The prediction of grape composition is becoming more important due to the need of reducing the current levels of alcohol and pH of the wines, a problem that is exacerbated by climate change. This work presents a 3-year study of the spatial variability of grape composition in a rainfed Tempranillo vineyard located in Rioja (Spain). It is based on the acquisition of multispectral imagery at véraison (start of the ripening process); and zoning based on NDVI, to assess its performance for zonal management. The results reveal a high spatial variability within the plot, with a stable pattern over the years, even with very different climate conditions. NDVI was a good predictor of vegetative growth variables. However, the prediction of grape composition was more complex. Unexpectedly, anthocyanins were found to be higher in the highest vigor zone, which is probably related to the effects of climate change. This unexpected relationship is particularly discussed in the article.