550 likes | 559 Views
This presentation by Tony Hirst from CSIRO discusses the four main components of coupled climate models and how data is computed and simulated. It also highlights the limitations of the CSIRO Mk3.0 model and the improvements in the Mk3.5 and Mk3.6 models.
E N D
Coupled climate modelling at CSIRO • Presented by • Tony Hirst • CAWCR
Coupled climate models Four main components: the atmosphere, the land surface and biosphere, the ocean, and polar ice Data are computed in short (~15-minute) time-steps over a global grid for a period of months, years or centuries Models simulate daily weather and average climate patterns
CSIRO climate model grids Facilitated by improved computing power and optimised programming Mk3 grid Mk2 grid
CSIRO Mk3 climate model Temperature (oC)
Australian global climate models • BMRC global coupled climate modelling • BCM – contributed to IPCC TAR (2000) • also seasonal prediction focus - POAMA • CSIRO global coupled climate modelling • Mk2 – contributed to IPCC TAR (2000) • Mk3.0 – contributed to IPCC AR4/CMIP3 (2004) • Mk3.5 – contributed to CMIP3 (2006) • Mk3.6 – Control simulation under way Australian Community Climate and Earth System Simulator (ACCESS) – Under development
CSIRO Mk3 global coupled climate model Atmosphere:Grid T63 (1.9 x 1.9); 18 levels - hybrid ,p Rotstayn prognostic cloud scheme Ocean: MOM 2.2 code; Grid 0.95NS x 1.9EW; 31 levels (8 in top 100 m) Griffies (1998) skew-diffusion form of eddy-induced transport Sea Ice: Flato-Hibler cavitating-fluid rheology Semtner Thermodynamics (3 layer) Land surface: Soil model – 6 levels, 9 soil types 13 vegetation types Snow cover model – 3 layer Reference: Gordon et al., 2002 www.dar.csiro.au/publications/gordon_2002a.pdf
CSIRO Mk3.0 and Mk3.5 • CSIRO Mk3.0 model limitations • Cool bias in the climate and ongoing cooling trend • Poor Southern Ocean Circulation and stratification • Problems in tropical Pacific climate (mean state, ENSO) • Mk3.5 – Improved physics over Mk3.0 • Improved scheme for oceanic eddy-induced transport • Improved scheme for wind-forced mixing in the oceanic mixed layer • Upgraded sea ice numerics • Ocean surface current speed included in wind stress calculation • Improved river routing • Model carefully rebalanced
Control simulations – global mean surface air temperature change (5yr mean)
Sea ice extent – Southern Hemisphere (106 km2) Mk3.0 0 100 200 300 400 500 Mk3.5 0 100 200 300 400 500
SST Model – obs. Mk3.0 years 101-200 Mk3.5 years 101-200
SST Model – obs. Mk3.0 years 401-500 Mk3.5 years 401-500
Tropical SST (difference Model – Observed) Annual Mean Mk3.0 Years 201 – 210 Mk3.5 Years 201 – 210 °C
Spatial pattern of ENSO SST anomalies(C)SST anomalies regressed onto1 stdev NINO 3.4 index Mk3.5 control (years 301-400) Mk3.0 control
Power spectrum of NINO 3.4 Index Mk3.5 control years 201-300 Mk3.0 control 6.4 yr 4.9 yr 5.5 yr 2.1 yr 3.5 yr 3.8 yr Observed 5.1 yr 3.5 yr
Power spectrum NINO 3.4 6.4 yr Mk3.5 control years 201-300 4.9 yr 3.8 yr 5.2 yr Mk3.5 control years 301-400
Correlation Between Rainfall and NINO3.4 – Mk3.5 model (years 301-400)
MJO variabilityPower spectrum for u200(eastward propagating) 45 days period
Mixed layer depth – September (depth in metres) Observation- based data set Mk3.0 Years 201 – 210 Mk3.5 Years 201 – 210
Mk3.5 simulations • Control simulation (Completed at 1300 years length) • Climate change simulations (following CMIP3/IPCC AR4 protocol) • CMIP simulation – 1% p.a. CO2 concentration increase for 80 years • Three 20th century simulations • SRES A2, A1B, B1, “COMMIT” • Climate change forcings applied (basic set for CMIP3 protocol) • Greenhouse Gas concentrations (as equivalent CO2) • Ozone concentrations (as a 3-D time dependent field) • Sulphate aerosol (direct effect only)
