Study sites and methods

The study sites comprise thirteen forests in the centreeast, the southeast, the southern coast, the centre-west, and the southwest of Côte d'Ivoire. The choice of these different forests was based on the intactness of the forest cover and the availability of species lists. The forests vary from semi-deciduous to evergreen (Figure 5.2), they are of various sizes (2590 to 300,000 ha), and have different soil and climatic conditions (Table 5.1). Three soil types can be distinguished: tertiary sands, clay soils derived from schist, or sandy soils resulting from granite weathering (De Rham 1971). The total annual rainfall ranges between 1400 and 2300 mm, while the water deficit is between 150 and 400 mm. The length of the dry season varies between 2 and 5 months (Perraud 1971).

Most of the data used in this research come from various studies, in which checklists are made for the forests (Table 5.1). Such inventories aim to identify all the plant taxa encountered while following pre-existing paths in the forest or paths created for this purpose. In addition, we carried out inventories in those forests that had no or

Table 5.1 The forests iticluded in this study (E = evergreen, SD = semi-deciduous) atid their environmental characteristics. "Abbr. "refers to the abbreviations usediti the figures. "Taxa" indicates the number of taxa found in the forests.

Geographical Zone

Forest

Abbr.

Forest type1

Area (ha)

Coordinates

Annual rainfall' (mm/yr)

Centre-East

Bossematie

Bm

SD

22,200

6° 20' - 6° 35'N 3° 20' - 3° 35'W

1400 - 1500

Southeast

Mabi

Mb

E

59,616

5° 51' - 6° 05'N 3° 22' - 3° 4l'W

1650 - 1700

Songan

S

E

38,189

5° 46' - 6° 12'N 3° 12' - 3° 26'W

1600 - 1650

Tamin

T

E

24,934

5° 40' - 5° 58'N 3° 11'-3° 25'W

1650 - 1700

Yaya

Y

E

23,877

5° 35' - 5° 54'N 3° 23' - 3° 46'W

1700 - 1800

South coast

Dassioko

D

E

11,203

5° 00' - 5° 07'N 5° 50' - 5° 57'W

1550 - 1600

Monog^ga

Mo

E

39,660

4° 44' - 4° 58'N 6° 14' - 6° 35'W

1650 - 1750

Port Gauthier

PG

E

2,590

5° 06' - 5° 11'N 5° 29' - 5° 34'W

1550 - 1600

Banco

B

E

3,300

5° 21' - 5° 25'N 4° 01'-4° 05'W

2000

Centre-West

Haut Sassandra

HS

SD

102,400

6° 22' - 7° 24'N 6° 59'-7° 10'W

1460 - 1680

Marahoue

Ma

SD

101,000

6° 53' - 7° 14'N 5° 46' - 6° 10'W

1400

Southwest

Haute Dodo

HD

E

236,733

4° 41' - 5° 19'N 7° 01' - 7° 25'W

1900 - 2300

Taï

Taï

E

300,000

5° 09' - 6° 09'N 6° 48' - 7° 26'W

1800 - 2200

1. Guillaume: &Adjanohoun (1969)

Subsoil and soil

Water deficit4 (mm/yr)

Taxa (#)

Source

Schist

350 - 400

611

Ake Assi (1992) Bakayoko (1999)

Schist

300

640

This study

Schist

300

591

This study

Schist

200

512

This study

Schis: 200 617 This study

Sand 450 719 Ake Assi (1997)

Sand 300 859 Kouame (1998b)

Sand 450 705 Kouame (1998b)

Sand 200 773 De Koning (1983)

Granite 400 843 Kouame (1998a)

Schist 350 - 400 475 Jongkind et ai (1999)

Granite 150 906 Kouadio (2000)

Kouassi (2000)

Granite 200 - 250 849 Ake Assi & Pfeffer (1975)

Figure 5.2 Map of Côte d'Ivoire with four main vegetation types (evergreen forest, semi-deciduous forest, Guinean savanna, sub-Sudanian savanna), the forest reserves (white polygons), and the 13 forests studied (dark grey polygons). The bold V-shaped line indicates the savanna intrusion in the forest zone, or V-Baoulé.

incomplete species lists. Thus, in the southeast forest area (Mabi, Tamin, Songan and Yaya), sixteen 2-km tracks, previously cleared by the forest management services (SODEFOR-GTZ), were used for species inventories. In Haute Dodo, due to the lack of specially cleared tracks, we randomly distributed several sites for inventories, following the pre-existing paths (Kouadio 2000). These inventories were accompanied by an inventory of all species in sample plots (Kouassi 2000). We are aware that the species lists analysed in this research are not exhaustive. In consequence, our discussions are based on the current state of knowledge of the flora of Ivorian forests.

The species lists of the forests were pairwise compared using S0renson's similarity index (1948). This allowed us to distinguish forests that were more similar in species composition and those that were more dissimilar. The maximum value of the index (100%) indicates that two forests have the same species composition. The minimum value (0%) indicates that two forests have no species in common.

Subsequently an ordination of the 13 forests and 2126 taxa was carried out using correspondence analysis (Jongman et al. 1987), based on the presence or absence of species in the species lists. The ordination score was related to the following environmental parameters: longitude and latitude, soils and subsoil, the annual water deficit and the total annual rainfall. The groups of forests produced by the correspondence analysis were pairwise compared using S0renson's similarity index. The taxa which were found in all the forests were considered as being the most common species of Ivorian closed forests. The taxa which were inventoried in all the forests of the same group, but which were not found in any other forest group, were considered as being typical for the group in which they were found.

