Planning tree species diversification in Kenya based on differences in tree species composition between farms I. Analysis of tree uses. Agroforestry Systems 67

composition Concerns exist about the limited diversity of tree species in agricultural landscapes. Complete tree inven-tories were carried out on 201 farms from four villages in western Kenya to establish whether significant differences in tree species composition existed between farms, and if so their magnitude and implications for new introductions and plantings. Tree species composition was interpreted to encompass elements of both tree identity and abundance. Tree identity was viewed from both taxonomic and function (e.g. fruit, timber, medicine) perspectives. Novel types of ordination using the Hellinger ecological distance and polynomial Redundancy Analysis indicated wide heterogeneity between farms with respect to tree species composition. For the 12 most prevalent functions of trees, the analyses showed significant differences (p<0.05). Par-titioning of variance identified that village location explained much of the differences between farms suggesting that farmers share tree species within villages more than between villages. Differences between farms were assessed on two-dimensional ordination graphs. For five important tree functions, including beverage, charcoal, construction, fodder and medicine, two species dominated the compositional differ-ences. For these functions, diversification can be achieved by village-to-village sharing even in the absence of any new species introductions. A general process to determine the degree of tree diversity at farm and landscape levels and steps to increase it are discussed