Simulation of climate anomalies on seasonal scales using general circulation model of the atmosphere M.A.Tolstykh 1,2) , D.B. Kiktev 2) , R.B.Zaripov 2) , M. Yu. Zaichenko 2) 1)Institute of Numerical Mathematics, Russian Academy of Sciences 2)Hydrometcentre of Russia Moscow
Seasonal forecast • Forecast of a mean seasonal anomaly of atmospheric circulation with respect to climate. • Usually for 4 months with 1 month lead time • Ensemble technology is commonly used • Computationally expensive => requires efficient atmospheric model
Motivation • WMO requirement for a World Meteorological Center to produce seasonal forecasts • 2003-2004: PCMDI SMIP-2, SMIP/HFP intercomparison projects • 2006-2007: WMO/WCRP : Seamless prediction, TFSP initiative. • TFSP requires the full model of climate system (i.e. atmosphere + vegetation, soil + ocean + ice + …).
SL-AV atmospheric model, seasonal version • Global semi-Lagrangian finite-difference model. • Semi-Lagrangian advection enables large time steps (~4-5 CFL) • Horizontal resolution 1,40625° х 1 lon-lat,125°, 28 vertical levels • Dynamic core of own development (vorticity- divergence formulation on the unstaggered grid; 4 th order finite differences). Validated in Held-Suarez test (3yr integration) • Subgrid-scale parameterizations from French model ARPEGE/IFS. No vegetation in the old version, ISBA scheme in the new version • The model contributes to the multi-model ensemble of APCC. Forecasts are at http://www.meteoinfo.ru/season
Validation issue • The forecast lead time is too long to enable reliable statistics in reasonable time • Two kinds of forecasts are considered: - historical forecasts (hindcasts), e.g. starting from reanalyses - real time forecasts starting from RHMC analyses (size of prognostic ensemble -10, breeding is used to generate this ensemble)
Historical seasonal forecasts with SL-AV model Seasonal models intercomparison project (http://www-pcmdi.llnl.gov) • Experiments conducted for 1979-2003 • Forecasts for four months • Four seasons evaluated (winter, spring, summer, autumn) Potential predictability (SMIP-2) Potential predictability • Size of prognostic ensemble – 6 (using initial data from NCEP/NCAR reanalyses with 12-hours shift) Practical predictability (SMIP-2/HFP) Practical predictability ( • Size of prognostic ensemble– 10 (using initial data from NCEP/NCAR reanalyses-2 with 12-hours shift); • Boundary condition – preserving initial SST anomaly
Potential predictability (old version of the model) SMIP-2 (Kiktev et al, Russian Meteorology and Hydrology, 2006)
T850. ACC. SL-AV model. Months 2-4. Potential predictability. 1979-2002. DJF MAM JJA SON
ROC scores for historical seasonal SL-AV model forecasts SMIP-2 protocol Period: 1979-2003. T850. Months: 2-4 Regions with ROC < 0.55 shown in white. Regions with statistically significant signal are in black ( α =0.05)
T850. ROC scores for 3 categories Normal Period: DJF (Months 2-4) 1979-2002 Potential predictability Below Normal Above Normal SL-AV Model
T850. ROC scores for SL-AV. Region 20N-90N. Months: 2-4. 1979-2002. Season Below Normal Above All normal normal categories Winter 0.624 0.517 0.619 0.588 Spring 0.604 0.507 0.618 0.560 Summer 0.611 0.517 0.613 0.583 Autumn 0.628 0.529 0.628 0.597
T850. ROC scores for SL-AV. Region: 20S-20N. Months: 2-4. 1979-2002 Season Below Normal Above All normal normal categories Winter 0.762 0.625 0.769 0.724 Spring 0.606 0.569 0.686 0.608 Summer 0.712 0.584 0.740 0.683 Autumn 0.701 0.569 0.713 0.665
Potential predictability (old version of the model) SMIP-2/HFP, Independent validation
International Cooperation
Multi- -Institutional Cooperation Institutional Cooperation Multi National Aeronautics and Space Administration USA Meteorological Service of Canada National Centers for Environmental Prediction USA National Climate Center/CMA China International Research Institute Institute of for Climate Prediction Atmospheric Physics USA China Center for Ocean-Land-Atmosphere Studies Central Weather Bureau USA Chinese Taipei Hydrometeorological Centre of Russia Japan Meteorological Agency Main Geophysical Observatory Korea Meteorological Administration Russia Meteorological Research Institute Korea
