This paper studies the monitoring of oak dieback in forests of the Centre-Val de Loire region, France. Due to climate change, drought-induced forest dieback has become a major concern in temperate forests, including our study area, where oak is a key species. In order to better assess and adapt the actions needed to mitigate these impacts, access to accurate and regularly updated maps of forest health has become essential. In this context, the main objective of the study is to evaluate the interest of multispectral satellite time series for operational monitoring of forest dieback. Thanks to the in-situ data collected from 2017 to 2022 on about 2700 oak plots, a multiyear mapping of the analyzed region was performed using a supervised classification approach with the Random Forest algorithm. The results show that it is possible to detect oak dieback accurately (average overall accuracy = 80% and average balanced accuracy = 79%). More importantly, this study highlights the importance of measuring the temporal stability of the classification model in addition to standard cross-validation metrics. In this respect, the samples used for training are selected using data augmentation and balancing techniques to achieve better generalization over years. The learned model can also be used for predictive mapping of forest dieback in the coming years, for which the balanced accuracy is slightly reduced to about 70%. A feature analysis is also performed and shows that the shortwave infrared (SWIR) part of the spectrum is the most important for mapping forest dieback. In addition, using the red-edge portion of the spectrum can increase the stability of the model over time. We show that using only two vegetation indices based on continuum removal (CR) of the red edge (CRre) and shortwave infrared (CRswir) parts of the Sentinel-2 spectrum is sufficient to efficiently capture forest dieback. Overall, both in situ data and model predictions showed evidence of forest decline in many areas of the study region. Moreover, our results show that large areas of forest can decline over short periods of time, highlighting the interest of satellite data to provide timely and accurate information on forest status at large scales. This encourages the use and improvement of such approaches in the future.