What explains the distribution of rare and endemic West African plants

Introduction

Understanding species distribution patterns is a crucial step for the conservation and management of plant communities and ecosystems. This is particularly true for species with small distribution ranges that grow and reproduce under a very particular set of environmental conditions. West African forests, although less species rich than some other tropical forests, seem to be relatively rich in endemic species (e.g. Myers et al. 2000, White 2001). About 22% of the forest species in Upper Guinea are restricted to this region (Jongkind & Wieringa Chapter 11). Because of the rapid rate at which these forests are disappearing (e.g. Chatelain et al.chapter 2), it is urgent to know the characteristics of the species distribution and the mechanisms responsible for them.

In Upper Guinea distinctive forest types occur along a west-east coastal gradient from Guinea to Togo. There is also a clear zonation of forest types from the coastal forests towards the northern savanna (Hall & Swaine 1976, Martin 1991, Bongers et al. chapter 4). The change in forest types along both gradients has been explained mainly by a strong rainfall gradient (Hall & Swaine 1976, van Rompaey 1993). At the species level, we also know how some forest species are distributed along environmental gradients. In most cases, species occurrence or density has been found to be significantly correlated with water availability (Bongers et al. 1999), although soil fertility (Swaine 1996) or, at a smaller spatial scale, light availability (Veenendaal et al. 1996c, Agyeman et al. 1999a), are sometimes better explanatory variables.

The interpretation of species distribution patterns along environmental gradients is difficult. This is partly because environmental gradients tend to covary. For example, it is hard to separate the effects of changes in rainfall and soil fertility because under high rainfall conditions, nutrient leaching causes a loss of soil fertility as well (e.g. Swaine 1996). Rainfall effects can be also complicated by the interaction with altitudinal and topographical gradients. For example, tree species largely confined to high rainfall forests can be found in forests with lower rainfall, but are restricted to the lower and therefore wetter, topographical places (Guillaumet & Adjanohoun 1971, de Rouw 1991, van Rompaey 1993). Field experiments can play a crucial role in separating the relative importance of different environmental variables. In

Figure 7.1 Cola attiensis, a small understorey tree species with a disjunct distribution between Upper and Lower Guinea.

general, they support the predominant role of seasonal drought over soil fertility to explain the limits of species distribution in these forests (e.g. Veenendaal & Swaine 1998). Clearly, the distribution of any species along these environmental gradients will be partly the result of its ecophysiological tolerances combined with the effects exerted by the interaction with co-occurring species.

The actual distribution of species along present environmental gradients may also reflect historical changes on longer time scales. The Upper Guinea region has been exposed to strong climatic changes and severe human-induced disturbances (Fairhead & Leach 1998). Rainforest cover has periodically retracted during glacial periods, when the climate was drier and cooler, and expanded during warmer and wetter interglacial episodes, to cover surfaces even greater than the present (Livingstone 1982). In this region, glacial-interglacial climatic changes were probably stronger than in any other tropical forest around the world because of the relatively low rainfall here compared to the other continents (Richards 1973, Morley 2000, Maley 2001, White 2001). Furthermore, these West African forests have been exposed to a longer, and recently faster, rate of deforestation and degradation than other rainforests around the world (FAO-UNEP 1981, FAO 2001). These past forces must have left a deep imprint on the composition and structure of these West African forests. Climate change and deforestation have produced not only a change in the general climatic and edaphic conditions for which the ecophysiological tolerances of the species might differ. But they have also increased the fragmentation of the landscape and we could expect differences in dispersal capabilities between species to play a fundamental role in explaining the extension of their distribution range. In summary, the present distribution of a particular species would be the combined result of ecophysiological tolerances, the opportunities to disperse, the interactions with other species, and the effects of past changes in climate and disturbance.

Most of our knowledge on the distribution of plant species in the Upper Guinea forests comes from the outstanding work performed in Ghana (e.g. Hall & Swaine 1976, Hawthorne 1996, Swaine et al. 1997), and to a lesser extent from other countries of the region (e.g. Guillaumet 1967, Schnell 1977, Ake Assi 1984, van Rompaey 1993). The only work covering the whole Upper Guinea region is the Flora of West Tropical Africa (Keay 1954-1963), which has a strong taxonomical focus. In general, most of the emphasis has been on taxonomical descriptions or studies confined to one country. A renewed regional perspective can bring new insights into the amplitude of the species distributions and the mechanisms behind them. In this chapter, we describe the spatial distribution patterns and discuss the potential responsible mechanisms for a large set of endemic and rare plant species in the Upper Guinea region. We use herbarium collections from the whole region and relate the inferred distribution patterns to the main environmental variables and species life-history traits.

Methods

Species selection

Approximately 2800 species of vascular plants have been recorded in the forests of Upper Guinea, 22% of which are endemic to this region (Jongkind, Chapter 11). As a first step we made a selection of 1000 endemic and rare species of these forests based on the second edition of the Flora ofWest Tropical Africa (Keay 1954-1963), forest inventories, and taxonomic revisions. The emphasis was on woody species (trees, shrubs and lianas), although some herbs were included as well. From this selection, we prepared a short-list of 600 species that have been taxonomically well described and for which there is enough certainty about their correct identification in the different herbaria. From this short-list we selected a subset of 286 species that are individually described in chapter 9, and for which the distribution patterns are discussed here. In selecting this sample we aimed at preserving as much taxonomical diversity as possible. This means that we selected species from all families taking a smaller fraction of species from the large families (with many species in this area). The 286 species of this final subset represent c. 10% of the total forest flora of Upper Guinea (with 223 species considered endemic to Upper Guinea). They belong to 75 different families. Amongst these species, there are 126 trees, 35 shrubs, 79 lianas, 36 herbs, nine epiphytes, and one saprophyte. This means that our sample is relatively rich in trees, since in the Upper Guinean flora the number of trees, shrubs and lianas is rather similar (W.D. Hawthorne, personal communication).

Species data

We entered into a database (Brahms 4.8, Filer 2000) all herbarium specimens collected from Senegal to Togo, for all 600 short-list species (including our 286 species). For some non-endemic rare species, we also included the herbarium specimens collected in Lower Guinea (the Congo Region). We checked and entered all specimens from Herbarium Vadense (Wageningen, The Netherlands), National Botanical Garden of Belgium (Meise, Belgium) and Kew Botanical Garden (Kew, Great Britain). In addition, we added the collections from Côte d'Ivoire present in the Geneva herbarium database (Conservatoire et Jardin Botaniques de la Ville de Genève, Switzerland), the collections from Côte d'Ivoire and Guinea present in the Paris herbarium database (Muséum National d'Histoire Naturelle, Paris, France), and the collections of 100 species present in the Herbarium of the University of Ghana (Legon, Ghana). These data were complemented by distribution data from taxonomic revisions. In total the database contained 48,000 records from Upper Guinea, of which about 6900 records concern our 286 rare or endemic species. This database is not exhaustive, but gives a fair indication of the state of knowledge for the species in the region.

Environmental data

Species distributions are usually strongly shaped by water availability, soil fertility, and altitude. Maps of these environmental variables were prepared and included in a Geographical Information System (ArcView, ESRI Inc.).

Water availability is determined by rainfall, soil water holding capacity, and the presence of rivers, lakes, and proximity to the coast. Annual rainfall data were compiled from different sources (Myers & Staff 1981, ANAM 1987, ICCARE 1994, Global Historical Climatology Network). In total, 578 weather stations were included from Senegal to Nigeria, Burkina Faso, Mali and Niger. From these data a rainfall map was made, using the inverse distance weighting interpolation method in ArcView. The soil water holding capacity was derived from the FAO Digital Soil Map of the World (FAO 1995b). The soil map has a scale of 1 : 5,000,000. Therefore small-scale differences in soil types due to topographical variation are not visible on this map. For each soil type, we calculated the water holding capacity of the soil (in mm water/m soil) based on the depth and the texture of the dominant soil type. The position of perennial rivers, lakes, and coastline was obtained from the Digital Chart of the World (ESRI 1984).

As a measure of soil fertility we used the sum of exchangeable cations (CMK; Ca2+, Mg2+, and K+, in cmolc/kg). For each unit of the FAO Digital Soil Map we used a median CMK value of the dominant soil type. The CMK values were based on a review of the chemical properties of soil profiles by Batjes (1997). Elevation data were derived from the Digital Chart of the World (US Geological Service, EROS Data Center, 1984). The elevation map has a spatial resolution of approximately 1 km.

In addition, we calculated the percentage of records of each species found in open habitats and in moist habitats. Using the herbarium collector's notes that had been added to our database, we searched for keywords for open places (e.g. gap, forest edge, roadside, etc.) and moist places (riverside, close to lake, stream, etc.). To estimate the proportions in particular habitats we used only the subset of records for which clear habitat descriptions were present.

Individual species occurrence along environmental gradients

A logistic regression analysis was carried out to model the species occurrence in Upper Guinea as a function of four environmental factors; altitude, annual rainfall, soil water holding capacity, and available cations. We used stepwise forward regressions with the simple and quadratic variables, but without including interaction terms. A significant simple variable indicates an increasing or decreasing probability to find the species with an increase in the environmental variable. A significant quadratic variable indicates that the species shows a bell-shaped response curve towards that environmental variable (Jongman et al. 1987). Since the species had not been collected at random and since we had only presence data (herbaria records), we needed to assume absence under certain conditions. For this we selected for Upper Guinea only those half-degree grid cells with 50 or more records of all species occurring in the area (thus including only cells that have been botanically well explored). This resulted in a total of 145 cells. For some cells there were no data for all four environmental variables. This reduced the total number of cells to 140. We assumed that a species was absent when no records of this species were found in those cells. A logistic regression analysis has only sufficient resolution to detect significant patterns, when there are sufficient numbers of presence and absence. In order to have a reasonable number of cells with presence, we ran the analysis only on those species present in ten or more cells (i.e. > 7% of the total amount of cells). This resulted in a total of 112 species.

Species distribution, range, and commonness along environmental gradients

The species were classified into different distribution types based on the continuity of their distribution, their range, and their commonness. We distinguished between species having a continental disjunct distribution (disjunct between Upper and Lower Guinea), a regional disjunct distribution (disjunct populations within Upper Guinea), and a continuous distribution (a more or less continuous occurrence within its range). The classification was based on a visual interpretation of the species distribution map, and a histogram of multiple neighbour distances of the species. To this end, we calculated for every species the distance between each collection point and all the other collection points of that species. For a species with five observations this would result in ten neighbour distances.

A histogram of the frequency of these neighbour distances showing a unimodal distribution was interpreted as an indicator for a continuous distribution (Fig. 7.2A). On the other hand, species with a bi-modal histogram were interpreted as having a disjunct distribution (Fig. 7.2B,C). Every mode or peak represents a clump, with the height corresponding to the number of collections within a clump. The distances between the peaks correspond to the distances between the clumps. Obviously, continental disjunct species (e.g. with a population in Liberia and a population in Gabon) usually have a much larger distance between the two peaks than regional disjuncts (e.g. with a population in Liberia and a population in Ghana). The classification of spatial patterns based on the multiple neighbour distance histograms was in close agreement with a visual interpretation of the maps.

For species with a continuous distribution, we estimated the species range within Upper Guinea based on the maximum distance found between collection points. We classified the species range as being very local (maximum distance < 100 km), local (100-300 km), regional (300-600 km), or widespread (>600 km). Our very local and local categories fit the most widely used criterion of tropical plant species endemism (Gentry 1992).

The commonness of a species in Upper Guinea was expressed in two different ways. The abundance is the total number of records found in the region from Senegal to Togo. This measure is potentially influenced by collector's bias. Rare species are likely to be over-represented, as they might be more interesting to collect by botanists. A more robust measure of commonness is the frequency of half-degree grid cells in which the species is collected. If a species had been collected many times in the same forest reserve close to a capital or research station, this would result only in one observation of a half-degree grid cell in which it is present. In our dataset, both measures are highly correlated (Pearson's r = 0.92, p < 0.001, n = 286) and therefore provide consistent estimates of a species commonness. For subsequent analysis of commonness, we only present the results for the total number of records.

In order to find general patterns among species with different distribution types, ranges, and commonness we performed a series of statistical analyses and selected different species subsets of the database depending on the questions asked. We compared both species distribution

Figure 7.2 Types of species distribution patterns. A) Continuous distribution (Combretum grandiflorum), B) continental disjunct distribution (Gymnosiphon longistylus), C) Upper Guinea disjunct distribution (Acanthus guineensis). Distribution maps indicate the locations of the species collections (dots), the potential forest cover (grey), areas above 500 m altitude (dark grey), and country boundaries. Next to the map is the histogram of distances between each collection and all other collections for that particular species.

Figure 7.2 Types of species distribution patterns. A) Continuous distribution (Combretum grandiflorum), B) continental disjunct distribution (Gymnosiphon longistylus), C) Upper Guinea disjunct distribution (Acanthus guineensis). Distribution maps indicate the locations of the species collections (dots), the potential forest cover (grey), areas above 500 m altitude (dark grey), and country boundaries. Next to the map is the histogram of distances between each collection and all other collections for that particular species.

types (i.e. continuous versus disjunct) for several environmental variables. No distinction was made between Upper Guinea and continental disjuncts in order to increase the sample size for the disjunct group. The variables for each species belonging to either distribution group were: rainfall (minimum, maximum, median, range,

10-percentile), altitude (% of records of a species found higher than 500 m), habitat openness (% of records found in disturbed open places), habitat moisture (% of records found in moist places). For each variable, we performed a t-test to compare the mean between both distribution groups. We used two criteria to select the species to be used in the comparisons. For the rainfall and altitude data we used species with more than five records in the database. For the habitat openness and moisture, we used all species for which we had more than ten records in the database with enough habitat information provided by the specimen collector and calculated the percentage of records in either disturbed open places or moist places.

For all subsequent analyses, relating a species range and commonness, we confined ourselves to species that have a continuous distribution, that are endemic to Upper Guinea (i.e. only occurring between Senegal and Togo), and that have more than five records in the database (n = 163). We used a multiple regression to relate a species range (maximum distance between collections) with the various environmental variables. We included only the environmental variables that were not strongly correlated with each other, namely minimum rainfall, maximum rainfall, altitude, percentage of records in open places, and percentage of records in moist places. To select the variables in the final model, we used both stepwise forward and backward selection. In both cases we obtained the same results. In addition, we used one-way Anovas to compare the four distribution range categories (widespread, regional, local, and very local) in relation to each environmental variable.

We related species commonness (a species total number of records, and the number of half-degree cells in which the species has been sampled) with the environmental variables using multiple regression analysis. We used the same environmental variables and data selection criteria as for the range analysis.

