In the last 35 years, the population of sub-Saharan Africa has increased nearly threefold and is expected to reach 681 million by the year 2000, with nearly 50% of the population living in urban centres. Such population pressures, exacerbated by a range of social and political factors, have already resulted in widespread land degradation in areas of high population densities and the expansion of agriculture on to marginal and sloping land. Declining soil fertility and soil erosion are increasingly threatening the sustainability of small scale farming systems throughout Africa, and affordable external nutrient inputs are seldom available to farmers. In addition, shortages of wood for construction and fuel and high-quality dry-season fodder for livestock are widespread and serious constraints to farm productivity.

Agroforestry, the deliberate integration of woody perennials into crop and livestock systems, has the potential to mitigate many of these constraints through both the service and production functions played by trees. In recent decades much agroforestry research has been undertaken in sub-Saharan Africa. In this review we focus specifically on research which addresses the potential of agroforestry systems to enhance soil fertility, prevent soil erosion, provide high-quality dry-season fodder or generate much needed income through the production of high-value goods.

Much emphasis has been placed on a wide range of agroforestry systems for the maintenance of soil fertility and the prevention of soil erosion losses, and encouraging results, both in technical performance and farmer enthusiasm, have occurred. However, it is clear that agroforestry solutions to land degradation are always likely to be location-specific in their relevance, performance and farmer acceptability. It is essential that farmers are included as research partners to determine what is appropriate for their conditions.

Good progress has also been made on identifying fast-growing leguminous trees and shrubs for high-quality livestock fodder supplements. Where livestock enterprises, such as peri-urban milk production, are market-oriented the adoption and impact of such systems have been high. Given population and urbanization projections, it is likely that fodder trees and shrubs will have a major role to play in meeting future feed demands for both milk and meat production. Research on the potential of high-value indigenous and exotic trees to generate income has been less extensive in Africa, although the huge potential of this approach has been clearly demonstrated by farmers in south-east Asia. We suggest that there is a need for increased research emphasis on the domestication of high value indigenous trees, and their integration into more sustainable, diverse and intensive land use systems.

We conclude that, although good progress has been made in agroforestry research in Africa and farmer adoption is occurring, future population projections pose a clear challenge. Agroforestry systems which provide solutions for today's land degradation problems will need to evolve in both diversity and intensity if they are to remain relevant and effective for tomorrow's Africa.

Biophysical and socio-economic variability have several impacts on farmers' cropping strategies in dryland areas of Africa. This study examines data from a communal area in southern Zimbabwe over a seven year period. Variation in the output of different crops, in the pattern of crop mixes, in the diversity of cropping patterns and the level of crop sales is explored. Through contrasts between ecological zones and household wealth status, the analysis helps to explain the range of farm-level risk management strategies employed. Finally, the importance for research and extension approaches of understanding variability, through a differentiated analysis of cropping strategies, is emphasized.

Preparation of small catchment areas on rainfed valley floor soils in highland Balochistan, Pakistan, is a low-cost method of generating run-off and increasing crop yields within the cropped areas. The effect of different proportions of water catchment area to cropped area were investigated by comparing a control treatment with the entire area planted to the crop (traditional rainfed agriculture); a 1:1 treatment, with one half of the area used for water catchment and one half for planting; and a 2:1 treatment, with two thirds of the area used for water catchment and one third for planting. Results from six seasons of trials using wheat (Triticum aestivum) showed that the 1:1 treatment had 23% higher net benefits than the control, with a 19% reduction in the coefficient of variation. The 2:1 treatment had 29% lower net benefits than the control and reduced the variation in net benefits by 8%. By contrast, four seasons of trials using barley (Hordeum vulgare) showed that the l: l treatment yielded 25% lower net benefits than the control but increased by 4% the variation in net benefits. Treatment 2:1 had 36% lower net benefits than the control and 18% more variation. Even though the gross revenue from wheat under the 1:1 treatment was less than that from the control, the reduction in total costs in the 1:1 treatment resulted in larger net benefits than the control. Water-harvesting in the valley floors is not a net yield-increasing technology. Land suitable for cultivation is limited and the increases in yields in the cropped area resulting from water-harvesting are offset by the opportunity costs of the catchment area. However, wheat grown in a 1:1 ratio of cropped to catchment area can increase farmers' income and decrease its variation. For barley, farmers are better off using their traditional management practice than giving up part of their cropped area to create water catchments.

