CHAPTER III
BIOPHYSICAL FACTORS IN PARKLAND MANAGEMENT

Biophysical interactions between trees and crops have shaped the physical configuration of parkland agroforestry and are one of the primary determinants of the way farmers manage trees in their fields. This chapter examines the available qualitative and quantitative information on biological processes governing the biophysical influence of parkland trees on soils and crops. While the focus is on cultivated plants in parklands, reference is also made to the interaction of savanna trees and wild herbaceous species so that similarities and contrasts with parklands can provide a better understanding of the biophysical dynamics involved.

Tree-soil relations

The potential contribution of trees to soil improvement is one of the major assets of agroforestry in general and, more specifically, of agroforestry parklands.

Influence on soil chemistry

The enhancement of soil fertility by trees is conspicuous in studies which compare productivity of crops grown on soils formed under tree canopies and on control soils in open sites. For instance, Verinumbe (1987) found that maize and sorghum produced higher dry matter on soils collected in tree plantations than on ordinary field soil. Highest crop yields were obtained on soils from under Azadirachta indica, followed by Prosopis juliflora and Eucalyptus camaldulensis. Total biomass of millet plants grown on soils sampled under Hyphaene thebaica and Faidherbia albida in Niger was 63 percent higher than of those grown on soils sampled away from trees (Moussa, 1997). In Senegal, similar trends were reported for millet and groundnut under Cordyla pinnata and in the open (Samba, 1997). In parklands, one may also note that crops produce better around tree stumps than elsewhere in fields. Differences in soil fertility as demonstrated by in situ crop productivity differences at varying distances from the tree are discussed later in this chapter.

Savanna tree species

Soil chemical analyses beneath savanna trees in semi-arid zones reveal a common pattern of higher soil fertility under isolated tree canopies than in the open. Fertilizer addition in Kenyan savannas resulted in significantly increased herbaceous productivity in open sites and showed that these soils were nutrient-limited. In contrast, there was no effect on productivity in sub-canopy zones. This suggests that savanna trees were responsible for nutrient enrichment on sites previously poor in nutrients (Belsky, 1994). Nutrient values in relation to distance from woody plants are presented in Breman and Kessler (1995: 156-9). Most studies reported significantly higher content of organic carbon, nitrogen, phosphorus and potassium in the sub-crown environment than in the open (Belsky et al., 1989; Dunham, 1991; Kamara and Haque, 1992). Concentrations of cations such as magnesium, calcium and potassium, and cation exchange capacity (CEC) were also higher under tree cover (Belsky et al., 1993; Isichei and Muoghalu, 1992). In general, concentrations of these elements gradually decline with increasing distance from the tree bole, as well as with increasing soil depth (Barth and Klemmedson, 1978; Bernhard-Reversat, 1982; Kellman, 1979; Vetaas, 1992).

Parkland tree species

Nutrient concentrations under F. albida are generally significantly higher than in the open (Table 3.1). The tree effect is variable, however, and subject to several factors including land form, soil type, tree density and management practices. First, it is generally more pronounced in upland sites where soils tend to be more gravelly, shallower and less fertile than in the richer bottomland locations (Depommier et al., 1992; Saka et al., 1994). On the rich alluvial soils of the lakeshore plain of Malawi, soil nutrient status showed no difference under and away from F. albida trees, probably because the influence of trees is masked by the high natural site fertility and the effect of tillage practices (Rhoades, 1995). In the deep eutrophic soils of Dossi, Burkina Faso, the lack of tree-related nutrient difference was attributed to the regular and abundant application of livestock manure, as well as the fairly high and regular density of large Faidherbia tree crowns, all of which contribute to spreading out the fertility influence related to trees (Depommier, 1996a). Soil fertility under crowns also varies with the intensity of tree pruning, estimated to remove 50 to 100 kg of dry matter of leaves and twigs, some of which is recovered under tree cover in animal droppings.

Table 3.1 Improvement (%) of soil nutrient content under Faidherbia albida canopies compared to controls in the open

n.s. = not significant
OM = Organic matter

Nutrient enrichment under other parkland trees may be less remarkable than for F. albida but is also common. Kater et al. (1992), for example, found higher carbon, available magnesium and potassium contents in the upper soil layers under both P. biglobosa and V. paradoxa canopies than in the open. However, phosphorus availability was greater away from trees. This may reflect a lower phosphorus uptake due to a lower root density (Tomlinson et al., 1995). Jonsson (1995) reported higher exchangeable potassium and higher pH under V. paradoxa trees than in open field controls in Saponé, Burkina Faso. In addition, the fraction of carbon derived from C₃ plants (which reflects the effect of trees (C plants) in C-plant dominated systems)¹ was significantly higher in the proximity of both Vitellaria and Parkia trees than in treeless areas. The C₃-plant carbon contribution was 30 percent higher under P. biglobosa than in the open (Jonsson, 1995). Areas outside Vitellaria canopies were also characterized by a slight but significant higher organic carbon, potassium and pH than zones at mid-distance between trees in southern Burkina Faso (Boffa et al., forthcoming). In Mali, Vitellaria sites were associated with higher content of exchangeable calcium and magnesium than treeless locations, while the opposite was true for aluminium (Diakité, 1995). In a sample of 54 trees located in Burkina Faso and Nigeria, total nitrogen and available potassium were significantly improved under P. biglobosa (Tomlinson et al., 1995). Organic carbon, total nitrogen, available phosphorus, exchangeable calcium and CEC were respectively 57, 61, 47, 22 and 15 percent higher under Cordyla pinnata cover than outside in Senegal (Samba, 1997). In Kareygorou and Say, Niger, soil pH, organic matter content, total nitrogen, available phosphorus, exchangeable potassium, magnesium and calcium were significantly higher under Hyphaene thebaica (and F. albida) than in the open and decreased with distance from trees (Moussa, 1997). Soils under Prosopis africana contained a higher concentration of total nitrogen and exchangeable bases (particularly magnesium and potassium), and had a better pH and CEC (Bernard, 1996).

¹ C₃ and C₄ plants have slightly different photosynthetic (carbon dioxide fixation) mechanisms, with the C₄ process particularly adapated to high light intensities and temperatures. Most C₄ species are monocotyledons and they include the majority of savanna grasses and cereal crops (except rice), while trees and other dicotyledonous species tend to use the C₃ mechanism.

Trees significantly influence the fertility of tropical soils by maintaining soil organic matter. Cation exchange capacity is primarily determined by clay content and mineralogical composition. However, organic carbon provides much of the CEC of tropical soils which are based on kaolinite clays rich in iron and aluminium oxides and hydroxides (Jones, 1971; Pichot, 1975; Kater et al., 1992). Furthermore, organic carbon also has a greater influence on the size of the exchange complex and soil cations on sandy soils than on fine-textured ones (Campbell et al., 1994).

