Abstract ID: 156
A weather regimes approach for identifying increased predictability in the subseasonal prediction of European winters
Lead Author: Ignazio Giuntoli
CNR-ISAC, Italy
Keywords: Weather-regimes, Sub-seasonal prediction, S2S, ECMWF
Abstract: Forecasting surface variables at the sub-seasonal scale has gathered increasing attention in recent years: reliable and accurate subseasonal forecasts could benefit human livelihoods and a wide range of economic and agricultural activities. However, for a given start date, the skill of a forecast system lowers with time and drops particularly beyond week two. The skill may vary considerably depending on the season and synoptic circulations, with windows of opportunity for which the forecast system has greater chance of yielding a good prediction. Identifying these windows a priori is nontrivial, but a possibility in this direction is to exploit through a weather regime approach the information coming from atmospheric fields. In this context, we analyse systematically how well the reforecasts of the ECMWF S2S forecasting system predict cold spells in Europe and how weather regimes can help in gathering information preemptively on forecast skill. We consider 20 extended winters (1999-2018) and use the ERA5 reanalysis as the reference. After extracting weather regimes from reanalysis and reforecasts a similarity index is introduced to define how well the reforecasts regimes reproduce the reanalysis’ ones. The reforecasts with good similarity are then carried forward and their ability to predict weekly surface temperature is assessed, following the hypothesis that when the forecast system captures the weather regimes of the reanalysis the prediction skill may improve (i.e. a possible window of opportunity is caught). Generally, there is a substantial improvement of the forecast skill of reforecast subset over of the whole reforecast set. This occurs particularly when a given regime persists in time for over a week. Finally, we identify the composite circulation patterns of good (bad) similarity reforecasts to describe the initial conditions that favour (hinder) extended predictability. This type of information could be transferred into an operational context to improve confidence in real-time forecasts.
Co-authors:
Daniele Mastrangelo (Institute of Atmospheric Sciences and Climate (CNR-ISAC))
