Floristic composition and turnover analysis in Dahomey Gap and the surrounding sub‐humid Togolese mountain minor forest refuges: Importance for biogeography and biodiversity conservation in sub‐Saharan Africa

Abstract The origin of Dahomey Gap (DG) flora is one of the central biogeographical questions in sub‐Saharan, which has been addressed in several studies. However, floristic evidence based on representative samples from the DG seems to be lacking in the scientific debate. The present study was conducted to fill this gap. Specifically, we assessed Togolese mountain riparian forests as minor forest refugia, examined their contribution to larger sub‐Saharan forest refugia, and discussed the significance of these findings for biogeography and biodiversity conservation. Southwest Togo, West Africa, and sub‐Saharan Upper Guinea Region Floristic data were collected in riparian forests through an intensive botanical inventory (n = 198; 50 × 10 m2). A comparative analysis was performed based on the floristic evidence related to forest refuges. The results showed significantly high species richness (868 species) and a high gamma and beta diversity associated with spatial turnover patterns. They also showed a high affinity between the study forests and large sub‐Saharan forest refugia. Riparian forests share about 60% of their current species richness with large sub‐Saharan forest refugia and contained refuge bio‐indicator species. The floristic evidence, consistent with those of other studies, suggests that Togolese mountains would have very ancient origins and have experienced paleoclimatic events. The studied riparian would have served as refuges during recurrent climatic episodes. Our results support the minimal forest fragmentation hypothesis (network of refugia along rivers). However, they seem to be incompatible with the idea that the DG flora may be essentially a relic of the early Holocene. In sub‐Saharan Africa, where maintaining a vast area of natural forest is difficult due to human pressure, efforts to preserve maximum species diversity should include a focus on the conservation of minor forest refuges, particularly in sub‐humid mountain riparian zone.


| INTRODUC TI ON
Forest refugia are crucial as retreats for many plant and animal species, which would have become extinct during the past climate change coinciding with the glacial cycles (Broecker & Denton, 1989;Husemann et al., 2014;Maley, 1980). They have played an important role in structuring the biotic and genetic diversity of these organisms (Carlson et al., 2012;Ley et al., 2016;Nicolas et al., 2010). According to ecologists, forest refugia can again function similarly during current and future climate change and are a high conservation priority as key areas for the long-term persistence of species and genetic diversity (Médail & Diadema, 2009;Montade et al., 2014). Hence, it is important to identify ancient climate refugia where possible and protect them at multiple scales (Eeley et al., 1999cited in Noss (2001).
Despite the importance of minor forest refuges for their impact on maintaining current and future diverse gene pools, they have received little attention in sub-Saharan Africa , probably due to the previous perception of forest refugia in this area. It has long been suggested that in the late Quaternary Period, the African rainforests were restricted to several large refugia (Maley, 1989;Maley & Brenac, 1998). Moreover, the sub-Saharan African refugia located south of the equator are highly developed, while those located north of the equator are rather smaller (Linder, 2001). However, recent phylogenetic studies in sub-Saharan Africa areas have contradicted these ideas by supporting that there are many small, rather than a few large, forest refugia in tropical Africa (Bohoussou et al., 2015).
In sub-Saharan Africa, the Pleistocene forest refugia hypothesis is supported by palynological, climatological, and phylogenetic evidences. Major climatic refugia would probably have been located near Cape Palmas and the Nimba Mountains in Liberia, Cape Three Points in Ghana , Benin-Ghana, southwest Cameroon, southern Gabon, northern Gabon, eastern Democratic Republic of Congo, and western Uganda (Maley, 1989;Nicolas et al., 2010). However, minor refugia have not been identified elsewhere and studied in terms of biogeography and biodiversity conservation. Few studies on sub-Saharan tropical landscapes, however, locate and assess the contribution of minor forest refugia to the distribution and conservation of local or regional plant diversity (Hall, 1973(Hall, , 1981. To the best of our knowledge, the only case studies on the biogeography of minor forest refugia focus on the "ecological islands of south-eastern Nigeria" and the southern slopes of Mount Cameroon (Hall, 1973(Hall, , 1981. Based on the floristic evidence, Hall (1981) has showed that the ecological islands of southeastern Nigeria served as a rainforest refuge during the Pleistocene climate change. This approach also allowed him to estimate the minimum size of a viable conservation area at 10 km 2 , for the purposes of conserving floristic diversity in the small Pleistocene refugial forests of Nigeria.
The hypothesis that riparian forests acted as refugia for flora in sub-humid tropical landscapes covered by savannah or deciduous forest during the arid Pleistocene period has been proposed to explain the relationships between the species composition of riparian and upland (i.e., non-riparian) forests. In addition, it was postulated that tropical forests re-expanded rapidly in the early Holocene originated from these refugia. Evidence for these assumptions has also been demonstrated in the temperate forests and in the Neotropics forests (Aide & Rivera, 1998;Meave & Kellman, 1994;Silva de Miranda et al., 2018). According to Meave and Kellman (1994), if the significant number of upland rainforest species can survive in these riparian forests, they could provide a plausible explanation for the survival of regional forest biotas during drier Pleistocene phases and their rapid reappearance in the Holocene pollen record. However, such studies have not been conducted in sub-Saharan Africa.
Although riparian forests are frequently studied in sub-Saharan tropical Africa, much less information is available on their biogeography, particularly in mountain areas. Previous studies were conducted in South Africa (Hood & Naiman, 2000), Benin (Natta, 2003), Togo (Adjossou, 2004;Kokou et al., 2008), Central Africa (Gautier-Hion & Brugiere, 2005), the Volta Sub-Basin of Ghana (Boakye et al., 2019), Southern Sierra Leone (Fayiah et al., 2020), and South East Kenya (Habel & Ulrich, 2021) have shown that riparian areas are generally richer than other forest ecosystems. However, the floristic similarity between riparian and "Terre firme" forests has not been investigated to estimate the proportion of regional humid relict species in riparian forests, thus allowing the discussion of their status as forest refugia.
The biogeography and history of the Guinean-Congolian rainforest have received a great deal of attention. White (1979White ( , 1983 defined the Guinean-Congolian rainforest as a phytogeographic domain comprising about 8000 plant species, of which 80% are endemic. These forests have been subdivided into three sub-centers of endemism (Fayolle et al., 2014;Hardy et al., 2013;Linder et al., 2012;Poorter et al., 2004): (1) the Lower Guinea forests, which extend from Nigeria to the eastern border of Gabon-coinciding with the separation of the Congo and Ogooué basins, (2) the Congolese

