Crop load expressed in terms of intercepted photosynthetically-active radiation can be used as a covariate to compare peach tree performance

Summary When fruit tree performance is evaluated in rootstock, thinning or other management trials, tree size and fruit number (or crop load) are the most important factors that confound the real effects of the treatments that are being evaluated. An experiment to assess the effects of tree size and crop load on yield efficiency and fruit size was conducted in a ‘Ross’ peach orchard on ‘Nemaguard’ rootstock in Malloa, Chile. Two groups of 15 trees each were selected according to tree size and were hand-thinned at the beginning of pit hardening to a wide range of crop loads within each tree size group, but keeping the average crop load similar between the two groups of trees. Fruits per tree (crop load) was either normalised or not normalised for tree size, assessed either as cm–2 of trunk cross-sectional area (TCSA) or the fraction of above-canopy photosynthetically-active radiation (PARf) intercepted by the canopy at harvest. At harvest, all the fruits were counted and weighed, and the average fruit weight calculated. Analysis of variance and covariance (with and without using crop load as a covariate), and regression analysis were performed and the results were compared. Differences in fruit size, yield efficiency (yield normalised by tree size) and total production between the two tree size groups were detected by ANOVA; but, by selecting the appropriate covariate, no differences between groups were detected. Thus, for a correct interpretation of the effect of treatment, in this case tree size, on tree performance measured as fruit size, yield, or yield efficiency, the differences in crop load (number of fruits per tree) must be removed by performing covariance analysis. Normalising crop load (fruits per tree) to account for tree size was essential for proper interpretation of the data. Calculating crop load using the fraction of light intercepted at harvest as the normalising factor was better than using TCSA as the normalising factor.