ContentsWhat role for Agroforestry?
The water balance of an agroforestry system
Water use efficiency in tree/crop mixtures
Resource capture: complementarity or competition?
Competition for below ground resources: root structure and function
Trees as ‘generators’ or ‘consumers’ of rainfall
Verslagen van Landbouwkunige Onderzoekingen
Resource Capture by Crops
Tree-crop interactions - a physiological approach
Arch Me Geophys Bioclim Ser B
Agricultural and Forest Meteorology
C. K. Ong
International Centre for Research in Agroforestry
Much of the future increase in food and wood production in the tropics will have to be achieved from existing land and water resources. This focuses ICRAF’s research agenda on the challenges of improving the efficiency with which land and water are currently used, especially in the semiarid areas. Agroforestry has the potential for improving water use efficiency in two ways. This can be achieved directly, via more water being used as tree or crop transpiration, or indirectly, via improvements in the transpirational water use efficiency of the tree and/or crop components. Indirect improvements in water use efficiency are a consequence of modifications to the microclimate of the crop due to the presence of the tree canopy.
The hypothesis that agroforestry may improve productivity by capturing a greater proportion of the annual rainfall has gained support in recent years. Evidence from semiarid India and Kenya showed that the annual transpiration of agroforestry systems was double the water use of the most productive intercrop system. Almost half of the total water use occurred between the dry season when cropping was impossible and the rest was extracted from soil reserves.
In recent years, the replacement of native forests by fast-growing trees and deforestation on major catchments has created hydrological concerns in several tropical countries. Native forests are popularly believed to be ‘generators’ of rainfall while exotic forests are often regarded as ‘consumers’ of rainfall. The most notable problem is the high demand for water by fast growing exotic trees such as eucalyptus in semiarid areas. On the other hand, removal of deep rooting native vegetations has resulted in salinity and rising water-table, especially in Australia. Consequently, it is essential to consider the implications of increased water use on the medium and longer-term water budgets.
This paper describes some of the technical approaches that can be used in agroforestry to improve land and water management, the role of land use and its relation to hydrology, and the challenges for rational land use decision- making.
Much of the future increase in food and wood production in the humid tropics (and elsewhere), necessary to meet the needs of increasing populations and to reduce hunger and poverty, will have to be achieved from land and water resources already in use. Field observation shows that the extent of the ‘land balance’ – land which could be used for productive purposes but is not currently in use – is very limited. Estimates by FAO and associated organisations appear to show substantial areas, which are cultivable but not presently cultivated. However, the validity of these estimates has recently been challenged, suggesting that the ‘land balance’ may be 50% or less of that in the official estimates (Young, 2000). Moreover, a large proportion of the ‘land balance’ is under forest, for example in Brazil, Congo Democratic Republic (formerly Zaire), Indonesia, Peru and Venezuela, clearance of which is strongly opposed for reasons of environment and biodiversity.
This focuses the associated research agenda on the challenge of improving the efficiency with which existing land and water resources are used. Over the past half-century, great progress has been achieved in this respect. In agriculture, the advances generally referred to as the green revolution; in forestry, it has been brought about through a variety of improvements in forest management systems, including fast-growing high-yielding plantations, and by means of genetic improvement. In the early stages of the green revolution, research was directed mainly at plant breeding, fertiliser use and plant protection. However, the pace of advances by these means is slowing; the annual increase in cereal yields in developing countries, which over 1967-82 was 2.9%, has fallen to close to 1%.1
As a consequence, more attention has recently been directed at greater efficiency in the use of land and water resources, for example through nutrient recycling and water conservation. A further powerful incentive in this direction has come from considerations of sustainability. Applied to land and water, sustainability means meeting the production needs of present land users whilst conserving for future generations the resources on which that production depends.
Agroforestry offers one promising option for efficient and sustainable use of land and water. In simplified terms, agroforestry means combining the management of trees with productive agricultural activities. Agroforestry has been practised by farmers for centuries, but its recognition as a scientific discipline, and hence a basis for research, dates from 1977.
The term ‘forest conversion’ is sometimes used as a euphemism for forest clearance. However, agroforestry provides opportunities for forest conversion in the true sense of the term, replacement of natural forest with other tree-based land use systems. There are also opportunities to use agroforestry for the prevention or reversal of land degradation in the humid tropics (Cooper et al., 1996).
There are a wide range of potential benefits which agroforestry systems can achieve ranging from diversification of production to improved natural resource utilisation. The key benefits in term of natural resource use are:
A recent review by Wallace et al (2002) has described the above benefits while this paper will focus on water utilisation.
Can agroforestry increase efficient use of water?
Successful plant mixtures appear to be those which make ‘better' use of resources, by using more of a resource, by using it more efficiently, or both. In terms of the water use of an agroforestry system a central question is, therefore, does intercropping woody and non-woody plants increase total harvestable produce by making more effective use of rainfall? It is possible, at least theoretically, that a mixture of trees and crops may improve the overall rainfall use efficiency either directly, by more rain being used as transpiration, or indirectly by increasing the water use ratio (or transpiration efficiency), i.e more dry matter produced per unit of water transpired. This theoretical possibility requires systematic study of the water balance of agroforestry systems such as that carried out by Ong. et al. (2000) for an agroforestry system in a sub humid part of Kenya.
The complexity of the water balance of an agroforestry system is shown schematically in Figure 1. Some of the gross rainfall, Pg, is intercepted by the tree and crop canopies, giving rise to interception losses from the trees, It, and crop, Ic . The input of rainfall to the ground beneath the trees, Pt, can be different from that beneath the crop, Pc. Pt, and Pc include both throughfall and stemflow. However, in widely spaced trees stemflow is usually a very small fraction of Pg , for example, ~ 0.7 % in the grevillea/maize agroforestry system studied by Jackson (2000). In crops such as maize, with much higher planting densities, stemflow can be between 2 and 4% of Pg (van Dijk and Bruijnzeel, 2001b). Rainfall reaching the ground may infiltrate at different rates below the trees (Ft) and crop (Fc), producing different rates of surface runoff, Rt and Rc respectively. In some circumstances Ft may be sufficiently high not only to absorb Pt , but also any runoff (Rc) from the cropped area. In this case the total runoff from the entire plot would be negligible, as observed in agroforestry studies of runoff on sloping land at Machakos, Kenya where runoff was < 2% of annual rainfall (Kiepe and Rao, 1994).
