Review of Recent Results, Ongoing Development

1
Review of Recent Results, Ongoing
Development TestingPlans For implementation

• Bo Cui1, Zoltan Toth2, Yuejian Zhu2, Richard
  Verret3
• 1SAIC at Environmental Modeling Center, NCEP/NWS
• 2Environmental Modeling Center, NCEP/NWS
• 3Canadian Meteorological Centre, Meteorological
  Service of Canada
• Acknowledgements
• David Unger, Stéphane Beauregard, Dingchen Hou,
  Richard Wobus

2
Review of Recent Results

• First moment correction
• Previous results kept reinitializing the prior,
  based on 40-day flat average difference
• Current system keeps cycling the bias estimate
  after initializing the prior, which starts from
  July 1, 2003. Choose decaying weight 10, 5, 2,
  1, 0.5 and 0.25, respectively, and apply on
  500 mb height of NCEP CMC ensemble
• Northern and Southern Hemisphere the smaller
  weigh is better for longer lead time, and larger
  weight is better for shorter lead time
• Tropical region 2 is the best one among the
  six weight factors
• Bias correct CMC member individually bias
  correct CMC member in
• 2 groups ( 8 SEF member 8 GEM member) due
  to CMC multi-
• model ensemble and each model member has
  its own physics parameterization
• applying the bias correction scheme on each
  member is the better approach though the
  differences are small between the two methods
• Combined ensemble, use equal weight for all
  members ( 5 NCEP 5
• CMC member)

3
Ongoing Development Testing Plans for
Implementation

• Bias correction
• First moment correction
• choose a fixed weigh factor (2 as a default),
  or vary it as a function of lead time and
  location ( how to determine variations?)
• apply bias correction scheme to 35 variables (
  NCEP CMC )
• apply bias correction on 1 x1 degree ensemble
  data (NCEP CMC )
• apply bias correction on 00z and 12Z (NCEP CMC,
  06 18Z for NCEP )
• Second moment correction
• may not be included in next spring operational
  implementation
• Weighting
• BMA method only tested for surface temperature
• Dave Ungers scheme based on skill measure
• If 1 or 2 dont improve skill, use equal weight
  for all members in the combined ensemble for next
  spring implementation

4
List of Variables for Bias CorrectionCMC NCEP
Ensemble
5
NCEP RPSS 500mb Height, Northern Hemisphere
2004 Annual Mean
http//www.emc.ncep.noaa.gov/gmb/wx20cb/Bias_Corre
ction_Algorithm/1st_2nd_Moments/Training_1month/Pl
ot_Comb_Post/z500_2004_ncep_annual/
6
NCEP PAC 500mb Height, Northern Hemisphere
2004 Annual Mean
7
NCEP RMS 500mb Height, Northern Hemisphere
2004 Annual Mean
8
NCEP RPSS 500mb Height, Southern Hemisphere
2004 Annual Mean

9
NCEP PAC 500mb Height, Southern Hemisphere 2004
Annual Mean
10
NCEP RMS 500mb Height, Southern Hemisphere 2004
Annual Mean

11
NCEP RPSS 500mb Height, Tropical 2004 Annual
Mean
12
NCEP PAC 500mb Height, Tropical 2004 Annual
Mean
13
NCEP RMS 500mb Height, Tropical 2004 Annual
Mean
14
CMC RPSS 500 mb Height, Northern Hemisphere
March, 2005 May, 2005


15
CMC RMS 500mb Height, Northern Hemisphere
March, 2005 May, 2005

16
CMC PAC 500mb Height, Northern Hemisphere
March, 2005 May, 2005

17
CMC RPSS 500 mb Height, Northern Hemisphere
March, 2005 May, 2005


18
CMC RPSS 500 mb Height, Northern Hemisphere
June, 2005 July, 2005


19
Combined RPSS (5 NCEP 5 CMC) 500 mb Height,
Northern Hemisphere, Jan 15 2005 Feb 28 2005


Data from old system
20
Combined RPSS (5 NCEP 5 CMC) 500 mb Height,
Southern Hemisphere, Jan 15 2005 Feb 28 2005


Data from old system
21
Questions and Comments?
22
NCEP PAC 500mb Height, Northern Hemisphere
March, 2005 May, 2005
23
NCEP RPSS 500mb Height, Northern Hemisphere
March, 2005 May, 2005

24
NCEP RMS 500mb Height, Northern Hemisphere
March, 2005 May, 2005

25

Correlation Between the Observed Anomalies and
Fcst Errors

• Preliminary Results
• The ens. mean fcst. error is a function of
  lead time. The correlation between the
• observed anomalies and the ens. mean fcst error
  is very high. This suggests that
• the ens. mean fcst error is dominated by the
  observed verifying anomalies. The
• time mean errors may not be closely related to
  systematic errors.
• Future plan
• remove the observed anomaly from the error fields
  before they are used as
• estimates of the bias.
• Method
• decompose the total error into (a) component
  parallel to obs. anomaly
• (b) residual error, orthogonal to obs.
  Anomaly ( M. Wei ).
• remove error component along obs. anomaly from
  total error and work with residual component for
  bias estimation.