T A X O N O M Y C L A S S I F I C A T I O N
Biodiversity ecology, Biogeography, Botany, Community ecology, Environmental sustainability, Evolutionary ecology, Restoration ecology, Taxonomy forests properly so called, which are confined to the Congo watershed and (3) the Upper Guinea forests, stretching from Ghana to Guinea and separated from the other entities by Dahomey Gap, the savannah band in Togo and Benin. The origin of the biogeographical feature of African rainforests has been the subject of interesting literature reviews (Hardy et al., 2013;Poorter et al., 2004).
According to the authors, about 2.4 million years ago, the polar ice caps expanded to the point where they extended into temperate zones. Cold and temperate periods thus followed one another according to the periodic oscillations of the Earth in its orbit around the sun (Milankovitch cycle). Warming occurred 16,000 years ago, which caused our current climate. The ice ages changed the spatial arrangement of the climate zones and thus, influenced the environmental conditions. Species could either follow the new organization of the climate zones, adapt to the new conditions, or become extinct. These ice ages caused a great drought in the African continent.
Tropical forests survived in a few refuges, which impoverished the tropical forest flora. About 6000 years ago, rainforest occupied a much larger area than presently and extended across the Dahomey Gap, thus connecting Upper and Lower Guinea (Maley, 1980(Maley, , 1997. So, the biogeographic history of sub-Saharan tropical Africa is rich, especially around the Dahomey Gap. The origin of the DG flora is one of the central biogeographic questions in sub-Saharan Africa, which has been addressed in recent phylogenetic and floristic studies (Demenou et al., 2016;Duminil et al., 2015;Fayolle et al., 2014;Koffi et al., 2011;Linder et al., 2012;Marshall et al., 2021).
Phylogeographic data suggest that the forest flora of the DG may be essentially relics of the early Holocene when the geographical distribution of the Guinean-Congolian forest reached its maximum (Demenou et al., 2016;Duminil et al., 2015;Koffi et al., 2011).
However, the floristic evidence for these conclusions, based on representative samples from the DG, seems to be absent from previous reports on this topic. It is true that there is floristic evidence based on regional data that has allowed discussion of the biogeography of sub-Saharan Africa as a whole (Fayolle et al., 2014;Linder et al., 2012).
But these data were not representative of the flora of the Dahomey Gap, as the plots sampled were selected elsewhere across the forest blok. A very few plots (1 or 2) were selected in the DG plain. This was also true for some species used in phylogenetic studies such as Distemonanthus benthamianus (Demenou et al., 2016), which were not from the Dahomey gap but from the Togo mountains. Till date, no study based on large floristic data collected in the DG has addressed biogeographical issues. Phylogeographic and genetic data should therefore be compared to those of the representative floristic data, taken in and around the DG, to prove their value in understanding the biogeographic history of the Dahomey Gap.
The mountains of Togo form the main mountainous area adjacent to the DG in sub-Saharan Africa. Due to this position, the vegetation of the Togolese Mountains is often described as dry vegetation similar to that of the DG (Hall & Swaine, 1976). The vegetation of the Togolese Mountains has been the subject of several botanical surveys (Aké Assi, 1971;Akpagana, 1989Akpagana, , 1992Brunel, 1977Brunel, , 1978Guelly, 1994).
Nevertheless, there is a significant lack of information specifically concerning the riparian forests of the region, in which new species of Togolese flora have recently been discovered (Adjossou, 2004(Adjossou, , 2009Adjossou & Kokou, 2009;Kokou, 1998;Kokou et al., 2008). Although several studies have focused on the vegetation of the Togolese Mountains, little information on this vegetation has been published in international journals (Adjossou & Kokou, 2009;Akpagana, 1992;Kokou et al., 2008); consequently, information on this vegetation is not easily accessible to a wide scientific audience.
We test the hypotheses that (1) during the opening of the DG, the riparian zones of the Togolese mountains would have served as a mini-refuge for forest flora during restrictive climatic regimes; the populations persisting in these zones would have recolonized the surrounding landscape when conditions favorable to their survival and reproduction returned; (2) the DG would have been subjected to repeated openings (in the Pleistocene and then in the Holocene); during these crisis periods, forest fragmentation would have been minimal or would have taken the form of a network of refuges (along rivers and/or in the mountains) where most species would have survived; (3) the forest flora of the DG could be essentially relics of the early Holocene period when the geographical distribution of Guinean-Congolian forest reached its maximum (Demenou et al., 2016;Duminil et al., 2015;Koffi et al., 2011).
We aimed to comparatively study the flora of the Togolese Mountains riparian and those of the large tropical climatic forest ref- uges. Specific objectives were to assess Togolese riparian as minor forest refugia, examine their contribution to larger sub-Saharan forest refugia and discuss the importance of these results for biogeography and biodiversity conservation in sub-Saharan tropical Africa.
We base our study on the floristic evidence and attributes commonly used in the field of biogeography. According to the ecologists and biogeographers, areas that have served as historical refugia during restrictive climatic regimes currently harbor higher levels of biodiversity than areas that have not experienced refugia (Boyer et al., 2016;Russell et al., 2009). They are also characterized by a high incidence of rare species, and a high level of beta and gamma diversity associated with species turnover (Cowling & Lombard, 2002). These are complex areas with a large number of endemic species, typical species of many forest types (Do Prado et al., 2014;Fiaschi & Pirani, 2009), and a significant proportion of tropical rainforest species (Meave et al., 1991;Meave & Kellman, 1994). These areas contain bio-indicator taxa of rainforest refugia (Blan, 1993;Tchouto et al., 2009) and share similar vegetation (White, 1983). All these floristic evidence were evaluated and compared with those of other studies (palaeobotanical, palynological, phylogeographical, and geological).

