This will be needed for all of the grid-related processing functions. It would be troublesome to have to generate the grid every time.
This could either be a new dataset from an existing grid or we can grid something ourselves and use it. Not sure what's best.