Global Climate and Weather Modeling

1
Global Climate and Weather Modeling

• John H. Ward
• NCEP Production Suite Review
• December 10, 2008

2
Outline

• GCWMB Responsibilities
• Global Model the Production Suite
• Recent Changes
• Upcoming Changes
• Dropouts
• Future Plans

3
GCWMB Responsibilities

• GCWMB presentation will be broken into four
  separate parts
• Overview of recent changes and test results from
  upcoming implementations for
• Global Forecast System
• Global Data Assimilation
• Ensemble Forecast System
• Long Term planning
• Data Assimilation John Derber
• Climate Forecast System Hau-Lu Pan
• Ensemble Forecast System Zoltan Toth

4
NWS Seamless Suite of ForecastProducts Spanning
Climate and Weather
Forecast Uncertainty
Years
Seasons
Months
Climate Forecast System
2 Week
North American Ensemble Forecast System
Climate/Weather Linkage
1 Week
Global Forecast System
Short-Range Ensemble Forecast
Days
Ocean Model Hurricane Models
North American Forecast
Hours
Rapid Update Cycle for Aviation

• GFDL
• WRF

Dispersion Models for DHS
Minutes
Health
Aviation
Recreation
Ecosystem
Agriculture
Commerce
Hydropower
Environment
Maritime
Fire Weather
Life Property
Energy Planning
Reservoir Control
Emergency Mgmt
Space Operations
5
NOAAs NWS Model Production Suite
Oceans HYCOM WaveWatch III
Climate CFS
Hurricane GFDL HWRF
Coupled
MOM3
1.7B Obs/Day
Satellites 99.9
Dispersion ARL/HYSPLIT
Regional NAM WRF NMM
Global Forecast System
Global Data Assimilation
Severe Weather
WRF NMM/ARW Workstation WRF
Short-Range Ensemble Forecast
North American Ensemble Forecast System
Air Quality
WRF ARW, NMM ETA, RSM
GFS, Canadian Global Model
NAM/CMAQ
Rapid Update for Aviation
NOAH Land Surface Model
6
Recent Changes

• Data Assimilation
• Global GSI Upgrades 1Q08 4Q08
• Ensemble Forecast System
• NAEFS 1Q08
• Climate
• CFS/GODAS 1Q08
• Global Forecast Model
• Physics upgrade 3Q08 ? Withdrawn

7
3Q08 GFS Change

• GSI Bundle (see slide 10)
• RRTM1 long-wave radiation
• More accurate absorption coefficients
• RRTM shortwave radiation
• Major improvement in upper stratospheric cloudy
  skies
• Maximum-random cloud overlap for shortwave
• Unifies cloud treatment for long shortwave
  radiation
• Hourly long-wave radiation
• Better definition of diurnal cycle. Makes scheme
  consistent with short-wave radiation

8
3Q08 GFS Change(cont)

• Generalized aerosol treatment
• Code efficiency. Aerosol algorithms will not
  have to be rewritten when radiation schemes
  change.
• Realistic CO2
• Model had been using old climatology. Now uses
  observed CO2
• Retuned mountain blocking
• Update parameters for T382
• Shallow convection up to sigma 0.7
• Permits shallow convection over high terrain
• Convective gravity wave drag
• Improve stratospheric tropical winds

9
Summary of Subjective Evaluation

• The proposed GFS implementation shows, at best,
  little improvement over the operational GFS, and
  has several shortcomings
• Major concerns
• Increased feedback from grid scale overturning on
  mass fields and downstream QPF
• Less run-to-run consistency
• Loss of 6-12 h of skill in NH 500-mb anomaly
  correlation die off curves for days 5-7
• While the parallel does show signs of improved
  performance at medium ranges, this is offset by
  lack of run-to-run consistency

10
Decision to Cancel Implementation

• Based on subjective review and a lack of an
  overwhelming improvement in objective statistics
  the proposed change was withdrawn
• Because of the complexity in testing and
  evaluating GFS changes it was also decided to
  separate data assimilation changes from future
  GFS bundles
• This will make it possible for more frequent
  changes in the assimilation system
• 4Q08 GSI implementation was a modified version
  from the GFS package. Components retained
• Observations
• assimilate WindSAT surface winds (new dataset)
• add assimilation of profile ozone quality flag 7
  SBUV/2 observations
• reintroduce assimilation of AQUA AMSUA channels 6
  and 8-13
• Code
• ability to distinguish between GPSRO data from
  various sources
• improved computational efficiency
• ability to directly read / use quadratic grid
  surface guess files

