The demography of anchovy fishes in the Gulf of Biscay seems to be related to the presence of socalled "retentive" hydrodynamical structures, that keep fish eggs and larvae in a favorable environment. To verify this hypothesis, an automatic detection tool is needed to process a database of thousands of hydrodynamical maps to be compared with biologists observations and fisheries statistics. The results could be used to decide fishing quotas or bans for the sake of preserving the natural resource. We propose two evolutionary schemes, one based on ant systems, the other on induction of genetic programming filters, in order to automatically detect these retentive structures. Our heuristics are shown to be competitive with human experts and are compared to other detection schemes.