Flow sensing has recently gained attention of the robotics community, as it can complement the conventional sensing modalities of vision and sonar in underwater robotics. There is increasing literature in flow sensing for robotics focusing on performing tasks such as object detection and positioning, and robot's orientation estimation, most commonly under idealized laboratory conditions. In this paper, using recent advances and methodologies for bioinspired flow sensing, we propose a map-based localization technique that employs simulated hydrodynamic maps. The proposed conceptual idea could be an interesting complement to perform localization in those underwater environments with structural maps and a heterogeneous hydrodynamic, such as dams, harbour structures, fishways, caves, swers or any other drowned structure. To demonstrate its performance, computational fluid dynamic models are used to generate flow-speed maps of a structured underwater environment. Later, during off-line experiments, pressure data acquired using a flow sensing probe fitted on a Cartesian robot is transformed into speed information, and used inside a particle-filter to perform localization within the simulated flow-speed maps. The proposed technique has been tested using multiple scenarios with varying particle densities and motion command error levels. The results show filter convergence for all studied scenarios, inducing motion errors up to 0.20 m, suggesting that flow based information could be used to improve the navigation and localization abilities of underwater robots.