Abstract: Anthropogenic climate change is amplified in the Arctic, where less sea ice enables energetic wave climates while higher air and soil temperatures increase tundra erodibility. These changes are likely to exacerbate retreat of coastal tundra yet remain poorly constrained on timescales relevant to storm wave impacts. A stochastic weather generator is used to create 1,000 synthetic hourly time series of waves, water levels, offshore sea ice concentration, and air temperatures used as forcing for an efficient coastal tundra model. The ensemble set of morphological change simulations provides a probabilistic perspective on the range of tundra retreats and the relative effects of each environmental forcing. Ensembles show as the depth of the erodible layer increases, the style of tundra retreat shifts from a consistent recession to intermittent events with large magnitudes and a factor 2 range in outcomes. Model scenarios highlight shallower thaw depths narrows the range of retreats and reduces individual extreme events, but a dynamic feedback between beach slopes, wave runup, and thermally limited erosion volumes ultimately increases the number of storm events associated with retreat. The minimum tundra retreat is governed by background shoreline change and the specifics of the topographic profile dominate underlying changes in the future wave climate statistics and open water season. As the Arctic continues to warm, the change in retreat style will have significant ramifications for coastal resilience.