CHAPTER I
AGROFORESTRY PARKLAND SYSTEMS

Parklands:
A review of definitions and terminology

Parklands are generally understood as landscapes in which mature trees occur scattered in cultivated or recently fallowed fields (Pullan, 1974; Sautter, 1968, cited in Raison, 1988). In the ICRAF Agroforestry Systems Inventory, agroforestry parklands are included in the very general category of ‘multipurpose trees on farmlands’ (Nair, 1985). Livestock production may be a significant or secondary component in these systems. Because of the variety of field realities encompassed, the notion of parklands has been interpreted widely and its terminology is still very much under discussion.

On the one hand, parklands have been referred to as a vegetation type. They are physiognomically comparable to ‘tree savannas’ and have been referred to as such. The terms ‘savanna parkland’ and ‘park savanna’ or ‘parklike savanna’ are sometimes used by phytogeographers. Specifically, Cole (1986) defined savanna parklands as “tall mesophytic grassland (grasses 40–80 cm high) with scattered deciduous trees (less than 8 m high)” in a savanna classification first proposed in 1963. This vegetation type is intermediate between savanna woodlands, “deciduous and semi-deciduous woodland of tall trees (more than 8 m high) and tall mesophytic grasses (more than 80 cm high)”, and savanna grasslands, defined as “tall tropical grassland without trees or shrubs”. In Cole's analysis, savanna parklands occur in Australia and in Africa, mostly in Central and Southern Africa, but not in South America, with the exception of the region of the High Pantanal. Forests with a very open canopy and a history of frequent fire in Western Canada and northwestern United States are also called parklands (British Columbia Ministry of Forests, 1991).

The term ‘parkland’ as used in this report specifically applies to landscapes derived from human agricultural activities.

However, while tree savannas can occur naturally or as a result of edaphic features, fire and grazing in the absence of cultivation, the term ‘parkland’ as used in this report specifically applies to landscapes derived from human agricultural activities (Pullan, 1974). In parklands the composition and density of the woody vegetation is altered in order to facilitate its use. Most often parklands are not the product of a single agricultural season, but reflect a slow process of species selection, density management, and tree growth over one or several decades. Parklands, in the strict sense of the word, are specific to permanently cultivated fields or fields where fallow duration is shorter than necessary for the regeneration of a second-growth forest (Seignobos, 1982; Raison, 1988). They would not, therefore, include the relic of a natural forest temporarily left standing by frontier farmers (Pélissier, 1980a).

Parkland trees stand out as an important component of the spatial structure of the landscape (Sautter, 1968, cited in Raison, 1988; Seignobos, 1982). Parkland attributes include a regular distribution of relatively even-aged trees or shrubs and a low tree density so that tree cover is never continuous. Their name derives from their resemblance to urban or rural recreational parks with large scattered trees in expanses of grass. Examples of parklands are more common in the semi-arid or subhumid tropics, and particularly in West Africa which will be the main subject of this report as it is of most of the literature reviewed in it.

Parklands are not, however, limited to the Sahel and Sudan zones of Africa. While they may not generally be called parklands, systems with scattered trees in fields with similar appearance and purpose are also widespread in Zimbabwe (Campbell et al., 1991) and Malawi (Maghembe and Seyani, 1991) and elsewhere in southern Africa. Several systems in Asia, Oceania, and Latin America would qualify as agroforestry parklands by definition. Information on practices in these areas is more limited, however (Baumer, 1994; Raison, 1988). In India, the well-known Prosopis cineraria is commonly protected in fields planted with millet and legumes, and occurs on fallows and grazing lands in the semi-arid zone of Rajasthan (Mann and Saxena, 1980). Farmers value its high ecological combining ability and suitability for pruning and fodder as well as its socio-cultural significance. Other species found in these systems include Ziziphus nummularia, Acacia nilotica var. cupressiformis and var. radiane, Acacia leucophloea, and Salvadora persica (Shankarnarayan et al., 1987). In the state of Tamil Nadu, several tens of multipurpose tree species are naturally regenerated or planted on farmlands (Jambulingam and Fernandes, 1986). To cite only a few, Borassus flabellifer is grown with cereals and pulses, A. leucophloea with millet and horsegram, and A. nilotica in rice fields.

Parkland or related systems also exist in temperate regions. Dehesa (Spanish) or montado (Portuguese) systems are centuries old, mainly silvipastoral systems in southwestern Spain and southeastern Portugal where holm oak (Quercus rotundifolia, syn. Quercus ilex) and cork oak (Quercus suber) are scattered in pastures or cereal (oats, barley, wheat) fields (Joffre et al., 1988; Janick et al., 1987). Trees are sometimes seeded and systematically pruned for better fruit and wood production. They provide highly nutritious acorns for fodder and timber, charcoal, tannin, cork (Q. suber), etc. With a density of 20 to 40 trees per hectare, the tree cover may be 5 to 20 percent. This system extends over an estimated 5 million ha in Spain and more than 500 000 ha in Portugal. Similar systems with the same or different species (olive, carob, etc.) also exist in other Mediterranean countries including Morocco, Algeria, Tunisia, southern France (mainly Corsica), Italy (Sardinia) and Greece (Joffre et al., 1988). Moreover, the Acacia caven silvipastoral system is widespread in the semi-arid and subhumid Mediterranean climate zone of Chile, covering 1.5 million ha (ODEPA, 1968, cited in Ovalle and Avendano, 1987). Walnut plantations in the Touraine and Berry regions of France may be included in the parkland family (Raison, 1988). The oak savannas of the eastern United States (New York State) may also be a case of parklands with a silvipastoral focus.

The term ‘agroforestry parkland’ emphasizes the multiple forms and purposes of agroforestry systems.

Among ‘parcs arborés’, the usual French term for parklands, Baumer (1994) distinguished between systems primarily used for cultivation and those used for pastoralism. He suggested ‘forêt-parcs’ as the term for parklands resulting from a high degree of human influence, where trees tend to be monospecific, even-aged, regularly spaced, and form a cover ranging from 1 to 25 percent, and where annual crops, generally cereals, are cultivated, sometimes manured, and grazed after harvest. These trees usually have a positive ecological role (soil fertility or wind reduction) and a strong economic significance. In contrast, ‘parcs arborés’ would describe systems with a parkland appearance but which have not been heavily manipulated by human beings, and are primarily used for pastoralism and gathering of tree products. Examples of such vegetation types include cattle-raising areas in Brazil, such as the open forests of the Sao Paulo and central-east regions, highly forested grasslands of southeastern Queensland in Australia, Miscanthus grasslands under forest relics in Indonesia, and the Daniellia oliveri open dry forests of East Africa. Although the term ‘parcs arborés’ only draws attention to the tree component of the landscape, the term ‘forêt-parcs’ is more problematic. ‘Forêt’ suggests a low degree of human intervention and conceals the major distinction between this and other systems, namely that canopies are spatially scattered rather than forming a closed cover (Depommier, 1996a).

The major English writings on this topic have used the word ‘parkland’ (Pullan, 1974; Bonkoungou et al., 1994, 1996). Sometimes, however, it is used to describe lands in parks, which bear no relation to the agroforestry parklands considered in this report. The term ‘park’ is sometimes found as a direct English translation of the French ‘parc’, but this is even less explicit than ‘parkland’ and is easily confused with the more common meaning of park as an enclosed or delimited land area managed for preservation or recreation. The term ‘farmed parkland’ is used by Pullan (1974) to encompass parklands being farmed as well as land lying fallow, and does not exclude a pastoralist component. Other labels such as ‘farm parkland’, and ‘cultivation parkland’ have also been used but are less common. The term ‘agroforestry parklands’ incorporates these systems in the fast-growing discipline of agroforestry. It was chosen for this study because it emphasizes the multiple forms and purposes of these systems, and thus includes various schools of thought. As defined by Bonkoungou et al. (1994), agroforestry parklands “are land-use systems in which woody perennials are deliberately preserved in association with crops and/or animals in a spatially dispersed arrangement and where there is both ecological and economic interaction between the trees and other components of the system”. This emphasis on interactions is positive and helpful for a finer understanding of the systems by the research and development (R&D) community, but may be extraneous from a farmer's holistic point of view.