Orientation sometimes has a significant impact on soil nutrient content due to the action of dominant winds or asymmetrical crown distribution on accumulation of organic matter. For instance, organic carbon, pH, calcium and magnesium were all significantly higher in soils located to the west of V. paradoxa trees than to the east in Mali, where northeastern harmattan winds prevail (Diakité, 1995). Higher moisture in an Ethiopian vertisol was also recorded on the west side of F. albida crowns, probably as a response to wind-accumulated litter (Kamara and Haque, 1992). In Malawi, nitrogen was mineralized more rapidly on the north and east sides of large F. albida canopies compared to the south, due to the fact that canopy volume is greater on the north side of trees where the sun trajectory is during the growing season (Rhoades, 1995).

Tree species affect the nutrient content of soils in different ways. For instance, in south Mali, soils (0 to 40 cm horizon) under P. biglobosa trees had a lower pH and less available calcium and magnesium than under V. paradoxa trees (Kater et al., 1992). Likewise, higher Mg content and acidity were reported on Adansonia digitata sites than under Acacia tortilis in Kenya (Belsky et al., 1989). Nutrient content of leaves, the extent of nutrient reabsorption prior to leaf abscission, and amounts of litterfall influence nutrient concentration in litter (Vitousek and Sanford, 1986, cited in Rhoades, 1995).

Fig. 3.1 Vitellaria paradoxa (right) and Parkia biglobosa (left) parkland in the Bassila region of Benin
K. Schreckenberg

Influence of trees on mineral content of understorey plants

The generally higher soil nutrient status under tree cover is reflected in the mineral content of understorey herbaceous species. In northern Senegal, nitrogen content of aerial herb parts was higher under Acacia senegal and Balanites aegyptiaca than in the open. It also decreased more slowly during the rainy season under trees than in the open (Bernhard-Reversat, 1982). Under F. albida in Senegal, concentrations of all mineral elements except insoluble ash and sulphur in millet leaves were 25 to 40 percent higher than in the open. Protein content of millet grain increased by 32 percent under F. albida, and by 242 percent on a kg/ha basis, due to the grain yield increase related to the presence of trees (Table 3.2) (Charreau and Vidal, 1965). Nutrient concentrations were also higher in millet grown under F. albida than in the open in N'dounga, Niger (ICRAF, 1996). A 54 percent increase in nitrogen and a smaller but significant difference in available phosphorus were recorded. In Nigerian savannas, however, no significant differences in mean crude protein, fibre and lignin content were observed between forb species growing under tree canopy and in the open (Muoghalu and Isichei, 1991). It should be remembered that, in order to monitor the tree effect correctly, account should be taken of the variation in understorey biomass, which depends on moisture availability, to some extent irrespective of mineral plant nutrition. Mineral elements can often be diluted with an increase in biomass.

Table 3.2 Crop yields under and outside Faidherbia albida canopies

Notes: a 1=open controls; 2= sub-canopy plots
b Difference = ((sub-canopy yield - open field control yield) / open field control yield)) × 100
c Average over the whole sub-canopy area
d Total biomass
e Although smaller than West African values in relative terms, Poschen's (1986) sorghum and maize yield improvements in Ethiopia are higher in absolute terms because his control values are about twice as high as West African ones.

Mechanisms of soil fertility improvement

Comparative increases in nutrient content are highly significant under Faidherbia albida trees, and less remarkable though still common in other parkland species.

Several factors contribute to the fact that soils under the cover of parkland trees have a higher fertility status than in the open field. These include soil microbial activity, atmospheric inputs, nitrogen fixation, dung deposition, pre-existing soil fertility and soil management practices.

Soil microbial activity and structure

The primary processes held responsible for the formation of high fertility islands around trees relate to enhanced biological processes associated with the seasonal and long-term return of nutrients accumulated in trees to the soil through litterfall, root decay and exudation, and their mineralization, as well as leaching of nutrients stored in canopies.

Increases in organic matter and improved microclimatic conditions under parkland (and savanna) trees enhance soil microbial and enzymatic activity, decomposition and physical characteristics. Compared to open sites, biological activity is two to three times higher under F. albida which loses its leaves at a time when conditions for microorganisms are favourable (Jung, 1970). In Kenya, Belsky et al. (1989) found 35–60 percent higher soil microbial biomass-carbon, lower bulk density of top soils, and higher water infiltration rates under Adansonia digitata and Acacia tortilis crowns than in the open. Increased macrofauna and soil organic matter under tree clumps were associated with improved soil macroporosity and lower bulk density in Côte d'Ivoire (Mordelet et al., 1993, cited in Rhoades, 1997). Higher activity of macrofauna such as termites may also be one of the factors responsible for changes in textural properties of soils at tree sites (see Box 3.1).

Rates of nitrogen mineralization are also higher under tree canopies than in the open, with an intense flush during the first few weeks of the rainy season (Jung, 1970; Belsky et al., 1989). Soils under F. albida in Malawi contained 7 times more plant-available nitrogen than in the open during the first month of the rainy season, and 1.5 to 3 times more during the rest of the cropping season (Rhoades, 1995). This phenomenon was also observed by Depommier (1996a) early in the rainy season in Burkina Faso. However, although the early rainy season peak of mineralized nitrogen, which is predominantly in the highly mobile nitrate form, is potentially beneficial to young plants, it may be leached to a large extent before crop roots have developed sufficiently to absorb it. Rhoades (1997) therefore recommends mixing F. albida leaf litter with lower quality (high carbon:nitrogen) plant material in order to obtain a more gradual nutrient release.

Atmospheric inputs

Fine soil lost through wind erosion may be intercepted by trees and deposited by throughfall and stemflow. It was calculated that an annual rate of 1.2 kg/ha of phosphorus is deposited with dust from the Sahara 1 000 to 1 200 km away (McTamish and Walker, 1982, cited in Kessler and Breman, 1991). Dust deposition may be particularly important in species retaining their foliage during the dry season when strong harmattan winds prevail. Studies comparing nutrient concentrations in rainwater in the open and in throughfall detect differences. For instance, nitrogen enrichment of 0.53 g/m² was measured beneath Acacia senegal (Bernhardt-Reversat, 1982). In contrast, no difference was noted in mineral nitrogen content between rainwater collected under F. albida canopies and in the open (Jung, 1969). Similar tests in Saria, Burkina Faso, revealed that V. paradoxa canopies had absorbed nitrogen and maybe phosphorus, from rainwater and released potassium, organic carbon, calcium and magnesium (Roose et al., 1974). These studies do not, however, specify whether these nutrients are recycled through leaching or added to the system by atmospheric deposition.

Nitrogen fixation

Trees also increase soil nitrogen availability due to N-fixation. However, Dommergues (1987) assessed the N-fixation potential of legumes in West Africa to be lower than expected. One reason is that few woody species have been reported to nodulate naturally in the Sahel and Sudan zones of West Africa. While nodulation is common for F. albida seedlings, N-fixation is rarely reported in adult trees (Dunham, 1991; Giffard, 1971). Therefore, the contribution of increased soil nitrogen under F. albida due to N-fixation is limited (CTFT, 1988). An absence of nodules was also noted in roots of mature Acacia senegal trees (Bernhardt-Reversat, 1982). N-fixation in forest stands may in reality be important when trees are young and is considerably reduced with increasing stand age when the pool of soil nitrogen is sufficient to supply them (Dommergues, 1995). One could hypothesize that some degree of N-fixation continues into adult tree age in agroforestry parklands, where nitrogen is extracted by crops and mostly not returned to the soil (Harmand, 1998).

