Post on 26-Jan-2016
description
Warm Season Precipitation Predictions over North America with the Eta Regional Climate Model
Model Sensitivity to Initial Land States and Choice of Domain Size
The 30th Climate Diagnostic Prediction Workshop (Oct.24 - 28, 2005) State College, PA
Rongqian Yang and Kenneth Mitchell
Environmental Modeling Center
National Centers for Environmental Prediction
This development is sponsored by the GAPP program of NOAA/OAR/OGP
Presentation Outline
Brief Introduction of the Eta RCM
Model Predictions using CFS Hcst data
• Initial starting time (10 members from CFS, Mid April to Early May
out to 6 months)
• Land States from Reanalysis II and Regional Reanalysis
• Big vs Small Domain
2004 results (NAME period, Precip, 500h, T2m etc.)
Preliminary Conclusions
Future Work
The Eta Regional Climate Model Development at NCEP
• Developed: uses very recent version of Eta Model physics
• As implemented in NCEP Regional Reanalysis
• As implemented in operational NCEP Eta on 24 Jul 01
• Virtually exact match to Eta model in Regional Reanalysis (R/R)
• R/R domain and grid (32-km, 45-levels, large R/R domain)
• R/R Eta model physics, e.g. Noah Land Model 2.3, with 4 soil layers
• Daily updates of several surface boundary fields
• Daily CFS predicted SST (or observed 1-deg weekly Reynolds/Stokes SST )
• Satellite NDVI-based 0.15-degree monthly greenness (NESDIS)
• Seasonal 1.0-deg snow-free albedo climatology (NASA)
• Initial land states of soil moisture and soil temperature
• Soil moist/temp from :
• Regional Reanslysis
• Global Reanalysis II
• Snow depth: USAF operational 47-km daily global snow depth
The Eta Regional Climate Model Development at NCEP (Cont’d)
• Tests of different initial land states (GR2 or R/R)• Test with both big and small domain size
• Predicted Lateral Boundary Conditions (CFS Hindcast)
• Predicted SSTs (CFS Hindcast)
• 10 summer members (mid-April/early May through mid-November)
• Currently 2004 (summer of NAME field campaign)
• 2000, 2001, 2002, and 2003 under way (to achieve 5yr model climo)
• Ultimate goal is to execute 10-20 yrs
Model Prediction (Experiments)
The Big Domain
The Small Domain
Focus Issues
• Model's sensitivity to initial land states
• Choice of domain size
• Full prediction mode (i.e. using CFS predicted LBC and SSTs)
• To see if the regional model shows any skill in seasonal prediction mode, if any, where (in ensemble sense)?
• Focusing on precipitation over the CONUS domain and whether the skills are better than CFS
10 Member Mean June Precipitation
RCMRCM/R2CFS Hcst
CPC AnalysisRCM/RR
All missed this heavy rainJust a HintBig Domain
CFS Hcst
CPC Analysis
RCM/R2
RCM/RR
10 Member Mean June Precipitation
All missed this heavy rain with small domain tooNothing HereSmall Domain
NAM
NAM
10 Member Mean July Precipitation
RCM/R2
RCM/RR CPC Analysis
CFS Hcst
NAM: North American Monsoon
Big Domain
NAM is weaker compared to Big Domain
CFS Hcst
CPC Analysis
RCM/R2
RCM/RR
10 Member Mean July Precipitation
Small Domain
Sustained NAM
10 Member Mean August Precipitation
RCM/R2
RCM/RR
CFS Hcst
CPC Analysis
Big Domain
CFS Hcst
CPC Analysis
RCM/R2
RCM/RR
10 Member Mean August Precipitation
Small Domain
Dry in Southern Great Plains
RCM/RR
RCM/R2CFS Hcst
CPC Analysis
10 Member Mean JJA Precipitation
Big Domain
Less dry in Southern Great Plains than the results using big domain
CFS Hcst
CPC Analysis
RCM/R2
RCM/RR
10 Member Mean JJA Precipitation
Small Domain
Eta-RCM and CFS Temperature is somewhat warmer in central US than the RR verification
RCM/R2CFS Hcst
RRRCM/RR
10 Member Mean June 2m Temperature
Orography signatures are much better revealed in RCM than in Global CFS Big Domain
Eta-RCM temperature is somewhat warmer than RR verification
CFS Hcst
RR Analysis
RCM/R2
RCM/RR
10 Member Mean June 2m Temperature
Small Domain
500mb GPH is low compared to RR, RCM/R2 Better
RCM/R2 CFS Hcst
RR Analysis
RCM/RR
10 Member Mean June 500mb GPH
GPH: Geo-potential Height
Big Domain
500mb GPH is high compared to RR verification
RCM/R2 CFS Hcst
RR Analysis
RCM/RR
10 Member Mean June 500mb GPH
Small Domain
Diffs in Initial Land States from One Member (April, 23)
Comparison of the two Initial Land States
Soil T Temp
Total Soil Moisture
0-100cm percentage of Soil Saturation
0-10 cm Soil Moisture
GR2 cooler
GR2 drier
(Difference Fields: EtaRCM with GR2 Land States minus EtaRCM with RR Land States)
Comparison of May Predicted Fields
Latent Heat 500mb
GPH
200mb GPH Precip
Comparison of Area-avgd Precipitation Timeseries (over CONUS)
Area Avgd Precip is low at the beginning of integration with big domain
Big Domain Small Domain
Green: GR2 land states Yellow: R/R land states
Other Comparisons between the two domains using GR2 and RR land states
Ensemble Mean Kinetic Energy
Big with R2 LS
CFS Hindcast
Big with R/R LS
Small with R2 LS
Small with R/R LS
Regional Reanalysis
Model Losing Kinetic Energy in May/June ?
Ensemble Mean Ratio of ACPCP/APCP
R/R
Small with R/R LS
Small with R2 LS
Big with R/R LS
CFS Hcst
Big with R2 LS
Small domain has the highest percentage of convective Precip
Preliminary Conclusions• Domain size choice is crucial to model results.
• Diffs caused by different land landstates is secondary.
• Big Domain EtaRCM yields better results than other combinations in general.
• The EtaRCM shows skills in warm season precipitation predictions (Compared to Obs and CFS hindcast, especially with features associated with NAM system), still problems?
LBC errors plus physics?
Future Work
• More tests on different years (under way, 2000-2003). To drive the Eta-RCM, CFS hindcasts need to be re-run (saved sigma files). It takes a lot CPU time (so only 2004 finished so far).
• Establishing model climatology to evaluate a relative dry/wet year with respect to model climatology.
• Further testing Land states from Regional Reanalysis and its impact on warm season precipitation.
Thanks to Suranjana Saha Wanqiu Wang Cathy Thiaw Jun Wang Kingtse Mo