The smallholder agriculture sector in East Africa is the dominant economic and social activity for millions of farm households who are often resource-poor, food-insecure and most vulnerable to climate change. In this region, population pressure has led to shorter fallow periods or continuous cropping even on hillslopes causing erosion and leading to reduced soil organic matter content and nutrient mining without replenishment. Consequently, poor agricultural productivity has led to food shortages and these problems are likely to intensify in the region, as the human population is growing faster than in other parts of the world. Agroforestry, a low-input technology, was shown to contribute to the enhancement of food production while ensuring sustainability in sub-Saharan Africa. Agroforestry may improve food security by increasing soil fertility and providing additional income from tree products. Thus, agroforestry is now receiving increasing attention as a sustainable land-management option and some countries in East Africa (e.g. Rwanda) have pledged to restore up to 100% of their agricultural land mainly through agroforestry by the year 2020. Nevertheless, crop yields reduction in agroforestry are frequent due to competition for resources among trees and crops. In recent studies, tree canopy and root pruning were tested to improve light availability and resource use efficiency but studies that tackle crop management and tillage options to optimize crop productivity in the agroforestry systems are scarce.This thesis aims to assess the importance of agroforestry across Rwanda and its implication on crop productivity and food security of farm households, explore and recommend the maize varieties and tillage options that could minimize tree-crop competition in the equatorial savannah of Rwanda and Ethiopia. The approach combined household survey on the contribution of trees on household income and food security in six agroecologies of Rwanda, experiments on the microclimate and fertility effects of trees on crops in sub-humid region of Rwanda, maize variety testing in agroforestry systems and trials on conservation agriculture with trees in the equatorial savannah of two East African countries: Rwanda (Bugesera site) and Ethiopia (Meki site). The survey in Rwanda found that food security increases with increasing farm size and farmers with more trees tended to be wealthier (e.g. with larger land and more often higher crop and livestock income) and therefore tended to be more food secure in half of the agroecologies. The proportion of household income that came from tree products was the least among sources of income suggesting that most tree products are not sold but kept by farmers for their own use. Yet tree income was important for about 12% of the farmers, contributing more than 20% of their overall income. Households having low food security relied more on income from tree products than those with higher food security status. Therefore, income from tree products can be seen as a ‘safety net’ for the poorest households.Experiments in the sub-humid environment of Rwanda assessed the effects of mature Alnus acuminata (Kunth) and Markhamia lutea (Seem.) on maize at different distances from tree trunk for four consecutive seasons. Nutrients availability was higher under A. acuminata compared with M. lutea, because of higher litter fall but maize nutrient uptake increased only under A. acuminata 3 m from tree trunk during a wetter season. None of tree species affected water availability for maize in the topsoil. Total solar radiation, photosynthetically active radiation (PAR), and day air temperature were reduced by both tree species. Whereas crops consistently underperformed in M. lutea system, the competitive effect of A. acuminata for light was to some extent compensated by extra N input in the wetter seasons (2015 A and 2015 B) at 3 m but not at 1 m from the tree trunk. In an APSIM modelled scenario under low N fertilization, larger N input from trees could compensate for yield loss caused by reduction in radiation and temperature in about 60% of the seasons. This study suggested that adequate pruning and high leaf litter recycling can reduce the negative effect of shade in low intensity farming systems. The low competition of A. acuminata with crops was also perceived by Rwandan farmers, who ranked this tree species as the least competing among all the other upper story trees grown on-farm in the highlands. Experiments compared the performance of four maize hybrids and four OPVs was compared in sole crop and under mature Grevillea robusta and Senna spectabilis – in Bugesera, Rwanda or Acacia tortilis – in Meki, Ethiopia. In Bugusera, grain yields of hybrids (2 t ha-1) was significantly better than OPVs (1.5 t ha-1). Further, the presence of trees significantly reduced maize grain yield and total biomass in both hybrids and OPVs in the same manner. However, trees reduced harvest index significantly more in OPVs than in hybrids, suggesting that competition had a greater impact on grain yield of OPVs than on biomass production. In the experiments in Meki, the grain yield of OPVs (2.08 t ha-1) and hybrids (2.04 t ha-1) did not significantly differ and the presence of trees reduced their grain yields in the same manner. We concluded that agroforestry farmers could benefit from growing hybrids in the equatorial savannahs of Rwanda, but not in the equatorial savannahs of Ethiopia. It appears that the relevance of using either hybrids or OPVs in agroforestry systems depends on local conditions and the comparative advantages in seed costs. Experiments in the same regions of Rwanda and Ethiopia were carried out to assess the effect of conservation agriculture with trees (CAWT) on crop productivity as compared to conventional tillage with trees (CTWT) in the equatorial savannah. Crop emergence was significantly reduced under CAWT compared with CTWT. Maize emergence rates in CAWT and CTWT were respectively 46.9% and 70.1%, compared with 74.7% and 79.8% in sole maize under conservation agriculture (CA) and conventional tillage (CT). Grain yield in CAWT and CTWT were respectively 0.37 t dry matter (DM) ha-1 and 1.18 t DM ha-1 as compared with 1.65 t DM ha-1 and 1.95 t DM ha-1 in CA and CT. It was concluded that CAWT likely exacerbates tree-crop competition for water and nutrients and reduce crop yields and was therefore not considered as a viable alternative to CTWT or to CT in the studied systems.Overall, this study found that mixing trees and crops produced a worthwhile, if somewhat reduced, crop yield, and that on-farm trees can provide substantial income for the poorest households of Rwanda.