Tree line dynamics and forest densification in the European Alps revealed by Landsat images and machine learning: a case study in the Senales/Schnals Valley

Abstract

The alpine tree line represents one of the most climate-sensitive ecological boundaries, where multiple interacting factors determine vegetation distribution at its upper limit. This study investigates the spatio-temporal dynamics of the tree line in the Senales Valley (South Tyrol, Italy) between 1985 and 2023, combining multi-temporal Landsat imagery, Random Forest (RF) classification, and visual orthophoto interpretation performed by manually delineating the forest boundary to assess both spatial and elevational shifts. Climatic variables (temperature, precipitation, snow cover, and growing season length) were analysed using linear models (LM) and generalized additive models (GAM) to identify long-term trends and potential drivers of tree line migration. The results reveal a consistent increase in forest cover in all 16 study areas, averaging +44%, with the largest expansions occurring on slopes facing W. Elevational advances were recorded in 15 of 16 areas, averaging +32 m for Landsat-derived data and +45 m for orthophotos. Elevated minimum temperatures during spring and autumn, alongside warmer summers and a significant rise in precipitation during the same season, created conditions which maintained soil moisture and reduced water stress—factors known to facilitate tree line advancement. Wind exposure from the N-NW sector and associated föhn effects appeared to limit tree line expansion on S-SE facing slopes. Comparison between manual and RF-derived tree lines revealed overall high agreement, with deviations below one Landsat pixel (30 m) in most cases. This confirms that Landsat imagery combined with RF algorithms provides a robust, cost-effective method for assessing long-term tree line dynamics in heterogeneous alpine environments.