Climate Simulation and Modelling at Ministry of Earth Sciences A.K.Sahai Indian Institute of Tropical Meteorology (IITM) First annual NKN workshop, IIT-B, 31 Oct. -2 Nov., 2012
Major Objectives of MoES To provide the country best possible weather forecast (short range ) and climate prediction (long range ) To conduct the R & D required to improve the skill of both weather and climate forecasts To conduct regional climate change research to provide reliable projection of monsoon under climate change First annual NKN workshop, IIT-B, 31 Oct. -2 Nov., 2012
Climate: A statistical description of weather Climate Weather First annual NKN workshop, IIT-B, 31 Oct. -2 Nov., 2012
Global climate Modeling Basic equations for Weather and Climate Models in pressure coordinate system D V ˆ f k V F --------------(1) Dt RT --------------(2) p p --------------(3) . V 0 p . T RT T Q V . T --------------(4) t C p C p p F = Friction (turbulent dissipation) . Q = Non Adiabatic Heating = Net Radiation + Latent heat (clouds) + Sensible heat First annual NKN workshop, IIT-B, 31 Oct. -2 Nov., 2012
Complexities involved in a Climate Modelling System
Calculation of Heating Distribution Radiation Stratospheric Incoming SW Aerosols Chemistry Outgoing LW • Direct Rad Eff • Heating • Indir eff thr clouds • Stability AGCM/OGCM NS Eqs. Global 3-D + Boundary Layer Land-Surface time Turbulence Processes • Fluxes • Vegetation • Mixing • Soil moisture • Dissipation Clouds • Convective • Startiform First annual NKN workshop, IIT-B, 31 Oct. -2 Nov., 2012
Key Uncertainties for Climate : High Clouds: Dominant effect is that they Trap heat (warm) More Clouds=Warming Fewer Clouds=Cooling
Source: Schär
Source: Schär
Scale Interaction and Parameterization All physical processes involved in heating arise from complex small scale processes, that need to be parameterized in the model Accuracy of paramereization determines heating distribution and hence weather and climate More complex paramerization requires more computation Improvement of parameterization need R & D First annual NKN workshop, IIT-B, 31 Oct. -2 Nov., 2012
First annual NKN workshop, IIT-B, 31 Oct. -2 Nov., 2012
Horizontal Resolution of the Contemporary AGCM/OGCM 500 km 300 km 150 km 75 km
Simulations at 200km and 50 km Resolution is of key importance for the representation of hydrological Cycle and extreme rainfall events First annual NKN workshop, IIT-B, 31 Oct. -2 Nov., 2012
Source: Kinter First annual NKN workshop, IIT-B, 31 Oct. -2 Nov., 2012
Weather & Climate Prediction Chaotic System; Probabilistic prediction Large number of ensemble of each prediction Initial value problem; 4-D data Assimilation Variational Assimilation ; Adjoint of the model; Extremely computation intensive; It is found that preparation of the I.C. for operational weather prediction at high resolution takes more computer time than actually making the prediction! First annual NKN workshop, IIT-B, 31 Oct. -2 Nov., 2012
First annual NKN workshop, IIT-B, 31 Oct. -2 Nov., 2012
The Butterfly Effect • Discovery of the “butterfly effect” (Lorenz 1963) • Simplified climate model… When the integration was restarted with 3 (vs 6) digit accuracy, everything was going fine until… Time First annual NKN workshop, IIT-B, 31 Oct. -2 Nov., 2012
The Butterfly Effect • Solutions began to diverge Solutions diverge Time First annual NKN workshop, IIT-B, 31 Oct. -2 Nov., 2012
The Butterfly Effect • Soon, two “similar” but clearly unique solutions Solutions diverge Time First annual NKN workshop, IIT-B, 31 Oct. -2 Nov., 2012
The Butterfly Effect Chaos • Eventually, results revealed two uncorrelated and completely different solutions (i.e., chaos) Solutions diverge Time First annual NKN workshop, IIT-B, 31 Oct. -2 Nov., 2012
The Butterfly Effect • Ensembles can be Chaos used to provide information on forecast uncertainty • Information from the ensemble typically consists of… (1) Mean Solutions diverge (2) Spread (3) Probability Time Ensembles useful in this range! First annual NKN workshop, IIT-B, 31 Oct. -2 Nov., 2012