11
Upcoming Changes

• Data Assimilation
• Global GSI Upgrades 2Q09
• Ensemble Forecast System
• GEFS 3Q09
• Climate
• Reanalysis Reforecasting ongoing
• Global Forecast Model
• Coupled Ocean-Atm 3Q08 ? Withdrawn
• Uncoupled physics upgrade being tested

12
2Q09 GSI Upgrade

• Addition of METOP/IASI data
• Improved atmospheric analyses, especially in S.H.
• Variational Quality Control
• Improved quality control of observations
• Eliminate need for OIQC step
• Change in land/snow/ice skin temperature variance
• Improved use of near surface radiance
  observations
• New Background error covariances
• Consistency with current version of system.
  Improved upper atmosphere
• Reduce number of AIRS water vapor channels
• Improved convergence and improved analysis

13
2Q09 GSI Upgrade

• New version and coefficients for CRTM
• Improved simulation of radiance observations
• Modification of height assignment for height
  based wind observations
• Improved use of these observations
• Retune observational errors
• Removed results of impact of removal will be
  shown
• Modification of surface land use file
• Removes a few permanent (12) glacial points to
  improve surface temperature forecasts
• Situation Dependent Background Variances

14
As is 500 hPa streamfunction (1e6) background
error standard deviation Valid 2007110600
New flow-dependent adjusted background error
standard deviation
15

• Surface pressure background
• error standard deviation
• fields
• with flow dependent re-scaling
• without re-scaling
• Valid 2007110600

16
2Q09 GSI Schedule

• 30-Day evaluation 16 December
• Evaluation ends 13 January
• Management Brief 22 January
• Implementation 27 January

17
Red New GSI Black Operational
New GSI better than current operational through
day 12 in Northern Hemisphere
18
Red New GSI Black Operational
New GSI generally better than current operational
through day 11 in Southern Hemisphere
19
Red New GSI Black Operational
Threat Score better than current operational
Bias better than current operational
20
GFS vs. Proposed GSI
Storm Tracks Derived from GFS forecast
New GSI produces better storm tracks
21
GFDL vs. GNEW
New GSI produces slightly better GFDL storm
tracks
22
HWRF vs. H047
New GSI produces slightly degraded HWRF storm
tracks
23
Global Ensemble Forecast System
24
3Q09 GEFS Upgrade

• Continue using current operational (n-1) GFS
  model
• Upgrade horizontal resolution from T126 to T190
• 4 cycles per day, 201 members per cycle
• Up to 384 hours (16 days)
• Use 8th order horizontal diffusion for all
  resolutions
• Improved forecast skills and ensemble spread
• Upgrade to ESMF (Earth System Modeling Framework)
  for GEFS Version 3.1.0r
• Allows concurrent generation of all ensemble
  members
• Needed for efficiency of stochastic perturbation
  scheme
• Add stochastic perturbation scheme to account for
  random model errors
• Increased ensemble spread and forecast skill
  (reliability)
• Add new variables (26 more) to pgrba files
• Based on user request
• From current 52 (variables) to future 78
  (variables)
• For NAEFS ensemble data exchange

25
3Q09 GEFS Schedule

• 30-Day evaluation will begin on Power6 system
  when NCO has completed a sufficient portion of
  the transition to free up staff resources
• Package would run in parallel on the P6 and
  become operational when the P6 becomes
  operational in mid-July

26
NAEFS future configuration Updated December 2008
27
Horizontal resolution change Ensemble control
only (deterministic) From T126 to T190 NH 500hPa
geopotential height
26 4
Gains from short waves
28
Horizontal diffusions Ensemble controls only
OPR(T126)-4th order NHD(T126)-8th order
May 2007
November 2007
29
Resolution and Diffusion for Global Ensemble
E20s T126 4th for all 16d (oper.) E20x T190
8th out to 16d E20e T190 8th (0-180h), then
T126 4th
When reducing resolution from T190 (8th order) to
T126 (4th order), the ensemble forecast
probabilistic skill score tends to t126
immediately, the example shows here for tropical
850hPa temperature. 8th order diffusion for t126
somewhat improves performance (not show here).
Therefore, both the resolution and diffusion play
an important role here.
30
Latest retrospective run (full package)
NH 2-m temperature RMSE Spread
NH 500hPa height RMSE Spread
E20s T126L28 E20g T190L28 (0-180 only)
SH 500hPa height CRPSS
NH 500hPa height CRPSS
31
CRPSS for NH 850hPa temperature
OPER
OPER
EXP
EXP
Extend current 5-day skill to 6-day
Extend current 5-day skill to 6.5-day
32
NH Anomaly Correlation for 500hPa HeightPeriod
August 1st September 30th 2007
GEFSg is better than GFS at 48 hours
GEFSg could extend skillful forecast (60) for 9
days 24 hours better than current GEFS 48 hours
better than current GFS
33
Conclusion

