Prospects for Wintertime European Seasonal Prediction Warwick Norton Walker Institute (University of Reading) & PCE Investors
North Atlantic Oscillation
Outline Predictability/persistence of the NAO on intraseasonal timescales (10-30 days). Predictability of the NAO on seasonal timescales with 1-6 month lead time. Examine reanalysis datasets and HiGEM - a new high resolution coupled atmosphere/ocean model. Contrast winter & summer
HiGEM Model Based on the new Hadley Centre HadGEM1 climate model HiGEM Atmosphere: horizontal resolution 1 (N144) 38 z-levels non-hydrostatic semi-Lagrangian dynamics aerosols HiGEM Ocean: horizontal resolution 1/3 40 z-levels with analytic stretching for 2nd order accuracy HiGEM Ice: EVP scheme with 5 ice categories 50 year control integration has just been completed
The UK in HiGEM
1st EOF of winter (DJF) monthly MSLP (NAO) ERA40 - explains 38% of total variance HiGEM - explains 34% of total variance
Projection of NAO pattern onto daily MSLP gives daily NAO timeseries ERA40 HiGEM Lack of persistence at days 10-30 in HiGEM
1st EOF of summer (JJA) monthly MSLP (NAO) ERA40 - explains 30% of total variance HiGEM - explains 28% of total variance
Projection of NAO pattern onto daily MSLP gives daily NAO timeseries HiGEM ERA40 Less persistence of NAO in summer, HiGEM ok
Middle Atmosphere GCM experiment where middle/upper Stratosphere is damped (Norton 2003) North Pole Temperatures at 30 km damped control No stratospheric warmings in strong drag experiment
Autocorrelation of daily surface AO index Control experiment Strong drag experiment Reduction in memory of surface AO from less variable stratosphere Supports the hypothesis that lack of NAO persistence in HiGEM is due to inadequate representation of the stratosphere.
Seasonal DJF Mean NAO Significant low frequency variability with correlation at lag 1 year. What is the source of this? Is it predictable?
Seasonal JJA Mean NAO Smaller amplitude than in winter Very little low frequency variability
Correlation of May SSTs with following Dec-Jan Central England Temperatures Non-ENSO years only (33) All years 1948-2004 (see also Rodwell & Folland, 2002; Czaja & Frankignoul, 2002)
May Sea Surface Temperatures Composite of: 1958, 1962, 1966, 1969, 1980, 1995 (region inside solid contour is different from mean at 95% confidence interval) 2005
Dec-Jan “Composite” Forecast Atlantic SST anomaly has persisted Cold temperatures centred over Eastern Europe Dry conditions in the North-West Atlantic extending into the UK
Dec-Jan 2005/6 Atlantic SST anomalies have persisted Cold temperatures centred over Eastern Europe Very warm over Canada - influence of weak La Nina? Dry conditions in the North-West Atlantic extending into the UK
2m Air Temperatures - Berlin Years: 1962, 1966, 1969, 1978, 1980, 1981, 1995, 1996, 2005/6 (Long term trend removed)
Dec-Jan Atmosphere GCM Experiment (low resolution HadAM3) Reinforcement by cold SSTs . . Circulation anomaly (anticyclone) Precipitation anomaly Rossby wave propagation into Europe
How does HiGEM do in getting the SST signal? Correlation of May SSTs with following Dec-Jan Central England Temperatures HiGEM Observations No signal in the subtropics
Correlation of November SSTs with following Dec-Jan Central England Temperatures HiGEM Observations No signal in the subtropics
Conclusions Representation of the stratosphere appears to be important for persistence of NAO in winter. Atlantic SSTs can force the NAO in ENSO neutral years. Coupled atmosphere/ocean models only partially represent the NAO/Atlantic SST relationship. Getting the Pacific SSTs right is important but an ENSO event happens only ~50% of years. The tropical/subtropical Atlantic SSTs are also important for seasonal prediction (and hurricanes).