Table 1.1 Scientific, English and French names of common parkland trees

As mentioned in the introduction, agroforestry is an ancient practice but a relatively new science. In order to assess the significance of the parkland system, it is important to characterize it physically, determine its zone of occurrence, and identify and describe existing parkland types. The next section sets the geographical stage, beginning with a review of the main historical studies of parklands. This is followed by a presentation of the main modes of investigation and existing data, on the basis of which a tentative picture of the wide spatial extension of agroforestry parklands and the distribution and density of prominent parkland species (see Table 1.1) is drawn.

Parklands in West Africa

Being linked to human activities, parklands occur in various latitudes and are not confined to specific agroecological zones. Nonetheless, the most well-known and described parklands are located in semi-arid or subhumid zones where tropical savannas, broadly defined as tropical grasslands with scattered trees (Bourlière and Hadley, 1983), occur.

We are indebted to the early explorers for the first references to African parklands (Pullan, 1974). Mungo Park described the frequent occurrence of Vitellaria parklands along the Niger River from Segou towards the east in Bambaraland (Miller, 1954), while Caillié (1830) noted the various tree species around houses and Vitellaria and Parkia trees in the surrounding agricultural landscape. On his travels to Hausaland in Nigeria, Clapperton (1829) described well-maintained Vitellaria, Parkia and Tamarindus trees located in agricultural lands. The early reports also focused on how products from these species were prepared for various human and animal uses. In the late 1800s, scientific and commercial interest in savanna and parkland species grew. Plant collections were undertaken and tropical African floras prepared, while tree products were sampled and investigated for commercial use in colonial countries of the North.

Their origin in people's agricultural activities has given parklands a peculiar scientific status amongst plant communities. Pullan (1974) noted that plant ecologists have generally neglected the study of vegetation in cultivated zones. Thus the degree to which parklands were recognized as distinctive vegetation types, or had their importance in successional stages of vegetation regrowth acknowledged, was variable among early plant scientists. Recognizing the widespread occurrence of Faidherbia albida¹ in the vegetation of southwestern Senegal, Trochain (1940) characterized it as a climax association (or, more specifically, a stable plant community of anthropogenic origin, which replaces the climax vegetation, including various forms of open forest found in the Sudan zone, on silica and clay-rich soils), developed after the species was introduced by nomadic people and their livestock. Roberty (1956) established distribution maps of parkland types in western Senegal (Fig.1.1), distinguishing major species in part based on a classification he established in the Middle Niger Valley. Pélissier (1953, 1966), Savonnet (1959), and Gallais (1967) also described F. albida parklands in Senegal, southern Burkina Faso and the inland delta of the Niger River, respectively.

¹ The scientific name of Faidherbia albida has been discussed for many years. The species was first named Acacia albida by Delille in 1813. Baillon observed in 1863 that its staminal characteristics departed from the pattern usually found in the Acaciae tribe. This distinction which relates it more closely to the Ingeae tribe, as well as variations in phenology, leaf, cotyledon, pollen and wood anatomy reviewed in CTFT (1988), led Chevalier in 1934 to propose a new monospecific genus Faidherbia located between the Acaciae and the Ingeae tribes. This name has been adopted by many authors since, but not unanimously. In order to emphasize its divergence from members of the Acacia genus, the use of Faidherbia albida has been recommended until its taxonomic position is clearly established (CTFT, 1988). Faidherbia albida is therefore used in this report.

Fig. 1.1 Distribution and types of farmed parkland in western Senegal (Source: after Roberty, 1956, shown in Pullan, 1974).

Lely (1925) recognized a parkland type dominated by Parkia, Vitellaria, Afzelia, Tamarindus and Acacia species in Nigeria. Vegetation maps for northern Nigeria (Clayton, 1962, 1963) took into account Keay's (1959) contribution to the recognition of local parkland types. The Samaru Soil Survey Bulletins (1956-71) also mention parklands in northern Nigeria including Parkia, Vitellaria, Ficus, Balanites, Adansonia, Hyphaene and Borassus species and F. albida. Three types of parklands, dominated by F. albida, Vitellaria-Parkia, and Parkia species, are described by FAO (1969) in northwestern Nigeria. In the northeastern part of the country, Leeuw and Tuley (1972) mapped F. albida, Adansonia and Parkia parklands, while Ceiba parklands were reported by Jackson (1970) around Zaria.

In northern Ghana, Taylor (1960) identified Vitellaria-Parkia-Tamarindus and Adansonia-Tamarindus parkland associations, with small parkland patches of F. albida. Irvine (1961) also referred to mature F. albida, Parkia and Borassus parklands, and Vitellaria-Parkia communities were described by Ramsay and Innes (1963). Wills (1962) also referred to farmed parklands in this area. Early citations for Niger include Dundas (1938) and Fairburn (1945), yet the relationship of the described vegetation to parklands was not clearly made (Pullan, 1974). Parkland communities dominated by Vitellaria, Parkia, F. albida, Borassus and Hyphaene were reported by Pias (1955) in the Chari-Logone lowlands of Chad and Cameroon.

Parklands did not become a major object of study until fairly recently, perhaps because their definition did not fit squarely into any single discipline. Thus, as noted by Pullan (1974), they fell outside of uncultivated plant communities, the ‘classical’ field of study of plant ecologists. Similarly, in the early days there were only isolated scientific investigations of parklands by agronomists or foresters, the latter having focused on natural forests and industrial or village plantations. The first in-depth system descriptions were in fact undertaken by geographers (Sautter, 1968, cited in Raison, 1988; Pélissier, 1964, 1966). It was only the emergence of agroforestry as a scientific discipline in the 1980s that finally established the study of parklands in their own right.

The Semi-Arid Lowlands of West Africa (SALWA) Programme of the International Centre for Research in Agroforestry (ICRAF) has contributed to gathering and generating information on parklands in its member states: Burkina Faso, Mali, Niger and Senegal. The Macro- and Micro-Diagnostic and Design exercises conducted in the first half of the 1990s advanced the identification, description and general analysis of the various agroforestry land-use systems on a national basis (see for instance ICRAF, 1990). National parkland reviews carried out within the African Research Network on Agroforestry (AFRENA) in SALWA countries (Sall, 1996; Cissé, 1995; S.J. Ouédraogo, 1995; Ounteni, 1998) represent an important step in synthesizing knowledge on these systems. Maps and parkland typologies have also been produced by ICRAF (1996) for the Dori area, Burkina Faso, and the middle Bani-Niger river basin and Gondo-Mondoro region of Mali.

Areal extent

Parklands constitute the predominant agroforestry system in semi-arid West Africa.

Parklands occupy a vast land area, representing a large part of the agricultural landscape under subsistence farming in the tropics and constituting the predominant agroforestry system in semi-arid West Africa (Nair, 1993; Bonkoungou et al., 1994). In Mali, the agroforestry parkland system occupies about 90 percent of the agricultural land area (PIRL, 1988, cited by Cissé, 1995) and is practised by an estimated 2.5 million people on the Mandingue and Koutiala plateaux and the Moyen-Bani-Niger, High Dogon Plateau, Seno, Gondo, Bélédougou, Wenia, Falo and Central Delta zones (Djimdé, 1990). In Burkina Faso, parklands are found throughout settled zones where agriculture is practised, i.e. most of the country with the exception of the extreme North, East, and parts of the South and Southwest where human population density is low (S.J. Ouédraogo, 1995). The parkland system is also recognized as the most common production system in Katsina State in northern Nigeria (Otegbeye and Olukosi, 1993). A discontinuous cover of scattered trees in crop fields is traditional in northern Ghana (Rudat et al., 1996). These references highlight the local or national significance of parklands but also serve to illustrate the lack of a coordinated quantitative assessment of this land-use system at the regional level.

Characterization by dominant species

Nowadays, agroforestry parklands are most often characterized by the dominance of one or a few species. Species composition is generally more diverse and variable, however, in areas located farther away from villages and only occasionally cultivated. In parklands, one or a few dominant species may prevail on a local scale or across large land areas with substantial variations in relative abundance, frequency and overall species composition. Thus, parklands are often described by their dominant species (Pullan, 1974; Weber and Hoskins, 1983). Table 1.2 provides a tentative list of the main species in each climatic zone. Seignobos (1982), however, notes that some parklands, such as those in the northern Mandara mountains in Cameroon or around Kimré in southern Chad, include a large diversity of species without apparent dominance. Otegbeye and Olukosi (1993) also report that parklands with a species mix without specific dominance are the most common types in Katsina State, northern Nigeria.