Weeds are generally considered to be undesirable plants as they can cause yield losses but fodder weeds are traded as a valuable feed in Niger, Vest Africa. In vitro analyses of fodder weeds indicated that they are a suitable crude protein supplement to low quality millet stover, a major feed during the dry season. Dry matter production of fodder weeds increased with a crop residue mulch. When weeds are used as feed, farmers may generate additional income through an increased availability of energy for animal traction or for dairy and beef production. Species with the highest crude protein content, usually the most limiting factor in feed for animal production, had the highest market prices. The sale of fodder weeds may provide farmers with cash. Agricultural research programmes should be encouraged to consider the use of fodder weeds to improve both the quality and quantity of feed in the West African Sahel.

Soil erosion and the productivity of cassava (Manihot esculenta Crantz) cropping systems, including forage legume intercropping, were monitored under natural rainfall conditions on Inceptisols in southwest Colombia. Trials were conducted over two cropping periods on moderate slopes. The best erosion control and cassava fresh root yields were achieved by sole cropping on contour ridges. Erosion was greater with forage legume intercropping than with cassava sole cropping in the first cropping period but once well established and uniformly distributed, undersown legumes controlled erosion effectively, though yields of cassava were depressed by 40% or more. Dense stands of grass contour barriers decreased erosion to tolerable levels. Lower cassava yields due to reduced cropping area and competition from the grasses could be compensated for by the production of valuable forage and by increasing cassava planting density.

A series of researcher-managed wheat fertilizer trials was conducted on representative farmers' fields across northwest Syria between 1986 and 1990. Wheat grain and straw yields were strongly correlated with seasonal (October-May) rainfall, almost irrespective of soil fertility, crop sequence or fertilizer rate, with a highly significant response to nitrogen fertilizer which increased with increasing rainfall and decreasing initial soil mineral-nitrogen values. These results were summarized in regression equations which express yield in terms of fertilizer rates, seasonal rainfall and their interactions. The equations with applied nitrogen and seasonal rainfall were the most appropriate for determining fertilizer needs. Economic analysis indicated that all fertilizer treatment rates were profitable under existing price conditions and that fertilizer use would still be beneficial for a nitrogen price up to three times higher than that of the price of grain (weight for weight) with a seasonal rainfall of 250 mm, and up to six times higher with a seasonal rainfall of 450 mm.

Earlier work has shown that, in cereals in general, grain yield (Y) and protein content (P) are strongly negatively related. It is shown here that the expected positive expected relationship between grain yield and protein yield also holds, with (YP) given approximately by YP = 0.141 + 0.0969 Y. All cereal breeding in which high protein content is desired therefore faces a compromise with grain yield.

The yield response to drought and temperature of six contrasting tea clones was studied in a line-source irrigation experiment in Southern Tanzania. The selected clones, all commercially and/or scientifically important in eastern Africa, embrace a range of morphological and physiological types. The bushes were planted in August 1988 and differential drought treatments were imposed for 16 and 13 weeks towards the end of the dry seasons in 1990 and 1991, respectively. The resulting soil water deficits were successfully simulated using a water balance model. Under well-watered conditions Clone S15/10 (from Kenya) gave the highest yield of dry tea, reaching 5600 kg ha−l in the fourth year after planting (1991/92) compared with 3640–4420 kg ha−1 for the other five clones. During the cool season Clone SFS150 (from Malawi) yielded more than Clones 1, 207, 6/8 and K35. Although annual yields decreased curvi-linearly as the maximum soil water deficit increased, single values for the drought sensitivity of each clone could be derived by using stress time as an index of drought. On this basis Clones S15/10 and 207 were identified as being the most sensitive to drought; Clones SFS150 and 1 were drought resistant. The reasons for these differences in yield responses and the importance of determining drought sensitivity over an appropriate time period are discussed.