Previsão da produção em duas castas brancas: análise da variabilidade espacial e das relações entre componentes do rendimento

Mestrado em Engenharia Agronómica - Instituto Superior de Agronomia

Yield prediction determines an essential competitive advantage in viticulture. There are many methods to predict the yield, being the one based on yield components the most used. Recently there have been many studies for the purpose of maximize prediction efficiency exploring the automated yield prediction with autonomous vehicles. This work aims to estimate the yield of two white varieties using the based on yield components method to evaluate their usefulness in image analysis. For this purpose, two samples of 30 vines of Alvarinho and Viosinho were selected to quantify their yield components during their cultural cycle of 2015. Despite Viosinho have been more productive, in a general way both varieties showed a spatial variability of the same magnitude order (C.V. between 10 and 40%), being the inflorescence number, the flower buds number and the cluster number the ones which showed a greater variability. After the elaboration of relationships between those yield components, three mathematical models were created to predict the production per vine, the cluster weight and the berry number. In the first model the inflorescence number allowed to explain about 78% of the variability of production per vine in Alvarinho, enabling an anticipation of the forecast compared to the model based on the cluster number. In the second model the cluster volume (Alvarinho) and the berry weight (Viosinho) explained practically all the variability of cluster weight (R2 = 99%), while in the third model the cluster weight explained most of the percentage of variability of the berry number (R2 = 95%) in both varieties. The pre-image analysis showed the existence of high correlated coefficients between the different measured variables, allowing to predict the use of this tool to estimate the yield components