Effective shorebird conservation requires a greater understanding of use and availability of high-tide roosts in coastal non-breeding areas. In this paper, we explore 1) variation in use of roosts; 2) landscape and environmental correlates of roost use; and 3) responses to predators and humans at roosts, to evaluate roost availability for a wintering population of Dunlin Calidris alpina pacifica at Humboldt Bay, CA, USA. For four winters (2002–2005), Dunlin use of particular roosts was highly variable at the population and individual level, even at the most-used roosts. In any given day, week, or month, most Dunlin roosts were unused, and we continued to record new roosts even in the fourth year of study. Although roost use was influenced by landscape attributes, time of day, and environmental conditions, these variables left most variation in roost use unexplained. Dunlin departures from roosts were associated with the presence of avian predators, but not with human activity. Collectively, these observations revealed no strong evidence that roost availability was limited. This contrasts with findings of other studies, even after considering differences in methodology and spatial scale of analysis. We discuss implications for the interpretation of movement data, protection and/or creation of roosting habitat, and assessment of roost quality.