Water will evaporate directly from the soil surface at potentially different rates Et and Ec and the trees and crop may also have different transpiration rates Tt and Tc. The water contents of the soil zones beneath the trees, t, and the crop, c, is therefore likely to be different, due to the different surface inputs, soil evaporation and transpiration rates. This in turn may lead to different drainage rates, Dt and Dc. Where the agroforestry system is grown on a hillslope, there may also be some lateral sub-surface water movement, Rs, particularly in high rainfall areas where the soil may saturate for significant amounts of time or where perched water tables occur above soil horizons of low permeability. The total amount of water transpired by the tree and crop mixture is therefore,
In a monoculture (tree or crop) on the same area of land, the equivalent total transpiration is
where the superscript m indicates that these terms can be different in the absence of a companion species. The hypothesis that agroforestry systems can use rainfall more effectively can therefore be expressed mathematically as
This defines the way in which the different water balance components need to be managed in order to benefit from the mixture of trees and crops. For example, soil evaporation, runoff and drainage should be minimised in the agroforestry system and this could be achieved via the utilisation of as much as possible of the locally ‘non-productive' components of the monoculture water balance (e.g. Emc, Rmc and Dmc). Where agroforestry is used in the upper reaches of a catchment, the effects of reducing runoff and drainage on the water supply ‘downstream' may need to be taken into account. In the following section we will look in more detail at how agroforestry systems can change each of the water balance components.
The interception process in agroforestry systems differs from that in other forests in two main ways. Firstly, many agroforestry systems tend to have relatively sparse tree densities and secondly, additional complexity is introduced by the crop component of the system with its rapidly varying canopy cover. Van Dijk and Bruijnzeel (2001a) have developed an interception model that deals explicitly with variable crop cover. The sparse nature of the tree component of agroforests affects two key factors that influence the interception, i.e. the amount of water stored on the tree canopy and the rate of evaporation from the tree canopy. For a given species, the storage of water on an individual tree is broadly independent of tree density (Teklehaimanot and Jarvis. 1991, Gash et al. 1995). However, it is the storage of water per unit area of ground which affects interception and this is directly related to canopy cover (Gash et al. 1995) and tends to zero as cover decreases. The rate of evaporation from the wet canopy is primarily determined by the weather conditions during rainfall, but the tree density can affect this rate with more efficient exchange in sparser more well ventilated canopies (Teklehaimanot and Jarvis, 1991).
Figure 2 shows calculations of the ratio of annual interception loss to annual rainfall (the interception ratio) for the trees in a Grevillea robusta agroforestry system in Kenya, made using the sparse canopy interception model described by Gash et al. (1995). The model assumes a canopy storage per unit area of cover of 0.8 mm, a mean evaporation rate during rainfall of 0.2 mm h-1and a mean rainfall rate of 2.3 mm h-1, consistent with the values estimated for this site by Jackson (2000). Figure 4 shows that with complete canopy cover the annual interception loss is ~ 20% of rainfall and this fraction decreases slightly as rainfall increases. This interception loss is intermediate between values reported for (dense) forests in temperate climates (e.g. ~ 40%, see Calder and Newson, 1979) and tropical climates (e.g. ~10%, see Lloyd et al., 1988 and Shuttleworth, 1988). This is a consequence of the mean rainfall rate at the Kenyan site being higher than that in temperate areas and lower than that in the humid tropics. The tendency for the annual interception fraction to decrease as rainfall increases from ~ 400 to 1000 mm has also been reported in temperate climates by Calder and Newson (1979).
Figure 2 also shows the modelled interception ratios (It /Pg) for sparse canopies, more typical of those in semi-arid agroforestry systems (i.e. 10 to 50% cover). In this case the annual interception loss is between 3 and 10% of rainfall. The 10% interception loss for a cover of 50% is similar to the observations of interception made in the same agroforestry system by Jackson (2000). Other sparse forest stands also have interception losses around 10% (e.g. Valente et al. 1997). Higher interception losses have been reported in the much denser multi-storey agroforestry systems in Costa Rica (>30%), by Imbach et al (1989), where the rainfall is higher and more intense than at the Kenyan site described here. Furthermore, high interception losses have also been reported for montane forests in humid tropical regions, e.g. ~ 35% by Hopkins (1960) and as much as 50% by Schellekens et al (1999). The main reason put forward for these high forest interception losses in humid regions is the advection of energy from nearby oceans. Further information and discussion of interception by crops and crop mixtures is given by van Dijk and Bruijnzeel (2001 a,b).
Significant quantities of water can be lost as evaporation from the soil surface, particularly in tropical regions with frequent rainfall, high radiation and sparse ground cover. In agroforestry systems the presence of a tree canopy decreases the radiation intensity at the ground thereby reducing soil evaporation. This is because soil evaporation theory and several field studies of soil evaporation suggest that total soil evaporation is determined (at least in part) by the radiant energy reaching the soil surface. This ‘first phase’ of soil evaporation could therefore be reduced by canopy shade. However, after the first phase is over, soil evaporation rates are determined by the soil hydraulic properties and should therefore be independent of shade. Total evaporative loss from the soil is the sum of losses in the energy limited first phase and the hydraulically limited second phase. The net effect of shade on cumulative soil evaporation over periods of several weeks or more will therefore depend on the total amount of time the soil spends in first and second stage drying. This will be a function of soil type and the frequency with which the surface is re-wetted by rainfall.
The effect of shade on soil evaporation has been studied in an agroforestry system in Kenya by Jackson and Wallace (1999) and Wallace et al. (1999). Direct measurements of soil evaporation made using mini-lysimeters showed large reduction, up to 30%, in soil evaporation due to the presence of the tree canopy. The data obtained were used to calibrate a simple soil evaporation model, which was used to calculate shaded and bare soil evaporation over a period of 18 months, Figure 3. During this period soil evaporation was reduced from 59% of rainfall in completely bare soil to 41% of rainfall for soil directly beneath the tree canopy. The mean annual reduction in soil evaporation due to full canopy shade, 157 mm or 21% of rainfall, was therefore very significant. The reduction in soil evaporation is smaller in sparser tree canopies, 15% of rainfall when cover is ~ 0.5 and 6% of rainfall when cover is ~ 0.2 (Wallace et al. 1999). This analysis is for the tree component of the agroforestry system, however, during the part of the season when the crop canopy is present, there will be additional shading of the soil which may further decrease soil evaporation.
Clearly the reductions in soil evaporation produced by tree canopy shade can help offset the losses of water associated with the tree canopy interception. This is illustrated in Figure 6, where the annual saving in soil evaporation and the annual interception loss are plotted as a function of annual rainfall. The rainfall data used in this figure are from two sites in Kenya (Machakos and Kimakia) and both soil evaporation and interception were calculated assuming a tree cover of 10%, 50% and 100%. This analysis indicates that when the annual rainfall is low the saving in soil evaporation due to canopy shade may be greater than the interception loss. However, once rainfall exceeds ~ 700 mm per annum, the reverse is true with interception losses exceeding saving in soil evaporation. The exact point at which the two effects cross over will depend mainly on rainfall intensity and soil type.