The lack of difference in soil nitrogen beneath Adansonia digitata and the leguminous Acacia tortilis indicated that nitrogen enrichment under these trees was not related to N-fixation (Belsky, 1994). Other leguminous parkland species such as P. biglobosa do not nodulate in the field or in the greenhouse. However, 85 percent of this species' roots are infected with endomycorrhizae regardless of location and tree size, suggesting an increased nutrient uptake capacity of the tree (Tomlinson et al., 1995). N-fixation is mostly limited by low availability of phosphorus, which is one of the most common nutrients limiting plant production in semi-arid regions (Penning de Vries and Djitèye, 1991). Finally, Breman and Kessler (1995) argue that no evidence points to an increase of nitrogen yields or concentrations in associated herbs and crops with increased cover of woody legumes over a north-south regional gradient, and conclude that fixed nitrogen is mostly utilized by the woody plants themselves. More recently, experiments in Cameroon have demonstrated that this process may occur on a local scale (Harmand, 1998).

Dung deposition

Increased fertility under trees may also be due to bird droppings and, in parkland systems which integrate livestock, dung deposition by animals which rest and feed under tree shade (Belsky et al., 1989). In Burkina Faso, the average amount of faeces deposited by cattle under F. albida crowns was higher than in the open field by only 11 percent in Dossi but up to 180 percent in Watinoma (Depommier, 1996a). The spatial variability of organic inputs by cattle, compounded by horizontal movements through runoff and floods which depend on topography, tends to generate differences in nutrient content observed between sub-canopy and open sites. Where livestock is a significant component of farming systems, its contribution to the overall soil fertility improvement of Faidherbia tree is probably considerable and should be specifically assessed. However, nutrient enrichment under F. albida was also demonstrated in the absence of livestock influence in Senegal (Charreau and Vidal, 1965). The tree effect may be more pronounced where livestock is excluded than in natural agrosylvipastoral systems (Poulain, 1984, cited in CTFT, 1988).

Pre-existing soil fertility

Higher fertility of tree sites has also been attributed to pre-existing soil fertility variations, which can occur at several scales. Variations at the village level are consistent with farmer practices of identifying fertile soils to be cleared for cultivation based on local vegetation characteristics. A high density of individual woody species, such as Isoberlinia doka, Pterocarpus erinaceus, Detarium microcarpum, Prosopis africana, Piliostigma sp. and Pteleopsis suberosa, (Bagnoud, 1991 and 1992, in Cissé, 1995), as well as grasses, and green and dense foliage are typical fertility indicators. In addition, in a F. albida trial in Niger, Geiger and colleagues (1994) observed that areas of good F. albida growth were associated with favourable microsite soil physical and chemical conditions. They suggested that these fertility characteristics, present before tree establishment, may be partly responsible for the high productivity found under mature F. albida trees. In Malawi, the positive impact of large F. albida trees on nitrogen mineralization rates, soil moisture increase and content of exchangeable cations was not noted for small trees. This led Rhoades (1995) to conclude that improved crop yields under large trees resulted from the combined tree effect on microclimate and the influence of litter/root inputs on soil nutrient availability, rather than patterns of pre-existing soil fertility. While it has mostly been studied for F. albida, pre-existing soil fertility could also apply for other species.

Soil management practices

Finally, fertility variations under and outside of trees can be related to human activities. Due to the absence of, or limited, cropping, there may be a higher soil fertility status under P. biglobosa (Kessler, 1992). Weed production is generally increased under parkland trees and more intense weeding is needed (Libert and Eyog-Matig, 1996).

Nutrient enrichment with increasing tree size

Small trees induce little fertility change in their soil environment. Kellman (1979) found that no significant nutrient enrichment could be found under Pinus caribaea saplings in a neotropical savanna of Belize. The process of soil amelioration under trees results from the trees' ability to establish a plant-litter-soil nutrient cycle which increases with time (Kellman, 1979). In Malawi, small (6.6 m in crown diameter) F. albida trees did not increase net nitrogen mineralization rates relative to open sites. In contrast, larger canopies (24 m) resulted in 170 percent more nitrogen production during the growing season than in the open field (Rhoades, 1995). With less than one-tenth of the canopy area of larger trees, small trees produced significantly less organic litter and root turnover inputs. Unlike larger trees, small ones also had no dung deposited beneath them.

Fig. 3.2 Crop and clearing residues are gathered and burned before the next agricultural season
J.-M. Boffa

Other reports also suggest that nutrient enrichment by trees increases with tree size. Bernhard-Reversat (1982) showed a clear, positive relationship between tree diameter and carbon and nitrogen content in soil under crowns of Acacia senegal and Balanites aegyptiaca in northern Senegal. Higher values of soil chemical variables were also detected under as opposed to outside canopies of 10-year and older saplings in a tropical dry forest of Mozambique (Campbell et al., 1990). In savannas of northwestern Nigeria, soils under trees above 7 m in height had higher concentrations of organic matter, exchangeable cations, clay and silt than under trees smaller than 7 m (Isichei and Muoghalu, 1992). Concentration of exchangeable cations (calcium, magnesium, potassium and sodium) around small F. albida trees was considerably lower than in the proximity of large trees in Malawi (Rhoades, 1995). In Tanzania, F. albida had no visible effect on intercropped maize and bean yields from one to six years after establishment, at which point it averaged 9 m in height (Okorio and Maghembe, 1994). Once F. albida trees mature, tree size has a favourable effect and is correlated with grain yield productivity, probably due to a larger input of litter (Depommier et al., 1992). The ‘Faidherbia albida effect’ may require 20 to 40 years to become evident, depending on the growth rate of individual trees (Poschen, 1986). Improvement of nitrogen and potassium content under P. biglobosa crowns also increases with tree size (Tomlinson et al., 1995).

The process of soil amelioration under tree results from the tree' ability to establish a plant-litter-soil nutrient cycle.

These references show that young trees do not seem to influence the size of the nutrient pool significantly, and that the nutrient concentration of sub-canopy soils expands with tree size. More specific information is needed on the dynamics of soil fertility with increasing tree size in relation to the performance of associated crops, and recommendations on size/age and related conditions of tree stands from which increased nutrient availability can potentially generate enhanced crop yields.

Parkland density and soil fertility

Most studies reviewed in this chapter take place at the scale of individual trees. However, a comprehensive assessment of the system also requires comparisons at the scale of a plot with two or more trees, a field or a land form unit. This section presents this level of analysis with respect to soil fertility, with crop yields and microclimate being dealt with in later sections. While nutrients may be concentrated in tree sites through various pathways, processes such as litterfall, decay of lateral roots, activity of macrofauna, transport of organic debris through water runoff and wind, as well as lateral movement of parkland-derived organic matter through cattle droppings and soil management practices, may involve the movement of nutrients from trees to the open field.

