Weather and climate: Operational Forecasting Systems and Climate Reanalyses M.A. Balmaseda Tim Stockdale, Frederic Vitart, Hao Zuo, Kristian Mogensen, Dominique Salisbury, Steffen Tietsche, Saleh Abdalla, Jean Bidlot, Giovanna di Chiara, Dinard Schepers, Bruce Ingleby Seamless forecasting systems From days to years Reanalyses as integral part of forecasting systems Ongoing developments Coupled Data assimilation Coupled Reanalyses Summary, Requirements, Priorities Slide 1
Real Time (re-)analysis of Atmosphere/ Ocean/SeaIce /Land Atmosphere Lead time Seamless Probabilistic Prediction S2S data base Land Waves Ocean / Sea Ice C3S database TIGGE database Same model and initial conditions for different lead times. Resolution changes as a function of lead time. Main advantage: simplicity and cost Each prediction is an ensemble of N members (N~50) System Lead Time Prod Frequency Medium Range 15 days twice daily Monthly 46 days twice weekly Seasonal 7 months monthly Annual 12 months quarterly Implication: Ocean and Sea-Ice model components are part of the weather forecasting systems MAGDALENA A. BALMASEDA TAOS KICKOFF MEETING. PORTLAND 8-9 FEB 2018 Slide 2 2
Reanalyses as integral part of forecasting systems Reforecasts are needed for Calibration: dealing with model error Detection of Extreme Events Skill estimation -Tailored products (health, energy, agriculture) Ocean/Atmosphere/Sea-ice/Land reanalyses Real time Probabilistic Coupled Forecast time Reforecasts require historical reanalyses for initialization, consistent with real-time initial conditions Ocean observations are needed in support of weather and climate operational activities: Initialization-model development-verification MAGDALENA A. BALMASEDA TAOS KICKOFF MEETING. PORTLAND 8-9 FEB 2018 3
Scale interaction North PDO AMO Equator South Seasonal: trends-decadal-interannual-intraseasonal-diurnal. Wind and thermohaline circulations. Boundary Currents Subseasonal: interannual-intraseasonal-diurnal. Mixed layer processes. Wind driven. Boundary Currents Weather: diurnal-wave-wind. Planetary Boundary Layer. Boundary Currents
Argo Operational Observation Monitoring Moorings 5m 200m 600m 1000m Number of Temperature Observations 2000m MAGDALENA A. BALMASEDA TAOS KICKOFF MEETING. PORTLAND 8-9 FEB 2018 5
SEASONAL FORECASTS 1: Impact of assimilating Ocean Observations on skill Assim NoAssim Persistence Eq and South ATL Model overestimates variability Assim improves RMS and anomaly correlation. North Atllantic Assimilation does not improve skill. Predictability limit reached?
SEASONAL FORECASTS II: observations for process understanding and model improvement Bias DJF Anomaly Correlation SST Precip 7
SEASONAL FORECASTS III: observations for validation of and system improvements AMOC related decadal variability in the Ocean initial conditions affect the skill and biases of seasonal forecast in North Atlantic. Before Argo period, AMOC is not constrained in ocean renalyses. Florida Transport cable measurements are being very useful to diagnose problems Florida Strait cable transport measurements SST Bias Before 2000 After 2000 SEAS5 8
Subseasonal The Madden Forecasts Julian ( Oscillation up to 45-60 days) MJO is corner stone for predictability at the subseasonal time scales. Role of the Atlantic n amplification and propagation of the NAO? Topical Atlantic relevant in phases 1-2 and 8 What is the role of the mixed layer processes in the tropical atlantic? What is the impact of the Ocean Initial conditions.? The impact of ocean observations on this time scale has been little explored (or not at all). This will be a strong research focus in the next few years. 9
W eek 3-4 averages R NCE Impact of the MJO on precipitation JFM RPSS W eek 4 0.2 0.8 0.6 0.6 0.4 0.4 0.2 0.2 Raw 0.4 (b) 0 0 0 0 0.5 0.2 0.4 0.6 0.8 1 1 Forecast frequency Forecast Forecastfrequency probability ERA Phase 45 JFM RPSS W eek 4 CMA MME 0.6 0.8 W eek 3-4 averages (a) RMM1/2 & NINO3.4 anomalies MME RPSS 0.8 Observed frequency ERA Phase 23 Vigaud et al 2017 ERA Phase 67 (d) JFM Week3+4 MME RPSS vs MJO RMM1/2 & NINO3.4 Model Phase 81 JFM Smoothed Era-Interim Model Phase 45 (c) 999 Below 2010 normal MME Xval Below normalb)week3+4 class a)week3+4 Class J Week 3JFM Week 4 1 1 1 Observed frequency Smoothed Week 2 JFM Week3+4 MME RPSS vs MJO RMM1/2 & NINO3.4 F. Vitart JFM RPSS W eek 3 Model Phase 67 JFM RPSS W ee ERA Phase 81 0 MME Week MME week ECMF Wee ECMWF we NCEP CFS Week3 week Week3 CMA week 3 0 0.2 0.4 Foreca M ore skill JFM comp R to week 3 & we Smoothed NCEP MME Model Phase 23 Week 1 ECMWF Model CMA JFM RPSS W JJA eek 3 M axim um JFM RPSS coi with Niño3.4 & M JO pe (e) RMM1/ N ote: no strongmelore Ni to we
Impact on Tropical Cyclone Density (Summer) Vitart, GRL 2009
Sub-Projects Sub-seasonal to Seasonal (S2S) Prediction Project Teleconnections Madden-Julian Oscillation Monsoons Africa Extremes Verification and Products Research Issues Predictability Teleconnection O-A Coupling Scale interactions Physical processes Modelling Issues Initialisation Ensemble generation Resolution O-A Coupling Systematic errors Multi-model combination Needs & Applications Liaison with SERA (Working Group on Societal and Economic Research Applications) S2S Database
NWP : In situ data valuable for verification on wind speed and waves, complementing altimeter Mean Sentinel-3A surface wind speed (13 Dec. 2016 12 Dec. 2017) Standard deviation of the surface wind speed difference between Sentinel-3A and ECMWF model (13 Dec. 2016 12 Dec. 2017) Courtesy of Saleh Abdalla MAGDALENA A. BALMASEDA TAOS KICKOFF MEETING. PORTLAND 8-9 FEB 2018 October 29, 2014 13
NWP: Comparison with GTS buoy data in the Tropical Atlantic Ocean, north of 15 N Wind speed Wave height But sufficient number of buoys lacking in the South Atlantic EUROPEAN CENTRE FOR MEDIUM-RANGE WEATHER FORECASTS Courtesy of Jean Bidlot 14
The value of MSLP in Buoys for NWP Ingleby and Isaksen 2017 conclude of SLP from Buoys significantly improves NWP. Pressure measurements from buoys better than pressure measurements from ship Yet, about 40% of buoys no not report Pressure. Recommendation: equipe buoys and moorings with barometers. Report to GTS. MAGDALENA A. BALMASEDA TAOS KICKOFF MEETING. PORTLAND 8-9 FEB 2018 15
NWP: Coupled feedback under Hurricane IRMA Kristian Mogensen EUROPEAN CENTRE FOR MEDIUM-RANGE WEATHER FORECASTS 16
Number of observations of SST diurnal cycle Large gaps in tropical Atlantic. PIRATA moorings do not seem to report on diurnal cycle? TAO moorings do.. EUROPEAN CENTRE FOR MEDIUM-RANGE WEATHER FORECASTS Dom Salisbury 18
Coupling with the ocean improves the weather forecast scores. RMSE reduction (blue) in Day+5 weather forecast by coupling to the ocean in the tropics Large benefit across the tropics. Some degradation in Gulf of Guinea. Need observational sudies to understand why Courtesy of Kristian Mogensen 19
ERA-CLIM2 - Selected description of work Produce global reanalyses to reconstruct the past climate/weather of the earth system CERA-20C: A coupled reanalysis of the 20 th century based on conventional surface and subsurface observations deliver long time series of Essential Climate Variables (ECVs) CERA-SAT: A coupled reanalysis at higher resolution based on conventional and satellite observations evaluate the impact of a higher resolution on the coupled processes October 29, 2014
Air-sea coupling surface fluxes assimilation increments In uncoupled DA the surface fluxes not consistent with ocean observations From Laloyaux et al, submitted
Coupled assimilation - Atmospheric winds impact salinity Impact of scatterometer winds on ocean salinity Improved fit to in-situ observations Mean departure rmse SCATT No SCATT Courtesy of Giovanna De Chiara October 29, 2014 24
Computer capabilities and data bases Orchestrating research and production of Earth System Reanalyses Community Research and Development Earth System Reanalyses Refining Uncertainty Refining Resolutio n Sectorial Applications Regional aspects Evaluation consistent earth system reanalyses as far back as possible, with increasing accuracy and reliable uncertainty estimates 25
Summary Summary One observation contributes to many aspects of the forecasting system: initialization, model, calibration, and verification Objective evaluation is difficult. Scientific expertise and judgement should be a guiding principle for observing system. Panorama for TAOS is wide: from diurnal to decadal and beyond. Modelling and initializing the coupled O-A boundary layer is becoming increasingly important. In-situ observations of the diurnal cycle needed for SST, ocean and atmosphere PBL Need of long continuous time series for verification of reanalyses, extreme events/severe weather warning, and calibration of seasonal forecasts Moorings is the main observing system currently providing collocated ocean and atmosphere information. Model & DA errors in both ocean and atmosphere remain a serious obstacle. Observations needed for vericationprocess studies leading eventually to model improvement. Scientific exploitation of public data bases of multi-system forecast and reanalyses is encouraged. Slide 26
Some Recommendations Sustain the observing system in its current spatial-temporal resolution. Support developments on O-A coupling (model and DA). Support long reanalyses activities: For calibration of extreme events, monthly and seasonal (> 30-50 years) For detection/prediction of low frequency variability (>50 years) Enhancements Barometers in drifting buoys Diurnal cycle at mooring sites Wave buoys in the vicinity of PIRATA, especially South of Equator Support Fiducial observations for SST analyses Atlantic Specific Florida Strait transport cable measurements very valuable Salinity (surface and subsurface) of particular important for the Atlantic ocean Looking ahead: integrated systems to support fluxes of carbon, aerosols -Observations to support estimates of cean-atmosphere carbon-fluxes -example: iron as indicator of aerosol deposition/seeding marine productivity. Slide 27