How stable are “stable” grasslands? Confronting Landsat-based habitat maps with long-term vegetation resurveys in the Western Alps

Earth observation (EO) products are increasingly used to derive biodiversity indicators and support national reporting, yet their ability to capture fine-scale ecological change in complex mountain landscapes remains uncertain. We present an ongoing comparison between Landsat-based land cover and habitat maps for Gran Paradiso National Park (Western Italian Alps) and a unique set of resurveys of historical vegetation plots collected 40 years apart. 

Our EO analysis suggests that ~88% of grasslands have remained stable in terms of land-cover class over recent decades. However, field resurveys reveal marked changes in species composition and diversity, including the rapid transformation of snow-beds into grasslands. Structural changes, including woody encroachment, is well detected from both survey methods. These asynchronous signals of change raise critical questions about what is actually being measured, e.g. land cover, habitat state, or biodiversity status, when EO products are translated into indicators and impact metrics. Focusing on montane to alpine grasslands, we (i) quantify agreement between land-cover and habitat classes and in situ plot assignments, (ii) explore detectability thresholds for grassland-rock mosaics that are highly interspersed and prone to misclassification at high elevation, and (iii) identify where EO “stability” masks substantial compositional and functional change detected by botanists. We relate these patterns to class hierarchy, thematically aggregated versus detailed legends, and spatial resolution constraints, and discuss the consequences for derived indicators such as habitat extent, integrity and turnover. 

By comparing the viewpoints of EO specialists and botanists, this contribution provides a concrete case study of the opportunities and limitations of merging satellite time series with vegetation resurveys. We outline practical recommendations for linking measurements to indicators, including the role of uncertainty analysis, cross-walking between land-cover and habitat typologies, and targeted field validation in areas of low detectability. Our results support the design of robust, scale-aware indicators for the Kunming-Montréal Global Biodiversity Framework and for adaptive management in protected mountain areas.