Quantifying metapopulation dynamics is a challenging task. Difficulties particularly arise in species that possess unobservable resistance forms that bias the estimation of colonization and persistence rates. Here, we develop a general multistate occupancy model that allows estimation of species persistence for both normal and resistant forms, even when the latter are not detectable. We apply this model to an 11-year data set on the tropical freshwater snail Drepanotrema depressissimum in a network of 229 ponds. These ponds frequently dry out, and the snails can persist by aestivating in the ground, where they are not detected. Our model adequately captures this feature because it infers a high persistence rate for the resistant form, while models ignoring aestivation tend to overestimate extinction and colonization. In addition, we find that, surprisingly, colonization and persistence are even higher in sites prone to desiccation and during years with low rainfall than in more humid sites and years, suggesting that D. depressissimum favors unstable sites where competitors are rare and where it can rely on aestivation to persist. Our model has the potential to provide valuable insights into the metapopulation dynamics of many species that otherwise could hardly be studied because of the existence of undetectable life forms.