Attributing Predictable Signals at Subseasonal Timescales (Invited Presentation)

Subseasonal forecasts offer significant economic value in the management of energy infrastructure and through the associated financial markets. Models are now accurate enough to provide, for some occasions, good forecasts in the subseasonal range. However it is often not clear what are the drivers of these subseasonal signals and if potentially the forecasts could be made more accurate with better representation of physical processes? Also what are the limits of predictability in the subseasonal range?  To address these questions we have rerun the ECMWF monthly forecast system over the 2015/16 winter with a set of 6 week ensemble integrations initialised every week. In these experiments we have relaxed the band 15N to 15S to reanalysis fields. Hence we have a set of forecasts where the tropics is constrained to what actually happened and we can analyse the changes in predictability in middle latitudes - in particular in regions of high energy consumption like North America and Europe.  Not surprisingly, the forecast of some periods are significantly improved while others show no improvement. We discuss what events/patterns have extended range predictability and also the tropical forecast errors which prevent the potential predictability in middle latitudes from being realised.