Abstract:
The objective of this work was to estimate the repeatability coefficients of mangaba (Hancornia speciosa) fruit traits, in order to define the number of fruit needed for an accurate selection of superior genotypes, as well as to conjecture about the nature of the phenotypic variation of these traits. Evaluations were performed for 160 fruit of 16 genotypes from two native H. speciosa populations of Goiás Velho and Padre Bernardo, in the state of Goiás, Brazil. Repeatability was estimated by the analysis of variance, principal component analyses based on covariance and on the correlation matrix, and structural analysis based on the correlation matrix. Repeatability estimates for fruit weight, diameter, and length, as well as seed number and weight, were of low magnitude, from 0.02 to 0.62, indicating low heritability. Repeatability estimates for ºBrix, titratable acidity, and ºBrix/acidity ratio were higher, from 0.34 to 0.91, indicating a low to potentially moderate heritability. The number of fruit for an effective selection of the best genotypes for titratable acidity, ºBrix, ºBrix/acidity ratio, and fruit weight is four for a 0.85 accuracy level. However, seven fruit would allow 0.90 accuracy for the same traits, and 0.85 accuracy for fruit length and diameter. The number and weight of seed per fruit are not effective for predicting the real value of a genotype.
Index terms:
coefficient of determination; mangaba; repeated measures; savannah; selection efficiency