Repeatability and number of harvests for selection of Amazon nut genotypes in an agroforestry cultivation system

- The objective of this work was to estimate the repeatability coefficients of variables related to fruit and nut production, the minimum number of crops required for genotype selection, and the permanent phenotypic correlation coefficients between variables in a cultivated population of Amazon nut. A total of 40 genotypes from seeds, belonging to an agroforestry system implemented in 1995, in the state of Roraima, Brazil, were evaluated during ten years for six variables related to fruit and nut production. The genetic-statistical analyses were performed using the maximum restricted likelihood and best linear unbiased prediction (REML/BLUP) method, based on the basic model of repeatability, without an experimental design, of the Selegen-REML/BLUP software. Genetic variability between genotypes and a moderate regularity in the repetition of the studied variables between different harvests were detected. The number of harvests required to reach a coefficient of determination of 80% ranged from seven, for the variables average nut weight per fruit and number of fruits per plant, to nine, for average nut weight. The correlation results indicate the possibility of indirect selection for nut production through the evaluation of the number and weight of fruits per plant.