Distribution type, range and commonness in relation to a species life history traits

Based on the literature, we classified the species according to life form (herbs, trees, shrubs, lianas, and epiphytes). The main dispersal mechanisms were: animal, wind, water, explosive (active expulsion of the seeds), and barochore (simply falling). We also classified the species into guilds as pioneer, non-pioneer, and shade bearers based on the reported shade tolerance in Ghana (Hall & Swaine 1981, Hawthorne 1995a). Pioneers are species with the highest light demand, of which seedlings are only found in gaps and older plants are absent or very rare in the forest understorey. Non-pioneer light demanders are species with an intermediate light demand, of which seedlings are common in the understorey whereas adults are not (i.e. they require a gap to grow). Shade bearers are those species for which both young and older plants are frequently found in the shaded forest understorey. We compared the percentage of records found in open places in our database with the classification made by Hawthorne (1995a). Indeed, we found that pioneer species had a higher abundance in open places, but we did not find a difference between shade bearers and light-demanding species (t-test, t = -2.37, p = 0.026). Based on this, we classified the species lacking a literature description on shade tolerance using the proportion of herbarium records found in open places. We classified them as pioneers (> 65% records in open places), non-pioneer light demanding species (30-65% records in open places), shade bearers (< 30% in open places).

We related a species' life history traits (life form, dispersal mechanism, and shade tolerance) with the occurrence in particular habitats (e.g. % of records in moist places) using one-way Anovas. The effect of these life history traits on the distribution type was done through Chi2 frequency analysis, and the relationship with the range and the abundance was done through one-way Anovas.

Results

Species distribution types

Three different types of species distributions were recognised (Fig. 7.2); continuous, continental disjunct, and Upper Guinea disjunct. Combretum grandiflorum is an example of a species with a continuous distribution; it has a regular occurrence all over its range, and a neighbour histogram without clear peaks (Fig. 7.2A). Gymnosiphon longistylus is an example of a species with a continental disjunct distribution. It occurs in Upper Guinea and in Cameroon-Gabon, and has a neighbour histogram with two clear peaks, spaced 2000 km apart (Fig. 7.2B). Acanthus guineensis is a species with an Upper Guinea disjunct distribution. It occurs around Liberia, and in Ghana, and does not occur in between. The neighbour distance histogram shows two clear peaks, which are spaced approximately 1200 km apart (Fig. 7.2C). From a total of 270 species, 86% showed a continuous distribution, 10% a continental disjunct distribution, and 3% an Upper Guinea disjunct distribution. For 16 species, there were not enough data points (< 5 collections in Upper Guinea and a neighbour distance greater than 100 km) to classify their distribution.

There were four distinctive patterns among the 28 species with a continental disjunct distribution (sensu White 1979, Table 7.1). The most frequent group of "Guinea wide" (eleven species) occurred in both Upper and Lower Guinea (Fig. 7.3C). A second group of "near endemics" (nine species) occurred mostly in Upper Guinea and very scarcely in Lower Guinea, (from the Bight of Biafra to Gabon) (Fig. 7.3D). A third group of "satellites" (three species) showed the opposite pattern, with very few records in Upper Guinea (concentrated at the borders of Côte d'Ivoire with Liberia and Ghana) and a high concentration in Lower Guinea (Fig. 7.3E). Finally, a fourth group of "Guineo-Congolians" (five species) occurred in all three subcentres of endemisms (Upper Guinea, Lower Guinea, and the Congolian Region) (Fig. 7.3B).

Nine species had an Upper Guinea disjunct distribution (Table 7.1, Fig 7.3A). Most of them have a population at the border between Côte d'Ivoire and Ghana, and a second population at the border between Côte d'Ivoire and Liberia (Cassipourea hiotou) or further to the west (Monocyclanthus vignei, Pierreodendron kerstingii, Schumanniophyton problematicum, Strephonema pseudocola). Both populations, especially the one between Côte d'Ivoire and Ghana, are concentrated close to the coast. Two other species (Acanthus guineensis and Memecylon aylmeri) show a comparable pattern although their populations tend to be more inland. There are two species with a completely different distribution; Uapaca chevalieri has a disjunct distribution in Liberia and Sierra Leone that seems to be related to montane habitats, and Anisophyllea laurina has two isolated populations in Sierra Leone and Guinea-Bissau.

Species distribution range

We calculated the distribution range for all those endemic Upper Guinean species with a continuous distribution type and more than five collections in the database (n = 163). Species differed considerably in their distribution range, varying from widespread to very local

Table 7.1 Species or varieties with disjunct distribution patterns. Life form (H=herb, S=shrub, T=tree, WC=woody climber, SH=saprophytic herb), Distribution (UGD= Upper Guinea Disjunct, CD= Continental Disjunct), country of occurrence (S=Senegal, GB=Guinea Bissau, Gu=Guinea, SL= Sierra Leone, L=Liberia, CdI=Cote d'Ivoire, Gh=Ghana, T=Togo) and distribution pattern are indicated.

Table 7.1 Species or varieties with disjunct distribution patterns. Life form (H=herb, S=shrub, T=tree, WC=woody climber, SH=saprophytic herb), Distribution (UGD= Upper Guinea Disjunct, CD= Continental Disjunct), country of occurrence (S=Senegal, GB=Guinea Bissau, Gu=Guinea, SL= Sierra Leone, L=Liberia, CdI=Cote d'Ivoire, Gh=Ghana, T=Togo) and distribution pattern are indicated.

