Typology of cocoa-based agroforestry systems of the semi-deciduous forest zone in Togo (West Africa)

In the tropical zone, cocoa-based agroforest systems (CAFS) are considered as a mean to maintain and conserve biodiversiy. In the sub-humid zone of Togo (West Africa), agroforest plant species are key components of the landscape and agricultural lands. Cocoa and coffee agroforest systems contribute directly and indirectly to the livelihoods of an estimated one million people in Togo. Despite this fact, there is only few informations regarding their structure, and typology. The current study assessed the typology, tree structure and diversity of cocoa-based agroforest systems. 213 random plots across the study area were sampled using variable areas (25 × 25 m 2 , 50 × 50 m 2 and 100 × 100 m 2 ) for the survey. 4766 non-cocoa trees belonging to 195 plant species, 140 genera and 47 families were identified. Only woody trees were recorded during this study. The estimated average tree density was 159.21 ± 97.58 trees/ha, whereas the basal area was 54.19 m²/ha. Based on the Importance Value Index (IVI), the floristic composition, and the frequency of species, six groups (from G1 to G6) were discriminated. Each group was a particular type of CAFS. These results are similar to those obtained in the CAFS of West, Central Africa and other tropical zones, confirming CAFS key role in forest trees diversity conservation.


INTRODUCTION
Agroforests are qualified as agricultural lands of forest zone covered by more than 10% of woody trees (Zomer et al., 2009). In humid tropical areas, they are complex multispecies cropping systems whose value for farmers is often hard to assess (Jagoret et al., 2014). According to Correia et al. (2009), their multi-strata structure is shaped by a diversity of biological form and habitat (seedlings, woody trees, lianas and herbaceous) in order to make their ecological functions similar to that of natural forests (de Foresta et al., 2000).
The important role of agroforest systems in general and of cocoa-based agroforest sytems (hereafter referred to as "CAFS") in particular is well known over the world.
providing vegetables in the form of leaves and fruit from trees, but possibly also through the ecosystem services provided within agroforest systems likely support availability of wild and cultivated vegetables by providing the microclimates needed for vegetables to grow and other ecosystem services. Many studies demonstrated the direct link between CAFS and in situ biodiversity conservation in humid forest zones (Sonwa et al., 2007;Jagoret et al., 2008;Dumont et al., 2014;Wartenberg et al., 2017). CAFS are rich and diversified in terms of forest tree species (Sonwa et al., 2007;Sonwa et al., 2014). They control key ecological functions such as carbon storage, water and air regulation, soil protection and fertility maintenance, micro-climate regulation and shading. From smart agriculture perspective, it is widely accepted nowadays that CAFS in general could serve as useful and pratical models to mitigate climate change issues in tropical zones, by reconciling productivity increase of lands, biodiversity conservation and use of forest biodiversity (Schroth et al., 2016). From food safety, nutritional and rural livelihood perspectives, the potential of agroforestry products to contribute to incomes, and value for domestic consumption is wellrecognized (Cerda et al., 2014). Data on local flora of CAFS contribution in nutrition, human well-being and household incomes were published for different agroforestry products such as cocoa, timber, fruits, and other food crops (Akinnifesi et al., 2008;Cerda et al., 2014). In Togo CAFS are found in the sub-humid zone of the country, and play key important socioeconomic and ecological roles (Adden, 2017). Despite their importance, only few studies assessed their typology, structure and floristic diversity. Most of existing studies did not consider tree diversity and species richness even if some of them recognized the conservation value provided by CAFS and qualified them as potential biodiversity conservation areas (Wembou et al., 2017;Sodjinou et al., 2019). Their assessment is often discussed in terms of productivity, yields and soil properties (Adden et al., 2016) and cocoa diseases (Oro et al., 2012) without qualitative and quantitative data on multi-species which shape CAFS multi-strata. Past studies in the sub-humid zone in Togo were focused on forest investigation (Akpagana, 1989;Adjossou, 2009;Issa, 2018) with less emphasis on CAFS. Up to now, research on biodiversity in agroforest systems are related to coffee based agroforest systems (Koudjega and Djiekpor, 1997;Koda et al., 2019).
Consequently, knowledge about CAFS relationships with biodiversity is still scarce compared to those in West Africa and other tropical cultivation zones. This gap causes a lack of information in the management of CAFS, which contrasts with their ecological and socioeconomic potential.
To meet both productivity needs, pest and disease management, Koudjega and Tossah (2009) and Wegbe and Agbodzavu (2013) recommended to associate cocoa Djiwa et al. 271 trees with particular forest tree species such as Albizzia spp., Erythrophleum guineensis, Eleais guineensis, Citrus spp., Cola nitida, Khaya spp., Milicia excelsa, Samanea saman, Terminalia spp., in Togo. This introduction of trees species needs to be based on a good understanding of the current species composition and typology of CAFS. The objective of this study is to analyze the tree diversity within and among CAFS in Togo and to determine their structure and typology as related to sustainable management practices.

