New clustering products

THE ECMWF clustering is one of a range of products that summarise the large amount of information in the Ensemble Prediction System (EPS). The clustering gives an overview of the different synoptic flow patterns in the EPS. Based on the similarity between their 500 hPa geopotential fields over the North Atlantic and Europe, the members are grouped together. EPS cluster products have been produced operationally since 1992. In recent years, due to the continuous improve- ments of the EPS (in particular reduced spread, consistent with decreasing ensemble mean error), these products only occasionally produced more than one cluster. The require- ment for cluster products was recently reviewed with ECMWF's Member and Co-operating States, particularly during the annual Forecast Products Users' Meeting. Although some countries do their own clustering (for specific parameters or areas of interest), there was a clear requirement for ECMWF to continue providing a general cluster product from the EPS. Therefore, based on the feed- back from the Member and Co-operating States, a new EPS clustering application was developed. The new clustering was endorsed by the TAC Subgroup on Verification Measures as part of their review of product development and user requirements. The new system includes two components: u A daily clustering of the forecast fields from the EPS, similar in principle to the original EPS clustering but using a different algorithm. u A set of four fixed climatological regimes. The daily clustering summarises the range of synoptic flow patterns in the current EPS. Each cluster is represented by the EPS member closest to its centre, referred to as the 'EPS scenario' for that cluster. Each EPS scenario is then attributed to one of the four climatological regimes. This shows the differences between scenarios in terms of the large-scale flow and provides information about the possible transitions between regimes during the forecast. This approach also enables the development of flow-dependent skill measures. The new cluster products were implemented operation- ally in November 2010. This article describes the new clustering, introduces the new cluster products and provides information on how to use them. Validation of the new clustering is also addressed.