Zonal mean H500 in SMIP2/HFP
Zonal mean Т 850 in SMIP2/HFP
Zonal mean precipitation
Drawbacks of the old seasonal version • Unrealistic high precipitation in tropics, wrong geographical distribution (lack of precipitation in continental tropics) • T850 too warm over Antarctida, too cold ( by 2 degrees) over tropics • H500 is 30-40 m lower All this was attributed to the absence of modern surface (soil-vegetation-snow) parameterization
New version of seasonal prediction model • In the old version, there was no vegetation; 100 % daily relaxation to climate values of deep temperature Tp and water content Wp • New version – parameterization of interaction between soil, vegetation, snow, soil ice and the atmosphere ISBA (Noilhan, Planton 1989, Giard, Bazile, 2000) + weak (1.e-2 per day) relaxation to climate values of Tp and Wp • Also some changes in PBL parameterization • Necessary requirement for ISBA to work – realistic initial conditions for soil water content of the deep (mean) layer. In seasonal context – also realistic water content climate field, appropriate for ISBA
Importance of proper soil water content field • Was shown on time scales form short- range weather forecast (Giard, Bazile; many others) to climate simulation (Lykossov, Volodin, PhAO, 1998) • The reason is its slow evolution: characteristic time is 3-4 weeks.
Assimilation of soil variables (N.N. Bogoslovskii) • The scheme (Giard, Bazile, 2000) developed especially for ISBA was implemented in medium-range forecast version. This scheme uses increments of T2m and RH2m analyses to produce increments of soil variables using some restrictions. • Due to large biases in T2m and RH2m analyses, development and implementation of own analyses was required • Now we have assimilated soil water content field for more than 1 yr, which can be used as climate
First guess errors for 5-18th June 2007 Black lines: old model, no soil assimilation, RHMC analyses for T2m and RH2m Red lines: model with ISBA + soil variables assimilation + new analyses for Т 2m и RH2m Temperature 2m, 6h forecast ( region lat. 35 n. - 80 n., lon. 0 - 144 e. ) 7 6 5 4 3 2 1 C 0 -1 -2 -3 -4 -5 -6 5 6 7 8 9 10 11 12 13 14 15 16 17 18 -7 day old var var
Validation of the new version of seasonal model • Ensemble forecasts with 10 members for initial data of 28/07/07, 28/10/07, 28/01/08, 28/04/08. All 4 ensembles showed similar improvements. • Study EOFs for H500 and MSLP fields
Zonal mean T850
Zonal mean H500
Zonal mean precipitation
Precipitation for Sep-Nov 2007:
Precipitation for Sep-Nov 2007:
Importance of correct Wp climate field
Some scores of old and new forecast starting from 28/07/07 • H500 • Old : region N20, RMSE: 46.47 • New: region N20, RMSE: 36.91 • Old : Region: Tropics, RMSE: 15.35 • New: Region: Tropics RMSE: 11.14 • T850 • Old: Region N20, RMSE: 2.29 • New: Region N20, RMSE: 1.54 • Old: Region Tropics, RMSE: 2.40 • New: Region Tropics, RMSE: 1.75
EOF analysis based on real data forecasts (new version)
Conlusions • Implementation of soil-vegetation-snow parameterization allowed to significantly reduce systematic biases in precipitatoin, H500 and T850 fields in seasonal forecasts • The use of own-produced deep soil water content field for relaxation helped to further reduce spurious precipitation over deserts • First EOFs of model circulation seem to be in a agreement with observations
Future work • Redo SMIP2/HFP forecasts • Calculate EOFs based on reanalyses hindcasts (huge computational work) • Coupling with the INM ocean general circulation model
Thank you f or at t ent ion ! Supported by RFBR grant 07-05-00893
Recommend
More recommend
