Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology MeteoSwiss Verification results of COSMO at MeteoSwiss in the year 2008 and experiences with CVS COSMO General Meeting 2008 WG5 Parallel Session 15 September 2008
Jan 01, 2016
Federal Department of Home Affairs FDHAFederal Office of Meteorology and Climatology MeteoSwiss
Verification results of COSMOat MeteoSwiss in the year 2008
and experiences with CVS
COSMO General Meeting 2008
WG5 Parallel Session 15 September 2008
Federal Department of Home Affairs FDHAFederal Office of Meteorology and Climatology MeteoSwiss
Experiences with CVS
Pirmin KaufmannMeteoSwiss
3 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
Installation hints
• Installation does not work without trailing slash in INSTALLATION_PATH
• For 64 bit EMOS, insert –Dbit64 after pgf90 in all make scripts: make_*_linux.sh
• Newer EMOS libraries need JSEC2=4096 instead of JSEC2=64 in *vs_bufr_decode.F
• Large BUFR files need DO ILOOP =1,500000 instead of DO ILOOP =1,50000 in gvs_bufr_decode.F
• Example files and documentation are sometimes inconsistent, e.g.• -UTC missing in example file fcs_date.par• UWIND ,VWIND instead of WMOD, WDIR and vice versa
4 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
Running CVS
• BUFR files for GVS must not contain any reports with another date, e.g. no 00 UTC reports from following day
• BUFR files for HVS must contain 23 UTC sounding from previous day
• GRIB files for GVS must not contain any other fields with the same GRIB element number, e.g. no T, U, V on other levels
• GRIB files for HVS must contain T, U, V on all required pressure levels
5 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
Fixed Bugs
• Height correction for temperature was applied with wrong sign
• Grid points with land fraction < 1.0 (later 0.55) are considered water points and excluded from the association of the grid points to stations. Correct value now 0.5.
• Irrelevant for practical application: The conversion of K to deg C was done with the subtraction of 273.16 (triple point) instead of 273.15 (freezing point).
6 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
Remaining Bugs
• HVS calculates only out to +66 h lead time, all scores for the +72 h lead time are written as 0.00
• Grid point selection is neither old nor new standard as decided last GM in 2007
7 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
Desirable Features
• GVS extraction of obs data should control date of BUFR message, same for GRIB extraction
• HVS extraction of obs data should look for soundings with 23 UTC reference time in file of previous day
• Test consistency of extracted and verified lead times• Currently bogus +54 h to +66 h scores are calculated without
warning even when fcs data have only been extracted out to +48 h
• Test consistency of land mask and orography with model GRIB files
• Remove write(555,*) commands in gvs_grib_decode.F
Federal Department of Home Affairs FDHAFederal Office of Meteorology and Climatology MeteoSwiss
Verification results of COSMOat MeteoSwiss in the year 2008
Francis Schubiger, Pirmin Kaufmann,Christophe Hug, Vanessa Stauch
MeteoSwissCOSMO General Meeting 2008
WG5 Parallel Session 15 September 2008
9 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
• Operational:• Surface (3h-steps SYNOP and 1h-steps SMN [Swiss Met Net])
• psred, ps, T2m, Td2m, 10m-wind, cloud cover, precipitation and since Spring 2008 also wind gusts
• Upper-air (TEMP)• T, RH, wind
• Ongoing in experimental modus:• Windprofiler Payerne and Schaffhausen• Precipitation with RADAR (fuzzy verification, Ebert’s package)• Fluxes at Payerne
• New / Next developments:• Global radiation at SYNOP stations• VAD (locations of swiss radars)
• Phased out:• integrated water vapor content with GPS data• weather-type dependant verification with RADAR and TEMP
=> All these packages should be integrated in VERSUS 2
COSMO: Verification packagesfor COSMO-7 and COSMO-2
10 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
Surface pressure bias (mean error)
in winter 07/08 hig positive bias in Spring 08 high negative biasbut observed pressure in Spring 08 is 9.6 hPa lower than Winter 07/08see results of U. Damrath: low ps (obs) -> negative bias
high ps (obs) -> positive bais
Winter 07/08 Spring 08
11 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
2m temperature diagnostics - Impact study: COSMO-7 forecast of 08.05.08
Zermatt1600m
Gd St Bernard2400m
EvoleneValais 1800m
12 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
2m temperature diagnosticsimpact on mean diurnal cycle
Summer 2007 Summer 2008
The (well known) errors of:- too strong temperature increase in the morning- maxima reached ~ 1.5-2 h too earlyis removed with the new 2m temperature diagnostics (introduced operationally 09.06.08)
13 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
2m temperature over Switzerlandbias Summer 2008 (CVS-Software)
300-800m > 800m
14 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
2m dewpoint (mean error)
Spring 08 Summer 08
Spring 08: bimodal pattern, negative (dry) bias along coasts, positive (wet) bias at inland stations
Summer 08: the dry bias discontinues the strongly increased wet bias of the summers 2006 and 2007
15 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
Other results in 2008:Precipitation, windspeed, cloudiness• precipitation:
• extreme high positive bias at the borders of the model (N and E) since Autumn 2006, disappeared in Spring 2008
• since Winter 07/08 less precipitation, especially the strong amounts are now underestimated (due to new microphysics and Runge-Kutta: introduced opr in November 2007 and in testchain from 14.7.-31.08.07) Summer 2007 2008 change
opr testchain opr 07 > 08mean obs 1.45 1.29 1.28 -12%mean mod 1.69 1.04 0.95 -44%fbi(1mm/12h) 1.16 1.00 0.93fbi(5mm/12h) 1.02 0.73 0.66fbi(10mm/12h) 1.00 0.64 0.55fbi(50mm/12h) 2.16 0.63 0.52
• Windspeed: negative bias along the coast and on mountains, positive bias inland
• Cloudiness: partial coverage underestimated
16 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
Precipitation over Switerland Spring 2008Frequency bias (12h-sums) of CVS-software
17 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
Precipitation over Switerland Summer 2008Frequency bias (12h-sums) of CVS-software
18 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
Comparison COSMO-2 and COSMO-710m windspeed: standard deviation of errors Spring 2008 over the COSMO-2 domain
about 400 stations
Forecast lenght
ST
D (
m/s
)
COSMO-7COSMO-2
19 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
• stronger over-estimation especially in narrow valleys
• over Swiss Middleland lower overestimation (or even low under- estimation)
• Lower understimation over mountain stations
• Standard deviation of errors lower (better)
COSMO-2
COSMO-7
Comparison COSMO-2 and COSMO-710m windspeed Winter 07/08
20 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
COSMO-2
COSMO-7
12h - forecasts
STD (m/s)
Dru
ck (
hPa)
Comparison COSMO-2 and COSMO-7Verification Payerne soundings:windspeed (standard deviation of errors)
December 07 / January 08
21 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
windgust: old vs new for 16.01.-17.03.08
22 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
windgust: old vs new for 16.01.-17.03.08
23 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
windgust: old vs new for 16.01.-17.03.08over Switzerland
new old
25 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
Windgust: number of obs Summer 2008
26 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
Windprofiler vs. COSMO-2 Wind speed (ff) on 2008-08-19
WindprofilerSchaffhausen assimilated
28 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
Verification global radiation: Kloten
29 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
Global radiation at Kloten 12 UTCCOSMO DMO vs error in 2006
30 Verification results of COSMO at MeteoSwiss in 2008 and experiences with CVSWG5 Parallel Session of COSMO GM, 15.09.2008
Verification global radiation: Kloten