Whether or not there is a dominant species, parklands usually host a wide variety of tree and shrub species. For instance, 22 and 39 species were recorded in cultivated fields in two sites around Kano, northern Nigeria (Cline-Cole et al., 1990), 43 and 46 in north-central and southern Burkina Faso (Gijsbers et al., 1994; Boffa, 1995), and 46 in northern Côte d'lvoire (Bernard et al., 1996). Species diversity increases when fallows are included.

There is very little quantitative information regarding the relative representation of major species throughout the parkland range which could help to prioritize conservation and development efforts in a rigorous way. Faiderbia albida parklands have received considerable attention because of the generally observed positive effects of the tree on soil fertility and crop production. It is likely, however, that V. paradoxa and P. biglobosa parklands occupy the largest land area among parkland types. Breman and Kessler (1995) indicate that V. paradoxa may be the most common parkland species in semi-arid zones of West Africa. Prioritization exercises conducted by ICRAF (Franzel et al., 1996) in selected sites have been useful in assessing the importance of parkland species in the eyes of farmers and thus guiding domestication programmes. There is a definite need for more quantitative data on the main parkland species at the regional level, including their distribution, stocking rates and dynamics, to improve the formulation of parkland research and development activities. Some methods for assessing parkland resources are reviewed in Box 1.1

Table 1.2 Dominant species in West African parklands by climatic zone

(Source: adapted from Pullan, 1974)

The rest of this section reviews recent information on individual parkland species throughout their zone of occurrence. Species are reviewed in order of their importance in the literature. Where parkland studies included information on tree densities, these are listed in Tables 1.3, 1.4 and 1.5 for F. albida, V. paradoxa, and P. biglobosa respectively. Other species mentioned may seem less widespread but are equally important to farmers. Reports of particular species in agroforestry parklands are generally localized, but their wider distribution is by no means limited to these few geographical references.

Faidherbia albida

Faidherbia albida parklands occur throughout the Sahel and Sudan zones of West Africa, as well as in eastern and southern Africa where it is strongly associated with alluvial soils along perennial or seasonal watercourses. In West Africa, F. albida still shows a preference for alluvial soils but occurs widely as a result of human activities on deep light sands or sandy clays. The African distribution of the species and the two geographical races identified by Brenan are shown in Figure 1.2 below. Because of its unusual characteristic of reverse foliation (i.e. bearing leaves during the hot dry season), it is water-demanding and has a deep fast-growing taproot to reach the aquifer. It occurs in areas with between 500 and 800 mm rainfall and is also found on lateritic soils where it can reach the water table through an opening in the shallow indurated pan. It also makes incursions into the northern Sahel and Sahara in moist sites or areas with a good water table (CTFT, 1988; Wickens, 1969).

Fig. 1.2 Distribution of Faidherbia albida (Source: adapted from Wickens, 1969)

Faidherbia albida, as a parkland species, is present in virtually all of Senegal (Fig. 1.3) from the Atlantic coast to the Falémé river and from the Senegal river to the Guinea Bissau border (Giffard, 1964). A highly integrated form of the F. albida parkland system among the Sérer in Senegal is described in depth by Pélissier (1953, 1966). This system is also present among the Wolof (Seyler, 1993) and the Mandingue of Casamance (CTFT, 1988). F. albida parklands are most common in the western part of the country on sandy soils within the Thiès-Louga-Kaolack triangle (Portères, 1952; Seyler, 1993; Sall 1996). In Guinea Bissau, they occur among the Brame and Mandjak ethnic groups (Pélissier, 1980a). In Mali, they extend over an estimated area of 8 780 km², or 17 percent of the country's total estimated parkland area. They are located in the rainfall range between 500 and 1 400 mm in the areas of Gondo-Mondoro, the Bandiagara-Hombori, Koutiala, and Mandingue Plateaux, the Central Niger Delta and Hod (Diallo, 1988, cited in Cissé, 1995). In terms of density and size, particularly outstanding F. albida parklands are found in the Dembéré-Douentza valley and extending into the Seno plain (Gallais, 1965), while Pageard (1971) reported the occurrence of well developed F. albida parklands in the Niger Valley between Bamako and Mopti, and particularly around Segou. A description of those in the inland Niger Delta is also available (Gallais, 1967).

The most dense and well developed F. albida parklands in Burkina Faso are seen among ethnic groups located near to the Bambara, Dogon and sedentarized Peul cultivators who have a strong F. albida tradition (Pageard, 1971). The F. albida parkland system is dominant in the Bwa and Samo regions extending to the Senoufo, Lobi, Dagara, Birifor and northern Côte d'lvoire regions. It is also well represented in Bissa country and in Yatenga as practised by the Mossi, though most of the Mossi Plateau is Vitellaria-dominated. It is found only in small, scattered areas in the Southeast and is rare in eastern Burkina Faso. In S.J. Ouédraogo's (1995) parkland classification, F. albida is represented in parklands throughout the country except for the extreme Southwest. For instance, he reported their existence around Markoye, Oursi, and Dori in the North, in Kokologo on the Central Plateau, and in Dossi and Boni in the West. Pure stands are reported in the Bulkiemdé Province west of the capital (Yélémou et al., 1993).

Fig. 1.3 Densities of Faidherbia albida in western Senegal (Source: after Portères, 1952, in Giffard, 1964)

In northern Côte d'lvoire, a parkland system including F. albida has been described in Dolekaha. This is one of the most southern areas of F. albida parklands described so far (Bernard et al., 1995).

In Niger, F. albida parklands are limited to the South and Southwest. Well-known examples are the highly stocked and permanently cultivated parklands in the ‘3M’ (Matameye, Myrriah and Magaria) region, resulting from the enforced protection of the species during the rule of Tanimoun, Sultan of Zinder (Montagne, 1986; CTFT, 1988). References to F. albida parklands in southern Niger also include the zone between Damergou (Tanout) and the Nigerian border (Bertrand, 1991), and Madaroumfa next to Maradi (Montagne, 1996). In the Southwest, F. albida parklands occur in Tilly near Sadoré and in Guilleny (Maï Moussa et al., 1993), as well as in the Dosso Department between the Dallol Bosso and the Dallol Maouri (Montagne, 1984). Faidherbia albida densities in parklands in the Dosso Department are lower than in the Zinder area (Montagne, 1996).

In Nigeria, F. albida is an important farm tree throughout the savanna zone where rainfall is above 1 000 mm (Sanusi, 1993). An early reference was made to F. albida along with Balanites, Borassus, Khaya, Anogeissus and Ziziphus species in fields on the hillsides of the northern Mandara mountains in the Northeast (White, 1941), and the terraced F. albida system was described later by Hallaire (1976). F. albida is also present in acha (Digitaria exilis Stapf) and millet fields of the Biron group on the Jos Plateau in association with the cattle of Fulani herders (Gosden, 1978, cited in Miehe, 1986), as well as in the North among Hausa farmers (CTFT, 1988). Further east in northern Cameroon, a strong tradition of F. albida parklands has survived among the Toupouri and Massa (Seignobos, 1982). Stands are densest in the highly populated zones around Golompui and

Table 1.3 Densities of Faidherbia albida in agroforestry parklands

Datcheka and some stands can also be observed around settlements among the Dowayo in the vicinity of Poli. Faidherbia parklands were widespread in the Diamaré area before their destruction during the Fulbe conquest in the nineteenth century. They are gaining ground again due to an increase in human population density (Seignobos, 1982). F. albida has the highest representation among the tree species present in 30 parkland samples within 30 km of Maroua (Triboulet, 1996). In Chad, relics of F. albida parklands are found among the Sara around the town of Koumra and well developed Acacia communities exist among the Kera around Fianga Lake (Seignobos, 1982).

Elsewhere in Africa F. albida parklands also occur in Sudan in the Fur farmlands in the Jebel Marra Highlands of Western Darfur (Miehe, 1986). The species dominates montane habitats between 1 800 and 2 300 metres above sea level, but F. albida parklands are also conspicuous in the Wadi Azoum alluvial system in the pediplains west of the Jebel Marra massif (Radwanski and Wickens, 1967). Faidherbia albida occurs along with other tree species in fields in the Nuba mountains in southern Kordofan province (Miehe, 1986). It is widespread in the Hararghe highlands of eastern Ethiopia (Poschen, 1986) and is reported further west around the town of Debre Zeit (Kamara and Haque, 1992), as well as associated with Cordia abyssinica in the Rift Valley among the Galla and Arussi groups (Miehe, 1986). Faidherbia albida also used to be maintained in arable areas of the Upper Jordan valley (Karschon, 1961).

Some F. albida parklands are found in Southern Africa. They have been studied in Tanzania (Fernandes et al., 1984; Okorio and Maghembe, 1994; Wickens, 1969), in Malawi's Lakeshore plain and upland central region (Saka et al., 1994; Rhoades, 1995), and in Zimbabwe (Campbell et al., 1991).

Reported F. albida densities in parklands throughout its zone of distribution are listed in Table 1.3. Compilations of information on various aspects of the species are presented in van den Beldt (1992), CTFT (1988) and Wickens (1969), among other sources.

Fig. 1.4 Flowering Vitellaria paradoxa, Sankpala, Ghana P. Lovett

Vitellaria paradoxa

Vitellaria paradoxa is probably the most common parkland species in semi-arid West Africa (Breman and Kessler, 1995). There are two subspecies in the genus, the subspecies paradoxa occurring from Senegal to the Central African Republic and the subspecies nilotica in southern Sudan and Ethiopia, Uganda and northeast Zaire (see Fig. 1.4). V. paradoxa is generally found in wooded grassland and farmland within the rainfall zone of 600 to 1 400 mm, annual potential evapotranspiration from 1 400 to 2 300 mm, and a dry period of 5 to 7 months. It preferentially grows on colluvial soils of reasonable depth (> 30 cm above parent rock or indurated layer), free drainage, and a predominantly sandy topsoil texture (Hall et al., 1996). Like other parkland species, its distribution has also been shaped by human influence.

A few estimates of the surface area occupied by Vitellaria species are available for Mali. Ruyssen (1957) has Vitellaria parklands occurring south of the line going through Bafoulabe, Segou and Bandiagara with an estimated distribution area of 19.4 million ha, revised to 22.9 million ha by Maïga (1990, cited in Cissé, 1995). Ruyssen (1957) estimated that 943 000 ha had Vitellaria as a significant tree component on land cultivated annually with rainfed crops. Based on a five-year rotation (one cropping and four fallow seasons), the area of productive Vitellaria parkland was estimated at 4.7 million ha. Clamagirand and Bruffaerts (1983, cited in Raison, 1988) state that at least 3.6 million ha in Mali have Vitellaria with a density of over 40 trees/ha. The species was represented in parklands covering 68 percent of the 415 700 ha middle Bani-Niger river basin mapped by ICRAF (1996).

Vitellaria parklands occur throughout Burkina Faso south of 14° latitude (S.J. Ouédraogo, 1995). In Bulkiemdé Province, Vitellaria has the highest density among parkland species (Yélémou et al., 1993). In Niger, the distribution is irregular and linked to human activities (Ounteni, 1998), with Vitellaria occurring in the south central and southwestern regions of the country (Hall et al., 1996). Vitellaria paradoxa and Parkia biglobosa are the most common indigenous parkland species in northern Ghana (Rudat et al., 1996) and spread through northern Togo and Benin where they have been studied in the Bassila and Parakou areas (Schreckenberg, 1996; Agbahungba and Depommier, 1989). They are also referred to in the north of Côte d'lvoire (Bernard et al., 1995; Louppe and Ouattara, 1996).

Fig. 1.5 Distribution of Vitellaria paradoxa subsp. paradoxa and V. paradoxa subsp. nilotica in relation to mean annual rainfall. (Source: Hall et al., 1996)

In Nigeria, Vitellaria parklands cover large areas. They were observed around Birnin Kebbi in the Northwest, south of Bida (southwest of Kaduna) in an area of old Nupe settlements (Pullan, 1974) and west of the Mandara mountains in the Northeast (Seignobos, 1982). Vitellaria paradoxa was cited among the most important farm trees during farmer interviews around the towns of Saki, Ilorin, Minna, Jos, Kaduna and Kano (Teklehaimanot et al., 1995). In northern Cameroon it is represented in many bushfallow communities outside F. albida parkland zones, in Toupouri country for example, but there are few well-developed Vitellaria parklands except on terraces of the Benoue valley west of Garoua (Seignobos, 1982). In southern Chad, there are large expanses of Vitellaria parklands (Seignobos, 1982) with tree densities higher than in F. albida parklands. References are also made to Vitellaria parklands among the Ngambay (area of Moundou in the Southwest) and around Koumra (west of Sarh, northern part of Sara country) mixed with F. albida (Seignobos, 1982). Vitellaria also dominates parklands south of Domo Dambali in Moussey country.

In Ouham province in the northwest of the Central African Republic, agriculture is traditionally practised under the parkland cover of V. paradoxa associated with P. biglobosa in high densities (Depommier and Fernandes, 1985). Further east, V. paradoxa subsp. nilotica is reported in fields in northern Uganda, where its density is highest in Otuke county in Lira district (Masters and Puga, 1994). Reported V. paradoxa densities in parklands throughout their distribution zones are listed in Table 1.4. Additional information on the species can be found in several monographs (Hall et al., 1996; Bonkoungou, 1987).

Table 1.4 Densities of Vitellaria paradoxa in agroforestry parklands

Parkia biglobosa

Parkia biglobosa is often found in parklands in association with V. paradoxa. This species is mostly present in areas with between 800 and 1 500 mm of rainfall, and 1 400 and 2 100 mm of potential evapotranspiration, and is generally associated with a dry season of 5–7 months. Its range is similar to that of V. paradoxa but extends further south. Parkia biglobosa typically occurs on mid-toposequence positions on deep soils and sometimes, through farmer protection, on well-drained soils in floodplains and riparian sites. It is absent from depressions where soil drainage is impeded (Hall et al., 1997). The species naturally occurs in the dry forests of the Sudano-Guinean zones where it is associated with Pterocarpus erinaceus (Sall, 1996). In Senegal, it is found on various soils in the Sine, Cayor, Laghem, Saloum, Tambacounda and Niololo Koba areas, but not in the North (Sall, 1996). Parkia biglobosa in parklands is found in the southern tip of Gaya in Niger (Ounteni, 1998), in south Mali (Kater et al., 1992; Bagnoud et al., 1995b), throughout Burkina Faso except north of the south Sahel zone (Teklehaimanot et al., 1997), in northern Côte d'lvoire (Bernard et al., 1995), northern Ghana (Rudat et al., 1996), northern Togo among the Kabre (Enjalbert, 1956), and the northern half of Benin (Schreckenberg, 1996; Agbahungba and Depommier, 1989). In Nigeria, Parkia genus was reported in fields in Bomo, north of Zaria (Pullan, 1974), and was studied in sixteen sites throughout the lowland forest zone up to the Sudan savanna zone (Teklehaimanot et al., 1996b). Pure stands were also observed in northern Cameroon (Seignobos, 1982) and the species pervades parklands in the Ouham province of northwestern Central African Republic (Depommier and Fernandes, 1985). Information regarding the taxonomy, biology, ecology, management, and uses has recently been compiled by Hall et al. (1997). Reported P. biglobosa densities in parklands throughout their distribution zones are listed in Table 1.5.

Fig. 1.6 Distribution of Parkia biglobosa in relation to mean annual rainfall (Source: Hall et al., 1997)

Table 1.5 Densities of Parkia biglobosa in agroforestry parklands

Other important parkland species

Overall, F. albida, V. paradoxa and P. biglobosa are probably the most widespread parkland species in the Sahel and Sudan zones of Africa. This explains their predominance in the existing literature. However, this should not conceal the importance of a large number of other parkland species, whose representation may be more restricted geographically but which may, on a local scale, be more abundant and more economically valuable. It is worth emphasizing that one of the major assets of agroforestry parklands is their biological (and genetic) diversity. This implies a wide diversity of uses and applications with significant economic value and a variety of management techniques and objectives (both discussed in later chapters).

The baobab (Adansonia digitata) is one of the best known and often reported tree species in semi-arid Africa, due to its large size and associated mythical and spiritual powers. Adansonia parklands throughout the zone are associated with both old and recent settlements. The species is characteristic of plant communities of the Sudano-Zambezian lowlands with 200–800 mm annual rainfall, but it has extended into higher rainfall areas, possibly with human assistance (Wickens, 1982). It is very common in the intensively managed and permanently cultivated fields around residential compounds. High densities are present in western Senegal around Dakar, Bargny (densest stands) and Thies, north of the Gambia river between Kaolack and Tambacounda, as well as in the Kedougou area in the Southeast (Sall, 1996), and were used as a defensive barrier against attacks by men on horseback (Baumer, 1994). Adansonia parklands are found in Burkina Faso's Yatenga and Bam provinces (S.J. Ouédraogo, 1995), and the species is represented in parklands on 12 percent of the middle Bani-Niger river basin in Mali (ICRAF, 1996). It occurs on old settlement sites in parklands of the southern Mandara, in northwestern Nigeria (Seignobos, 1982) and in northern Togo among the Kabre (Enjalbert, 1956). Additional review information on the species, including its distribution and ecology by regions, is provided in Wickens (1982).

Fig. 1.7 Distribution of Adansonia digitata in Africa and neighbouring areas Key:

• 1 Distribution based on Herbarium and flora records;

• 2 Specimens known to be cultivated or introduced;

• 3 Distribution based on published and unpublished photographs;

• 4 Distribution based on the Kew ‘Baobab Survey’ information;

• 5 Records obtained from travel literature, maps etc.

(Source: Wickens, 1982)

Fig. 1.8 Adansonia digitata parklands, Senegal R. Faidutti

Borassus aethiopum, or fan palm (rônier in French), also has a wide distribution. It is found on a variety of soil types but demands a high water holding capacity and shallow groundwater tables. Its development is favoured by fertile and weeded soils as found in parklands (Niang, 1975; Weber and Hoskins, 1983; Cassou et al., 1997). Among other places, it occurs along all the major rivers and other water bodies of the Sahel. According to Pélissier (1980a), the most beautiful stands are in villages around the mouth of the Soungrougou river (Ziguinchor and Kolda), Casamance in Senegal. Other stands exist in the North in the Goumel forest next to Dagana, in the Northeast in the Matam département, in the Southwest near Tabacounda and on the Gambia river, as well as west around the towns of Thiès, Dakar and Fatick. Very high densities of 50–100 and 115 trees/ha in almost pure stands were reported in the Leolgeou area in Mali (Gallais, 1965) and Wolokonto, Burkina Faso. The species forms one of the ten parkland types classified in Mali (Cissé, 1995). In Burkina Faso, they are most common in the southwestern provinces of BoboDioulasso and Banfora in village or compound fields (S.J. Ouédraogo, 1995). In Niger, B. aethiopum occurs in six major populations estimated at 27 000 ha in the Dallol Maouri, Gaya Department, and 3 000 ha along the Niger river and on Leté island (Ounteni, 1998).

The species occurs in northern Ghana associated with F. albida. This combination is also found in Nigeria near Foggo, southwest of Azare, an area of Hausa and Fulani settlement (Pullan, 1974), and in northern Cameroon around Mboum settlements in the Guidar region (Seignobos, 1982). Borassus aethiopum is also observed in the North Cameroon inlet of Chad between and along the Logone and Chari river valleys, for example among the southern Kotoko and Mousgoum people around Goffa, Holom, Marmay and Pouss, and among the Massa in Ziguey, Goufka, Domo, Bosgoye, Geme, Nahaide and Dana (Seignobos, 1982). In Chad, the Borassus area extends east and southeast from the Logone river along the southern fringe of the Baguirmi region going through the towns of Logone Gana, Morno, Ngam (Kwang group) to the Sarwa region and even into the Day area (Seignobos, 1982).

Fig. 1.9 Distribution of Balanites aegyptiaca in relation to mean annual rainfall. The 400,800 and 1400 mm isohyets are shown. The hatched area is the ‘main range’. (Source: Hall and Walker, 1991)

Dominant species in the Sahelian (northern) parklands include Acacia raddiana (syn. Acacia tortilis), Acacia senegal, Balanites aegyptiaca and Hyphaene thebaica. In Senegal, A. raddiana parklands are located throughout the North (Sall, 1996). Trochain (1940) also mentioned mature A. raddiana trees in groundnut fields in the Sahelian part of the country. Acacia raddiana is found in the Sahel zones of Mali (Cissé, 1995) and Burkina Faso, sometimes in pure stands (S.J. Ouédraogo, 1995). The species occurs almost exclusively in cultivated or agro-pastoral areas, in contrast to A. senegal which is characteristic of rangelands in West Africa. In Sudan, virtually pure stands of A. senegal are systematically rotated for crop and fallow/gum production and are sometimes referred to as ‘gum gardens’ (Seif El Din, 1981; Jamal and Huntsinger, 1993).

Balanites aegyptiaca is a spiny, fairly short tree, occurring most frequently on the 400–800 mm rainfall zone. The species avoids shallow and gravelly soils and prefers deep sands, sandy clay loams or clays in lowlands. As a shrub, it is an important species on aeolian sands in the Sahel (Hall and Walker, 1991). It is associated with A. raddiana in northern Senegal (Nizinski and Grouzis, 1991; Sall, 1996), Mali (Cissé, 1995) and Burkina Faso (S.J. Ouédraogo, 1995). It is also well adapted to the climatic conditions prevailing in Niger where it is widespread (Ounteni, 1998). It is the most frequent species in parklands in Léré, Koro arrondissement (district) in the Fifth Region of Mali (PRSPR, 1993), and is reported in Vitellaria-dominated parklands in Birnin Kebbi, northwestern Nigeria (Pullan, 1974). Additional data on the species are available in Hall and Walker (1991).

Hyphaene thebaica, or dum palm, is another economically significant Sahelian species which is easily recognized by its dichotomous branching habit. It occurs in the Sahelian and Sudano-Sahelian zones with 200 to 600 mm rainfall, even extending to the southern edge of the Sahara, generally on light soils with shallow aquifers, in dune depressions or gallery forests but not on rocky soils (Raison, 1988; von Maydell, 1983). It is mentioned in parklands in the Ferlo river valley, Senegal (Freudenberger, 1993b), in the Leolgeou area, Mali (Gallais, 1967), in Tilly, southwestern Niger (Maï Moussa et al., 1993) and in many dallol and goulbi valleys and along the Niger river (Ounteni, 1998). It also occurs in northern Nigeria (Sanusi, 1993) and is characteristic of parklands in the southern Sahelian region of Burkina Faso (S.J. Ouédraogo, 1995).

Other important parkland species include the oil palm, Elaeis guineensis, found in Benin (Schreckenberg, 1996; Raison, 1988) and northern Togo (Enjalbert, 1956). The exotic species, Azadirachta indica (neem), has expanded widely in Sahelian countries, to the point of sometimes being considered an invasive pest (Ganaba, 1996). It thrives in the Bulkiemdé province of Burkina Faso, especially in compound fields (Yélémou et al., 1993) where it is the second most common parkland species after V. paradoxa (Yélémou, 1993). Prosopis africana is a major, yet declining parkland species along the Chad-Cameroon border north of Fianga Lake and further east in Ngam in Chad (Seignobos, 1982; Bernard, 1996). It also occurs in the southern part of the Dosso and Tillabéry Departments as isolated individuals and as monospecific stands in the Matameye area, Zinder Department in southern Niger (Ounteni, 1998).

Fig. 1.10 Cordyla pinnata fruit, Nioro du Rip, Senegal P. Danthu

Ceiba pentandra parklands have been noted around Zaria, Nigeria (Jackson, 1970) and in northern Togo (Enjalbert, 1956). Cordyla pinnata (dimb) was ranked second in relative abundance after F. albida in Sob, Senegal (Lericollais, 1989), and after V. paradoxa in Mali (Ohler, 1985). It thrives in farming systems with high population density in the Sine Saloum area of Senegal because of its nutritional contribution, the kernels being used as a substitute for meat (Niang, M., 1998). Cordia africana dominance was reported in fields devoid of F. albida in the Alemaya area, Ethiopia (Poschen, 1986).

