New Seasonal Prediction System JMA/MRI-CPS2 (JMA/MRI-CGCM2) Tamaki Yasuda (tyasuda@met.kishou.go.jp) Japan Meteorological Agency (GPC Tokyo) WGSIP17, 13-14 September 2015, Norkörrping, Sweden

History of JMA Coupled Prediction System 2015 JMA/MRI-CPS2 Atm: TL159 (~110 km, 1.125 deg.) L60 (~0.1hPa) Ocn: 1x0.5-0.3 L52+BBL 　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　2010 JMA/MRI-CPS1 for Seasonal Prediction 2008 JMA/MRI-CPS1 for ENSO Outlook Atm: TL95 (~180 km, 1.875 deg.) L40 (~0.4hPa) Ocn: 1x1-0.3 (30N-30S) L50　(top: ~1m) 2003 JMA-CGCM02 (GSM0103) for ENSO Outlook Atm: T63 (~180 km, 1.875 deg.) L40 (~0.4hPa) Ocn: 2.5x2-0.5(10N-10S) L20 (top: ~10 m) 1999 JMA-CGCM01 (GSM8911) for ENSO Outlook Atm: T42 (~250 km) L21 (~10 hPa) Ocn: 2.5x2-0.5(10N-10S) L20 Takaya et al., in prep. Takaya et al.: Japan Meteorological Agency/Meteorological Research Institute-Coupled Prediction System version 1 (JMA/MRI-CPS1) for operational seasonal forecasting, to be submitted to Climate Dynamics. WGSIP17, 13-14 September 2015, Norkörrping, Sweden

Updated Seasonal Ensemble Prediction System JMA/MRI-CPS1 (Previous) JMA/MRI-CPS2 (New) Atmosphere (JMA-AGCM) TL95L40, ~180km, Up to 0.4hPa TL159L60 , ~110km, Up to 0.1hPa Stochastic Tendency Perturbation GHG forcing in RCP4.5 scenario Ocean (MRI.COM) (Tsujino et al 2010) 1.0º (lon) x 0.3-1º (lat) L50 75ºS-75ºN Ocean Sea-ice climatology 1.0º (lon) x 0.3-0.5º (lat) L52+BBL Global Ocean with Tripolar Grids Sea-ice model Coupler (Scup) (Yoshimura and Yukimoto 2008) 1-hour coupling interval Momentum and heat flux adjustments No flux adjustment Initial Condition Atmosphere: JRA-25 Land: Climatology with ERA-15 forcing Ocean: MOVE/MRI.COM-G T, S&SSH (Usui et al. 2006) Atmosphere: JRA-55 Land: JRA-55 land analysis Ocean: MOVE/MRI.COM-G2 T, S & SSH Ensemble Size 51 (9 BGMs, 6 days with 5-day LAF) (13 BGMs, 4 days with 5-day LAF) *Hindcast: 10 member ensembles (5 BGMs, twice a month, 1979-2014) WGSIP17, 13-14 September 2015, Norkörrping, Sweden 3

New Sources of Predictability Global ocean domain Dynamical sea ice simulation Fully covered stratosphere (Top: 0.1 hPa) Land initialization with JRA-55 More sophisticated description of GHGs (6 gases prescribed with RCP4.5 scenario) The new system is capable of incorporating a wide range of potential sources of the predictability. Atmo-sphere Strato-sphere Sea Ice Ocean Land WGSIP17, 13-14 September 2015, Norkörrping, Sweden 4

SST Standard Deviation (DJF) Observation (COBE-SST) - Too large amplitude of SST interannual variability is reduced. JMA/MRI-CPS2 ; LT=1 month (initial: Nov) JMA/MRI-CPS1 ; LT=1 month (initial: Nov) JMA/MRI-CPS2 ; LT=4 months (initial: Aug) JMA/MRI-CPS1 ; LT=4 months (initial: Aug) [ K ] WGSIP17, 13-14 September 2015, Norkörrping, Sweden

WGSIP17, 13-14 September 2015, Norkörrping, Sweden ENSO Prediction ACC of NINO.3 SST RMSE of NINO.3 SST - Better prediction skill of ENSO for longer lead time in the new system WGSIP17, 13-14 September 2015, Norkörrping, Sweden

ACC for 3-month forecast 2-m temperature averaged in the Northern Hemisphere (20-90N) Precipitation averaged in the Tropical Region (20S-20N) WGSIP17, 13-14 September 2015, Norkörrping, Sweden

Forecast of Arctic Sea Ice Arctic sea-ice extent (Initial: May) Arctic sea-ice extent in September Observation (COBE-ICE) Lead time of 2 months ACC=0.73 Lead time of 5 months ACC=0.68 Color: Individual forecast Black: Observed climatology (COBE-ICE) (Grey shaded) minimum and maximum - Seasonal variation of Arctic sea ice - Interannual variability and reduction trend of Arctic sea-ice extent WGSIP17, 13-14 September 2015, Norkörrping, Sweden 8

Improvement with Land Initialization Anomaly correlation of monthly 2-m temperature over land (lead time: 0 month) Land Initialization (JMA/MRI-CPS1: Climatology with ERA-15 forcing) JMA/MRI-CPS2: JRA-55 land analysis (Land A) Additional experiment: Climatology (1981-2010) of JRA-55 land analysis (Land C) Global Northern Hemisphere Eurasia ACC Initial Month Initial Month Initial Month Land A: slightly better predictions than Land C Global: 90S-90N,0E-360E NH: 20N-90N,0E-360E Eurasia: 20N-90N,0E-180E WGSIP17, 13-14 September 2015, Norkörrping, Sweden

