1 / 50

Initialization Techniques in Seasonal Forecasting

Initialization Techniques in Seasonal Forecasting. Magdalena A. Balmaseda. Outline. The importance of the ocean initial conditions in seasonal forecasts A well established case: ENSO in the Equatorial Pacific A seasonal forecasting system Ocean Model initialization

ham
Download Presentation

Initialization Techniques in Seasonal Forecasting

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Initialization Techniques in Seasonal Forecasting Magdalena A. Balmaseda

  2. Outline The importance of the ocean initial conditions in seasonal forecasts A well established case: ENSO in the Equatorial Pacific A seasonal forecasting system Ocean Model initialization The value of observational information: fluxes, SST, ocean observations The difficulties The ECMWF Ocean Initialization system ORAS4 Assessment of Initialization strategies Full Initialization, Anomaly Initialization Ensemble generation

  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 SF • How long in advance?: from seasons to decades • 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,… • Sea-Ice

  4. End-To-End Seasonal forecasting System COUPLED MODEL Atmosphere model Atmosphere model Atmosphere model Ocean model Ocean model Ocean model PROBABILISTIC CALIBRATED FORECAST ENSEMBLE GENERATION Forward Integration Forecast Calibration Initialization Tailored Forecast PRODUCTS OCEAN

  5. Dealing with model error: Hindcasts Ocean reanalysis Real time Probabilistic Coupled Forecast time Coupled Hindcasts, needed to estimate climatological PDF, require a historical ocean reanalysis Consistency between historical and real-time initial conditions is required. Hindcasts are also needed for skill estimation

  6. 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. • Emphasis on the thermal structure of the upper ocean • Predictability is due to higher heat capacity and predictable dynamics

  7. Need to Initialize the subsurface of the ocean

  8. A decade of progress on ENSO prediction Half of the gain on forecast skill is due to improved ocean initialization S1 S2 S3 Initialization into Context • Steady progress: ~1 month/decade skill gain • How much is due to the initialization, how much to model development? Balmaseda et al 2010, OceanObs

  9. Initialization Problem: Production of Optimal I.C. • Optimal Initial Conditions: those that produce the best forecast. Need of a metric: lead time, variable, region (i.e. subjective choice) Usually forecast of SST indices, lead time 1-6 months In complex non linear systems there is no “objective searching algorithm” for optimality. The problem is solved by subjective choices and physical insight. • Theoretically: I.C. should represent accurately the state of the real world. I.C. should project into the model attractor, so the model is able to evolve them. In case of model error the above 2 statements may seem contradictory • Practical requirements: If forecasts need calibration, the forecast I.C. should be “consistent” with the I.C. of the calibrating hindcasts. Need for historical ocean reanalysis • Current Priorities: • Initialization of SST and ocean subsurface. • Land/ice/snow potentially important. Atmospheric initial conditions play a secondary role.

  10. 1982 1993 2001 XBT’s 60’s Satellite SST Moorings/Altimeter ARGO Information to initialize the ocean • Ocean model Plus: SST Atmospheric fluxes from atmospheric reanalysis Subsurface ocean information Time evolution of the Ocean Observing System

  11. How do we initialize the ocean? To a large extent, the large scale ocean variability is forced by the atmospheric surface fluxes. Different ocean models forced by the same surface fluxes will produce similar tropical variability. Daily fluxes of heat (short and long wave, latent, sensible heat), momentum and fresh water fluxes. Wind stress is essential for the circulation. Constrained by SST: Fluxes from atmospheric models have large systematic errors and a large unconstrained chaotic component Constrained by SST+ Atmos Observations: Surface fluxes from atmospheric reanalysis Reduced chaotic component. But still large errors/uncertainty Constrained by SST+AtmosObservations+Ocean Observations: Ocean reanalisys

  12. Uncertainty in Surface Fluxes:Need for Data Assimilation Equatorial Atlantic: Taux anomalies Equatorial Atlantic upper heat content anomalies. No assimilation Equatorial Atlantic upper heat content anomalies. Assimilation ERA15/OPS ERA40 • Large uncertainty in wind products lead to large uncertainty in the ocean subsurface • The possibility is to use additional information from ocean data (temperature, others…) • Questions: • Does assimilation of ocean data constrain the ocean state? YES • Does the assimilation of ocean data improve the ocean estimate? YES • Does the assimilation of ocean data improve the seasonal forecasts. YES

  13. Challenges on Ocean Initialization using data assimilation In the production of the reanalysis • Ocean model and surface forcing error • Changing observing system Need to take into account model bias. Balance Multivariate Relationships In the production of the forecast • Coupled model error can cause initialization shock and drift, even with perfect initial conditions

  14. Data coverage for Nov 2005 Ocean Observing System Data coverage for June 1982 Changing observing system is a challenge for consistent reanalysis Today’s Observations will be used in years to come • ▲Moorings: SubsurfaceTemperature • ◊ ARGO floats: Subsurface Temperature and Salinity • + XBT : Subsurface Temperature

  15. PIRATA Impact of data assimilation on the mean Assim of mooring data CTL=No data Large impact of data in the mean state leading to spurious variability This is largely solved by the introduction of bias correction

  16. The Assimilation corrects the ocean mean state Equatorial Pacific (x) Mean Assimation Temperature Increment z Data assimilation corrects the slope and mean depth of the equatorial thermocline Free model Data Assimilation

  17. Improves the Interannual Variability ASSIM TS T+S+Alti CONTROL (no ASSIM) Correlation with altimeter-derive sea level data.

  18. No Data Assimilation Data Assimilation No Data Assimilation Data Assimilation Impact of Data Assimilation Forecast Skill Ocean data assimilation also improves the forecast skill (Alves et al 2003) This was with S1. It has been subsequently repeated with S2, S3, S4. But it is not trivial

  19. Ocean Initialization at the ECMWF Ocean Reanalysis System 4 (ORAS4) Main Objective: Initialization of seasonal forecasts Historical reanalysis brought up-to-date => Useful to study and monitor climate variability • Operational ORA-S4 NEMO-NEMOVAR • ERA-40 daily fluxes (1958-1989) and ERA-Interim thereafter • Retrospective Ocean Reanalysis back to 1958, 5 ensemble members • Multivariate offline+on-line Bias Correction (pressure gradient, Temp,Sal, offline from recent period ) • Assimilation of SST, temperature and salinity profiles, altimeter sea level anomalies an global sea level trends • Balance constrains (T/S and geostrophy) • Sequential, 10 days analysis cycle, Incremental Analysis Update

  20. NEMOVAR • Variational data assimilation system for the NEMO ocean model (follow up of OPAVAR) • Collaborative project with several institutions: CERFACS, MetOffice, INRIA, ECMWF • Multiple loops. Adjoint and Tangent linear exist. • NEMOVAR in ORAS4: Multivariate Incremental 3Dvar FGAT IAU • Flow dependence background errors: • T error depends on vertical gradients • T/S relationship: linearized vertical profile displacement • Sea Level and density: vertical profile displacement taking into account stratification • Geostrophy • It assimilates T/S profiles and along track altimeter. • Automatic QC, thinning, supperobbing • See Mogensen et al 2012. ECMWF techmemo 668 • http://www.ecmwf.int/publications/library/do/references/show?id=90389 • OUTSIDE NEMOVAR: • SST is used to correct surface heat fluxes • Global Sea Level corrects the fresh water flux • Bias Correction

  21. Bias Correction Algorithm Slow varying term, estimated online from assimilation increments dk Seasonal term, estimated offline from Argo Period Bias online: Time evolution • Need to determine: • Offline bias correction • Time evolution of on-line bias: a(memory) andb(updating factor) • A(y): Partition of bias into T/S and pressure gradient. • Function of latitude. At the equator the bias correction is mainly adiabatic (pressure gradient) • Refinement of Balmaseda et al 2007, Dee 2005, Bell et al 2002 The offline bias correction is estimated from Argo period. The correction is applied since 1957-00-01 to present. It is a way of extrapolating Argo information into the past.

  22. Time correlation with altimeter SL product CNTL: NoObs NEMOVAR T+S ORAS4 T+S+Alti CNTL NEMOVAR TS ORAS4 (TS+Alti)

  23. Impact on Seasonal Forecast Skill Consistent Improvement everywhere. Even in the Atlantic, traditionally challenging area ORAS4 CNTL

  24. Quantifying the value of observational information in seasonal forecasts • The outcome may depend on the coupled system • In a good system information may be redundant, but not detrimental. If adding more information degrades the results, there is something wrong with the methodology (coupled/assim system) • Experiments conducted with the ECMWF S3 Balmaseda and Anderson 2009, GRL SST (SYNTEX System Luo et al 2005, Decadal Forecasting Keenlyside et al, 2008) SST+ Atmos observations (fluxes from atmos reanalysis) SST+ Atmos observations+ Ocean Observations (ocean reanalysis)

  25. Impact of external “real world” information ALL ATMOS+SST SST only Central Western Pacific Eastern Pacific SST only produces a cold drift in both eastern and western Pacific. Large interannual variability (not shown) ATMOS+SST does not have drift in the Eastern Pacific, but its bias cold in the Western-Central Pacific ALL produces warm bias in the Eastern Pacific, no bias in the Western-Central Pacifi. The warm bias in the eastern Pacific is a consequence of the unbalanced initialization: Weak winds in coupled model -> downwelling Kelvin wave->warming in Eastern Pacific What about the forecast skill?

  26. Impact of “real world” information on skill: Reduction in Error (MAE) in SST SF by adding observational information WINDS = [(ATMOS+SST) – SSTonly ]/ SSTonly OCDATA = [ALL - (ATMOS+SST)]/ (ATMOS+SST) OC+WINDS = [ALL - SSTonly]/SSTonly The additional information about the real world improved the forecast skill, except in the Equatorial Atlantic Still, optimal use of the observations needs more sophisticated assimilation techniques and better models, to reduced initialization shock

  27. Perceived Paradigm for initialization of coupled forecasts Real world Model attractor Medium range Being close to the real world is perceived as advantageous. Model retains information for these time scales. Model attractor and real world are close? Decadal or longer Need to initialize the model attractor on the relevant time and spatial scales. Model attractor different from real world. Seasonal? At first sight, this paradigm would not allow a seamless prediction system. • Seasonal traditional approach: • Full initialization + A aposteriori calibration of forecast • Not clear how to achieve initialization in model attractor • Anomaly Initialization (decadal forecasts, Smith et al) • Coupled with observed SST interface (SST only): This means neglecting observational information. Keenlyside et al 2008. As shown before, this is not optimal • Other more sophisticated (EnKF, coupled DA, weakly coupled DA)

  28. Initialization Shock and Skill e Model Attractor (MA) c non-linear interactions important b phase space Real World (RW) Forecast lead time Initialization shock d a: perfect initialization and perfect model b: no initialization shock. Best skill c: Initialization shock. Good skill until lead time L d: model attractor ini. No initialization shock e: initialization shock + Non linearities Different convergence a L

  29. Initialization Shock and non linearities Model Attractor (MA) non-linear interactions important phase space Empirical Flux Corrections Real World (RW) Forecast lead time b a

  30. CLIVAR EasyINIT workshop on Initialization Inventory of strategies (see table) Need to be assessed and evaluated Nudging of anomalies from other reanalysis may not the best, but a practical solution to study sensitivities and to get started Relation between initialization strategy and forecast strategy Coupled initialization probably the long term solution, but more difficult to start with. Use all possible observational information unless there is a good reason why not. http://www.knmi.nl/samenw/easyinit/

  31. Assimilation mainly of ocean observations. Not intention to initialize the fast time scales of the atmospheric component. The atmospheric observations are either neglected or binned in long windows. They can also be used indirectly via nudging to existing atmospheric reanalysis. The aim is to produce better (more balanced) initial conditions rather than an accurate estimation of the ocean variability. Observations are used to correct both the mean and the variability. Existing efforts Coupled 4D-var (Suguira et al 2008). Both ocean and atmospheric observations (binned). 9 months assimilation cycle. Control vector: coupling coefficients and ocean initial conditions. Coupled EnKF (Zhang et al 2007). Only ocean observations are used directly. Atmospheric information is nudged during the integration. Ocean Data Assimilation with a coupled model. (Fujii et al 2009) Atmospheric model is free (AMIP). Spectral control of SST variability. Free at time scales < 1month. Coupled Data Assimilation: MEAN+ ANOMALY

  32. Observational information is used to initialize only the anomalies, which are superimposed into the model climate. It assumes quasi-linear regime. Observational information is used either directly or from existing reanalysis. Usually only the ocean component is initialized. Background given by the coupled model To obtain observational anomalies an observational climatology is assumed. In poor observed areas the time sampling for the climatology may be limited Two flavours One-Tier anomaly initialization (Smith et al 2007). Ocean observations are assimilatated directly. Background error covariance formulation derived from coupled model. Emphasis on large spatial scales Two-Tier anomaly initialization (Pohlmann et al 2009). Nudging of anomalies from existing ocean re-analysis. The spatial structures are those provided by the source re-analysis. Useful to compare different ocean reanalysis with the same model (Zhu et al 2012, GRL). Widely used in decadal forecasts Special case when only SST anomalies are used. (Keenlyside et al 2008) 2) ANOMALY INITIALIZATION

  33. Comparison of Strategies for dealing with systematic errors in a coupled ocean-atmosphere forecasting systemas part of the EU FP7 COMBINE project Nature climate Flux correction Normal initialisation Anomaly initialisation Model climate Magnusson et al. 2011 ECMWF Techmemo 658 Magnusson et al. 2012 ECMWF Techmemo To appear

  34. Coupled model error 10m winds: model - analysis T2m bias: model - analysis Part of the error comes from the atmospheric component (too strong easterlies at the equator) The error amplifies in the couped model (positive Bjerkness feedback) Possibility of flux correction

  35. Coupled Ucor Coupled UHcor Ucor: surface wind is corrected when passed to the ocean UHCor: surface wind and heat flux are corrected when passed to the ocean Different mean states Analysis Coupled Free

  36. Comparison of Forecast Strategies: Method A) Raw Forecasts Example of decadal forecasts: Global SST B) Bias Estimation C) Corrected Forecasts Analysis Full Initialization Anomaly Initialization U Correction U+H correction

  37. Comparison of Forecast Strategies: Seasonal Nino 3 SST Drift 1-14 month forecast Analysis Full IniAnomaly IniU Correction U+H correction

  38. Nino3.4 SST forecasts November 1995 – November 1998 Full Initialization Anomaly Initialisation U-flux correction U- and H-flux correction 99 96 97 98 Linus Magnusson et al.

  39. Impact on Forecast Skill (SST and Precip) Persistence Full IniAnomaly IniU Correction U+H correction Depending on the variable and region, the forecast skill is more or less affected by Initialization and forecast strategy. Where the non linearities are strong, having a good mean state helps. In these situations, anomaly initialization underperfoms (Precip in Nino3.4, for instance)

  40. ENSEMBLE GENERATION Representing Uncertainty without disrupting Predictability Seasonal versus Medium Range In Seasonal, the main source of error/uncertainty is error formulation It uses random (but realistic) perturbations as opposed to optimal perturbations to the initial conditions There has been some research on optimal perturbations (breeding vectors, stochastic optimals, empirical singular vectors), but it will not be covered here (some information is given as hidden slides at the end)

  41. Ensemble Generation In the ECMWF Seasonal Forecasting System • Uncertainty in initial conditions: Burst ensemble: (as opposed to lag-ensemble) 50-member ensemble forecast first of each month Uncertainty in the ocean surface 50 SST perturbations Uncertainty in the Ocean Subsurface 5 different ocean analysis generated with wind perturbations + SV for atmospheric initial conditions Impact during the first month • Uncertainty in model formulation: Stochastic physics Multi-model ensemble (EUROSIP)

