Know about this problem with griddata(...) function and NaN values. It is one of the reason I switched to delaunay mesh. It is possible to workaround byt setting NaN value to zero, it is possible to find indexes with isnan(...) function and find(...) function might be useful.
Thank you! Once I realized that the issue was just that I was trying to interpolate outside my sample data, I just expanded the region that I was sampling, and I no longer have the NA issue and everything is resolved. At least it makes sense that there would be problems when trying to interpolate outside the sample data.
If you want to try Kriging, have a look at the stk package:
https://octave.sourceforge.io/stk
I used it with satisfaction few years ago
Thank you! I'll save this for future reference.