May 2006 Upgrade of the NCEP Global Ensemble Forecast System (NAEFS)

1
May 2006 Upgrade of theNCEP Global Ensemble
Forecast System(NAEFS)

• Yuejian Zhu,
• Zoltan Toth, Richard Wobus, Mozheng Wei and Bo
  Cui
• Environmental Modeling Center
• NOAA/NWS/NCEP
• Acknowledgements
• DingChen Hou and Ken Campana EMC
• David Michaud, Brent Gorden and Luke Lin NCO

2
Planned Changes - Summary

• Increasing the number of perturbation runs
• 14 (20 later) perturbation runs for each cycle
• Adding control runs for 0600, 1200 and 1800
  cycles.
• Use Ensemble Transform (ET) breeding method
  instead of breeding method
• ET breeding method will create initial orthogonal
  vectors instead of independent vectors
• NAEFS new products from NCEP-CMC joint ensemble
• Bias corrected forecast
• Forecast anomalies
• Weights

3
GEFS configurations
Current Plan
Model GFS GFS
Initial uncertainty BV ETBV
Model uncertainty None None
Tropical storm Relocation same
Daily frequency 00,06,12 and 18UTC same
Hi-re control (GFS) T382L64 (d0-d7.5) T190L64 (d7.5-d16) same
Low-re control (ensemble control) T126L28 (d0-d16) 00UTC only T126L28 (d0-d16) 00,06,12 and 18UTC
Perturbed members 10 for each cycle 14 (20) for each cycle
Forecast length 16 days (384 hours) same
Implementation August 17th 2005 May 30th 2006
4
Planned Changes - 1

• Increasing the number of perturbation runs
• This change is intended to improve ensemble based
  probabilistic forecast over all and to support
  NAEFS (North American Ensemble Forecast System)
  project.
• Results
• Improving probabilistic skills
• Slightly improving ensemble mean skills (seasonal
  dependent)

5
Planned Changes - 2

• Adding control runs for 0600, 1200 and 1800
  cycles.
• This change is intended to enable for relocation
  of perturbed tropical storm, to respond the
  ensemble initial perturbation changes from pairs
  to one side only. After this implementation,
  there will be complete ensemble package with
  control for each cycle. It is useful to compare
  different cycles and leg forecast.

6
Planned Changes - 3

• Use Ensemble Transform (ET) breeding method
  instead of breeding method
• There is no pair anymore after this
  implementation (see next slide for details)
• This change is intended to improve probabilistic
  forecast skills for all lead-time.
• Results
• Slightly reducing ensemble mean track errors
  (retrospective runs) for 12-96 hours
• Improving probabilistic forecast skills

7
Ensemble Transform Bred Vector (Plan)
Bred Vector (Current)
Rescaling
Rescaling
P1 forecast
P2 forecast
P1
ANL
ANL
N1
P3 forecast
P4 forecast
tt1
tt0
tt2
tt0
tt2
tt1
P, N are the pairs of positive and negative P1
and P2 are independent vectors Simple scaling
down (no direction change)
P1, P2, P3, P4 are orthogonal vectors No pairs
any more To centralize all perturbed vectors (sum
of all vectors are equal to zero) Scaling down by
applying mask, The direction of vectors will be
tuned by ET.
P2
ANL
N2
8
Planned Changed 4 NAEFS Post Products

• NAEFS basic product list
• Bias corrected members of joint MSC-NCEP ensemble
• 35 of NAEFS variables
• 32(00Z), 15(06Z), 32(12Z) and 15(18Z) joint
  ensemble members
• Bias correction against each centers own
  operational analysis
• Weights for each member for creating joint
  ensemble (equal weights right now)
• Weights dont depend on the variables
• Weights depend on geographical location (low
  precision packing)
• Weights depend on the lead time
• Climate anomaly percentiles for each member
• 19 of NAEFS variables
• 32(00Z), 15(06Z), 32(12Z) and 15(18Z) joint
  ensemble members
• Use NCEP/NCAR 40-year reanalysis
• Considering the difference between current
  analysis and reanalysis
• Non-dimensional unit, allows downscaling of
  scalar variables to any local climatology

