Rapid mass movements such as landslides, debris flows and snow avalanches pose a considerable danger to infrastructure in mountain areas. The accurate assessment of their destructive power is thus crucial in hazard management. To date, the impact pressure caused by snow avalanches on infrastructure is calculated with simple analytical models, and recently the analogy between snow avalanches and granular flows has been used to improve the quantification of the impact pressure using increasingly sophisticated numerical methods like the Discrete Element Method (DEM). In particular recent 3D simulations using DEM offered important new insights into pressure build-up mechanisms during the interaction between cohesive granular flows and obstacles. However, no plastic compaction was considered in these simulations and only small systems were studied due to computational reasons. To address these shortcomings, we use and further develop a more computationally efficient model based on the Material Point Method (MPM) in which snow is described as a compressible, elastoplastic continuum. Validation is performed by comparing MPM results with impact pressure data from non-generic obstacles from the Vallèe de la Sionne snow avalanche test site. This leads to a powerful numerical laboratory for investigation of various scenarios of full-scale avalanches interacting with various types of structures. Particular attention is devoted to analyzing phenomena like material run-up, non-steady state effects like accelerating or decelerating avalanches, confinement effects imposed by varying flow widths as well as the effect of compressibility. Furthermore, the high resolution of the simulation allows for the evaluation of the spatial distribution of the impact pressure on obstacles, which is important for engineering applications. This work enables us to better understand avalanche-obstacle interaction and will ultimately contribute to improving empirical impact pressure guidelines used for avalanche risk management.