Agenda of the 1st COSMO General Meeting

At ARPA-SMR, Bologna, Italy 23-25 September 1999

Short Summary of Presentations

EU-Activities
(J. Steppeler, DWD)

The two EU proposals within COSMO (LM-NET and COSMO-NOW) were described.

Lead Centre of Nonhydrostatic Modelling
(J. Steppeler,DWD)

The Lead centre activities are the organization of the third internationl workshop on nonhydrostatic modelling , 25.-27. October 1999 in Offenbach, The invitation to model comparison and the accomodation of guest scientists from outside COSMO.

Work Package List for R&DArea 1: Data Assimilation
Coordinator: Schraff, D (DWD)

                      c: commitment for work
      Main            i: interest, but only marginal resources
Work  Responsibility     D: Germany , CH: Switzerland , I: Italy
Pack. with Firm          GR: Greece
Required
No.   Commitment         D  CH I  GR  Activities (and Deliverables)                 Status / Milestone   req. for WP.
----- -------------- -------- -------------------------------------------------------
                                        |  |  |  |
                                                              technical aspects:
                                                              -----------------
1.1   DWD: Schraff       c     i              operational diagnostic and data supervision     03.99: 1st version       1.2
                                                             package for DWD/SMI nudging

                                                             evaluation / testing:
                                                             --------------------
1.2   DWD: Schraff       c  c                evaluation of pre-operational DWD/SMI nudging        11.99
1.2.1 SMI: Bettems           c                nudging operational with the DM/SM

                                                             assimilation of surface-level data:
                                                             ----------------------------------
1.3                                i     i                vertical propagation of surface-level data
1.4                                i     i                data selection for surface stations

                                                             assimilation of upper-air data:
                                                             ------------------------------
1.5                               i  i  i                assimilation of Wind Profiler & VAD radar winds
1.6                               i  i     c             vertical structure functions (WP, VAD, AIRCRAFT)

                                                            use of radar reflectivity (COST 717):
                                                            ------------------------------------
1.7   DWD: Koepken  c     i               latent heat nudging, 2D approach                      03.99: 1st experiments
1.8                                i  i  i              (latent heat nudging, 3D approach                      not starting before 2000)

                                                            other topics
                                                            ------------
1.9   DWD: Hess       c        i              soil moisture assimilation (2DVAR)                        03.99: 1st version
1.10                            i  i  i  i             data quality control for nudging
 

Diagnostic Package for nudging / LM (WP1.1)

1st part:

Establishment of BUFR data base (alt. formatted files) containing.Used for:

1. Daily monitoring of data assimilation scheme
2. Observation monitoring (monthly; special: aircraft data)
3. Verification (vertical profiles: TEMP; aircraft data)
4. Statistical evaluation for experimentation
5. Statistics for tuning nudging scheme (correlation functions)
6. data base as complete as possible:
   1. All reports from AOF (conventional observations), also passive reports and data
   2. Almost all observed variables
   3. Flags
   4. Deviations to model runs: analysis, ?first guess?, forecasts

Status: program finished, but not in operation or available for experiments due to missing scripts.

2nd part:

Reading BUFR file, computation of statistics, graphical display

Status: started in 06.99 for verification /monitoring using upper-air data

Evaluation of nudging (WP1.2)

at DWD (LM):

Statistical results required, should be done with forecasts from a parallel assimilation cycle

Status: severely delayed.

Problems:

1. Parallel assimilation / forecast cycle not possible since 01.99 (various technical problems)
2. Diagnostics package not available for experimenting
   1. Only pre-operational verification,plus case studies from forecasters, not specific to data assimilation
   2. Tuning not started
3. at SMI (SM): Evaluation of 4 months assimilation cycle completed in 11.98. Upper-air verification not yet started.

Nudging operational with SM (WP1.2.1)

Status: cancelled.

Problems:

1. Processors of new machine too slow
2. Short-range data base erroneous Implementation of LM, with nudging at Manno.

Vertical structure functions (WP1.6)

No activities reported by HNMS. (DWD could not provide required LM-based data.)

Developments on data assimilation at DWD, part 1
(Christoph Schraff, DWD)

A test implementation of the latent heat nudging method has been developed to use 2D radar reflectivity processed to rain rates as proxy data for the vertically integrated latent heat release. To distribute the desired change of the integrated heating in the vertical, model latent heat profiles are filtered and scaled according to the ratio of the observed to the simulated rain rate. For non-precipitating model grid points with observed rain, a profile search is done. Standard parabolic vertical profiles are used only in the absence of any precipitating grid point located sufficiently close.