9
List of Variables for Bias Correction,
Weightsand Forecast Anomalies for CMC NCEP
Ensemble
10
NCO parallel

• Start from 01/15/2006
• 14 perturbed runs and control for each cycle
• With new file structures
• Start from 02/01/2006
• Adding ET scheme (03/07/2006)
• Tuning TS relocation (?)
• Start from 04/01/2006
• Use GEFS new system (May implementation)
• Create bias correction forecast
• Forecast anomalies
• Evaluations

11
Statistic results

• NCO real-time parallel verification statistics
  are posted at
• http//wwwt.emc.ncep.noaa.gov/gmb/yzhu/html/opr/pr
  x_daily.html (available now)
• Updated every morning
• Retrospective experimental verification
  statistics are posted at
• http//www.emc.ncep.noaa.gov/gmb/yzhu/html/opr/et1
  4m_daily.html (available now)
• Updated as required

12
Northern Hemisphere
Southern Hemisphere
Early studies for ET Winter of 2002-2003 ROC
scores for 32 cases ENS-o ? control runs ENS-s ?
ET-20 members ENS-x ? ET-10 members
Tropical
13
Summary of Retrospective Runs

• Period 08/20/2005 09/30/2005
• Statistics for
• Hurricane track errors
• Atlantic-, East Pacific-, West Pacific- basins,
  total basins
• AC scores (Northern Hemisphere)
• RMS errors (Southern Hemisphere)
• Outlier for Northern Hemisphere
• ROC scores for NH and SH
• Conclusion
• Tropical mean of track error (slightly
  improved)
• Improved (48-, 72-, 96-hours over all)
• NH mean (improved), probabilistic (improved)
• SH mean (slightly), probabilistic (improved)

14
Hurricane Track Errors (Atlantic Basin
08/20-09/30/2005)
AEMN-operational ensemble ETIM-retrospective runs
ETIM is better than AEMN
Hours
Cases
174
48
75
101
128
141
157
15
Hurricane Track Errors (East Pacific Basin
08/20-09/30/2005)
Ensemble need to improve in East Pacific Basin in
the future
Hours
Cases
181
69
85
109
135
149
165
16
Hurricane Track Errors (West Pacific Basin
08/20-09/30/2005)
ETIM is better than AEMN
Hours
Cases
177
44
66
101
129
145
161
17
Hurricane Track Errors (All Basins
08/20-09/30/2005)
Overall, ensemble mean beat GFS from/after 72
hours for hurricane tracks, it is similar to
NH/SH 500hPa height rms errors
Hours
Cases
532
161
226
311
392
435
483
18
Improving skills from/after day 3
65 AC scores useful skill Ens. extended
another 20 hours
Much better than GFS after 72 hours
ENS_s operational ensemble ENS_x
retrospective runs
19
Reduced errors after day 6
Reduced initial spread Growing faster than
operational
20
Outlier zero is perfect
Due to reduce initial perturbation
Better after day 4
21
ENS_s operational ensemble ENS_x
retrospective runs
Improved ROC scores for Northern Hemisphere
22
Improved ROC scores for Southern Hemisphere
23
Summary of NCO Parallel

• Period 03/01/2006-current (more than 50 days)
• Statistics for
• RMS errors (Northern Hemisphere)
• AC scores (Southern Hemisphere)
• AC scores for tropical
• ROC scores for NH, SH and tropical
• Conclusion
• Tropical no significant changes
• NH mean (even), probabilistic (improved)
• SH mean (improved), probabilistic (improved)

24
Rms errors are slightly better for short lead time
Less spread at initial, but growing faster
25
Significant improvement for Southern Hemisphere
Ensemble mean beat GFS from/after day 3
26
There is no big effect in Tropical by apply ET in
generally
27
Improving for all lead time
Most considerable improvement for medium range
28
Improvement for all lead time
It is very similar to NH
29
There is no big difference for Tropical region
30
Perturbation versus Error Correlation Analysis
(PECA)
Retrospective runs
NCO parallel runs
31
Summary of NAEFS new products- Post process