When rainfall reaches the soil surface some of it will normally infiltrate into the soil. If the rainfall rate is greater than the infiltration rate the excess water starts to collect at the surface and when the surface storage is exceeded, runoff will occur. Infiltration is therefore a dynamic process which changes during the course of a rainstorm depending on the soil characteristics, slope of the land and the rainfall intensity. Where the intercropping of woody and non-woody plants alters any of these factors, then the infiltration and runoff may be affected (Kiepe 1995a).
Soil characteristics that affect infiltration are surface crusting, surface storage, saturated hydraulic conductivity and the presence or absence of plant residues. Vegetation cover generally increases infiltration and reduces runoff by altering one or more of these factors. For example, in Senegal runoff decreased from 456 mm in bare soil to 264 mm in cultivated land and further to 200 mm in fallow land containing a mixture of shrubs and herbs (Lal, 1991). Vegetation cover can affect surface infiltration via the canopy modification of the rainfall kinetic energy which may alter soil particle detachment and crust formation. Another effect is via a reduction in surface crusting and improved soil hydraulic conductivity as a result of the incorporation of plant residues into the soil (Kiepe and Rao, 1994). This effect of plant residues was found to be dominant in contour hedgerow systems (see section 2). Mulching is widely used in the tropics for conserving soil water and reducing soil erosion (e.g. see Stigter 1984) and the distribution of plant residues in one form or another is usually a major feature of many agroforestry systems. An additional beneficial effect of mulching is through increased activity of soil fauna, further improving the soil structure and water holding capacity (Lavelle et al., 1992).
There are a number of agroforestry practices which are designed to conserve water and to reduce runoff by the direct effect trees can have on soil slope. Planting of trees or hedgerows on the contours of sloping land can have the effect of forming natural terraces as water and soil are collected on the up-slope side of the hedgerow. The barrier effect of the hedgerow not only reduces soil loss but also runoff, commonly to the order of one third of its value without hedges. Measurements by drip infiltrometer at Machakos, Kenya, showed that on a lixisol (alfisol) with a 14% slope, rates of infiltration were measured as 69 mm h-1 under hedgerows, compared with 8-11 mm h-1 under the cropped alleys (Kiepe, 1995a, 1995b). This increased infiltration rate also reduced runoff in these contour hedgerow systems.
Drainage is the component of the water balance which is most difficult to measure directly. Most deductions about drainage are therefore made from observations of soil water content. This is illustrated in Figure 7 which shows soil moisture content data below a Grevillea/maize agroforestry system in Kenya and compares this with equivalent data from a sole maize crop. In December 1994 the soil moisture deficit over the entire depth of the soil profile on this site (ca 1.6 m) was low and similar in both the crop and agroforestry plots. As the season progressed the soil moisture deficit under the agroforestry system developed more rapidly than in the sole crop and by April 1995 was ~ 100 mm greater in the tree/crop mixture. Figure 4(b) and (c) shows how soil moisture content varied throughout the soil profile towards the beginning and end of this time series. These observations show that the soil water contents decreased by a much greater amount in the tree/crop combination, especially at depth. Further details of these data are given by Jackson et al (2000) and they have concluded that drainage from the tree/crop mixture was much less than in the sole crop.
Another way in which trees can affect soil moisture is via the possibility of ‘hydraulic lift’, in which water taken up by plant roots from moist zones of soil is transported through the root system and released into drier soil (Dawson, 1993). Rainfall captured through stem flow, especially by a woody canopy, can be stored deep in the soil for later use when it is returned to the topsoil beneath the canopy by hydraulic lift. Recently, the opposite of hydraulic lift has been reported in Machakos and elsewhere, in which water is taken from the topsoil and transported by roots into the subsoil (Burgess et al., 1998; Smith et al., 1999a). This mechanism, termed ‘downward siphoning’ by Smith et al., would lead to the opposite effect of hydraulic lift and would enhance the competitiveness of deep-rooted trees and shrubs.
The likely effect on each of the water balance components of the combination of trees with a crop compared to growing the crop alone are summarised in Table 1.
Table 1. The change in water balance component between an agroforestry system with 50% tree cover and a monocrop.
Interception losses are around 10% in semi-arid areas, but can be between 10 and 50% in humid tropical climates, depending on whether the location is continental, montane or coastal. This loss will be completely compensated for by a decrease in soil evaporation in a semi-arid climate, but only partially in a humid tropical climate. Runoff, soil moisture and drainage are all likely to decrease in an agroforest in either climatic regime, with the amount varying according to soil type, slope and species. The extra canopy and the ability of tree roots to exploit water at depth in the soil will lead to a general increase in transpiration in the agroforestry system.
The water use efficiency of any crop or tree/crop mixture can be improved by increasing the water use ratio, ew (i.e. the amount of carbon fixed per unit of water transpired). This is inversely proportional to the mean saturation deficit of the atmosphere, d (Monteith, 1986),
where k is a physiological characteristic specific to a given species. Total dry matter production (W, per unit area in a given time) is simply the product of Tc or Tt and ew, where Tc and Tt are the crop and tree transpiration respectively. Theoretical considerations and experimental studies have shown that (at least under fairly idealised conditions) the product ew d is quite conservative among species groups ( Ong et al., 1996 ). For example, in C3 species (e.g. rice, beans and trees) ew d is ~ 4 kg mm-1 kPa and about twice this (8 kg mm-1 kPa ) in C4 species such as maize . The net effect of atmospheric humidity on any given species is therefore one of the most important factors affecting productivity, since dry matter production per unit of water transpired decreases by a factor of two as saturation deficit increases from ~ 2 kPa in moist temperate climates to ~ 4 kPa in semi-arid areas (Squire, 1990). For example, experiments in India under similar mean saturation deficits (2.0-2.5 kPa) provided season-long values of 3.9 and 4.6 g kg-1 for millet, compared to 1.5-2.0 g kg -1 for groundnut (Ong et al., 1987). However, ew is not always higher in C4 species, since similar values have been reported for drought tolerant C3 species such as cowpea and cotton and relatively drought-sensitive cultivars of the C4 species, sorghum and maize.
Equation (5) shows that there are two ways that overall production could be increased by increasing ew. The first is by increasing k, the physiological characteristic which depends on the biochemistry controlling the photosynthetic processes in plant cells. This may be achieved by plant selection (e.g C3 or C4 species), or by breeding or genetically engineering crops with a higher value of k. The second way to increase ew is to reduce d, either by manipulating the micro-climate, or growing plants in a more suitable macro-climate. This means that agroforests growing in humid tropical regions, where the air is more humid (i.e. low d), will have higher water use ratios.
