The metapopulation concept is a cornerstone in the recent history of ecology and evolution. However, determining whether a natural system fits a metapopulation model is a complex issue. Extinction-colonization dynamics are indeed often difficult to quantify because species detectability is not always 100%, resulting in an imperfect record of extinctions. Here, we explore whether combining population genetics with demographic and ecological surveys can yield more realistic estimates of metapopulation dynamics. We apply this approach to the freshwater snail Drepanotrema depressissimum in a fragmented landscape of tropical ponds. In addition to studying correlations between genetic diversity and demographical or ecological characteristics, we undertake, for the first time, a detailed search for genetic signatures of extinction–recolonization events using temporal changes in allele frequencies within sites. Surprisingly, genetic data indicate that extinction is much rarer than suggested by demographic surveys. Consequently, this system is better described as a set of populations with different sizes and immigration rates than as a true metapopulation. We identify several cases of apparent extinction owing to nondetection of low-density populations, and of aestivating individuals in desiccated ponds. More generally, we observed a frequent mismatch between genetic and demographical/ecological information at small spatial and temporal scales. We discuss the causes of these discrepancies and show how these two types of data provide complementary information on population dynamics and history, especially when temporal genetic samples are available.