Simulation of climate anomalies on seasonal scales using general - - PowerPoint PPT Presentation

Simulation of climate anomalies

• n 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°х1 lon-lat,125°, 28

vertical levels

• Dynamic core of own development (vorticity-

divergence formulation on the unstaggered grid; 4th

• rder 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 Potential predictability (SMIP-2)

• Size of prognostic ensemble – 6 (using initial data from

NCEP/NCAR reanalyses with 12-hours shift) Practical predictability ( Practical predictability (SMIP-2/HFP)

• 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.

JJA MAM DJF 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

Period: DJF (Months 2-4) 1979-2002 Potential predictability SL-AV Model

Below Normal Normal Above Normal

• T850. ROC scores for SL-AV.

Region 20N-90N. Months: 2-4. 1979-2002.

0.597 0.628 0.529 0.628 Autumn 0.583 0.613 0.517 0.611 Summer 0.560 0.618 0.507 0.604 Spring 0.588 0.619 0.517 0.624 Winter All categories Above normal Normal Below normal Season

• T850. ROC scores for SL-AV.

Region: 20S-20N. Months: 2-4. 1979-2002

All categories Above normal Normal Below normal Season 0.665 0.713 0.569 0.701 Autumn 0.683 0.740 0.584 0.712 Summer 0.608 0.686 0.569 0.606 Spring 0.724 0.769 0.625 0.762 Winter

Potential predictability (old version of the model)

SMIP-2/HFP, Independent validation

International Cooperation

National Climate Center/CMA China Institute of Atmospheric Physics China Central Weather Bureau Chinese Taipei Japan Meteorological Agency Korea Meteorological Administration Meteorological Research Institute Korea Main Geophysical Observatory Russia Hydrometeorological Centre

• f Russia

Center for Ocean-Land-Atmosphere Studies USA International Research Institute for Climate Prediction USA National Centers for Environmental Prediction USA National Aeronautics and Space Administration USA Meteorological Service of Canada

Multi Multi-

• Institutional Cooperation

Institutional Cooperation

Zonal mean H500 in SMIP2/HFP

Zonal mean Т850 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

• wn 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 Т2m и RH2m

Temperature 2m, 6h forecast ( region lat. 35 n. - 80 n., lon. 0 - 144 e. )

• 7
• 6
• 5
• 4
• 3
• 2
• 1

1 2 3 4 5 6 7 5 6 7 8 9 10 11 12 13 14 15 16 17 18 day

C

• ld

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