CLIVAR ENSO WS, Nov 17-19, 2010 ENSO simulation in MIROC: Perspectives toward CMIP5 M. Watanabe1, M. Chikira2, Y. Imada1, M. Kimoto1 and MIROC modeling team 1: Atmosphere and Ocean Research Institute (AORI), The Univ. of Tokyo 2: Research Institute for Global Change (RIGC), JAMSTEC, Japan Watanabe et al. (2010, JC in press.) Motivation (or triggering) Obs.(ProjD_v6.7&ERA40) MIROC3. T42 Collins et al. (2010, Nature Geo.) Improvements in an update (MIROC5) Obs.(ProjD_v6.7&ERA40) MIROC3. T42 impact of resolution impact of new model physics MIROC3. T213 MIROC5. T85 ENSO in CGCMs ENSO diversity in CMIP3 models -> Controlling ENSO in complex system is still challenging ENSO diversity in CGCMs is likely due to the atm. component - Schneider 2002, Guilyardi et al. 2004, 2009 In particular, convection scheme potentially has a great impact • CMT - Wittenberg et al. 2003, Kim et al. 2008, Neale et al. 2008 • Entrainment (incl. cumulus triggering) • - Wu et al. 2007, Neale et al. 2008 Low clouds - Toniazzo et al. 2008, Lloyd et al. 2009 Perturbing cumulus convections Entrainment rate () Conventional A-S scheme: prescribed C-S scheme: state dependent w2 w2 aB 2a(1 - l ) B l 2 , z w Altitude [eta] Chikira-Sugiyama convection scheme: Vertical profiles of in a single column model Mixture of A-S and Gregory schemes C-S A-S Efficiency of the entrainment controlled by l (large l -> suppress deep clouds) Cloud type Chikira and Sugiyama (2010, JAS) Sensitivity experiments w/ T42 MIROC5 exp l Length L500 0.5 85 L525 0.525 85 L550 0.55 85 L575 0.575 85 * l0.53 is the default value in the official T85 CTL ENSO in MIROC5 L500 GCM L525 Reality? Obs. L550 L575 artificial? CP El Niño? Comparison of the ENSO structure Nino3-regression along EQ a L575 Precipitation Nino3 SST Std Dev L500 m Zonal stress longitude Lloyd et al. (2009) As ENSO amplifies, maximum in both precipitation and x anomalies be stronger but shifted to the western Pacific -> reduction in the effective Bjerknes feedback Mean state differences Deviations from the ensemble mean SST precip. L500 L550 ENSO amplitude L525 L575 Larger l (efficient cumulus entrainment) -> drier & colder mean state in E. Pacific <-> weaker ENSO ENSO metric in MIROC5 Cold tongue dryness (CTD) index AGCM experiments (5yrs each) exp Remark l L500a 0.5 L525a 0.525 L550a 0.55 Coupled feedbacks L575a 0.575 Direct effect of convection L500b 0.5 l=0.575 over Nino3 L575b 0.575 l=0.5 over ITCZ SST & ice from CGCM ensemble mean Coupling always works to reduce the precipitation contrast Mechanism of convective control Wet cold tongue -> enhanced effective Bjerknes feedback Dry cold tongue -> reduced effective Bjerknes feedback Summary & remarks In MIROC5, a parameter for the cumulus entrainment (l) greatly affects the ENSO amplitude ENSO controlling mechanisms involve: Direct changes in convective systems over the E. Pacific Coupled feedback (incl. ENSO structural change) The mean meridional precipitation contrast over the E. Pacific is a relevant indicator of the ENSO amplitude in MIROC. * the former is not necessarily the cause of the latter!! Generality? Similar experiments with the other GCMs desired Implication for the future change of ENSO CTDI-ENSO in CMIP3 models CTL or 20C MIROC5 GDFL CM2.1 (by J-S Kug) CMIP3 Axes of the parametric and structural uncertainties are quite different!! CTDI-ENSO in CMIP3 models 2xCO2 or A1b Sensitivity to increasing CO2 agrees well with the axis of the parametric uncertainty in MIROC5 → by chance? What’s the issues for CMIP5/AR5? “KNOWN” & UNKNOWN Relatively robust: mean change (weakening of trades / shoaling of thermocline / warming in the e. Pacific) Not robust: ENSO property changes (amplitude/preference etc) TODO Theory & GCM (e.g. BJ index -> CMIP3/CMIP5 outputs) Verification of convective processes using TRMM Combined analyses to AMIP+20C Single param. perturbed experiments -> PPE Climate sensitivity and ENSO changes Extensive use of near-term predictions (assimilation/hindcasts) What’s the issues for CMPI5/AR5? Nino 3.4 SST std dev [K] Result from the Hadley Centre PPE ? Toniazzo et al. (2008) Equilibrium climate sensitivity [K] Does this occur only when the model’s ENSO is controlled by low clouds? But, it seems consistent with MIROCs, too … backup RR2002 2003 EarthSimulator Simulator2 Earth “Kakushin” 2007 2008 2009 2010 AR4 AR5 AR5 data submission MIROC history MIROC3.2 T42+1deg (med) T106+1/4x1/6deg (hi) 2013 MIROC4.0 (bug fixed version of 3.2) T42+1deg (med) T213+1/4x1/6deg (hi) MIROC-ESM T42L80+1deg MIROC4.1 (prototype new model) MIROC5.0 T85+1deg (med) Introduction ENSO diversity in CMIP3 models -> Controlling ENSO in complex system is still challenging MIROC3 (for AR4) -> MIROC5 (for AR5) Most of the atm. physics schemes replaced Std resolution: T85L40 atm. 0.5x1 deg ocean MIROC5 MIROC3med ENSO was greatly improved Guilyardi et al. (2009) Mechanism of the convective control What is likely to be happening in MIROC5: Large l (effective entrainment) → deep cumulus suppressed (→ more congestus in ITCZ → drying the cold tongue due to subsidence) → strong north-south moisture contrast in the eastern Pacific (mean state change) → precip./x response to El Nino confined to the western-central Pacific → weaker effective Bjerknes feedback → weak ENSO Feedback to the mean state New version of MIROC Atmos. Dynamical core MIROC3 (for AR4) MIROC5 (for AR5) Spectral+semi-Lagrangian Spectral+semi-Lagrangian (Lin & Rood 1996) (Lin & Rood 1996) V. Coordinate Sigma Eta (hybrid sigma-p) Radiation 2-stream DOM 37ch 2-stream DOM 111ch (Nakajima et al. 1986) (Sekiguchi et al. 2008) Diagnostic (LeTreut & Li 1991) + Simple water/ice partition M-Y Level 2.0 Prognostic PDF (Watanabe et al. 2009) + Ice microphysics (Wilson (Mellor & Yamada 1982) (Nakanishi & Niino 2004) Prognostic A-S + critical RH (Pan & Randall 1998, Prognostic AS-type, but original scheme (Chikira & Sugiyama 2010) Cloud Turbulence Convection & Ballard 1999) MYNN Level 2.5 Emori et al. 2001) simplified SPRINTARS SPRINTARS + prognostic CCN (Takemura et al. 2002) (Takemura et al. 2009) Land/ River MATSIRO+fixed riv flow new MATSIRO+variable riv flow Ocean COCO3.4 COCO4.5 Sea-ice Single-category EVP Multi-category EVP Aerosols Entrainment rate () Vertical profiles of in a single column model C-S A-S eta New convection scheme Mixture of A-S and Gregory scheme Conventional A-S scheme: prescribed C-S scheme: Cloud type dependent upon buoyancy and cloud-base mass flux What’s the consequence? Chikira and Sugiyama (2010) Deep cumulus altitude Strong w’ -> large Shallow cumulus Weak w’ -> small Both work to increase middle level cumulus that was less in A-S Not necessary to use empirical cumulus triggering function ENSO in MIROC5 A-O coupling strength MIROC3med MIROC5 Guilyardiet al. (2009) Mean state differences SST precipitation Obs. model Narrow warm pool, but the single ITCZ is well reproduced over the e. Pacific Mean state differences Model clim. L575-L500 w Qcum More congestus? Feedback coefficients Both differences in a and m do not explain the different ENSO amplitude! Comparison of the ENSO structure Contour: regression of Eq. temperature anomaly on to Nino3 (per 