European larch (Larix decidua) forests of the western Alps form extensive cultural landscapes whose resilience to global changes is currently unknown. Resilience describes the capacity of ecological systems to maintain a "stable state", i.e. constant functions, processes, structure, and identity despite disturbances, environmental changes and internal fluctuations. Our aim is to explore the resilience of larch forests to changes in climate and land use in the western Italian Alps. To do so, we assumed that mountain forests ecosystems can exist under alternative stable states. To describe quantitatively the larch forest state we used species tree basal area data obtained from field forest inventories in combination with topography, forest structure, land use, and climate information. To infer the resilience of larch forests relative to that of other forest states we applied three different probabilistic methods: frequency distributions, logistic regressions, and potential analyses. We found patters indicative of alternative stable states: bimodality in the frequency distribution of the percent of larch basal area, and the presence of an unstable state, i.e., transient mixed larch forests, in the potential analyses. We also found: (1) high frequency of pure larch forests at high elevation, (2) the probability of pure larch forests increased mostly with elevation, and (3) pure larch forests were a stable state in the upper montane and subalpine belts. Likewise, in the upper montane belt open canopy cover and high grazing pressure increased the frequency of larch forests. Our study shows that the resilience of larch forests may increase with elevation, most likely due to the altitudinal effect on climate. Subalpine larch forests may be more resilient, and natural succession after land abandonment, e.g., towards Pinus cembra forests, seems slower than in montane larch forests. In contrast, in the upper montane belt only intense land use regimes characterized by open canopies and forest grazing may maintain larch forests. We conclude that similar approaches could be applied in other forest ecosystems to infer the relative resilience of tree species.