A methodology for evaluating a location as a selection site for an international plant breeding program

A method for comparing locations as selection sites based upon their abilities to predict yield and disease reaction over a target region is proposed. The probability of coincidence in selection for a site is defined as the probability for a line selected at the site to be selected at other sites within the region. The probability of divergence in rejection is defined as that associated with regional selection of a line given that is discarded by the site where selection is being conducted. The ideal selection site would maximize the probability of coincidence in selection and minimize the probability of divergence in rejection. The method is illustrated using a set of data from the rice yield nurseries of the International Rice Testing Program for Latin America planted under the rainfed conditions of Central América and México during the period 1978–1984. Five locations were compared for their predictive ability in selecting for the rainfed rice growing region, based on yield and disease reaction. Selection for yield was defined as performance superior to the best check in each location. Selection for disease reaction was based on an index derived from the Standard Evaluation System for Rice for diseases of regional importance. Locations varies 10–15 percent in their selection coincidence with the region for both yield and disease selection criteria applied independently.