• Based on two sets of retrospective runs (summer
  and winter 2007)
• New package improved the forecast skill (score)
  significantly
• For deterministic (ensemble mean)
• For probabilistic (ensemble distribution)
• The better results is benefited from
• Increase horizontal resolution (include
  diffusion)
• Stochastic perturbation scheme
• Better initial condition (analysis)
• Better forecast model (GFS)

34
Experimental Coupled Ocean-Atmosphere Model

• No GSI changes
• RRTM1 long-wave radiation
• More accurate absorption coefficients
• RRTM shortwave radiation
• Major improvement in upper stratospheric cloudy
  skies
• Maximum-random cloud overlap for shortwave
• Unifies cloud treatment for long shortwave
  radiation
• Hourly long-wave radiation
• Better definition of diurnal cycle. Makes scheme
  consistent with short-wave radiation
• Generalized aerosol treatment
• Code efficiency. Aerosol algorithms will not
  have to be rewritten when radiation schemes
  change.

35
Experimental Coupled Ocean-Atmosphere Model (Cont)

• Realistic CO2
• Model had been using old climatology. Now uses
  observed CO2
• Retuned mountain blocking
• Update parameters for T382
• Shallow convection up to sigma 0.7
• Permits shallow convection over high terrain
• Convective gravity wave drag
• Improve stratospheric tropical winds
• Improved low level Stratus withdrawn early
• Enthalpy thermodynamic prognostic variable
• Improved Mesosphere for SWPC
• Coupling with MOM4 Ocean Model
• Improved week 2 forecast

36
Operational GFS Proposed GSI Coupled GFS
Proposed Coupled System shows reduced Threat
Score and an increase in Precipitation Bias
for larger amounts compared to GSI
37
Physics Package without coupling shows only a
slight reduction in Threat Score but increase in
Precipitation Bias remains
38
Operational GFS Proposed GSI Couled GFS
Proposed Coupled System Shows reduction in NH 500
MB Anomaly Corr compared to GSI
39
Physics Package without coupling produces a
larger degradation in NH 500 mb Anomaly
Correlation
40
Removing Enthalpy change (PRC8) shows An
improvement in Threat Score, but still shows
increase in Precipitation Bias for higher amounts
41
Removing Enthalpy change (PRC8) produces an
improvement in 500 mb Anomaly Correlation
42
Removing Enthalpy change produces improvement in
RMS error for upper level tropical winds
43
Removing Enthalpy change produces improvement in
NH RMS temperature error
44
Removing Enthalpy change produces improvement
in NH RMS height error
45
Removing Enthalpy change produces improvement
in tropical wind bias
46
Preliminary Summary

• Objective analysis indicates significant
  improvement in a number of areas, but
  precipitation bias remains an issue.
• Do significant improvements in Height, Threat,
  Hurricane Track outweigh the increase in
  precipitation bias?
• Testing of the Physics Package will continue with
  hopes of 3Q08 implementation

47
Preliminary Summary

• Objective analysis indicates significant
  improvement in a number of areas, but
  precipitation bias remains an issue.
• Do improvements in Height, Threat outweigh the
  increase in precipitation bias?
• Testing of the Physics Package will continue with
  hopes of 3Q08 implementation

48
Future Approach

• Continued development of major model upgrades
• Semi-Lagrangian Model
• Increased Model Resolution
• Identify know model weaknesses and develop
  targeted changes to address those short comings
• Identify poor forecasts from subjective scores
  and focus on diagnostics

49
Poor forecasts or Skill Score Dropouts lower
GFS performance.
20081001-19
Significantly reduced skill in AC or RMS (not
shown) occur a little more than once a
month. Shown above are the GFS 00, 06, 12, and
18 Z cycle, and the ECMWF 00 and 12 Z cycle
anomaly correlation skill scores for 20-80 NH for
the first half of October 2008 .
50
GFS Diagnostics of Dropouts
51
GFS Diagnostics of Dropouts (cont)
52
Tools to compare ECMWF and NCEP Dropouts

