Finally got CF geostationary projections working, thanks to Tom Rink at SSEC. Click on image to see the big picture:

Here's a picture from another program for comparison (thanks to Randy Horne):
Due to the current gap in continued funding from the U.S. National Science Foundation (NSF), the NSF Unidata Program Center has temporarily paused most operations. See NSF Unidata Pause in Most Operations for details.
CF geostationary projection now in CDM
29 August 2014
Hi John,
I've been trying out this support for geostationary projection, and I've had success with GOES data (which I specify in the grid variable with sweep\_angle\_axis = "x"). But I haven't had success with Himawari 8 data (sweep\_angle\_axis = "y"). It looks like the pixel locations are just a bit off when computed by the CDM grid mapped projection code. I haven't looked at the code, but do you know if the sweep angle axis functionality is implemented? It does't seem to matter which axis I specify in my Himawari data, I get the same issue.
If it's needed, I put a test dataset at the URL. It contains mostly empty data variables, except the l2p\_flags data which has a land mask as one of the flags. That's how I've been checking accuracy, by comparing against know coastline data. I made a PNG file in the same directory to give you an idea of what I'm seeing when I overlay coastline data, it's over New Zealand and I see an offset of about 5 pixels from the land mask data. It seems the issues in the projection occur diagonally far away from the satellite sub-point lat and lon.
Peter
Posted by Peter Hollemans on October 23, 2018 at 09:55 AM MDT #