1. Integrating Climate Science into Adaptation Actions Alberto Arribas (alberto.arribas@metoffice.gov.uk)alberto.arribas@metoffice.gov.uk Kuala Lumpur, November 2012

2. Linking forecasting timescales and model development - Weather prediction to provide the high spatial and temporal detail (e.g. 4 km, hourly data) at short-range. Atmosphere-only models - Monthly-to-Seasonal predictions to provide early warnings with lower spatial and temporal detail (e.g. 50-100 km, weekly-monthly data). Coupled models (ocean, atmosphere, sea-ice, land-surface, etc)

3. And longer … projected changes 2040s-2100s (IPCC SREX report, 2012) 20 th Century Return period (years) More frequent extreme precipitation events

4. © Crown copyright Met Office Over the next 20-30 years, climate variability is even more important Climate variability can greatly amplify or oppose any trend: Tropical Floods during 2010/11 Russian heatwave 2010 African Drought 2011 Recent Cold European winters… Temperature Time (years) Climate Change Climate Change + Variability Flooding at Toowoomba, Australia, 2011Barcelona, Spain, March 2010Dry Water Pan, Kenya, 2011

5. © Crown copyright Met Office Initial (weather) and Boundary (climate) problem:  Predictability comes from slowly varying processes (ocean, soil moisture, sea-ice, green house gases, etc) … We need more complex models than for weather prediction  ALL relevant processes and teleconnections have to be well represented and initialised to have useful skill Seasonal Forecasting: A complex forecasting problem crucial for model development

6. Similarly to NWP, we need to initialise and run a forecast model but: (a)We need a coupled model (ocean/atmosphere/land- surface/sea-ice) (b)We need to initialise all components (c)We need to run the model for longer: (c.1) Larger spatial / longer temporal averages (c.2) Ensemble prediction (probabilistic forecast) (c.3) Output needs to be bias corrected (c.4) Skill needs to be estimated Hindcast: what is it? How do we do monthly- seasonal forecasts?

7. © Crown copyright Met Office How do we do monthly- seasonal forecasts? 5 th Nov. 1996 9 members 1997 9 members 2010 9 members 5 th Nov. 2012 40 members Plenty of model simulations, every week:

8. Nino 3.4 SST

9. Hindcast: a collection of forecasts of the past (1996-2009) Model Bias: SST in JJA

10. Hindcast: a collection of forecasts of the past (1996-2009) Bias corrected forecast obs Forecast members obs Hindcast mean Forecast members All Hindcast runs: ~ 12 members 14 years (96-09) Hcst mean Raw forecast

11. © Crown copyright Met Office Are monthly-to-seasonal forecasts good enough for early warnings and disaster risk management?

12. Seasonal Forecasting: A new strategy to increase skill Hindcast length Frequency of system upgrades Centre's priorities NCEP (USA)~ 40 yr8 yr Link to re-analysis ECWMF~ 25 yr5 yr Med-range UK Met Office * 14 yr 1 yr Link to model development * Arribas et al., 2011: GloSea4 ensemble prediction system for seasonal forecasting. MWR. 139, 1891-1910

13. A recent history of improvements at UK Met Office -Summer 2009: New generation prediction system (linked to model development) becomes operational -Nov. 2010: -Vertical high-res (L85 stratosphere. / L75 ocean) -Sea-ice assimilation -May 2011: -Extension to Monthly system -Nov. 2012: -Horizontal high-res (50 km atm. / 0.25 ocn) -NEMOVAR – 3d-Var Ocean Data Assimilation

14. Representation of orography ~ 120 km ~ 50 km

15. GloSea5 operational system Model version: HadGEM3 GA3.0 Resolution: N216L85 O(.25)L75 (~50 km atm.) Simulations length: 7 months Model uncertainties represented by: SKEB2 stochastic physics (Tennant et al. 2011) Initial conditions uncertainties represented by: Lagged ensemble

16. Initialisation of the system Forecast (initialised daily): - Atmosphere & land surf: Met Office NWP analysis (4d-Var) - Ocean & sea-ice: NEMOVAR (3d-Var joint system for ocean, med-range, monthly and seasonal) 14-year Hindcast (1996-2009): - Atmosphere & land surf: ERA-interim - Ocean & sea-ice: Seasonal ODA reanalysis - Fixed start dates of 1 st, 9 th, 17 th, 25 th of each month - 3 members per start date

17. Ensemble: lagged approach Seasonal Forecast: - 2 members run each day. - Seasonal forecast updated weekly by pulling together last 3 weeks (i.e. 42 members) Hindcast (for monthly-seasonal): 14 year hindcast run in real time ( 42 members run each week = 14 years x 3 members) Monthly Forecast: - 2 additional members run each day. - Monthly Forecast updated daily by pulling together last 7 days (i.e. 28 members)

