Minutes of the meeting of COSMO Working Groups 4 and 5

Bologna, 10-11 April 2001

Contents

In this page you will find overwiews and download material of the COSMO Working Groups 4 and 5 meeting, in Bologna, Italy, at 10-11 April 2001.

• Presentations
• Discussion of WG4
• Discussion of WG5
• Plenary discussion
• Agenda
• Note for WG5

Presentations

The meeting started at 9.00 of 10/4/2001 with the presentations of the workpackages of the Working Groups 4: Application of LM, chaired by Pierre Eckert of MeteoSwiss.

Six talks (see attached agenda) have been given. The first three talks concerned the ARPA-SMR activities in the field of Limited Area Ensemble Predictions (LEPS). Andrea Montani (ARPA-SMR) and Chiara Marsigli (University of Bologna) presented methodology, case studies results and probabilistic evaluation of LEPS, Tiziana Paccagnella (ARPA-SMR) presented the opportunities to start an operational chain of COSMO-LEPS, based on LM, to be implemented at ECMWF.

Pierre Eckert (Meteoswiss) discussed the results of the use of Kalman filtering technique applied to SM model and about the use of the operational use of ECMWF-EPS, based on a neuronal approach for weather classification.

Norbert Raderschall (Swiss federal institute for snow and avalanche research) presented an interesting study about the downscaling of weather elements directed produced by SM. The study of systematic error of DMO and of correlation of different parameter against observation (essentially temperature and precipitation) allow to hypothesise different approach for statistical downscaling (or dynamical/statistical post processing) of DMO in order to optimise the use of these products in the weather offices.

Finally, the last talk of the WG4 was of U. Damrath of DWD who presented an interesting study about the statistical interpretation of LM. In particular the discussion on the best use of QPF was focused on the use of some new probabilistic quantities (quantiles) deduced by the distribution occurred.

In the afternoon of the same day the presentations of the WG5 (verification and case studies) started. The first talk was done by Ulrich Damrath (DWD) who presented the results of operational verification of LM at DWD. The main attention has been given to the verification of surface parameters (2mt T., precipitation, rel. Hum., 10mwind, cloud cov.) as regards the spatial and time distribution of errors.

A similar talk has been presented by Francis Schubiger (Meteoswiss) who showed the results of operational verification of LM done. Detailed results of the diurnal cycles of the errors were shown. Francis gave also a quick view of the interesting activities at Meteo Swiss about the verification of precipitation against radar data (work done by Andrea Rossa), of cloud coverage using satellite data (Zelenka) and of a verification for different weather types.

After the coffee Massimo Ferri (Italian Weather Service, UMG) showed the results of verification of LM precipitation during the MAP period that was done using the high resolution data set ITAMAP 1.2 of non-gts data available in the Northern Italy for the MAP period (3 month of hourly data on more than 250 station available at the MAP web site).

Federico Grazzini (ARPA-SMR) presented the results of LM verification in Italy using all the Italian synop data for the period December-March 2001. The main parameters have been investigated as regards the diurnal cycle of the errors for four different classes: island, coastal, lowland and mountain areas.

Enrico Minguzzi presented a very interesting talks about the results of LM precipitation forecasts used in Piedmont Region on Northern Italy. The very good non-gts data set available in that region (195 raingauges used in a area of less than 30.000 Km2) allowed a detailed verification of LM QPF inside all the hydrographic basin of the region. Comparison of scores of ECMWF and LM models have been shown.

Finally the last talk of the WG5 (and of the day) was shared between Stefano Gallino and Giulio Contri (CMIRL-Liguria and University of Genova). Stefano showed the state of advancements of the study on objective clouds classification using IR and visible meteosat data. Giulio showed interesting results of an exercise of inter-comparison between different LAM QPF forecasts (Bolam, LM and ECMWF models) operating during the MAP field phase.

The second day (11 April), parallel sections of WG4 and WG5 started at 9.00 and then followed a common discussion.

Back to top

Discussion of WG4

We discussed the way to implement LEPS at ECMWF and decided to proceed in steps starting by opening a common account using units from the COSMO members. Then the IFS-LM should be implemented with a resolution of 10 km and on a domain covering the four members' areas. The MAP cases will be studied afterwards. Based on encouraging results, an operational suite will be implemented. MeteoSwiss will also provide comparison point based on a statistical method applied directly on the ECMWF EPS.

Various postprocessing methods will be tried including MOS, Kalman filtering, computation of instability indices,... A questionnaire will be established by Pierre Eckert to list the type of postprocessing used or wished by the members. These postprocessing can be internal (temperature on a pressure level, PV,..) or external (MOS, filtering,...).

It has been decided that the WG4 will give guidelines on the type of parameters which should be presented to the forecasters and how to interpret them. The graphical presentation should be left to the WG6.

According to these decisions the work plan has been updated and can be consulted on the COSMO web site.