Species

Family

LF

Distrib.

Countries

Pattern

S

GB

Gu

SL

L

CdI

Gh

T

Acanthus guineensis

Acanthaceae

H

UGD

+

+

+

+

+

submontane

Anisophyllea laurina

Anisophylleaceae

T

UGD

+

+

+

unclear

Cassipourea hiotou

Rhizophoraceae

T

UGD

+

+

coastal

Memecylon aylmeri

Melastomataceae

S

UGD

+

+

+

+

+

inland

Monocyclanthus vignei

Annonaceae

T

UGD

+

+

coastal

Pierreodendron kerstingii

Simaroubaceae

T

UGD

+

+

+

+

unclear

Schumanniophyton problematicum

Rubiaceae

T

UGD

+

+

+

coastal

Strephonema pseudocola

Combretaceae

T

UGD

+

+

+

+

inland

Uapaca chevalieri

Euphorbiaceae

T

UGD

+

+

+

+

montane

Englerina gabonensis

Loranthaceae

S

CD

+

+

Guinea congolian wide

Guaduella macrostachys

Gramineae

H

CD

+

Guinea congolian wide

Illigera vespertilio

Hernandiaceae

WC

CD

+

+

+

+

Guinea congolian wide

Lasiodiscus mannii

Rhamnaceae

S

CD

+

+

Guinea congolian wide

Vernonia titanophylla

Compositae

T

CD

+

+

+

+

Guinea congolian wide

Anisophyllea meniaudii

Rhizophoraceae

T

CD

+

+

+

+

Guinea wide

Calvoa monticola

Melastomataceae

H

CD

+

+

+

+

Guinea wide

Chytranthus cauliflorus

Sapindaceae

T

CD

+

+

+

Guinea wide

Cola attiensis

Sterculiaceae

T

CD

+

Guinea wide

Dorstenia turbinata

Moraceae

S

CD

+

+

+

+

Guinea wide

Guaduella oblonga

Gramineae

H

CD

+

+

+

+

Guinea wide

Gymnosiphon longistylus

Burmanniaceae

SH

CD

+

+

+

+

+

Guinea wide

Magnistipula zenkeri

Chrysobalanaceae

T

CD

+

+

+

+

Guinea wide

Manotes macrantha

Connaraceae

WC

CD

+

+

Guinea wide

Mapania rhynchocarpa

Cyperaceae

H

CD

+

+

+

+

Guinea wide

Puelia olyriformis

Gramineae

H

CD

+

+

+

+

Guinea wide

Cassipourea afzelii

Rhizophoraceae

S

CD

+

+

+

+

+

near endemic

Combretum bipindense

Combretaceae

WC

CD

+

+

+

near endemic

Dracaena ovata

Liliaceae

S

CD

+

+

+

+

near endemic

Euadenia eminens

Capparaceae

T

CD

+

+

+

+

near endemic

Hemandradenia chevalieri

Connaraceae

T

CD

+

near endemic

Marattia odontosora

Marattiaceae

H

CD

+

+

near endemic

Okoubaka aubrevillei

Santalaceae

T

CD

+

+

+

+

near endemic

Pyrenacantha glabrescens

Icacinaceae

WC

CD

+

+

+

near endemic

Tarenna vignei var. subglabra

Rubiaceae

S

CD

+

+

+

+

near endemic

Begonia hirsutula

Begoniaceae

H

CD

+

satellite

Begonia mildbraedii

Begoniaceae

H

CD

+

+

satellite

Hymenocoleus axillaris

Rubiaceae

S

CD

+

satellite

I f f

f -

* 1 % «

«

E

Figure 7.3 Patterns of Upper Guinea disjunct (A) and continental disjunct (B-E) distributions.

A) Upper Guinea disjunct (9 species),

B) Guineo-Congolian (5 species) with individuals found in all three subcentres of endemism,

C) Guinea wide (11 species),

D) near endemic (9 species) with the largest population in Upper Guinea,

E) satellite (3 species) with the largest population in Lower Guinea. Distribution maps indicate the locations of species collections (dots), the potential forest cover (grey) and country boundaries.

Figure 7.3 Patterns of Upper Guinea disjunct (A) and continental disjunct (B-E) distributions.

A) Upper Guinea disjunct (9 species),

B) Guineo-Congolian (5 species) with individuals found in all three subcentres of endemism,

C) Guinea wide (11 species),

D) near endemic (9 species) with the largest population in Upper Guinea,

E) satellite (3 species) with the largest population in Lower Guinea. Distribution maps indicate the locations of species collections (dots), the potential forest cover (grey) and country boundaries.

(Fig. 7.4): 67% of the species had a widespread distribution (range > 600 km), 14% of the species had a regional distribution (range 300-600 km), 14% a local distribution (range 100-300 km), and 5% had a very local range (< 100 km, Table 7.2). In addition there were 13 other species with a very local range (< 100 km) but with only two to five records in the database. An extreme example of a highly endemic subspecies with a very small range is Impatiens nzoana ssp. nzoana of which all five known botanical collections were found within a range of 7 km on Mount Nimba (Table 7.2). Most species with a continuous distribution had a range between 800 and 1600 km (Fig. 7.5).

Species commonness and rarity

Twenty-six species were collected only once or twice and can be considered extremely rare (Table 7.3). Most of these species are found in Liberia (eight species), Côte d'Ivoire (seven species), or both countries (two species). The rare Liberian species occur often in three locations; around Monrovia and Bomi Hills (e.g. Ancistrocladus pachyrrachis), southern Greensville (e.g. Begonia fusicarpa), and in the mountains (e.g. Sericanthe adamii). The rare Ivorian species are frequently found in two locations, near the border with Liberia (e.g. Clerodendrum sassandrense) and around Abidjan (e.g. Argocoffeopsis lemblinii).

Figure 7.4 Types of distribution ranges. A) Widespread (Berlinia tomen-tella), B) regional (Aphanocalyx microphyllus ssp. compactus), C) local (Begonia quadrialata ssp. nimbaensis). Distribution maps indicate the locations of species collections (dots), the potential forest cover (grey), areas above 500 m altitude (dark grey) and country boundaries.

Figure 7.4 Types of distribution ranges. A) Widespread (Berlinia tomen-tella), B) regional (Aphanocalyx microphyllus ssp. compactus), C) local (Begonia quadrialata ssp. nimbaensis). Distribution maps indicate the locations of species collections (dots), the potential forest cover (grey), areas above 500 m altitude (dark grey) and country boundaries.

Individual species occurrence in relation to the environment Of the 112 species for which we had enough data to run the logistic regressions, we found significant relationships with the environment for 88 species. The variation explained by the models was moderate to low

(average r2 = 0.27, range 0.05-0.63). The species occurrence could be explained by a single environmental factor (57 cases), or by a combination of factors (31 cases). Rainfall was the most important environmental factor (significant for 71% of the species) followed by altitude (36%), water holding capacity (23%) and available cations (14%). The combination of significant factors found most frequently were rainfall and altitude (16 cases), and water holding capacity and altitude (11 cases).

Distribution type, range and commonness in relation to the environment

We compared the environmental variables between species with continuous versus disjunct distributions. We found that species with disjunct populations had a lower number of records in open places than species with continuous distribution and thus tended to be found in relatively undisturbed habitats (Fig. 7.6, t-test, p = 0.001, t = 4.0, nc = 97, nd = 10). We did not find any significant relationship with rainfall (all indicators), altitude, or habitat moisture.

For species with a continuous distribution, we compared the environmental conditions among different distribution range categories (widespread, regional, local, and very local), and found significant differences for all rainfall variables (minimum, maximum, mean, range, and 10-percentile) (one-way Anovas, n = 198, p = 0.001). We did not find significant differences between species range categories for altitude, habitat openness and habitat moisture. We performed a multiple regression analysis to compare the relative contribution of all environmental variables on a species' range. We found that a species' range was positively correlated with the maximum rainfall and negatively correlated with the minimum rainfall where the species occurs (multiple regression, regression coefficient b maximum for rainfall = 0.4, p < 0.001; b minimum rainfall = - 0.74, p < 0.001; n = 163; r2 = 0.66). Thus species with larger distribution ranges where found in a larger amplitude of rainfall conditions, and were found also in drier places. We found the same pattern among all life forms (i.e. trees, shrubs, herbs, lianas). The relationship between species range and either minimum rainfall or maximum rainfall did not vary among life forms (Anovas, no significant interaction between life form and rainfall covariate) (Fig. 7.7).

We found significant correlations between a species' total number of records and all rainfall variables (Pearson's r, p < 0.001 in all cases), and also with the percentage of records found in open places (Pearson's r, p = 0.01). When we tested the relative contribution of the environmental variables, we found comparable results as with species range, namely that a species frequency was positively related to the maximum rainfall and negatively related to minimum rainfall (multiple regression, b maximum rainfall = 0.018, p < 0.001; b minimum rainfall = - 0.027, p < 0.001; n = 163; r2 = 0.37). This relationship was consistent among different life forms.

Discussion

Species distribution: the role of environmental factors

Water availability is probably the most important factor explaining the distribution of individual plant species in West Africa. We found that the probability of occurrence of 71% of these endemic forest species in Upper Guinea was related to annual rainfall. These results are in agreement with the findings at smaller scales in the region. For example, in Liberia and Côte d'Ivoire Bongers et al. (1999) found that the occurrence and abundance of 10 of their 12 selected canopy tree species could be explained fairly well by water availability. Also field studies and experiments in Ghana, comparing the effects of rainfall and soil fertility, have found that water availability is the most limiting factor in the distribution of tree species, although soil fertility might be important for some species (Swaine 1996, Veenendaal & Swaine 1998). Not surprisingly, the distribution of forest types and species richness has been explained mainly by changes in rainfall conditions (see Bongers et al. chapter 4, Kouamé et al. chapter 5, Wieringa & Poorter chapter 6). Beyond rainfall, other environmental factors can play a role for particular species. For example, we found that an altitude effect was significant in 36% of the species tested. This probably

Table 7.2 Highly endemic species or varieties with a continuous distribution pattern, and a distribution range smaller than 100 km. The life form and country of occurrence are indicated. The species are ordered based on their maximum distribution range. For abbreviations see Table 7.1.

