I'm trying to obtain a gridded bathymetric dataset of a certain region for use in a model, but all of the openly accessible products that I'm aware of (GEBCO, SRTM+) are very poorly constrained by in-situ observations in this region so are mainly based on gravity data. Unfortunately, I know as a fact that this has introduced some major errors (e.g. islands that don't exist).

I have point depth observations for this region from various historical studies which have not been incorporated into the above data products, and I would like to use them to modify the GEBCO grid to nudge depths in incorrect regions but I was wondering if there's an established method for doing this without introducing serious artefacts.