Results

The floristic diversity of the thirteen forests studied comprises 150 families, 857 genera and 2126 species. 624 species are found in only one of the forests, while 41 species are found in all the forests (Table 5.2). The number of species shared by the forests is inversely

Table 5.2 Taxa that are common to the 13 forests, or taxa that are characteristic for each of the four forest groups (Fig. 5.4).

Species common to the 13 forests

Aganope leucobotrya Agelaea paradoxa Agelaea pentagyna Aidia genipiflora Alstonia boonei Amphimas pterocarpoides Baphia nitida Baphia pubescens Buchholzia coriacea Bussea occidentalis Calycobolus africanus

Cola nitida Costus afer

Craterispermum caudatum Culcasia barombensis Diospyros soubreana Funtumia africana Glyphaea brevis Grifonia simplicifolia Klainedoxa gabonensis Landolphia hirsuta Landolphia owariensis Microdesmis keayana Myrianthus arboreus Myrianthus libericus Napoleonaea vogelii Nephrolepis biserrata Neuropeltis acuminata Ochthocosmus africanus Palisota hirsuta Parinari excelsa Piptadeniastrum africanum Psychotria peduncularis Pycnanthus angolensis Strophanthus gratus Strychnos aculeata Treculia africana Ventilago africana Xylopia quintasii Xylopia villosa

Semi-deciduous forests (Group I)

Acroceras gabunense

Aneilema umbrosum

Bridelia atroviridis

Clerodendrum polycephalum

Cyrtococcum chaetophoron

Desmodium adscendens var.robustum

Dichapetalum madagascariense var.

madagascariense

Diospyros abyssinica

Eugenia tabouensis

Grewia carpinifolia

Khaya grandifoliola

Lagenaria breviflora

Landolphia landolphioides

Melochia melissifolia

Mischogyne elliotianum var. glabra

Psychotria kitsonii

Psydrax manensis

Simirestis dewildemaniana

Strychnos congolana

Strychnos splendens

Telfairia occidentalis

Vitex ferruginea subsp. ferruginea

Coastal forests (Group II)

Ancistrocladus barteri Bulbophyllum imbricatum Eugenia whytei Eupatorium microstemon Heterotis rotundifolia Salacia pallescens Salacia whytei Tapinanthus belvisii

Southwest forests (Group III)

Angraecum podochiloides Anthoclitandra nitida Bertiera fimbriata Bolbitis heudelotii Brieya fasciculata Cercestis ivorensis Clappertonia minor Dalbergia albiflora Delpydora gracilis Didelotia brevipaniculata Drypetes klainei Garcinia granulata Gilbertiodendron robynsianum Gynura sarmentosa Lomariopsis rossii Mapania minor Millettia lucens Mussaenda landolphioides Pauridiantha hirtella Polystemonanthus dinklagei Premna grandifolia Psychotria subglabra Renealmia maculata Scleria vogelii Selaginella versicolor Strychnos icaja Tarenna gracilis Trichilia heudelotii Vitex ferruginea

Southeast forests (Group IV)

Aframomum alboviolaceum Buforrestia mannii Cecropia peltata Costus englerianus Crotonogyne craterviflora Friesodielsia enghiana Guibourtia copallifera Guibourtia tessmannii Licania elaeosperma Marantochloa filipes Memecylon polyanthemos Rutidea dupuisii subsp. occidentalis Sabicea discolor correlated to the number of forests compared (Figure 5.3). Nine families are each represented by more than 50 species (Table 5.3). Rubiaceae and Euphorbiaceae are the most speciose families. The richest genera are Psychotria and Ficus with 43 and 34 species respectively; they top the list of the six genera represented by at least 20 species each (Table 5.3).

The similarity index is lowest (29%) between Banco and Marahoue (Table 5.4), which reveals the floristic dissimilarity of these two national parks. The similarity index is highest (77%) between Songan Forest Reserve and Tamin Forest Reserve, which resemble each other most. The highest values are found either between forests belonging to the same geographical zone (Tai/Haute Dodo, Songan/Tamin, Mabi/Songan, Tamin/Yaya) or between a forest of the southeast and a forest of the southwest (Haute Dodo/Mabi, Haute Dodo/Yaya). The similarity index varies from 38% (Dassioko/Marahoue) to 53% (Bossematie/Songan) for the forests with the same longitudes but different latitudes.

The first and second axis of the correspondence analysis explain together 31% of the variation in species composition. 44% of the variation in species composition is explained by the six environmental variables. Four groups of forests can be distinguished (Figure 5.4). Group I, made up of semi-deciduous forests (Bossematie, Haut Sassandra, Marahoue), is characterised by a high water deficit (Figure 5.5A) and a high latitude (Figure 5.5B). The coastal forests (Banco, Dassioko, Port Gauthier and Monogaga) make up group II, which is characterised by a low latitude, intermediate rainfall, and occurrence on tertiary sandy soils (Figure 5.4). Group III corresponds to the forests of the southwest (Haute Dodo, Tai), while group IV corresponds to the forests of the southeast (Mabi, Songan, Tamin, Yaya). Groups III and IV are

Table 5.3 The most common families and genera from all 13 forests. Families with more than 50 species each are shown, and the genera represented by at least 20 taxa. Psychotria is principally herbaceous and Ficus includes mainly shrubs. The genera Salacia, Combretum, Strychnos and Dichapetalum are essentially lianescent.

Table 5.3 The most common families and genera from all 13 forests. Families with more than 50 species each are shown, and the genera represented by at least 20 taxa. Psychotria is principally herbaceous and Ficus includes mainly shrubs. The genera Salacia, Combretum, Strychnos and Dichapetalum are essentially lianescent.

