Improving early-season estimates of olive production using airborne pollen multi-sampling sites

Abstract

The aim of this study was to improve the accuracy of aeropalynological models to forecast yields in areas with heterogeneous characteristics by applying principal component analysis to integrate the airborne pollen sampled from more than one trap. The sampling was performed during the past seven years (1998–2004) in the main northeast olive regions of Portugal. Annual crop production was forecasted on the basis of airborne pollen concentration measured at flowering, comparing the performance of three different independent variables: total airborne pollen concentration sampled in each trap and a derived variable that was determined by principal component analysis of the total airborne pollen concentration sampled. The best predictive results were obtained using a logarithmic relationship with airborne pollen concentration principal component scores describing about 97% of olive fruit production variability over the last seven years. The use of this technique improved the ability of pollen to explain the production interannual variations by about 13%. The comparison between actual reported and the adjusted production showed an average spread deviation of 5%.