Oceanographers from the IFREMER institute have an hypothesis that the presence of so-called “retentive” meso-scale vortices in ocean and coastal waters could have an influence on watery fauna’s demography. Up to now, identification of retentive hydro-dynamical structures on stream maps has been performed by experts using background knowledge about the area. We tackle this task with filters induced by Genetic Programming, a technique that has already been successfully used in pattern matching problems. To overcome specific difficulties associated with this problem, we introduce a refined scheme that iterates the filters classification phase while giving them access to a memory of their previous decisions. These iterative filters achieve superior results and are compared to a set of other methods.