1. Impacts of Rossby Wave Packets on Forecast Uncertainties and Errors
Brian A. Colle, Edmund K.M. Chang, and Minghua Zheng
School of Marine and Atmospheric Sciences
Stony Brook University
Stony Brook, New York, USA
NROW 14
10 December 2013

2. Outline Motivation Relationship between RWPs and Forecast Errors
-- for relatively large East coast medium range errors
-- evolution of errors relative to a moving RWP
-- ensemble spread starting with RWP over Pacific
Ensemble Wave Packet Page and Updates to Ensemble Sensitivity Page
Summary and Ongoing Research

3. What are Rossby Wave Packets? An example (adapted from Chang 1993)
Contours: 300 hPa Z D E

4. A B A B C B C D C D E
Contours: 300 hPa Z
Shades: 300 hPa v D E

5. Hovmoller (Time-Longitude) diagramB C D E

6. Climatology of RWPA (Souders et al. 2013 –submitted to MWR)
Max: ~18m/s across NA
Min: < 8m/s central
Asia
Unit: m/s 6

7. From THORPEX International Science Plan
(Shapiro and Thorpe, 2004)
RWPs have frequently been linked to high impact weather downstream
Cyclogenesis near Japan
Flooding over Europe

8. Taken from U.S. TPARC science plan. Adopted from Hakim (2005)
Initial analysis error structure
12-hr forecast uncertainty
24-hr forecast uncertainty
Taken from U.S. TPARC science plan. Adopted from Hakim (2005)

9. Another example showing the similarity between RWPs and ensemble sensitivity signals (2010 Christmas snowstorm)
Sensitivity
Magenta: region for calculating EOF of ensemble meridional wind v300.
State field: v300
Wave packet
Downstream development
Group V > Phase V
Unit: m/s
From Zheng et al. (2013)

10. Forecast Errors
Data 1:
GFS analysis and 7-day control run forecast for 5 cool seasons ( )
Parameters: geopotential height at 300 hPa (Z300)
Data 2:
RWP amplitude: calculated by Matthew Souders using u,v and Z at 300 hPa
Method
Composite method
RMSE of Z300
N: total grids number
Absolute error

11. Preliminary results: verification region and large error cases selections
Verification region (VR):
22°N-63°N, 99°W-46°W
Inter Quartile Range (IQR):
Third Quartile - First Quartile
Drop-out cases: RMSE>=mean+1.2*IQR (75 cases)
Good forecasts: RMSE<=mean-1.0*IQR (79 cases)

12. Composited RWPA anomaly for large error cases
Initial Positive RWPA
Anomaly Develop and
Propagate into VR
Unit: m/s
The purple contour corresponds to 95% significance level

13. For 46 most robust wave packets over NAmer (2007-2013):
Composited RWPA (shaded, [m/s]) and the normalized 300 Z forecast error (Magenta Contours, forecast error divided by maximum error,[%]). Errors at day -3 to +3 correspond to 0.5 day to 6.5 day forecast errors.

14. For 46 most robust wave packets over NAmer (2007-2013):
Composited RWPA (shaded, [m/s]) and the normalized 300 Z forecast error (Magenta Contours, forecast error divided by maximum error,[%]). Errors at day -3 to +3 correspond to 0.5 day to 6.5 day forecast errors.

15. Wave packet relative errors for same 46 cases, which includes the composited RWPA and errors with the centers representing the weighted wave packet centers at each step.
Wave packet amplitude (shaded, [m/s]) and forecast errors (magenta contour, [%]). The composites are centered at the weighted centers of RWPA at each step.

16. 25 coherent RWPS starting from central Pacific are selected to calculate the RWPA mean and 300Z normalized spread divided by max spread (%) using the 20-member GEFS from day 0 to day 6.5

17. Days 3-5

18. Days 6-6.5

19. Discussions: sensitivity to VR―composited RWPA anomaly for large error cases (EUROPE)
Day 0
Day 4
Day 1
Day 5
Very strong Positive
RWPA Anomaly
Develop and
Propagate into VR
Day 2
Day 6
Unit: m/s
The green contour corresponds to 95% significance level
Day 3
Day 7

20. (http://wavy. somas. stonybrook. edu/wavepackets/home. html)
( Figure 8 shows the cover of the page, in which users can

21. Updates to Ensemble Sensitivity Web Page
). Figure 8 shows the cover of the page, in which users can

22. Summary
Composited RWPA for GFS large error cases over U.S. East Coast shows a positive RWPA anomaly originating from eastern Asia and propagating across the Pacific into the verification box. Large error cases over Europe also show a propagating RWPA positive anomaly upstream.
-- These results suggest that large error cases are associated with the presence of aenhanced RWPA upstream.
Early in the forecast over the Pacific, the largest errors tend to be in the middle of the RWP, but as the packet spreads across North America in the medium range, the largest errors (and ensemble spread) occur more around the leading edge of the packet (**Could be an important signal to look for while forecasting…)

23. Ongoing Research Examine ensemble forecasts using TIGGE data
To study the relationship between RWPs and forecast uncertainties
Study how forecast uncertainties and errors grow in multi-model
Compare ensemble outputs from different operational models