Manfred Kurz, Deutscher Wetterdienst

During the Sixth Meeting of the WG CEF in Innsbruck in September 2000 the participants presented the technical means they have available in their Services for the presentation of observations as well as NWP output. In this connection it was stated that forecasters must be always be aware of the fact that the excellent visualization or even animation of the model output that most graphical systems allow, does not automatically mean that the model products are always perfect and could be used without careful evaluation by the forecasters themselves and by the continuous monitoring and comparison with the actual weather. The question arose, which methods are used in the different Services in the cases in which the model forecasts prove to be wrong. Another interesting question was, how to reach consensus between forecasters of different centres when the opinions on the weather development are significantly different. The author offered to address both questions with the aid of a questionnaire.

The questionnaire was sent out to all members of the WG CEF in November 2000. Although there were only nine responses, an evaluation was made showing the following results:

Question 1: We all know that numerical models sometimes fail to simulate the weather development correctly. That is especially important in cases of hazardous weather. Which methods are used in your Service in such a case with regard to issue of warnings, general forecasting, forecasting for aviation and shipping etc.?

1. For nowcasting and very-short-range forecasting (0-12 hours) all observations including radar and satellite images are basic instruments to detect possible model failures. Support by a competent SATREP analysis was mentioned.

2. Monitoring of the model output is made by comparing the surface analysis of the model with the manually constructed surface map and by comparing the first steps of the forecasts with the synoptic observations.

3. A manual intervention into the numerical analyses can be performed in the Met Office, UK, but no scheme to use PV-inversion in order to correct all fields in a consistent manner has been incorporated as yet.

4. It is recognized that when there are large errors from the start of a model run then manual adjustments have to be made. However, mainly subjective methods are mentioned to make the adjustment. Only in the Met Office, UK, the Chief forecaster in the Guidance Centre can make the required adjustments in the short-term by using the method of On-Screen Field Modification (OSFM) based on PV-inversion. In Deutscher Wetterdienst only the surface forecast maps can be adjusted interactively, but in a subjective manner using extrapolation and diagnostic reasoning.
5. The methods used for correcting the numerical output or replacing it by alternate forecasts are not explicitly described in most responses. The importance of a sound knowledge of dynamics and of experience as a deciding factor is stressed. Altogether the forecasters should develop their own 'mental weather picture' and follow it in cases of model failures.

6. In most Services the operational forecasters have access to a variety of models. They have therefore the freedom to decide, which if any, gives the best guidance in any particular weather situation.

7. The application of existing diagnostic studies dealing with analogous cases in the past (Greece) and of special diagnostic packages in order to fully exploit the output of NWP models (e.g. 'High resolution isentropic diagnosis' used in Croatia) was mentioned. In Deutscher Wetterdienst a combination of a quasi-geostrophic diagnosis based on model data and application of 'Conceptual models' based on observations is recommended for the daily work and, of course, especially for cases with model failures.

8. The responsibility for adjustments or corrections in the case of model failures rests in most Services with the Chief forecaster in the Central Forecasting Unit. With the aid of internal conferences and special bulletins the necessary information and co ordination with the Regional Offices is performed.

9. It is recognized as very important that not only warnings, but all forecasts should reflect the adjustments made in case of model failures. This requirement, however, can be hardly realized, since in many Services a lot of products are derived and transmitted fully automatically.

10. An interesting proposal made by E.v.Loock (Belgocontrol): Centres should issue 'warnings' to attract the attention of other forecasters on important differences between models.

Question 2: It is an important task for every forecaster to make his/her own synoptic diagnosis of the actual weather with regard to the potential for future developments (the so-called 'mental weather picture'). Sometimes it may happen that the opinions on the weather development between forecasters in different shifts or between the forecasters in the Central Office and the Regional Offices are significantly different. Since only one uniform forecast should be issued by the Met Service, a consensus has to be reached. What are the ways you follow in your Service to reach this consensus?

1. The common way to overcome or avoid different opinions on the weather development is by discussion and regular interaction between the forecasters on duty and by phone conferences between Central and Regional Offices, or even between different Services as, for example, in Belgium.

2. It is recognized that reaching a consensus when opinions differ significantly can be a problem. In all Services, however, provisions are made for the rather rare events of disagreement. The Supervisors or Chief forecasters have to decide within an Office, whereas in the discussion between Central and Regional Offices the view of the forecaster at the Central Office prevails.

3. An important point was made by Liisa Fredrikson: 'Sometimes a forecaster issues differing forecasts but does not mention this during the conference. This is something we should continuously remind the forecasters not to do. The struggle for the consensus itself often is fruitful. With no discussion, important points of view might be missed.'