Rapid species extinctions and the loss of other biodiversity features worldwide have prompted the development of a systematic planning framework for the conservation of biodiversity. Limited resources (~ 40 million USDannually) are available for conservation, particularly in the developing countries that contain many of the world’s hotspots of species diversity. Thus, conservation planning problems are often represented as mathematical programs in which the objective is to select sites to serve as conservation areas so that the cost of the plan is as small as possible and adequate habitat is protected for each species. Here, we generalize this approach to allow for uncertainty in the planning process. In particular, we assume that the species to be protected disperse after the conservation areas are established and that planners cannot anticipate with certainty the species’ future locations when selecting the conservation areas. This uncertainty is modeled by including random variables in the mathematical program. We illustrate the approach by designing a network of conservation areas for birds in southern Quebec.