The Butterfly Effect Chaos Ensembles extend predictability • Ensembles extend predictability… • A deterministic solution is no longer skillful when its error variance exceeds climatic variance • An ensemble remains skillful until error Solutions diverge saturation (i.e., until chaos occurs) Time Ensembles especially useful in this range! First annual NKN workshop, IIT-B, 31 Oct. -2 Nov., 2012
Data Assimilation 1975 1985 1990 1997 1999 2005 - Univariate SI Multivariate SI 3DVAR 4DVAR 4DVAR/EnKF 4DVAR/EnKF IR sounders IR/MW sounders Adv. sounders Adv. sounders Adv. sounders Scatterometer Scatterometer Scatterometer Scatterometer Increasing TRMM TRMM TRMM complexity Rainfall Rainfall Vast increase in data assimilation assimilation Mesoscale assimilation Chemical species
Success story of Numerical Weather Forecasting! Great Improvement in medium-range forecast skill. Note the convergence of skill in NH and SH From ECMWF 12-month running mean of anomaly correlation (%) of 500 hPa height forecasts First annual NKN workshop, IIT-B, 31 Oct. -2 Nov., 2012
Improving the Forecasts Must improve the MODEL and Data Assimilation Ex: Consider seasonal prediction with a CGCM 100 yr integration for testing mean climate Hindcast experiments to test prediction skill; 25 member ensemble x six month prediction x 20 years = 250 year integration Must turn around within a few days so that other improvement could be tested! First annual NKN workshop, IIT-B, 31 Oct. -2 Nov., 2012
First annual NKN workshop, IIT-B, 31 Oct. -2 Nov., 2012
Evolution of Climate Models in last 5 decades at renowned climate centers India’s Present status Leading climate centers’ status
[Range: Assumed efficiency of 10-40%. 0 - Atmospheric General Circulation Model (AGCM; 100 vertical levels) 1 - Coupled Ocean-Atmosphere-Land Model (CGCM; ~ 2X AGCM) 2 - Earth System Model (with biogeochemical cycles) (ESM; ~ 2X CGCM)] Computational requirements of climate models. Range: Assumed efficiency of 10-40%. 0 - Atmospheric General Circulation Model (AGCM; 100 vertical levels) 1 - Coupled Ocean-Atmosphere-Land Model (CGCM; ~ 2X AGCM) 2 - Earth System Model (with biogeochemical cycles) (ESM; ~ 2X CGCM)] First annual NKN workshop, IIT-B, 31 Oct. -2 Nov., 2012
How a code in a coupled model works? Source: Anne Roches & Piccinali
International Centres HPC Current Capacity NERSCC, USA Cray XT5 ~1.17PF (peak) UKMet Office IBM P6 ~150TF IBM P7 ~900TF (by 2011) NCAR, USA IBM P5/P6 ~80TF NCEP, USA IBM P6 ~90TF German Met Office IBM P6 ~165TF ECMWF IBM P6 ~300TF IBM P7 ~1PF (by 2011) JAMSTEC Earth Simulator ~131TF KMA Cray XT5 ~600TF National Supercomputing NUDT ~4.7PF Center in Tianjin China Oak Ridge National Laboratory Cray XT5 ~2PF USA Supercomputer(JA GUAR) These centres are also having additional HPC for operational/other usage .
HPC at ECMWF Phase 1 Number of Clusters 2 compute clusters (272 nodes each) Compute Nodes 272 x 32-core POWER6 (SMT) Peak Performance ~300TF (Total) Sustained Performance ~20TF First annual NKN workshop, IIT-B, 31 Oct. -2 Nov., 2012
In India where do we stand? National Weather HPC Current Capacity and Climate Centres NCMRWF, Noida IBM P6 ~23TF IMD, New Delhi IBM P6 ~15TF INCOIS, Hyderabad IBM P6 ~7TF IITM, Pune IBM P6 ~70TF 2010 July Ranking 94 th 2010 November Ranking 137 th 2011 November Ranking 403 rd
No. of Systems in Top 500 from Different countries India:2 (0.4%) USA:263 China:74 Source: Top 500 list, Nov. 2011
Dynamical Seasonal Prediction of Indian Monsoon JJAS Rainfall – 2010 (CFS V1.0) Issues in April IITM CFS T126 IITM CFS T62 T126L64 T62L64 Central Indian drought predicted by CFS model IMD Above normal rainfall over southern peninsular India
Dynamical Seasonal Prediction of Indian Monsoon With Initial Conditions generated within India at (INCOIS & NCMRWF) JJAS Rainfall – 2011 (Issued in March) IITM CFS T62 IITM CFS V2.0 T126 T126L64 T62L64 Central Indian above normal rain predicted by CFS model IMD Below normal rainfall over Upto 10 th September southern peninsular India
Predictions for 2012: Predicted Vs. Observed Actual Rainfall Departure (IMD) Monsoon Performance = 92 % Monsoon Performance = 100 ± 4.5 %
Recommend
More recommend
