GMAOs SeasonaltoInterannual Forecasts Potential contributions to MME Forecasts

  • Slides: 11
Download presentation
GMAO’s Seasonal-to-Interannual Forecasts & Potential contributions to MME Forecasts Max Suarez Global Modeling and

GMAO’s Seasonal-to-Interannual Forecasts & Potential contributions to MME Forecasts Max Suarez Global Modeling and Assimilation Office (GMAO) Michele Rienecker, Randy Koster, Siegfried Schubert Christian Keppenne, Rolf Reichle, Shu-Chih Yang CTB SAB Meeting August 28, 2007 30 Nov 2005

CGCMv 1 Ensemble Forecast System AGCM (AMIP forced with Reynolds SST; NCEP Analyses) Atmospheric

CGCMv 1 Ensemble Forecast System AGCM (AMIP forced with Reynolds SST; NCEP Analyses) Atmospheric state perturbations: ’s randomly from previous integrations 12 month Coupled Integrations: 6 -30 ensemble members Ocean state estimate perturbations: ’s randomly from snapshots or from En. KF Ocean DAS (Surface wind analysis, GPCP precipitation; Reynolds SST, Temperature profiles; synthetic salinity profiles; altimetry) AGCM: NSIPP 1 AGCM, 2 x 2. 5 x L 34 LSM: Mosaic (SVAT) OGCM: Poseidon v 4, 1/3 x 5/8 x L 27, with embedded mixed layer physics CGCM: Full coupling, once per day ODAS: Optimal Interpolation of in situ temperature profiles; Ensemble Kalman Filter “LDAS”: Offline forced land states (recalibrated) 30 Nov 2005

CGCMv 1 Forecasts initialized 1 Aug, 2007 30 Nov 2005

CGCMv 1 Forecasts initialized 1 Aug, 2007 30 Nov 2005

Forecast skill (ACC) from CGCMv 1 SST anomaly Initialized 1 March, 1993 -2006 En.

Forecast skill (ACC) from CGCMv 1 SST anomaly Initialized 1 March, 1993 -2006 En. KF OI-TS 1 -month lead 3 -month lead 6 -month lead 30 Nov 2005

Forecast skill (ACC) from CGCMv 1 Heat content anomaly in upper 300 m Initialized

Forecast skill (ACC) from CGCMv 1 Heat content anomaly in upper 300 m Initialized 1 March, 1993 -2006 En. KF OI-TS 1 -month lead 3 -month lead 6 -month lead 30 Nov 2005

July Forecast Anomaly Correlation CNT Mean BV FEB start MAY start AUG start NOV

July Forecast Anomaly Correlation CNT Mean BV FEB start MAY start AUG start NOV start 30 Nov 2005 Mean BV 1 Mean BV 2

GEOS-5 CGCM Forecast System - Status & Plans q GEOS-5 AGCM, Catchment LSM, MOM

GEOS-5 CGCM Forecast System - Status & Plans q GEOS-5 AGCM, Catchment LSM, MOM 4 • Simulation for several decades, still tuning q Configuration • AGCM: 1° 1. 25° 72 L • MOM 4: 0. 25° 0. 5° 40 L (telescoping grid in equatorial band - NCEP configuration) q Forecast Initialization • MOM 4 initialized by ODAS-2 multivariate assimilation (En. KF) • LSM initialized by offline (LIS) forcing • AGCM initialized by reanalysis (NCEP, MERRA? ) q Hindcast strategy • 1993 -2007 • Use low-resolution MERRA for atmosphere and also for ocean forcing • Coupled En. KF? q Ensemble Strategy possibilities • mimic NCEP, initialization every day • mimic NCEP (except initialization every 3 days) + 3 -member ensemble from breeding • all ensemble members initialized 1 st of the month 30 Nov 2005

GEOS-5 CGCM- contributing to the MME q GEOS-5 timeline • Q 2 -FY 08

GEOS-5 CGCM- contributing to the MME q GEOS-5 timeline • Q 2 -FY 08 Begin ODAS • Q 3 -FY 08 Begin hindcasts • Q 4 -FY 08 Contribute selected G 5 hindcast months to CTB MME q Proposed Interim Strategy • Use CGCMv 1 (ensembles initialized 1 st month, 1993 - present) • Q 1 -FY 08 contribute En. KF system for selected months as test of MME 30 Nov 2005

CGCMv 1 Forecasts initialized 1 Aug, 2007 30 Nov 2005

CGCMv 1 Forecasts initialized 1 Aug, 2007 30 Nov 2005

30 Nov 2005

30 Nov 2005

Ensemble forecast system performance ACC in Nino 3 Mean SST error in Nino 3

Ensemble forecast system performance ACC in Nino 3 Mean SST error in Nino 3 SST ensemble spread vs. SST error In Nino 3 region SST Ensemble spread in Nino 3 • Forecast errors show a strong dependence on seasons/starting month error • Ensemble spread doesn’t have the season-dependent characteristic shown in forecast errors spread • Ensemble spread is too small compared to the forecast error, especially at early months Forecast month 30 Nov 2005 Shu-Chih Yang