Development of a method for vine yield estimation using within field historical database

Abstract

Yield estimation is an important task for any vine grower to organize logistical operations such as harvest, wine storing and also sell it on the market. Some researches as well as results from vine growers in the central zone of Chile have reported errors about 15 to 40% in yield estimation using classical field manual sampling. Consequently, there is an urgent demand for improving yield estimation. This is a difficult labour due to the inherent variability from vine to vine, from zones to zones at the within-field level and also from field to field. This study aims at addressing the question of yield estimation before harvest. It also wants developing a methodology to estimate grapevine yield from punctual measurements of the year and a model build-up from within field historical yield database. The model assumes a temporal stability of the within field variability. That last is derived from the main factor of a principal component analysis (PCA). Thus, this approach was tested on seven seasons in a Syrah vine field under non-irrigated conditions in Narbonne region, France, and another database on four seasons in a Cabernet Sauvignon vine field under irrigated conditions in Maule region, Chile. Yield measurements at harvest were used to evaluate the model. The results showed a linear correlation between standardized within-field yield values over the different seasons, for both vine fields. This supports the use of a linear model resulting from the PCA to summarize the yield within variability and use this information to best locate the measurement site over time. Thus, it was possible to estimate yield for 6 seasons in the Syrah field and 3 at the Cabernet Sauvignon field with r² greater than 0.93 in both cases, using only six representative sampling sites within the vine fields. Also, the results indicated that the approach was able to estimate the mean yield of the fields with an error about 5%.