Regridding fine to coarse and missing lon/lat coords in fine GOES-16 FOV grid

I have been trying multiple regridding software available in Python (i.e xESMF, xarray coarsen, pyresample, ESMFPy, etc…) and otherwise (cdo, nco ncremap, ncl ESMF_Regrid, etc…).

I have the file of the GLM field of view longitudes/latitudes. I used Brian Blaylocks’s script here: goes2go/create_standard_GOES_latlon_grids.ipynb at master · blaylockbk/goes2go · GitHub to generate the the GOES 16 FOV. As you know it is 5424 x 5424 file and it also has missing values

I presume that the missing values in these files come from the projection of the FOV onto a sphere.

With most of the regridding software (cdo, ncl ESMF Regrid, nco ncremap, etc), a netcdf file needs to include variables that contain the a variable of 4 corners of each of a 2-D lon lat or 2 bounds of a 1-D lon/lat. So I have added those variables. In these cases within the GOES FOV file, a subset of grids will either have all or 1-3 of the 4 are missing values… In other words, the grid is not well defined.
The missing values are usually printed out to an automatic fill_value number of 1e+20 or 9.96e+36, hence the calculation this throws the regridding calculation off. And I usually try to do a conservative regridding with any of the software where that option is available since I am going from a very fine 5424 x 5424 to 128 (1.25 degree Lon) to 162 (1 degree lat).

Would it be wise to do some type of interpolation to estimate those missing coordinates with the GOES-16 FOV? Or should I use a fixed coordinate system where the coordinates are (0,0) to (5424,5424)? Do I need to mask out the missing coordinates?

I am taking a stab in the dark at this point. I don’t know if the missing values are my problem but trying to get to a solution.


1 Like