It could be hypothesized that the ‘parkland effect’ (i.e. more than one tree) on soil fertility in the open field may become detectable over a given range of tree density. This range may be intermediate between treeless agricultural soils whose stock of organic matter and fertility are gradually impoverished under continuous cultivation (Taonda, 1995), and soils under fallow which allow for the biological pumping of mineral elements and the restoration of organic matter in topsoil (Harmand, 1998). For instance, Depommier (1996a) argued that the high density and regular arrangement of F. albida trees in Dossi could, among other factors, contribute to the lack of difference in nutrient concentration in subcanopy and open soils.

Data relating tree density and soil nutrient status in savanna or parkland situations are relatively scarce and inconclusive. Total nitrogen in the 0–15 cm horizon was significantly positively correlated with tree density in Nigeria (Sanford et al., 1982). The fact that soil productivity in the Peanut Basin in Senegal depends almost entirely on soil organic matter content led Seyler (1993) to postulate that farmers manage the woody component in fields to manipulate the organic matter and nutrient content of soils. Based on samples in the 0–20 cm horizon at 20 m from F. albida crowns on four soil types, no significant relationship was found between cover of F. albida or other woody species in sample fields and soil organic matter in the open. When considering total parkland cover, the relationship was only slightly strengthened. The effect of tree density on soil organic matter content in open areas was not apparent in this study of limited sample size.

The high variability in fertility among tropical soils, even within a field, and the two-dimensional scale of sampling units including more than one tree at a time pose an important challenge in studying the fertility-density relationship. GIS and modelling technologies may facilitate processing and representation of such data needed to advance understanding of this particular area.

There may have a positive ‘parkland effect’ on crop production linked to the spatial arrangement of scattered parkland trees.

Kessler and Breman (1991) caution that attempts to increase parkland density should evaluate processes which influence nutrient availability in these systems (see next section). Where trees improve understorey soil nutrient content through spatial redistribution, maximum tree densities are determined by the overall size of the system's nutrient pool.

Nutrient redistribution versus enrichment

Whether trees only act to redistribute nutrients already available in the system or actually increase nutrient availability is central to determining when and how integrating trees in cropping systems is beneficial. While the way each component may relate to the processes at play in overall nutrient dynamics is conceptually well established, the quantification of these processes remains limited.

Some researchers think that higher fertility close to trees results from nutrient redistribution and spatial concentration around woody plants through lateral uptake by roots, animal deposition, and wet and dry atmospheric deposition (redistribution on a larger scale) (Kessler and Breman, 1991; Kater et al., 1992; Tomlinson et al., 1995). A large number of investigators prefer to emphasize the importance of enhanced biological recycling of organic matter in higher system productivity. The only experiment identified in the literature reviewed, which challenges nutrient redistribution through lateral root uptake, originates in Matopos, Zimbabwe, where similar fertility patterns were found at natural savanna sites and on soils which have been experimentally managed with trenches since 1963. From this, Campbell and collaborators (1994) concluded that fertility under trees was not at the expense of fertility decline in the surface soils of the zone around trees. This probably depends on the rooting habits and density of tree species and the characteristics of soil profiles.

Trees may also increase system productivity by reducing nutrient losses through leaching in deep soil, and reduced soil erosion. However, the data necessary to document this process are difficult to obtain. In Mali, the majority of fine Acacia seyal and Sclerocarya birrea roots were found below those of herbaceous plants. Soumaré et al. (1994) therefore suggested that trees might capture nutrients which would have been lost in their absence, and improve efficiency of nutrient use. While leaching is common in humid conditions, however, the process is limited to run-on locations (valleys and water courses) in semi-arid zones (Kessler and Breman, 1991). The available yet limited data from northern Cameroon suggest that parkland trees do not reduce soil erosion significantly (Libert and Eyog-Matig, 1996).

Finally, trees may increase overall system productivity by increasing nutrient availability through N-fixation (as discussed earlier) and deep rooting, and their enlarged absorptive capacity associated with mycorrhizal and fungal infection. However, even though these processes may be important in particular sites with appropriate soil conditions and water availability, there are limitations to these processes in the Sahel and Sudan zones.

The deep penetration of tree roots is often highlighted as being important in the role of trees as nutrient pumps (Young, 1989). Tree roots are assumed to capture nutrients in deep soil layers made available by weathering of the bedrock and those leached down from upper layers. Yet this process may not always be significant. The highest concentrations of available nutrients, for example, are found in the topsoil; the subsoil's nutrient content is low (Kellman, 1979). Few nutrients are made available following weathering in already intensely weathered soils (Breman and Kessler, 1995). In the case of phosphorus, which limits primary production and whose mobility is low due to its high apparent adsorption constant, Breman and Kessler (1995) argue that, if weathering and subsequent uptake by deep tree roots were significant, one would find a correlation between phosphorus yield in herbage biomass and woody cover. Such a correlation was not found along a transect from the northern Sahel to the southern Sudan zones. Furthermore, the impact of leaching decreases in importance as aridity increases (Kessler and Breman, 1991). Finally, water infiltration to the subsoil is limited, and shallow plinthite pans (CILSS, 1981 and 1982, cited in Breman and Kessler, 1995) and chemical root barriers (acidity, salinity, aluminium or manganese toxicity, and phosphorus and calcium deficiency) are common in semi-arid areas (Szott et al., 1991).

Fig. 3.3 Cattle grazing on Prosopis africana pods in cotton fields, Holom, northern Cameroon
C. Bernard

Underground tree-crop interactions

Understanding the nature of interactions between trees and crops is of major importance in determining approaches to tree management in parklands. Savanna tree-grass interactions have been understood in a two-layer model which emphasizes that grasses have a root system confined to the upper soil and tree roots extend to deeper soil layers, thus minimizing the competition between the two (Walter, 1971; Walker and Noy-Meir, 1982). Other studies further suggested facilitation between woody and herbaceous plants, whereby trees improve soil nutrient content and herbage production, and affect the composition of herbaceous species nearby (Radwanski and Wickens, 1967; Kennard and Walker, 1973). It is now recognized that grasses and trees have a respective competitive advantage in capturing water and nutrients from the upper and lower soil horizons, yet herbaceous roots also extend into deep soil layers (Knoop and Walker, 1985). This suggests that competition and facilitation occur together in agroforestry, with variations in outcome according to habitat and tree species.

Numerous studies report the co-occurrence of tree and grass/crop roots throughout soil profiles (Belsky, 1994; Knoop and Walker, 1985; Jonsson et al., 1988). Generally speaking, the lateral spread of roots tends to be concentrated in the canopy area in humid zones, whereas they extend far beyond this area in more arid zones in order to acquire adequate supplies of soil moisture (Breman and Kessler, 1995). In a dry area of Kenya (450 mm rainfall), severing roots in the canopy zone by trenching did not alter understorey productivity substantially. But this was probably due to the fact that the sub-canopy roots were large rather than fine absorptive roots and did not compete with the herbaceous vegetation. Since trenching also had no effect in the open, Belsky (1994) believed that areas explored by fine absorptive roots located at the end of large, long roots in mature trees might have been overlooked in view of the spatial dispersion of these roots. Overall, the positive soil nutrient and microclimate effects of trees on understorey productivity exceeded the depressive effect of competition. In contrast, in higher rainfall (800 mm) areas, tree roots may explore a smaller area restricted within or shortly outside tree crowns where soil fertility is high. Trenching under tree crowns in mesic conditions often results in the increase of understorey productivity (Belsky, 1994).