In theory, the potential of agroforestry to improve ew is limited compared to intercropping, as the understorey crops are usually C4 species and the overstorey trees are invariably C3 species in agroforestry. Improvement in ew is most likely if the understorey crop is a C3 species, which are usually light saturated in the open, so partial shade may have little effect their assimilation. However, the shade will reduce transpiration with the result that ew increases. This may explain why cotton yield in the Sahel is not reduced by the heavy shading of karite (Vitellaria paradoxa) and nere (Parkia biglobosa) in parklands, while yields of millet and sorghum were reduced by 60% under the same trees (Kater et al., 1992). The same reason may explain the observation in the South and Central American savannas that C3 grasses are found only under trees and never grow in open grassland dominated by C4 grasses.
There is also the potential for micro-climate modification in agroforestry systems due to the presence of an elevated tree canopy. This may alter not only the radiation, but also the humidity and temperature around an understorey crop. Some evidence for this has been found where crops have been grown using trees as shelterbelts, and decreases in d have been reported for several crops (Brenner, 1996). Data from an agroforestry trial in Kenya also show that the air around a maize crop growing beneath a Grevillea robusta stand is more humid than the free atmosphere above the trees (Wallace et al., 1995).
In Kenya, Belsky and colleagues (Belsky and Amundson ,1997; Rhoades , 1997). observed improved microclimate due to the presence of trees, along with higher soil biotic activity and N mineralisation, higher infiltration rate and greater beneficial effects in more xeric environments. Plants grown in the open sites were more nutrient-limited than under the tree canopy but artificial shade generated smaller increases in understorey vegetation. However, Belsky was unable to conclude decisively whether microclimate changes or nutrient enrichment were more important in increasing understorey productivity. It is significant to note that the positive tree effects on understorey vegetation are limited to certain sites and species combinations, including both nitrogen and non-nitrogen fixing trees. It is also difficult to determine precisely whether the tree-grass interactions in tropical savannas are typical of the competitive category (Figure 5) as the literature evidence is primarily based on plot level analysis. If the improvement in soil fertility is due to redistribution in the landscape, this would be an example of the neutral category in Figure 5.
Wallace and Verhoef (2000) have developed a multi-species interaction model (ERIN) that can be used to quantify the effect of tree cover on the water use ratio (carbon fixed per unit of water transpired) of an understorey crop. Since fluxes of heat and water vapour from the understorey crop (and soil) can also affect the micro-climate around the overstorey trees, this may alter the water use ratio of the trees. Figure 6 shows the results of an ERIN model simulation of both these effects in an agroforesty system with tall (4m) C3 trees over a 1m high C3 crop. The model predicts that both the crop and tree water use ratios will increase by ~ 25% as the tree cover increases from 0 to 1. The total system water use ratio includes the evaporation from the soil, so it is lower than the water use ratios of the component species. However, this also increases by ~ 25% in this simulation. Clearly, this improved micro-climate is only of benefit to the crop as long as there is adequate light for crop growth and water in the soil to meet both the tree and crop requirements. This highlights the need to identify the tree/crop mixtures and soil and climate combinations within which this may be the case.
Evidence from a series of shade cloth trials on maize and beans at Machakos confirmed the small but beneficial effects of shading on crop temperature and crop production when rainfall is inadequate for crop production (Ong et al 2000) but, unlike the savanna situations, the crops failed because below ground competition consistently outweighed the benefit of shade. In contrast, Rhoades (1997) reported increased soil water (4 to 53% greater than in the open) in the crop root zone beneath Faidherbia albida canopies in Malawi. In theory, trees can increase soil water content underneath their canopies if the water ‘saved’ by their shade effect on reducing soil evaporation and rainfall redistribution e.g. funnelling of intercepted rainfall as stem flow, exceeds that removed by the root systems beneath tree canopies (Ong and Leakey 1999). At high tree densities, the proportion of rainfall ‘lost’ as interception by tree canopies and used for tree transpiration would exceed that ‘saved’ by shading and stem flow, resulting in drier soil below the tree canopy. Van Noordwijk and Ong (1999) expressed this as the amount of water used per unit of shade. This may be one of the most important factors for the observed difference between savanna and alley cropping systems.
Can agroforestry mimic the ecological functions of natural ecosystems?
It is often assumed that appropriate agroforestry systems can provide the environmental functions needed to ensure sustainability and maintain micro-climatic and other favorable influences, and that such benefits may outweigh their complexity (Sanchez, 1995). Second, it is also assumed that agroforestry might be a practical way to mimic the structure and function of natural ecosystems, since components of the latter result from natural selection towards sustainability and the ability to adjust to perturbations (Van Noordwijk and Ong, 1999). It is this opportunity for agroforestry to mimic the interactions between trees and other plants in natural ecosystems that led to the recent redefinition of agroforestry, in which different agroforestry practices are viewed as stages in the development of an agro-ecological succession akin to the dynamics of natural ecosystems (Ong and Leakey, 1999). Recent reviews of agroforestry findings have, however, highlighted several unexpected but substantial differences between intensive agroforestry systems and their natural counterparts, which would limit their adoption for solving some of the critical land use problems in the tropics (Rhoades, 1997., Ong and Leakey, 1999., van Noordwijk and Ong, 1999). The most intractable problems for agroforestry appear to in the semi-arid tropics. In this section, we describe recent insights into the physiological mechanisms between trees and crops in agroforestry systems and how they might be employed to reduce the tradeoffs between environmental functions and crop productivity i.e. retain the positive effects of trees observed in natural ecosystems.
The principles of resource capture have been used to examine the influence of agroforestry on ecosystem function i.e. the capture of light, water and nutrients (Ong and Black, 1994) and to better understand the ecological basis of sustainability of tropical forests. The concept of complementary resource use is not new in ecological studies and it was proposed by de Wit (1960) and others that mixtures of species may have greater capacity to exploit growth resources and hence be more productive than monocultures ( Sinoquet and Cruz, 1995). However, the recent re-interpretations of published results indicate that increased yield of combinations of annual crops was not always associated with greater resource capture or utilisation. Nevertheless, current ideas on agroforestry interactions continue to be rooted in complementary resource use concept. For example, Cannell et al. (1996) proposed that successful agroforestry systems depend on trees capturing resources that crops cannot. The capture of growth resources by trees and crops can be grouped into three broad categories to show competitive, neutral or complementary interactions, Figure 8. In the neutral or trade-off category, trees and crops exploit the same pool of resources so that increases in capture by one species result in a proportional decrease in capture by the associated species. If trees were able to tap resources unavailable to crops, then the overall capture would be increased as shown by the convex curve i.e. complementary use of resources. In the third category, negative interactions between the associated species could result in serious reduction in the ability of one or both species to capture growth resources (concave curve). It is important to bear in mind that tree-crop interactions may change from one category to another depending on the age, size and population of the dominant species as well as the supply and accessibility of the limiting growth resources.