Improvement with GHGs Trend 2-m temperature Trend over Land (JJA) (Initial: May) VarCO2 - JMA/MRI-CPS1 - CO2 Trend VarGHG - JMA/MRI-CPS2: - CO2, CH4, N2O, CHC-11, CHF-12, HCFC-22 (GHGs) Trend ConstGHG (Additional experiment) - JMA/MRI-CPS2 - Constant GHGs VarCO2 ConstGHG VarGHG JRA-55 * The land-sea mask is different between JRA-55 and model Linear trend of 2-m temperature: VARCO2 < ConstGHG < VarGHG < JRA-55 WGSIP17, 13-14 September 2015, Norkörrping, Sweden

Summary (JMA/MRI-CPS2) JMA’s seasonal EPS has been updated in June 2015. 1. Improvement of JMA/MRI-CPS2: - Enhanced horizontal / vertical resolution - New sources of predictability: such as global ocean, stratosphere, sea ice and GHGs - New initial conditions: JRA-55 for atmosphere and land surface MRI.COM/MOVE-G2 for ocean 2. Improvement of prediction skill: - ENSO amplitude of interannual variability and prediction skill - 3-month forecast (such as 2-m temperature, precipitation) - Sea-ice interannual variability and reduction trend - Warming trend of 2-m temperature over land - MJO amplitude (not shown) WGSIP17, 13-14 September 2015, Norkörrping, Sweden

New Ensemble Initialization System for Seamless Climate Predictions For WGSIP Sep, 2015 Activity of MIROC group New Ensemble Initialization System for Seamless Climate Predictions Masayoshi Ishii Meterological Research Institute (MRI), Japan Meteorological Agency (JMA) with colleagues of AORI, NEIS, JAMSTEC, and MRI (This system is used in MRI and MIROC groups.)

WGSIP17, 13-14 September 2015, Norkörrping, Sweden Purposes To provide less uncertain information on global warming projection and prediction. To realize seasonal-to-decadal predictions seamlessly for future climate variations. To reproduce past climate changes and variations for more than 100 years for understanding past global warming and for improvement of climate prediction models. To understand mechanisms of climate variations on various spatio-temporal scales. WGSIP17, 13-14 September 2015, Norkörrping, Sweden

New Initialization System & 150-yr Reanalysis Coupled Model + EnKF From 1850 to present Ocean Obs. Quality Control Objective Analysis w/ EnKF CGCM Ensemble Run Atmos. Obs. Computational Node Reformation Spec. (ver. 1) LETKF (Hunt et al. 2007) Loosely coupled, full DA Assimilation interval: 6 hours for Atmos. and 5 days for Ocean Surface Pressure Data (ISPD v.3.2.8) Typhoon bogus (IBTrACS v03r05) SST: COBE-SST2 + SST perturbations Gridded subsurface ocean T and S (Ishii and Kimoto 2009) A new system configured specifically for long-term climate reanalysis The same initialization system is used by MRI and MIROC groups. WGSIP17, 13-14 September 2015, Norkörrping, Sweden

500hPa Geopotential height at 12UTC, February 20th, 2005 ERA-Interim EnKF Spread Pattern CC after subtracting zonal means Pattern Correlation Time Preliminary Run with MIROC3m (T42L20 AGCM and 1-deg OGCM)

Exercise on Nino-3.4 SSTA Prediction Raw predictions Training period of prediction models is now extended to the beginning of the 20th Century or more. Prediction biases removed.

Thank you for your attention. WGSIP17, 13-14 September 2015, Norkörrping, Sweden

Improved Madden-Julian Oscillation Initial: Nov-Apr Wheeler-Hendon Index: RMM1, RMM2 Correlation JMA/MRI CPS2 JMA/MRI CPS1 JMA 1-month EPS (AGCM) [day] larger amplitude Amplitude Error smaller amplitude - Forecast time of 0.5 correlation extends roughly 2 days. - Too small MJO amplitude is improved. WGSIP17, 13-14 September 2015, Norkörrping, Sweden

ACC for 3-month forecast 2-m temperature JMA/MRI-CPS2 2-m temperature averaged in the Northern Hemisphere (20-90N) (Initial: May) JJA GLB:0.414 NH:0.375 (Initial: Nov) DJF GLB:0.405 NH:0.323 WGSIP17, 13-14 September 2015, Norkörrping, Sweden

ACC for 3-month forecast Precipitation JMA/MRI-CPS2 Precipitation averaged in the Tropical Region (20S-20N) (Initial: May) JJA GLB:0.167 TRP:0.342 (Initial: Nov) DJF GLB:0.216 TRP:0.362 WGSIP17, 13-14 September 2015, Norkörrping, Sweden

T2 and 300hPa Geopotential height February, 2010 EnKF ERA-I Preliminary Run with MIROC3m (T42L20 AGCM and 1-deg OGCM) Shade: SAT Contour: Z300

Monthly Mean Precipitation July 2010 GPCP July 2010 GPCP EnKF EnKF July 2010 Internannual variability (1980-2010)

Atmosphere and Ocean Data Rescue Recently processed SLP data (1890 – 1940; Kubota et al. JAMSTEC) Upper air observations to be digitized  Expected data distribution (red) of marine met. Obs. by the Japanese imperial navy (1903-1944, ~ 1 million) Participating in and collaborating with international activities: ACRE, ICOADS, ICA&D, IQuOD Collaborating with ACRE, ICOADS, and IQuOD …