An overview of the NCEP Ensemble Prediction Systems EPSs

1
An overview of the NCEP Ensemble Prediction
Systems (EPSs)

• Richard H. Grumm
• National Weather Service Office
• State College, PA 16803

2
OVERVIEW

• Ensemble Prediction Systems (EPS)
• Focus on the Primary NCEP Systems
• Short-range ensemble forecast system (SREF)
• Global Ensemble Forecast System (GEFS)
• North American Ensemble Forecast System (NAEFS)
• Includes CMC-Ensemble Forecast system
• Super ensemble? all of these plus the higher
  resolution deterministic runs from the NAM and
  GFS
• Configurations and data.

3
Ensemble Nomenclature

• Ensemble Prediction System ?EPS
• Ensemble Forecast System ?EFS
• Familiar systems
• NCEP
• short range ensemble forecast system (SREF)
• Global Ensemble forecast system (GEFS)
• Other systems and combined systems
• Canadian Ensemble forecast system (CEFS)
• North American Ensemble Forecast System (NAEFS)?
  joint effort.
• European Center for Medium-range weather
  forecasts (ECMWF).
• KMA, CMA, and JMA run systems as does Brazil.

4
The SREF

• Configuration and set-up
• Multi-model EPS (strength)
• Varied resolution (32-45 km) of members
  (weakness)
• Older versions of some operational models
  (weakness)
• Future plans
• Upgrade WRF members
• Go to higher resolution
• More post processed data sets
• Applications
• Strengths and weaknesses

5
NCEP Short-range ensemble forecast system (SREF)
6
SREF Future

• Future plans
• Upgrade to the WRF members
• Higher resolution (need value)
• More post processed data sets
• Mean/spread and probabilities availabel
• Bias corrected data is currently made for each
  member? the low hanging fruit as we can gain
  information here
• Downscaled data sets.

7
SREF Strengths

• Mesoscale resolution EPS
• Timeliness
• Available 4 times daily
• At off times so can be ensembled with the older
  and then newer high resolution deterministic
  runs.
• 3 hourly output
• Good set of parameters available
• Severe weather,Winter weather, Heavy rainfall
• Aviation ? NCEP and those with good bandwidth can
  make products.
• Multi-model ensemble

8
SREF Weaknesses

• Not as fine resolution as the operational NAM
• There is some value to the high resolution and
  more accurate model, especially at shorter ranges
• Resolution of the model is an issue
• Lags NAM and not much better than GFS
• Will improve but really need 15-20 km mesoscale
  EPS
• Perturbations

9
The GEFS

• Configuration and set-up
• Complex as members are low resolution compared to
  the control
• And then, control resolution changes!
• Future plans
• Model improvements and resolution changes
• Applications
• Strengths and weaknesses

10
GEFS ConfigurationSee COMET Matrix
11
GEFS Strengths

• Global EPS
• Timeliness
• Available 4 times daily
• 6 hourly output
• Good set of parameters available
• Good for large scale weather and some higher
  end/larger mesoscale patterns and events.
• Good set post processed data
• Bias correction
• Anomalies
• GFE downscale winds and 2m temperatures.
• RMOP data and images (Relative Measures of
  Predictability)

12
GEFS Weaknesses

• Single Model Core
• WFOs
• Need more short-term forecast data these data are
  longer in arriving and 6-hour increments
• Coarse data for mesoscale features.

13
GEFS Future

• Future plans
• Upgrades GFS implemented into the GEFS over time
• Slow move toward higher resolution
• More post processed data sets
• Mean/spread and probabilities availabel
• Bias corrected data is currently made for each
  member? the low hanging fruit as we can gain
  information here
• Downscaled data sets? several for use in GFE
  already.

14
CMC EPShttp//www.weatheroffice.gc.ca/ensemble/in
dex_e.html

• Meteorological Service of Canada
• Twice a day
• 20 perturbed members 1 control with the GEM
  model
• 20 models have varied physics (radiation and
  convection), varied diffusion, data assimilation,
  and perturbed ICs
• GEM core
• A global model 400x200 points (0.9 degree
  resolution about 100.19km at 0, 76.5 40N).
• 28 vertical levels.

15
CMC-EPS Configuration
16
CMC EFS Strengths

• Multi-model core
• GEM is main the core but
• Varied physics provides model diversity to the
  system
• Convection large scale and shallow
• Diffusion
• And other phsyical parameterizations varied in
  the model
• Diverse ICs
• CMC has mix of different ICs for each model
  core.
• Twice daily 42 member EPS
• 0000 and 1200 UTC
• 20 members plus 1 control

17
The NAEFS and the CMC-EFS

• Configuration and set-up
• NCEP GEFS and CMC-EPS
• Future plans
• Applications
• Strengths and weaknesses

18
NAEFS Strengths

• Multi-model cores
• NCEP has 1 core
• CMC-EFS has diverse set of GEM cores
• Diverse ICs
• NCEP has 20 members all with unique ICs
• CMC has mix of different ICs
• Twice daily 42 member EPS
• 0000 and 1200 UTC
• Has nearly ideal number of members (50) and can
  make very useful probability diagrams.
• Diverse sets of post-processed products produced
• Bias corrected and downscaled data of 2m temps
  and winds
• Two websites to access data CMC and NCEP

19
NAEFS Weakness

• Timeliness
• Takes time to move and process the data
• Only 2X daily at 0000 and 1200 UTC

20
Post Processed Data

• Bias correction?
• low hanging fruit
• Better to use for forecast products (GFE) than
  unfiltered model data
• Downscaled data?
• NAEFS has this with 42 member 0000 and 1200 UTC
• GEFS has this 21 members 4X daily
• Downscaled to GFE grids

21
Critical role of EPSs NOAA/NWS Science Infusion
Bulletin

• Ensembles add the most value in cases and times
  of high uncertainty,
• Never try to chose a model or a single EPS
  member, when one of comparable skill is
  available, in highly uncertainty events,
• In quiescent weather or areas of high confidence,
  the ensemble has minimal value.
• Science Infusion document on uncertainty.

22
All EPS and Combinations

• Facilitate
• Producing Probabilistic output of critical fields
• Viewing spread and uncertainty
• Use to quantify uncertainty and aid in
  reliability issues
• RMOP
• Expected verse current forecast spread.

23
REVIEW

• We focused on the Primary NCEP Systems
• Short-range ensemble forecast system (SREF)
• Global Ensemble Forecast System (GEFS)
• North American Ensemble Forecast System (NAEFS)
• Super ensemble? all of these plus the higher
  resolution deterministic runs from the NAM and
  GFS
• Configurations and data.
• We looked at how these systems will all improve
  over time.

24
Some take away points

• NCEP has
• two local EPS in the SREF and GEFS.
• leverages the CMC EFS to make the NAEFS 2X daily.
• The deterministic runs are also viable ensemble
  members.
• Post processed data
• To aid in the forecast process ? GFE
• Remove bias and systematic errors.