Estratégias para a seleção em um programa de melhoramento de eucalipto

Abstract

One of the limitations of progeny testing is the low correlation between individual performances in progeny testing and their performance in the clonal testing. Thus, breeders must seek strategies that provide favorable scenarios for selecting the best individuals. The present study was conducted with the following objectives: to optimize clonal and parent selection in eucalyptus by modeling additive and non-additive effects using different kinship structures; and to evaluate the minimum number of individuals needed to represent a half-sibling progeny in eucalyptus. For this purpose, two research studies were conducted using a dataset provided by the company Suzano S. A. Data referring to two half-sib progeny tests (TP1 and TP2) was used, both implemented in a randomized complete block design with four replications, plots of four rows with five plants, and 80 individuals per progeny. The first study focused on optimizing clonal and parent selection by modeling genetic effects based on different pedigree structures. Data from a progeny test were used to estimate the average annual wood increment, pulp yield, and wood basic density, all at six years. Genotyping was performed on 1,740 out of 2,160 plants, and analyses were conducted using two model structures, with and without non-additive effects, combined with three pedigree structures: original, corrected, and hybrid. The coincidence index was estimated for selecting the 30 best individuals for parents and the 200 best for clones. In the second study, the focus was on estimating the minimum number of individuals per progeny to be used in a half-sibling progeny test while maintaining a scenario with good experimental precision. Data from two half-sibling progeny tests were used, estimating the average annual wood increment at three years of age. The number of individuals tested per progeny ranged from 8 to 76, with 1,000 non-replacement resamplings for each scenario. Each resampling provided estimates of additive genetic variance between progenies, within-plot variance, residual variance, population mean, selective accuracy, genetic coefficient of variation, and residual coefficient of variation. For the first study, the results showed that non-additive genetic effects were significant only for IMA. The use of the hybrid pedigree was more efficient than others, yielding high accuracy estimates. The models did not influence the selection of individuals for parentage, but there were changes when selecting individuals for cloning. In the second study, the number of individuals per progeny had little or no influence on most parameters. Its greatest impact was on accuracy estimates, with minimal increments beyond 60 individuals, indicating that accurate selection could be achieved in this scenario.