IPCC SRES scenarios for AR4 simulations Atmospheric CO2 concentration Historical “COMMIT”
Mk3.5 simulations Global average surface air temperature
Mk3.5 simulations – Output data • Output data summary • Atmospheric fields: monthly, 6 hourly • Oceanic fields: monthly • Most data now in standard IPCC AR4 format • Key contact – Mark Collier (mark.collier@csiro.au) • Availability • 1. PCMDI (Mk3.0, Mk3.5 available now)http://www-pcmdi.llnl.gov/ • 2. OPeNDAP from CSIRO server
Recap on Mk3.5 • Mk3.5 developed and displays significant improvement in several key attributes • Drift largely eliminated • Southern Ocean circulation and stratification improved • Character of ENSO variability modestly improved • Set of climate change simulations (IPCC AR4 protocol) is nearly complete with extensive saving of monthly and 6 hourly data • Project aims to support usage of model output in other stakeholder projects
CSIRO Mk3.6 • Mk3.6 – Improvements over Mk3.5 • Physics • Comprehensive prognostic aerosol scheme • Upgraded radiation code • Upgraded boundary layer formulation • Solution • Significant improvements in simulation of Australian climate including variability • Simulations: 1. Present-day control performed – 70 years • 2. Pre-industrial control under way • – currently at 400 years
Mk3.6 – first mode of rainfall variability Modelled – Mk3.6 vs CMIP3 models Observed CSIRO Mk3.6
ENSO Impact on Australian aerosols – Mk3.6Correlation aerosol concentration and NINO3.4 SST anomaly Mineral dust Sulfate Carbonaceous Mk3.6
Future simulations with CSIRO Mk3.6 • Mk3.6 – Future simulations • Simulations to be performed as part of partnership with QCCCE • Pre-industrial control following CMIP5 specification • Historical (~20th century) simulation following CMIP5 specification • All four RCP experiments to 2100 • 20th century simulations with individual agents • Ensemble simulations • To be performed on new QCCCE machine – computational resources should be adequate. • Set of simulations may readily be extended to include the remaining core CMIP5 (IPCC AR5) experiments
Australian Community Climate and Earth System Simulator (ACCESS) • Joint initiative of the Bureau of Meteorology and CSIRO, with university sector involvement to: • Develop anational approach to model development for climate and weather prediction • Focus on the needs of a wide range of stakeholders
ACCESS - Aims • Produce a modelling system capable of supporting: • NWP – Bureau operations • Seasonal prediction – Bureau operations • Provision of model-derived climate information • Climate and climate change simulation • Contributes to IPCC Assessment Reports • Includes full carbon cycle • Includes scales appropriate to support decision making • Spatial scales (regional) • Temporal scales (decades – centuries) Morton and Love (2005) (paraphrased)
OPS ACCESS Modelling System M e t O f f i c e VAR What we are aiming to develop as the ACCESS Earth System Model Dynamic Vegetation Atmosphere Assimilation Atmospheric Chemistry Land Surface Coupler Sea Ice Ocean Carbon cycle Ocean BODAS Dynamic Ocean Primary Prod. OBS
ACCESS – First version of global coupled model Atmosphere UM 7.3 Land Surface CABLE Coupler OASIS 3.2.5 Sea Ice CICE 4.0 Ocean GFDL MOM4p1
ACCESSCurrent Computing Infrastructure • The main computing facility for Mk3 and ACCESS development has been the HPCCC (located at the Bureau), using two NEC SX-6 machines. • The main computing facility for ACCESS development is now the NCI NF (located at the ANU, Canberra), using two SGI machines (total cores 3200). • NCI machine upgrade (to 12,000 core Sun Constellation) is due in December 2009.
