The inclusion of the observations with mixed surface type (water percentage, sfcpct<1) into GSI/NSST analysis

[XuLi-NOAA]

It has been found out some in situ observations are not assimilated since their mixed surface type, i.e., sfcpct < 1 (water percentage determined with the surface mask from background surface file) in the GSI/NSST foundation temperature analysis.

It happens near the coast areas, such as Pamlico Sound.

It is not an issue for the satellite observations, since the surface type dependent signal from the observation can be assimilated.

One solution to the inclusion of the in situ observations with mixed surface type has been proposed and is in test.

[XuLi-NOAA]

The observations from ships, coast buoys and others (GSI type 196, a type of moored buoy) are actually not used even if they pass all the Quality Control tests, because of a if block in setupsst.f90, where only a observation with dominant water surface type, see:
if ( isli == 0 ) then
nobs_qc = nobs_qc + 1
call intrp2a11(dsfct(1,1,ntguessfc),dsfct_obx,dlat,dlon,mype)
else
dsfct_obx = zero
endif
this means that a datum is used only if the surface type is completely water (dominated by water).
Along the coast, for a specific grid, the water percentage, sfcpct(0) is usually between 0 to 1.
In order to use the data without perfect water coverage, the code can be modified as follows:
if ( isli == 0 ) then => if ( owpct > 0.05_r_kind ) then (use the data if the water percentage > 5%).

[XuLi-NOAA]

A cycling run, oprcst10, starting from operational initial conditions at 2021102400, with the operational GFS version plus the NSST package and the above change on the inclusion of the mixed surface type in situ sea water observations, has been running. about 14-day has been done and the too warm NSST Tf analysis over the Pamlico Sounds area are gone. The more critical period, around the middle of November of 2021 is coming.
One concern is over the coast area southern to Pamlico Sound, the Tf analysis is still warmer than OSTIA by about 2 K and a little bit warmer than the operational. The preliminary diagnosis suggested that the issue seems due to the too tight QC for the in situ observations. A test will be done.

[XuLi-NOAA]

A more detailed diagnosis has shown that the threshold of (O-B)/oberr is too tight (corresponding O - B 1.5 K), and it is the cause some coast in situ data are rejected and led to too warm NSST analysis near the coast of Pamlico Sound and nearby areas. The test with the looser QC (corresponding O - B becomes 3.0 to 3.5 K) shows the too warm issue (in a dramatically cooling period).