The list of species occurring in parklands reviewed so far is by no means exhaustive. The following, among many others, can be added: Sclerocarya birrea, Parinari macrophylla, Cola cordifolia, Anogeissus leiocarpus, Bombax costatum, Tamarindus indica, Sterculia setigera, Ziziphus mauritiana, Lannea microcarpum, Lannea acida, Pterocarpus erinaceus, Diospyros mespiliformis, Detarium microcarpum, Combretum glutinosum, Celtis integrifolia, Piliostigma reticulatum, Cordia myxa, Khaya senegalensis, Blighia sapida, Tectona grandis.

Factors used in establishing parkland typologies

As noted earlier, savanna landscapes have evolved under the influence of natural processes such as fire and grazing. In contrast, parklands reflect deliberate human manipulation of trees in agricultural production systems. Various classification systems attempt to capture regional and local variations in parkland structure and composition according to factors such as the degree of human intervention, their main functional uses, their physical structure, and their reflection of the different natural resource management systems of diverse ethnic groups.

Degree of human intervention

Geographers first classified parklands in relation to the degree of human intervention that contributed to their formation, an approach which was particularly useful in revealing gradations of management intensity and giving them a broad, inclusive representation among human-manipulated vegetation types. Categories first established by Pélissier (1964), and later adapted by Seignobos (1982) and Raison (1988), are presented below and an attempt made to examine how effective they are in reflecting actual and significant field realities.

The least elaborate and most transient type is ‘residual’ parklands. Species that are either useful or too hard to cut are left after clearing, but their density does not necessarily reflect the importance of their use. Species richness is relatively high in this type and is bound to decline with further selection. These parklands will only persist where low population density allows long time intervals between clearings. Cordyla pinnata parklands observed by Pélissier (1980a) between Saloum and Gambia in Senegal are one example. They are primarily used for grazing and harvesting of tree fruits (Baumer, 1994).

The second type develops with conditions created by ‘clearing’. Such parklands are generally composed of shade-intolerant species, released after clearing, and species which become dominant due to their ability to regenerate quickly from root-suckers (Seignobos, 1982). Typical species include some acacias and Prosopis africana, which benefit crop production and provide excellent fodder. Prosopis africana, however, also persists in old parklands (Bernard, 1996). Elaeis is a similar case (Pélissier, 1964). The two kinds of parkland presented above are often short-lived, but their recognition may be useful to situate their current stage and potential for evolution into a more elaborate and stable type.

Karité and néré parklands have been deliberately protected in the field. They reflect old, stable and relatively dense settlements.

‘Selected’ parklands are composed of trees which were part of the initial vegetation and have been deliberately protected in fields owing to their various uses including food, fodder or soil fertility. Their composition and density correspond to family and community needs. They reflect old, stable and relatively dense settlements as well as landscapes more deeply influenced by human beings than the previous two types. Classic examples are the karité (Vitellaria paradoxa) and néré (Parkia biglobosa) parklands. Given the slow growth rate of Vitellaria, such parklands indicate a certain degree of permanence. Systems based on the fan palm, Borassus aethiopum (source of palm wine and seedlings eaten during famines), and to some degree the oil palm, Elaeis guineensis, are other examples of selected parklands.

Quantitative evidence of differences in tree density and species composition in fields as compared to the original savanna vegetation illustrates the impact of human selection on parkland formation. Farmers tend to reduce the total number of stems and species and favour preferred species. In Vitellaria and Parkia parklands of southern Burkina Faso, a third of the species found in uncultivated woodland savanna were not found in adjacent agricultural fields, while nine out of ten woody individuals were thinned during clearing and cultivation. At the same time, the relative abundance of Vitellaria had increased five times from 16 to 83 percent, and that of Parkia nine times from 0.4 to 3.5 percent (Boffa, 1995). Similarly in Benin, V. paradoxa and P. biglobosa represented 10.6 and 0.7 percent of trees in their original savanna vegetation and 39.2 and 3.6 percent in fields, a 3.7 and 5.4 fold increase respectively (Schreckenberg, 1996).

Fig. 1.11 Borassus aethiopum parkland, Senegal R. Faidutti

The fourth type, ‘constructed’ parklands, is made up of species which may not always be present in the initial vegetation, at least not in the densities observed after farmer selection. They are more elaborate as trees are not only protected but also pruned and tended in order to reach large height and crown dimensions. The best example is that of F. albida which is naturally bushy but which will develop into tall trees when pruned early on. This type has also been called substitution parkland in the case of F. albida (Seignobos, 1982), because the species, which can be absent from climax communities, may partially or completely replace spontaneous vegetation (Pélissier, 1980b).

The fifth type of parkland displays species disseminated by people in a type of ‘proto-arboriculture’ (Seignobos, 1982). Such parklands are strongly linked to settlements. For example, the baobab, Adansonia digitata, the fruits, leaves and bark of which are systematically gathered, is a familiar sight in Sahelian villages. Years after settlements relocate, its presence may still indicate areas used for cultivation. Borassus aethiopum is another such species. Among others, it was associated with the Baïnouk people and is found along their migration paths from Senegal to north Cameroon (Pélissier, 1980b).

Finally, ‘planted’ parklands have the highest degree of human intervention and may more appropriately be referred to as intercropped orchards. Such is the case, for instance, of orchards of mango trees interplanted with food crops. However, whether such stands of specially planted trees qualify as parklands is debatable (Raison, 1988).

The lack of clear definitional boundaries tends to make the different parkland types less operational than desired. Selected and constructed parklands can overlap as in the case of Vitellaria and Parkia parklands. Both are selected from the pre-existing vegetation, but they are also constructed through farmer management. As suggested by the above data, tree size and relative abundance are greatly enhanced through the selective process and their density in parklands may also be increased through fallow enrichment, transplanting or planting. The same can be said of the Acacia senegal (and Acacia laeta) parklands of Sudan, often called gum gardens². Present in the original vegetation, they are selected and constructed through various cycles of cultivation consisting of coppicing, fallow and gum collection. As parkland trees become more valuable and are managed more intensively by manuring, trimming and pruning practices, the distinction between selected and constructed parklands will probably become less pronounced.

Similarly, distinctions between constructed, proto-arboricultural and planted parklands can appear artificial. For example, in areas where land use becomes sedentary, A. digitata parklands may be considered to be constructed rather than products of proto-arboriculture. The case of Azadirachta indica, which is both planted and disseminated by birds, encompasses all three categories. Similarly, the notion of substitution of the original vegetation by one species, as in F. albida parklands, is at least potentially present in the (trans-) planting of parkland species. There is no clear consensus on the inclusion of intercropped orchards (mango, citrus, palm) in the parkland family, yet the ‘artificialization’ of tree cover in parklands is likely to become more pronounced as research and development efforts give increasing recognition to trees in these systems. One may speculate that the dissemination and planting of domesticated parkland trees will further challenge the available classification and may increasingly divert attention from categories for whole systems towards a parkland mosaic of niches specific to particular species. Clearly, there are opportunities for refining and operationalizing parkland classifications.

² In contrast to the gum gardens of Sudan, Acacia senegal parklands in the northern fringe of West Africa would not belong in the constructed parklands category as they are less intensively managed and tend to be exclusively pastoral.

Functional classification

By allowing certain trees to persist in their fields, local populations have shaped the tree component of their farmland to fulfil specific needs. Among the most impressive and probably insufficiently recognized characteristics of agroforestry parklands are the diversity of tree species they contain and the variety of products and uses they generate. This has led researchers to develop functional classifications for parklands as in the one adapted from Seignobos (1982) in Table 1.6. Some of these categories, such as the use of tree fibre for clothing, appear less relevant today. In addition to the productive roles presented in Table 1.6, trees are maintained in fields for less obvious purposes, such as the provision of shade, fodder storage (crop residues like cereal stalks and cowpea stems stored in branches of tall trees), and a vast array of medicinal products. Chapter 6 elaborates on the quantitative and qualitative diversity of parkland products and services.

