Neural meshing of a geographical space in regard to oceanographic data location

Presents a method to mesh a geographical area in regard to oceanographic data locations. The meshing is achieved using Kohonen´s networks. The idea is to adapt the position of the nodes (neurons) by successively considering the location of each data and moving the closest node and its neighbours toward this location. The authors have successfully tested this method on a database of the northeast Atlantic Ocean. The authors have constrained the meshing in regard to some practical considerations (minimal size of the meshes, discrete positioning of the nodes, ability to create, remove or move nodes during the adaptation...). Kohonen´s networks appear as an efficient and a robust meshing tool