Validation of a FAO land evaluation method by comparison of observed and predicted yields of five food crops in Burundi

Abstract. A land evaluation using unsophisticated data successfully predicted yield ranges of various crops in Burundi. Yields of wheat, pea, bean, maize and potato predicted from data for climate, soil and land use technology were compared with observed yields from farm trials and from seed station and research station trials. The predicted range of yield for each crop suitability class enclosed the mean farm yields 13 times out of 16 and yields on seed stations and research stations 10 times out of 15 and 14 times out of 21 respectively. The variability of the observed yields exceeded the predicted range of yields. The method is considered as validated. The method appears to be applicable for multi-year studies at a broad scale, but yield variation according to the weather from year to year is not accounted for. The variances of the yields on farms are greater than the variances in seed station and research station trials.
Since the method predicts correctly the mean regional farm yields, it could be useful for land use planning, research into optimal regional cropping specialisation, studies on food policy, and for evaluation of economic return and sustainability of different crops.