ECMWF has begun to make a new range of forecast charts based on the Sub-seasonal to Seasonal predictions (S2S) database available to the public, at www.ecmwf.int/en/research/projects/s2s/charts/s2s/.
The charts can be used to monitor the S2S data and assess the quality of the forecasts. They can also serve as a testbed for the development of new products, for example by helping to identify signals for extreme events at the sub-seasonal timescale. The charts include ensemble mean anomalies for a range of meteorological parameters, the Extreme Forecast Index (EFI) for 2-metre temperature, and forecasts of the Madden–Julian Oscillation (MJO). Since S2S is a research project, the forecasts are available with a three-week delay. They are not intended for operational use.
Currently the S2S charts are limited to six models. In future all 11 S2S models will be included and the range of products will be extended. Some centres produce their sub-seasonal forecasts on a daily basis while others produce theirs on a weekly basis. The horizontal and vertical resolution of the models and the ensemble size also vary greatly from one centre to another, and the forecasts cover a different number of years. Constructing products from very different forecasting systems is a challenge.
The S2S project
The Sub-seasonal to Seasonal prediction project (S2S), launched by the World Weather Research Programme (WWRP) and the World Climate Research Programme (WCRP), aims to improve forecast skill and understanding of the sub-seasonal to seasonal timescale and to promote its uptake by operational centres and its exploitation by the applications community. As part of this project, ECMWF is hosting a data portal for S2S forecasts containing near-real-time ensemble forecasts and re-forecasts up to 60 days from ten forecasting centres (11 by the end of 2016). Most of the forecasts are created by a coupled system (an atmospheric model coupled to an ocean model), and some include an active sea-ice model. For more information, visit http://www.s2sprediction.net.
In order to make the products comparable, we use the largest common period available (1999–2010) to estimate the model climate. Forecast products are issued once a week (every Thursday) to keep the number of charts to a minimum. Forecasts produced on a daily basis typically have a small ensemble size, but it is possible to construct a larger ensemble by pulling together forecast ensembles initiated on a number of subsequent days. How many daily ensembles should be included in one larger ensemble depends mainly on the forecast application. In order to keep product generation simple, we consider only the ensemble initiated on Thursdays.
The S2S chart web page contains charts for all forecasts since January 2016. Every week new forecasts are added to the list. With this forecast history it is possible to evaluate forecast performance for specific events (e.g. heat waves and wet spells); analyse consistency between different models; and test the benefits of a multi-model approach.
