This paper proposes a methodology aiming at using historical yield data to improve yield sampling and yield estimation. The sampling method is based on a collaboration between historical data (at least three years) and yield measurements of the year performed on some sites within the field. It assumes a temporal stability of within field yield spatial patterns over the years. The first factor of a principal component analysis (PCA) is used to summarize the stable temporal patterns of within field yield data and it represents a large part of the variability of the different years assuming yield temporal stability and a high positive correlation between this factor and the yield. This main factor is then used to choose the best sites to sample (target sampling). Yield measurements are then used to calibrate a model that relates yield values to coordinates on the first factor of the PCA. This sampling method was tested on three vine fields (Vitis vinifera L.) in Chile and France with different varieties (Chardonnay, Cabernet Sauvignon and Syrah). For each of these fields, yield data of several years were available at the within field level. After temporal stability of yield patterns was verified for almost all the fields, the proposed sampling method was applied. Results were compared to those of a classical random sampling method showing that the use of historical yield data allows sampling sites selection to be optimized. Errors in yield estimations were reduced by more than 10% in all the cases, except when yield stable patterns are affected by specific events, i.e. early frost occurring on Chardonnay field.