Inconsistent masks for areas for 850mb plot

[chengzhuzhang]

Reported by @PeterCaldwell
"the bias calculation for 850mb seems to use inconsistent masks for areas where 850mb is below the land surface, causing rings of massive bias values. See:

http://portal.nersc.gov/project/e3sm/coupled/tuning_tests/20180410.A_WCYCL1950_HR.ne120_oRRS18v3_ICG.theta/e3sm_diags_yr6-10/viewer/latitude-longitude/merra2/t-850mb-global-merra2/ann.html
"

"I’m using this file to regrid: map_ne120np4_to_181x360_aave.nc"

[chengzhuzhang]

After reviewing some test data, we found that the inconsistent mask only occur when conservative method is applied when to regrid higher res data (obs data in this case) to low res data (model data) before comparing.
Results using conservative method,
https://acme-viewer.llnl.gov/shaheen2/all_sets_cdat8_esmf_7.1.0r-TRY2/lat_lon/MERRA2/MERRA2-T-850-ANN-global.png
Results using bilinear regridding method,
https://acme-viewer.llnl.gov/shaheen2/all_sets_cdat8_esmf_7.1.0r-linear/lat_lon/MERRA2/MERRA2-T-850-ANN-global.png

The treatment of masking in conservative method should plays a role in this issue. @dnadeau4 would you maybe comment on this issue? Thanks!

[dnadeau4]

@taylor13 and I are working on a solution for conservative masking. The one I implemented from NCAR does not satisfied your need. (BTW I cannot access these pictures...)

[chengzhuzhang]

Thank you denis! here is the access info to the pics..
Username: acme
Password: AwesomeModel

[PeterCaldwell]

@chengzhuzhang - thanks for working on this! I'm watching but don't have anything to contribute right now. Let me know if there's anything specific you'd like me to do.