Species

Family

LF

Countries

Range (km)

Abundance (#)

Gu

SL

L

CdI

Gh

Uvaria dinklagei

Annonaceae

WC

+

4

2

Sericanthe adamii

Rubiaceae

WC

+

6

2

Impatiens nzoana ssp. nzoana

Balsaminaceae

H

+

+

7

5

Clerodendrum sassandrense

Verbenaceae

S

+

10

2

Pseudocalyx libericus

Acanthaceae

WC

+

11

2

Dichapetalum dictyospermum

Dichapetalaceae

WC

+

11

53

Albertisia cordifolia

Menispermaceae

WC

+

13

14

Alafia parciflora

Apocynaceae

WC

+

+

22

3

Sabicea arachnoidea

Rubiaceae

WC

+

23

3

Millettia leonensis

Leguminosae-Pap.

T

+

31

2

Begonia quadrialata ssp. nimbaensis

Begoniaceae

H

+

+

+

46

17

Tapinanthus praetexta

Loranthaceae

S

+

46

11

Ixora liberiensis

Rubiaceae

S

+

47

2

Alsodeiopsis chippii

Icacinaceae

S

+

60

6

Tapura ivorensis

Dichapetalaceae

T

+

+

65

3

Gilbertiodendron robynsianum

Leguminosae-Caes.

T

+

68

6

Cola umbratilis

Sterculiaceae

T

+

68

10

Suregada ivorensis

Euphorbiaceae

T

+

73

2

Zanthoxylum psammophilum

Rutaceae

WC

+

87

2

Strychnos odorata

Loganiaceae

WC

+

89

4

Beilschmiedia caudata

Lauraceae

T

+

92

6

Distribution type, range and frequency in relation to species life-history traits

Continuous and disjunct distribution types differed in the proportion of species that are shade bearers or light demanding (i.e. non-pioneer and pioneer) (Chi2 analysis, X2 = 6.6, df = 1, p < 0.01, n = 200). There were more shade tolerant species with disjunct populations than one would statistically expect if there was no association between distribution type and species guild. We found an association between life form and distribution type (Chi2 analysis, X2 = 11.1, df = 3, p = 0.01, n = 261). Herbs had relatively more species with disjunct distribution type. We found no association between distribution type and dispersal mechanism.

There were no significant relationships between a species' range and the life-history traits (i.e. life form, guild, and dispersal mechanism). Neither did the abundance of a species vary with its life form or dispersal mechanism. However, light demanding species (i.e. nonpioneer light demanding ones and pioneers) were more abundant than shade-bearers (t-test, p = 0.02, nsb = 70, nu = 66).

We did find that a species' percentage of records at open places was significantly different among dispersal syndromes. Species dispersed by wind had a higher percentage of records in open places compared to species dispersed by animals or with explosive seeds (Fig. 7.8, oneway Anova, p = 0.02).

reflects the fact that altitude represents a complex environmental gradient where different resources (temperature, irradiance, soil fertility, water availability) change simultaneously.

General patterns of distribution and driving forces

We found that species with continuous spatial distributions had a higher proportion of records in disturbed open habitats than species with disjunct distributions. In fact, species with disjunct distributions were predominantly shade tolerant. Species with larger ranges (i.e. maximum distance between collections) were found in a larger range of rainfall conditions and had a larger drought tolerance indicated by the minimum rainfall where the species has been collected. Also the commonness of a species (i.e. number of records) was positively correlated with the amplitude of rainfall conditions at which it was found and the openness of the habitat. Light demanding species were more abundant than shade tolerants. We also found that species having a higher proportion of records in open places were mainly wind dispersed. These patterns indicate that how continuous, widespread, and abundant a species is, depends on how successful a species is in dispersal and in tolerating open habitats, and a wide range of rainfall conditions.

These results suggest the success of a ruderal strategy behind the main patterns of plant distribution in these forests. This is an interesting result because even within the set of endemic or rare forest species of Upper Guinea the most successful species tend to be ruderal, a pattern that is more commonly found in drier ecosystems. Our results are in line with the early observations of Chevalier (1917) and Mildbraed (1922), who pointed out that a very large proportion of the African forest species had wide ranges (Richards 1973). These patterns likely reflect the importance of disturbances in shaping the composition and distribution of species in West Africa. The climatic fluctuations over the past 10 million years, which have periodically led to expansions and contractions of these rainforests (Morley 2000), might have contributed to the selection of ruderal species able to disperse and colonise new areas. Certainly the long and severe deforestation in this region (e.g. Martin 1991) must have reinforced this pattern. As a result of deforestation, the landscape has become drier both in terms of the microclimate of open

Table 7.3 Twenty-six extremely rare forest species (or varieties) with only one or two collections in Upper Guinea in the database.