| Study area
The study area (6°15′-8°20′N; 0°30′-1°E) covers 6441 km 2 in the southern part of the Atakora Mountains, southwest Togo, on the border between Togo and Ghana ( Figure 1). It is adjacent to the DG, an extension of the savannah woodlands from the Sahel to the Gulf of Guinea, which separates the Upper Guinean forests from the rest of the African rainforests . The climate in this area is mountainous (Papadakis, 1966) and characterized by a long rainy season (8-10 months). The average annual temperature and total annual rainfall ranges are 21-25°C and 1250-1900 mm, respectively. The study area presents a strong topographic and geomorphological heterogeneity, which is named Togo Mountains, of Togo, and Buem sandstones that were extensively folded and metamorphosed during the Pan-African Cambrian orogeny (Hall & Swaine, 1976 (Akpagana, 1989).
For the past four decades, the forests in the Togolese Mountains have been affected by agriculture, particularly coffee and cocoa production, which has led to repeated clearing. Charcoal production by local people in the 1990s, after the reduction in coffee and cocoa prices, has tended to aggravate the situation. Thus, the remaining F I G U R E 1 Map showing the hydrographic network of Togo highlands and the location of the 198 study plots.
forest is highly fragmented today and mainly located in hard-toreach areas and along rivers (Adjossou, 2004). The high population growth rate in Togo (3.3% per year) also affects the forest remnants, in contrast to those along the rivers, which are still relatively untouched by the local population partly for traditional reasons.

| Site selection and plant survey
A small-scale topographic map (1:200,000) available for the study area was used to identify the hydrological network and potential sites with riparian forests for selecting the sampling sites. The sites were selected according to a general strategy to cover the habitat diversity (Mueller-Dombois & Ellenberg, 1974). However, sites located on steep slopes, which are very difficult to access, were not selected Africa (Kokou, 1998;Natta, 2003). Following the riverbed, usually upstream, the survey plots were located using GPS. At each site, the presence/absence of vascular plant taxa (trees, shrubs, lianas, herbs, and epiphytes) was noted. To have a complete floristic composition, these surveys were completed by including the species not observed in the 198 samples but observed in the riparian vegetation during the inventory. Taxon identification in the field is based on the floras reported by Aubréville (1959), Brunel et al. (1984), and Hutchinson et al. (1954and Hutchinson et al. ( -1972. Specimens of species that could not be identified in the field were collected and identified later at the herbarium of the University of Lomé (Togo). To standardize the species names, the African plant database was used.

Species and taxonomy richness
Species richness is the number of species recorded for a specific group of organisms during a given period. To calculate the species richness of the riparian forests of the Togolese Mountains, a general floristic list was established by compiling the inventories of the 198 floristic surveys, completed by the species not identified in the 198 surveys but observed in the riparian forests during the field surveys. However, only the species recorded in the plots were used in the statistical analyses. The inventory of the accepted species was based on those reported by Aubréville (1959), Brunel et al. (1984, and Hutchinson et al. (1954Hutchinson et al. ( -1972. To avoid duplication, the accepted species were listed alphabetically. The synonyms for each species were checked using the African plant database. The total number of species on this general list was taken as the species richness. Then, the resulting species richness was compared to those of the richest sites in sub-Saharan Africa considered as forest refuges, such as the Ziama forest in south-eastern Guinea, the Taï forest in south-eastern Côte d'Ivoire, the closed forests in southern Ghana, the Monts Doudou forests in southwestern Gabon, the Campo-Ma'an forests in southwestern Cameroon, etc. (Table 1). To calculate taxonomic richness, the species recorded were grouped according to their taxonomic level (such as genus and family). Within each group, the species richness was calculated. The floristic list of the riparian forests of Monts Togo was prepared according to the APGIII classification system to compare it with the available floristic lists on the vegetation of sub-Saharan Africa, mainly established according to the APGIII classification system (Fayolle et al., 2014; African plant database https://www.ville -ge.ch/musin fo/bd/cjb/ afric a/index.php?langu e=an).
According to Kindt and Coe (2005), we recorded the number of species for several sites in most situations. We did not cover the whole area of interest. Accordingly, we do not know the species composition of the unsampled area. However, as species richness depends on sample size, we cannot expect to have recorded all species present in the study area. Richness estimators were therefore used to estimate the total number of species in the riparian forests surveyed. The approach used is as follows: based on the presence/ absence matrix (198 surveys × 828 species), species accumulation curves were drawn to determine whether the surveyed sites were sufficiently sampled. Next, first and second-order jackknife richness estimators (Jack1, Jack2) were used to estimate the total number of species in the surveyed riparian forest. Different methods of mathematical richness extrapolation exist, but the jackknife technique was chosen to predict total richness because it has been shown to outperform other estimators). Jack1 and Jack2 estimate the minimum and maximum number of species, respectively.

Species diversity and habitat heterogeneity analysis
To assess the habitat diversity and heterogeneity level, gamma, alpha, beta diversity, and Shannon's diversity index were calculated. Gamma, alpha, and beta diversity were used as basic diversity indices (Whittaker, 1972). Gamma diversity, the diversity at the landscape scale, is calculated as the total number of species or total richness across plots. Alpha diversity is calculated as the average species richness per plot. Beta diversity is a measure of data heterogeneity or habitat diversity (Mccune et al., 2002). The classical metrics were used to determine beta diversity: βw = (Sc/S) − 1, where βw is beta diversity, Sc is the number of species in the composite sample (the number of species in the dataset), and S is the average species richness in the sample units. If βw = 0, then all sample units have all species. Dataset heterogeneity increases with increased βw. As a rule, in this context, βw < 1 is rather low and βw > 5 can be considered high. The maximum value of βw is Sc − 1, which is obtained when the sampling units do not share any species. The Shannon's diversity index, derived from information theory, considers the number of species present in survey i (n i ) and the relative cover (R ij ) of the different species j in survey i (Shannon & Weaver, 1949). This index quantifies the heterogeneity of the biodiversity of a study area and thus indicates changes over time. The index decreases (≈0) when the number of species is low and only few species are dominant.
H′: Shannon biodiversity index; i: a species in the study area; S: species richness; p i : proportion of a species i in relation to the total number of species (S) in the study environment, which is calculated as follows: where is the number of individuals for species i and N is the total number of individuals; log2: logarithm to base 2.

Species rarity analysis
The rank-frequency curve has been used to analyze rarity. This curve is generally represented by a graph, where the ranks of the species are plotted on the x-axis in order of decreasing abundance, and the absolute or relative frequencies in the collection examined are plotted on the y-axis in a logarithmic or semi-logarithmic scale, which gives a more complete picture of the population than simple diversity index. Three types of rank-frequency curves exist. However, we have chosen the curve cited by Frontier (1976) because it provides information on the diversity of the populations in the environments surveyed. It is based on a distribution represented by a straight line in semi-logarithmic coordinates (rank in arithmetic scale; frequencies in logarithms). In the model based on semi-logarithmic scales, we mostly observe not a single alignment, but a succession of rectilinear segments, suggesting the coexistence of several "stands," useful for interpreting forest refuges. Generally speaking, the rarest species constitute a tail of distribution with rapidly decreasing numbers. However, 1% frequency threshold corresponding to 1 or 2 species-occurrence was chosen in this study; all species with ≤1% frequency were considered rare.

Species turnover analysis
To assess whether plant assemblages in the Togolese highland riparian forest show a spatial turnover pattern and to Country, geographical zone, refuge forest sites, area, annual rainfall, subsol, and taxa comparison between riparian forests and sites considered as forest refuges throughout tropics Africa region. Note: The total number of species recorded in our study is of the same order as those of sub-Saharan Africa forest refuges.
understand the cause of the processes underlying species richness, Baselga's (2010) beta diversity partitioning method was used. According to Baselga, beta diversity can reflect two different phenomena: nesting and spatial turnover. Nesting of species assemblages occurs when biotas at sites with few species are subsets of biotas at rich sites, reflecting a nonrandom species loss process as a result of any factor favoring orderly assemblage disaggregation. In contrast to nesting, spatial turnover involves the replacement of some species due to environmental screening or spatial and historical constraints. The method, based on dissimilarities between multiple sites, combines three classical multiplicative metric formulations: total beta diversity (bSOR), spatial turnover (bSIM), and nesting (bNES

| Representativeness of rainforest species and forest type analysis
The relative proportion of rainforest species in the riparian forests of the Togolese Mountains was assessed according to the procedure reported by Meave and Kellman (1994), who considered the species of not only the rainforest block but also the xeric environments surrounding the rainforest zone. In this study, these xeric species include the GC-SZ regional or forest-savanna transition species. To ensure that the species inventoried in the forests of the Togolese mountain riparian areas were from moist forests, they were compared with the list of moist forest species of West Africa (Holmgren et al., 2004). For ambiguous species, they were compared with the African plant database.
Forest types in vegetation can be identified based on the occurrence of typical tree species (Langenberger et al., 2006).
In sub-Saharan Africa, forest types and their characteristic species are known from previous studies (Akpagana, 1989;Hall & Swaine, 1976;White, 1983). To assess the forest types present in the riparian forests studied, the forest type classification of White (1983) and Hall and Swaine (1976), reported by Jongkind et al. (2006), was used. It should be noted that not all species in the general floristic list were systematically assessed according to the forest types to which they belong, only a few typical tree species were selected to show the presence of forest types in the riparian forests studied.

| Similarities between riparian vegetation and large sub-Saharan forest refuge analysis
To assess the similarity between the riparian vegetation studied and the large sub-Saharan forest refuges, the shared species were analyzed. This method was used because, in ecology, the number of shared species is a standard measure of similarity between two communities, and shared areas in species distribution are assumed to indicate a unique shared biological history (Lenormand et al., 2019). First, we checked whether riparian forests, and not the whole study area, were used as forest refugia.
Therefore, the floristic list of riparian forests was first compared with the floristic list of previous inventories for vegetation lists of non-riparian forests only, i.e., "Terre firme" forests, and vegetation matrix lists of whole forests. The non-riparian forest vegetation lists were derived from the "Terre firme" forest inventory (Adjossou, 2009;Akpagana, 1992), while the whole forest matrix vegetation lists were derived from the compilation of available floristic lists and herbarium data on the whole forest vegetation of the study area (Adjossou, 2004(Adjossou, , 2009Akpagana, 1989Akpagana, , 1992Brunel, 1977Brunel, , 1978Brunel et al., 1984). Second, the floristic list  Kokou (1998Kokou ( , 1999

| Floristic diversity and turnover
The results show an exceptional species richness in the riparian forests of the Togolese Highlands, amounting to 868 vascular species (Appendix S1) divided into 506 genera and 115 families.
Rubiaceae (88 species) and Fabaceae (87) (Figure 3), indicating a rich and diverse flora associated with high habitat heterogeneity. The rank-frequency curve shows that a significant number of species (31%) could be considered rare as they were observed once or twice, i.e., a frequency of around 1% (Figure 3).
Species distribution according to phytogeographical origin showed that taxa from the GC represents 60% of the listed species, of which 54% was common to the western and eastern Guinean-Congolese forest block, 5.30% was endemic to the western forest (GCW) including one endemic species from DG, and 0.70% was endemic to the Eastern forest (GCE). GC-SZ species also occupied a relatively high proportion (21%; Figure 4).  Table 3). The forests studied contain most of the characteristic species of the Guinean-Congolian Mixed Moist Forest, common to sub-Saharan forest refuges sites ( Table 4).
The results show a significant contribution of the studied forests to the current flora of large sub-Saharan forest refuges.
Indeed, The riparian vegetation studied shares 41% with the flora recorded in southwest Côte d'Ivoire, 76% with that of the humid forests of the study area, 100% with that of the entire matrix of the study area, 50% with that of the forest patches of the coastal plain of Togo, 41% with that of the riparian forests of Benin and 32.64% (dbh = 10 cm) with that of the humid forests of northern Congo (Appendix S1). The studied riparian vegetation shares 71% of its 868 species with that of the "terre firme" forests of the whole study area. This means that 29%, or 255, species are not common to both forests. These species are, mostly, those whose distribution is confined to the riparian corridor, probably due to historical events ( Table 5).

| Evidence based on species diversity and rarity
Our results showed that the species richness in the studied riparian forests is significantly high. The number of species listed in this study (868 species) is higher than those listed by Akpagana (1992) in the "terre firme" forest of the same area (648 species), and Kokou et al. (1999) in the forest patches of the whole coastal plain of DG in southern Togo (649 species). The total number of taxa listed in this study is rather of the same order as places considered to be very rich in sub-Saharan Africa ( Table 1). Very few data sets cover tropical riparian vegetation comprehensively and are therefore suitable for comparison. The only study using the same plot size was conducted by Natta (2003) in riparian forests of the Benin, Sudanese-Guinean zone, who analyzed 198 vascular plant plots (500 m 2 ), which TA B L E 2 The 25 most common plant families and genera recorded in Togo mountain riparian forest community according to the classification systems of Cronquist and APGIII. recorded 556 species. In this study, with the same number of identically sized plots, 828 species were recorded, thus much higher than the number found by Natta (2003). These results support the biodiversity analysis made by Wieringa and Poorter (2004)

for Upper
Guinea, which considered Togo plateau as an extension of the mountain high diversity belt.
The riparian forests studied showed a high incidence of rare species, and a high level of beta and gamma diversity associated with species turnover, which is the evidence of forest refugia according to ecologists (Cowling & Lombard, 2002). The observed high beta diversity is an indication of a heterogeneous and complex environment, which is also characteristic of forest refuges the high number of rare species (30%) observed. However, the high incidence of rare species observed is not specific to the forests studied and is similar to those reported in the rainforests of the Amazon basin of Venezuela (Uhl & Murphy, 1981), on the northern slopes of the Barva Volcano in Costa Rica (Lieberman et al., 1996), in the Western Ghats of India (Amarnath et al., 2003), in the Philippines (Langenberger et al., 2006), and in West Africa (Holmgren

| Evidence based on the presence of bioindicator species of rainforests refuges
In southwest Cameroon refugia, bio-indicators species of rainforest refugia such as slow-dispersing taxa (Begonia, Rinorea spp., etc.),

F I G U R E 2
Area-species curve of the Togolese mountain riparian vegetation. The area-species curve did not stabilize with the sampling effort indicating that not all species were identified during the surveys. Estimates (first and second order jackknife richness estimators: Jack1, Jack2) show that at least 101 and at most 165 species remain to be identified, thus increasing the species richness of the surveyed riparian forests to at least 969 and at most 1033 species.

F I G U R E 3
Rank-frequency curve of the Togolese mountain riparian vegetation. The curve is represented by a graph, where the ranks of the species are plotted on the x-axis in order of decreasing abundance and the relative frequencies are plotted on the y-axis in semi-logarithmic scale, which gives a more complete picture of floristic diversity than a simple diversity index. In the model based on semi-logarithmic scales, we observe not a single alignment but a succession of rectilinear segments, suggesting the coexistence of several "stands" or "habitats". The rarest species constitute a tail of distribution with rapidly decreasing numbers. However, 1% frequency threshold corresponding to 1 or 2 species-occurrence was chosen in this study; all species with ≤1% frequency were considered rare. The vertical line intercepting the distribution marks the 1% relative frequency threshold, at which species are considered relatively frequent. The diversity indices are projected onto the graph. The Beta.SOR and Beta.SIM indices measure beta diversity on a standard scale of zero to one.
African rainforest refugia (Tchouto et al., 2009). These evidences were also present in the forests studied. For example, the Rinorea species were collected in the present study. The Begonia species were not collected in the present study but were included in Akpagana (1992) (Rull, 2005).
The presence of small groups of fragile plants with low dispersal power in the studied forests supports the existence of refuges in Togo Highlands. According to the biogeographical hypotheses based on the diversification of the groups of small fragile plants with low dispersal power (Blan, 1993), these would have diversification and abundance formerly more marked. The current species would have then been relict and testified to long wet past periods during which these groups would have specialized. The current climate, equally humid and favorable to these fragile species, would have been preceded by dry climates, during which these groups would have undergone strong regressions and persisted only in refuges (mountains, waterfalls,). According to Blan (1993), the current presence of these plants at low altitudes in areas surrounded by mountains, as is the case in Cameroon, Congo, and Gabon, suggests that the current wet period has been long enough to allow these low dispersal species to recolonise the lowland forests.

| Evidence based on the representativeness of forest types and proportion of humid tropical species
Studied riparian comprise a typical element of many tropical forests types. Out of the 11 known forest types in the GC forest block, 10 are represented in the studied forests by their characteristic species. Only the ultra-wet evergreen forest is not represented in the riparian forests of the Togolese Mountains. Thus, we can say that almost all sub-Saharan tropical forest types were represented in the studied riparian forests by their typical tree species. Furthermore, our results show that 93.89% of the 835 native species listed in the studied riparian forest were typical of tropical rainforests. This percentage is slightly higher than that reported (85.3%) by Pither and Kellman (2002) in the fragments of riparian forests in savannas of the Mount Pine Ridge and agreement with the hypotheses on forest refugia studied in sub-humid tropical landscapes worldwide (Meave et al., 1991;Meave & Kellman, 1994;Pither & Kellman, 2002).
In addition, the studied riparian vegetation shares 71% of its 868 species with that of the "terre firme" forests of the whole study area. This means that 29%, or 255, species are not common to both forests. This is because many species are restricted to the riparian corridor (Table 5). Simultaneously, some species identified in the "mainland" forests were not found in the riparian forests. We do not believe that these species are restricted to the "mainland" forests, as information on the ecology of some of them shows that they could also inhabit the riparian forests. It is therefore possible that these species exist, particularly as estimates show that there are several unidentified species in the riparian forests surveyed. It is clear from these analyses that it was not the mountains of Togo as a whole that served as refuges for tropical rainforests, but it was the riparian forests from which the post-Pleistocene recolonization took place.

F I G U R E 4
Phytogeographical origin of the sub-humid Togolese mountain minor forest refuges flora: Cosmopolitan (Cosm), Pantropical (Pan), Paleotropical (Pal), Afro-American (AA), Afro-Malagasy (AM), Guinean-Congolese regional center of endemism (GC), Upper Guinean sub-center of endemism including Western Guinean-Congolese forest block (GCW), Lower Guinea and Congolese sub-center of endemism (GCE), Dahomey Gap (DG), Sudanese-zambezian regional center of endemism (SZ), Guineo-Congolian-Zambezi regional transition zone and Guineo-Congolian-Sudanese (GC-SZ), Introduced origine (I), Indeterminacy (Ind). Note: Out of the 11 known forest types in the GC forest block, 10 are represented in the studied forests by their characteristic species. Only the ultra-wet evergreen forest is not represented in the riparian forests of the Togolese Mountains. The ultra-humid evergreen forest type may not be represented in the riparian forests of the Togolese Mountains due to the influence of the DG drought. a Forest type adapted from Hawthorne and Jongkind (2006).

