Mountain pastures are part of an ecosystem that hosts a rich plant biodiversity organized in sub-communities. An accurate long-term monitoring of this ecosystem, going beyond the sole ground survey, is of primary importance for forage production planning. Within the GrassSense project, we develop a novel analytical framework to study the vegetation variability of mountain pastures based on the joint statistical analysis of ground data and satellite imagery. The driving research questions are whether it is possible to monitor the vegetation variability of pastures using commonly available satellite imagery, what satellite product types, and which kind of analytical workflow are suitable to track different vegetation associations in relation to mapped data. We investigate two study zones: a mid-mountain pastoral environment in the Toggenburg district (St. Gallen canton) and mid-to-high mountain pastures surrounding the Swiss National Park in the Grisons canton. Both environments are ground mapped for pasture vegetation associations, including wetlands, fertile pastures, improvable, and dry units. The preliminary results show that using publicly available images from the Landsat program (NASA), with a 30-m spatial resolution and sub-monthly revisit time, can catch a clear difference between wet and dry units and, to some extent, between different fertile and improvable ones. This is observed by computing the Normalized Difference Vegetation Index (NDVI), indicating living vegetation and proportional to photosynthetic activity. Moreover, NDVI shows a different intensity of vegetation activity for different months, reflecting the phenological cycle of the units. The temporal difference among units is barely visible since Landsat has a too coarse (15-day) revisit time, suggesting that a higher temporal resolution, e.g. using commercial satellite products, could improve the temporal analysis. The results also show that shadow cast by mountains is one main disturbing factor of the statistical relation between the vegetation variability and the observed spectral properties. As a solution, cast shadow can be computed for every image to select non disturbed data or apply a radiometric correction. This research proceeds in the direction of the joint spatial analysis of NDVI and other spectral properties in relation to altitude, aspect, and climatological conditions, with the goal of identifying a set of explanatory variables to track the vegetation variability in these highly complex systems.