Underground competition between unpruned F. albida and crops is thought to be small because of its dry-season physiological cycle of growth and its deep tap roots. This is consistent with the crop productivity gains under this species reviewed later. But the number of below-ground interaction studies remains limited. Deep rooting is common for this species as its growth has to rely on dry-season reserves. A three-year old F. albida excavated near Kano, Nigeria, had a vertical root shaft of over 10 m (Weber and Hoskins, 1983). In Tanzania, F. albida had no effect on maize and beans during the first six years of intercropping even at 4 × 4 m spacing (Okorio and Maghembe, 1994). In northern Cameroon, F. albida roots were not present in the 80 cm soil depth of cotton root extension. Available water content in soils measured by moisture probes was higher under trees at the beginning and end of the season but was little affected by the trees. Libert and Eyog-Matig (1996) concluded that the negative impact of F. albida on cotton production was due to shade rather than root competition.

Extended foliation during the rainy season caused by intense pruning (Box 3.2) prolongs Faidherbia's growth season (Depommier and Detienne, 1996) and may increase underground competition with crops for nutrients. However, given that farmers made no mention of it, any negative (above and underground) impact of trees in leaf during the rainy season on understorey crops may be insignificant compared with its otherwise positive fertility effects (Depommier, 1998). More detailed investigations into the effect of pruning on underground interactions would be desirable.

Lateral root systems of species with normal leaf phenology are likely to cause competition with crops in parklands. Large roots of V. paradoxa and P. biglobosa were found 60 m away from the trunk (Jonsson, 1995). The extensive lateral root system of P. biglobosa covered an area eight times that of the crown (Tomlinson et al., 1995). This species appears to display a more superficial rooting than V. paradoxa and results in more intense competition with crops (Kater et al., 1992). In Niono, Mali, the root system of Acacia seyal had a thin, deep taproot which could reach 6 m in depth, while lateral roots were concentrated in the upper 40 cm of soil and extended a distance of 5.6 times the average canopy radius. Sclerocarya birrea roots extended 2.4 m in depth and 7.4 times the average crown radius in the top 60 cm of soil (Soumaré et al., 1994). So far, the depressive effect of tree-crop competition and its spatial patterns have not been clearly measured and demonstrated in agroforestry parklands. Understorey crop (especially cereal) production seems to suffer more from the negative effect of shade from trees other than leafless F. albida. The specific effect of underground competition in tree-crop interactions needs to be isolated. Additional knowledge about the spatial patterns and functions of tree roots in parklands of distinct rainfall regions will help define the spatial variation of underground competition.

Relations between woody and herbaceous plants

This section reviews available data on the influence of woody plants on the biomass and grain production of herbaceous plants. Box 3.3 reviews some of the experimental designs used by different authors.

Influence of trees on biomass productivity

In savannas, herbaceous biomass under woody plants can be higher than in the open. Productivity close to trees and in the interspaces may vary by a factor of 1.5 to 4.5 (Tiedemann and Klemmedson, 1977; Barth and Klemmedson, 1978; Weltzin and Coughenour, 1990; Akpo and Grouzis, 1996; Belsky et al., 1989). A lower productivity under trees has also been reported (Grunow et al., 1980; Jameson, 1967; Somarriba, 1988), however, and other authors observed that removing woody plants increased vegetation yield (Dye and Spear, 1982; Beale, 1973; Ward and Cleghorn, 1964). Belsky (1994) noted that increased understorey productivity is the most common pattern in tropical tree communities with low density, low rainfall and moderate soil fertility. While common for low densities of isolated trees, the pattern of higher understorey herbaceous biomass changes as the density of woody plants changes (see below).

In general, understorey biomass decreases with increasing distance² from the bole (Barth and Klemmedson, 1978; Weltzin and Coughenour, 1990). A more pronounced effect of cardinal directions is occasionally reported. For instance, in the study by Coughenour et al. (1990), south and west directions were more productive.

Plant biomass of cereals under F. albida is significantly higher than in the open (Table 3.2). For cotton only a limited amount of information is available. Height of cotton plants under F. albida was greater than in controls away from trees, but plant survival was lower under canopies (Libert and Eyog-Matig, 1996). Overall sub-canopy biomass was not reported, but weed biomass in sub-canopy plots was two to six times higher than in open controls. The difference in productivity between tree-covered and treeless sites is substantially reduced when fertilizer is applied.

² Plant biomass according to distance from the bole was described by the following equation: Herbage dry matter m^-2 = 74.4 D^-0.42, where D = distance. The regression coefficient (r²) = 0.97 (Akpo and Grouzis, 1996).

Above-ground biomass (straw) productivity under Vitellaria and Parkia crowns tends to be less depressed than grain yields and sometimes positively influenced. Kessler (1992) reported that plant height of sorghum and millet was negatively affected under both Vitellaria and Parkia canopies but plant weights were not determined. Biomass measurements under medium-sized Vitellaria crowns and outside tree canopies were not significantly different, even though sorghum plant height was significantly lower under crowns (Boffa et al., 1999). Jonsson (1995) found that millet biomass was slightly but not significantly higher under Parkia and Vitellaria trees than in control areas in the open located 50 to 60 m from the trees. Variation in crop biomass was also lower under canopies than in the open. Moreover, in Mounzou, Mali, sub-canopy straw yields were higher than in the open on average, yet this was more obvious under the smaller trees (Diakité, 1995). Reduction of total millet biomass was only 0.5 percent under a combination of pruned and unpruned Cordyla pinnata trees in the southern part of the Peanut Basin in Senegal (Samba, 1997). In addition, grass production was 6.9 t/ha under P. biglobosa canopies compared to 3.3 t/ha outside (Sabiiti and Cobbina, 1992). These results support the fact that plant biomass productivity, especially for C₄ crops, may not be significantly lower under than away from parkland trees, as is often observed for natural vegetation in savannas. The outcome of the interaction is also influenced by several site conditions and tree characteristics.

Among C₃ plants, Louppe (1993) reported that cotton plants were taller and bolls were larger under Vitellaria crowns than in the open, much like under Prosopis africana (Bernard, 1996), whereas Kater et al. (1992) found that the height of cotton plants growing under tree crowns declined slightly.

Influence of trees on grain yields of crops

Because of its unique reverse foliation behaviour and its high potential for Sahelian agrosystems, interest in the effect of F. albida on crop performance has been ongoing in the last three decades. More recently attention has been extended to other major parkland species such as V. paradoxa, P. biglobosa and Azadirachta indica.