Study area
This study was carried out in the sub-humid and mountainous zone in Togo, located at the south-western part of the country and bordering Ghana. It extends between 6°57 and 7°35 latitude North and 0°30 and 1°08 longitude East ( Figure 1). This zone belongs to ecological zone IV (Ern, 1979) and is characterized by a transitional subequatorial climate. However, this type of climate is more Sudanian in the north of the study area because of harmattan and Foehn effect. It is the wettest area of the country, with mean annual rainfall comprising between 1,250 and 1,500 mm and temperatures varying between 22.5 and 26°C. The long rainy season period is about eight (8) months and extends from March to October. Geologically, the study area is part of the Atakorian mountains and the main structural unit is composed of epimetamorphic rocks. The dominant soils are ferruginous tropical soils, ferrallitic soils and hydromorphic soils, according to the French classification system (CPCS, 1967). The vegetation is dominated by a mosaic of relics of dense forests, savannahs, fallows and agroforests (Adjossou, 2004). Cocoa and coffee agroforests mainly dominate these agroforests. Riparian forests and remnant semi-deciduous forests of the study area are still the biodiversity hot-spot in Togo (Kokou et al., 2008;Issa, 2018 ;Sodjinou et al., 2019).
Agriculture is the main activities in the area due to soil fertility and moist climate conditions. This favours cash crops, food crops, fruit plantation and gardening (Adden, 2017;Koglo et al., 2018). However, the area is facing agronomic and environmental issues, which are soil degradation and crop yield decrease (Koglo et al., 2018). Illegal wood logging of forest trees species (Milicia excelsa, Khaya grandifoliola, Terminalia superba, Triplochiton scleroxylon, Antiaris toxicaria, etc.), and charcoal production are some factors of forest lands degradation (Kedjeyi et al., 2013). The area is part of the Region Plateaux (the highest producer of cash crops such as coffee, cocoa, cotton as well as vegetables), which is also one of the regions where there is intensive use of chemical pesticides for agricultural production enhancement in Togo (Kolani et al., 2017).

Data collection
Vegetation data such as names of woody plant species, trees height and diameter were collected across eight (8) Prefectures in 62 villages; approximately three representative cocoa agroforests are selected randomly in each village. 213 random plots across the study area were sampled using a variable area method (25 × 25 m 2 , 50 × 50 m 2 , 100 × 100 m 2 ). The variable sampling method was used as basis of woody species inventory in CAFS in Côte d'Ivoire (Vroh et al., 2015). According to the author, this method has the advantage of analyzing results independently to the sampling size. Accordindg to DSID (2019), in the sub-humid zone of Togo, the size of CAFS varies from 0.125 ha to 5.5 ha. For the current study, in CAFS with size less than 1 ha we surveyed 39 plots of 25 × 25 m 2 ; while 41 plots of 50 × 50 m 2 were sampled in CAFS with 1 to 2 ha size. In CAFS with size higher than 2 ha, 134 plots of 100 × 100 m 2 were sampled. The total surveyed area was 157.81 ha. For practical purposes, when the farm's size exceeded 2 ha, the sample area was divided into elementary plots of 25 × 25 m 2 , which were randomly located in the agroforests, according to Sonwa et al. (2014) and Vroh et al. (2015).
In each elementary plot, trees (excluding cocoa trees and including exotic fruit trees) with a diameter at breast height (dbh) ≥ 10 cm were recorded. Cocoa trees were also measured for their dendrometric characteristics assessment and their basal area calculation by considering a dbh ≥ 5 cm within sub-plots of 10 m × 10 m (100 m²) inside 25 × 25 m 2 plots. Such a sub-plot size is adopted to take advantage of different ages of cocoa trees as focus species of this study.
The total height of recorded trees was estimated with a clinometer while the diameter was recorded using a diameter measuring tape. The nomenclature for plant species identification was based on Brunel et al. (1984)'s Flora of Togo, and Akoègninou et al. (2006)'s Flora of Benin, and in the same way by comparing collected samples with some specimens at the National Herbarium (University of Lomé). Only woody trees were recorded during inventories. The prospected sites were geolocated by a GPS Garmin 64S.

Plant species richness, occurrence, and α-diversity
The plant species list recorded in the CAFS was compiled and stored in the Comma delimited (CSV) format. Plant species richness (S), occurrence, frequencies (Fr) according to Mori et al. (1983), number of species per family and specific abundance were analysed through pivot table "plant species" versus "plot" in Excel ® .
The α-diversity was evaluated by computing the Shannon-Wiener index (Ish), the Pielou's evenness (Eq) and the species pool (S) (Hill, 1973 ;Kent and Coker, 1992 ;Magurran, 2004). Frequent and rare plant species in CAFS were determined by computing the Rarity Index (Ra) (Géhu and Géhu, 1980). Following Adomou (2005), species were considered rare when the Ra is lower than 80% and frequent when the Ra is higher than 80%.

Biological forms, life forms and phytogeographic spectra
The plant species were classified into their life forms and chorological affinities according to White (1983). The biological and phytogeographical spectra were computed. Table 1 summarizes

Pielou's evenness (Eq)
Where S is number of plant species recorded, ni is number of plots where plant species i is present, and n is total number of sampled plots.

Density of trees (D) in trees/ha
Basal area (Gi) in m 2 /ha Importance value index (IVI) in % Relative density (Der) in % Relative dominance (Dor) in % ∑

Rarety index (Ra) in % 100 -Fr
where N is total number of individuals (trees) recorded, A is area (ha), dbh is diamter at breat height (cm), Ni is number of indivuals (trees) of plant species i and Fr is relative frequency, Der is -----, and Dor is ----.
formulas used for the α-diversity calculation.

Structural parameters
The main structural parameters considered are density (D) and basal area (Gi). The importance value index (IVI) for each tree species was then calculated according to Curtis and Mcintosh (1950). Table 2 summarizes formulas use for structural parameters calculation. Horizontal and vertical distributions were obtained by grouping individuals into height and diameter class sizes by considering a pitch of 2 m and 5 cm, respectively, using Minitab (2000) and Excel. The distributions were adjusted with the Weibull theoretical function because of its relevance in structural parameter prognosis (Miguel et al., 2010). The 3-parameter Weibull theoretical density function was computed as follows: where x is tree diameter or height, a is location parameter, b is scale parameter and c is shape parameter of the structure.

Analysis of the typology of CAFS
A Gradient Analysis Method of direct Canonical Correspondence Analysis (CCA) in Canoco ® 4.51 was performed on the "plant species" versus "plot" binary table. The Ward method was applied and focused on inter-species distances (Ter Braak and Smilauer, 2002), and analysis based on Whittaker's Index of Association (WIA) was used to adjust results with Community Analysis Package (CAP ® ) (Pisces Conservation, 2002;Legendre et al., 2005). The inter-species distance approach offered more flexibility, allowing easy visual analysis of patterns in ordination biplots for response data. The WIA is expressed as the fraction of the total number of individuals in the sample to measure similarity distance and is computed as follows: where i is the number of plots holding species i and P is the total number of plots.

Richness, floristic diversity and species frequency of CAFS
In total, 4766 non-cocoa trees were recorded within  Tables 3 and 4 provide the summary of frequent woody species and the structural 6xcels6na6ic6c of the CAFS identified in this study.