The scheme has been extended by an adjustment of the model's specific humidity so as to preserve relative humidity when the latent heating is reduced, and to reach saturation over a nudging timescale when the heating is enhanced. Results of feasibility tests comparing runs with and without humidity adjustment to routine results showed, that the enhancement of humidity fields is an important factor. Since the latent heat nudging is observed to modify the wind field patterns, the method could be shown to have a potential to improve both precipitation patterns and amounts not only during, but also several hours after the end of the assimilation period. However, the method needs to be improved further before considering its operational implementation.(Work done by Christina Koepken)

Developments on data assimilation at DWD, part 2
(Christoph Schraff, DWD)

A variational soil moisture analysis scheme for the top soil layer has been implemented which minimizes a cost functional that expresses the differences between model-derived and observed screen-level temperature (and humidity). The soil moisture retrieval was found to work well in early spring for clear-sky conditions with strong radiative forcing, but imposes very large soil moisture increments with only a small impact on the temperature in other regions.

Hence, the method has been refined to include a background field based on the soil moisture analysis of the previous day, and a background error is computed in a Kalman-Filter-like way. The analysed soil moisture will then significantly deviate from the background only where the background error and (or) the soil-atmosphere coupling are sufficiently large. In this way, information from previous days is conveyed to the soil moisture analysis in those areas, where at that particular day the observed temperature contains little information about the soil moisture.

As a result, reasonable retrieved soil moisture fields without extreme values are obtained in a two-week test in March, and both the derived soil moisture analysis error and the error of the predicted screen-level temperature are reduced. Since the moisture of the second soil layer contributes to the surface evaporation by influencing the top soil layer through capillary effects and by evapotranspiration of plants in late spring and summer, the scheme has been extended to analyze the soil moisture in the top two layers, and the testing and tuning of this version has started. (Work done by Reinhold Hess)

Work Package 2: Numerical Aspects and Case Studies
(J. Steppeler, DWD)

Wp 2.1: ARPA/Paccagnella; MAP-Cases

The Piedmonte 4.11.9 was begun with LM7km. Both Piedmont and Brig Cases are expected to be ready before the COSMO-meeting

Status: ongoing work. Transfer to WP6?

Wp 2.2: DWD/Bitzer,Steppeler; Reference cases 7 km 2.5 km

One case (Brig) with resolution LM 2.5km is ready, a list of further interesting cases exists

Status: ongoing work

Wp 2.3: SMI/Bettems; Truely horizontal diffusion

Experiment in 2-d at DWD (Gassmann)

Status: No work (3-d) has been done. Will be taken up

Wp 2.4: SMI/NN ;Boulder case

Status: No work has been done. Delete?

Wp 2.5: SMI/Schubiger; 5 test cases LM 2.5 comparison with MC2

Status: No work has been done. Delete?

Wp 2.6: NN; Systematic comparison to MC2 during MAP SOP

Status: Responsible person has not yet been found. Delete?

Wp 2.7: DWD/Doms; Test of 2-timelevel-scheme by Wicker/Doms

Most tests had positive results, but more work is needed before operational application is possible

Status: Ongoing work

Wp 2.8: DWD/Rissmann; Test of 2-way nesting option

Two cases were computed showing no problems in respect of the meteorological results, but problems remain making the scheme portable to all computer platforms.

Status: Ongoing work

Wp 2.9: NN; Evaluate NWV-applications for 2-way nesting

Status: No work has been done. Delete?

Wp 2.10: DWD/Minotte; Evaluate Mesingers step coordinate

This coordinate was found not to produce proper gravitational waves. Alternative schemes which do not have this defect have been tested in 2-d

Status: To be completed by end 1999; Follow up work to be decided

Wp 2.11: DWD/Steppeler; Plan for next generation numerics

Report can be given at COSMO meeting

Status: First version ready; needs further discussion

Wp 2.12: DWD/Herzog Evaluation of 3-d semi-implizit version at very high resolution (100m)

The expected speedup for high resolution runs was not realized, due to convergence problems encountered. Further evaluation is needed

Status: Ongoing work

Wp 2.13: DWD/Schubert; Run LM at ECMWF

A system to run a very high resolution version of LM at ECMWF has been installed successfully. The use for general LM-experiments will require further work and planning.

Status: Work completed; follow up work to be decided

Wp 2.14: DWD/Doms; Evaluation of LM for dx=7km

Will lead to operational use at DWD by end 1999

Status: Completed by end 1999; follow up work to be decided

Wp 2.15: Reformulation of dynamic equations

Status: No work has been done. No responsible person has been found, delete?

Wp 2.16: Two cases using SM boundary values

SM- boundary values have been delivered to Bologna

Status: ongoing work (?),  delete?

Wp 2.17: Test explicite positive vertical advection schemes

Status: new task

Wp 2.18: Interface to reanalysis ECMWF, NCEP (Wp 7, Wp 5)

Status: new task

Wp 2.19: simple Legoland tests

Status: Continuation of work

Greece has indicated interest to take part in wp 2.10 and wp 2.13

Results of the Pre-operational Test of LM at DWD
(G. Doms, DWD)

A new nonhydrostatic limited area model, the "Lokal-Modell" (LM) has been developed at DWD. It is designed as a flexible tool for various applications on a wide range of scales. The LM will replace the current high resolution model DM for short range numerical weather prediction by the end of 1999. Using a higher horizontal and vertical resolution, it will be nested directly into a new global model (GME).

In order to test and upgrade the new model system a one-year pre-operational trial starting in October 98 has been initiated. The basic set-up for LM uses 7 km horizontal resolution on a 325x325 grid with 35 vertical layers covering the integration domain of DM. The timestep is 40 s and two daily 48 h forecast runs starting from 00 UTC and 12 UTC are performed.

The initial conditions are supplied by a data assimilation cycle using the nudging analysis technique (since January 99). The boundary data are obtained from corresponding GME forecasts.

During this test phase the model was shown to run stable, robust and efficient on a CRAY T3E and no blow-ups have been recorded up to now. Also, no severe problems related to numerics, nonhydrostatic dynamics or physics have been encountered. The new data assimilation stream based on nudging runs stable and no drifts compared to the DM analysis cycle have been observed. However, a moist bias with too large convective precipitation amounts in the assimilation cycle occurred during the summer months. This problem is currently worked on.

As a basic result the pre-operational trial has shown that the overall quality of the LM-forecasts is very similar to DM. This had to be expected since we are still on a hydrostatic scale and the present physics routines do not differ much from DM. But nevertheless it is not easy to beat an established, well tuned operational model. In general, the the forecasts of LM and DM are very similar but, as expected, LM gives a better regionalization with more dynamical and topographical details. In most cases where larger differences in the forecasts occurred, the reason could be traced back to larger differences in the boundary conditions specifying the synoptic-scale forcing.

The subjective evaluation of NWP products by forecasters reveals advantages for LM in many weather situations. E.g., most of the winter storms as well as heavy precipitation events from mesocyclones during winter and spring have been well predicted by LM. However, up to now no clear advantage for LM could be recognized for unorganized convection as airmass thunderstorms within the warm sector ahead of cold front. In such situations, both models either fail or succeed to predict the onset and duration of convection at the right place, indicating a basic deficiency of the convection scheme used.

The statistical evaluation of surface weather elements shows advantages of the LM for pressure, wind and especially for temperature with a significantly reduced bias. More or less the same skill scores are obtained for cloud cover by both models. The area average precipitation in LM is in general somewhat smaller than in DM and compares better to observed monthly climatologies. However, there are problems with the precipitation scores, especially for detecting precipitation events with more than 5mm/24hrs.

Numerical Developments
(J. Steppeler,DWD)

The currant state of the LM-numerics was described. There are currently three alternative time integration schemes available for LM: three time level split explicit, two timelevel split explicit and 3-d implicit. The prospects of these methods were discussed and their potential role in capturing future chances of development were described. In respect of ongoing work at DWD the problems associated with the step mountain coordinate were described.

At the second international SRNWP workshop on nonhydrostatic modelling J. Klemp reported on problems which have been encountered at NCAR when trying to imlement the step mountain coordinate. Some of the existing step-approaches do not produce proper gravitational waves over mountains. These problems made an immediate implementation of the step coordinate in LM impossible.

Instead two dimensional experiments were performed at DWD in order to bring these problems to a solution. In cooperation of DWD with the University of Trento some success has been achieved. Therefore the introduction of the step coordinate into LM now seems feasible.

Most work into numerical aspects has longer term character. Therefore the need was seen to discuss the potential numerical developments for LM. The actual work to be performed can then be chosen according to resources available and priorities.

The current list of potential numerical tasks is the following:

1. Supervision and improvement of LM at resolution 7 km
2. Developments in connection with LM at a resolution 2.5 km
3. Prognostic rain / advection (DWD)
4. 3-d implicit method
5. 2 timelevels (DWD)
6. Step mountains
7. Semi-Lagrangian Method for CFL>1
8. 2 way nesting
9. Splitting and interface to physics scheme
10. Cloud ice scheme/advection
11. Adaptive meshes
12. Shallow convection by numerical process
13. Splitting, inclusive interface to physics
14. Investigtion of the performance of LM at resolution 2.5 km (DWD)
15. Investigation of special NWP-applications of LM (such as Olympic games version)

Studies of the Brig strong Precipitation Case using the LM non Hydrostatic Model
(H.W. Bitzer and Dr. J. Steppeler)

The verification of model results against observations is indispensable for development, improving the quality and for sensitivity studies of nonhydro-static models. For this reason extreme weather events are favourable as test cases. The current investigation concerns the Brig flood from 22nd September 1993 until 24th September 1993. This case was taken in order to verify the LM.

