1. Development and Testing of NCEP's Coupled Climate Forecast System
Stephen Lord (EMC/NCEP)
Presented by Huug van den Dool (CPC/NCEP)
S. Saha, S. Nadiga, C. Thiaw, J. Wang, W. Wang, Q. Zhang

2. Overview Introduction Development of CFS Prediction of extreme events
Simulation
Reforecast results
Prediction of extreme events
Possible projects

3. NCEP Seasonal Forecast System Prior to August 2004
Developed
Atmospheric Seasonal Forecast Model (SFM)
First NOAA operational seasonal forecast model
Ocean model and data assimilation
Equatorial Pacific domain
Provides 4 initial weekly ocean states
TOGA/TAO, XBT, ship, altimeter data

4. NCEP Seasonal Forecast System Prior to August 2004 (Cont)
Coupled model provides [ensemble] SST forecasts
1995 NCEP atmospheric model
MOM V.1 Pacific Ocean model
Anomaly flux coupling
SFM ensemble runs from SST ensemble realizations

5. NCEP’s NEW CFS Components for S/I Climate
T62/64-layer version of the current NCEP atmospheric GFS (Global Forecast System) model
Model top 0.2 mb
Simplified Arakawa-Schubert convection (Pan)
Non-local PBL (Pan & Hong)
SW radiation (Chou, modifications by Y. Hou)
Prognostic cloud water (Moorthi, Hou & Zhao)
LW radiation (GFDL, AER in operational wx model)
GFDL Modular Ocean Model, version 3 (MOM-3)
40 levels
1 degree resolution, 1/3 degree on equator
Global Ocean Data Assimilation (GODAS)
Fully coupled atmosphere-ocean (no flux correction)

6. NCEP Global Ocean Data Assimilation System (GODAS)
Real time global ocean data base
ARGO (1000 reports/month), altimeter, XBTs, buoys, SST
Community access to ocean data
Standardized formats with embedded QC meta data
Global ocean data assimilation system
Upgraded ocean data analysis
Reanalysis (ODASI)
Salinity analysis (improved use of altimeter observations)
Implemented September 2003

7. CFS Simulation Study 38 year ‘free’ run
Fully coupled system (NO FLUX CORRECTION)
64 level atmospheric model
Sensitivity result using 28 level atmospheric model
Initial conditions
GODAS
1 January 2002
Verification
Observed Fields : NCEP/NCAR Reanalysis/CDAS

8. 64 Level (0.2 hPa) vs 28 Level (2.0 hPa) Atm.
CFS Simulations
64 Level (0.2 hPa) vs 28 Level (2.0 hPa) Atm.
ENSO SST cycles
Nino 3.4 SST Anomalies
Observed
28 Level Atm
Coupled
Red: monthly bias
64 Level Atm

9. Coupled Model Simulation
SST Interannual Variability
Observed
28 Level Atm
64 Level Atm

10. Coupled Model
Simulation
38 Year Mean
SST Bias

11. Examples of ENSO events
Simulated El Nino
Simulated La Nina
Real El Nino
Real La Nina

12. Tropical Precipitation Performance
AC=.86
AC=.80
AC=.43

13. Re-Analysis
AMIP
Coupled
28 Level Atm
64 Level Atm

14. CDAS Chi anomalies

15. 64 layer model Chi anomaly with climatological SST

17. 200 mb Velocity Potential Anomaly AMIP runs & Opnl Verif.

18. Easterly waves in the AMIP run

19. Easterly waves in the observations

20. AMIP run: Rotated EOF (Nov-Mar) Z200
NCEP Reanalysis AMIP
NAO
PNA

21. Hindcast Skill Assessment
Atmosphere
15-member ensemble over 25 years from
Monthly mean forecasts for all 12 months
9 month forecasts
Initial states 0000 GMT + or - 2 days from ocean state for each month
Reanalysis-2 archive forces both historical and real time forecasts
Operational system continues updating model climatology
Ocean
NCEP Global Ocean Data Assimilation System (GODAS)
Initial states 0000 GMT for 1st, 11th, 21st of each month
GODAS operational September 2003
Global ocean state 1 week behind real time

22. Hindcast Skill Assessment (cont)
Estimated after doing a bias correction for each year
Uses model climatology based on the other years
Anomaly correlation skill score
Nino 3.4 region SST prediction
Standard atmospheric variables such as temperature, precipitation
Skill maps
Anomaly of model vs its own climatology in coupled mode
Comparisons with CMP14 (former operational system) and CASST (CPC statistical technique)

23. Ensemble Mean
CASST
CMP14
April IC

24. Observed
6 Month Lead
(November)
from April IC
SST anomaly
for
Note Amplitudes

25. CASST
Ensemble Mean
January IC
CMP14

26. Observed
6 Month Lead
(August)
from January IC
SST anomaly
for
Note Amplitudes

27. SST Anomaly Correlation
Hindcast
Monthly Averaged
SST Anomaly Correlation
April IC
June-September
Left: New Coupled System
Right: CMP14

28. SST Anomaly Correlation
Hindcast
Seasonally Averaged
SST Anomaly Correlation
January IC
Left: New Coupled System
Right: CMP14

30. 1st and 2nd modes of REOF for SST

31. U. S. Surface Temperature Hindcast Skill (left) 3 Month Averages
CFS
U. S. Surface Temperature
Hindcast Skill (left)
3 Month Averages
April IC
Comparison with CPC
CCA Method (right)
Note: Coupled System skill
Has different geographical
Distribution than CCA

32. U. S. Surface Temperature Hindcast Skill (left) 3 Month Averages
CFS
U. S. Surface Temperature
Hindcast Skill (left)
3 Month Averages
January IC
Comparison with CPC
CCA Method (right)
Note: Coupled System skill
Has different geographical
Distribution than CCA

33. Note: Coupled System skill
CFS
U. S. Precipitation
Hindcast Skill (left)
3 Month Averages
April IC
Comparison with CPC
CCA Method (right)
Note: Coupled System skill
complementary to CCA

34. Note: Coupled System skill
CFS
U. S. Precipitation
Hindcast Skill (left)
3 Month Averages
January IC
Comparison with CPC
CCA Method (right)
Note: Coupled System skill
complementary to CCA

35. Seasonal Forecast for Tropical Vertical Wind Shear (Chelliah & Saha)

36. Performance of the NCEP CFS Forecasts for Severe Weather Events
Suranjana Saha
Environmental Modeling Center
NCEP/NWS/NOAA/DOC

37. CFS Performance for Extreme Events
What is an extreme?
Large departure from normal, for example in temperature and/or precipitation, we have heat waves, cold spells, droughts, floods, etc.
Given how important the effect of extremes is on society (life, property and the economy), did the CFS predict these events ?

38. CFS Performance for Extreme Events (cont)
We evaluate skill in CFS predictions only on occasions when an extreme occurred in observations.
“Probability of detection”
Using monthly mean data, we define an
extreme = |value| of anomaly of variable >= 2 or 1.5 times local standard deviation.

39. CFS Performance for Extreme Events
Two initial cases
Mississippi flood of 1993
Midwest drought of 1988
Lead times of 1-5 months
Time series of extreme events over the U S Midwest
1 month lead time

40. CFS Performance for Extreme Events (cont)
1993 Flood
1988 Drought

41. EXTREME EVENTS IN TEMPERATURE (Reanalysis-2 used for validation)

42. One Month Lead NW NE SW SE

43. Lead Time vs Season

52. Conclusions (for Temperature)
1. US : Modest skill mainly in late spring
2. Europe : No skill
3. India : Modest skill mainly in winter
4. Africa : Modest skill mainly Northern winter
5. South America: Moderate skill throughout the year.

53. EXTREME EVENTS IN PRECIPITATION (Xie-Arkin Precip used for validation)

54. NE NW SW SE

55. U. S. Forecasts of Extreme Precipitation Events

63. Conclusions (for Precipitation)
US : Modest skill mainly in winter
Europe : No skill at all
India : Modest skill mainly Feb-May
Africa : Modest skill mainly Aug-Jan
South America: Modest skill throughout the year only for lead-1.
(Keep in mind there are complications when precipitation is skewed, or standard deviation is small (like deserts).

64. Conclusions and Possible Collaborations
The NCEP CFS displays realistic behavior for monthly and seasonal forecasts
An accompanying reforecast data set
Is used operationally to define forecast skill
Can be used in research mode for a wealth of climate studies of predictability
Suggestions for future work
Extending predictability work of Saha
Continued evaluation and understanding of more detailed system performance
Tropical, topographic interactions, AO….
Development, assembly and testing of next CFS

65. Conclusions and Possible Collaborations (cont)
Study the impacts of :
vertical resolution. 28, 42 and 64 levels
convection in different vertical resolutions running RAS
pbl impact by running with an older version of the PBL
prognostic cloud scheme versus diagnostic cloud scheme.
Impact of sub-grid scale orography with different mountain variance
Impact of new longwave and shortwave radiation in the CFS
Impact of new ice model for polar regions
Stratus deficiency
Testing Noah 3.0, GLDAS, NLDAS for Climate applications, including application to Drought Monitor
Mitigate ocean model biases and develop advanced ODA techniques and investigate impact of MOM-4
Test sigma-p and sigma-theta hybrid coordinates
Sensitivity experiments for tuning the ocean mixed layer
Investigation of simulated ocean - atmosphere modes of variability at the subseasonal timescale and assessment of their realism at different lead times; improvement of relevant parameterizations.
Estimation of the realism of simulated scale interactions (subseasonal to seasonal time scales)
Modification of physics to enhance PNA, AO, MJO, NAO, AAO indices

66. Conclusions and Possible Collaborations (cont)
5-45 day forecast project
Output from 45 day runs has been saved twice-daily data from the entire CFS hindcast set from1981-present, nearly 24 years.
15 members per month
Selected subset for
Evaluation of subseasonal skill for Regional Climate Model experiments
15 additional members/month (1/day)
6 hourly output
Possible new monthly product
Requires GODAS (data assimilation) run in real time instead of the current 7-day lag
Enhanced ensemble size
Hindcasts will still have to be done to provide calibration

67. Questions and Discussion

68. The latest forecast

69. Ocean Data Assimilation - Impact of Salinity

70. Subsurface Temperature Anomalies
Examples of ENSO events
Subsurface Temperature Anomalies
At the Equator
Observed (GODAS)
Simulated
Time
Depth
Pacific
Ocean
Pacific
Ocean

74. Priority Cloud-radiation interaction Orographic forcing
Mesoscale forecast
Hurricane forecast
Seasonal forecast
Week-2 and monthly forecast

75. The Weather Point of view
We decide on the physics upgrade based on model performances in the synoptic forecasts. We try to get the most realistic forecasts of mid-latitude as well as tropical systems in the 0-14 day time range
We evaluate the forecasts both on the correlations with analysis but also with observations

76. Our current activities
Orography
Separation of grid resolvable part and sub-grid part
Improvement of the sub-grid block effect
Cloud
Testing of Ferrier cloud scheme
Merging of RAS and SAS

77. Other activities PBL Radiation
Song-You Hong’s new YSU pbl scheme
Radiation
AER shortwave scheme
Working with meso group to test GFS physics at high resolutions

78. Low hanging fruits
Shallow convection
Cloud fraction