• ECM Analysis Use ECMWF analysis to generate
  Pseudo Observations for input to the Gridded
  Statistical Interpolation (GSI) to make analyses,
  for initial conditions, and integrate GFS
  forecasts.
• Compare these ECM runs with GFS runs made in the
  same way and w/ production.
• A Climatology of NH and SH dropouts what are
  the systematic differences
• Interpolate ECM and GSI analyses to observations
  to determine comparative strengths and
  weaknesses.
• using statistics of observation type fits
• stratify by pressure, type, and difference
  magnitude
• Goal Diagnose Quality Control problems,
• Implementation of real time QC
  detection/correction and improvements to analysis
  system algorithms

53
GFS forecasts from ECMWF IC alleviate
production dropouts
New GSI is improvement but does not usually
alleviate dropout
GFS Operational (Production) GFS ECMWF
Their Operations ECM Pseudo Obs from
ECMWF analysis using the GSI as a Grand
Interpolator CNTRL Updated GSI system with
GDAS observations (GSI details change often
across the dropout accumulation
period) InterpGES Updated GSI with previous 3,
6, 9-hour ECM forecasts as background guess
plus GDAS observations
54
Results from SH Dropout Experiments

• 90 ECM success in alleviating SH dropouts.
• Adding GDAS Obs (ecmanlGES) lowers the success
  rate to 75.
• GSI upgrades (CNTRL GFS Operations plus all Obs)
  improves GSI but not enough to alleviate dropouts
• InterpGES (Background Guess from ECM plus all
  Obs) runs also improves GSI but not enough to
  alleviate dropouts
• Composites show low level SH difference near
  850mb in temperature.
• Adding the SH observations and the 3, 6, 9-h
  background guess degrade the ECM forecasts (from
  runs with ECM as the background guess)

55
Summary and Future Work

• ECM analysis show dropouts can be alleviated in
  GFS forecasts.
• Running the operational GSI with an ECM
  background guess offers a systematic approach for
  assessing observation types by removing
  observation types from control analyses
• There are systematic height differences between
  all models and ECMWF perhaps from the satellite
  window coverage of ECMWF (12-h) vs others (6-h)
• Work continues to analyze what is the optimal fit
  of analysis to observation types and to determine
  an implementable algorithm for QC, bias
  correction, and analysis weights

56
Know Model Weaknesses(from internal evaluation)

• Large growth in rms error in first 24 hours in
  midlatitudes
• Large growth in rms error in first 48 hours in
  tropics
• Too much surface downward short wave radiation
• Warm low-level bias over western US in summer
• Negative height bias over Antarctica
• Poor scores in Southern Hemisphere
• Lack of stratus clouds in fcsts displacement of
  stratus clouds from coasts
• Higher precipitation than independent estimates
• Cold bias over US in winter
• Negative moisture
• Relatively low QPF scores
• Failure to maintain rising motion over Indonesia
  in the forecasts
• Failure to capture early stages (TD) of tropical
  systems

57
Know Model Weaknesses(from user feedback)

• High bias for areal coverage of low QPF amounts
  (lt 0.10 in.)
• Grid scale overturning induces spurious mid-level
  vorticity maxima, particularly in the warm
  season.
• Too cold with the melting layer in wintry
  precipitation events
• Too flat with strongly amplifying troughs/closed
  lows in the western U.S. resulting in the model
  being too fast to eject these systems out to the
  east
• Too far east with the track of East Coast/Gulf
  Stream cyclones
• Lack of terrain resolution adversely impacts QPF
  in the western U.S. and in other areas of complex
  terrain, such as southern Mexico.
• GFS forecasts conditions favorable for heavy lake
  effect snow but has low amounts of QPF
• Routinely places a trough over the eastern Brooks
  Range in Alaska
• Too weak and fast with well-developed Pacific
  systems more than a couple of days into the
  forecast period
• Too aggressive bringing energy into the western
  side of ridges
• Cold bias in the longer ranges over the eastern
  U.S./western Atlantic
• Surface lows coming out of the west into the
  Plains are too far north The model tends to focus
  too much QPF in the cold sector of cyclones

58
Schedule for Future Upgrades

• Power6 transition impacting possible FY09
  implementations
• Physics upgrade on Power6 when it becomes
  operational in July 2009
• The go-operational date is during Hurricane
  Season, which would produces implementation of
  both GFS and HWRF during Hurricane Season.
• Next window would be 1Q10

59
Schedule for Future Upgrades

• Earliest possible implementation of major upgrade
  package would be mid to late FY10
• Minor changes to improve efficiency and
  streamline post processing and product generation
  can take place at any time after the Power6 is
  operational
• Changes will not impact forecast product quality

60
Questions