Dynamical Season Prediction: A Personal Retrospective of the Past 30 Years (1975-2004), and Conjectures about the Future

1. Dynamical Season Prediction: A Personal Retrospective of the Past 30 Years (1975-2004), and Conjectures about the Future J. ShuklaGeorge Mason University (GMU)Center for Ocean-Land-Atmosphere Studies (COLA)with contributions from:J. Kinter (COLA)Symposium on the 50th Anniversary of Operational Numerical Weather PredictionUniversity of Maryland, June 14-17, 2004

2. Outline • Historical Overview: The 50 years Preceding JNWP50 • International Contributions to NWP • The First 90-day Integration of the NMC Forecast Model • DERF: NMC-COLA Collaboration (1983-1984) • From NWP to DSP to Coupled Model Prediction • Dynamical Seasonal Prediction: The Current Status • Dynamical Seasonal Prediction: Future Prospects • Conclusions, Conjectures and Suggestions

3. The 50 Years Preceding JNWP50 • V. Bjerknes (1904) Equations of Motion • Father of J. Bjerknes, son and research assistant of C. Bjerknes (Hertz, Helmholtz) • L. F. Richardson (1922) Manual Numerical Weather Prediction • Military background, later a pacifist, estimated death toll in wars • C. G. Rossby (1939) Barotropic Vorticity Equation • First “Synoptic and Dynamic” Meteorologist; Founder of Meteorology Programs at MIT, Chicago, Stockholm • J. Charney (1949) Filtered Dynamical Equations for NWP • First Ph.D. student at UCLA; Chicago, Oslo, Institute for Advanced Study, MIT • N. A. Phillips (1956) General Circulation Model • Father of Climate Modeling; Chicago, Institute for Advanced Study, MIT

4. Global Contributions Towards Research on Predictability and Prediction of Weather • USA: Predictability: Charney et al., Lorenz; NWP: Cressman, Phillips, Miyakoda • Canada: Numerical methods: Robert; Data assimilation: Daley • Australia: Spectral model: Bourke • France: Data assimilation: Talagrand • U.K.: Theory: Eady; NWP: Sutcliff, Sawyer • Germany: Theory: Ertel; NWP: Hinkelmann • Norway: Theory: Eliassen • Russia: Theory: Obukhov, Monin, Kibel; Adjoint: Marchuk Data assimilation: Gandin • Japan: NWP: Fujiwara, Syono, Gambo • Sweden: Initialization: Machenauer; NWP: Bengtsson

5. Weather Predictability and Prediction • Predictability and theory: Charney et al., Lorenz, Eady; Ertel, Eliassen, Obukhov, Monin, Kibel • NWP: Cressman, Phillips, Miyakoda, Hinkelmann, Sutcliff, Sawyer, Syono, Gambo, Bengtsson • Numerical methods: Robert, Bourke, Marchuk • Data assimilation: Daley, Talagrand, Gandin • Initialization: Machenauer, Baer and Tribbia • Physical parameterizations - Convection, Radiation, Boundary Layer, Clouds, etc. • Ensembles: Farrell, Kalnay, Palmer, Toth

6. The First 90-day Integration of the NMC Forecast Model DERF: NMC-COLA Collaboration (1983-1984) • Meeting with Bonner, Rasmusson, Phillips and Brown (3 Oct 1983) • Statement of Intent for NMC-COLA Work on DERF (14 Feb 1984) • Acronym “DERF” created by Gerrity (24 Aug 1984) • NMC Committee on DERF created • Tracton Named CAC DERF Project Leader (11 Jun 1985) • Large Number of NMC Scientists Involved in DERF • Major Logistical Arrangements Required to Make 90-day Run First 90-day Run of NMC Model Approved by Brown (30 Sep 1985)

11. Monthly and Seasonal Predictability and Prediction • Dynamical Predictability: Shukla (1981, 1984), Miyakoda, Gordon, Caverly, Stern, Sirutis, and Bourke (1983) • Boundary-Forced Predictability: Charney and Shukla (1977, 1981), Shukla (1984) • Theory: Hoskins and Karoly (1981), Webster (1972, 1981) • Programs: PROVOST (Europe); DSP (USA); SMIP (WCRP)

12. Simulation of (Uncoupled) Boundary-Forced Response: Ocean, Land and Atmosphere INFLUENCE OF LAND ON ATMOSPHERE • Mountain / No-Mountain • Forest / No-Forest (Deforestation) • Surface Albedo (Desertification) • Soil Wetness • Surface Roughness • Vegetation • Snow Cover INFLUENCE OF OCEAN ON ATMOSPHERE • Tropical Pacific SST • Arabian Sea SST • North Pacific SST • Tropical Atlantic SST • North Atlantic SST • Sea Ice • Global SST (MIPs)

