1. Research Needs for Decadal to Centennial Climate Prediction: From observations to modelling Julia Slingo, Met Office, Exeter, UK & V. Ramaswamy. GFDL, Princeton, USA

2. Climate Change Projections and Uncertainties IPCC AR4

3. 2020’s2080’s Winter rainfall in south east England Improved model physics e.g. clouds Increased understanding of earth system processes – more uncertainty? Benefits of initialisation for near-term projections Quantifying uncertainties – setting research priorities Higher resolution global models

4. Challenges for Centennial Projections: Earth System Modelling

5. Land physics and hydrology Ocean circulation Atmospheric circulation and radiation Land physics and hydrology Ocean ecology and biogeochemistry Atmospheric circulation and radiation Allows interactive CO 2 Ocean circulation Plant ecology and land use Climate Model Earth System Model Sea Ice Moving from Climate to Earth System Models: Balancing the carbon cycle

6. Carbon-climate feedback and centennial climate change

7. More Earth System Modelling challenges How can we reduce the uncertainties in current estimates of the carbon-climate feedback? How do missing or poorly represented processes such as the nitrogen cycle, plant adaptation to climate change, vegetation dynamics, and plankton dynamics affect current model results? What other biogeochemical feedbacks involving methane, ozone and aerosols play a significant role on the centennial timescale? How can Earth System Modelling inform decision-making when climate change is one of many drivers for environmental change (e.g. food security, water resources and quality, biodiversity, air quality)?

8. Earth System Modelling: Combining the needs of adaptation and mitigation

9. 2020’s2080’s Winter rainfall in south east England Improved model physics e.g. clouds Increased understanding of earth system processes – more uncertainty? Benefits of initialisation for near-term projections Quantifying uncertainties – setting research priorities Higher resolution global models

10. Challenges for Decadal Prediction: Initialisation and Evaluation

11. Decadal predictions of global annual mean surface temperature Observations Forecast/hindcast Forecast from 2008 Forecast from 2009 Smith et al., 2007

12. Impact of initialisation on hindcast skill 5 year mean (JJASON) surface temp 15x15 degrees DePreSys-NoAssim correlationDePreSys anomaly correlation HadCM3 9 member perturbed physics ensemble Starting every Nov from 1960 to 2005

13. Improved predictions of multi-year Atlantic hurricane frequencies Normalised anomaly DePreSysNoAssim 5-year mean JJASON number of model storms (coloured) and observed hurricanes (black) Skill comes from SSTs in tropical Pacific and N. Atlantic sub- polar gyres, and from wind shear in hurricane development regions

14. Sub-surface ocean observations: A limiting factor in estimating skill and predictability 1980 1960 2007 Need historical tests to assess likely skill of forecasts Far fewer sub-surface ocean observations in the past Doug Smith, Met Office Hadley Centre

15. Temperature at 300m : June 2007 from 1960 observational base Analysis using all obsAnalysis using sub-sampled (1960) obs June 2007 obs June 1960 obs

16. Variability versus Anthropogenic Forcing of the Physical Climate System

17. 20 centuries of NINO3 SSTs annual means & 20yr low-pass

18. Major uncertainty in Chemistry-Climate Interactions

19. Land OceanSea Ice Mixed-Layer Deep Ocean SST Surface Flux Clear SkyCloudy Sky Aerosols Droplets Activation SW Radiation LW Radiation Evaporation Precipitation Atmosphere Coupled Chemistry-Aerosol-Climate model Aerosols and Climate Global Air Quality and Climate

20. Aerosol-Cloud Interactions in GFDL’s Newest Physical Climate Model (CM3) ‏ 20 CM3CM2.1 Direct effects – Sulfate and organic carbon ~0 (assuming internal mixing of sulfate and black carbon) -1.3 (external mixing) Direct effects - Black carbon 0.5 (external mixing) Indirect effects -1.3 Not included Radiative Flux Perturbation w/m2 Comparison of Simulated Aerosol Properties with Observations Observations (AERONET) MODELMODEL MODELMODEL

21. Capturing High-Resolution Phenomena

24. Atlantic Hurricanes in a Warming World

25. Most recent GFDL downscaling study (Bender et al, Science, 2010) see https://www.gfdl.noaa.gov/21st-century-projections-of-intense-hurricanes Uses two downscaling steps: Global CMIP3 models => regional model of Atlantic hurricane season regional model => operational GFDL hurricane prediction system

26. Conclusion: Best estimate is for doubling of cat 4-5 storms in Atlantic by end of century, despite decrease in total number of tropical cyclones Much of the uncertainty arises from global model input

27. Conclusions I Emerging need for centennial and decadal projections. They pose similar and differing challenges. Earth system processes potentially increase uncertainty in centennial projections, especially in the upper range of warming. Initialising decadal projections can reduce uncertainty at least for a few years ahead.

28. Conclusions II Observations of the sub-surface ocean and the full earth system may limit our ability to provide more confident projections. Natural variability in the context of forced change is challenging. High resolution modelling is opening up new avenues for more detailed projections of regional climate change and high impact phenomena.