Can the ecosystem mimic hypotheses be applied to farms in African savannahs?

The first ecosystem mimic hypothesis suggests clear advantages if man-made land use systems do not deviate greatly in their resource use patterns from natural ecosystems typical of a given climatic zone. The second hypothesis claims that additional advantages will accrue if agroecosystems also maint ain a substantial part of the diversity of natural systems. We test these h ypotheses for the savannah zone of sub-Saharan Africa, with its low soil fe rtility and variable rainfall. Where annual food crops replace the natural grass understorey of savannah systems, water use will decrease and stream a nd groundwater flow change, unless tree density increases relative to the n atural situation. Increasing tree density, however, will decrease crop yiel ds, unless the trees meet specific criteria. Food crop production in the pa rkland systems may benefit from lower temperatures under tree canopies, but water use by trees providing this shade will prevent crops from benefiting . In old parkland trees that farmers have traditionally retained when openi ng fields for crops, water use per unit shade is less than in most fast gro wing trees introduced for agroforestry trials. Strong competition between p lants adapted to years with different rainfall patterns may stabilise total system productivity - but this will be appreciated by a farmer only if the components are of comparable value. The best precondition for farmers to m aintain diversity in their agroecosystem hinges on the availability of a br oad basket of choices, without clear winners or 'best bets'.