netcdf model input

[bjarnebuchmann]

I have a few questions and comments regarding netcdf model data.

First late me state, that I am a huge fan of the open source and data. As I am also a fan of open data standards, netCDF happens to be a goto data standard for me. For the present project, the OTIS format is obviously useful for sharing "black-box data", but it is hardly convenient nor flexible. Thus, I really appreciate the flexibility of pyTMD to use netcdf both for models, for input and for output.

For the actual models, netcdf viewers, such as ncview, provides a good way to get a quick glance at the data.

But more importantly, the might also be a performance issue: If I understand things correctly, then pyTMD ahs to read entire model fields into memory before it can start the interpolation of constituents to target points , which is necessary before computing elevations and currents. With netcdf it should be rather easy for a user to "crop" (ncks, hyperslap) models to a smaller area, thus greatly reducing the fields. Obviously, care should be taken, and there might be some file naming issues, but it should be doable(?)

My use case - boundary conditions for an operational ocean model - would require to compute tidal predictions for a given set of points in a small regions many, many times [eg 4x/day for years]. As we have to do it over and over, computational efficiency becomes an issue.

This leads me to my main points: I should very much like support for tpxo9.5 global-file version including especially w/ netcdf support, see also #109
I haven't tried the atlas-versions yet, but I will as soon as I understand some other technical details (separate issues).

Further, I have heard (from the author's mouth) that the Arc2kmTM and Gr1kmTM models will be released in netcdf versions in some hopefully not to distant future. For these models, it might then also make sense to crop the models to a smaller region.

I will be glad for any kind of dialogue/discussion on this topic.
/Bjarne