We address the problem of optimal selection of sites to constitute a nature reserve which ensures that a given set of species has fixed survival probabilities. This classic problem has already been considered in the literature of conservation biology. The originality of this article is to consider that the values of the survival probabilities of each species in each potential site may be subject to a certain error while assuming that the number of sites where these probabilities are wrong is limited. We thus define a set of possible survival probability values in each site. We then show how to determine, by solving a relatively simple mixed-integer linear program, an optimal robust reserve, i.e., a reserve which ensures that each species has a certain survival probability whatever the values taken by the survival probabilities in each site, in the set of possible values. The fact of being able to formulate the search for an optimal robust reserve by a mixed-integer linear program provides an easy way to take into account additional constraints on the selection of sites such as, for example, spatial constraints. We report some computational experiments carried out on many hypothetical landscapes to illustrate the concept of robust reserve and show the effectiveness of the approach.