1K) Shade: difference from the grand ensemble mean White contour: 19,20,21 degC mean isotherms Mean state differences RH in the eastern Pacific Wet Dry Contour: annual mean clim. Shade: diff from the ensemble mean RH-precipitation relationship RH600 histgram Wet (dry) mid-troposphere is less (more) frequent in Nino3 region for larger l Composite Pr. wrt RH600 “Rich-get-richer” for larger l ? Mechanism of convective control Composite cumulus heating wrt CAPE in AGCM Large l (efficient entrainment) works to prevent deep cumulus convection Opposite direction of change in congestus clouds Question Small but cooler cold tongue (=larger zonal SST gradient) for large l: is it consistent with weaker ENSO? A simple tropical climate model (Jin 1996, Watanabe 2008) Te -T (Te - Tr ) - ( w) rbL hw - rhw 2 Te - Tse Hm w -ad 0 - m (Tr - Te ) he hw bL 1 - tanh ( H he - z0 ) / h* Tse Tr - (Tr - Tr 0 ) 2 Stationary solutions Question Radiative heating Obs. Mean Te Range of mean Te in four runs Larger l ? Std of J96 Larger l ? Bjerknes feedback efficiency Cooler cold tongue & weaker ENSO can coexist if l-1 ∝ bL Can feedback factors explain the model’s diversity? ENSO parameters in CMIP3 models r > 0, may be consistent with what a means r < 0, inconsistent with what m means Nino3 SST Std Dev a (net heat flux damping) m (Bjerknes feedback) Lloyd et al. (2009) Convective control of ENSO? Most of the recent studies point out the role of cumulus parameterization in ENSO simulations CCSM3 : Cumulus convection (Neale et al. 2008) GFDL CM2: Cumulus convection (Wittenberg et al. 2006) IPSL: Cumulus convection (Guilyardi et al. 2009) SNU: Cumulus convection (Kim et al. 2008) HadCM3: Low cloud (Toniazzo et al. 2008) What is meaningful with MIROC5? ー ENSO controlled by a single parameter (1D phase space) ー mean state changes are not large (but large for the TRH) Generality ? ー diff model has diff bias, so the mechanisms may not be unique Mean state (SST) Mean state (precipitation) seasonal cycles over the eastern Pacific CMAP Model EM Diff L575-L500 Watanabe et al. (2010) Mean state and ENSO seasonal cycles of clim SST & ENSO amplitude Nino3 SST mean seasonal cycle Nino3 SST std dev Mean state differences SST Contour: annual mean clim. Shade: diff from the grand ensemble mean SST is warmer in E. Pacific when ENSO is stronger, but the difference is quite small (less than 2 %) Mean state differences Precipitation Contour: annual mean clim. Shade: diff from the grand ensemble mean Wetter in E. Pacific for larger ENSO The absolute difference is quite small (less than 1mm/dy), but relative difference is quite large (more than 50%!) ENSO in MIROC5 SST mode or thermocline mode? Guilyardiet al. (2006) Convective control of ENSO New version of MIROC (MIROC4.5) State-dependent entrainment in cumulus scheme (Chikira 2009) Assumption between the entrainment rate and updraft velocity w (Gregory aB w2 w2 l 2 , 2a(1 - l ) B w z The parameter l is found to control the frequency of deep cumulus clouds (l->large, suppress deep clouds) hence affect ENSO amplitude l=0.5 l=0.55 l=0.525 MIROC3.2 Guilyardi et al. (2009) 2001) Convective control of ENSO Mean climate is quite similar to each other; nevertheless, ENSO amplitude is different with factor 2!! l=0.55 SST T along Eq. Pr/SLP/ Regression with Nino 3 index l=0.5 Implication to 20th century trend 20C runs MIROC3 MIROC5 Cl trend (%/100y) SST trend (K/100y) Decrease (-0.28%/100y) Increase (+0.47%/100y) Tropical Cl (30S-30N) Likely due to fast response (but change is much slower) (CO2 increase; abrupt vs gradual) -> (fast response)?