• Early studies
• NCO parallel 04/01/2006-current
• No new stats yet

32
Ensemble size 10 members
33
(No Transcript)
34
(No Transcript)
35
(No Transcript)
36
ENSEMBLE 10-, 50- (MEDIAN) 90-PERCENTILE
FORECAST VALUES (BLACK CONTOURS) AND
CORRESPONDING CLIMATE PERCENTILES (SHADES OF
COLOR)
Example of percentile forecast in terms of
climate percentiles
Proposal future NDGD products
37
Background !!!!!
38
RAW BASIC PRODUCT AVAILABILITY
2005, 2006, 2007, 2008
39
END PRODUCTS

• End product generation
• Can be center specific
• Need to conform with procedures/requirements
  established at different centers
• End products generated at NCEP
• Based on prioritized list of requests from NCEP
  Service Centers
• Graphical products (including Caribbean, South
  American, and AMMA areas)
• NCEP official web site (gif NA, Caribbean, SA,
  AMMA)
• NCEP Service Centers (NAWIPS metafile)
• Gridded products
• NAWIPS grids
• NCEP Service Centers (list of 661 products)
• GRIB2 format
• Products of general interest (Possible ftp
  distribution, no decision yet on products)
• NDGD (10-50-90 percentile forecast value
  associated climate percentile)
• End products generated at MSC
• TBD
• End products generated jointly
• Experimental probabilistic Week-2 forecast
• Fully automated, based on basic products bias
  corrected, weighted climate anomalies

40
ENSEMBLE PRODUCTS - FUNCTIONALITIES
List of centrally/locally/interactively generated
products required by NCEP Service Centers for
each functionality are provided in attached
tables (eg., MSLP, Z,T,U,V,RH, etc, at
925,850,700,500, 400, 300, 250, 100, etc hPa)
FUNCTIONALITY CENTRALLY GENERATED LOCALLY GENERATED INTERACTIVE ACCESS
1 Mean of selected members Done
2 Spread of selected members Done
3 Median of selected values Sept. 2005
4 Lowest value in selected members Sept. 2005
5 Highest value in selected members Sept. 2005
6 Range between lowest and highest values Sept. 2005
7 Univariate exceedance probabilities for a selectable threshold value FY06?
8 Multivariate (up to 5) exceedance probabilities for a selectable threshold value FY06?
9 Forecast value associated with selected univariate percentile value Sept. 2005 - FY06?
10 Tracking center of maxima or minima in a gridded field (eg low pressure centers) Sept. 2005, Data flow FY06?
11 Objective grouping of members FY08?
12 Plot Frequency / Fitted probability density function at selected location/time (lower priority) FY07?
13 Plot Frequency / Fitted probability density as a function of forecast lead time, at selected location (lower priority) FY07?
Potentially useful functionalities that need
further development - Mean/Spread/Median/Ranges
for amplitude of specific features -
Mean/Spread/Median/Ranges for phase of specific
features
Additional basic GUI functionalities - Ability
to manually select/identify members - Ability to
weight selected members Sept. 2005
41
ENSEMBLE PRODUCT REQUEST LIST NCEP SERVICE
CENTERS, OTHER PROJECTS
42
NDGD FORECAST UNCERTAINTY - DOWNSCALING

• Ensemble uncertainty information
• Sent on NDGD grid for convenience (if no big
  overhead)
• Valid on model grids (32km for regional, 110 km
  for global ensemble)
• How to bridge gap between model and NDGD grids?
• Anomaly uncertainty information proposed
  methodology
• Establish reanalysis climatology
• In progress for global (NAEFS), methods can be
  transferred to regional reanalysis
• Bias correct ensemble forecasts (wrt operational
  analysis)
• Take 10-50-90 percentile values from bias
  corrected ensemble
• (For establishing anomaly forecasts, adjust
  10-50-90 percentile values to look like
  re-analysis)
• Check climatological percentile corresponding to
  10-50-90 forecast percentiles
• Provide climatological percentiles corresponding
  to 10-50-90 percentile forecast values as second
  set of guidance products

43
Hurricane Track Errors (AtlanticEast Pacific
Basins 08/20-09/30/2005)
Hours
Cases
355
117
160
210
263
290
322
44
Track errors and spreads2004 Atlantic Basin
(8/23-10/1)
From Timothy Marchok (GFDL)
Reduced mean track errors and spreads
45
Hurricane track errors2 basins (Atlantic and
e-Pacific)
Percentage improvement to operational ensemble
Track errors (miles)
Period 20040824-20040930 (53-103 cases)