Family

Species (#)

Genus

Species (#)

Rubiaceae

260

Psychotria

43

Euphorbiaceae

107

Ficus

34

Fabaceae

86

Salacia

28

Apocynaceae

77

Combretum

27

Orchidaceae

76

Strychnos

21

Annonaceae

70

Dichapetalum

20

Caesalpiniaceae

69

Moraceae

52

Hippocrateaceae

51

centres of greatest floristic diversity in the lowlands in Côte d'Ivoire. They are characterised by an intermediate latitudinal position, quite high rainfall and a low water deficit (Figure 5.4). They differ fundamentally in the type of their subsoils. The subsoil of the southwest is essentially of granite, whereas that of the southeast is composed of schists.

The annual water deficit and the latitudinal position are the environmental factors that are best correlated with the first axis (Figure 5.5, Table 5.5). The second axis has a strong positive correlation with tertiary sandy soils, a strong negative correlation with schists (Table 5.5), and a weak negative correlation with latitude (Figure 5.4B). The similarity coefficients between the groups of forests produced by the correspondence analysis are always higher than 50%. This indicates that these groups of forests have many species in common. The southwest and southeast groups have the highest index of similarity (65%). Seventy-two species are characteristic for one of the forest groups: 22 species are typical for the semi-deciduous forests, 8 for the coastal forests, 29 for the southwest forests and 13 for the southeast (Table 5.2).

Table 5.4 Matrix of similarities between forests using S0rensen's index (1948). The figures in brackets correspond to the cumulative species richness of the two forests that are compared. The similarity indices larger than 50% (indicating that two forests have more than 50% of the species in common) are underlined and those larger than 60% are given in bold.

Table 5.4 Matrix of similarities between forests using S0rensen's index (1948). The figures in brackets correspond to the cumulative species richness of the two forests that are compared. The similarity indices larger than 50% (indicating that two forests have more than 50% of the species in common) are underlined and those larger than 60% are given in bold.

Forest

Banco

Bossé-matié

Dassioko

Haut Sassandra

Gauthier

Mabi

Mara-houe

Mono-gaga

Songan

Taï

Tamin

Bossématié

(1382)

Dassioko

(1495)

(1328)

Haut Sassandra

42 (1618)

(1447)

48 (1561)

Haute Dodo

(1685)

(1518)

(1632)

(1751)

Port Gauthier

(1482)

(1313)

(1427)

49.5 (1546)

(1617)

Mabi

(1417)

(1249)

(1363)

(1482)

(1553)

(1348)

Marahoue

(1250)

50 (1082)

(1195)

(1318)

(1385)

40 (1182)

(1117)

Monogaga

(1635)

(1467)

(1581)

(1700)

(1771)

(1566)

(1502)

(1335)

Songan

(1367)

(1200)

(1314)

(1433)

(1504)

(1299)

(1235)

(1067)

(1453)

Taï

(1625)

(1457)

(1571)

(1690)

(1761)

(1556)

(1492)

(1325)

(1710)

(1443)

Tamin

(1289)

46 (1121)

(1235)

(1354)

(1425)

49 (1220)

(1156)

(989)

(1374)

(1107)

(1364)

Yaya

(1392)

(1224)

(1338)

(1457)

(1528)

(1323)

(1259)

(1092)

(1477)

(1210)

(1467)

(1131)

Table 5.5 Pearson's correlation coefficients between the environmental factors and the first two ordination axes of the correspondence analysis. ns= not significant; * = P < 0.05; ** = P < 0.01.

Variable

Axis 1

Axis 2

Latitude

0.70

**

- 0.61

*

Longitude

- 0.38

ns

- 0.38

ns

Rainfall

- 0.39

ns

0.45

ns

Water deficit

0.70

**

0.15

ns

Granite

0.14

ns

- 0.04

ns

Sand

- 0.11

ns

0.90

**

Schist

- 0.02

ns

- 0.80

**

Figure 5.3 Relationship between the number of species in common and the number of forests compared. Spearman's rank correlation coefficient is given.

Figure 5.4 Ordination diagram of 13 Ivorian forests, showing their position on the first two axes of the correspondence analysis. Forest sites are indicated with filled symbols, and environmental factors with arrows. The forests are arranged in four groups. Abbreviations are given in Table 5.1.

Figure 5.3 Relationship between the number of species in common and the number of forests compared. Spearman's rank correlation coefficient is given.

Figure 5.4 Ordination diagram of 13 Ivorian forests, showing their position on the first two axes of the correspondence analysis. Forest sites are indicated with filled symbols, and environmental factors with arrows. The forests are arranged in four groups. Abbreviations are given in Table 5.1.

Discussion

Floristic diversity

The 9700 km2 of forests included in this research, represent less than 50% of the forest cover of the country before 1985 (Davis et al. 1994) and less than 3.6% of the national territory. However, 2126 species of vascular plants were found in these forests, which represents 58% (out of 3660 species) of the total Ivorian flora (Heywood & Davis 1994). This indicates that the 13 forests contribute to a large extent to the flora of Côte d'Ivoire. The high contribution can be explained by the fact that all principal lowland forest types in Côte d'Ivoire were included in the study. The remaining 42% of the Ivorian flora is found in savannas, the upland evergreen forests, forest islands, gallery forests, granitic domes and human environments.

The species richness of these forests resembles that of certain countries such as Benin, Liberia, Senegal or Togo. It is richer than Sierra Leone, whose flora is evaluated at 1700 species (Heywood & Davis 1994). It represents almost two-thirds the diversity of Guinea and Ghana. The relative paucity of flora in Benin, Togo and Senegal can be attributed to their low rainfall and to their limited forest cover. Liberia is considered to harbour the former glacial forest refuges (Morley 2000, Wieringa & Poorter chapter 6). The relatively low richness of Liberia and Sierra Leone may be attributed to the lack of knowledge of the flora of these countries. Haute Dodo is the richest forest in Côte d'Ivoire with at least 906 species.

The Rubiaceae (260 species) and Euphorbiaceae (100 species) were the most species rich families. At the national level, these forests contribute to respectively 85 and

75% of the species richness of these families (Ake Assi 1984). The Fabaceae, represented in the Guinean savannas by 103 species (Kouame 1993, Banninger 1995), are not uncommon in closed forests, where they account for 86 species.

The families of Dioncophyllaceae, Hoplestigmata-ceae, Medusandraceae, Octoknemataceae, Pandaceae and Scytopetalaceae are present in the 13 forests, and represent nearly 70% of the endemic families in the Guineo-Congolian region (White 1986). Numerous endemic Guineo-Congolian genera such as Afrobrunnichia, Amphimas, Anopyxis, Anthonotha, Antrocaryon, Aubrevillea, Buchholzia, Calpocalyx, Chidlowia, Coelocaryon, Coula, Crotonogyne, Cyclodiscus, Decorsella, Didelotia, Discoglypremna, Distemonanthus, Duboscia, Heckeldora, Hymenostegia, Gilbertiodendron, Grossera, Monocyclanthus, Ophiobotrys, Tieghemella and Turraeanthus have also been found in these forests.

Relationships among forest blocks

The distinction in forest types is governed by water availability, which is the combination of the total amount of rainfall, the rainfall distribution over the year, and the water holding capacity of the soil.

The largest distinction in species composition in our ordination analysis is between semi-deciduous forests on the one hand, and the evergreen and coastal forests on the other hand (Figure 5.4). The semi-deciduous forests clearly have a lower rainfall which is unevenly distributed over the year. Although the coastal forests have a similar high rainfall as the other two evergreen forest groups, the water availability to the vegetation is substantially lower, because of a high water deficit, and sandy soils with a low water holding capacity. The more subtle separation between the two evergreen forest groups is mainly caused by a higher rainfall amount in the southwest.

it ei

i'mi"

VUi •ta

■a

tfc

V

r

A

Witv doUfri^M

P D

□ Ù

U'.i

*D

i'-Q.pT-

07

■ T.a

-U

kO „

• L'-

-Lin

f*h

44

B

Figure 5.5 Relationship between A) the first axis score of the forest and the water deficit, B) the second axis score of the forest and the latitudinal position. The regression lines, coefficients of determination, and significance level are given.

species occurs in only one of 13 forests and only 2% of the species occur in all sites.

The floristic similarity between two forests varies from 29 to 77%, and is on average around 50%. In general, the floristic similarity is related to the distance between sites, their similarity of environmental conditions, or a combination of the two. All three cases were found in this study. For example, the neighbouring forests of Bossématié and Songan share almost 50% of the species despite obvious differences in the environment. The forests of the southeast and the southwest that are 400 km apart have a high floristic similarity, because of a high similarity in water availability. Subtle differences in rainfall and water deficit in these two zones are attenuated by the nature of their subsoils. The clay soils of the southeast (Perraud & Souchère 1970) have a greater water retention capacity than the sandy soils of the southwest (De Rham 1971), thus compensating for the lower rainfall in this region. As a consequence, plants experience a similar water availability in these two zones. Finally, forests which are geographically close and have similar soil and climatic conditions, as is the case for Mabi, Songan, Tamin, and Yaya, have the highest floristic similarities (Table 5.4).

The paleoclimatic history is another reason why two forests might be similar. It has been postulated that forests in the extreme southwest and southeast Côte d'Ivoire have been part of two former glacial forest refuges (Guillaumet 1967, Wieringa & Poorter chapter 6). Wet forest species might have survived in these refugia during the dry glacial periods. Many species with regional or continental disjunct distribution patterns have populations in both of these two forest blocks (Holmgren et al. chapter 7). Summarising, the differences among Ivorian forests are governed by an intricate interplay between climate, soil and history.

Although Banco has a high rainfall, and is relatively far from the coast (13 km) it still has been grouped with the coastal forests (Figure 5.4). The high rainfall is mediated by the low water holding capacity of its sandy soils, resulting in a low water availability. The most important environmental factor that determines this floristic group is the soil; the tertiary sands which underlie the forests, lead to a low water and nutrient availability.

Each forest group possesses a fairly well defined group of characteristic species (Table 5.2), e.g., Khaya grandifoliola in semi-deciduous forests, Ancistrocladus barteri in the coastal forest, Mapania minor in the southwest forests, and two species of Guibourtia in the southeast forests. It is striking that almost one third of the

Acknowledgements

We would like to thank SODEFOR for having allowed us to carry out research in its forest reserves. We are grateful to Claude Amani, Jean Assi, Saturnin Dougoune, Amadou Fofana and Patrice Mabea for their assistance in the data collection. We thank Profs Laurent Ake Assi for identifying the taxa, and Dossahoua Traore and Frans Bongers for their comments, which have allowed us to improve this chapter.

J.J. Wieringa and L. Poorter

Biodiversity hotspots in West Africa; patterns and causes

Introduction

The rainforests of West Africa have been earmarked as one of the world's hotspots of biodiversity (Myers et al. 2000). These forests extend from Togo to Senegal, and are referred to as the Upper Guinean forests (White 1983). The Upper Guinean forests are separated from the rest of the African rainforests by the Dahomey gap; a woodland savanna which extends in Togo and Benin from the north to the Gulf of Guinea.

Upper Guinean forests harbour a large number of endemic plant and animal species. It is estimated that about 2800 vascular plant species can be found in the Upper Guinean forests, of which 22% are endemic to the region (Jongkind, chapter 11). These forests are disappearing rapidly because of logging activities, shifting cultivation, and conversion into plantations (Chatelain et al. chapter 2). For an effective conservation policy, information is needed on the distribution of rare and endemic species in Upper Guinea, and the places in which they are concentrated. A problem of many tropical countries is that such botanical background information is scarce, or highly fragmented. To rapidly generate the necessary information one may carry out botanical surveys, in which selected areas are screened for their species composition. Hawthorne and Abu-Juam (1995) used such an approach in Ghana. Based on sample plots, systematically distributed over the whole forest zone, they were able to demarcate areas with a high share of endemic species. However, such an approach is labour intensive, and is only possible if a restricted floristic and geographic range is covered with a team of well-trained botanists. Another option is to use existing herbarium collections as a data source. Botanical collections have the advantage that they provide an existing source of information, cover a large geographic range, and that they are likely to be identified correctly. The latter is important, as especially the rare species are not easily recognised by tree spotters in the field. A disadvantage is that collection efforts are not evenly distributed over the area.

Several authors have used the distribution patterns of small groups of plants or animals to indicate areas with a high biodiversity (Aubreville 1949, 1962, Hamilton 1976, Grubb 1982, Sosef 1994, Lovett et al. 2000). These studies typically point at three areas in Upper Guinea: the interior

Figure 6.1 The wet evergreen forest of Cape Three Points, Ghana. Cape Three Points is one of the three postulated Pleistocene forest refuges in Upper Guinea.

of Liberia with a diversity centre around Mount Nimba, Cape Palmas at the border between Liberia and Côte d'Ivoire, and Cape Three Points and its surrounding area in southwest Ghana. The exact location of the hotspots may vary, depending of the taxonomic group under concern (Conservation International 2001). A few biodiversity analyses on a larger number of species exist (e.g. Pomeroy 1993, Linder 2001), but they have a rather low resolution. Up to now a detailed, quantitative analysis based on a large number of species is therefore lacking for Upper Guinea.

Both environmental and historical factors affect spatial variation in species richness. Species richness is known to vary along environmental gradients of rainfall (Hall & Swaine 1976, Currie 1991, O'Brien 1993), altitude (Hall 1973), and soil fertility (Hall & Swaine 1976, Huston 1980). In general, better site conditions lead to an increased primary productivity, more individuals and niches, and hence, a higher species richness. Species richness may decrease at very productive sites, giving rise to an unimodal relationship between richness and site productivity (Grime 1973, Huston 1979).

Current floristic patterns are strongly shaped by large-scale climatic disturbances in the past. Throughout the Quaternary, the area of rainforest waxed and waned with climatic fluctuations (Hamilton & Taylor 1991). During the dry and cool glacial periods, precipitation levels declined, and the rainforest contracted to small patches. As rainfall levels were low in Africa compared to other continents, the effect of the glacial period has been felt more strongly here than elsewhere (Richards 1973, Morley 2000) and it is likely that its imprint has lasted for a longer time. Testimony of the dynamic changes in vegetation cover are savanna relicts within the rainforest zone in Côte d'Ivoire (Gautier & Spichiger chapter 3), and the observation of fossils of Guineo-Congolean forest trees in Ethiopia (Bonnefille & Letouzey 1976). Some species could not keep up with the rapid expansion, and are still found in a narrow range around former forest refuges. The terrestrial forest herb Begonia mildbraedii, for example, has a disjunct distribution, with two isolated populations in Côte d'Ivoire and Ghana, being as far as 1800 km apart from its main, more widespread, distribution in central Africa (Holmgren et al. chapter 9, Sosef 1994).

This chapter focuses on patterns and causes of plant biodiversity hotspots in West Africa. Hotspots are defined as centres with a high richness of rare and endemic species. The chapter explores how, and to what extent herbarium collections can be used to define hotspots of biodiversity. Then it relates biodiversity to environment and distance to postulated former forest refuges, and weighs the importance of environment and history in the current distribution of biodiversity.

Methods

Data collection

Based on the 2nd edition of the Flora of West Tropical Africa (Keay 1954, 1958b, 1963, Hepper 1968, 1972), inventories in Ghana (Hall & Swaine 1981, Hawthorne 1995a), taxonomic revisions, and new herbarium collections, we made a compilation of just over 1000 species, which are rare or endemic to the closed forests of Upper Guinea (see Jongkind & Wieringa, chapter 11). All life forms were included (trees, shrubs, lianas, herbs, parasites, saprophytes and epiphytes). 640 species were selected for a shortlist to analyse biodiversity patterns. Care was taken to include species from different families, and to include species with different distribution patterns or ecology. Of these 640 species, herbarium specimens were entered for c. 510 species, and this was complemented by distribution data from taxonomic revisions for c. 130 species.

We entered into a database all herbarium specimens collected from Senegal to Togo. For some non-endemic rare species, we also included the herbarium specimens collected in Lower Guinea. We entered all specimens from Herbarium Vadense (Wageningen, The Netherlands),

Figure 6.2 Relationship between the observed number of species in a cell and the number of herbarium collections in that cell. The cells have a size of55 X 55 km (n = 329). Two cells with a collection intensity larger than 500per cell are not shown.

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) (only as far as the data was owned by Geneva), 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). For many non-endemic species, specimens from lower Guinea were entered as well. In total the database contained 48,000 records from West Africa, of which over 12,500 records correspond to rare or endemic species from our shortlist.