ACCESSKey Timeline required for AR5 • Climate Change simulation • 2008/2009Complete technical coupling • 2009/2010Test and tune coupled system • 2010 2nd halfPerform and submit CMIP5/AR5 simulations • Initial aim is to perform minimum set of simulations for CMIP5/IPCC AR5
CMIP5 and IPCC AR5 • Coupled Model Intercomparison Project 5 (CMIP5) • International project featuring clearly defined model experimental design including prescribed emissions and concentrations scenarios and model output data protocols. • IPCC AR5 will use the model results submitted to “CMIP5” • Will continue past the AR5 • Experimental design now set (Taylor et al. ,2009) following extensive reviews. • Three experimental suites (a particular model may enter any or all) • “long-term” simulations • “near-term” (decadal) hindcast/prediction simulations • “atmosphere-only” (prescribed SST) simulations – for especially computationally demanding models • Experiments are designated as “Core”, “Tier 1”, “Tier 2”
Long-term experiments • The principal long-term experiments are: • 500+ year control coupled model - preindustrial • Historical (1850 – 2005) (ensemble optional) • Representative concentration pathways (RCPs) 2005 – 2100 • RCP8.5 (Core) • RCP6 (Tier 1) • RCP4.5 (Core) • RCP2.X (Tier 1) • Extension of RCP8.5, RCP4.5 and RCP2.X to 2300 (Tier 1, 2) • Several idealised experiments
CMIP5 – RCP CO2 emissions and concentrations • Key to projections are four “representative concentration pathways” • RCP8.5, RCP6.0, RCP4.5, and RCP2.X • The number (8.5, etc) indicates approximate W m-2 radiative forcing by 2100. • However, these four pathways will have concentrations of GHGs specified (though corresponding emissions series will be available for use in certain coupled climate-carbon experiments for those groups that have that capability). • The RCP8.5 and RCP4.5 are to be “core”, with RCP2.X the next highest priority
Near-term experiments • The principal near-term experiments are: • Simulations initialised at 1960, 1965, …., 2005, each of 10 years duration. Ensemble size of 3 (optional 10) members. • Extend simulations initialised at 1960, 1980, 2005 to 30 years duration. RCP4.5 is to be used for the period 2005 – 2035. The method of initialisation is to be left entirely to the modelling groups. (Initialisation is one of the major scientific questions.)
Climate models and IPCC AR5 timelines • Current “rule of thumb” is that models will need to complete submission of their simulations by late 2010 to have maximum impact in the IPCC AR5 • Impact fades steadily thereafter until no impact perhaps by late 2011 • ACCESS has plans as to what may be delivered to CMIP5 by certain times. These are quite rubbery as model development time scales are uncertain. • ACCESS aims to deliver the core long-term experiments in to the CMIP5 in time for use in AR5. • Australia will not be delivering to the near-term (decadal prediction) part of the AR5. No initialisation scheme readily at hand. • In collaboration with CAWCR SP Research Group, potential to develop capacity for decadal predictability work in time for delivery to CMIP5 later
Contact CSIRO Phone: 1300 363 400+61 3 9545 2176Email: enquiries@csiro.auWeb: www.csiro.au Contact Tony HirstTitle: Stream leader, ACCESS Stream Phone: (+61 3 9239 4531)Email: tony.hirst@csiro.auWeb: www.cmar.csiro.au
CSIRO Mk3.5 • Mk3.5 – Improvements over Mk3.0 • Physics • Improved river routing • Upgraded sea ice numerics • Ocean surface current speed included in wind stress calculation • Improved scheme for oceanic eddy-induced transport • Solution • Much reduced drift over that in Mk3.0 (< 0.05C/century) • Much improved Southern Ocean circulation • Full set of CMIP3 simulations performed • (including 1300 year control)
Surface air temperature change for CMIP3 models Idealised CMIP forcing (1% p.a. CO2 concentration change) + 3 + 2 + 1 0 - 1
Mk3.5 simulations – Output data • Availability • 1. PCMDI (Mk3.0, Mk3.5 available now) • http://www-pcmdi.llnl.gov/ 2. Direct copy on CSIRO server cherax cherax:~IPCC/data/M/E/T/R/file_name.nc Where directory is constructed with: M model version (mk3.0/mk3.5) • E experiment (PICntrl, 20C3M, SRESA2 etc.) • T table (A1a, A2b, O1e etc.) • R ensemble (run1, run2, run3) • 2. OPENDAP • (i) Browser: • http://hpsc.csiro.au/cgi-bin/OpenDAP/CMAR_mk3/nph-dods/ • Access via a username and password. • (ii) Ferret: • Usehttp://CMAR_mk3:ipcc4@hpsc.csiro.au/cgi-bin/OpenDAP/CMAR_mk3/nph- • dods/ipccdata/M/E/T/R/file_name.nc • (iii) DODS enabled ncks; • Use ncks http://CMAR_mk3:ipcc4@hpsc.csiro.au/cgi-bin/OpenDAP/CMAR_mk3/nph- • dods/M/E/T/R/file_name.nc
Mk3.6 – second mode of rainfall variability Modelled – Mk3.6 vs CMIP3 models Observed CSIRO Mk3.6
Spatial pattern of ENSO SST anomaliesSST anomalies regressed onto 1 stdev NINO 3.4 index Observed Mk3.5 Mk3.5A