Summary In many places in Europe, trees are currently far below their theoretical altitudinal limit because of past land use. However, under the current crisis affecting mountain agriculture, trees are likely to recolonize the subalpine belt. Addressing the societal issues associated with such landscape change requires predictive tools to assess vegetation dynamics in relation to management strategies. This study aims to analyse the factors determining tree dynamics at the subalpine ecotone and to evaluate the impact of land‐use change on landscape vegetation patterns. We developed a hierarchical scaling approach and applied it to a European larch (Larix decidua Mill.) colonization in the French Alps. Our landscape case study focused on subalpine south‐facing grasslands of Villar d’Arêne (France). First, we used a habitat‐suitability model to delineate the climatic suitability habitats of larch in the French Alps. Then we used a landscape model (LaMoS) to determine whether larch is able to colonize these grasslands predicted to be climatically suitable, considering land use and local vegetation. The sensitivity of landscape colonization patterns to land‐use scenarios (abandonment, delayed mowing or traditional mowing) and biological attributes of larch were analysed with a factorial simulation experiment. The accuracy of the habitat‐suitability model at the French Alps scale allowed for the prediction that the grasslands of Villar d’Arêne were highly suitable for larch. Simulations highlighted the effect of land use on larch establishment at the local but not at the landscape scale. On the other hand, larch attributes such as dispersal capacity, juvenile tolerance to light interaction (competition and facilitation) and its capacity to survive on infertile soil were shown as essential for tree spatial dynamics.  Synthesis and applications. Combining a habitat‐suitability model with a spatially and temporally explicit landscape model enhances the capacity to account for environmental factors acting at different overlapping scales. This modelling strategy increases the robustness and accuracy of predictions, a prerequisite for landscape management in a global change context. We conclude that ongoing and future agri‐environmental policies have to be quickly adapted to protect biodiversity and ecosystem services provided by subalpine grasslands.