  42. SST Perturbations Uncertainties in the SST -Create data base with errors of weekly SST anomalies,arranged by calendar week: Error in SST product: (differences between OIv2/OI2dvar) Errors in time resolution: weekly versus daily SST -Random draw of weekly perturbations, applied at the beginning of the coupled forecast. Over the mixed layer (~60m) -A centred ensemble of 50 members

  43. 1-3 months decorrelation time in wind Wind perturbations +p1/-p1 Effect on Ocean Subsurface (D20) ~6-12 months decorrelation time in the thermocline Uncertainties in the ocean Subsurface -Create data base with errors in the monthly anomalous wind stress, arranged by calendar month: (differences between ERA40-CORE) -Random draw of monthly perturbations, applied during the ocean analyses. -A centered ensemble of 5 analysis is constructed with: -p1 -p2 0 +p1 +p2

  44. Can we reduce the error? How much? (Predictability limit) • Can we increase the spread by improving the ensemble generation? Is the ensemble spread sufficient? Are the forecast reliable? Forecast System is not reliable: RMS > Spread How to improve the reliability of the ensemble: a) Sampling model error: multi-model, physical parameterizations b) Calibrating the forecast a-posteriori

  45. RMS error of Nino3 SST anomalies Persistence ECMWF ensemble spread EUROSIP 2.1) Sampling model error: The Real Time Multimodel EUROSIP ECMWF-UKMO-MeteoFrance

  46. 2.2) Sampling model error: The Real Time Multimodel RMS error of Nino3 SST anomalies Bayesian Calibration Persistence ECMWF ensemble spread EUROSIP EUROSIP ECMWF-UKMO-MeteoFrance

  47. Summary: Initialization • Seasonal Forecasting (SF) of atmospheric variables is a boundary condition problem. • Seasonal Forecasting of SST is an initial condition problem. • Assimilation of ocean observations reduces the large uncertainty (error) due to the forcing fluxes. Initialization of Seasonal Forecasts needs SST, subsurface temperature, salinity and altimeter derived sea level anomalies. • Data assimilation changes the ocean mean state. Therefore, consistent ocean reanalysis requires an explicit treatment of the bias • The separate initialization of the ocean and atmosphere systems can lead to initialization shock during the forecasts. A more balance “coupled” initialization is desirable, but it remains challenging. • Initialization and forecast strategy go together. The best strategy may depend on the model. The anomaly initialization used in decadal forecasts can have problems in seasonal ”

  48. Summary: Ensemble generation • The ensemble techniques used in the Medium Range can not be applied directly to the Seasonal Forecast System (SFS)(since the linear assumption would not hold in the atmosphere model for optimization times ~>1month) • The ECMWF SFS uses random sampling (as opposed to optimal sampling) of existing uncertainties, mainly in the initial conditions. • Results suggest that model error is the largest source of forecast error. • There is a variety of techniques to sample model error: • Stochastic Physics • Multi-model • Perturbed parameters

  49. Stochastic forcing 2.1) Uncertainties in deterministic atmospheric physics? ECMWF stochastic physics scheme:  is a stochastic variable, constant over time intervals of 6hrs and over 10x10 lat/long boxes Buizza, Miller and Palmer, 1999; Palmer 2001 The Stochastic Physics samples neither uncertainty in the parameters, nor model error

  50. ST SP From Vialard et al, MWR 2005 Ensemble Spread Wind Perturbations (WP) SST Perturbations(ST) Stochastic Physics (SP) Wind Perturbations No DA (WPND) All(SWT) Lag-averaged(LA) • The spread by different methods converge to the same asymptotic value after after 5-6 months. • SST and Lag-averaged perturbations dominate spread at ~1month lead time. • With DA, the wind perturbations grow slowly, and notably influence the SST only after 3m. • Without DA, the initial spread (<3m) is larger. The asymptotic value is slightly larger • But the level of spread is not sufficient. Need to sample model error

More Related