The Canadian Ensemble Prediction System CEPS Comments from forecasters

1
The Canadian Ensemble Prediction System (CEPS)
Comments from forecasters clients

• Lewis Poulin, Kent Johnston
• Meteorological Service of Canada
• NCEP ensemble workshop
• 13-15 may, 2008

2
Outline

• 1 - Operational products currently available
• 2 Operational use of the CEPS Various
  comments from users
• 3 Notes from the Client Services desk
• 4 Development of an automatic EPS public
  forecast generator
• 5 EPS products and renewable energy Ideas for
  the future?
• 6 - Conclusion

3
1 - CEPS products

• Available on publicly accessible web server
• 10-day mean temperature anomaly
• spaghetti plots
• Calibrated probability of equivalent
  precipitation
• Accumulated quantity of precipitation
• Sea level pressure centres
• GZ 500 maps
• Ensemble spread of trial fields
• NAEFS Products (North American Ensemble Forecast
  System)
• Basic products.
• Mean and standard deviation.
• Spaghetti charts.
• EPS-grams.
• Probabilistic products.
• Probability of exceedance
• Bayesian Model Averaging (BMA)

4
2 Operational use of CEPSGeneral comments

• Late product availability limits the use of CEPS
• (about 7 hours after 00 and 12UTC)
• Use varies from center to center and within
  centers
• Users still view EPS for mainly long term
  forecasts
• Users still thinking in deterministic mode
• To increase use of EPS data
• Workshop developed to train operational
  meteorologists
• Working on automated EPS forecast
• Earlier availability at next implementation 45
  min earlier.

5
Main message in training workshop
Uncertainty is a fundamental characteristic of
weather forecasts and no forecast is complete
without a description of its uncertainty. Worksh
op was given in all regional offices Note Users
can access the content (french and english) of
the workshop at these addresses http//collaborat
ion.cmc.ec.gc.ca/cmc/ensemble/Formation-Training/L
isez-moi.html http//collaboration.cmc.ec.gc.ca/c
mc/ensemble/Formation-Training/Read-me.html
6
2 - Comments from CMCs main forecast desk

• Recent interests Winter precipitation, snow
  melt, flood
• Not all use EPS data when preparing bulletins
• Forecasters look at areas, single point products
  not used
• Those that use EPS in graphical FX look at map
  of averages, 500 mb flow, precipitation averages
• EPS data arrives too late, use 12-hour old EPS
  data
• Eager to use NAEFS clustering, not yet fully
  exploited
• Scenarios often chosen quickly, by eyeballing
  data
• Having CMC (main desk) include EPS info in GFX
  helps regional offices by providing a high level
  outlook

7
2 Comments collected from regional forecasters

• Very busy at desks, EPS not integrated in routine
  tasks
• EPS data not ingested (yet) in automated forecast
  system (SCRIBE)
• Some believe EPS still only for long term, do not
  use EPS information even for certain Day-2
  situations
• A need for more user friendly tools, to highlight
  trends
• Some are eager for probabilistic and pdf
  information

8
2 EPSgrams users would like

• Clickable thumbnails of EPSgrams, zooming-in
  potential
• Like NOAA page at http//www.nco.ncep.noaa.gov/pm
  b/nwprod/analysis/namer/gfs/00/index_pcp_m_loop.sh
  tml
• EPSgrams produced in two portions
• Day 1-5 with high temporal resolution
• Days 6-16 with less temporal resolution
• Improve EPSgram readibility
• Spatial maps they can click on to select points
  of interest
• instead of picking them from a list
• A verification line drawn onto the EPSgrams

9
Suggestion / Example to better highlight
bi-modalitéInstead of packaging info in
quartiles, package clustering of members using
color coded boxes
10
2 - User of EPS Maps (would) like

• Clickable thumbnails with zoom-in potential
• Colours that can be read by all
• Better quicker access between naefs products
• Probability products at 12hr intervals
• Clearly labelled calibrated vs non-calibrated
• Some pefer maps with range of values instead of
  standard deviation only (ex snow 15-25 cm, 80).
  

11
2 - User of EPS maps (would) like(continued)

• Click on averages/standard deviation map to
  access pdf
• Animation of thumbnails, especially with 40
  members
• Make their own maps with own colours, contours
• Products with easy to read and clear geography
• Compare EPS debiased products to non-debiased
  ones
• More info. products on precipitation and type
  of precip

12
3 From Client ServicesHow clients are using EPS

• BC River Forecast Centre ingesting NAEFS data for
  use in forecast hydrographs
• BC Hydro using NAEFS temperature forecasts and
  short range EPS for precipitation forecasts
• Aviation industry has expressed desire to ingest
  probabilistic forecasts into decision support
  systems
• Concordia Solar Lab investigating how to
  integrate EPS probablity information into their
  solar buildings management systems

13
3 From Client Services Training Issues

• Many private sector companies in Canada limited
  by lack of training resources for EPS
• Many meteorologists lack basic statistics
• Very limited instruction on EPS in universities
• Problem even greater in less developed countries

14
3 From Client Services Calibration and
Averaging

• Different success measures for different users
• In some cases raw model output might be better
• BMA technique will tend to miss major regime
  changes (dry to wet, warm to cool)
• these are most critical to many users
• Calibration may have similar shortcomings
• what if sub grid scale processes dominate one
  year but not in another?

15
4 - Automated EPS public forecasts

• All public forecasts in Canada are produced
  through a forecast production system called
  SCRIBE
• Automatically generates written forecasts from
  raw model data
• A basic matrix is produced, then packaged for
  users
• We are developing new SCRIBE matrices using EPS
  output
• Goal Produce 1-16 day forecasts experimentally
• Weather parameters included
• cloud covers,
• precipitation occurrences and types,
• winds and temperatures.

16
5 Exploring new EPS productsR.E. Forecasts?

• DSWRF from the model
• Sample 2-day forecast of the no. of 40-gal hot
  water tanks from a solar water heater
• EPS could be used to produce probability type RE
  forecasts
• Solar water, solar pv, wind, water collection

17
- 5 -http//collaboration.cmc.ec.gc.ca/cmc/CMOI/P
ublications/REFcst_SolarH2O.pdf

• Prototype EPS RE forecast
• Probability of receiving more than 250 (L) of
  solar heated water in the next 5 days
• Similar products can be produced for 1 or 2 day
  forecasts
• To help support R.E. based lifestyles, personal
  energy management

18
6 - Conclusion

• ltlt no forecast is complete without a description
  of its uncertainty. gtgt EPS can answer that need.
• A number of mainly graphical EPS products are
  presently available
• Need to improve product delivery times (coming
  soon)
• Users have many ideas
• Considering user feedback will help improve
  quality, use of products
• EPS still underused in Canada, to correct that
  situation
• A workshop to train operational meteorologists
  was given.
• A prototype automated EPS forecast for public use
  is being developed
• Growing number of external users interested in
  EPS data
• EPS can be used in creative opportunities, ex
  R.E. forecasts

19
Thank You

• Questions?