Such ideas on capture of deep water and nutrients coupled with recent innovations in instrumentation (mini-rhizotrons, sap flow gauges) have stimulated a resurgence in root research (Van Noordwijk and Purnomosidhi, 1995; Khan and Ong, 1996 ) and increased attention on spatial complementarity in rooting distribution and the potential beneficial effects of deep rooting. Agroforestry is also considered as critical for maintaining ecosystem functioning in parts of Australia, where deep-rooted perennial vegetation have been removed and replaced by annual crops and pastures, leading to a profound change in the pattern of energy capture by vegetation, rising water-tables, and associated salinity (Lefroy and Stirzaker, 1999). The Australian example showed that compared to the natural ecosystem it replaced the agricultural system is 'leaky' in terms of resource capture. Recent investigations in West Africa suggest that a similar magnitude of ‘leakiness’ is possible when native bush vegetation or woodland, which provide little runoff or groundwater recharge (Culf et al., 1993), is converted into millet fields. The expectation is that agroforestry systems will be able to reduce this leakiness because of its extensive tree root systems. Earlier research on South African savannas has shown that tree roots extend into the open grassland, providing a ‘safety net’ for recycling water and nutrients and accounting for 60% of the total below- ground biomass (Huntley and Walker, 1982).
One of the earliest detailed studies of resource capture in agroforestry systems was that described by Monteith et al. (1991) in semi-arid India (Hyderabad) for a C4 crop, millet (Pennisetum americanum) and C3 tree, Leucaena leucocephala. Total intercepted radiation during the rainy season was 40% greater in the alley crop than in sole millet, primarily because the presence of leucaena increased fractional interception during the early stages of the growing season. The sole leucaena and alley leuceana intercepted twice as much radiation again during the following long dry season when cropping was not possible. The evidence from this study shows that the main advantage of alley cropping was in extending the growing period into the dry season and increasing the annual light interception. However, interception by the more efficient C4 crop was reduced to only half that of the sole millet. This system falls within the lower end of the complementary curve (Figure 5).
The alley crop produced 7 t ha-1 biomass compared to 4.7 t ha-1 of sole millet despite the high amount of light interception because of the low photosynthetic rate or conversion coefficient of a C3 species. The conversion coefficient (er) is defined here as the ratio of biomass production to intercepted light per unit area provides a measure of the "efficiency" with which the captured light is used to produce new biomass; the alternative term radiation use efficiency is also commonly used. This and other studies by Ong and co-workers showed that the less efficient C3 overstorey (tree) component dominated the total light interception while the increased er of the understorey (crop) component was insufficient to compensate for the reduced light interception. These results are typical of many alley cropping studies where the tree populations were so high that reduction in crop yield was inevitable since the trees captured most the resources at the expense of the crops. Although crop yields were seriously reduced, these are examples of complementary interactions which are often misinterpreted as competitive as the sole tree controls are not available.
As for light, agroforestry offers substantial scope for spatial and temporal complementarity of water use resulting from improved exploitation of available water. However, the opportunity for significant complementarity is likely to be limited unless the species involved differ appreciably in their rooting patterns or duration. Recent findings for a range of tree species at Machakos, Kenya showed that when rainfall was low (250 mm) maize yield is linearly and negatively related to the amount of water used by the trees. This relationship breaks down when rainfall exceeds 650 mm. This example illustrates that the trees were using water from the same soil profile as the maize i.e. neutral response.
Early studies of spatial complementarity in agroforestry began by examining the rooting architecture of trees and crops grown as pure stands. For example, Jonsson et al. (1988) described the vertical distribution of five tree species at Morogoro, Tanzania, and concluded that the tree root distribution and maize were similar except for Eucalyptus camaldulensis, which had uniform distribution to 1m. Thus, they concluded that there is little prospect of spatial complementarity if these trees and crops were grown in combination. Recent reviews of the rooting systems of agroforestry systems by Gregory (1996) and Ong et al (1999) essentially supported the earlier conclusion of Jonsson et al. (1988).
What is the extent of spatial complementarity in water use when there is such a considerable overlap of the two rooting systems? Results at Machakos, Kenya consistently showed that there was no advantage in water uptake when there was little water recharge below the crop root zone ( Jackson et al., 2000). However, when recharge occurred following heavy rainfall tree roots were still able to exploit more moisture below the rooting zone of the crops, even when there was a complete overlap of the root systems of trees and crops. This is an example of temporal complementarity, which demonstrates that soil water distribution, rather than root distribution, is the controlling variable.
Direct measurement of tree function was facilitated by the availability of robust sap flow gauges, which offer a unique opportunity for quantifying the amount of water extracted from the crop rooting zone, and hence for assessing spatial complementarity. Experiments in which the lateral roots were progressively severed or excavated indicated that three year old trees were capable of extracting up to 80% of their water requirements from beneath the crop rooting zone (Lott et al. 1996, Howard et al., 1997). This demonstrates the potential for taproots to extract large amounts of water from deep in the soil, however, this does not mean that such large amounts of deep water abstraction will occur when the lateral roots are present. As the trees grew larger in Lott and Howard’s experiments, they depleted the soil water and became more and more dependent on current rainfall and severe shoot pruning was necessary to improve infiltration of soil water and below ground complementarity (Jackson et al. 2000).
The lack of spatial complementarity in alley cropping was highlighted by Van Noordwijk and Purnomosidhi (1995), who observed that repeated prunings of trees in alley cropping had the danger of enhancing below-ground competition by promoting the proportion of superficial roots. They imposed three pruning heights (50, 75, 100 cm) on five tree species (Paraserianthes falcataria, Gliricidia sepium, Peltophorum dasyrachis, Senna siamea, and Calliandra calothyrsus) at a sub-humid site at Lampung, Sumatra, Indonesia. Recent measurements of the long term (3 years) effects of pruning on two species, S. spectabilis and G.sepium, at Machakos, Kenya, confirmed the findings in Lampung and showed that rooting depths of both pruned trees and crops were almost identical in the alley cropping treatment. The evidence so far suggests that below ground competition is inevitable in alley cropping systems where water is limiting.
The most remarkable example of temporal complementarity in water use is the unusual phenology of the Sahelian tree, ^ , which retains its leaf shedding habit in the rainy season even when planted in the Deccan plateau of India, where the water-table is too deep for tree roots. One of the few deliberate experiments in which F. albida is compared with a tree with 'conventional' leaf phenology is that reported by Ong et al. (1996) at Hyderabad, India. Comparison of sap flow rates of F.albida and a local Indian tree, Albizia lebbek, show that transpiration of F. albida begins in August when the understorey crops have developed a full canopy. In contrast, A. lebbeck produces a full canopy in May, well before the onset of the rains and sheds its leaves when the F. albida starts to develop its canopy. Crop yield beneath both trees were about the same suggesting that both tree species were utilising water from the same soil profile. This is clearly an example of a neutral category (Figure 5). Thus, phenology on its own is not adequate for complementary use of resources.