Associated with an upper level cutoff cyclone that moved slowly from Spain to the gulf of Genoa, while intensifying, several mesoscale convective systems along the surface coldfront developed. The low tropospheric advection of warm moist air from the mediterranean sea, orograhic forcing and synoptic scale vor-ticity advection favoured the long lasting lifting of air, that resulted in intense precipitation events. Widespread rain maxima of 100 mm within 24 hours (isolated significantly more than 100 mm) were recorded. To forecast the important local effects correctly is generally very difficult. The simulation of these effects as well as the exact location of precipitation is a big challenge for the nonhydrostatic model LM. The results of two LM simulations are presented:

1. Model run with horizontal grid size dx = dy = 2.8 km, ke = 55 layers in the vertical
2. Model run with horizontal resolution of 7 km, ke = 35 layers in the vertical.

The initial and boundary data are provided from the model chain GM-EM-DM. Finally the interpolation of data from DM to the fine mesh of the LM is done. Both model runs are a 54 h forecast starting 22nd September 1993 00 UTC until 24th September 06 UTC. Model results of the LM are compared with surveyings, gained via Internet from the MAP DATA-centre. In the synoptic scale the flow pattern and distribution of precipitation can be well represented by the LM.

However significant differences against recordings are detectable in the regional scale (meso beta). Beyond the alpine crest (north of it) no model precipitation occures in the lee of the mountains. On both days of the simulation the results show up with a significant suppression of rainfall in the lee areas, while the luv area precipitation almost always exceeds the recordings slightly. In addition the severe precipitation in Saltina river's catchment above Brig cannot satisfyingly be simulated. A flood in Brig cannot be predicted according to the model results. A model run with the convection scheme switched on cannot improve the results as well, as is expected, since running the model with a horizontal resolution of 2.8 km. The lack of precipitation in the lee areas is rather a result from the dynamic part of the model and from the horizontal diffusion. With the 7 km run the precipitation in the lee area can be well simulated. But still the exact location of severe rainfall over the area of interest (Saltina river's cathment above Brig) cannot be captured. The vertical cross section (yz-plane located through Brig) shows sufficient liquid water for precipitation north of the alpine crest, while in the 2.8 km run the available liquid water is completely missing.

Conclusions

Precipitation structure and amount is basically captured. These precipitation areas are strongly correlated to the vertical velocity, whereas the vertical velocity is correlated to the gradient of orography. A considerable overestimation of the lee effect of the LM is obvious. A possible improvement could be the introduction of prognostic rain and a step coordinate.

Report of Work-Package Coordinator WP 3
(E. Heise, DWD)

I. Status of Work Packages

In the following the status of the different work packages from the final list (March 1999) is given:

WP 3.1: Level 2.5 turbulence parameterization (D: c)

A basic version of the level 2.5 parameterization scheme has been implemented in LM-version 1.30. The scheme replaces the diagnostic determination of turbulent kinetic energy in the operational turbulence parameterization by a prognostic treatment. The scheme (optionally) also replaces the former treatment of the Prandtl-Layer and (optionally) uses a laminar sublayer for coupling of the Prandtl-Layer with the soil surface. A statistical scheme for the determination of subgrid-scale clouds is formulated consistent with the turbulence scheme and (optionally) replaces the former relative humidity approach.

status: in time

WP 3.2: New soil model AMBETI (D: c, GR: i)

The implementation of AMBETI proved to be a very hard and time consuming task, which has by far not been finished yet. Therefore, the operational soil model TERRA has been extended to include the Penman-Monteith formulation of plant transpiration and a bare soil evaporation formulation based on the concept of maximum sustainable moisture flux in the soil.

status: significant delay, but provisional solution available

WP 3.3: Ice phase (CH: c, D: c)

The effect of the prognostic equation for the ice-phase has been tested by SMA. Parallel runs for February 1999 showed generally little differences with one notable exception: Total cloud cover increased by 2 octas when the ice scheme was used. This problem has not yet been solved, and the ice phase has not yet been switched on in the preoperational LM runs.

status: delayed

WP 3.4: Convection (CH: c, D: i)

A convective available energy (CAPE) closure has been implemented and tested by DWD. The closure improves the distribution of convective precipitation in cases of air mass convection.

status: partly in time, CH-contribution postponed ?

WP 3.5: LM-runs with level 2.5 scheme (D: i, GR: i)

LM-runs were used during the development of the level 2.5 scheme for a general testing and tuning of the scheme. Since July 1st nearly daily parallel runs are conducted at DWD for further testing and tuning. (These runs also include the modified soil model and the CAPE closure for convection.) The runs resulted in the detection of some remaining problems in special results of the new parameterizations. This required a retuning of some parameters. In general the results look promising.

status: in time

WP 3.6: LM-runs with AMBETI (D: i, GR: i)

status: postponed

WP 3.7: 1D long runs with Lindenberg-data for validation of level 2.5 scheme and AMBETI (D: c)

A lot of short-range (30h) runs for testing the level 2.5 scheme has been performed. This was prerequisite for tuning the scheme. Long runs and runs with AMBETI have not yet been made.

status: partly in time,  partly postponed

WP 3.8: Use of CORINE-data for 2.8 km LM (D: c)

The use of CORINE-data was implemented in the program-system for the determination of external parameters. Fields for the 7 km version of LM have been produced, and 2.8 km data can be produced at short notice.

status: in time

WP 3.9: Snow albedo for forests (D: c)

A version based on the respective ECMWF-scheme was implemented in the operational Global-Modell of DWD providing the expected results. The version can be used in LM, too, but has not yet been implemented. Can be used at short notice.

status: in time

WP 3.10: Partial convective cloudiness (CH: c, D: i)

status: CH-contribution postponed ?

II. Critical issues

The communication could be better. COSMO-members having finished some part of a work package or see problems causing delays should inform the work package coordinator.

III. Objectives and priorities

1. The level 2.5 turbulence parameterization is well in progress
2. The implementation of AMBETI needs support. Perhaps a solution in DWD is possible
3. Most important seems to be an improvement of penetrative convection. Two points are of special interest:
   1. improvement of the daily cycle of convection
   2. convection simulation in the 2.8 km version (parameterized versus explicit simulation)
4. parameterization of shallow convection should be developed in close connection with the turbulence scheme (DWD can take the responsibility)
5. A lake model seems to be necessary in the near future

The new physics package of the LM
(E. Heise and G. Doms)

A 1-year preoperational trial of the new GME/LM system has been started in October 1998. During this test phase, a number of new components of the LM are also evaluated. Current work concentrates on a new LM physics package for the operational application. It is made up of

1. a new scheme for turbulent transports based on a prognostic equation for turbulent kinetic energy including subgrid-scale condensation and the impact of subgrid thermal circulations
2. a refined surface layer parameterization including a laminarboundary layer
3. a new version of the soil model allowing for three soil moisture layers and a Penman-Monteith type transpiration parameterization
4. a modified version of the mass-flux convection scheme using a convective available energy (CAPE) closure and
5. a new grid-scale cloud and precipitation scheme including cloud ice as an additional prognostic variable besides cloud water

Tests with the new physics package including the TKE-scheme, the revised soil model and the CAPE closure for convection have been performed nearly daily since July 1990 in a parallel suite. The results from these integrations in general look promising, especially with respect to the 2m-temperature. In most cases the CAPE closure performed better than the standard closure based on moisture convergence, both with respect to the area covered by convective rainfall and the maximum amounts of precipitation. It is planned to put the new physics into operations in a step by step procedure until the end of this year.

The cloud ice scheme suffers from non-adequate initial and boundary conditions for specific humidity. Thus, the scheme has to be implemented in the global model GME and a number of modifications have to be done in the analysis schemes of and LM.

COSMO: Working Group 5: Verification
(C. Cacciamani, ARPA-SMR)

WORK DONE SINCE BAD SAECKINGEN 1999 IN THE R&D WP: VERIFICATION (WG5) WP 1.1: ARPA-SMR

Implementation of a "high resolution verification" of both LM and SM precipitation forecasts in the Emilia Romagna region of Northern Italy (accumulated precipitations in 6 and 24 hours) performed against precipitation analyses (5 Km grid) and standard surface observations (GTS data, SYNOP reports and local non-GTS data) available at ARPA-SMR.

1. The precipitation analyses (called "GIAS") are obtained using raingauge and calibrated Radar data (the radar is located at San Pietro Capofiume,Bologna). The verification period covers the 6 month period from February '99 to July '99 as regards LM and the 8 months period from December '98 to July '99 as regards SM .

2. The LM GRIB1 data have been obtained from DWD through a daily ftp transfer to ARPA-SMR (many thanks to Dr. Damrath of DWD).A data base of LM-GRIB1 forecasts, the necessary Fortran codes to manage these data (extraction, decoding, graphics, etc) and finally all the verification routines have been produced at ARPA-SMR;

3. GRIB1 data of the Swiss Model (SM) are transmitted via normal mail (monthly delivery) by SMA and are available at ARPA-SMR 10-15 days after the end of the previous month (many thanks to E. Zala and F. Schubiger of SMA). A similar archive of SM GRIB data have been produced. The Fortran routines to manage these data are essentially the same used for the LM model.

4. Puntual verification: evaluation of contingence tables using the nearest LM grid points to the available "station" points using different precipitation thresholds. Contingence tables evaluated for different forecast times (+6 and +24 hours), considering all the "stations" of the whole Emilia Romagna region and also some sub-regions of it. From the contingence tables has then been possible to evaluate standard accuracy indices (i.e.: hit scores, false alarm, bias, etc..).

5. Areal verification: evaluation of spatial averages of observed and forecasted precipitation, for different forecast times and accumulated periods. All the "stations" and LM grid points belonging to the total Emilia Romagna area and/or sub regions of it are considered and spatial averages of observed and simulated precipitation are compared. Time trend of forecast error (for the different forecats times and geographical areas) and standard quality indices (i.e.: mean absolute error, rms error) are evaluated.

Notes: Daily ftp-transfer from DWD to ARPA-SMR is fully operative only since the end of January 1999. During December '98 and January '99 problems occurred in the ftp procedure due to the big size of the LM GRIB files. For this reason and after request of ARPA-SMR, DWD created smaller GRIB files which could be transmitted in a few minutes from Offenbach to Bologna.

A big hardware fault of the ARPA-SMR radar system occurred during March '99 and no radar data are available from March to July 1999. For this reason the precipitation analyses (GIAS) in that period are not available and only the verification against observations have been performed.

WP 1.2: ARPA-SMR/IMS

A Verification of Lokal Model by other Regional Weather Services (RWS) of Piedmont administrative region, Liguria region and Marche region which have dense networks of non-GTS stations available.

A common verification work plan have been defined; at the moment only the LM data, received from DWD at ARPA-SMR, are forwarded to these RWS and there are verified against local non-GTS observations.

WP 1.3: IMS Italy

Verification of LM precipitation forecast (and probably also other parameters) against non-GTS data available in the Northern Italy during the MAP field phase

1. Creation of all the procedures to receive and archive, during the MAP field phase (from September to November 1999), a large number of non-GTS hourly weather observations, gathered in real time from several RWS operative in the Northern Italy and from the National Hydrographic Service.

2. This data set will be used at the IMS for the verification of the LM surface parameters forecasts.

Notes: At the moment, the total number of available station is over 240 and the observations are relative to the following parameters: precipitation; surface pressure, temperature relative humidity, wind velocity and direction.

As regards the LM forecast, the GRIB data can be delivered from DWD (or forwarded from ARPA-SMR only as regards precipitation) to IMS after or during the MAP field phase.

WP 1.4: SMA

Implementation of the verification on an hourly basis of the LM with observations of the automatic network of SMI (ANETZ) and (on these gridpoints) the comparison of LM with SM.

1. Archive of the LM-GRIB files on the ETHZ-CRAY SILO and program for the extraction of the GRIB1-data. Some problems occured with the transmission of the LM-GRIB files (see below under Notes).

2. Correspondence of LM-gridpoint to ANETZ-observations: it is done in a same manner as for SM.

3. The verification of cloud cover has been extended to all grid points with a radius of 30 km around the observation, i.e. to 41 grid points of LM (for SM there were 13 grid points).

4. The verification of precipitation has shown that many gridpoints were almost dry, so the verification scheme had to be extended. It is done now (for precipitation) for one gridpoint, as well as for 5 LM-gridpoints (i.e. giving about the same area as for SM) and for 13 LM-gridpoints around the observation station.

5. Comparison of LM and SM: Attention has been paid that strictly the same forecasts of SM and LM are used in the verification. The results are available for each location of ANETZ-station / SM resp. LM gridpoint(s), and also as a mean for all gridpoints and the ANETZ-stations of three height classes. As the height of the LM-gridpoints are different from those of SM, the comparison for the different height classes (< 800m, 800-1500m>1500m) is done with the reference height of the observation (for common comparison).

6. At the COSMO meeting the results for Spring 1999 (February-May) and Summer 1999 (June-August) have been shown for 2m-temperature, 10m-wind, precipitation (hourly intervals) and for cloud cover (3-hourly).

