MSC Ensemble Prediction System an update

1
MSC Ensemble Prediction System an update

• Richard Verret, Normand Gagnon,
• Stéphane Beauregard, Jacques Hodgson,
• Benoit Archambault
• Canadian Meteorological Center
• Meteorological Service of Canada

2
Outline

• Current and proposed EPS set-up at CMC.
• Basic products.
• NAEFS products.
• Conclusions.

3
Canadian EPS - current set-up
96 perturbed analyses based on Ensemble Kalman
filters
One 16-day control integration with SEF model
Mean
Selection of 16 initial conditions
T149 L28
Eight 16-day integrations with SEF model
Spread inflation
Products
resolution 150 km
Eight 16-day integrations with GEM model
1.2 L28
Integration done twice a day (00 and 12 UTC)
4
Canadian EPS - proposed set-up
96 perturbed analyses based on Ensemble Kalman
filters
One 16-day control integration with GEM model
Mean
Selection of 20 initial conditions
resolution 100 km
Spread inflation
Twenty 16-day integrations with GEM model
Products
0.9 L28
Integration done twice a day (00 and 12 UTC)
5
GZ-500 hPa
6
TT-850 hPa
7
CRPS TT-850 hPa January 2006
CRPS GZ-500 hPa January 2006
8
CRPS UU-500 hPa January 2006
CRPS VV-500 hPa January 2006
9
May 12 2006 00 UTC 120-h Valid May 17 2006 00 UTC
10
Probability of 24-h precipitation amounts
May 7 2006 00 UTC 96- to 120-h Probability of
24-h precipitation amounts Valid 00 UTC May 11
2006 00 UTC May 12 2006
calibrated
11
Bayesian Model Averaging
Prepared at 00 UTC May 19 2006
12
(No Transcript)
13
NAEFS Grand Ensemble (principles)

• Global ensembles
• NOAA, MSC, NMS of Mexico official accord signed
  in November 2004.
• ECMWF and UKMetO, via Thorpex research program.
• Other partners (FNMOC, AFWA, JMA) may join at a
  later time.
• Advantages
• Larger ensemble allowing better PDF definitions
  (super-ensemble).
• Improved probabilistic forecast performance.
• Seamless suite of forecast products across
  international boundaries and across different
  time ranges (1-14 days).
• Minimal additional costs levering computational
  resources.
• Synergy with NCEP on RD work.
• Collaborative product development.
• Contingency with another national NWP Centre.
• Problems
• Combination of multi-model ensembles into a
  super-ensemble.
• Real time exchange (operational considerations).

14
NAEFS Grand Ensemble (deliverables)

• Raw data exchange (00 and 12 UTC runs).
• 00 and 12 UTC production runs.
• 50 selected variables.
• 6-hourly output frequency.
• GRIB2 format.
• Basic products
• Using same algorithms/codes.
• Bias correction algorithm.
• Weighted combination of members.
• Forecast products in terms of climatological
  anomalies.
• Week 2 (days 8 to 14) forecasts based on the
  combined MSC/NCEP ensembles.
• Center specific end products.
• Evaluation and feedback for improvements
• Verification using same approaches

15
NAEFS set-up

• CMC
• 16 members 1 control
• Multi-model
• 8 GEM 1.2 L28
• 8 SEF T149 L27
• Perturbed Kalman filter data assimilation cycles.
  
• Integration done two times a day (00 and 12 UTC)
  out to 16 days.
• To come
• 20 GEM members
• 0.9 L28

• NCEP
• 14 members 1 control
• Single model
• GFS T126 L28
• Ensemble Transform breeding method.
• Integration done four times a day (00, 06, 12, 18
  UTC) out to 16 days.
• To come
• 20 GFS members
• T190 L28

16
EPS-grams
CMC
NCEP
NAEFS
17
Mean and standard deviation
CMC
NCEP

• Available for
• 24-h precipitation amounts
• 10 m wind speed
• 2 m temperature
• Mean sea level pressure
• 500 hPa geopotential heights
• 1000-500 hPa thickness
• 200 hPa wind speed

May 10 2006 00 UTC 24- to 48-h 24-h precipitation
amounts Valid May 12 2006 00 UTC
NAEFS
18
Probability of exceedance over 24-h periods at
least once during the specified forecast range
CMC

• Available for
• Precipitation amounts
• lt 0.2 mm
• gt 2 mm
• gt 5 mm
• gt 10 mm
• gt 25 mm
• 10 m wind speed
• gt 15 km/h
• gt 50 km/h
• gt 90 km/h
• Minimum temperature
• lt 0C
• lt -15C
• lt -30C
• Maximum temperature
• gt 0C
• gt 15C
• gt 30C

NCEP
non calibrated
May 18 2006 00 UTC 0- to 120-h Probability 24-h
precipitation amounts exceeding 10 mm at least
once Valid 00 UTC May 18 2006 00 UTC May 23 2006
NAEFS
19
Probability of total precipitation amounts during
the specified forecast period
CMC
NCEP

• Available for
• Precipitation amounts
• lt 0.2 mm
• gt 2 mm
• gt 5 mm
• gt 10 mm
• gt 25 mm
• gt 50 mm

non calibrated
May 10 2006 00 UTC 0- to 336-h Probability
precipitation amounts exceeding 50 mm Valid 00
UTC May 10 2006 00 UTC May 24 2006
NAEFS
20
Bias correction algorithm
May 14 2006 00 UTC 120-h 500 hPa GZ CMC
ensemble average (contour) / Average bias
correction (color) Valid 00 UTC May 19 2006
For multi-model ensemble, the bias correction is
applied on each member separately
21
Week 2 product

• Methodology
• Get temperatures from members debiased toward
  CDAS reanalyses.
• Average four instantaneous temperatures (00, 06,
  12 and 18 UTC) each day for each member.
• Average daily temperatures over seven days (day 8
  to 14) for each member.
• Get percentile from reanalysis weekly climatology
  for each member.
• Count members in each tercile to get
  probabilities of  above  and  below .
•  normal  1 ( above   below ).
• Calibration of forecasts.

22
(No Transcript)
23
End products
NAEFS products
(link to products)
24
(No Transcript)
25
(No Transcript)
26
(No Transcript)
27
(No Transcript)
28
End products
Products Quebec LAb
29
(No Transcript)
30
(No Transcript)
31
Where will we be Tomorrow?

• Increased number of members and resolution of
  models
• 20 members,
• 100 km.
• Stochastic physics.
• Extension of public forecasts
• Days 6-7 using EPS outputs.
• Usage of EPS outputs for days 3-7.
• Wider usage of NAEFS Grand Ensemble
• Improved bias correction scheme (1st moment).
• Correction of 2nd moment.
• Member weighting scheme.
• Forecast products in terms of climatological
  anomalies.
• More end products.

32
Where will we be after Tomorrow ?

• Regional EPS
• 20 members GEM-LAM.
• 25 km resolution.
• Model perturbations (CAPE).
• Initial condition perturbations.
• Products for decision making processes
• EPS provides the best source of probabilistic
  information for decision making.
• Reliability of probabilistic forecasts is
  required.
• Extreme forecast index.
• Products statistical post-processing
• Verification
• Calibration

33
(No Transcript)