Current dynamical downscaling activities at CSIRO John McGregor, Jack Katzfey, Kim Nguyen and Marcus Thatcher Climate Adaptation Flagship Aspendale, Melbourne, Australia UNFCCC NWP Technical Workshop Apia 3 March 2010
CSIRO Climate Adaptation Flagship Under the NWP action pledges we have 4 major project themes: 1. AusAID projects - Mekong Futures; Eastern Indonesia with a focus on rural livelihoods and coastal ecosystems; climate adaptation and urban water systems in Vietnam and Indonesia. 2. Pacific Climate Change Science Program - PCCSP 3. New project in Laos, Cambodia, Bangladesh and India focussing on adaptation in smallholder farming systems 4. Scoping work on rural livelihoods and root crop staples in the Pacific Government, planners and industry are increasingly calling for more accurate climate projections at finer scales of resolution IPCC climate change projections are available globally but there is limited ability to utilise this information on a regional scale as the information provided is too coarse Few Asian or Pacific Island countries have the ability to dynamically downscale this information for their own regional purposes CSIRO dynamical downscaling activities are focussed over Australia, Indonesia and Pacific Island countries
Dynamical downscaling difficulties for tropical regions A major difficulty comes from biases in the host coupled GCMs, especially for SSTs, for example the common coldtongue bias over the equator a) These biases produce errors in the trade winds over the South Pacific and in the location of the South Pacific Convergence Zone (SPCZ) - leads to incorrect downscaled tropical rainfall
Pacific MSL pressure patterns for DJF from some IPCC AR4 models - note deficiencies in their MSLP patterns, and implied trade winds over Fiji (for example) Observed MRI Fiji GFDL 2.1 UK - GEM GFDL 2.0 UK CM3
b) These biases can also lead to very poor downscaled simulations of tropical-cyclone-like vortices. One group has found they must correct for Atlantic SST biases (and winds) in highresolution cyclone studies. c) In the ongoing Regional Model Intercomparison Project (RMIP3) over Asia, all groups are getting unacceptable downscaled rainfall over N China due to GCM biases (despite good rainfall when downscaling from reanalyses, and host rainfall being reasonable). SST bias correction seems necessary. CORDEX domain for east Asia Also being used for RMIP3
CSIRO dynamical downscaling methodology Big demand for downscaling from an ensemble of AR4 GCMS. We usually select those with best ENSO behaviour. Regard the most credible output of coupled GCMs to be their changes in SST patterns and sea-ice. Methodology is to run a quasi-uniform 200 km (or modestly stretched) CCAM atmospheric GCM simulation driven by the monthly-bias-corrected SSTs of a range of AR4 CGMs Only small datasets are needed. Could re-run original host GCM with bias-corrected SSTs, but usually not feasible. The coarse run is then downscaled to finer resolution by running CCAM with a stretched grid, but applying a digital filter every 6 h to preserve large-scale patterns of the coarse run, Quasi-uniform C48 CCAM grid with resolution about 200 km Stretched C48 grid with resolution about 20 km over eastern Australia
Present-day rainfall from a 20 km simulation downscaling Mk 3.5 for 1961-2000 Obs 20 km Produces good present-day rainfall with generally small biases. Also good max/min temperatures. 20 km biases
Rainfall trends 1961-2100 mm/day DJF MAM JJA SON ANN Produces similar Mk 3.5 broad-scale patterns of changes between 200 km and 20 km runs Also gives broadly similar changes to Mk 3.5, but less so in tropics in DJF All runs show drying over Murray Darling Basin in most seasons 200 km CCAM 20 km CCAM
Some CCAM climate simulations performed in 2009/2010 1) Ensemble of 60 km C64 runs over Australia for Climate Futures Tasmania project, (1961-2100) 140 years (14 runs) Mk3.5 A2 B1 plus 2 other A2 runs GFDL 2.1 A2 B1 GFDL 2.0 A2 B1 ECHAM5 A2 B1 HADCM3 A2 B1 MIROC-Med A2 B1 2) Ensemble of 14 km C48 runs over Tasmania (1961-2100) 140 years downscaled from above 60 km CCAM runs The high-resolution CCAM simulations provide the main climate-change information for a large integrated assessment project for Tasmania 3) Simulations over Indonesia (60 km and 14 km) 4) RMIP3 model intercomparison over east Asia downscaling from ECHAM5 5) PCCSP very large global 60 km runs from 1971-2100, then downscaled to around 8 km for individual countries
CCAM regional climate simulations for Indonesia 6 long simulations were driven by 6 different IPCC AR4 coupled GCMs:- from 1971-2000, 20412060, 2081-2100 for the A2 emission scenario Used monthly bias-corrected SSTs from the 6 GCMs First ran 200 km quasi-uniform CCAM simulations Final grid resolution is about 60 km Will downscale over Lombok to 14 km resolution Stretched C48 grid with resolution about 60 km over Indonesia
DJF maximum and minimum temperatures
Simulations of present-day rainfall (mm/day) for DJF Observed and host GCMs In top row CCAM 60 km downscaled runs In bottom rows The 60 km runs produce reasonable rainfall, better than the 200 km runs, but still with room for improvement.
Rainfall changes 2080-2100 from 1971-2000 60 km Indon runs Host GCMs - tendency to become drier over Java - tendency to become wetter over Sumatra - mixed results over Borneo
8 km simulations over Fiji C48 grid Model orography a) Earlier study downscaling NCEP reanalyses for 10 years (Lal et al., Climate Dynamics 2009) acceptable temperature and rainfall climatology and seasonality b) Pilot simulations downscaling from Mk 3 SSTs via 200 km CCAM
Simulated present-day Jan and July for Fiji - pilot run downscaling from Mk3 via 200 km C48 CCAM (from NCEP simulation note colours reversed)
Simulated future Jan and July rain in 2050 - pilot run downscaling from Mk3 via 200 km C48 CCAM Winds become more northerly less rain on east coast Winds become stronger more rain on east (and west) coast
Regional downscaling collaborations Groups now running CCAM for regional climate simulations: Univ. Pretoria and CSIR, South Africa BMKG Indonesia Univ. Hanoi, Vietnam PAGASA, Philippines
Thank you!