7. In a second stage (later on) the verification of last winter 1998/1999 will be done for 2m-temperature, precipitation and cloud cover (10m-wind was not available at that time ).

8. Results are available as mean diurnal cycle, bias, rms, standard deviation, % correct within different thresholdes. For precipitation and cloud cover contingency tables are computed and for precipitation scores for different thresholdes (0.1, 2, 5, 10, 30 and 50 mm/6h).

WP 1.5 - SMA

Low PBL verification of T, W up to 110 m. with masts of swiss nuclear powerplants. Work pending. Start during 1999.

WP 1.6 - SMA

Daily verification of LM/SM cloudiness with the Meteosat VIS Chanel at 12 GMT. Activity to be started after summer 1999.

WP 1.7. DWD:

Verification of surface weather elements observed at synop stations

1. Verification results are available for all months starting with March 1999 (RMSE, BIAS, percent of correct cases within certain thresholds, TSS for 6-h-precipitation amounts)

2. Special investigations are made concerning precipitation; graphics are prepared with 24-h-precipitation amounts observed in a high density network (about 4200 Stations in Germany) and forecasted by LM and DM.

3. On the basis of synop station measurements over Germany statistics were made (probability of detection, false alarm rate and TSS).

WP 1.8 DWD

As WP 5. Work pending. Start during 1999.

WP 1.9 and WP 10: Greece WS

Unfortunately the Greek area is outside the LM integration at DWD. For this reason it was not possible to achieve the planned verification activity.

CRITICAL ISSUES, OBJECTIVES AND PRIORITIES FOR FUTURE ACTIVITIES IN THE WP5
(hints for discussion)

Critical Issues:

During this first year each Institution worked essentially by its own, without a strong coordination. That has been inevitable because some of these Institutions had to setup from scratch all the procedures to receive LM GRIB data (and also SM in the case of ARPA-SMR), archive and finally use them. It has been really impossible to dedicate more time and manpower to have technical restricted meetings to exchange ideas and define common methods to work.

It is now necessary for future to increase the level of coordination and to improve the collaboration between the Institutions involved in this WG. A better level of coordination will avoid useless duplications of similar activities supporting, on the contrary, a good exchange of ideas and also application software.

It is recognised that when the horizontal resolution of the LAMs becomes very high (7 Km or even less in the case of LM), and very often much higher than the density of the observing network, the standard verification methods (evaluation of bias, rms, contingence tables, etc.) could be no more satisfactory to judge the quality of simulated fields, especially when these have a very high spatial variability (for example precipitation). New verification methods (i.e., pattern correlation with different spatial lags between observed and simulated fields) should be discussed and probably proposed.

Priorities:

To define a common (at least for D, CH and I) non-GTS network of real time observations available for the verification of high resolution numerical outputs of LM. The MAP data set will be a very rich archive and in future should be carefully used (even) for this purpose;

To discuss in detail common verification procedures, in order to obtain results that can be compared each other. If this point cannot be considered essential as regards the verification against standard observations (i.e. surface GTS or non-GTS data), it becomes instead really crucial when the model outputs are compared with remote sensed data, like, for example, satellite or radar data. An exchange of ideas regarding methods, techniques and (probably) software to use these "non standard" observations can really facilitate the activities and improve the level of the results.

For short time periods (1-2 weeks) personnel could be exchanged between the different Institutions, in order to acquire experience and direct knowledge of the activities performed by other partners.

Work Package List for Working Area 6: Reference Version and Implementation
Coordinator:Majewski, D (DWD)

                         c: commitment for work
      Main            i: interest, but only marginal resources
Work  Responsibility     D: Germany , CH: Switzerland , I: Italy
Pack. with Firm          GR: Greece
Required for
No.   Commitment         D  CH I  GR  Activities (and Deliverables)                       Status / Milestone                      Work Package
----- --------------     -----------  --------------------------------------------------------------
                         |  |  |  |
 6.1                     c  c  c      Preparations of topographical data sets (GRIB1)            complete by 26/02/99
                                          Domains I to V, 28, 21, 14, 7, 2.8 km

 6.2                                    Work at SMI
 6.2.1                   i  c         Installation of SM on Cray SV1 at ETH                          start at 01/99
 6.2.2                   i  c         Installation of GME2HM/GME2LM on SV1 at ETH    start at 04/99
 6.2.3                   i  i         Installation of LM_S on SV1 using MPI                          start at 09/99
 6.2.4                   i  c         Transmission of GME data to SMI                                  start at 09/99

 6.3                                    Work at ITAV
 6.3.1                   i     i      Installation of IM on DEC Alpha 8200                            start at 07/99
 6.3.2                   i     c      Installation of GME2HM/GME2LM/HM2LM on DEC Alpha   start at 03/99
 6.3.3                   i     c      Installation of LM_I on DEC Alpha 8200                       start at 07/99
 6.3.4                   i     c      Transmission of GME data to ITAV (via ECMWF?)     start at 03/99                                          6.4.3
 6.3.5                   i     c      Creation of AOF file from IMS data base                       start at 04/99

 6.4                                    Work at SMR-ARPA
 6.4.1                   i     c      Installation of LM_I on Cray T3E (Bologna)                  start at 02/99
 6.4.2                   i     c      Installation of GME2LM/HM2LM on Cray T3E           start at 03/99
 6.4.3                   i     i      Transmission of GME data to SMR-ARPA via ITAV   start at ?
 6.4.4                         i      Transmission of IM  data to SMR-ARPA from ITAV   start at ?
 6.4.5                   i     i      Creation of AOF file from ARPA data base for LM_I   start at 09/99
 6.4.6                   i     i      Quasi-operational use of LM_I at SMR-ARPA             start at ?

 6.5                                   Work at HNMS
 6.5.1                   i        i   Installation of LM_G on Convex (Athens)                       start at ?
 6.5.2                   i        i   Installation of GME2LM on Convex (Athens)                 start at ?
 6.5.3                   i        i   Transmission of GME data to HNMS                              start at ?
 6.5.4                   i        i   Creation of AOF file from HNMS data base for LM_G start at ?
 6.5.5                   i        i   Quasi-operational use of LM_G at HNMS                      start at ?

 6.6                      i  c         Programming and testing of LM2LM                              start at 06/99

 6.7                     c            Reference version of LM packages (LM, GME2LM, LM2LM)
 6.7.1                   i            Set up new mailing list of "operational" LM users          02/99
 6.7.2                   i            Provide direct access to master libraries (ftp)                   02/99
 6.7.3                   i  c  i  i   Define procedure for updating the reference version      04/99
 6.7.4                   i  c  i  i   Regular update of reference version                                 start at 04/99
 
 6.8                        c         Set-up of common plotting package based on METVIEW  start at 04/99
 

Model Design and Code Maintenance for the LM Package
(U. Schaettler, DWD)

In the first part, an overview of the different models now in use was given. Two of them, the interpolation program GME2LM and the LM itself form the LM-package. The software design for the programs, which takes especially care of portability and modularity, was explained. The second part described the work and activities done in the R&D Area 6.

Minutes of the strategic discussion

1. The participants confirmed the three main directions into which the efforts have to be pursued in 2000

(a) To put the Local Model in operation with

• a modular structure of the software pieces
• a modular structure for the parametrisation of physical and numerical aspects
• a documented reference version of the model

(b) To improve the collaboration between the working groups including

• an agreement on the delivery transferts
• a better communication
• informal travelling

(c) To be customer oriented with

• a better coordination of the verification activities
• the preparation of guideline for the forecasters
• a better information about the model

2. Draft Agreement

A discussion occured about the current status of the draft agreement (25.9.99). Few corrections were proposed, mostly from Switzerland:

• Art. 3: "that will not be unreasonably witheld" is quite a weak formulation
• Art. 9: One should read "Data and Products"
• Art. 10: Each NMS-Partner should define a "Licencing Agent", responsible for the licensing conditions as they are defined in the Agreement (Same practice as EUMETSAT)
• Annex A: Bullet 3: defines conditions for the COSMO members, but implies that recipient not being a member of COSMO may use the data and products commercialy. The definition of a license contract should be given with an example.

3. The steering committe decided to open a WEB-site devoted to COSMO

Greece proposed himself as the initiator of this project. The issue whether only static information or some dynamical data -forecasts- should be presented remained open. A decision has to be taken during the second meeting of the steering Committee.

4. Dr. Majewski was designated as the representative of COSMO at the EWGLAM - Meeting

5. Contribution to EUCOS

The discussion showed that only SMI had a strong commitment with the SM. Possibility to use LM Data remaind an open issue, to be discussed at the second steering Committee.

6. Dr. Frühwald presented the EU-Project "European Flood Forecasting System" favourably evaluated.

7. Dr. Marco Arpagaus was designated as the coordinator of the work package Physical Aspects

8. Dr. Günther Doms was elected as the Scientific coordinator of COSMO.

It was agreed that this charge should last 3 years.

9. J. Ambühl was designated as the Chairman of the Steering Committee for the year 2000

10. A mailing list was organized

Jacques Ambühl, 24.1.00