TA B L E 3
Eleven sub-Saharan African forest types and their typical species found in the riparian forests of the Togo Mountains.

| Evidence based on endemism
Our results showed that 60% (517 species) of the species studied have a GC endemic distribution pattern indicating very high endemism levels in the studied forests. These results are also consistent with the Pleistocene refuge model, which predicts that areas with a high number of endemic species probably served as refuges for tropical forest taxa during the expansion of savannas during the Ice Age in Europe, corresponding to the period of high heat in the tropics (Fiaschi & Pirani, 2009).

| Evidence based on affinity with vegetation and flora of forest refuges
The vegetation of the studied riparian forests have an affinity with those of the forest refuges in sub-Saharan Africa. White (1983) described a forest type in sub-Saharan Africa qualified as Guinean-Congolian mixed moist forest. It seems that this forest type is the characteristic forest of refuge sites. According to him, most Guinean-Congolian rainforests belong to this type. It occurs on well-drained soils throughout the Guinean-Congolian region except in the wettest and driest extremities. It is relatively undeveloped in West Africa due to the rapid increase in severe drought (It is easily understood that the ultra-humid evergreen forest type may not be Genera are necessarily older than species and families are older than genera. Consequently, the higher the taxonomic level on which the subdivision into biogeographical units is based, the further back in Earth's history the events that gave rise to them. Empires and regions would be characterized by families and even orders, and would thus be a copy of continental drift. The New Zealand flora is often cited as being of Gondwanan origin. Comparison of our results with those obtained in New Zealand rainforests (Jaffre et al., 1994) shows a perfect similarity between these two floras with regard to the representativeness of the families and genera ( Table 6). Fossil data reported by Craw et al. (1999), which has been cited by Craw et al. (1999), show that the Marattiaceae family dates back to the Carboniferous and the Cyatheaceae to the Cretaceous. These taxa are also present in the forests studied.
The Paleobotanical research has provided information on the nature of the various tropical vegetation patterns that developed through central and northern Africa from the Cretaceous to the end of the Tertiary and has brought to light several stages in the development of the rainforest (Elenga et al., 1991Maley, 1991Maley, , 1996. According to these sources, it is in the upper Eocene, about 45 million years ago, that the floristic composition of the south Cameroon vegetation really begins to resemble its present state. It is at this time that many taxa still living today appear. For instance; Bombax buonopozense, Pentaclethra, Symphonia globulfera, etc. The presence of these elements in the current flora indicated that the humid tropical flora would have been differentiated long before the Pleistocene era in the DG and around. We rejected any idea of seed transport of these species, especially by water, as the Togo Mountains are TA B L E 4 The characteristic tree species of the Guinean-Congolian Mixed Moist Forest, common to sub-Saharan forest refuges sites (White, 1983), also present in the studied forests relatively isolated from the rest of the rainforests and are at a higher altitude than the surrounding drier forests. These results are in line with geological studies that suggest that the mountains of Togo date from the Late Precambrian (Hall & Swaine, 1976). The mountains of Togo can therefore bear witness to very ancient historical events.
Indeed, the distribution of current forests and savannas in West and Central Africa is thought to be the legacy of the long-term history of climate and human impacts. Palaeo-environmental reconstructions suggest that West and Central African forests have experienced a succession of contraction and extension in response to dry and humid periods since the Last Glacial Maximum (LGM, ~21 cal ka BP; Bengo & Maley, 1991;Daniau et al., 2020;Elenga et al., 1991Elenga et al., , 1992Elenga et al., , 1994Giresse et al., 1991;Maley, 1992 (Maley & Brenac, 1998).
Polyscias fulva, currently found on Mount Cameroon and further north on the Cameroonian Ridge, are also present in the forests studied. Polyscias fulva lives in association with Parinari excelsa at high altitude in the Togo mountain. The occurrence of low dispersal elements in these forests gives also evidence of that humid phase. Sosef (1994Sosef ( , 2004 had noted the presence of seven species of Begonia occurring in the Monts Doudou Reserve, which are recognized as indicators of a Pleistocene forest refuge. This has been confirmed by phylogenetic data. Molecular dating of the African genus Begonia indicated that these taxa originated during the Pleistocene (Plana et al., 2004). Tree species of the genera Carapa (Meliaceae) and Coffea (Rubiaceae) originated during the Pleistocene (Brée et al., 2020;Koenen et al., 2015) were also found in the studied forests.

| Contribution of sub-Saharan minor forest refugia to larger sub-Saharan forest refugia
The forests studied would contribute about 60% of their rich flora to the large sub-Saharan forest refuges above mentioned. The riparian vegetation studied shares 41% of its 868 species with the flora recorded in southwest Côte d'Ivoire (Adou et al., 2005), which is located in the sector of the three major Pleistocene forest ref-  Kokou (1998Kokou ( , 1999 in the forest patches of the Togolese coastal plain and Natta (2003) in the riparian forests of Benin but are recorded by Kimpouni et al. (2013)