13. From Numerical Weather Prediction (NWP) To Dynamical Seasonal Prediction (DSP) (1975-2004) • Operational Short-Range NWP: was already in place • 15-day & 30-day Mean Forecasts: demonstrated by Miyakoda (basis for creating ECMWF-10 days) • Dynamical Predictability of Monthly Means: demonstrated by analysis of variance • Boundary Forcing: predictability of monthly & seasonal means (Charney & Shukla) • AGCM Experiments: prescribed SST, soil wetness, & snow to explain observed atmospheric circulation anomalies • OGCM Experiments: prescribed observed surface wind to simulate tropical Pacific sea level & SST (Busalacchi & O’Brien; Philander & Seigel) • Prediction of ENSO: simple coupled ocean-atmosphere model (Cane, Zebiak) • Coupled Ocean-Land-Atmosphere Models: predict short-term climate fluctuations

14. Evolution of Climate Models 1980-2000 Model-simulated and observed rainfall anomaly (mm day-1) 1983 minus 1989

15. Evolution of Climate Models 1980-2000 Model-simulated and observed 500 hPa height anomaly (m) 1983 minus 1989

16. Vintage 2000 AGCM

17. Observed and Simulated Surface Temperature (°C)

18. Cross-Validated CCA of Z500 & SST (Observed and Modeled)

20. Variance of Model-Simulated Seasonal (JFM) Rainfall (mm2)

21. Predictability of the Coupled Climate System

22. Standard Deviation of Monthly Equatorial Pacific SSTA Observations Forecast (JUL ICs) Simulation COLA Predictions (1980-1999) COLA Coupled Simulation (250 years) GFDL MOM3 ODA (1980-1999)

23. “Operational” ENSO Prediction with Coupled A-O GCMs Courtesy of A. Barnston, IRI and B. Kirtman, GMU/COLA

24. “Operational” ENSO Prediction with Coupled A-O GCMs Courtesy of A. Barnston, IRI and B. Kirtman, GMU/COLA

25. Current Limit of Predictability of ENSO (Nino3.4) Potential Limit of Predictability of ENSO 20 Years: 1980-1999 4 Times per Year: Jan., Apr., Jul., Oct. 6 Member Ensembles Kirtman, 2003

26. Impact of Ensemble Size

27. Factors Limiting Predictability: Future Challenges

28. Challenges Conceptual/Theoretical Modeling Observational Computational Institutional Applications for Benefit to Society

29. Challenges Conceptual/Theoretical ENSO: unstable oscillator? ENSO: stochastically forced, damped linear system? (The past 50 years of observations support both theories) • Role of weather noise? Modeling • Systematic errors of coupled models - too large • Uncoupled models not appropriate to simulate Nature in some regions/seasons: CLIMATE IS A COUPLED PROCESS • Atmospheric response to warm and cold ENSO events is nonlinear (SST, rainfall and circulation) • Distinction between ENSO-forced and internal dynamics variability

30. Challenges Observational • Observations of ocean variability • Initialization of coupled models Computational • Very high resolution models of climate system need million fold increases in computing • Storage, retrieval and analysis of huge model outputs • Power (cooling) and space requirements-too large

31. Challenges Institutional • Development of accurate climate (O-L-A) models, assimilation and initialization techniques, require a dedicated team with a critical mass of scientists (~200) and resources (~$100 million per year: $50M computing; $30M research; $20M experiments) • Climate modeling and prediction efforts should be 10 times NWP but is currently only ~10% of NWP Applications for Benefit to Society • Educate the consumers about the limits of predictability (uncertainty and unreliability) • Decision making and risk management using probabilistic predictions

32. Inconsistency of SST and Precip in the W. Pacific - Prescribed SST

33. Models Today Weather T254: 5 d/hr on 144 CPUs T511: 2.5 d/hr on 288 CPUs Climate T85/ 1°: 2.0 yrs/d on 96 CPUs 2°X2.5°/1°: 5.25 yrs/d on 180 CPUs Models in 2014 Weather T3800 (5 km): 4 d/hr (2,160 CPUs) - or - T825 (25 km): 4 d/hr (468 CPUs) Climate T420/ 0.5°: 2.4 yrs/d (2,500 CPUs) -or- T420/0.5°: 2 mo/d (2,500 CPUs) Climate Modeling and Computing {Moore’s Law (43%/yr)-OR- 10%/yr}

