Ecosystem service (ES) mapping has been developed with the aim of supporting ecosystem management, but ES maps often lack information about uncertainty and risk, which is essential for decision-making. In this paper, we use a risk-based approach to map ES in mountain forests, which are experiencing an increasing rate of natural disturbances, such as windthrow, bark beetle outbreaks, and forest fires. These disturbances affect the capacity of forests to provide essential ecosystem services, such as protection from natural hazards, wood production, and carbon sequestration, thus posing a challenge for forest management. At the same time, disturbances may also have a positive effect on certain services, e.g. by improving habitats for species that rely on dead wood. We integrate forests’ susceptibility to natural disturbances into probabilistic Bayesian Network models of a set of ES (avalanche protection, carbon sequestration, recreation, habitats, and wood production), which combine in-formation from remote sensing, social media, and in-situ data, existing process-based models, and local expert knowledge. We use these models to map the level of the services and the associated uncertainties under scenarios with and without natural disturbances in two case study areas in the Swiss Alps. We use clustering to identify bundles of risk to ES, and compare the patterns of risk between the non-protected area of Davos and the strictly protected area of the Swiss National park with its surroundings. The spatially heterogeneous pattern of risk to ES reflects topographic variability and the forest characteristics that drive disturbance susceptibility, but also the demand for ecosystem services. In the landscape of Davos, the most relevant risks to ES are related to decreases in the protection against avalanches and carbon sequestration, as well as some risk to wood production and recreation. In the strictly protected Swiss National Park, the overall level of ES risk is lower, with an increase in habitat quality under the disturbance scenario. This risk-based approach can help identify stands with high levels of ES that are particularly susceptible to disturbances, as well as forests with a more stable ES provision, which can help define priorities in forest management planning.