Fruit growth models are used, among other applications, to describe the evolution of fruit size throughout its development period. For forecasting, these models allow estimate in advance the size that will achieve the fruits at harvest. To have growth curves that provide unbiased estimate of the fruit size at harvest is necessary to design an appropriate sampling at the data collection stage for its construction. The objective of the present work was to determine the optimal sample size, consisting of trees (n) and fruits (m), to establish the growth patterns of orange fruits 'Valencia late' to estimate the size distribution at harvest. It was used the equatorial diameter of fruits before harvest, from two orchards located in the province of Corrientes, Argentina, during three seasons. Through mixed models were estimated components of variances between trees and fruits, and then from two types of methodologies was determined optimal sample size. The variability among fruits was higher than the variability between trees. To determine the growth pattern of orange fruit 'Valencia late' by sampling in two-stages, it is suggested to select 7 trees and 30 fruit per tree, for the estimative to reach equatorial diameter of fruit with an accuracy between 2 and 3%.
biometric; citrus fruit; fruit sampling; growth curves