Spatial analysis

As a result of habitat dispersion, land-use patterns, social groupings and defence needs, as well as economy of movement, farmland is often arranged in concentric rings around settlements in rural Africa. Management intensity levels decrease as distance from the compound increases (Morgan, 1969). In the Sudan zone of West Africa, three major concentric rings characterized by different cropping and soil management activities have been recognized (Prudencio, 1993; Sautter, 1962): (i) a zone of home gardens or compound fields (champs de case), (ii) a zone of permanently cultivated village fields, and (iii) a zone of bush fields.

Table 1.6 Main productive functions fulfilled by agroforestry parklands

(Source: adapted from Seignobos, 1982)

‘Compound fields’ are usually located within a 50 m radius of the compound. They are heavily fertilized with household wastes and manure produced by small livestock and fowl kept close by. Soil fertility and moisture conditions are generally higher here than on other farmland types and can support crops such as maize and red sorghum. Vegetables and other crops used daily are grown near the house for convenience.

‘Village fields’ extend within a ring several hundred metres wide. Depending on distance and fertility, they may be manured and can be fertilized with chemical fertilizers and permanently cultivated. In many parts of semi-arid West Africa, migrant Fulani herders arrange with local farmers to manure these fields in exchange for various benefits. In areas of high population density where land degradation has taken place, erosion control measures such as rock bunds may have been applied on these and compound fields. Farmers generally cultivate a variety of crops, such as white sorghum or millet intercropped with cowpeas and groundnuts in village fields. Cotton, tuber crops and Bambara groundnut (Voandzeia subterranea) may also be found.

‘Bush fields’ further out are where the bulk of staple crops such as sorghum, millet, cowpeas, and groundnuts (on relatively sandy soils) are cultivated. Historically, this is a zone of shifting cultivation where the area of fallow grass is much greater than that under crops (Morgan, 1969), the ratio being highly dependent on population density. In recent decades, there has been a strong tendency to shorten or eliminate fallow periods. Fertile and moist lowland areas, where farmers grow vegetable crops, rice, sorghum, etc., may occur regardless of distance from settlements.

Tree selection and conservation are closely linked to agricultural production activities over a span of years or decades. The spatial structure of parklands is influenced, therefore, by the specific management of each of the above-mentioned cultivation rings. In fact, a terroir villageois (village land) is often said to harbour three distinct zones of tree cover, namely a ‘core’ and a ‘margin’ with an intervening zone of ‘plain parkland’ (Seignobos, 1982). The core or residential zone has species with particular uses which may not be found elsewhere such as Mangifera indica and other fruit and shade trees, Ficus thonningii, Azadirachta indica, Eucalyptus spp., Cassia siamea, Adansonia digitata, Ceiba pentandra, Tamarindus indica, etc. (Savonnet, 1959; Pullan, 1974). The plain or pure parkland zone is found on permanently or repeatedly cultivated fields and visually represents the agricultural system practised. It is usually dominated by one or a few species, such as F. albida, or V. paradoxa. In Birnin Kebbi, northwestern Nigeria, this zone is characterized by the presence of mature trees of species including Vitellaria, Tamarindus, Vitex, Ficus and Balanites (Pullan, 1974). In the Jebel Marra massif of western Sudan, the most extensive F. albida parklands are primarily found in limited areas of very dense and continuous habitation (Miehe, 1986). In parkland margins, a greater variety of species may be selected or selection may be less precise.

This general threefold zonation is reflected in several studies which record the different names given to each field type in local languages. For instance, among the Bwaba-Bobo-Oulé of Burkina Faso, ka fields around compounds, and extending 25 to 50 metres out, are heavily manured and enclosed with fences, and contain C. pentandra, baobab and tamarind trees (Savonnet, 1959). Outside the ka fields, there are wa fields corresponding to the area covered with F. albida trees in a radius of up to one kilometre. Still further out are the more intensive ma or bush fields, which undergo fallowing and are dotted with V. paradoxa, P. biglobosa and Ficus spp. Likewise, Gallais (1967) notes that the ring of permanent fields is called so-foro in Bambara, or fio by the Bwa people in the inland Niger delta region in Mali, while bush fields are called kongodian-foro and mwese respectively by these two ethnic groups. In Wolokonto, Burkina Faso, compound gardens include mango, citrus and cashew trees, while Borassus aethiopum parklands develop on permanently cultivated village fields which are themselves subject to different intensities of land use. A decreasing gradient of Borassus tree density of 155, 127 and 87 trees/ha was found in three adjacent rings of village fields located at increasing distance from the village. Beyond these, bush fields are characterized by Vitellaria and Parkia parklands and the absence of Borassus (Cassou et al., 1997).

Fig. 1.12 Faidherbia albida parklands around Dolekaha, Côte d'lvoire. Monospecific stands of F. albida are located almost exclusively in a belt around the village, while Vitellaria paradoxa and Parkia biglobosa parklands are located at a greater distance. Note sacred forest in lower right hand corner. D. Louppe

Not only do parkland composition and structure result from the type and management intensity of agricultural activities, but they are also influenced by demographic changes. Reflecting such trends, Gallais (1967) outlined stages in the spatial expansion of F. albida parklands in the eastern part of the Kounari region in Mali. The continuum included villages with ‘non-existent’, ‘limited’ and ‘extensive’ parklands, as well as the separate case of ‘distended’ parklands. Villages in this region were devastated by wars and abandoned during the nineteenth century, after which access was open to exploitative wood cutters and herds. More recently, they have been recolonized. In places, trees were cut down and these lands may still remain treeless. In others, degraded parklands were regenerated, but where human density was low, the zone of so-called ‘limited’ parkland is now surrounded by a growing belt of permanently and extensively cultivated fields. Elsewhere in areas of higher population density and tight settlements, ‘extensive’ parklands ranging from the village houses through the cultivated zone were rapidly regenerated to support an intensively-managed farming system. Also, in agropastoral communities Gallais observed ‘distended’ parklands where an inner treeless zone used for cattle pens in the settlement area was surrounded by a parkland ring. The intensive use of F. albida branches for fencing pens and cattle pathways through the cultivated fields to outer grazing zones explained the low tree density in the village centre. Such a spatial pattern was also reported by Pullan (1974) in Birnin Kebbi in northwestern Nigeria.

Adjacent parkland areas with stands of trees of different sizes have often been reported (Pullan, 1974). One may find a central zone of cultivation with large scattered trees surrounded by an outer zone of smaller farm trees in far higher density, the latter being a zone of expansion of cultivation where conditions for natural regeneration are favourable. The condition of the woody cover on these newly cultivated areas depends on the intensity of wood collection, burning, and grazing. Gallais (1967) also noted variations in tree size ranging from 10–15 cm to 70–80 cm in diameter in the F. albida parkland stands of Mali which revealed their dynamic condition. These observations indicate a relationship between the size of parkland trees, which is often reported as relatively uniform within stands, and time of village occupancy or at least duration of cultivation. In southern Burkina Faso, average diameter at breast height of Vitellaria in farmers' fields was highly correlated with the number of years fields had been cultivated (Boffa, 1995).

The above parkland categories linked to soil management patterns or general land-use changes make up only one general component of the spatial analysis of parklands. Finer trends at various lower scales of analysis should also be pursued. Local situations generally hold a large diversity of different parkland types, which may overlap or contradict the proposed patterns (Depommier, pers. comm.). These locations may, however, yield significant information on a potentially wide range of variations in land use due to land saturation, land tenure modifications or economic constraints, etc., which are central to the dynamism of the systems. The age structure, density and composition of woody cover reflect compound management strategies at the ethnic, village, household, sub-household and individual levels in response to various external factors which are analysed in this report. In addition, parklands vary spatially in response to ecological constraints such as position in the toposequence and soil type. In order to capture the dynamics of parkland systems properly, researchers should be concerned with all of these levels. This has methodological implications. Firstly, characterization techniques need to cover all the different scales from the region, terroir, and parkland unit, down to the field, field section and even individual trees. Secondly, these agroforestry systems demand a strong multidisciplinary approach and wide collaboration between various sciences including morphogeology, soil science, agronomy, forestry, geography, history, socio-economics, animal science, remote sensing, Geographic Information System (GIS) technology and others.

