05.08.2005 - 1 -

COSMO General MeetingZürich, 20 - 23 Sept 2005 1christoph.schraff@dwd.de

Christoph Schraffchristoph.schraff@dwd.de

Revision of Quality Control for Radiosonde Humidity

24-h precipitation analysisfor 1 August 2004 , 6 UTC

relative humidity profile at Stuttgart31 July 17 UTC

LM ass without use of profileref LM assobs

Motivation

‘ref’ LM assimilationopr. humidity QC setting

Stuttgart

> 100 mm

05.08.2005 - 2 -

COSMO General MeetingZürich, 20 - 23 Sept 2005 2christoph.schraff@dwd.de

threshold quality control for individual observations (first guess check)

• relative humidity observation RHo rejected if : | RHo – RHb | > RHthr

z

T

crit To

k

Tok+1

sT

• further revision to increase threshold in stable conditions and in particular at inversions where model errors are known to be larger ( limit RHthr(3) (to)

≤ 70 % )

b b * (1 + fstable + finvers ) , where

for lapse rate > crit = – 0.0065 K/m :

fstable = ¼ * (1 + sT /(1+sT)) * (1 – min{,0} / crit )

• operational (‘ref’) : RHthr (to) = 70 % , RHthr (t[h]) = RHthr (to) * (1 + 0.1* | to – t | )

• revision (analogous to revised GME scheme):

RHthr(1,3) (to) = min { (o2

+ b2

) ½

, 2 b } * cflag(1,3) , where

observation error o = 10 % ( 15% for To < 233K , 20% for RHo < 20% )

background error b = 10 % ( 15% south of 30 N )

constant cflag(3) = 3.1 ( cflag(1) = 1.8 for flag 1 as used in multi-level check )

for lapse rate > 0 : finvers = 0.2 * Tinvers * (1 + min{2, / 0.05[K/m]} )

both factors increase with increasing stability and with increasing T

05.08.2005 - 3 -

COSMO General MeetingZürich, 20 - 23 Sept 2005 3christoph.schraff@dwd.de

multi-level check

• define ‘analysis layers’: as standard layers, except p = 50 hPa below 700 hPa p = 25 hPa below 800 hPa

(boundaries of standard layers: ps , 925, 775, 600, 450,

• criterion: if 4 or all consecutive standard layers contain observations with flag > 0 ,set these standard layers to rejected

• follow-up criterion: each analysis layer within these standard layers are rejectedif they contain observations with flag > 0

all observations within these rejected analysis layers are rejected

05.08.2005 - 4 -

COSMO General MeetingZürich, 20 - 23 Sept 2005 4christoph.schraff@dwd.de

spatial consistency check of integrated water vapour

• derive observation increments of IWV from radiosonde humidity profiles and optionally from ground-based GPS ZPD data

• at each observation location, an‘analysis increment’ is computedfrom neighbouring observations

1,max

),(2

2

kj kj

jjkj kjaik

w

tIWVIWVwIWV

x

, kaithr

aikkk IWVIWVtxIWVIWV

• analysis increments used to correct model background value to obtain a better estimateof ‘truth’ revised first guess check

many observations used for analysis increment better estimate of truth decrease threshold

large analysis increment = disagreement of model and obs large uncertainty increase threshold

aikj kjothro

aithr

k IWVwtIWVtIWV 1,2.0min2.01)()( 2

satmoothr IWVtIWV 15.01)( [mm]

• no neighbouring observations (no GPS) : first guess check on IWV

many neighbouring observations : checks consistency of current obs to them

05.08.2005 - 5 -

COSMO General MeetingZürich, 20 - 23 Sept 2005 5christoph.schraff@dwd.de

why all this ?

• if threshold RHthr too small too many good observations near inversions are rejected

13 Feb 2003, 0 UTC

(old) GME QC thresholds

too tight → GME thresholds revised

larger stability-dependent thresholds

thresholds too small (no IWV-QC)

new QC incl. IWV, larger thresh.

ok

LM ass., opr. QC (ref)

01:53

NOAA IR

05.08.2005 - 6 -

COSMO General MeetingZürich, 20 - 23 Sept 2005 6christoph.schraff@dwd.de

results, low stratus period

• low-stratus period 8 – 13 Feb 2003, assimilation cycle with forecasts (GPS used / not used)

- multi-level check rejects 4 % of profiles (completely or partly, above ps … 700 hPa)

- IWV check rejects 1 % of profiles (completely)

- << 1 % of remaining individual observations rejected by first guess check

- > 80 % of rejected data are close to lateral domain boundaries (within ~ 50 grid points)

boundary fields too moist (GME-OI)

background field too moist rejection of (good) profile

new QCref QC

obs

12 UTC

Warsaw12 Feb 0 UTC

new QCref QC

obs

13 Feb 2003, 12 UTC

Warsaw boundary

east-west vertical cross sectionLM-ass. using Warsaw sounding

12 Feb 2003, 0 UTC

05.08.2005 - 7 -

COSMO General MeetingZürich, 20 - 23 Sept 2005 7christoph.schraff@dwd.de

Stuttgart30 July 23 UTC

new QCref QC

obs

Stuttgart19 July 23 UTC

new QCref QC

obs

24-h precipitation 30 – 31 July 2004, 6 UTC

> 100 mm

humidity profile rejected by spatial consistency check of IWV

> 100 mm

LM ass, ref QC LM ass, new QCobserved precipitation

24-h precipitation 19 – 20 July 2004, 6 UTC

humidity profile above 925 hPa rejected by first guess / multi-level check

05.08.2005 - 8 -

COSMO General MeetingZürich, 20 - 23 Sept 2005 8christoph.schraff@dwd.de

Stuttgart22 July 23 UTC

new QCref QC

obs

24-h precipitation 22 – 23 July 2004, 6 UTC

humidity profile not rejected !

Stuttgart23 July 17 UTC

new QCref QCobs

24-h precipitation 23 – 24 July 2004, 6 UTC

humidity profile rejected (only) above 850 hPa by first guess / multi-level check

observed precipitation LM ass, ref QC

75 mm

LM ass, new QC

50 mm

20 mm 30 mm

05.08.2005 - 9 -

COSMO General MeetingZürich, 20 - 23 Sept 2005 9christoph.schraff@dwd.de

summary

• introduced smaller, stability-dependent thresholds for radiosonde humidity

• adjusted multi-level check, rejects 2 % (summer period) to 4 % (low-stratus winter period) of humidity profiles

• introduced IWV spatial consistency check (both for radiosonde and GPS data; without dense GPS data IWV first guess check), rejects 1 % of profiles

• QC accepts almost all good data near inversions

• rejects most erroneous profiles of radiosonde Stuttgart (in particular the ones with worst impact)

• will go into next LM version and become operational