GEOS-ESM / swell

Workflow system for coupled data assimilation applications
https://geos-esm.github.io/swell/
Apache License 2.0
14 stars 4 forks source link

Surface pressure Hofx in Swell test suites #394

Open gmao-jjin3 opened 3 months ago

gmao-jjin3 commented 3 months ago

Compare JEDI hofX with GSI hofX when FV3-JEDI is not involved and is involved.

gmao-jjin3 commented 3 months ago

GSI hofx is reproduced by JEDI/UFO in the swell-ufo_testing suite. FV3-JEDI is not involved. stationPressure_i_hofxdiffbcQc_vs_gsihofxbcQc

gmao-jjin3 commented 3 months ago

There are large differences between FV3-JEDI hofx and GSI hofx in the swell-hofx suite. gsi_hofx_vs_jedi_hofx_diff_sfc_stationPressure

gmao-jjin3 commented 3 months ago

There are large differences between "surface_pressure" geoval made by GSI and those made by FV3-JEDI. See figures at https://github.com/GEOS-ESM/swell/issues/241#issuecomment-2248887905

gmao-jjin3 commented 3 months ago

Here are Hofx differences between GSI and JEDI versus latitude. The JEDI hofx is the hofx0 produced by the swell-3dfget suite with only surface station data (sfc). gsi_hofx0_vs_jedi_hofx0_diff_vs_Lat_sfc_stationPressure

gmao-jjin3 commented 3 months ago

Here are Jo differences between GSI and JEDI versus latitude. Results come from the swell-3dfget suite with only surface station data (sfc). Jo_diff_vs_Lat_sfc_stationPressure

rtodling commented 2 months ago

I've done experiments w/ PS only observations in GSI and JEDI:

Here is the increment from GSI (top) and the difference between the JEDI corresponding increment and GSI's:

psOnly-psInc

The counts of obs used are: GSI: 105,638 JEDI: 105,520

The total Jo's and Jo/n are quite close.

it should be noted that the JEDI analysis above used SWELL's setting pointing to x0049 which has c90 backgrounds. I will redo the JEDI analysis w/ backgrounds at c180 which is consistent w/ the experiment I showed before I travel. I believe results will be even closer.

I think this corroborate my suspicion that the diff in PS increments we see in the exp I discussed are due to some other observing system; but not directly the PS observations.

rtodling commented 2 months ago

Just for the record, this is a similar test when using only scatterometer winds - this show the corresponding (multivariate) GSI ps increment (top) and the difference of the corresponding JEDI ps increment w/ GSI's:

scatwOb-psinc

This is pretty good agreement.

rtodling commented 2 months ago

I am coming to think the issue in the surface pressure increment is associated to how JEDI and GSI difference when assimilating clear-sky MW observations.

The following figure - again showing GSI increment on top, and difference of JEDI increment w/ that of GSI on the bottom - is the result of runs done w/ GSI (576x361) and JEDI (c180) when only AMSUA and ATMS observations are used.

psinc-clearMWob

yes, the differences in the increments (bottom) are nearly as large as the increment itself.

rtodling commented 2 months ago

I see also considerably large differences in the stratosphere - in the experiment above when only clear-sky MW observations are used. Here is the increment and the difference in increments for temperature at 1 mb.

tinc_1mb_clearMW

But this has no correlation to what's happening near the surface.

gmao-jjin3 commented 2 months ago

Hofx are different for ATMS_N20 in the stratosphere. Here are comparison for its channel 14 made by swell-3dfget with ATMS_N20 data only. Sensitivity of this channel peaks at ~2 hPa. Surface characteristics in BKG should have little impact on its Hofx.

Jo_diff_vs_Lat_atms_n20_brightnessTemperature14 gsi_hofx0_vs_jedi_hofx0_diff_vs_Lat_atms_n20_brightnessTemperature14 ObsError_vs_latitude_atms_n20_brightnessTemperature14 rmsres_difference_atms_n20_brightnessTemperature14

gmao-jjin3 commented 2 months ago

Large differences in hofx at moisture channels.

gsi_hofx0_vs_jedi_hofx0_diff_vs_Lat_atms_n20_brightnessTemperature17 gsi_hofx0_vs_jedi_hofx0_diff_vs_Lat_atms_n20_brightnessTemperature18 gsi_hofx0_vs_jedi_hofx0_diff_vs_Lat_atms_n20_brightnessTemperature19 gsi_hofx0_vs_jedi_hofx0_diff_vs_Lat_atms_n20_brightnessTemperature20 gsi_hofx0_vs_jedi_hofx0_diff_vs_Lat_atms_n20_brightnessTemperature21 gsi_hofx0_vs_jedi_hofx0_diff_vs_Lat_atms_n20_brightnessTemperature22