Faidherbia albida, a tree with reverse leaf phenology

The positive influence of F. albida on cereal production is considerable and can result in yield increases of over 100 percent³ under its canopy compared to open field controls (Table 3.2). As assessed by yield differences between tree sites and interspaces, the tree effect was similar to the effect of the combined application of organic and mineral fertilizer on fertile soils in Bambey, Senegal (Charreau and Vidal, 1965), and was equivalent to improving fertility from that of a poor site to one of good productivity as commonly observed in Sob, Senegal (Louppe et al., 1996).

³ Tree effect is expressed in percentage as ((sub-canopy yield - open field control yield) / open field control yield))×100

Soil type, topographic position and fertilization

Cereal productivity increases under F. albida are generally more pronounced on soils of low fertility and water availability. In the Sahel and Sudan zones of West Africa, fertility and water availability are strongly related to position in the toposequence. In the Ouahigouya region of Burkina Faso, F. albida caused a 78, 64 and 18 percent increase in millet yield on upland lateritic soils, sloping lands, and lowland plots, respectively (GERES-CTFT, 1965). The increase in yields was greater when growing under F. albida on low productivity soils (113 percent) than on high productivity plots in Sob, Senegal (62 percent) (Louppe et al., 1996). The difference in sorghum yields between sub-canopy plots and controls was also relatively greater on upland plots than on lowland sites (Depommier et al., 1992). In this last experiment, yields were more variable under trees than outside.

Faidherbia's effect may be particularly remarkable in conditions of low soil fertility, as noted above, in combination with below-average rainfall years. In 1990, a drought year in Watinoma, Burkina Faso, upland sites under Faidherbia cover yielded almost three times more than locations in the open (Depommier, 1996a). The food security implications of this Faidherbia-based agroforestry practice are therefore highly significant to farmers and justify the strong interest aroused by this species.

The influence of Faidherbia trees is most noticeable in the absence of fertilizer application. When fertilizer (60 nitrogen, 80 phosphorus, 60 potassium, 15 sulphur) was applied, the F. albida effect on grain yield virtually disappeared and substantially decreased on biomass (Table 3.2) (Dancette and Poulain, 1969). Poschen (1986) also noted a comparative advantage of trees in the absence of manure application and where drainage problems prevailed. In N'dounga, Niger, the application of 180 kg nitrogen per hectare cancelled the F. albida effect observed on total millet biomass. This clearly indicates that enhanced nitrogen availability was the major parameter contributing to millet productivity under the tree (ICRAF, 1996).

Distance from the tree

There is no consensus on the distribution of yields in the canopy zone. This may be linked to variations in sampling design and/or pruning intensity. Subcanopy millet yields were highest closest to the trunk (73 percent increase compared to an average 46 percent under the whole crown area) in Sob and gradually decreased towards the crown edge (Louppe et al., 1996; Charreau and Vidal, 1965). In contrast, subcanopy plots furthest from the bole showed highest productivity in the Bazega province of Burkina Faso (Maïga, 1987) and in Malawi (Saka et al., 1994). A similar pattern with low yields next to Faidherbia trunks and highest yields at the crown edge and gradually diminishing from there on was observed on sorghum in Dossi and Watinoma (Depommier, 1996a). Extended foliation in the rainy season caused by pruning (Box 3.2), and resulting shade, were held responsible for this grain yield depression next to the trunk.

Fig. 3.4 Faidherbia albida parklands are a significant source of fuelwood, especially where trees are regularly pruned
R. Faidutti

Yields outside Faidherbia canopies are generally intermediate between the crown zone and the open field. The extent to which cereal production is influenced beyond the crown appears to depend on the intensity of pruning. Pruning was often more intense in large trees, and the tree effect more pronounced at the edge of smaller trees, in Watinoma (Depommier, 1996a). Based on a regression analysis of bole and crown surface areas on unpruned trees, Louppe et al. (1996) estimated that the crown cover of F. albida trees in their experiment was reduced to 38 percent of its original area by pruning. The beneficial yet statistically insignificant impact of Faidherbia was felt a few metres beyond crown limits. These authors assumed that intense pruning led to a reduction of biomass that would otherwise have been incorporated into soils, and thus affected crop production negatively.

Influence on yield components

Yield components are mutually dependent, and variables measured and reported in the various studies differ so that the tree effect on specific yield components is more difficult to determine. It is also likely to vary locally. In Senegal, Charreau and Vidal (1965) found that the increased number of millet heads per planting hole (86 percent compared with control outside F. albida) made a greater contribution than increased grain weight per head (32 percent compared with control outside F. albida) to overall yield improvements, but plant density was not measured. In Ethiopia, yield increases resulted from the accumulation of higher grain weight, number of grains/plant and slightly higher plant densities under trees (Poschen, 1986). Trees had a significant positive impact on head weight and weight of grains per head in Sob, Senegal (Louppe et al., 1996).

A rigorous study of yield components may provide information on the effect of trees during stand establishment, tillering, spikelet initiation and flowering and grain filling. Farmer management of initial crop density in relation to parkland trees may be insufficiently stressed. Planting density in the proximity of F. albida crowns tended to be higher than in the open in Alemaya for sorghum (Poschen, 1986) and in Sob for millet (Louppe et al., 1996). The opposite trend was reported under V. paradoxa trees in Thiougou, southern Burkina Faso, in addition to a relatively higher rate of plant failure under Vitellaria in Thiougou (Boffa et al., 1999), and under both V. paradoxa and P. biglobosa trees in South Mali (Kater et al., 1992), than in the open.

Available projections of F. albida's influence on total crop yields at field scale include the following:

• +3 percent for millet, -2.4 percent for groundnut in Sob, Senegal with a density of five F. albida trees/ha (Louppe et al., 1996);

• +11 to 17 percent in Watinoma, Burkina Faso, with a density of 7 to 19 trees/ha (Oliver et al., 1996).

Influence on cotton

In contrast to the relative abundance of observations of F. albida intercropping with cereals, only a few studies are available on cotton associations (Libert and Eyog-Matig, 1996; Ouldra Malai, 1990, cited in Libert and Eyog-Matig, 1996). As opposed to the systematically positive influence of F. albida on cereals, cotton's response depends on the fertility status of the sites considered. For isolated trees in Tokombéré, northern Cameroon, the tree effect on cotton growing on relatively poor soils was positive, while competition led to a negative tree effect on more fertile soils (Libert and Eyog-Matig, 1996). The average weight of cotton bolls explained 67 to 72 percent of the variability in yields. Plant height under tree crowns was 24 percent greater on average and less variable than in the open. Trees were also responsible for a 20 percent decrease in plant survival.

Influence on groundnuts

Data on the influence of F. albida on groundnut production paint a variable picture (Table 3.2). This is reflected in farmer perceptions. Senegalese farmers were much less unanimous about the positive effect of F. albida on groundnuts than in the case of millet (Seyler, 1993). In studies with large enough sample sizes to test statistical significance, pod yields were negatively influenced in Sob, Senegal (Louppe et al., 1996), but positively affected in Silane (Dancette and Poulain's data cited in CTFT, 1988). When positive, the influence of F. albida on groundnuts is nevertheless substantially lower than that usually reported on cereals in similar regions (Table 3.2). Groundnut response is related to the difference in soil nutrient levels under and away from trees (IRHO, 1966). Where phosphorus or potassium was shown to be deficient, groundnut yields were higher under F. albida than in the open, but they remained unchanged in fertile soils. Lower pod production under Faidherbia trees in Sob was attributed to higher nitrogen levels in the soil leading to an increase in stalk biomass at the expense of pods, and to the low potassium and phosphorus content due to insufficient returns of stalks to soils, as well as reduced sun radiation during flowering (Louppe et al., 1996).