| Implications for sub-Saharan Tropical African biogeography and biodiversity conservation
The origin of the DG is one of the central biogeographical questions in sub-Saharan Africa, which has been addressed in recent phylogenetic studies. Till date, no study based on large floristic data collected in the DG has addressed this question. Our work, based on large floristic data collected in the DG and compared with those of other studies suggests that the riparian forests of the sub-humid Togolese Mountains would have served as important minor forest refugia during recurrent climatic episodes.
Indeed, the Cross River biogeographic break, like other smaller breaks identified in other places is thought to be the result of past fragmentation of the forest block as a result of climatic changes. It appears, therefore, that the major breaks in the sub-Saharan forest block must have been repeated many times with relative geographical constancy since the late Pliocene (Maley, 1991;White, 1979).
Our results are in line with these conclusions and rejected the hypothesis of a single opening of the DG. The riparian forests of the Togolese mountains would have served as refuges during recurrent climatic crises during their history and would continue to play this role in future.
Our findings also seem to confirm the hypotheses that the riparian forests played the role of refuge for flora during the arid period of Pleistocene in sub-humid tropical landscapes covered with savannas or deciduous forests. During this severe drought period, the riparian zones would have been the only forest places maintaining the conditions necessary for most species, and the current flora would have been reconstituted and differentiated from these zones when moisture conditions became favorable for a forest cover. This process could explain why the flora of the tropical riparian forests often contains a significant part of the flora of the neighboring forests (Meave et al., 1991;Meave & Kellman, 1994). These results therefore support the hypothesis that the forest fragmentation was minimal or took the form of a network of refugia (along rivers and/or in mountains) where most species survived.
The few phylogeographic studies conducted in the DG have suggested that the forest flora of the DG may be essentially relics from the early Holocene, when the Guinean-Congolian forest reached its maximum geographical distribution (Demenou et al., 2016;Duminil et al., 2015;Koffi et al., 2011). Our floristic data do not seem to be compatible with these ideas. Prior to the last opening of DG, the differentiation of sub-Saharan rainforests would have taken place under exceptionally wet conditions, allowing their maximum development.
These conditions would probably coincide with the wet periods of the Late Pleistocene and Holocene (Salanville, 1992). We hypothesize that during this period of high humidity, the Rubiaceae family would be more differentiated than the Fabaceae family, since Rubiaceae would be more prevalent in the wetter forests (southwestern Côte d'Ivoire) than in the relatively drier ones (DG) where Fabaceae would be strongly represented. But the period of severe aridity following the last wet period would have impoverished DG in ancient forest species, which would be typical Guinean-Congolese species. Moreover, it is widely accepted that evolutionary changes would have occurred in isolated populations in forest refugia. Models of Fabaceae and Gramineae studied in North Africa also argue that isolated populations would favor the emergence of genotypes adapted to environmental conditions (Amirouche & Michet, 2009). Speciation would have therefore contributed to the creation of hybrid individuals adapted to the relatively precarious climatic and soil conditions in forest refuges.
Fabaceae and Gramineae would have been better adapted to the severe environmental conditions and served as pioneer and transitional species during the recolonization phase around forest refuges. They continue to play this role today in environments hostile to GC species.
Our results seem to indicate that, in the DG, there are more adapted ecotypes in current transitional forests than relict species of ancient forests. This is the case of Millettia thonningii (Schum. & Thonn.) Bak. Historically, the biogeographical theory suggested that large reserves are good for biodiversity conservation. In the real world of conservation planning, the opportunity to apply such guidelines is constrained by costs and patterns of land use history (Margules & Pressey, 2000). According to Margules & Pressey, if the area available for reservation is limited, a choice might be made between a few large reserves that favor the persistence of some species or many small reserves that together are more representative of the region's biodiversity but individually less effective for the persistence of some species.
In sub-Saharan tropical Africa, where maintaining a vast area of natural forest is difficult due to human pressure, efforts to preserve maximum species diversity should include a focus on the conservation of minor forest refuges, particularly in sub-humid mountain riparian zone. In this scenario, the findings of Bohoussou et al. (2015) constitute beneficial for managing biodiversity in sub-Saharan tropical Africa.

| CON CLUS ION
This study was undertaken with the aim of contributing to scientific debates on tropical biogeography and its importance for biodiversity conservation in the context of climate change. Our work, based on floristic evidence compared to those from palaeobotanical, palynological, phylogenetic, and geological studies, suggests that the mountains of Togo and their riparian forests would have very ancient origins. They would therefore have experienced paleoclimatic events as evidenced by the presence in these forests of paleoflora elements dating from the Cretaceous, Pleistocene, and Holocene.
The study rejects the idea of seed transport of these species, especially by water, as the mountains of Togo are relatively isolated from the rest of the humid forest blocks and are at a higher altitude than the surrounding drier forests.
According to ecologists the major breaks in the sub-Saharan forest block must have been repeated many times with relative geo- The phylogeographic studies conducted in the DG have suggested that the forest flora of the DG may be essentially relics from the early Holocene, when the Guinean-Congolian forest reached its maximum geographical distribution. Our floristic data do not seem to be compatible with these ideas.
In sub-Saharan tropical Africa, where maintaining a vast area of natural forest is difficult due to human pressure, efforts to preserve maximum species diversity should include a focus on the conservation of minor forest refuges, particularly in sub-humid mountain riparian zone.

AUTH O R CO NTR I B UTI O N S
Kossi Adjossou involved in conceptualization (lead), data curation

ACK N OWLED G M ENTS
This study was sponsored by several grants awarded to Dr. Adjossou

CO N FLI C T O F I NTE R E S T
The authors declare no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data supporting the results of this study will be archived in the DRAYD data directory (URL: https://doi.org/10.5061/dryad.w6m90 5qsf).