125 A NEW EVALUATION SCHEME FOR CASSAVA BREEDING AT CIAT Hernan Ceballos, J.C. Pérez, F. Calle, G. Jaramillo, J.I. Lenis, N. Morante and J. López 1 ABSTRACT One of the most important questions to be answered in plant breeding relates to the parents to be used in the generation of new germplasm. This is particularly true for cassava, given the time required to evaluate segregating progenies and the large genetic variation generated with each cross due to the highly heterozygous nature of the crop. In the past, cassava breeding at CIAT focused on properly identifying the best clones, from large, segregating progenies. However, the process was not designed to take advantage of all the potential information that could be generated. Significant steps have recently been taken at CIAT to modify the evaluation scheme, particularly during the first clonal evaluation stage, with the following objectives: a) obtain information that allows an approximation to the general combining ability of progenitors; b) shorten the length of the evaluation process; c) improve the probabilities of identifying superior germplasm; and d) detect new potential traits that can be incorporated into the selection criteria. The modifications have been implemented and improved during the past three years. The new breeding scheme has already produced important benefits. Parents are currently selected based on the quality of progenies they produce. Leaf retention at five month of age and in the absence of biotic or abiotic stresses, has proven to have a large effect on root yield. The evaluation cycle has been shortened by 16 months, and it is expected that the new scheme is more efficient in identifying superior germplasm. INTRODUCTION Cassava germplasm development at CIAT is centered on the development of improved gene pools for specific edapho-climatic zones with importance for cassava production. The most relevant ecosystems are the semi-arid and sub-humid tropics, for which the majority of efforts are devoted. The main selection activity is conducted in sites selected to represent the conditions of the target ecosystem. For each zone, a recurrent selection program with a progressive set of stages is followed. As the stages progress, more emphasis is given to traits of lower heritability, because more planting material for each genotype is available, and the evaluation can be conducted in bigger plots with replications. Certain selection criteria are of general importance across ecosystems (i.e. yield potential, dry matter content), while others are specific for each ecosystem (i.e. specific pests and/or diseases). The traditional evaluation and selection procedure has a few important drawbacks: a) breeding cycles were long; b) no data was taken at the early stages to allow estimates of general combining ability effects of the progenitors employed; and c) it took several steps in the selection process until replicated evaluations could be performed. Cassava has unique opportunities to increase its relevance in tropical agriculture and agro-industries. To take advantage of these opportunities a more dynamic and efficient breeding scheme is required to meet the new demands on this crop. This article describes the modifications introduced into the cassava breeding scheme at CIAT and some of the initial observations made upon their implementation. 1 Cassava Breeding Project, Centro Internacional de Agricultura Tropical (CIAT), Cali, Colombia.