Where groundwater is accessible to tree roots there is clear evidence for spatial complementarity. For instance, measurements of stable isotopes of oxygen in plant sap, groundwater and water in the soil profile of windbreaks in the Majjia valley in Niger showed that neem trees, Azadirachta indica, obtained a large portion of their water from the surface layers of the soil only after rain, when water was abundant but during the dry season tree roots extracted groundwater (6 m depth) or deep reserves of soil water (Smith et al., 1997). In contrast, at a site near Niamey, West Africa, where groundwater was at a depth of 35m, they found that both the trees and millet obtained water from the same 2-3 m of the soil throughout the year.
In recent years, the replacement of native forests by fast-growing trees and deforestation on major catchments has created hydrological concerns in several tropical countries ( Calder, 1997). Native forests are popularly believed to be ‘generators’ of rainfall while exotic forests are often regarded as ‘consumers’ of rainfall. The most notable problem is the high demand for water by fast growing exotic trees such as eucalyptus in semiarid areas. Much remains to be learned on how factors such as climate and topography influences water infiltration and stream flow in the landscape .On the other hand, removal of deep rooting native vegetations has resulted in salinity and rising water-table, especially in Australia. Consequently, it is essential to consider the implications of increased water use on the medium and longer-term water budgets.
In Kenya where forested catchments are located on the submontane and montane elevations there is a growing concern that deforestation was associated with the decline in river flows although there is no hard evidence to show that link between deforestation, rainfall and river flow. Nevertheless, there is evidence from elsewhere that montane or cloud forests have a vital role in intercepting moist air and maintaining low flow, which cannot be reproduced by planting fast growing trees such as pines and eucalptyus ( Hopkins 1960, Schellekens et al 1999). More research is clearly needed to determine ways to restore the hydrological functions of such vital catchments.
The understanding of the hydrological, ecological and physiological processes in alley cropping and other simultaneous agroforestry systems has advanced considerably during the last few years. Although much remains to be studied, we conclude that sufficient is now known to make broad recommendations for what types of agroforestry systems are suited to which climatic and soil conditions.
In the humid tropics, agroforestry systems offer opportunities for conversion of forested land to productive use, whilst retaining many of the beneficial effects of a tree cover. Multistrata systems (forest gardens, agroforests) and perennial crop combinations appear to be the most appropriate agroforestry systems for sustainable land use in the humid tropics, including on sloping land; these systems are commonly found acceptable by farmers. Managed tree fallows and biomass transfer systems provide opportunities for the retention of a tree component in land use systems directed primarily at annual cropping. Contour hedgerow systems offer a technically viable alternative for soil conservation; their acceptability to farmers is variable, although no worse than for conventional methods of conservation. Reclamation agroforestry offers opportunities for restoring degraded lands to productive use.
In semi-arid conditions the beneficial effects of savanna and parkland trees on soil properties are linked to trees with a high proportion of woody above ground structures. It would therefore take a long time (20 - 40 years) before the beneficial effects could be realised, since investment in woody structure slows tree growth. Such long time scales are well beyond the planning horizon of many farmers for the relatively small benefit in crop productivity and may well explain why farmers rarely plant these trees, even though they are well aware of tree species which are soil ‘improvers’. Instead, it may be more worthwhile to focus attention on selection of trees that provide more direct and immediate benefits to farmers (rather than selection for soil enrichment), with minimum loss of crop productivity. It is perhaps not surprising that farmers are already beginning to experiment with such systems. For example, in the drylands of eastern Kenya farmers have recently developed an intensive parkland system using a fast-growing indigenous species, Melia volkensii (Meliaceae), which provides high value timber in 5 to 8 years and fodder during the dry season without an apparent loss in crop productivity (Stewart and Bromley, 1994).
There are two main ways of increasing the efficiency with which water is used in agroforestry. The first is to convert more of the rainfall input into transpiration and second is to increase the transpiration water use ratio. The former may be achieved using a range of physical engineering, hydrological and agronomic techniques which reduce soil evaporation, runoff and drainage. The latter can be achieved either by plant selection or climate modification. Significant improvements (~ 10 to 25%) in the transpiration water use ratio appear to be possible in semi-arid climates, whereas there is less scope for this in the humid tropics. Savings in soil evaporation due to canopy shade may compensate significantly for losses due to interception of rainfall by the canopy. In humid climates interception losses will normally exceed soil evaporation savings, however, in semi-arid climates it is possible that savings in soil evaporation may exceed losses due to interception. Agroforestry systems can also ‘convert’ runoff and drainage into transpiration, thereby increasing rainfall use efficiency at the scale of the agroforestry plot. However, if the runoff and/or drainage used previously contributed to significantly to downstream water resources, this needs to be taken into account in an overall assessment of water use efficiency at a catchment scale.
The importance of obtaining more information using a catchment wide approach is underlined by pointing out that current understanding of resource capture by agroforestry systems is based on well-managed small plots, often in research stations, in which about 30-45% of the rainfall is used for transpiration. Such level of rainfall utilisation is rarely achieved in subsistence agriculture or on a watershed scale and there are still ample opportunities for increasing water use by incorporating trees in the landscape. For example, Rockstrom (1997) reported that only 6 to 16% of the total rainfall in a watershed in Niger is utilised by pearl millet for transpiration and the remainder is lost by soil evaporation (40%) or by deep drainage (33 to 40 %). In contrast, plot level studies at Machakos by McIntyre et al (1996) reported a rainfall utilisation of 40-45 % by maize and cowpea for transpiration and the rest was lost as soil evaporation, thereby limiting the opportunity for agroforestry. Thus, future opportunities for simultaneous agroforestry systems should be explored within the landscape as well as on under-utilised niches within and around the farms, such as boundary plantings.
Finally, although there is clearly great potential for agroforestry systems to conserve and improve resource use, it is by no means suggested that agroforestry automatically brings about all of the above benefits. In order to do so, an agroforestry system must be appropriate for the environment (climate, soil, etc.), practicable (within the local and on-farm constraints), economically viable, and acceptable to the farmer. Finally, as with any system of agriculture or forestry, to achieve the potential benefits an agroforestry system needs to be well managed. Provided that these conditions are fulfilled, there is considerable potential for agroforestry to combine production with conservation, and thus to achieve sustainable land use.
This paper is largely drawn from a review presented by Wallace, Young and Ong at the UNESCO meeting on Water, People and Forest, Kuala Lumpur, August 2001, Malaysia.
