Modelling grape and wine quality through PLS Spline statistical method

Started in 1994, this project intends to explain quality of grapes and wines using data of soil, climate and vineyard that are currently used in field trials. Firstly set at a national scale, it has been transferred to the Aquitaine region in 2000. The work has been conducted by the ITV institute thanks to many other partners. 2 cultivars have been considered: cvs. Merlot and cabernet sauvignon.
A set of data has been collected using different years and plots showing varied environnemental and cultural situations. Data mining used PLS Spline method. 4 models have been produced: sugar and total acids in musts, colour intensity and total polyphenolic compounds in wines. These models point out the variables that are most influent on quality and order them. A validation with plots that have not been used to build the models has been done in 2006. The prediction is of correct level and gives a potential-like result. At the same time, the models have been integrated into a better convenient tool called SPQV 1.1 software. It is aimed to farmers’s advisors.
The models do not give any prediction during the year the grapes are produced, because it uses post-harvest variables. Nevertheless they can be a helpful tool for potential zoning, plots selection or planting advising.