Parklands as a reflection of agrosystems and ethnicity

Ethnic groups differ in their settlement and land-management patterns, even in comparable arable landscape units. Variations in social organization may be reflected in general land-use patterns including parkland management strategies at village and regional scales, as illustrated in examples from Burkina Faso (Box 1.3) and Senegal (Box 1.4).

The structure and composition of parklands may reflect the way an ethnic group manages the vegetation in fields and fallows. The composition of dominant species indicates the general type of agricultural system while secondary parkland variables may be determined more by cultural and ethnic factors (Seignobos, 1982). This may explain why early explorers observed variations in parkland composition as they crossed West Africa (Kirk-Greene, 1962, cited in Pullan, 1974). For instance, Seignobos (1982) illustrated how F. albida and V. paradoxa parklands reflected two fundamentally different agrosystems and rural societies.

Faidherbia albida is the dominant species of the Sérer parklands in Senegal (Pélissier, 1966). The cultivated area is divided into at least two or three zones cropped and fallowed in rotation on a biennial or triennial basis. Elsewhere F. albida parklands are usually associated with permanent cultivation. During the cropping season, cattle are restricted to fallow areas and circulation to and from the outer zones of uncultivated forest is facilitated by live and dead fences around fields and along paths. At night, they are tethered and progressively moved in order to ensure homogeneous manuring of fields. After harvest, they are free to roam throughout the village, but remain secured at night.

Trees, cattle and crops are strongly interconnected in an intensively managed system. Trees contribute to the maintenance of soil fertility through nitrogen fixation and cycling of mineral elements from deep soil layers, but their pods and leaves also provide substantial quantities of fodder for cattle which, in turn, are a source of manure for soil productivity.

In contrast, V. paradoxa parklands may indicate an absence of cattle in the local farming system and little or no contact with pastoralists; Vitellaria kernels, from which a vegetable butter is extracted, compensate for the absence of milk and animal butter. This type of parkland is usually associated with a relatively stable settlement and a more extensive cropping regime than in the case of F. albida. Farmers typically clear and cultivate fields at the village periphery, beyond a limited zone of permanent cultivation around settlements, and these are shifted further out after three or four years of cultivation, while a fallow period of 12 to 15 years is applied. Land availability permitting, residence sites may be moved after two or three cycles to allow the land to rest longer. Over several decades, a stratum of selected trees, often dominated by V. paradoxa and P. biglobosa, develops. This land-management system requires a large area of land and population density associated with V. paradoxa parklands therefore tends to be lower than in F. albida parklands. Unlike F. albida, the Vitellaria system supports only one aspect of the economy rather than the whole agricultural system.

Some authors argue that ethnic specificities could result in the mutual exclusion of the two species. For instance, Pélissier (1980a) claimed that the Vitellaria genus is absent from communities which are either purely agropastoral or have maintained close relations with cattle raisers. This would explain why it is not found west of the Falémé river in Senegal. However, natural factors limiting the extension of the species could also be responsible. The presence of both species in the Dagomba area in Ghana as well as in other places in West Africa illustrates their compatibility (Sène, 1998). The two species are also interspersed where distinct ethnic groups used the same area either subsequently or simultaneously as a result of adoption (Marchal, 1980; Gallais, 1967; Savonnet, 1959; Dubourg, 1957). For example, in Dakola, Burkina Faso, stands of F. albida are found only in very old settlement sites established by Dogon and Samo populations before the Mossi conquest, which subsequently introduced V. paradoxa, P. biglobosa and Tamarindus indica (Kohler, 1971). This is also what Izard-Héritier and Izard (1959) and Marchal (1978) observed in Yatenga. Historical research in areas where the species occur together or exclude each other would provide highly interesting insights into the promotion of parkland species by various ethnic groups.

The influence of specific ethnic groups on the extension of parkland systems is also referred to by Pageard (1971). In Mali, the most elaborate and extensive F. albida parklands are located between Bamako and Mopti where sedentary cultivators of the Niger valley, namely the Bambara, the sedentary Peulh descendants of the Macina empire, and the Dogon people, live. In Burkina Faso, parklands and the emphasis on F. albida in myths and religious practices are mostly found among ethnic groups of the western provinces who were geographically closest to Malian groups associated with F. albida parklands. These parklands are rare in the Gourmanché villages of eastern Burkina Faso.

In northern Cameroon, from the Mandara mountains in the east to the Logone river in the west, F. albida parklands are most frequently encountered. However, Prosopis africana stands out as the most common species in some villages in the Bec de Canard region (Bernard, 1996). Interestingly, these stands are circumscribed in a land pocket occupied by the Musey ethnic group, whose largest area of settlement is in Chad and surrounded by the more spatially dominating Massa populations. This group had a strong warrior and hunter tradition with a well established reputation. They are known for having fought the neighbouring Massa and Fulani and their strong cultural identity persists today. Prosopis africana was actively maintained in parklands because it provided stakes for decorating tombs and recalling the war or hunting feats of old Musey warriors. Whereas the surrounding Massa villages emphasize livestock production, the Musey raise Logone horses used in combat and hunting. However, cultural and technological changes have made the practice of regenerating P. africana less significant among young segments of the Musey.

Whereas the above-mentioned agricultural systems stand out because of the prominence of one or a few species, this is not always the case. In the Mafa region of Cameroon, where terraced agriculture is practised on the steep slopes of the northern Mandara mountains, parklands include several species, none of which is predominant. In this remote area farmers are relatively poor and self-sufficient. They maximize the use of trees, while minimizing their negative effect on crop production (Seignobos, 1982).

Summary

Parklands (or ‘parcs arborés’ in French) are landscapes in which mature trees occur scattered in cultivated or recently fallowed fields. They are extremely variable, leading to continued discussion about appropriate terminology. They have been referred to as a vegetation type similar to ‘tree savannas’ but differ from these in that they are of specifically human origin, with the composition and density of their woody component manipulated in order to facilitate its use. Although occurring in other areas of the world, such as southern Africa and the Mediterranean, it is in the Sahel and Sudan zones of Africa that parklands are most widespread. No accurate figures exist at a regional level, but parklands constitute the predominant agroforestry system in semi-arid West Africa. They are most often characterized by the dominance of one or a few species. Faidherbia albida has received most attention in the literature because of its positive effect on soil fertility but it is likely that parklands dominated by Vitellaria paradoxa and Parkia biglobosa occupy the largest area.

Their origin in human agricultural activities meant that their study was generally neglected by early plant ecologists and the first in-depth system descriptions were undertaken by geographers in the 1960s. Several approaches have been used to distinguish types of parklands. First, they have been categorized according to the intensity of human management as residual, selected, constructed, proto-arboricultural or planted parklands. This perspective is useful in revealing the various degrees of management, but parklands characterized by a dominant species often do not fit exclusively into one or other of these categories. With increasing management intensity and dissemination of species and technology, the development of a mosaic of parkland types adapted to various niches may be expected.

Parklands have also been classified according to broad use categories. These systems have been established for the production of essential food complements, oils, wine, famine foods, wood, browse, crafts and for soil fertility restoration or a combination of these purposes.

The age structure, composition, and density of parklands have also been analysed in terms of spatial variation. This corresponds in part to soil management practices, which decrease in intensity from compound fields through village fields to bush fields. Thus Faidherbia albida may be favoured on permanently cultivated village fields, while alternating fallow and cultivation cycles in bush fields may promote the development of floristically more diverse parklands with a Vitellaria paradoxa and Parkia biglobosa dominance. Tree size is correlated with the time during which fields have been cultivated. Thus, contrasts in size between parkland stands reveal land-use changes sometimes associated with demographic events. The growth rate of trees in parklands is generally higher than in uncultivated locations due to more favourable conditions.

Finally, the literature reports on the influence of ethnic identity and social organization on land-use patterns and parkland types. Examples in Senegal and Burkina Faso illustrate how land-use systems and agroforestry practices of several ethnic groups have evolved and become distinct through a number of processes including wars, colonization, population growth, decreased land availability and agricultural commoditization. Ethnic specificities have also been responsible for the promotion of given parkland species in some geographical zones. Fertilizer policies may also have a significant impact on traditional fertility maintenance and tree management practices. There are indications that removal of agricultural input subsidies in the late 1980s fostered Faidherbia albida regeneration in Western Senegal.