The database covers all major herbaria that have collections from West Africa, with the exception of Paris which herbarium was closed for reconstruction during our data-entry period. For our area, we expect the Paris herbarium to contain mainly collections from Côte d'Ivoire and Guinea. Since we have abundant collections from Côte d'Ivoire from other sources, the real problem was the lack of data from Guinea. Cited collections made by Chevalier in Guinea were retraced using Chevalier (1911, 1920). Although our results for Guinea would have been more accurate with all data available, it is likely that they are good enough for those areas where the largest amount of forest species is to be expected (Fouta Djalon, Mt Ziama and Mt Nimba). The identification of specimens in different herbaria was cross-checked by C. Jongkind, who has ample knowledge of the taxonomy of West African forest species. The nomenclature was updated using Hawthorne & Jongkind (2004).

Box 6.1. Species-collection curve

To allow for comparison between regions, one may construct a species-collection curve, in a similar way as a species-area curve is made. For each grid-cell of half a degree latitude by half a degree longitude, we randomly drew collections, and plotted the cumulative number of species found, against the cumulative number of collections (Figure 6.3). In this way 100 species-collection curves were constructed and subsequently averaged. The number of species found in a given area increases with the number of collections made, until an asymptote is reached (Colwell & Coddington 1995, ter Steege 1998). Therefore we fitted per cell an asymptotic curve through the data, according to the equation Sn = (Smax* n) / (c+n), where Sn is the number of species in a sample of n collections, Smax the estimated maximum number of species in a given area, and c is a constant (Figure 6.3). For this analysis we confined ourselves to collections belonging to our shortlist of 640 species.

The biodiversity estimates become more accurate when many specimens (N) have been collected relative to the number of species found (Sob,). We consider our biodiversity estimate to be fairly satisfactory if for a cell the ratio of N / Sob, is larger than 1.5. If N / Sob, < 1.1, then the biodiversity estimates vary considerably, and are too unreliable to use. We only fitted a curve through the data, if we had eight or more collections. For eight cells as many species were found as collections made, so curve fitting was not possible. For another seven cells N / Sobs was smaller than 1.1. This reduced the number of cells with regression results to 154.

n nuG 4xo ud nr ioooc

n nuG 4xo ud nr ioooc

GotoctxroiH

Figure 6.3 Species-collection curve for the cell that includes Banco forest, Côte d'Ivoire. The species-collection curve is created by randomly drawing collections from the total collection pool in a 55 X 55 km area. The cumulative species number is plotted against the cumulative number of collections (circles). Subsequently a regression curve is fitted through the data using an asymptotic curve. By using the curve, the predicted species richness can be calculated (broken line), and the maximum number of species in the area (Smax) can be estimated.

How to analyse biodiversity patterns?

To describe large-scale patterns in biodiversity, Upper Guinea should be divided into areas of equal sizes. What grid size should be chosen to describe those patterns best? A practical consideration is that there should be sufficient collections in each cell to make meaningful comparisons. A more theoretical consideration is that the grid size should not be too small, else local site conditions are likely to overrule the large-scale picture. As a compromise we selected a gridsize of 0.5 X 0.5 degree.

Botanical collections have several advantages; they provide an existing source of information, are likely to be correctly identified (if the herbarium is not too small and if the species groups have been revised recently), and provide permanent records that can always be rechecked. A disadvantage is that specimen collection is not carried out in a stratified or random way. As a consequence, sampling efforts are not evenly (or randomly) distributed, and areas of high measured species density often coincide with areas of high collection intensity (Nelson et al. 1990). The same applies for the herbarium data from West Africa. If West Africa is divided in equal cells of 0.5 x 0.5 degree, then there is a large variation in collection intensity; the number of collections per cell for our shortlist species varies from 0 to 1303. The observed number of species per cell increases in a curvilinear way with the number of collections (Figure 6.2, second degree polynome, r2 = 0.96, P < 0.001, n = 329). Sites with a larger number of collections appear to be species-rich, but this can largely be attributed to higher sampling effort there.

The shape of the curve resembles the species-area curve (Gaston 1996), and the underlying sampling mechanism is somewhat comparable; the larger the area or number of collections sampled, the more species are found. To allow for comparison between regions, one may construct a species-collection curve, in a similar way a species-area curve is made (Box 6.1). From the species-collection curves several biodiversity measures are derived. The S 50 indicates the number of species found when 50 collections are made from the shortlist of 640 rare or endemic species (Box 6.2). The Smax is the estimated maximum number of species from our shortlist that can be found in a cell (Box 6.2). Finally, the rarity-weighted species richness (Srw) indicates areas with a large number of rare and endemic species (Box 6.3). For each of these three biodiversity measures maps have been made using the inverse distance weighting interpolation method in ArcView. Cells for which curve fitting was impossible (N = Sobs), or that were considered to be too unreliable to be used (N / Sobs < 1.1, see Box 6.1) were not included in

Box 6.2. S50 and Smax; two measures of species richness

The diversity of different grid cells can be compared, using a species-collection curve. Figure 6.4 shows such a species-collection curve for three different sites. In Tabou, the species number increases more rapidly with collection numbers than in Tai', which in turn is far more diverse than Banco. Tabou has therefore the largest diversity. Yet, this ranking in diversity is reversed, when one simply compares the total number of species found (Banco 192 species, Tai' 160, and Tabou 119). The reason is that in Banco far more collections have been made than at the other two sites. According to our calculations in Banco 84% of the actually occurring species have also been found, while for Tabou this is only 38%. As it would be far too complicated to compare many species-collection curves, we use as a biodiversity measure the total number of species found at a standard number of collections. To this end we calculated, with help of the regression equation, the expected number of species at 50 collections (S50). This reference value of 50 is close to the median number of collections per grid cell (median = 35, range 8 - 1303), so that not too many extrapolations have to be made. S50 can vary from 1 (if all 50 collections in a cell belong to the same species) to 50 (if all 50 collections belong to 50 different species).