36. Conclusions, Conjectures and Suggestions

37. Conclusions, Conjectures and Suggestions • The estimates of the growth rate of initial errors in NWP models is well known, and the current limits of predictability of weather are well documented. The most promising way to improve forecasts for days 2-15 is to improve the forecast at day 1. • The limits of predictability for short-term climate predictions (seasons 1-4), are not well known, because the estimates of predictability remain model-dependent. Our ability to make more accurate seasonal predictions is limited by: • Inadequate understanding of coupled dynamics • Insufficient observations • Inaccurate models • Insufficient computing • Inefficient institutional arrangements

38. Conclusions, Conjectures and Suggestions • During the past 25 years, the weather forecast error at day 1 has been reduced by more than 50%. At present, forecasts for day 4 are, in general, as good as forecasts for day 2 made 25 years ago. • With improved observations, better models and faster computers, it is reasonable to expect that the forecast error at day 1 will be further reduced by 50% during the next 10-20 years. Therefore, at that time, the forecasts at day 3 could be as good as forecasts for day 2 are today.

39. Conclusions, Conjectures and Suggestions • 25 years ago, a dynamical seasonal climate prediction was not conceivable. • In the past 20 years, dynamical seasonal climate prediction has achieved a level of skill that is considered useful for some societal applications. However, such successes are limited to periods of large, persistent anomalies at the Earth’s surface. Dynamical seasonal predictions for one month lead are not yet superior to statistical forecasts. • There is significant unrealized seasonal predictability.Progress in dynamical seasonal prediction in the future depends critically on improvement of coupled ocean-atmosphere-land models, improved observations, and the ability to assimilate those observations.

40. Conclusions, Conjectures and Suggestions • Improvements in dynamical weather prediction over the past 30 years did not occur because of any major scientific breakthroughs in our understanding of the physics or dynamics of the atmosphere • Dynamical weather prediction is challenging: progress takes place slowly and through a great deal of hard work that is not necessarily scientifically stimulating, performed in an environment that is characterized by frequent setbacks and constant criticism by a wide range of consumers and clients • Nevertheless, scientists worldwide have made tremendous progress in improving the skill of weather forecasts by advances in data assimilation, improved parameterizations, improvements in numerical techniques and increases in model resolution and computing power

41. Conclusions, Conjectures and Suggestions • Currently, about 10 centers worldwide are making dynamical weather forecasts every day with a lead time of 5-15 days with about 5-50 ensemble members, so that there are about500,000 daily weather mapsthat can be verifiedeach year • It is this process of routine verification by a large number of scientists worldwide, followed by attempts to improve the models and data assimilation systems, that has been the critical element in the improvement of dynamical weather forecasts • In contrast, if we assume that dynamical seasonal predictions, with a lead time of 1-3 seasons, could be made by about 10 centers worldwide every month with about 10-20 ensemble members, there would be less than5,000 seasonal mean predictionsworldwide that can be verifiedeach year • This is a factor of 100 fewer cases compared to NWP, so improvement in dynamical seasonal prediction might proceed at a pace that is much slower than that for NWP if we didn’t do something radically different

42. Conclusions, Conjectures and Suggestions • NWP (World Wide) • 10 Centers • 5-15 day forecasts each day • 5-15 ensemble size • 500,000 daily weather maps each year • DSP (World Wide) • 10 Center • 1-3 seasons predictions each month • 10-20 ensemble size • 5,000 seasonal maps each year DSP is a factor of 100 fewer cases than NWP

43. Conclusions, Conjectures and Suggestions Consumers could save $1 billion per year in energy costs if the average weather forecast could be improved by just 1º Fahrenheit. David S. Broder Washington Post, 22 April 2004 Excerpt from NOAA report in interview with Admiral Conrad Lautenbacher Under Secretary of Commerce, NOAA

44. Suggestion for Accelerating Progress in Modeling and Prediction of the Physical Climate System • There is a scientific basis for extending the successes of NWP to climate prediction • The problem is beyond a person, a center, a nation … • A multi-national collaboration is required

45. Suggestion for Accelerating Progress in Dynamical Seasonal Prediction Reanalyze and Reforecast the seasonal variations for the past 50 years, every year • Exercise state-of-the-art coupled ocean-atmosphere-land models and data assimilation systems for a large number of seasonal prediction cases and verify them against observations • Equivalent to producing reanalysis and 1-2 season dynamical forecasts for each month of one year, every week • Conduct model development experiments (sensitivity to parameterizations, resolution, coupling strategy, etc.) with the specific goal of reducing seasonal prediction errors

46. THANK YOU! ANY QUESTIONS?