In the field, plant identification was done using five-letter codes, including the first three letters of the genus and the first two letters of the species. In cases where the genus and species were not known or only the genus or family was known, arbitrary codes were generated to represent morphospecies. For unknown species (those that could not be identified and those partly identified or with doubtful identification), herbarium specimens were collected, labelled, pressed and dried for proper identification at the National Herbarium of Cameroon, in Yaoundé. Flowers and fruits were collected when the species was possibly new to science, endangered or endemic to the area. Identification in the herbarium was accomplished by comparing and matching specimens with existing collections and available floras and monographs. Plant classification followed species lists in the Angiosperm Phylogeny Group (APG III, 2009), with the Papilionaceae, Caesalpiniaceae and Mimosaceae merged into Fabaceae, and Sterculiaceae, Tiliaceae and Malvaceae merged into Malvaceae (APG III 2009;Judd et al. 1999).

Data analysis
Correlation analysis was used to assess the relationships between numbers of species and elevation, in PAST, version 2.17 (Hammer, Harper & Ryan 2001). Individuals not identified to the species level (1271 individuals with dbh ≥ 10 cm, 10.6%) and singletons (57 individuals, 0.5%) were excluded from inventory completeness calculations. Thus, 10 709 fully identified individual trees (dbh ≥ 10 cm, 88.9%) of 311 species were used in the analysis. Classification of the vegetation was achieved using twoway indicator species analysis (TWINSPAN; Hill 1979). Detrended correspondence analysis (DCA; Hill & Gauch 1980) was used to examine relationships between . Forest structure and composition were described using basal area, relative density, relative basal area, relative frequency and importance value index (Dallmeier 1992). Inventory completeness was assessed using EstimateS version 9.1.0 (Colwell 2013: http://purl.oclc.org/estimates), via the Chao2 estimator of expected species richness (S exp ), which is calculated from the number of species actually known from the site (S obs ) and frequency of detection of rare species; completeness was calculated as S obs /S exp , where S exp = S obs + a 2 /2b, a is the number of species detected only once and b is the number of species detected exactly twice. Completeness is then calculated as S obs /S exp .

Species accumulation and inventory completeness
Inventory completeness varied considerably among plots, from a low of 0.36 in lowland plot 2 to 1.0 in lowland plot 1, with an overall mean of 0.73 (Figure 2a and b).

Composition and floristic structure
We recorded 4086 individual trees in lowland evergreen rainforest, 3600 in mid-elevation evergreen forest, 1831 in lowland evergreen rainforest on basalt rocks, 1191 in submontane forest, 1066 in montane cloud forest and 263 in transitional forest. For all vascular plants ≥ 10 cm, the highest mean number of trees per hectare (596) was recorded in the submontane forest (range 542-649), whereas the lowest mean tree number (263) was obtained in the transitional submontane forest (Table 1). The mean number of shrubs per 0.05 ha varied as follows: 180 shrubs/0.05 ha in lowland evergreen rainforest (range 140-212) and 71/0.05 ha in montane cloud forest (59-83). The mean number of lianas ranged from no lianas in montane or transitional forest at dbh ≥ 10 cm to 6 lianas per hectare (range 0-12) in lowland evergreen rainforest (Table 2).