Figure 1. A schematic representation of the water balance of an agroforestry system on a hillslope. Gross precipitation Pg is intercepted by the tree and crop canopies, giving rise to interception losses from the trees, It, and crop Ic. Rainfall input to the ground beneath the trees, Pt, mat be different from that beneath the crop, Pc. Infiltration rates below the trees, Ft, and crop Fc, may produce different rates of surface runoff, Rt and Rc. Evaporation from the soil surface proceeds at rates Et and Ec beneath the trees and crop respectively. The water contents of the soil zones beneath the trees, t, and the crop, c, due to the different surface inputs and transpiration rates, Tt and Tc, may lead to different drainage rates, Dt and Dc. There may also be some lateral sub-surface water movement, Rs. ( from Wallace et al. 2001)
Figure 2. Estimates of the annual fraction of rainfall lost as interception (It /Pg) made using the Gash et al. (1995) sparse forest model with rainfall data for 1984 to 1988 from Machakos, Kenya. Different degrees of cover are input to the model to simulate dense (100%, ), intermediate (50%, ) and sparse (10%, ) canopies.( from Wallace et al 2001)
Figure 3. Annual saving in soil evaporation (Es) at Machakos () and Kimakia () in Kenya compared to annual interception loss (open symbols). Calculations have been made at 100% cover ( , ), 50% cover , (- - - - - , ) and 10% cover (_ . _ . _ , O ).( from Wallace et al 2001)
Figure 4. (a) Changes in total profile soil moisture deficit under maize () and maize plus ^ trees () at Machakos, Kenya. The water contents changes with depth in this profile are also given for the beginning (b) and end (c) of the period shown. ( from Wallace et al 2001)
Figure 5. Resource capture by tree and crop showing complementarity, competitive and neutral interactions, (1) parkland or savanna, (2) boundary planting, and (3) alley cropping ( from Ong and Leakey,1999)
Figure 6. Variation of water use ratio with tree fractional ground cover in an agroforestry system with a dominant tree species. (from Wallace and Verhoef, 2000)
Belsky, A. J. & Amundson, R. G. (1997) Influence of savannah trees and shrubs on understorey grasses and soils: new directions in research. In: Carbon and Nutrient Dynamics in Natural and Agricultural Tropical Ecosystems (Eds. Bergstrom, L. and Kirchmann, H.) pp 153-171, CAB International, Wallingford, UK.
Brenner, A. J., 1996. Micro-climatic modifications in Agroforestry. In: Tree-crop interactions - a physiological approach (Eds. Ong, C. K., & Huxley, P. A.) CAB International, Wallingford, UK, pp 159-187.
Burgess, S. S. O., Adams, M. A., Turner, N. C. and Ong, C. K. (1998) Redistribution of water within plant root systems. Oceologia 115, 306-311.
Calder, I. R. & Newson, M. D. 1979. Land-use and upland water resources in Britain-a strategic look. Water Resources Bulletin, 15, 1628-1639.
Cannell, M. G. R., Van Noordwijk, M. and Ong, C. K. (1996) The central agroforestry hypothesis: the trees must acquire resources that the crop would not otherwise acquire. Agroforestry Systems 34, 27-31.
Cooper, P. J. M., Leakey, R. R. B., Rao, M. R., and Reynolds, L. (1996) Agroforestry and the mitigation of land degradation in the humid and sub-humid tropics of Africa. ^ 32, 235-290.
Culf, A. D., Allen, S. J., Gash, J. H. C., Lloyd, C. R. and Wallace, J. S. (1993) Energy and water budgets of an area of patterned woodland in the Sahel. Agricultural and Forest Meteorology 66, 65-80
de Wit, C. (1960) On competition^ en 66, 1-82
Gash, J. H. C., Lloyd, C. R. and Lachaud, G. (1995) Estimating sparse forest rainfall interception with an analytical model. Journal of Hydrology, 170, 79-86.
Gregory, P. J. (1996) Approaches to modelling the uptake of water and nutrients in agroforestry systems. ^ 34, 51-65
Hopkins, B. (1960) Rainfall interception by a tropical rainforest in Uganda. East African Agricultural Journal 25, 255-258.
Howard, S. B., Ong, C. K., Black, C. R. and Khan, A. A. H. (1997) Using sap flow gauges to quantify water uptake by tree roots from beneath the crop rooting zone in agroforestry systems. Agroforestry Systems 35, 15-29
Huntley, B. J. & Walker, B. H. (ed) (1982) Ecology of Tropical Savannas. Springer Verlag, Berlin, pp 669.
Jackson, N. A. & Wallace, J. S. (1999) Soil evaporation measurements in an agroforestry system in Kenya. Agricultural and Forest Meteorology 94, 203-215.
Jackson, N. A. (2000) Measured and modelled rainfall interception loss from an agroforestry system in Kenya. Agricultural and Forest Meteorology 100, 323-336.
Jackson, N. A., Wallace, J. S. and Ong, C. K. (2000) Tree pruning as a means of controlling water use in an agroforestry system in Kenya. Forest Ecology and Management 126, 133-148.
Jonsson, K., Fidgeland, L., Maghembe, J. A. and Hogberg, P. (1988) The vertical root distribution of fine roots of five tree species and maize in Morogoro, Tanzania. Agroforestry Systems 6, 63-69.
Kater, L. J. M., Kante, S. and Budelman, A. (1992) Karite ( Vitellaria paradoxa) and nere (Parkia biglobosa) associated with crops in South Mali. Agroforestry Systems 18, 89-105
Kiepe, P. & Rao, M. R. (1994) Management of agroforestry for the conservation and utilization of land and water resources. Outlook on Agriculture 23, 17-25.
Kiepe, P. (1995a) No runoff, no soil loss: soil and water conservation in hedgerow barrier systems. Tropical Resource Management Papers 10. Wageningen, Netherlands: Agricultural University.
Kiepe, P. (1995b) Effect of Cassia siamea hedgerow barriers on soil physical properties. Geoderma 66, 113-120.
Khan, A. A. H. & Ong, C.K. (1996) A low cost heat pulse method for measuring tree root water uptake. Agroforestry For 7(2):19-22.
Lal, R. (1991) Current research on crop water balance and implications for the future. In: Soil Water Balance in the Sudano-Sahelian Zone (Eds. M. V. K. Sivakumar, J. S. Wallace, C. Renard & C. Giroux) Proc. Int. Workshop, Niamey, Niger, February 1991) IAHS Publ. no. 199, 31-44. IAHS Press. Institute of Hydrology, Wallingford, UK.
Lavelle, P., Spain, A. V., Blanchart, E., Martin, A. and Martin, S. (1992) The impact of soil fauna on the properties of soils in the humid tropics. In: Myths and Science of Soils of the Tropics. Soil Science Society of America Special Publication 29, 157-185.
Lefroy, E. C. & Stirzaker, R. J. (1999) Agroforestry for water management in southern Australia. Agroforestry Systems 45, 277-302.
Lloyd, C. R., Gash, J. H. C., Shuttleworth, W. J. and Marques, F. A.de O. (1988) The measurement and modelling of rainfall interception by Amazonian rain forest. Agricultural and Forest Meteorology 43, 277-294.
Lott, J. E., Khan, A. A. H., Ong, C. K. and Black, C. K. (1996) Sap flow measurements of lateral roots in agroforestry systems. Tree Physiol 16, 995-1001.
McIntyre, B. D., Riha, S. J. and Ong, C. K. (1996) Light interception and evapotranspiration in hedgerow agroforestry systems. Agricultural and Forest Meteorology 81, 31-40.
Monteith, J. L., (1986) How do crops manipulate water supply and demand ? Phil Trans Roy Soc London, B., 316, 245-259.
Monteith, J. L., Ong, C. K. and Corlett, J. E. (1991) Microclimatic interactions in agroforestry systems. For Ecol Manage 45, 31-44
Ong, C. K., Simmonds, L. P. and Matthews, R. B. (1987) Responses to saturation deficit in a stand of groundnut (Arachis hypogaea L.) 2. Growth and development. Ann Bot 59, 121-128.
Ong, C. K. & Black, C. R. (1994) Complementarity in resource use in intercropping and agroforestry systems. In: ^ (Eds. Monteith J. L., Scott R. K. and Unsworth M. H.) Nottingham University Press, pp 255-278 Loughborough, U.K.
Ong, C. K., Black, C. R., Marshall, F. M., and Corlett, J. E., (1996) Principles of resources capture and utilisation of light and water. In: ^ (Eds. Ong, C.K., & Huxley, P.A.) CAB International, Wallingford, UK, pp. 73-158.
Ong, C. K. & Leakey, R. R. B. (1999) Why tree crop interactions in agroforestry appear at odds with tree-grass interactions in tropical savannahs. Agroforestry Systems 45, 109-129.
Ong, C. K., Deans, J. D.Wilson, J., Mutua, J., Khan, A. A. H. and Lawson, E. M. (1999) Exploring below ground complementarity in agroforestry using sap flow and root fractal techniques. Agroforestry Systems 44, 87-103.
Ong, C. K., Black, C. R., Wallace, J. S., Khan, A. A. H., Lott, J. E., Jackson, N. A., Howard, S. B. and Smith, D. M. (2000) Productivity, microclimate and water use in Grevillea robusta based agroforestry systems on hillslopes in semi-arid Kenya. Agriculture, Ecosystems and Environment 80, 121-141.
Rhoades, C. C. (1997) Single-tree influences on soil properties in agroforestry: lessons from natural forest and savanna ecosystems. Agroforestry Systems 35, 71-94.
Rockstrom, J. (1997) On –farm agrohydrological analysis of the Sahelian yield crisis: rainfall partitioning, soil nutrients and water use efficiency of pearl millet. Ph.D thesis, University of Stockholm.
Sanchez, P. (1995) Science in agroforestry. Agroforestry Systems 30, 5-55.
Sinoquet, H. & Cruz, P. (1995) Ecophysiology of Tropical Intercropping. INRA. Paris. 483 pp.
Schellekens, J., Scatena, F. N., Bruijnzeel, L. A. and Wickel, A. J. (1999) Modelling rainfall interception by a lowland tropical rain forest in northeastern Puerto Rico. Journal of Hydrology 225, 168-184.
Shuttleworth, W. J. (1988) Evaporation from Amazonian rain forest. Proceedings of the Royal Society of London Series B, 233, 321-346.
Smith, D. M., Jarvis, P. G. and Odongo, J. C. W. (1997) Sources of water used by trees and millet in Sahelian windbreak systems. J Hydrol 198:140-153.
Smith, D. M., Jackson, N. A., Roberts, J. M. and Ong, C. K. (1999a) Reverse flow in tree roots and downward siphoning of water by Grevillea robusta. Funct Ecol 13(2), 256-264.
Smith, D. M., Jackson, N. A., Roberts, J. M. and Ong, C. K., (1999b) Root distributions in a ^ - maize agroforestry system in semi-arid Kenya. Plant and Soil 211, 191-205
Stewart, M. & Bromley, T. (1994) Use of Melia volkensii in a semi-arid agroforestry systems in Kenya. Comm For Rev 73(2), 128-131
Stigter, C. J. (1984) Mulching as a traditional method of microclimate management. ^ 35, 147-154.
Teklehanimanot, Z. & Jarvis, P. G. (1991) Direct measurement of evaporation of intercepted water from forest canopies. Journal of Applied Ecology 28, 603-618.
Valente, F., David, J. S. and Gash, J. H. C. (1977) Modelling interception loss for two sparse eucalypt and pine forests in central Portugal using reformulated Rutter and Gash models. Journal of Hydrology 190, 141-162.
Van Dijk, A. I. J. M. & Bruijnzeel, L. A. (2001) Modelling rainfall interception by vegetation of variable density using an adapted analytical model. Part 1. Model description. Journal of Hydrology 247, 230-238
Van Dijk, A. I. J. M. & Bruijnzeel, L. A. (2001) Modelling rainfall interception by vegetation of variable density using an adapted analytical model. Part 2. Model validation for a tropical upland mixed cropping system. Journal of Hydrology 247, 239-262
Van Noordwijk, M. & Purnomosidhi, P. (1995) Root architecture in relation to tree-soil-crop interactions and shoot pruning in agroforestry. Agroforestry Systems 30, 161-173.
Van Noordwijk, M. & Ong, C. K. (1999) Can the ecosystem mimic hypotheses be applied to farms in African savannahs? ^ 45, 131-158.
Wallace, J. S., Jackson, N. A. and Ong, C. K. (1995) Water balance of agroforestry systems on hillslopes - phase I. Final Report to the forestry Research Programme, UK Department for International Development pp40, Centre for Ecology & Hydrology Wallingford.
Wallace, J. S., Jackson, N. A., and Ong, C. K., (1999) Modelling soil evaporation in an agroforestry system in Kenya. ^ 94, 189-202.
Wallace, J. S., & Verhoef, A., (2000) Modelling interactions in mixed-plant communities: light, water and carbon dioxide. In: Leaf Development and Canopy Growth (Eds. Marshall, B., & Roberts, J. A.) Sheffield Academic Press, Sheffield, pp 204-250.
Young, A. (2000) How much spare land exists? Bulletin of the International Union of Soil Sciences 97, 51-55.
1 Discussion of the potential of genetically modified crops, and associated environmental questions, lies beyond the scope of this paper.