Back to top

Discussion of WG5

During the discussion it was debated about some common results that can be advanced about the LM performances.

Temperature

As regards 2m temperature (T2m), LM seems to be "too cold". Negative bias are evident during nightime, especially in winter. In the Alpine area this negative bias is evident also at upper level (into the whole PBL and even more).

It's evident also a diurnal cycle of the error: during the day LM performs better than during the night.

Humidity

LM seems to be "too dry". Negative bias in 2m dewpoint temperature (Td2m) are observed, especially during the night and more during winter.

In particular the negative bias in Td2m is higher (in absolute value) that the negative bias of 2mt. Temperature. The new scheme of PBL parameterisation should overcome this problem. The impact will be checked in future.

The vertical structure of relative humidity doesn't show a big negative bias, but that can be a result of "compensating errors" of T2m and Td2m errors.

Precipitation

In general LM seem to underestimate high values of precipitation and overestimate small values (in Switzerland and Germany). Anyway LM is able to produce somewhere in the integration domain very high (and non only isolated grid points) values of precipitation (especially near the orography);

The spatial distribution of precipitation is not correct. Much more precipitation is produced upstream the orography, less downstream. Forecasts with a filtered orography (operational at DWD and MeteoSwiss) and/or spatial averaged values seem to give more correct results.

By comparing LM QPF with the simulations of global models (ECMWF) it is clear that better results are observed for high thresholds of precipitation from LM. ECMWF performs better as regards small amount of rain.

Some hints concerning the best use of LM QPF came out:

• In general spatial averaged values (50x50 Km2) are quite good. Better results for higher values of precipitation
• Hourly precipitation are not usable. It is necessary to accumulate precipitation in time window of (at least) 3-6 hours to have realistic forecast

Information

Verification results of the different work packages must be sent to the WG coordinator C. Cacciamani every 3 months (DEC, MAR, JUN, SEP). Text in html or ASCII plain text. Figures in GIF/JPEG format.

The coordinator will organise the results in "chapters" like:

1. Surface parameter verification (Contribute of different partners...)
2. Vertical profiles verification...
3. .........

Than the coordinator will send this re-organised material to the COSMO web pages and this material will take part of the "member" COSMO web pages. User names and passwords are available through the WP5 coordinator

Back to top

Plenary discussion

After the parallel sections of the Working Groups, a plenary discussion has been done concerning the problem of the definition of a common verification package at ECMWF (WP15 of WG5).

Guenther Doms explained the needs to install a common verification package at ECMWF where LM has been already installed and can be used to run case studies. In order to verify the impacts of changes in the LM chain over all the countries where LM is runned, it's necessary to run LM on a larger area and that can be done only at ECMWF.

GTS data and non-GTS data must be available at ECMWF and a common verification package has to be installed at ECMWF to allow a quick view of the problem.

The discussion that follow concerned:

1. How to define case studies and fine longer periods to be used
2. The provision of non-gts data for the selected case studies and for the selected longer period.
3. The problem of "who can do this job" at ECMWF;

The main discussion was that the MAP IOP case studies should be considered as a starting point (15 case studies). This choice is fine because a lot of non-GTS data are available at the MAP data centre in Zurich and the work of acquiring new data doesn't start from scratch.

All the MAP cases are of great meteorological importance as regards the occurrence of mesoscale meteorological features (dry and wet MAP events) and are of common interest at least for Germany, Switzerland and Italy. In particular MAP IOPs 2, 8 and 15 are very important WET MAP cases and a large amount of data (GTS, non-GTS, remote sensed etc..) is available for these events at the MAP web site

Greece colleagues have to check if some of these MAP case studies are of interest also for them. If that will be the case they should be make available their data to the COSMO Consortium. In any case the Greece colleagues involved in COSMO should indicate other interesting cases, not included in the MAP IOPs cases, of interest for their country.

The problem related to the person that can be responsible of the organisation of non-GTS data, the set-up of software for the common verification package was debated but a clear candidate was not found. It may be that at ECMWF verification software is available. That will be checked.

A list of action was decided in order to achieve these objectives in the near future.

Back to top

Agenda

Back to top

Note for WG5

Discussion about WP 5.15 (Common Cerification Package at ECMWF). Detailed points discussed:

• Definition and installation of a common verification package for LM running at ECMWF. Problem: who can do this job ? How ?
• Provision of verification material (GTS and non-GTS data) as well as an automatic verification procedure to evaluate LM test cases
• Agreement on common verification products, a final version of product list must be defined during the meeting
• Interface to the ECMWF-MARS archive for LM verification

and also

• Organisation of a quarterly reporting cycle for the Verification results (December- March-June-September). Interface with COSMO-WEB and annual reporting
• Distributed Cosmo Verification system (are there political problems ?; Are there technical problems: data bases, special observing networks etc.?)

Back to top