The F. albida effect on stalks is more commonly positive than on pods for groundnuts. When soil fertility levels are improved through application of fertilizer (Dancette and Poulain's data, 1966, in CTFT 1988) or cattle faeces (Louppe et al., 1996), the difference between sub-canopy and control locations decreases or is reversed.

Vitellaria paradoxa, Parkia biglobosa and other species with typical leaf phenology

Unlike the situation with F. albida, the response of cereals to trees with typical leaf phenology, such as V. paradoxa and P. biglobosa, can be substantially negative (Table 3.3). Yields are generally lowest next to the bole and gradually improve with increasing distance from the trunk (Kessler, 1992). However, the fact that this general pattern does not always prevail raises questions. In Saponé, Burkina Faso, no significant differences were found in millet grain production growing under Vitellaria and Parkia canopies and in the open (Jonsson, 1995). Furthermore, in Mounzou, Mali, sorghum grain productivity was significantly superior under Vitellaria trees than at a distance of three times the canopy radius, but this was more conspicuous under the smaller trees in the experiment (Diakité, 1995).

Fig. 3.5 Vitellaria paradoxa parkland in the dry season, Thiougou, southern Burkina Faso
J.-M. Boffa

Table 3.3 Impact of Vitellaria paradoxa and Parkia biglobosa on sub-canopy crop yields (%)^a

Notes:^a Measured by the ratio ((subcanopy yield - open field control yield) / open field control yield))¥100. It should be noted that experimental designs in the above studies differed, including the location of ‘control’ plots to measure tree effect.
^b Difference between sub-canopy plots and plots at mid-transect locations between two trees.
^c Difference between yields of plots located at half crown radius and at a distance of three crown radii.
^d This positive ratio was obtaine don plots of higher soil fertility.
^e Difference between the average of the whole sub-canopy zone and the most distant concentric ring.
n.s. = not statistically significant.

Soil type

While F. albida gives rise to greater yield improvements on soils of low fertility, V. paradoxa has a more pronounced negative effect on crop yields on soils of low as opposed to high fertility. For example, cotton yields on low fertility sites are more depressed (-32 percent) than on more fertile locations (-14 percent). The same applies for groundnuts (-20 versus +7 percent) and for maize (Louppe, 1993). Prosopis-crop interactions may, however, be more favourable on sites of low soil fertility. While a majority of farmers in Holom, northern Cameroon, declared that the effect of P. africana on soils was positive, its effect on cotton was not significant as it was probably concealed by the regular application of fertilizer (Bernard, 1996).

Influence on groundnuts and cotton

The effect of Vitellaria and other trees tends to be less severe on C₃ plants than on cereals (Table 3.3). In Côte d'Ivoire, average groundnut yields were 6 percent lower under Vitellaria canopies than in the open (Louppe, 1993). Picasso (1984) also reported that Vitellaria trees had a negative (but unquantified) impact on groundnut yields within a 4 m radius of the bole (cited in Hall et al., 1996). A 25 percent reduction rate was reported for groundnut yields under Cordyla pinnata in Senegal (Samba, 1997). Cotton seems less affected by Vitellaria than by Parkia (Table 3.3) trees. In Holom, northern Cameroon, there was no significant difference in overall cotton yields under Prosopis africana trees and in the open (data collected by Libert in 1990 and analysed by Bernard, 1996).

Distinctions between tree species

The ecological combining ability of trees with given crops is a species-specific characteristic related to branch and root architecture. Prosopis biglobosa generally has a more pronounced negative effect on crop production than does V. paradoxa. This is the case for sorghum (Maïga in Kessler, 1992; Kessler, 1992; Kater et al., 1992) and to a lesser extent for millet (Maïga in Kessler, 1992) (Table 3.3). This is partly due to the larger size and shape of P. biglobosa and different rooting patterns. Vitellaria trees have an ascending architecture, while Parkia branches are low and extend further laterally. In addition, Kater et al. (1992) observed that average panicle weights of sorghum and millet at one crown radius away from the canopy boundary were lower on Parkia plots than on Vitellaria sites. This difference was dramatic with millet, suggesting that superficial rooting is more extensive in P. biglobosa, thus resulting in more intense competition with crops. While V. paradoxa hardly influenced cotton performance, cotton yields were negatively affected by P. biglobosa, mostly in the vicinity of its trunk. Because cotton did not suffer under cover of either species during establishment, Kater et al. (1992) concluded that the significant cotton loss under P. biglobosa was mostly due to tree/crop (possibly light) competition, and advised pruning experiments.

Crops are combined with some other tree species with little negative effect. The depressive effect of Azadirachta indica on sorghum is slight. In Burkina Faso, the sub-canopy sorghum yield was significantly lower by around 20 percent than at the edge of the crowns, but neither of these positions had significantly different yields from the open field (Tilander et al., 1995). Similary, there was no significant difference in sorghum production at four distance under and away from A. Indica countries (Zoungrana et al., 1993). In contrast to V. paradoxa, P. biglobosa and A. indica, Hyphaene thebaica had a significantly positive effect (2 to 2.5 times) on grain and straw yields of sub-canopy millet in Kareygorou and Say, southwestern Niger (Moussa, 1997). This is recognized by farmers and may be due to the palm's high and small crown and horizontally limited root system. Total dry matter production of millet and soil fertility were both even higher under H. thebaica than under F. albida trees in Kareygorou, though this was not true in Say (Moussa, 1997). Borassus aethiopum is also believed to associate with crops without intense competition (Cassou et al., 1997).

Cardinal directions

Cardinal direction does not usually result in significant variation of overall production. Nevertheless, occasional studies report trends on the influence of orientation. In a system with successive harvests, Louppe (1993) observed that the first cotton harvest was larger in the open than under Vitellaria trees, but this pattern was reversed for the second harvest. The southern orientation was associated with higher yields in the first harvest and lower yields in the second. Average sorghum grain and biomass productivity on plots to the western side of Vitellaria trees was higher than on the eastern side in Mounzou, Mali, but not significantly so (Diakité, 1995). A similar pattern with statistically significant differences was observed for soil content of organic matter, calcium, magnesium and pH, as well as soil moisture in September. This result, due to dominant north-eastern winds influencing the direction of litterfall, confirmed the importance of increased soil fertility in improving yields in west-oriented plots.