If we would make an infinite number of collections, we would approach the maximum number of species (Smax) in a cell. The Smax indicates how many rare species are present in an area. Smax and S50 show a strong exponential relationship (r2= 0.96, P < 0.001); therefore the larger S50, the larger Smax. At high levels of S50, a small increase in S50 will lead to a

Tmi«n»j

r^mij 1 > IV kppl

JT J* / y

/s

CMnll

Figure 6.4 Fitted species collection curves for 50 X 50 km areas around Tabou (continuous line), Tai (small broken line), and Banco (broken line). For each site the actual number of species observed in the area is indicated between parenthesis.

disproportionally large increase in Smax. The Smax is therefore a biodiversity measure that provides a large resolution at high levels of species richness. A disadvantage of Smax is that extrapolations need to be made. In incidental cases this may lead to an over- or underestimation of the real number of species present in the area.

the interpolation analysis. The interpolated biodiversity values are only shown for the forest zone.

Relationship between biodiversity and environment

Water availability, altitude, and soil fertility shape to a large extent the structure and composition of plant communities. Environmental variables used in the analyses were rainfall (in mm/yr), soil water holding capacity (WHC, in mm water/m soil), altitude (in m), soil fertility (Ca , Mg , K , in cmol cations per kg soil), soil pH and cation exchange capacity (CEC in %). A rainfall map was created based on a compilation of 578 weather stations in the region. Data on soil fertility and water holding capacity were calculated based on the FAO soil map of Africa, and a quantitative review of chemical analyses of soil profiles (Batjes 1997). For each grid cell an average soil fertility was calculated, based on the relative cover of the different soil types, and their median soil fertility. Similarly, for each grid cell an average soil water holding capacity was calculated, based on the relative cover of the soil types, their depth, and texture. It was presumed that sandy soils have a water holding capacity of 75 mm/m, loamy soils of 100 mm/m, and clayey soils of 125 mm/m. For a detailed analysis of the environmental metadata, see chapter 9. All spatial analyses were carried out using ArcView (Esri Inc.).

The S50, and Srw were related to environmental variables using a Spearman's rank correlation and stepwise multiple regression. In addition to these abiotic factors, also the distances to the three proposed glacial forest refuges were used as factors in the analyses. To some extent, it might seem circular to relate the richness of endemic species to the distance to proposed forest refuges, as endemism and refuge theory are closely connected. However, the existence and location of these forest refuges has been postulated mainly on the basis of pollen cores and paleoclimatic data (Maley 2001), and on species with disjunct distributions between Upper and Lower Guinea (Aubreville 1949, Guillaumet 1967). To a lesser extent it has been based on the occurrence of species with restricted distribution ranges.

Box 6.3. Rarity-weighted species richness

S50 and Smax are general measures of species richness that do not take the "specialness" of the species into account. For conservation purposes, one might be interested to know where species are concentrated that are rare, or have a limited distribution. To this end we calculated the rarity-weighted species richness (Srw). It is a measure of species diversity, that gives weight to cells with many rare or endemic species. Each species is weighted for the number of half-degree grid-cells in which it occurs. A common species, which occurs in 100 half-degree grid-cells receives thus a weighting score of 1 / 100 for each cell in which it is occurs. A rare species which occurs in 4 half degree grid-cells receives a weighting score of 0.25. The rarity-weighted species richness is then the sum of the weights of all the species occurring in a cell. Yet, of all the species that occur in a cell (Smax), only a part has been sampled (SobJ. To correct for this, we multiplied the weighted measure with Smax/Sob For example, if a grid cell A has 10 collections belonging to 4 species of which 1 species occurs in 1 cell, 2 in 2 cells and 1 in 4 cells, then the weighted richness of that cell equals (1 X 1 + 2 X 0.5 + 1 X 0.25) = 2.25. The predicted species richness (Smax) of the cell is 16, four times as much as actually observed. When correcting for sampling intensity, the rarity-weighted species richness equals

The rationale behind a rarity-weighted species richness is that, at a large scale (of Upper Guinea), each species is equally important, and receives a total weight of 1. We included 640 species in our analysis, so the sum of the rarity-weighted richness of our species in all cells world-wide equals

640. Because some species also occur outside Upper Guinea, the sum of all cells in Upper Guinea is 498. Since we subsequently multiplied each cell with Smax/Sobs, the total amount of points assigned to Upper Guinean cells became 1406. Because Upper Guinea has a conservation value of 498 species points, we rescaled the rarity-weighted species richness of each cell by multiplying it with 0.354 (= 498 / 1406), to arrive again at our original score of 498.

Table 6.1 gives an example of biodiversity calculations for two different cells. In cell A only a few, but very rare species have been found. In cell B more species have been found but they are also more common. Although less species have been found in A, its rarity-weighted richness is nearly four times as large as in B.

The rarity-weighted richness of a cell may vary from close to 0 (if only one species is found that occurs in many grid-cells) to 640 (if all 640 species happen to be found in the same cell and nowhere else). An interesting feature of this biodiversity measure is that it puts the floristic value of the cell into a regional, Upper Guinea-wide perspective. If a cell has a rarity-weighted richness of 10, it might indicate that it contains 10 species that are found only here, and nowhere else. Alternatively, it might contain 20 species, that occur only here, and in another cell somewhere in the world. Each cell has of course a different conservation value for different species, but the rarity-weighted richness is a good measure of the combined value of a cell for the conservation and long-term survival of individual species.

