1. Decadal Forecast Exchange
Doug Smith, Adam Scaife, G. J. Boer, M. Caian, C. Cassou, F. J. Doblas-Reyes, V. Guemas, E. Guilyardi, E. Hawkins, W. Hazeleger, L. Hermanson, C. K. Ho, M. Ishii, V. Kharin, M. Kimoto, B. Kirtman, J. Lean, D. Matei, W. J. Merryfield, J. Mignot, W. A. Müller, H. Pohlmann, A. Rosati, C. Severijns, D. Swingedouw, L. Terray, B. Wouters and K. Wyser
© Crown copyright Met Office

2. Many groups have now developed decadal predictions
Key experiments have been done for CMIP5
What about real time predictions?
15th session of the WMO Commission for Climatology recommended action to start the coordination and exchange of decadal predictions
Informal exchange of decadal predictions has been coordinated by Met Office since 2011
9 dynamical systems + 2 empirical
Annual mean fields of surface temperature, pressure and precipitation (to be expanded…)
Exchanged each year in March
© Crown copyright Met Office

3. Max Planck Institute (Germany)
2011
2012
2013
2014
Uninitialised
Hadley
UK Met Office 
EC-Earth anomaly
SMHI (Sweden)
EC-Earth full field
IC3 (Spain)
KNMI (Holland)
MRI
JMA (Japan)
MIROC
University of Tokyo
MPI
Max Planck Institute (Germany)
CCCMA
Canada
CERFACS
France
IPSL
GFDL
USA
RSMAS
University of Miami, USA
NCEP
NCAR
NorESM
Norway
NRL
USA (empirical)
Reading
University of Reading (empirical)
© Crown copyright Met Office

4.
© Crown copyright Met Office

5. © Crown copyright Met Office

6. Surface temperature predictions (five year means)
Skill of initialised predictions
Initialised - Uninitialised
Skilful almost everywhere (positive correlations)
Mostly due to external forcing
Initialisation gives improved skill mainly in North Atlantic and tropical Pacific
(Smith et al. 2010, also Doblas-Reyes et al 2013 for CMIP5 multi-model)
© Crown copyright Met Office
© Crown copyright Met Office

7. CMIP5 decadal predictions
Predictions (yr 2-5) from 6 CMIP5 systems
Initialized solid, uninitialised dashed
Observations grey bars
Global-mean T and Atlantic multi- decadal variability
Correlations and RMSE below
Initialisation improves skill, especially for AMV
Forecast lead time
Doblas-Reyes et al. (2013)

8. Physical basis for improved skill
No historical observations – must rely on models
Consistent signal: increase from 1960 to 1995, decrease thereafter
Agrees with related observations
Some skill in initialised predictions, but not in uninitialised predictions
Initialised hindcasts
Uninitialised hindcasts
© Crown copyright Met Office
(Pohlmann et al. 2013)
© Crown copyright Met Office

9. Atlantic multi-decadal variability (AMV)
Smith et al, 2011
© Crown copyright Met Office

10. Atlantic tropical storms
Initialised
Uninitialised
Persistence
Skill from both initialisation and external forcings
Improved through initialisation
Consistent with remote influences from improved SST predictions
1995 transition could have been predicted
(Smith et al. 2010, 2014)
© Crown copyright Met Office
© Crown copyright Met Office

11. Remote influences on tropical Atlantic atmosphere
Idealised predictions of rainfall 2 – 5 years ahead
Tropical Atlantic atmosphere is relatively predictable
Skill originates from sub-polar North Atlantic
© Crown copyright Met Office
© Crown copyright Met Office
Dunstone et al, 2011

12. North Atlantic sub-polar gyre (SPG)
Meridional heat transport
Overturning circulation
SPG 500m temp
Observations
Initialised (DePreSys)
Uninitialised (NoAssim)
Improved skill for 1995 rapid warming results from initialisation of increased Atlantic overturning circulation and meridional heat transport
(Robson et al. 2012, also Yeager et al. 2012, and Robson et al 2014 for 1960s cooling event)
© Crown copyright Met Office
© Crown copyright Met Office

13. Impacts of 1995 SPG warming Model Observations Temperature Pressure
Precipitation
© Crown copyright Met Office
(Robson et al. 2013)

14. Predicted cooling of SPG
SPG predicted to cool…
…in response to weakening of Atlantic overturning
Likely to cause climate impacts around the Atlantic basin
Not a reversal, but impacts associated with warm SPG less likely:
cold winters and wet summers in Europe less likely
fewer hurricanes than recent peaks
reduced Sahel rainfall
reduced risk of drought in SW USA
(Hermanson et al, submitted)
© Crown copyright Met Office

15. Observed weakening of AMOC
(Smeed et al, 2013)
© Crown copyright Met Office

16. Ratio of predictable components in reality and models (RPC), years 2-5
Temperature
Pressure
RPC<1 (blue)  models overconfident (agree with each other but not with reality)
RPC>1 (red)  models under confident (unexpected!)
Implies reality is more predictable than models  models respond too weakly to SSTs?
Members are not potential realisations of reality  affects skill and predictability assessment
Can make skilful predictions now, but need mean of large ensemble and to adjust variance
Higher skill possible with improved models
(Eade et al, in prep)
© Crown copyright Met Office

17. Summary
CMIP5 enables assessment of historical performance of decadal predictions
Need to focus on actual predictions as well
Informal exchange of decadal predictions started
Needs formalising under WMO to support GFCS
Encouraging skill in North Atlantic (including hurricanes), related to AMOC
Expect impacts over land, but too weak in models
Models underestimate predictability  respond too weakly to SSTs?
AMOC is weakening now (observations)
predicted to continue
climate impacts likely
© Crown copyright Met Office

18. Potential role of aerosols…
North Atlantic SST
Observations
Model
Booth et al.,
2012
Dunstone et al., 2013
© Crown copyright Met Office

19. Predicted cooling of SPG: JJA impacts
(Hermanson et al, submitted)
© Crown copyright Met Office

20. Impact of initialisation 2012-16
Initialised – uninitialised, stippled where not significant (Smith et al 2013)
© Crown copyright Met Office