Tree size

There are few data available on the influence of tree size on tree-crop interactions. Tree size is generally positively correlated with production gains under canopies of F. albida (Depommier, 1996a). The opposite may prevail with species with typical phenology, but the few reports do not provide a consensus opinion. Sorghum grain and biomass yields under Vitellaria trees with average crown diameters of 9.9 m were respectively 70 to 80 percent higher than under trees of larger diameter (15.4 m) (Diakité, 1995). This was attributed to the higher interception and evaporative demand of the larger canopy group. In a semi-arid (422 mm average rainfall) thornveld dominated by Acacia karroo trees ranging from seedlings to 3 m high mature trees, grass growth was more suppressed under trees taller than 2 m than under trees of 0.8 to 1.2 m height, possibly due to competition for water and nutrients (Stuart-Hill and Tainton, 1989). In contrast, in Nigeria (with > 1 000 mm rainfall), herbaceous production was depressed under trees smaller than 3 m, whereas it was unaffected or positively influenced under trees higher than 7 m (Sanford et al., 1982).

Parkland tree density and ‘parkland effect’

One of the primary objectives of biophysical research on parklands should be to indicate the appropriate tree densities necessary for sustaining or improving parkland management. Unfortunately, available information on relationships between parkland density and crop production is very limited. Probably due to logistical difficulties and the only recent awareness that more intensive tree management should be pursued in traditional Sahelian agroforestry systems (ICRAF, 1995b), studies so far have been mostly limited to individual trees.

The relationship between herbaceous and woody plant productivity is complex and needs to be analysed within a given vegetation type, because it is influenced to a great extent by climatic and edaphic conditions (Breman and Kessler, 1995). Within a given vegetation type, increasing tree densities are usually associated with lower herb productivity. Both water and nutrient availability influence the competitive relationship between herbaceous plants and trees. Sites of lowest productivity can be expected to show the greatest rate of decline in herbage production with increasing tree density. Trees can also have a more intense negative effect on herbaceous biomass production in years of low rainfall.

The concept of ‘critical canopy cover’ seems highly relevant to the determination of optimum agroforestry parkland density from a biophysical point of view.

In semi-arid conditions, understorey herbaceous productivity is highest with low tree densities and decreases with increasing density. For instance, in Zimbabwe, the average yield of the herbaceous layer was 4.5 t/ha under open tree canopies compared with 2.8 t/ha in closed canopy sites and 3.0 t/ha in open grassland sites (Kennard and Walker, 1973). In areas afforested with Albizzia lebbek, Prosopis cineraria, Tecomella undulata and Acacia senegal, Ahuja et al. (1978) noted that herbage production was most depressed under A. senegal which had been planted in the highest densities. In savannas of northwestern Nigeria, Sanford et al. (1982) also found that higher production was achieved under a light canopy than in the open or under a dense canopy.

Other studies report an increase in understorey productivity with the artificial reduction of density or cover of woody plants (Breman and Kessler, 1995). These suggest the existence of a critical level below which herbage production is no longer increased with further cover reductions. Below the critical canopy cover, factors other than tree cover determine herbage production and/or beneficial effects associated with trees prevail over competitive ones. As canopy cover exceeds this critical level, herbage production decreases due to tree-herb competition. Breman and Kessler (1995) hypothesized that the critical canopy cover averages 15 percent in the Sahel and Sudan zones of West Africa for woody plant communities with two leaf layers.⁴. This level would decrease going north as rainfall levels fall, and increase toward the south where annual rainfall is higher.

The concept of ‘critical canopy cover’ seems highly relevant to the study of agroforestry parklands (less so for the less competitive F. albida) and to the determination of optimum parkland densities from a biophysical point of view.

⁴ The average number of leaf layers in natural woody plant communities ranges from 1.5 in the northern Sahel region to 4 in the southern Sudan region (see Breman and Kessler, 1995).

Farmers seemingly attempt to reach a certain range of densities perceived as ‘ecologically-optimal’ as they clear woodlands for agriculture. For instance, farmers reported clearing a relatively higher number of Vitellaria trees where their natural density was high (Boffa, 1995). Vitellaria stump counts in recently cleared fields showed that the number of eliminated trees per unit area was highly correlated with the original Vitellaria density (r²=0.97). Farmers would spare a larger number of other useful species where V. paradoxa densities were originally low. They may also attempt to regenerate F. albida trees in fields of low stocking in order to reach a more productive density level, which is itself probably higher than the critical density level of Vitellaria parklands. Farmers are well aware that increasing tree density is desirable for crop production up to a certain level. In Holom, northern Cameroon, they reported that the biomass productivity of cotton increases to the detriment of cotton boll yields in the presence of too many P. africana trees (Bernard, 1996).

In northern Cameroon, a 32 percent increase in cotton production in plots located under a relatively dense F. albida parkland cover (37 trees/ha, 26 percent crown cover) was noted, compared with plots in an adjacent bare area (Libert and Eyog-Matig, 1996). This was unexpected because isolated trees tended to be associated with depressed cotton productivity. These results may suggest a positive ‘parkland effect’ linked to a spatial arrangement of scattered parkland trees, which would not exist in the presence of isolated individual trees and would consist in a synergetic increase in soil and air moisture, as well as less air circulation. Available evidence on this mechanism is presented in the next section. However, the observed difference could also be due to higher inherent soil fertility and better germination conditions of the area with trees (Geiger et al., 1992).

It would appear that cereal crops would benefit from high F. albida densities, yet the productivity gain may decline as critical densities are reached. While high stocking levels have been reported in the Sahel, the relation to yields is rarely established. This parkland type illustrates the practical limits to the notion of critical canopy cover, as production objectives other than crop-oriented ones probably result in lower densities than those that maximum crop yields would dictate. Available data show that the highest yield increase is not necessarily positively related to density. In Watinoma, Burkina Faso, plots with the lowest F. albida density among three study sites located in a bottomland area showed the highest yield increase/tree density ratio. Tree size, pruning (or absence of), as well as site productivity also account for this result (Oliver et al., 1996).

In V. paradoxa parklands of southern Burkina Faso on leached tropical ferruginous soils, the projected influence on sorghum productivity along canopy-wide transects between two trees was positive for densities of 12 and 30 trees/ha and negative for smaller trees at densities of 43 trees/ha (Boffa et al., 1999). Although based on a limited sample size and a single experimental season, these results suggest that the maximum recommended density for medium-sized Vitellaria may lie somewhere between 30 and 43 trees/ha in this particular area.

An alternative approach was used by Bertelsen and Kaboré (1993). They argued that, in spite of the well-known yield variability in space and time, tree-crop interaction studies tended to focus on limited areas under and around trees rather than whole fields, and that yields were generally measured in single agricultural seasons. During interviews, farmers from two villages of Burkina Faso's Central Plateau gave expected yields for a total of 87 fields. Densities of trees, broken down into ‘small’ and ‘large’ classes, were determined from aerial photographs. The use of expected yields better reflected the long-term nature of tree density decisions. There was a strong positive relationship between densities (V. paradoxa mostly) and yields suggesting that a 1 percent increase in tree density resulted in a crop yield improvement of 0.5 percent.

Such findings underline the important contribution of trees to overall soil productivity of agroforestry parklands in this region. It should be noted, however, that the biophysical parameters dealt with here are far from being the only ones influencing farmer decisions on optimal parkland densities. As illustrated in other chapters of this review, a range of cultural, socio-economic and institutional parameters, as well as objectives other than crop yields also come into play.