18. 20/06/2011 How the system runs, an example Atmos & land surf: NWP anal Ocean/sea-ice : Seasonal ODA Atmos & land surf: ERA-i Ocean: Seasonal ODA reanalysis 25/07/1996 (m1) 25/07/1997 (m1) 25/07/1998 (m1) 25/07/1999 (m1) 25/07/2000 (m1) 25/07/2001 (m1) Monday 21/06/2011 25/07/2002 (m1) 25/07/2003 (m1) 25/07/2004 (m1) 25/07/2005 (m1) 25/07/2006 (m1) 25/07/2007 (m1) Tuesday 26/06/2011 25/07/2004 (m3) 25/07/2005 (m3) 25/07/2006 (m3) 25/07/2007 (m3) 25/07/2008 (m3) 25/07/2009 (m3) Sunday Each week: 14x 7-month forecasts, 14x 2-month forecasts (for monthly forecast) and 42x 7-month hindcasts (1996-2009) 20/06/2011 21/06/2011 26/06/2011

19. Improving ENSO forecasts Obs The westward extension of Nino is a common error in many climate models. It affects remote regions. High-res model has better ENSO pattern and teleconnections Low resolution High resolution

20. Nino 3.4 SST:ACC / RMSE&Spread ACC higher (good) RMSE reduced (also good) May  JJANov  DJF GloSea5 (red) GloSea4 (blue)

21. © Crown copyright Met Office JJA DJF ForecastObserved Better ENSO Teleconnection: Prec. Nino - Nina

22. MJO

24. MJO correlations with lead time

25. NAO

26. Benefits of higher resolution: Improved Atlantic Blocking Gulf Stream Bias Wly wind bias => Blocking Deficit No Gulf Stream Bias No Wly wind bias => Good Blocking in N. Atl New Model Scaife et al., Geophys. Res. Lett., 2012. Low-res: 1 deg ocean High-res: 0.25 deg ocean

27. Significant skill for NAO prediction! First time we get significant skill (ACC 0.5) Our previous system had corr. values of 0.2 (Japan/ECMWF near 0)

28. PMSL anomalies (from Nov for DJF)

29. WNPSH

30. International collaboration to improve prediction systems Working with Chinese Meteorological Agency on West North Pacific Subtropical High

31. GPCP Composite rainfall with strong WNPSH Importance of West North Pacific Subtropical High

32. Obs Previous System New System The variability of the WNPSH is much improved in the latest system

33. SH index and rainfall Correlations with observations: Previous System =0.41 ---- New System=0.83 Skill predicting interannual variability of West North Pacific Subtropical High

34. SH index and rainfall Skill predicting interannual variability of rainfall over the Yangtse River Valley Correlations with observations: Previous System = 0.35 ---- New System= 0.69

35. Sector specific applications: Lake Volta, Ghana © Crown copyright Met Office Corr. = 0.69 June forecasts of total July-Oct. inflow Preceding rainfall and flow predictors plus seasonal forecast predictors Fcst Obs

36. Managing risk: precipitation over SE Asia, summer 1998 Obs 9 Forecasts

37. Managing risk: precipitation over SE Asia, summer 1998

38. An international prediction system KMA KMA (Rep. of Korea) Joint seasonal forecast system Shared workload and computing costs: possibility to extend hindcast and increase resolution NCMRWF NCMRWF (India) Implementing GloSea for research

39. Seamless system across timescales GloSea5 Med-range (2013) Project to merge with med-range in 2013 Aim is to have a single operational system (using coupled model at the highest possible resolution) for short-range ocean, med-range, monthly and seasonal at the end of 2013 GloSea5 Decadal (2014) System to be extended – in research mode - to decadal timescales in 2013 Seamless system med-range to decadal from 2014

40. Conclusions - Met Office new generation monthly- seasonal prediction system shows useful skill months ahead This is a problem in need of international solutions: - International collaboration to further improve prediction systems - Sustainable dissemination of information - In country development of sector-specific applications

41. Thanks

42. © Crown copyright Met Office How do we predict climate variability months ahead? Are these forecasts good enough? Can they be useful for risk management? 5 th Nov. 1996 9 members 1997 9 members 2010 9 members 5 th Nov. 2012 40 members Plenty of model simulations, every week:

43. © Crown copyright Met Office How do we predict climate variability months ahead? Are these forecasts good enough? Can they be useful for risk management? 5 th Nov. 1996 9 members 1997 9 members 2010 9 members 5 th Nov. 2012 40 members Plenty of model simulations, every week: