83rd Annual

Tuesday, 11 February 2003: 1:30 PM
An automated synoptic typing system using archived and real-time NWP model output
Robert R. Dahni, Bureau of Meteorology, Melbourne, Vic., Australia
Poster PDF (304.4 kB)
The Australian Bureau of Meteorology (BoM) has developed an automated synoptic typing system using archived and real-time NWP model output. Objective synoptic typing is performed using a pattern recognition scheme with fields of mean sea level pressure (MSLP). Real-time NWP model output is then automatically classified to generate synoptic type guidance for storage in a Forecast Database for accessibility and verification.

The synoptic types are derived using MSLP analyses over the Australian region from the BoMís archived METANAL grids and the NCEP re-analysis dataset. Principal components were first computed to represent the features of the MSLP fields. Reducing the dimensionality of the dataset, the derived variables based on these principal components were then fed to a K-means unsupervised clustering scheme to derive the synoptic types.

Real-time NWP model output extracted from the operational database is then classified using these synoptic types. Synoptic type guidance for a variety of NWP models is generated operationally and stored in the BoMís Forecast Database (under development), a comprehensive real-time database which allows forecasters to flexibly display, assess, manipulate and store all types of observational data, NWP data, and processed data (analyses and forecasts).

The real-time synoptic type guidance is accessible to new software applications used by forecasters, making the applications ďcontext-sensitiveĒ and allowing automatic presentation of statistical summaries of similar situations. Alerts for likely significant weather appropriate to the synoptic type may also be generated.

The BoMís synoptic typing system is unique in that no other operational system exists (to our knowledge) that incorporates both the development of synoptic types and generation of synoptic type guidance from the automatic classification of real-time NWP model output. The system continues to be under development.

Supplementary URL: