Ocean Observations for an End-to-End Seasonal Forecasting System Magdalena A. Balmaseda Yosuke Fujii Oscar Alves Tong Lee Michele Rienecker Tony Rosati Detlef Stammer Yan Xue Howard Freeland Michael McPhaden Lisa Goddard Caio Coelho ECMWF (UK) MRI (Japan) CAWCR (Australia) NASA JPL (USA) GMAO, NASA (USA) NOAA/GFDL (USA) University of Hamburg (Germany) NCEP/NOAA(USA) FOC (Canada) PMEL/NOAA (USA) IRI (USA) CPTEC (Brazil) OceanObs 09, Venice 21-25 September 2009 1
Overview Why do we want forecast at seasonal time scales? Societal applications End To End Seasonal Forecasting Systems Role of ocean observations. Initialization Achievements and challenges Development of model and assimilation systems: Process studies, multivariate relationships Calibration and skill assessment providing meaningful forecasts from the numerical output. Recommendations OceanObs 09, Venice 21-25 September 2009 2
There is a clear demand for reliable seasonal forecasts: Forecasts of anomalous rainfall and temperature at 3-6 months ahead For a range of societal, governmental, economic applications: Agriculture Heath (malaria, dengue, ) Energy management Markets, insurance Water resource management, Huge progress in the last decade: Operational seasonal forecasts in several centres Pilot/Research progress for demonstrating applicability (DEMETER,IRI,EUROBRISA, ) Build-up of community infrastructure (at WMO level) OceanObs 09, Venice 21-25 September 2009 3
The basis for extended range forecasts Forcing by boundary conditions changes the atmospheric circulation, modifying the large scale patterns of temperature and rainfall, so that the probability of occurrence of certain events deviates significantly from climatology. Important to bear in mind the probabilistic nature of climate forecasts How long in advance?: from seasons to decades The possibility of seasonal forecasting has clearly been demonstrated Decadal forecasting activities are now starting. The boundary conditions have longer memory, thus contributing to the predictability. Important boundary forcing: SST: ENSO, Indian Ocean Dipole, Atlantic SST Land: snow depth, soil moisture Atmospheric composition: green house gases, aerosols, Ice? OceanObs 09, Venice 21-25 September 2009 4
End-To-End Seasonal forecasting System COUPLED MODEL Forecast PRODUCTS OCEAN ENSEMBLE GENERATION PROBABILISTIC CALIBRATED FORECAST FORECAST CLIMATE 20 E 40 E 60 E 80 E 100 E 120 E 140 E 160 E 180 160 W 140 W 120 W 100 W 80 W 60 W 40 W 20 W 80 N 80 N 70 N 70 N 60 N 60 N 50 N 50 N 40 N 40 N 30 N 30 N 20 N 20 N 2 2.9 25.1 26.2 9.8 13.3 14.3 10.3 10 N 10 N 0 0 10 S 10 S 20 S 20 S 30 S 30 S 40 S 40 S NINO3.4 SST anomaly plume 50 S 50 S ECMWF forecast from 1 Jan 2007 60 S 60 S Monthly mean anomalies relative to NCEP adjusted OIv2 1971-2000 climatology 70 S 70 S 80 S 80 S System 3 20 E 40 E 60 E 80 E 100 E 120 E 140 E 160 E 180 160 W 140 W 120 W 100 W 80 W 60 W 40 W 20 W 2 2 No Significance 90% Significance 95% Significance 99% Significance Anomaly (deg C) 1 1 0 0-1 -1 JUL 2006 AUG SEP OCT NOV DEC JAN 2007 Produced from real-time forecast data FEB MAR APR MAY JUN JUL AUG SEP Initialization Forward Integration Forecast Calibration OceanObs 09, Venice 21-25 September 2009 5
Importance of Initialization Atmospheric point of view: Boundary condition problem Forcing by lower boundary conditions changes the PDF of the atmospheric attractor Loaded dice Oceanic point of view: Initial value problem Prediction of tropical SST: need to initialize the ocean subsurface. o Emphasis on the thermal structure of the upper ocean o Predictability is due to higher heat capacity and predictable dynamics A simple way: ocean model + surface fluxes. o But uncertainty in the fluxes is too large to constrain the solution. Alternative : ocean model + surface fluxes + ocean observations o Using a data assimilation system. o The challenge is to initialize the thermal structure without disrupting the dynamical balances (wave propagation is important) While preserving the water-mass characteristics OceanObs 09, Venice 21-25 September 2009 6
Dealing with model error: Hindcasts time Ocean reanalysis Real time Probabilistic Coupled Forecast Coupled Hindcasts, needed to estimate climatological PDF, require a historical ocean reanalysis OceanObs 09, Venice 21-25 September 2009 7
Impact of Data Assimilation Ocean data assimilation also improves the forecast skill (Alves et al 2003) Forecast Skill No Data Assimilation No Data Assimilation Data Assimilation Data Assimilation OceanObs 09, Venice 21-25 September 2009 8
A decade of progress on ENSO prediction NINO3.4 SST rms errors 64 start dates from 19870401 to 20021201 Ensemble sizes are 5 (0001), 5 (0001) and 5 (0001) Relative Reduction in SST Forecast Error ECMWF Seasonal Forecasting Systems 40 1 Fcast S3 Fcast S2 Fcast S1 Persistence 35 TOTAL GAIN 0.8 30 Rms error (deg C) 0.6 0.4 % 25 20 OC INI MODEL 0.2 15 0 0 1 2 3 4 5 6 Forecast time (months) 10 S1 S2 S3 5 Steady progress: ~1 month/decade skill gain How much is due to the initialization, how much to model development? 0 1 TOTAL GAIN OC INI MODEL Half of the gain on forecast skill is due to improved ocean initialization OceanObs 09, Venice 21-25 September 2009 9
Assessing the Ocean Observing System 1. No observing system is redundant Example: the Pacific, where Argo, moorings and altimeter still complement. Lessons for other basins. Implications of the missing TAO data for the on-going El Nino 2. The altimeter is the only OS contributing to the North Subtropical Atlantic. Argo is the only OS contributing the skill on the Indian Ocean. 3. There are obvious problems in the Eq Atlantic: model error, assimilation, and possibly insufficient observing system From Balmaseda and Anderson 2009 See also Fujii et al 2008 The assessment depends on the quality of the coupled model Sign of progress: a decade ago the OSES with Seasonal Forecasts were not considered a useful evaluation tool. Long records are needed for results to be significant: Any observing system needs to stay in place for a long time before any assessment is possible. So far impact on forecasts of SST only. Impact on atmospheric variables next OceanObs 09, Venice 21-25 September 2009 10
Importance of Real Time Ocean ReAnalyses From Xue et al CWP. Not operational yet OceanObs 09, Venice 21-25 September 2009 11
Ocean Observations & Model Development Understanding and revealing important processes. Examples are too many to mention. Role of the MJO on ENSO It has changed the conceptual models for ENSO with implications for predictability It has triggered intense activities on improving the model representation of tropical convection From McPhaden et al. GRL 2006 OceanObs 09, Venice 21-25 September 2009 12
Ocean Observations & Assimilation Development [Importance of] Multivariate relationships. Example: (T & S) 1999 time 2006 0m Salinity at 156 E From Fujii et al 2009 depth 5N 300m EQ 5S Tobs Tobs + S(T) Tobs+S(T)+Sobs Observations OceanObs 09, Venice 21-25 September 2009 13
Ocean Observation & Reliable forecast products 0.8 Forecast NINO4 Systems SST rms are errors generally not reliable (RMS > Spread) 252 start dates from 19870101 to 20071201 Ensemble sizes are 11 (0001), 11 (0001) and 11 (0001) Multi Model Calibration Single Model Rms error (deg C) 0.6 0.4 0.2 RMS Error Ensemble Spread A. Can we reduce the error? How much? (Predictability limit) B. Can we increase the spread by improving the ensemble generation and calibration? 0 0 1 2 3 4 5 6 Forecast time (months) Calibration and multi-model can increase the skill and reliability of forecasts. In a general case, even the multi-model needs calibration. Long records are needed for robust calibration and downscaling OceanObs 09, Venice 21-25 September 2009 14
Multi-Model Seasonal forecasts of Tropical Cyclones Multi-model Forecasts: 1 st June 2005: JASON 30 Obs July-November 20 E FORECAST 40 E 60 E 80 E 100 E 120 E 140 E CLIMATE 160 E 180 160 W 140 W 120 W 100 W 80 W 60 W 40 W 20 W 25 80 N 80 N 70 N 70 N 60 N 60 N 20 50 N 50 N 40 N 30 N 30 N 40 N 15 20 N 20 N 2.4 2.5 20.6 21.2 8.7 12.5 17.4 11.6 10 N 10 N 0 10 S 10 S 0 10 20 S 20 S 30 S 30 S 40 S 40 S 5 50 S 50 S 60 S 70 S 70 S 60 S 0 80 S 20 E 40 E 60 E 80 E 100 E 120 E 140 E 160 E 180 160 W 140 W 120 W 100 W 80 W No Significance Sig at 10% level Sig at 5% level Sig at 1% level NORMALIZED ACE 1.9 1.8 1.7 1.6 1.5 1.4 1.3 1.2 1.1 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 80 S 60 W 40 W 20 W EUROSIP multi-model seasonal forecast ECMWF/Met Office/Météo-France Western North Pacific Accumulated Cyclone Energy MJJA Forecast start reference is 01/03/YYYY Ensemble size = 37 (real time =123) Correlation= 0.83( 1.00) RMS Error= 0.26( 0.43) Forecast Observations +/- 1 Std. Deviation 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 Year (start of verification period) 1.9 1.8 1.7 1.6 1.5 1.4 1.3 1.2 1.1 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 W-Pac E-Pac Atl 1987-2004 2005 Vitart et al, GRL 2007 OceanObs 09, Venice 21-25 September 2009 15
What is the value of a long historical record? Example from the Medium Range Weather Forecasts (TIGGI) Impact of Increased ensemble size versus longer calibration period (Continuous Rank Probability Skill Score, T-2m Europe) A longer calibration period has larger impact than increasing the ensemble size. From Hagerdorn 2008 OceanObs 09, Venice 21-25 September 2009 16
Predicting for users: end-to-end 62 1 2 3 4 Climate 63 forecast 62 63 1 2 3 4 Downscaling 1 2 3 4 62 Application 63 model non-linear transformation 0 Forecast probability of T or PP 0 Forecasts probability of e.g. crop yield OceanObs 09, Venice 21-25 September 2009 17
Prediction of Dengue Risk transmission: 5 month lead time Forecast issued in Nov 1997, valid for Apr 1998 5-month lead fcst Obs Corr. skill From EUROBRISA http://eurobrisa.cptec.inpe.br/ Numerical Model+ Calibration + Dengue model OceanObs 09, Venice 21-25 September 2009 18
