The aim of this paper is to apply the REML/BLUP methodology in coffee breeding programs, using it in the parameter estimation and genotypic values prediction in a cultivar trial of Coffea arabica. A trial established in Selvíria-MS by UNESP/Campus de Ilha Solteira, was studied. The trial evaluated 12 cultivars developed by the Instituto Agronômico de Campinas in a complete block design with four replications and 10 plants per plot. The results showed a low genetic variability among cultivars for the traits height, stem diameter and number of primary branches, at age of 26 months. Only the cultivars Catuaí Amarelo, Icatu Vermelho and Catuaí Vermelho showed predicted genotypic values for stem diameter superior to the general mean of the trait. The accuracy of the genotypic evaluation of the cultivars for stem diameter was, in average, 76%. The use of an experimental design with two plants per plot and 20 replications can lead to an accuracy of 90% for stem diameter. The mixed model methodology (REML/BLUP) showed to be adequate for genetic parameter estimation and genotypic values prediction in coffee breeding and then, could be used routinely.
Coffea arabica breeding; mixed linear models; variance components; perennial plants breedin; prediction of random variables; selection
