In seasonal environments, the winter months can drive the structure and dynamics of vertebrate communities. Most ecological studies have focused on species above the snow, overlooking those within the subnivium (below-snow environment). The Great Lakes region of North America is characterized by seasonal snow cover and a diverse assemblage of small mammals, both of which are changing rapidly in response to a changing climate. To study the seasonal dynamics of small mammals, we sampled communities in three forest types (hardwood, hemlock-cedar, and spruce bog) in summer and winter over 2 years. We hypothesized that seasonal differences in communities would be moderated by forest structure in summer and snow in winter. We assessed community dominance via species diversity and the abundances of deer mice (Peromyscus maniculatus), white-footed mice (P. leucopus), and southern red-backed voles (Myodes gapperi). We estimated a Community Temperature Index (CTI; a metric of the balance between warm- and cold-adapted species in a community) and tested seasonal drivers of CTI. Species diversity was highest in spruce bogs in winter due to increased species evenness, but diversity was consistent year-round in hardwood and hemlock-cedar forests due to unchanging species dominance. Spruce bogs supported the coldest CTI in summer, but CTIs converged across forest types in winter. In summer, forest structure drove CTI; in winter, colder CTIs were associated with deeper snow and a warmer subnivium. Our work reveals that snow is a critical habitat and filter for cold-adapted species that ultimately structures small mammal communities. Given rapid changes in snow conditions due to climate change, we predict continued range contractions for subnivium-adapted species.