guoqing-noaa
commented
4 months ago @HuiLiu-NOAA Thanks for testing and documenting this! We would prefer to use a local lapse rate instead of of a constant lapse rate to do height difference adjustment. For regional applications, local lapse rate gives us a better results. Stan has a conference presentation talking about this. I will try to see whether I can find this.
For temperature, we adjust the observation not hofx . We bring the observation to the model terrain height and use 2mT as the background. So here we need both the height difference adjustment and 2mT as the background at the same time to get a reasonable result.
For 2mTd, we don't need to do any adjustments on observations, but we have to use the 2mTd as background to get a reasonable result.
HuiLiu-NOAA
spanNOAA
In a recent RDAS test, a warm bias (~1K) is seen in the OmB of ADPUPA (Type 181/187) surface temperature (T2m) observations using the vertical interpolation operator ("Vertinterp"), RDAS issue: https://github.com/NOAA-EMC/RDASApp/issues/107. This bias seems related to many T2m observations below the model surface level. The average height difference is ~20-100 meter.
For those T2m OBS below model surface level, however, the "Vertinterp" operator does not extrapolate from model surface level down to OBS height; instead, it directly uses the values from model surface level as simulated ones at the OBS height. This could lead to a cold bias in the simulations; thus a warm bias in the OmB. A few available empirical terrain height corrections may help study and reduce the bias issue.
The initial implementation and testing of the corrections in UFO may include two options: 1). UKMO scheme Reference: Ingleby,2013. UKMO Technical Report No: 582, Appendix 1.
To avoid impact of diurnal/local variations on the correction, integrate the hydro-static equation to obtain temperature at model surface height from model temperature at 2000 m; then extrapolate to the OBS height. The constant lapse rate (0.65 K/100m) is used in the calculation, which is consistent to the other UFO operator "SfcPCorrection" for surface pressure data assimilation. This would ensure no potential conflicts when both surface temperature and pressure observations are assimilated together.
2). GSL scheme Reference: gsi/gsd_terrain_match_surfTobs.f90 https://agu.confex.com/agu/fm22/meetingapp.cgi/Paper/1168054 First, calculate temp lapse rate between model surface level and level 5. Then use the height difference between station and model surface level to extrapolate the model simulation from model surface to OBS height, and then bring the observed value to model surface level. Another recent choice is to directly use T2m from background to obtain the simulation at OBS height.
We plan to implement these corrections into a new UFO SfcTQCorrected class and test these corrections in understanding and fixing the bias issue in the RDAS test case.
Comments and suggestions are very welcome. Thanks.
After the new UFO terrain height corrections for T2m are done, we plan to implement and developing UFO corrections for Q2m observations as well.