Session 4A.1 Tropical cyclone initialisation in the Met Office global model: Is it still necessary and can it be improved?

Monday, 28 April 2008: 3:30 PM
Palms GF (Wyndham Orlando Resort)
Julian T. Heming, Met Office, Exeter, Devon, United Kingdom

Presentation PDF (733.4 kB)

In 1994 the Met Office implemented a new method for the initialisation of tropical cyclones in its global model. In trials this resulted in a 30% drop in track forecast errors. In 1995 the Met Office global model produced better track forecast guidance than any model available to the National Hurricane Center for the active Atlantic hurricane season of that year. Tropical cyclone track forecast errors in the Met Office global model have continued to show a modest downward trend in the decade since the initialisation technique has been in operation. However, there has been a considerable increase in satellite data available to numerical models during this period. This raises the question of whether usage of the initialisation technique is still necessary to produce a good analysis and forecast of tropical cyclones.

In the first trial of its kind for more than a decade, the Met Office global model was rerun for a period of a month during August and September 2006 with tropical cyclone initialisation switched off. The results show that the initialisation technique still produces a mean decrease in track forecast error of 12.2%. In addition, the detection rate of tropical cyclones at day 5 in the forecast dropped from 96% to 75% and the mean forecast intensity of tropical cyclones dropped by 20.6% when the initialisation technique was not used.

Having established the value of the current initialisation technique, a trial was undertaken of a change to the technique to reduce the areal coverage of the ‘bogus' data produced for small tropical cyclones. This change was proposed in order to address a perceived weakness in the original formulation of the initialisation technique. The revised formulation produced a further reduction in track forecast errors of 4.6% for the August/September 2006 trial period. As a consequence of these good results, the revised formulation became operational in the Met Office global model in November 2007.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner