Problem found in processing the observation at the domain boundary

[xyzemc]

When try to assimilate goes_16 ABI radiance data in RRFS CONUS domain, there are error messages like:

61: CRTM_Surface_IsCoverageValid(INFORMATION) : Total coverage fraction does not sum to 1 +/- 1.000000000000000E-06 0.8427 78: CRTM_Surface_IsCoverageValid(INFORMATION) : Total coverage fraction does not sum to 1 +/- 1.000000000000000E-06 0.3020 70: CRTM_Surface_IsCoverageValid(INFORMATION) : Total coverage fraction does not sum to 1 +/- 1.000000000000000E-06 0.5450 64: CRTM_Surface_IsCoverageValid(INFORMATION) : Total coverage fraction does not sum to 1 +/- 1.000000000000000E-06 0.8539 0: CRTM_Surface_IsCoverageValid(INFORMATION) : Total coverage fraction does not sum to 1 +/- 1.000000000000000E-06 0.1657 59: CRTM_Surface_IsCoverageValid(INFORMATION) : Total coverage fraction does not sum to 1 +/- 1.000000000000000E-06 0.7820 42: CRTM_Surface_IsCoverageValid(INFORMATION) : Total coverage fraction does not sum to 1 +/- 1.000000000000000E-06 0.4165 14: CRTM_Surface_IsCoverageValid(INFORMATION) : Total coverage fraction does not sum to 1 +/- 1.000000000000000E-06 0.8745 46: CRTM_Surface_IsCoverageValid(INFORMATION) : Total coverage fraction does not sum to 1 +/- 1.000000000000000E-06 0.9874

After debugging, I found all those points are on the right side of domain edge. See plots below: Surface total fraction: land+water+ice+snow

Similar issue happened to other input variables required by CRTM: Air_temperature has 0 values on those points:

[TingLei-daprediction]

@xyzemc Would you try this draft PR https://github.com/JCSDA-internal/fv3-jedi/pull/1096 to see if it could avoid this issue? Or you can directly use the built branch on hera: /scratch2/NCEPDEV/fv3-cam/Ting.Lei/dr-jedi-bundle/jedi-bundle/build/bin/?

[xyzemc]

This issue has been solved.