Aims: The objective of this paper is to study the temporal stability of withinfield spatial variability of the Normalised Difference Vegetative Index(NDVI) at two time scales: intra-annual and inter-annual. This study aims to provide answers to the practical use of NDVI and, in particular, to determine whether it is possible (i) to advance the date of image acquisition in order to increase the time required for image analysis and interpretation before harvest and (ii) to verify if an image acquired in one year can be used to manage the vineyard in the following years. Methods and results: The study was conducted on 17 individual fields. The analysis of the intra-annual stability was performed with four images in 2007 and two images in 2006 that were acquired at different stages of vine development. The analysis of the inter-annual stability was performed with five images taken around veraison on five different years over a period of 10 years (from 1999 to 2009). For the 17 fields of the study, a sampling grid was defined to take into account the characteristics of image processing and the particular shape of each field. A rank coefficient of correlation (Spearman) was used to characterize the correlation between dates of acquisition (images). A Kendall test was implemented to individually characterize and identify the source of the observed temporal stability. Conclusion: In Mediterranean conditions, this study highlighted the temporal stability of within-field NDVI patterns both within a season or between seasons. Regarding the intra-annual scale, an image acquired from 15 to 20 days before veraison had a significant correlation (p < 0.05) with an image acquired at harvest. For earlier images (i.e., taken around flowering), the strength of the correlation decreased as the time lag between two images increased. This decrease was probably linked to summer pruning operations, the presence of an inter-row cover crop or a spring vigour that differed from the final vigour in some fields. Regarding the inter-annual scale, images acquired at veraison were all significantly correlated (p < 0.05) over the 10 year period regardless of the time lag between image acquisition. The degree of the correlation decreases continuously with time. For the 17 fields of the experiment, a decrease in the stability of NDVI between years was noticeable when significant changes in vine training (irrigation, replanting) occurred. For fields that did not undergo major changes, the spatial patterns of NDVI could be considered relatively stable over time for periods up to 3 to 5 years according to the age of the vineyard. Significance and impact of the study: This study showed that in Mediterranean conditions it is possible (i) to advance the date of image acquisition to at least 20 days before veraison if the objective is to highlight the spatial variability at harvest, (ii) to use information from an image acquired at veraison over several subsequent years if the field does not undergo major changes in management practices and (iii) to use earlyseason images (around flowering) as a potential source of information for managing other operations.