The development of models that allow forecasting yield tendencies is important to all sectors of the citrus industry. This work evaluated the influence of meteorological variables in different phases of the crop cycle in order to propose empirical models to estimate the number of fruits per plant (NFP) of 'Valencia' and 'Hamlin' sweet oranges. NFP sampling data from the citrus juice industry of the State of São Paulo, on the total of 15 harvests (1990/91 to 2004/05), classified into three age classes, and meteorological data of maximum and minimum air temperature and rainfall of Limeira, SP, Brazil, were utilized. Correlation coefficients were initially computed between the number of fruits per plant and each meteorological variable used for water balance and variables related to air temperature, in different periods. Linear multiple regression models were fit to describe the empirical relationship between NFP and the subsets of agrometeorological predictors that presented higher correlations in different phases of the crop cycle. The meteorological conditions during the phases of vegetative summer flush, pre-flowering, flowering and beginning of fruit growth influenced the number of fruits per plant. The proposed models presented adequate goodness-of-fit with determination coefficients varying from 0.72 to 0.87.
citrus; harvest; modeling; yield