Amélioration des modélisations de cultures de maïs à l'aide de la télédétection

Abstract

In the context of ever denser demography and climate change, forecasting yields is of paramount importance. This can be useful at the international, national or producer level for assessing agricultural market fluctuations. Knowledge of potential production enables the farmer to become a dynamic player on the cereal market. Remote sensing is a rapidly expanding tool, thanks to the analysis of satellite images; it is a huge source of information about crops. In Belgium, the BELCAM project uses data collection from remote sensing to estimate future returns. The results observed from the Aquacrop model suggest that there are shifts in growth curves due to differences in early maturity between maize varieties. Earliness is a key criterion in corn productivity when determing the level of earliness is an essential information in the estimation of yields. The objective of this work is to validate this hypothesis using a discriminant analysis that will lead to the development of a classification tree. The last allows to binder per groups of precocity the data of Aquacrop for the different plots observed. Late varieties have shown a surprising result in classing in very early varieties. Several hypotheses explaining this phenomenon have been developed. Even imperfect, the objective of this Master Thesis is reached because the decision tree will be used in the next studies of BELCAM.