Recommendations for providers of observational data 1. Sustainability of the current observing systems and completion of impending missions. 2. Complete implementation of the RAMA array in the Indian Ocean. Add moorings in the S. Eq. Atlantic, where PIRATA sampling is very sparse. 3. Collect observations of the ocean mixed layer (likely to benefit medium-range, monthly and seasonal forecasts) 4. Ensure availability of independent data (ocean currents from current meters, sealevel gauges and transports), important for validation. Semi-independent data (OSCAR currents or Argo-derived velocities), are also valuable. 5. Continue observations of sea-ice concentration and thickness, important to for wide range of time scales, from weeks to decades. 6. Enhance observations of surface salinity, in order to reduce the large uncertainties in the fresh-water budget over the oceans. OceanObs 09, Venice 21-25 September 2009 19
Recommendations to the modeling and data assimilation communities 1. Further develop models and assimilation methods to exploit existing observations. 2. The assimilation community should be ready for the timely use of imminent observing systems (gravity missions, surface salinity and RAMA). 3. Continue efforts on ocean re-analyses, aiming at providing long, climate-quality records of the history of the ocean. This includes efforts on observation retrieval and quality control, as well as the improvement of assimilation methods. 4. Improve forcing fluxes from atmospheric re-analyses, ensuring that the products continue in near-real-time 5. Closer interaction between the oceanic and atmospheric communities for the balanced initialization of coupled models 6. Work should continue on SST products and re-analyses. (long records, high spatial and temporal resolution. Diurnal cycle) 7. Work should continue on OSEs and OSSEs OceanObs 09, Venice 21-25 September 2009 20
THE END OceanObs 09, Venice 21-25 September 2009 21
Recommendations for providers of observation data 1. It is essential to maintain the current observing system in the years to come and complete observing systems still under development. 2. Complete implementation of the RAMA mooring array in the Indian Ocean. Also add moorings in the south equatorial Atlantic in regions where PIRATA sampling is currently very sparse. 3. Collect observations of the ocean mixed layer, needed for better representation of processes related with the air-sea interaction at intraseasonal time scales, such as the MJO. This is likely to benefit medium-range, monthly and seasonal forecasts. 4. Ensure availability of independent data, such us ocean currents from current meters, sea-level gauges and transport, which are important to validate results from the assimilation systems. Semi-independent data, such as the OSCAR currents or Argo-derived velocities, are also very valuable, since they often involve an independent methodology. 5. Continue observations of sea-ice concentration and thickness, which are likely to be important to for wide range of time scales, from weeks to decades. 6. Enhance the in situ network of surface salinity observations, to complement impending satellite salinity missions, in order to reduce the large uncertainties in the fresh-water budget over the oceans. OceanObs 09, Venice 21-25 September 2009 22
Recommendations to the modeling and data assimilation communities 1. Further develop models and assimilation methods to exploit existing observations. Special attention should be paid to those areas where existing observations appear to have a negative effect on forecasts, such as the Equatorial Atlantic. 2. The assimilation community should be ready for the timely use of imminent observing systems, such as those coming from gravity missions, surface salinity and the newly developed Indian Ocean observing system. 3. Continue efforts on ocean re-analyses, aiming at providing long, climate-quality records of the history of the ocean. This includes efforts on observation retrieval and quality control, as well as the improvement of assimilation methods. In particular, it is important to develop methodology to extrapolate observational information into the past, as to mitigate the spurious variability induced by the ever evolving ocean observing system. 4. Improve forcing fluxes from atmospheric re-analyses, ensuring that the products continue in near-real-time as needed for the production of historically consistent records of ocean initial conditions. 5. Continue efforts in the oceanic and atmospheric community to develop more balanced initialization techniques that mitigate the undesirable initial adjustments by taking into account the air-sea interaction processes. 6. Work should continue on SST products and re-analyses. Ocean and atmosphere reanalysis would benefit from historical SST reconstruction resolving time scales shorter than one week as far into the past as possible. Future SST analysis resolving the diurnal cycle will be of interest for model development and shorter range forecasting. 7. Work should continue on OSEs and OSSEs so as to evaluate current and future ocean observing systems as well as current and future assimilation methods. OceanObs 09, Venice 21-25 September 2009 23