Workshop of the COSMO priority project on
verification of very high resolution models
MeteoSwiss, Zurich, 23 May 2008
Participants:
- Pierre Eckert (chair and minutes)
- Pirmin Kaufmann (MeteoSwiss)
- Ulrich Damrath (DWD), Felix Ament (ZMAW, Univ. Hamburg)
- Adriano Raspanti (USAM), Maria-Stefania Tesini (ARPA-SIM),
- Elena Oberto (ARPA Piemonte)
- Rodica Dumitrache (NMA)
- Joanna Linkowska (IMGW)
Pierre Eckert: introduction
Get pdf document (130Kb)
P. Eckert first reminds that the end of the project is planed for September 2008 and that the
final presentations will be given at the COSMO general meeting in Cracow. The final report should
also be available until then.
A first workshop of the project has taken place in April 2007 in which it already has been noticed
that the optimal scale for fuzzy precipitation verification is many times the length of the grid.
A certain amount of properties of good scores have also been proposed.
A mini-Workshop has also taken place in December 2007 with Beth Ebert in which some "recommendations"
have been proposed: Avoid "leaking" scores, use illustrative and understandable scores. Use at least
upscaling and fraction skill score which also can be related to products: Upscaling σ regional
mean, Fraction skill score σ probability to exceed some threshold in neighbourhood. It also has
been recognized that on quite large scale (30-50 km) 2 km models are"better" than 7km model
Felix Ament: Precipitation verification during MAP D-PHASE.
get pdf document (600Kb)
All the models involved in MAP D-PHASE have been evaluated on the catchments in Switzerland. The
radar estimations have been calibrated with the existing rain gauges. Following conclusions can
be drawn:
- Observational error is considerably large.
- Quality of precipitation warnings is poor (D-PHASE set-up: ~50% POD, ~50% FAR).
- However COSMO models give very good results compared to other D-PHASE models.
On the other hand, a new fuzzy verification of the DOP has been carried out. Following
conclusions have been found:
- Results of Upscaling and Fraction skill score are reasonable
- Results are surprisingly stable, even on a monthly bases
- Overall better results for high resolution models
- This benefits shrinks slowly, if large lead times or accumulation periods are considered
- Great variation from month to month – results depend strongly on the weather during the
verification episode
Adriano Raspanti: Verification of COSMO ME vsCOSMO I2
Get pdf document (300Kb)
- C-IT tends to be noisier than C-ME, but when the threshold increases, it behaves better.
- In winter the performance of C-IT seems to be generally worse.
- Cosmo-IT upscaled tends to have a better behaviour than Cosmo-Me for "middle" thresholds,
worse for very low ones and almost the same for very high thresholds.
Ulrich Damrath: Verification of COSMO-EU and COSMO-DE using RADAR and SYNOP data
Get pdf document (2.8Mb)
Data from GME (resolution i192f ~ 40 km), COSMO-EU and COSMO-DE on the grid of COSMO-DE. The
verification has been brought to an operational stage. Some conclusions:
- Verification against SYNOP observations shows in most cases clear
signals that favorite COSMO-DE against COSMO-EU, especially for wind.
- Advantages of COSMO-DE occur in most cases at stations where
small scale variability is evident in observations.
- COSMO-DE is better in convective situations.
Pirmin Kaufmann: Neighborhood shapes and sizes for COSMO-2
Get pdf document ( 2.5Mb)
Shapes given by circles in space and a fixed interval in time (cylinder). Linearly fading weights.
- Proposed setting for COSMO-2: rxy=10, rf=10, rt=1h. The effective radius is 15 grid points=33 km.
Ulrich Damrath: Current configuration of the neighborhood method for COSMO-DE at DWD and some future plans.
Get pdf document ( kb)
Current configuration: horizontal search radius: 10 GP, temporal search radius: 1 hour (time steps:
T-1 h, T, T+1h). The system is planed to be used in an automated waning procedure (autowarn).
Following parameters will be predicted: Frost, Heavy rain, Extreme rain, Gusts, Heavy gusts, Extreme
gusts, Storm like gusts, Storm gusts, Extreme storm gusts, Thunderstorm, Heavy thunderstorm, Extreme
thunderstorms, Black ice. No verification is available yet.
Recommended properties of scores
The properties of scores proposed in 2007 have been reviewed. No great change needs
to be done and they read:
- Theoretical properties
- score should be illustrative and understandable
- score should give a real information of forecast quality
on the different scales
- monotonic behavior concerning
- scale (best values for large scales)
- frequency of occurrence (best values for high frequencies of occurrence)
- Represent some significant characteristics of the PDF (obs and forecast)
- Murphy’s properties of scores
Bias, Accuracy, Skill, Reliability, Resolution, Sharpness, Discrimination, Uncertainty
- Practical properties
- agreement between subjective and objective judgment
- possible help in decision making
- Corresponds to the needs of the user
- Should be able to provide a comparison between 2km and 7 km models
- Should not use a matching between prediction and observation because it would not allow
the generation of univocal products
Writing of the final report
Pierre Eckert will write the introduction and a synthesis of the conclusions. Each individual worker
will contribute with a chapter limited to 2-5 pages with figures. The accent will be put on strengths
and weaknesses of each score applied on very small scale models. Felix Ament, Pirmin Kaufmann,
Ulrich Damrath, Maria-Stefania Tesini (with Carlo Cacciamani), Rodica Dumitrache, Joanna Linkowska and
eventually Adriano Raspanti took a commitment to write such a text.
The texts are expected for the end of August 2008. Pierre Eckert will do the final editing.