First PEPS Workshop

6th of April 2005, Bologna (Italy)




21 participants representing the National Weather Services of the following 15 nations:

Austria, Belgium, Croatia, Denmark, Finland, France, Germany, Hungary, Italy (ARPA Emilia-Romania), Netherlands, Norway, Slovenia, Sweden, Switzerland, United Kingdom.


Better graphical representation

Products: Probability of maximum 10m wind gusts.

Only 8 NWS send wind gust forecasts. Thus the PEPS area (35N-70N, 30W-30E) cannot be entirely covered with this information.

On the plots, an area without information or an area with a 0% probability is uncoloured, which makes them undistinguishable.

This has to be changed: the area with 0% probability must be coloured and the area with no information must remain uncoloured.



Models give in general too much precipitation along their lateral boundaries, i.e. in the relaxation zone. For 4 models, the relaxation zone is not considered.

The point whether it would not be better to ignore the relaxation zone for all the models must be examined.


Minimum number of models for a forecast (mean or probabilistic)

Today, at grid points of the PEPS grid covered with 2 models only, mean and probabilistic forecasts are computed.

This is maybe acceptable for the mean forecasts, but not for the probabilistic forecasts.

It has been requested that probabilistic forecasts should only be computed where there is a minimum of 4 models.


Calibration of the model ensemble

The forecast ensembles should be calibrated and this will be done by the DWD with the use of the Bayesian Model Averaging (BMA) technique (see presentation of Michael Denhard).


Statistics of the model availability

It has been requested - and indeed it would be interesting - to have a monthly or at least a quarterly statistics of the mean number of models available for each point of the PEPS grid, at least for the 00 and 12 UTC integrations.


Extension of the forecast range

In today's PEPS forecasts, the lead time does not exceed +30 hours. The wish has been expressed to also have forecasts (mean and probabilistic) for the second day, i.e. for the period +30/+54 hours:

- 24 hour (+30/+54) accumulated total precipitations

- 24 hour (+30/+54) accumulated snow

- Maximum wind speed for the period +30/+54

- Maximum of the wind gusts for the period +30/+54

Presently, PEPS receives forecasts with a lead time of 54 hours or more from 6 models only.

Several models of the 23 models participating are integrated operationally only up to 48 hours or less.

The first step in this development is to find out how many models would be at disposal with a lead time of 54 hours.



With time passing, the need for a verification of the PEPS forecasts becomes always stronger. Unfortunately, the DWD does not have presently the capacity to realise a comprehensive verification of the whole PEPS area: it will limit itself to a basic verification. DWD will contact KNMI for the definition of the verification strategy. Besides this, each participating NWS should try to thoroughly verify the PEPS forecasts over its territory by using its high resolution observations.


Other variables

Michael Denhard has asked whether there is a need (next to total precipitation, snow, wind speed and speed of gusts) for other variables. This does not seem to be case presently. As the PEPS Project puts emphasis on disrupting and dangerous weather, the next variable could be a convective index.


Neighbourhood Ensembles

For each grid point of the PEPS grid, probabilities could be based not only on the values given by the different models, but also on the values at grid points in a neighbourhood. In this case, the resulting probabilities would be results of much larger ensembles.


PEPS data for research

For research works by NWS participating to the Project, PEPS results can be obtained from the DWD. If it would be necessary to also have individual forecasts, the DWD will have to ask the producers of the forecasts whether it can deliver them to the NWS wishing them.

Other Institutions - as Universities - wishing PEPS results or individual forecasts must ask for them through the NWS of their country. Individual forecasts can only be used for research works directly connected with the PEPS method.


Commercialization of the PEPS forecasts

As it is very probable that, soon or later, there will be opportunities to sell PEPS forecasts, Jean Quiby encouraged the participants from NWS members of ECOMET to go to their ECOMET Officer and inform him or her of this possibility.

As ECOMET did it for the radar composits, this organisation should develop formulae for the determination of the prices of the different PEPS forecasts as well as for the computation of the financial returns for participating NWS.



18 August 2005

J. Quiby (with the help of the notes of Michael Denhard)