Activities on model error at Météo- France Philippe Arbogast, Marie Boisserie (CNRM-GAME, Toulouse) With contributions by I. Beau, H. Douville, F. Bouyssel, CH. Lac, D. Ricard, Y. Seity, R. Honnert, L. Descamps 7-9 July 2010
Outline 1. Validation of GCM parameterizations 2. The use of Single Column Model (SCM) 3. Large Eddy Simulations (LES) to validate turbulence 4. Where are the sources of NAO predictive skill? using nudging 5. Computation of effective horizontal resolution of a model using spectra 3
Validation of GCM parameterization schemes (turbulence and convective schemes) on Western Africa; comparison of LAM and CRM simulations D. Pollack, J.F. Gueremy and I. Beau Image aladin Explicit simulations of convection / Parameterized simulations: (Méso-NH model) / (Aladin-Climat model) of observed case studies ALADIN-Climat simulations performed on the same domain, with the same initial and lateral conditions as Méso-NH. at: 10, 50, 125 and 300 km resolution and for 31 and 91 levels 4
Validation of GCM parameterization schemes (using Model to Sat. approach) M. D Errico, I. Beau, D. Bouniol, F. Bouyssel EUCLIPSE FP-7 project CloudSat Radar simulator 1.5 km 1.5 km CALIPSO 12.5 km Lidar simulator 12.5 km 5
Altitude (km) Altitude (km) Validation of GCM parameterization schemes (using Model to Sat. approach) CloudSat Radar simulator Reflectivity (dbz) Reflectivity (dbz) Lack of overshooting in the model.. 6
Outline 1. Validation of GCM parameterizations 2. The use of Single Column Model (SCM) 3. Large Eddy Simulations (LES) to validate turbulence 4. Where are the sources of NAO predictability? using nudging 5. Computation of effective horizontal resolution of a model using spectra 7
LES/SCM (single column model) setting for parameterization validation (J. Pergaud, S. Malardel, V. Masson) Validation of a Mass flux scheme for unified parameterization of dry and cloudy convective updraft LES SCM/1D GCM 8
LES SCM w' θl ' w' θl' ARM Case : part of the Eurocs project (1997) (Brown et al., 2002) Diurnal cycle of shallow cumulus convection over land. Intercomparison Study : Lenderink et al. (2002) 9
Outline 1. Validation of GCM parameterizations 2. The use of Single Column Model (SCM) 3. Large Eddy Simulations (LES) to validate turbulence 4. Where are the sources of NAO predictability? using nudging 5. Computation of effective horizontal resolution of a model using spectra 10
LES to develop and validate turbulence scheme (TKE) (R. Honnert PhD) What happens at intermediate horizontal scales? E(explicit)>E(subgrib) E(explicit)<E(subgrib) 11
LES to develop and validate turbulence scheme (TKE) (R. Honnert PhD) subgrid explicit 12
Outline 1. Validation of GCM parameterizations 2. The use of Single Column Model (SCM) 3. Large Eddy Simulations (LES) to validate turbulence 4. Where are the sources of NAO predictive skill? using nudging 5. Computation of effective horizontal resolution of a model using spectra 13
Motivation DEMETER2 DJF hindcasts (1958-2001): Poor predictive skill of the North Atlantic Oscillation index (e.g. Palmer et al. 2004) 14
Model and simulations Arpège-Climat atmospheric spectral GCM in its low-top configuration (T63L31) => only 4 levels above 100 hpa (model top at 10 hpa) Prescribed observed SST and radiative forcings (GHG, sulfate and volcanic aerosols) Ensembles of 5-member integrations from 1970 to 2000 (including a 1-yr spin-up): CT: Control (no nudging, observed SST) NS: Stratospheric nudging north of 25 N NCS: Tropospheric nudging between 25 S-25 N 15
Grid point nudging dx/dt = D(X) + P(X) l(x-x ref ) Nudging is applied: at each time step (every 30 min) towards linearly interpolated 6-hourly data to U/V and T using a 5-hour and 12-hour e-folding time respectively in a 3D domain with a smooth transition between the nudged and free atmosphere ERA40 16
Control experiment Nudging of the tropical troposphere 1971-2000 time series of DJF NAO principal components. Ensemble mean anomalies (thick red lines) are compared to ERA40 (in black) and spread is also shown (+/- 1 standard deviation in dashed red lines and minimum and maximum anomalies in solid red lines). R is the ensemble mean anomaly correlation coefficient with ERA40. Nudging of the extratropical stratosphere 17
Outline 1. Validation of GCM parameterizations 2. The use of Single Column Model (SCM) 3. Large Eddy Simulations (LES) to validate turbulence 4. Where are the sources of NAO predictability? using nudging 5. Computation of effective horizontal resolution of a model using spectra 18
Assessment of spectra / effective horizontal resolution checking Log k 19
20 Meso-nh (2.5 km): effective resolution is 4-6DX
21 Arome (2.5 km): effective resolution is 8-9DX
Kinetic energy Spectrum vs forecast range to address the spin-up (~3 hours) wavenumber 22
Summary 1. Comparison against global climatologies/advanced satellite observations 2. LES/SCM/CRM to tune, to choose the best formulation, to address the need of some schemes (convection or turbulence) 3. Effective resolution using spectra 4. Nudging within GCM together with process studies (to improve the understanding of the physics of teleconnections ) 5. Split forecast uncertainty in terms of initial condition error and model error : Marie s talk. 23