Multivariate analysis
The 25 1-ha plots were classified in the TWINSPAN analysis into six groups at 50% similarity ( Figure 4). Plots 1-8 corresponded to lowland evergreen forest (sensu Letouzey 1968Letouzey , 1985, characterised by an abundance of Oubanguia alata Our floristic dataset of 25 plots was also subjected to DCA analysis and plotted along Axes 1 and 2. Variation was expressed along Axis 1, with an eigenvalue of 0.772 and a gradient length of 4.183, which reflects high variation among vegetation types and species composition. Vegetation types 4, 5 and 6 (submontane, montane and transitional forest, respectively) separated toward the positive side of DCA Axis 1, whereas vegetation types 1, 2 and 3 (lowland, basalt and mid-elevation, respectively) separated toward the negative end ( Figure 5). DCA Axis 2 showed a weaker eigenvalue of 0.478, with a gradient length of 2.389 (Table 3). Figure 5 shows patterns suggesting that vegetation types 1-3 are more closely related than vegetation types 4-6.
A high species-environment correlation for Axis 1 indicates a strong association between vegetation type and elevation, which can be verified from the biplot record ( Figure 6). A Monte   Carlo permutation test (998 runs) with an eigenvalue of Axis 1 and significant at p < 0.001 confirmed the strong relationship between species composition and elevation ( Figure 6).

Vegetation patterns Lowland evergreen rainforest abundant in
Oubanguia alata Baker f.

Transitional submontane forest abundant in Trema orientalis (L.) Blume
This vegetation type is a mosaic of forest and grassland occurring along the eastern edge of the RHFR at 1300 m -1600 m, close to Dikome Balue, and is characterised by low diversity. We sampled 263 individual trees in 20 families, 31 genera and 32 species with dbh ≥ 10 cm and 61 individual trees in 16 families, 22 genera and 24 species with dbh < 10 cm. Dominant tree families with dbh ≥ 10 cm were Euphorbiaceae and Rubiaceae (4 species each); all other families were represented by only 1-2 tree species.

Discussion
The diverse and heterogeneous vegetation structure and composition in RHFR are likely related to the complex physical features, elevational differences (from 50 m to the top of Mount Rata) and climatic factors, such as the south-western monsoon (warm wet) winds that are weak in the dry season and strong in the wet season (Neba 1999;Ngwa 1978). Rainfall is highest in the south-western corner of the reserve; however, to date, no weather station has been installed to provide detailed climatic data for the reserve. We assume that, given its proximity to Korup National Park, with an average yearly rainfall of ~5 m (Chuyong, Newbery & Songwe 2000Newbery & Gartlan 1996;Thomas et al. 2003), RHFR also has high rainfall; temperature within the reserve is variable, with temperatures lowest at the top of Mount Rata.
The full set of characteristics of RHFR (i.e. its enclaved nature, importance as a watershed and lack of human settlements) are indicators of conservation importance. This importance is emphasised by our findings of high tree, liana and shrub diversity (Table 1) and the large number of species of high conservation priority (Online Appendix 1), as well as the fact that it occurs in a recognised biodiversity hotspot (Barthlott et al. 2005;Marchese 2015;Myers et al. 2000;Onana 2013;White 1983).

Vegetation patterns and floristic composition
Multivariate analyses (TWINSPAN and DCA) showed strong influences of elevation on forest types. TWINSPAN analyses classified RHFR vegetation into six types. Letouzey (1985) recognised seven vegetation types in RHFR: the Atlantic Biafran forest; Atlantic littoral forest (which in the current study is classified as lowland forest); piedmont forest; degraded submontane forest; submontane forest; highly degraded evergreen forest (which was not sampled in the present study); and submontane grassland (also not sampled), as well as various combinations of fallow, grazed and human-inhabited areas (Figure 1).
The RHFR is part of the chain of mountains of Cameroon and Nigeria that includes the Cameroon Mountains and associated highland biomes (Burgess et al. 2007;Cronin et al. 2014). It forms part of the Lower Guinea Forest, with high levels of species richness and endemism (Barthlott et al. 2005;Burgess et al. 2007;Plumptre et al. 2007). The occurrence of a mosaic of forest and grassland on the upper slopes of Mount Rata, at elevations of 1300 m -1600 m, is not surprising, as grassland savanna begins at 1500 m in the Takamanda Forest Reserve, in the South-West Region of Cameroon (Sunderland et al. 2003), and above 2000 m on Mount Cameroon (Richards 1963). Administratively, Mount Rata falls outside of the RHFR and yet represents the highest peak in the Rumpi Hills range.
Our results differ from patterns in lowland forest at other sites, like the Takamanda Forest Reserve (South-West Region, Cameroon), where lowland forest is dominated by Huaceae (Afrostyrax kamerunensis) and Irvingiaceae (Irvingia, Klainedoxa, Desbordosea) (Sunderland et al. 2003). However, the RHFR lowland forest showed the same dominance trends as the lowland forest in nearby Korup National Park (Thomas et al. 2003).
Forest structure changed from lowland evergreen forest (50 m -200 m), with some trees 35 m -55 m tall, to montane cloud forest (1778 m), with a lower and more even canopy 20 m -25 m tall, comprising trees with branches covered by Piperaceae, Orchidaceae, ferns, liverworts, lichens and so on. Our results agree with Letouzey (1985) that RHFR is composed of different vegetation types and show that these vegetation types demonstrate impressive variation in structure, species composition and distribution. Furthermore, RHFR contains a distinct montane vegetation type, as detected and defined by the TWINSPAN analysis, at elevations above 1600 m. This result concurs with Vallèrié (1971) and Thomas (1984), who both classified upper montane forest as starting from 1600 m in the Cameroon Mountains region.

Elevation
The effect of elevation on the vegetation of the RHFR was pronounced, as it influences vegetation pattern, vegetation structure, species diversity and species composition of the area (Figure 4) across an elevational range of 50 m -1778 m.
We documented marked changes in species composition with elevation: lowland evergreen rainforest on basalt and lowland evergreen rainforest rich in O. alata were relatively richer in species than the other vegetation types. The inverse relationship between species richness and elevation recorded in RHFR was consistent with results obtained in many studies (Chuyong et al. 2011;Hamilton 1975;Henrik et al. 2006). Dauby et al. (2013) investigated tree diversity patterns in communities of evergreen forest trees in five landscapes of western Gabon and concluded that mean alpha and gamma diversities were much higher in the hilly region, with differences in elevation explaining a significant part of species turnover. Decreased alpha diversity with elevation within the hilly region could be associated with mass effects, which are expected to enrich valleys and slopes (Dauby et al. 2013). Overall, in RHFR lowland forest is characterised by large trees with huge buttresses and lianas; at high elevations, shrubs and tree branches are covered with bryophytes and vascular epiphytes and tree boles and leaves are covered by moss and liverworts, with a minimal liana population.
Our detailed sampling across vegetation types and elevations within and near the RHFR makes our data useful both for ecological understanding and for guiding management decisions. Given that our plots are permanent, with GPSbased outlines of each sampling plot (http://hdl.handle. net/1808/25180), the opportunity arises to repeat these censuses in the future (e.g. every 5-10 years) to understand the dynamics of the forest (Condit 1998). Such detailed monitoring would allow a far more nuanced understanding of the status and condition of these forests, as well as of the effects of global change on their composition and structure.

Conservation implications
Conservation prioritisations usually involve a number of factors for designing conservation areas, incorporating aspects of species composition, structure, vegetation patterns and socio-economic and cultural importance of the sites. Our study confirmed that RHFR is a rich site in terms of vegetation; analysis of species composition segregated the community into six vegetation types, some of which are found only outside of the administrative boundaries of the reserve (e.g. montane cloud forest). Our records of rare species, such as Deinbollia angustifolia, Korupodendron songweanum, Gambeya korupensis and Oubanguia alata in lowland forest; Crateranthus talbotii in lowland basalt forest; Rhaptopetalum geophylax and Cylicomorphia solmsii in submontane forest; and Carapa oreophila, Oncoba lophocarpa and Xylopia africana in montane forest, may upgrade the conservation interest in the reserve. Particular species occurred only in specific vegetation types or in a few adjacent types. Hence, protection of each of the different vegetation types, including the unique vegetation types at higher elevation on Mount Rata, is paramount for conservation.