Table 6.1 Example of the calculation of rarity-weighted species richness for a fictive cell with a few but very rare species (cell A), and a cell with many common species (cell B). For eight species it is indicated how many individuals are found in cell A and B, in how many cells they occur, and their weighting score (weight).

Table 6.1 Example of the calculation of rarity-weighted species richness for a fictive cell with a few but very rare species (cell A), and a cell with many common species (cell B). For eight species it is indicated how many individuals are found in cell A and B, in how many cells they occur, and their weighting score (weight).

Cell A

Cell B

N

10

20

Sob.

4

6

Smax

16

15

2 weight

2.25

1.01

2 Weight X Smax/Sobs

9.00

2.53

rescaled Srw

3.20

0.89

Species

#ind

# cells

weight

#ind

#cells

weight

1

4

1

1.0

2

1

2

0.5

3

1

2

0.5

1

2

0.50

4

4

4

0.25

3

4

0.25

5

3

10

0.10

6

2

10

0.10

7

1

20

0.05

8

10

100

0.01

2

10

2.25

20

1.01

Figure 6.5 Distribution of herbarium collections (dots) of640 rare and endemic forest species in West Africa. The potential forest zone is shaded.

Figure 6.6 Sampling intensity (100 X Sobs/Smax) of640 rare and endemic forest species in West Africa.

Figure 6.5 Distribution of herbarium collections (dots) of640 rare and endemic forest species in West Africa. The potential forest zone is shaded.

Figure 6.6 Sampling intensity (100 X Sobs/Smax) of640 rare and endemic forest species in West Africa.

Results

Collection efforts; where are the white spots on the map?

Most collections of our shortlist of forest species are confined to the southern part of West Africa and indeed, closely follow the forest zone (Figure 6.5). Collection efforts have been particularly high near the capitals (Abidjan, Monrovia, Freetown) and near botanical research stations (see Appendix 5 for a map of West Africa). By far the highest number of collections have been made west of Abidjan. Both the Banco National Park and the research station in Adiopodoume fall within this half-degree square, resulting in 1303 collections belonging to 190 species from our shortlist. Also the cell north of Abidjan, that includes Yapo forest and Teke forest, and the cell containing the Ecological Station at Ta'i score rather high. In Liberia, next to the surroundings of Monrovia the area around Mt Nimba has been well collected. In Sierra Leone three cells have relatively many collections: the Peninsula (including the capital Freetown), the forestry research station at Njala, and the Kambui Hills Reserve. In Ghana the Ankasa Forest Reserve and the Atewa Range are the cells with the highest number of collections attributed to our shortlist species.

If we divide the observed species number by the predicted maximal species number, we get a measure of sampling intensity. Sampling intensity is particularly low (< 20%) in southeast Liberia (Figure 6.6). This area contains one of the largest remaining forest blocks in West Africa, and receives also a high amount of rainfall. It might therefore harbour a rich and unexplored flora, which definitely merits further attention. Southeast Sierra Leone, and southwest Ghana (north of Ankasa and Cape Three Points to Bia National Park) are two other large areas with a high potential species richness but few species sampled in our database. On a smaller scale it is striking that areas like the southeastern part of the Tai National Park and the Scio

Forest Reserve are so poorly explored that we can not even estimate the sampling intensity.

Hotspots of diversity

There is a north-south gradient in species richness, which coincides with the rainfall gradient; the S50 (Fig. 6.7A) and SmD (Figure 6.7B) increase from the Sahel towards the coast. A belt of high rare and endemic species richness is found about 50-100 km inland, starting in Sierra Leone, running through Liberia to southwest Côte d'Ivoire and then fading away towards Sassandra. The climax of this range lies in Liberia and southwest Côte d'Ivoire. Additional rich areas, east of this belt are found around Abidjan and Ankasa (Ghana). A second belt of high rare and endemic species richness can be found more inland around the montane area of Mt Nimba with extensions to Mt Ziama in Guinea and the montane area around Man in Côte d'Ivoire. The Atewa Range in Ghana could be seen as a far-out exclave of this species-rich montane belt.

Some areas in species-poor regions are relatively rich compared to their immediate surroundings or other areas with the same latitude. Examples are coastal areas in the Casamance region in Senegal and in Guinea-Bissau, the Fouta Djalon in Guinea, the Peninsula of Freetown in Sierra Leone, and Haut Sassandra Forest and Comoé National Park in Côte d'Ivoire (Figure 6.7A).

Patterns in rarity-weighted richness are comparable to the ones for species richness (Figure 6.7C), although the richness of montane areas becomes more pronounced compared to the coastal areas. Examples are Mt Nimba, Mt Ziama, Mt Tonkui and the Atewa range. Within the coastal rich band also the wettest forests become more pronounced, which is expressed in very high values at Tabou and Greenville.

Biodiversity vs. environment

What environmental factors give rise to a high species richness? S50 shows a curvilinear relationship with rainfall;

A S50

A S50

i ■ 13

Uïfl

»

» Jl

Jo -1+

*

li-JTi

V--J

Ii JTS

0

]T J . 4H>

m

Jtj ■

Ü

If. Ail-

1 U-4X.S

n

121 - 4S

H

1 If ■ If

hu [Mi*

f • * r

Figure 6.7 Biodiversity maps of West Africa showing A) the number of species at 50 collections (S;o), B) the maximum estimated number of rare and endemic species (Smax) and C) the rarity-weighted species richness (Srw). Biodiversity values of cells are interpolated over the whole potential forest zone of Upper Guinea.

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