Typology and characteristics of CAFS
Based on the hierarchical Clustering Analysis and applying agglomerative method of Ward, the 213 sampling plots were discrinminated into six groups of plots at the threshold of 3.0 Whitaker distances. The categorization of each group was also based on the Importance Value Index (IVI) and the 6xcels6na6ic structure of the CAFS (Figure 3).

CAFS with F. mucuso and Ceiba pentandra (G2)
The total number of plots of this group is 49. The type of CAFS is dominated by F. mucuso (IVI = 139.71 %) and Ceiba pentandra (IVI =83.60 %). In terms of diversity, the species richness of this CAFS type is 75 plant species; and the Shannon-Wiener index and Pielou's evenness were 5.31 and 0.85 bits, respectively. The computed structural parameters showed 226 trees/ha as tree density, and the basal area was 32.45 m²/ha. Considering phytogeographical affinities, the Guineo-Congolian species (52 %) were the most frequent. Microphanerophytes and mesophanerophytes were represented by 45.33 and 37.33 %, respectively ( Figure 5).   (Figure 4).

CAFS with P. americana and T. superba (G5)
This CAFS type is composed of 43 plots. The group is characterized by P. americana (IVI = 79.88 %) and T. superba (IVI = 79.11 %) as important plant species. The highest plant richness was recorded in this group and was 122 plant species. The Shannon-Wiener index and Pielou's evenness were 6.44 and 0.93 bits, respectively. The density was 58 trees/ha and the basal area was 24.12 m²/ha. Considering phytogeographical affinities, Guineo-Congolian (GC) and exotics (I) were the most represented. Their values are 33.60 and 21.31%, respectively. Microphanerophytes (57.37 %) followed by mesophanerophytes (24.59%) were the most represented life forms (Figure 4).

CAFS with C. pentandra and M. excelsa (G6)
This cluster is the smallest with 18 plots. The most      Tables 5 and  6.

Diameter class distribution
The distribution of CAFS trees by diameter class-size follows an "L"-shaped or "bell"-shaped curve ( Figure 5). The "L"-shaped distribution is mainly observed in G1 and G2; and is characterized by high density of trees of small diameter size (5 to 20 cm). The predominance of medium diameter size trees, between 20 to 50 cm diameter, was observed in the CAFS with large tree species such as Albizia spp., Ficus mucuso, M. excelsa and A. toxicaria. Large tree dimater size (up to 50 cm) is less represented (e.g., Adansonia digitata, C. pentandra, Cola gigantean, etc.). The calculated mean diameters of each of the six groups are in the following order: 42.36 ± 17.21 cm, 39.54 ± 23.45 cm, 31.54 ± 21.07 cm, 29.66 ± 11.54 cm, 28.32 ± 12.38 cm and 43.55 ± 19.23 cm.

Height class distribution
The distribution of CAFS trees by height class-size follow an "L"-shaped or "bell"-shaped curve ( Figure 6). This structure reflected the dominance of trees with small (less than 6 m) and medium height (6 to 12 m). The smal heights are manly found in CAFS type 1 and 2. The height class from 6 to 12 m is the most represented from G2 to G6. In general, heights greater than 20 m are less represented except in the CAFS dominated by C. nitida and A. adianthifolia (G3 and G6). In each of the six CAFS types the mean height is in order 9. 67 ± 2.22 m, 17.78 ± 7.53, 18.63 ± 5.04, 17.11 ± 6.21, 14.13 ± 6.24 and 16.29 ± 5.69 m.

G6 G5 G4
instance, the Shannon-Weiner index reported from cocoa agroforests in Ghana varied between 4.69 and 4.76 bits (Anglaaere et al., 2011). In the semi-deciduous forest zone of centre Côte d'Ivoire, the diversity was low.
The Shannon-Wiener index varied between 1.5 and 4.2 bits and the plant specific richness ranged from 56 to 97 plant species (Vroh et al., 2015). The number of species in this study is comparable to the reference values found in Nigeria. Bobo et al. (2006) and Zapfack et al. (2002) located the specific richness between 116 and 206 woody species. Osei-Bonsu et al. (2003) and Anglaaere et al. (2011) inventoried less woody species in sub-humid zone of Ghana comparable to our study. Their values were 116 and 82 woody species, respectively.
Togolese cocoa agroforests were characterized by the predominance of Moraceae, Mimosoideae and Caesalpinioideae. This is consistent with that of Vroh et al. (2015) in cocoa agroforests in Côte d'Ivoire. There was also a great similarity between species composition. Some species such as P. americana, C. nitida. Irvingia gabonensis, Margaritaria discoidea, Spondias monbin etc. were well represented in cocoa agroforests in Togo as well as in Ghana (Anglaaere et al., 2011).

Typology of cocoa-based agroforestry systems
Six types of CAFS were discriminated based on importance value indices, species richness, and structural parameters of species. However, some other ecological parameters such as temperature,, hygrometry, topology, soils etc., were under the control of this typology (Bongers et al., 2004). The drying-out from south to the north on sub-humid forest in the zone might also have had an influence (Akpagana, 1989) on the species composition, and the density of the defined groups. To avoid such drying-out on cocoa productivity, farmers maintained in their CAFS some native trees or introduced new exotic or native trees. CAFS with M. excelsa and P. americana (G1) and CAFS with C. nitida and A. adianthifolia (G3) were mainly shaped by multipurpose trees. Specialy, P. americana, C. nitida and other fruits trees found in the CAFS were not only present as shading species, but also for commercial purpose and for economic benefits. It is widely accepted that trees in CAFS had important value for farmers and were used as medicinal, nutritional plants, or as a source of indirect cash from the agroforests (Akinnifesi et al., 2008;Koda et al., 2016). This added value of plant species may also explain the predominance of exotic plant species in some discriminated CAFS. Our findings were similar to those obtained by Vroh et al. (2015). These authors also remarked about this co-dominance of exotic and plant species in old agroforests in Côte d'Ivoire. Besides fruits species, some CAFS, such as sub-systems with P. americana and T. superba (G5) were mainly characterized by timber-tree species.
Indeed, some species such as T. superba and Khaya spp. were recommended to be included in some density rate by agroforestry technical services during their training, awareness and sensitization. Some tree species were completely avoided due to the fact that they were hosts of diseases or infection vectors (Oro et al., 2012). At the same time, some species such as C. gigantea were avoided in some CAFS, because they could be shelter for rodents that are a serious threat to cocoa production.

Structure of CAFS
The density of non-cocoa trees in CAFS is relatively low and the distribution of diameter class-size revealed the predominance of individuals with small diameters. This might be explained by the fact that trees are removed progressivelly for use (house-construction, fuelwood, timber, etc.), to regulate shadow for cocoa trees or avoid windthrow that can damage cocoa trees. One of the consequences of illegal logging in CAFS could be drought severity leading to high death rate of young cocoa trees, especilay during the dry season. Sonwa et al. (2014), in this persective, demonstrated that human interventions such as frequent use of fire, intensive use of fertilizers and pesticides, and lack of tree cover are expected to cause negative effects on cocoa farms. According to famers, to prevent low cocoa productivitiy, or to avoid pest and diseases in cocoa trees, some species such C. gigantea, C. pentadra, M. excelsa, Cola nitida, etc. are removed from CAFS. These practices may explain the low density and the rarety of some woody species in some groups.

Conclusion
The species richness in cocoa agroforest systems in Togo was 195 plant species. Based on the Importance Value Index (IVI), six types of CAFS were identified. They were sub-sytems (from G1 to G6) shaped, respectively, by M. excelsa and P. americana trees (G1), F. mucuso and C. pentandra (G2), C. nitida and A. adianthifolia (G3), Albizia spp. and M. excelsa (G4), P. americana and Terminalia superba (G5), and C. pentandra and M. excels (G6). Tree species richness and diversity were high in the CAFS associated with P. americana and Terminalia superba (G5) (species richness = 122, Shannon-Wiener index = 6.44 bits); but much lower than in any other CAFS for the case of CAFS associated with C. pentandra and M. excelsa (G6) (Species richness = 32, Shannon-Wiener index = 4.36 bits). The highest (52.60 m 2 /ha) and lowest (11.67 m 2 /ha) basal areas were recorded, respectively, in CAFS dominated by M. excelsa and P. americana (G1) and CAFS associated with C. pentandra and M. excelsa (G6). These findings, and those of other published work referenced here provide