LAUSR

Abstract In the present study, bio-meteorological regression models for forecasting the fruit produced by the olive trees in Jaen (southern Spain), the province with the largest extension of olive groves in the world, were revised and improved. The new forecasting models were constructed using partial least-squares regression, taking the annual olive yield as the dependent variable and several aerobiological and meteorological parameters as the independent variables. The models were validated following a full cross-validation method. A 23-year period (1994–2016) was used. The number of days with pollen concentrations ≥400 pollen grains m−3 was revealed as a newfangled predictive variable to accurately predict the olive harvest in this area, being included in the forecasting model with the highest determination coefficient value (R2 = 0.89). Weather-related variables such as the cumulative precipitation from October to December of the previous year or the mean maximum temperature from January to March were also factors of particular importance on crop production. The new model proposed provides early and effective olive crop forecasting by using independent variables which can be easily obtained towards the middle of June, also incorporating to the model the phenological variability associated with changes in the local weather. The approach shown in this study readily could be applied to any potential new situation and to be extensible to other similar olive growing areas across the Mediterranean region.