Sorry to start a new thread, but I was afraid that my comment buried in the thread about zero-level grids was lost.
The grdpaste command seems to expect grids that share an entire boundary. Is it absolutely necessary that that be the case? I’d sure love it if grdpaste could paste whatever boundary is shared and thus permit the user to form L-shaped grids (even if the open section is filled with NaNs).
My guess is that it would take some very fancy (and smart) decision-making code to make that happen. Or maybe there is a restriction with NetCDF that prevents it.
Just hoping that this question might plant a seed for an enhancement, down the road…