This paper investigates the potential of time series of Sentinel-2 satellite images and derived normalised difference vegetation index (NDVI) to characterise vine water status. It assumes that in non-irrigated conditions, growth is the main factor affected by water restriction. Therefore, NDVI time series may be relevant to identify sites presenting differences in vine water restriction. The study was carried out in 7 plots and 2 reference plots across the Mediterranean basin. It shows that raw NDVI temporal series present some limitations. It then proposes an original methodology based on empirical cumulative distribution function of NDVI values observed over the whole season to account for the whole temporal dynamics of NDVI. Results showed that an approach based on cumulative distribution of NDVI values is relevant to classify plots in relation to their water restriction.