Quality Control for RADAR reflectivity Short report by
Quality Control for RADAR reflectivity Short report by Marián Jurašek
Challenges n Standard screening decision procedure reject too many observation and destroy the reflectivity profiles n Where to put decision about reflectivity rejection?
Too strong rejection n 1 st possible solution is to open screening rules for reflectivity Ø too dangerous n Need of more selective approach reflectivity profile consistency check Ø
Reflectivity profile consistency check n algorithm: if the single reflectivity didn’t pass classic rule check upper value in reflectivity profile Ø if we found reflectivity, which will pass standard rules and is not more far than Zconstant 2 from original reflectivity and all reflectivity in profile between original point are bigger than 0 and they are not far than Zconstant 2 of original value, we open screening rule by factor Zconstant 1. Ø if the check reflectivity will pass the new opened rule is considered as good, will be Ø
Reflectivity profile consistency check (2) n Remark: This check will not affect the rejection rule taking into account dynamic quality flag provided by “radar” people. n Results: After applying this new check we let pass around 5 -10% more observation the screening decision. n REMARK: All limits and constants should be object of investigation, when real assimilation will be ready, considering the impact on analysis.
Where put rejection? n We need two screenings in one: n for reflectivity itself n for the q, T profiles which will enter to minimization n Bad reflectivity should not enter to the baysienne inversion (creation of q, T profile)
Where put rejection? (2) n Solution (proposal): n Move REFLSIM from HOP to the HRETR before calling the Baysienne inversion, to let in HOP just part for q and T. n Move all checks for reflectivity from FGCHK and DECIS to the new subroutine REFLCHECK after calling REFLSIM and baysienne inversion, and let in FGCHK and DECIS just for q and T
Other results n First successful attempt to use more radars in screening: n n 4 radars: Arcis, Bordeaux, Falaise and Trappes “Nice” outcome: bug found in BATOR => problem solved n memory limits reached for such amount of data => thinning should be applied before processing data to the ODB (11 radars available online on cougar with more elevations) n
- Slides: 8