The life form, country of occurrence, abundance (# collections), and frequency (# half-degree grid cells in which the species occurs) are given. Byttneria dahomensis occurs only in Benin, so there is no "+" mark in the Countries list. For abbreviations see Table 7.1.

Table 7.3 Twenty-six extremely rare forest species (or varieties) with only one or two collections in Upper Guinea in the database.

The life form, country of occurrence, abundance (# collections), and frequency (# half-degree grid cells in which the species occurs) are given. Byttneria dahomensis occurs only in Benin, so there is no "+" mark in the Countries list. For abbreviations see Table 7.1.

Species

Family

LF

Countries

Abundance (#)

Frequency (#)

Gu

SL

L

CdI

Gh

Impatiens nzoana ssp. bennae

Balsaminaceae

H

+

1

1

Begonia fusicarpa

Begoniaceae

H

+

1

1

Begonia prismatocarpa ssp. petraea

Begoniaceae

H

+

1

1

Gilbertiodendron obliquum

Leguminosae-Caes.

S

+

1

1

Hibiscus whytei

Malvaceae

H

+

+

+

1

1

Diaphananthe suborbicularis

Orchidaceae

H

+

1

1

Malaxis melanotoessa

Orchidaceae

H

+

1

1

Argocoffeopsis lemblinii

Rubiaceae

S

+

1

1

Hymenocoleus axillaris

Rubiaceae

S

+

1

1

Sabicea bracteolata

Rubiaceae

WC

+

1

1

Tarenna vignei var. vignei

Rubiaceae

S

+

1

1

Byttneria dahomensis

Sterculiaceae

WC

1

1

Byttneria ivorensis

Sterculiaceae

WC

+

1

1

Pseudocalyx libericus

Acanthaceae

WC

+

2

1

Ancistrocladus pachyrrachis

Ancistrocladaceae

WC

+

2

1

Uvaria dinklagei

Annonaceae

WC

+

2

1

Suregada ivorensis

Euphorbiaceae

T

+

2

2

Leuccomphalos libericus

Leguminosae-Pap.

WC

+

+

2

2

Millettia leonensis

Leguminosae-Pap.

T

+

2

2

Ixora liberiensis

Rubiaceae

S

+

2

1

Keetia obovata

Rubiaceae

WC

+

+

2

2

Sericanthe adamii

Rubiaceae

WC

+

2

2

Zanthoxylum psammophilum

Rutaceae

WC

+

2

2

Byttneria guineensis

Sterculiaceae

WC

+

2

1

Clerodendrum sassandrense

Verbenaceae

S

+

2

2

Premna grandifolia

Verbenaceae

S

+

2

2

gaps within the forest area, as well as in terms of the general climate conditions experienced in the whole region. This is because of the feedbacks of vegetation on both micro and regional climate. In open gaps irradiance and temperatures are higher and relative humidity is lower compared to conditions under the closed forest canopy (e.g. Agyeman et al. 1999b). At a larger scale, tropical forests modify regional climate by recycling a large proportion of the rainfall mainly through the transpiration of forest trees. Deforestation disrupts this feedback, makes climate drier, and beyond a certain threshold could cause a collapse of forest ecosystems and their replacement by savannas (Da Silveira & Sternberg 2001). In fact, hydrological models predict that in the worst scenario of deforestation in West Africa, in which tropical forests are converted into savannas, we could expect a complete collapse of the monsoon system and a significant reduction of regional rainfall to about half of present conditions (Zheng & Eltahir 1997).

Certainly we should interpret our results cautiously because they have been derived from the analyses of non randomly selected samples. We can expect herbaria collections to be biased towards places easily approached by roads and therefore closer to relatively disturbed places. One may also argue that the correlation between a species' distribution range and the range of rainfall conditions is spurious because a species with a large range would cover a larger part of the rainfall gradient. We believe this is not the case because rainfall is the driving factor for occurrence in 71% of the species, and all life forms responded to it the same way. In general, the clarity of the patterns we found is remarkable in view of the great diversity of genus, life forms, and habitat of the analysed species.

Rare species and disjunct distributions: relicts from the past?

Most very rare species (with one or two collections) and disjunct species occurred near the coast at the border between Côte d'Ivoire and Ghana, at the border between Côte d'Ivoire and Liberia or further to the west. Others were confined to the mountains. These areas coincide with the three postulated Pleistocene forest refuges: Cape Three Points, Cape Palmas, and Mount Nimba (Fig. 7.3, se

Oplan Termites

Oplan Termites

You Might Start Missing Your Termites After Kickin'em Out. After All, They Have Been Your Roommates For Quite A While. Enraged With How The Termites Have Eaten Up Your Antique Furniture? Can't Wait To Have Them Exterminated Completely From The Face Of The Earth? Fret Not. We Will Tell You How To Get Rid Of Them From Your House At Least. If Not From The Face The Earth.

Get My Free Ebook


Post a comment