Throughout the world, fragmentation of landscapes by human activities has constrained the opportunity for large herbivores to migrate. Conflict between people and wildlife results when migrating animals transmit disease to livestock, damage property, and threaten human safety. Mitigating this conflict requires understanding the forces that shape migration patterns. Bison Bos bison migrating from Yellowstone National Park into the state of Montana during winter and spring concern ranchers on lands surrounding the park because bison can transmit brucellosis (Brucella abortus) to cattle. Migrations have been constrained, with bison being lethally removed or moved back into the park. We developed a state-space model to support decisions on bison management aimed at mitigating conflict with landowners outside the park. The model integrated recent GPS observations with 22 years (1990-2012) of aerial counts to forecast monthly distributions and identify factors driving migration. Wintering areas were located along decreasing elevation gradients, and bison accumulated in wintering areas prior to moving to areas progressively lower in elevation. Bison movements were affected by time since the onset of snowpack, snowpack magnitude, standing crop, and herd size. Migration pathways were increasingly used over time, suggesting that experience or learning influenced movements. To support adaptive management of Yellowstone bison, we forecast future movements to evaluate alternatives. Our approach of developing models capable of making explicit probabilistic forecasts of large herbivore movements and seasonal distributions is applicable to managing the migratory movements of large herbivores worldwide. These forecasts allow managers to develop and refine strategies in advance, and promote sound decision-making that reduces conflict as migratory animals come into contact with people.