Monitoring the greening of mountain pasture habitats based on satellite image analysis in response to elevation and seasonal weather change.

Abstract

Mountain pastures host rich plant biodiversity organized in various distinct habitats. An accurate long-term monitoring, going beyond the sole ground survey, is of primary importance for nature conservation and forage production planning. We develop here a novel analytical framework to identify and monitor plant communities in mountain pastures based on the joint statistical analysis of ground data and satellite imagery. We use commonly available satellite imagery (Sentinel-2) to track, for the first time to our knowledge, spatial and temporal changes of individual habitats composing the mountain pasture ecosystem, in relation to interannual hydroclimatic variations. We consider as study zone the mid-to-high elevation mountain pastures surrounding the Swiss National Park in the Grisons canton, Switzerland (approx. 100 sq. km), including fertile pastures, wetlands, dry plant communities, and shrubs. We couple the habitat map to the NDVI spectral index (Sentinel-2 images, ESA) to retrieve a proxy for living vegetation and its productivity. Then, by computing statistical parameters of the NDVI curves, we characterize the annual greening season for the different habitats, taking into account elevational changes and interannual variations of snow persistence, depending on autumn and winter rainfall. The different habitats show a marked difference in their productivity in function of their wetness until 2400 m a.s.l, while they seem to homogenize at higher altitudes. The greening in the 1-st season half is strongly controlled by snow persistence variations and partially compensated by an after-snowmelt quick growth. Conversely, the 2-nd half season greening is mainly linked to the season maximum NDVI. This workflow presents as an effective strategy to monitor the seasonal and long-term evolution of mountain pasture vegetation in the complex alpine domain.