Implementation of a vertical correlation function for the SYNOP two-meter temperature and relative humidity analysis.
Description:
The function accounts for the vertical distance between the observations and the model grid point, reducing the weight of the stations which are less representative of the model grid point elevation and increasing the weight of the stations closer to the model orography. This follows the approach done at Environment Canada and at Meteo-France in MESAN-SAFRAN.
The vertical correlation is expressed as a Gaussian function consistent with that used for snow depth analysis.
The branch also accounts for a few technical changes, such as a cleaned fetchobs for the use of IMS snow cover to account for the recent reprocessing that we archived on ECFS in a single place. It is also adapted to cope with cases where few SYNOP observations available before QC without any observation remaining after QC. A switch was introduced at the script level (L_VERT_CORR in ssaana) to open the possibility to switch off the usage of the vertical structure function. By default it is true which leads to use the vertical structure function.
Branch name: dap_CY42R1_t2m_vertical_struct_Lswitch
Control branch: dag_CY42R1_esuite
Note that this change only concerns observations for which the altitude is +/- 300m from the model grid point altitude. For more than 300m difference in the absolute value of the vertical distance between the model grid point and the observation, the observation is rejected, both in the control branch and in the revised branch.
Experiments:
1) Test the switch and BR when it is off:
Objective: Test that when we de-activate the use of the vertical structure function we get Bit-Identical results with the control branch from Gabor:
expid: gh9p (Tco639 dap_CY42R1_t2m_vertical_struct_Lswitch) with L_VERT_CORR=.false
ctrl: gh9h (Tco639 dag_CY42R1_esuite)
are bit-identical on 2014060100-2014060112
This shows that (i) the switch enables to switch off the use of the vertical structure function and (ii) the technical changes included in the branch are BI to the control dag branch.
2) Scientific evaluation when switch is on (use of vertical structure function):
CTRL: gh9h (Tco639 dag_CY42R1_esuite), EXPT: gh9i (Tco639 dap_CY42R1_t2m_vertical_struct_Lswitch) 20140601-20140930
CTRL: gh9q (Tco639 dag_CY42R1_esuite), EXPT: gh9r (Tco639 dap_CY42R1_t2m_vertical_struct_Lswitch) 20150601-20150930
For the EXPTs gh9i and gh9r the vertical correlation feedback was used (L_VERT_CORR=.true.) as specified by default in the script ssaana in /vol/ifs/sms
iver scores:
Winter (20150101 to 20150330): 3 months - expt still running:
file:///scratch/rd/dap/gh9q_gh9r/index.html
Summer (20140601 to 20140831) 3 months - expt still running:
file:///scratch/rd/dap/gh9h_gh9i/index.html
Impact summary:
- Neutral impact for upper air analysis (also neutral OBSTAT differences).
- Significant small impact very localised to near surface T2m (and to a lesser extent Tskin) and mainly in areas with orography (as expected):
Andes, Himalaya, Alpes, and ice capes: Greenland and Antarctica. - Similar impact on T2m for winter/summer, day/night.
Note that this change only concerns observations for which the altitude is +/- 300m from the model grid point elevation. Beyond 300m difference in vertical distance between th emodel grid point and the observation, the observation is rejected, both in the control and in the experiments.
Illustration of the near surface impact for winter (similar impact for summer, see iver links above) for Tskin and T2m:
Day (T+12) and Night (T+24):
