Predictability of High Impact
Weather during the Cool Season
over the Eastern U.S: CSTAR
Scientific Objectives
Brian A. Colle and Edmund Chang
Stony Brook University - SUNY
CSTAR Partners/Contributors:
HPC – David Novak et al.
EMC – Yucheng Song, Jun Du, and Jordan Alpert
OPC – Joseph Sienkiewicz et al.
WFO-OKX: Jeff Tongue et al.
WFO-PHI: Al Cope et al.
WFO-CTP: Richard Grumm et al.
WFO-PIT: Josh Korotky et al.
The operational community has made significant
progress in ensemble modeling and products
during the last decade
However, why aren’t ensembles
used more effectively in operations?
• Ensemble underdispersion and biases limit
ensemble skill.
• Ensembles have not been comprehensively
verified, especially for high impact weather.
• Forecasters lack tools to understand the
origin of ensemble spread and errors in
realtime.
• Forecasters have few ways to communicate
uncertainty in their forecast products
Novak, David R., David R. Bright, Michael J. Brennan, 2008: Operational Forecaster Uncertainty
Needs and Future Roles. Wea. Forecasting, 23, 1069–1084.
Short-Range Ensemble Systems
Stony Brook Univ. 13 Member Ensemble (00 UTC)
- 7 MM5, 6 WRFv2.2 members at 12 km grid spacing (NAM soil moisture and SST).
- IC: NAM, GFS, CMC and NOGAPS.
- CP: Grell, KF and Betts-Miller.
- PBL MM5: Blackadar, MRF, MY
- PBL WRF: MYJ and YSU.
- MP MM5: Sice, Reis2
- MP WRF: Ferrier, WSM3.
NCEP SREF 21 Member (21 UTC)
- 10 ETA members at 32 km grid spacing.
- 5 with BMJ CP and Ferrier MP.
- 5 with KF CP and Ferrier MP.
- 5 RSM members at 45 km grid spacing.
- 3 with SAS CP and Zhou GFS MP.
- 2 with RAS CP and Zhou GFS MP.
- 3 WRF-NMM members at 40 km grid spacing.
- 3 WRF-ARW members at 45 km grid spacing.
- IC's are perturbed using a breeding technique.
Daily NCEP/NCAR Composite of 500 mb Z Anomaly
For GFS F96 Cyclone SLP Mean Errors (> 1.5 stndev)
along the U.S. East Coast
Positive Cyclone SLP Error (28 cases)
GFS Underdeepened Events F96
Negative Cyclone SLP Error (35
cases)
GFS Overdeepened Events F96
Colle and Charles: Accepted to WAF
CSTAR Scientific Motivation
• Improve the understanding of high
•
•
•
impact weather predictability during
the cool season through objective
verification of cyclones and Rossby
wave packets (RWPs) in ensembles.
Integrate RWPs more in operations:
Better understanding of RWP
climatology, downstream impact of
targeted observations, and linkage
with extreme weather.
Better understanding of the
predictability of some mesoscale
phenomena (e.g., snowbands).
Better ensemble construction and
post-processing.
From THORPEX International Science Plan
(Shapiro and Thorpe, 2004)
Forecaster Motivation
• Forecasters need more knowledge
of RWPs and how the packets are
related to high impact weather,
regime changes, etc...
Schwierz et al (2004), Martius et al
(2007)
- Wave packet signal precedes Alpine
heavy precipitation events, especially
in DJF and SON
• Relationship between ensemble
sensitivity and wave packets may
alert forecasters to the potential
predictability of important weather
phenomena days in advance.
DJF
180 Longitude
180
• Forecasters need post-processed
ensemble data and better ways to
display uncertainty (ensemble)
information.
SON
Stony Brook Wave Packet Diagnostics for Winter TPARC
http://xs1.somas.stonybrook.edu/~chang/personal/Wave/main.htm
Stony Brook Univ. – NCEP-HPC collaboration:
Real-time Wave Packets in Operations
Courtesy: Mike Bogner
Relating Wave Packets to Medium Range Forecasts:
Large spread regarding Eastern U.S. trough verifying
1200 UTC 4 November 2010
Courtesy: Dr. David Novak
500 mb Analysis Diffs 00 UTC 29 Oct 2010
GFS Height (thin yellow)
ECMWF Height (thick yellow) and Isotachs (blue)
Winter Storm Reconnaissance Program
Operational since January 2001
Objective: Improve Forecasts of Significant Winter Weather Events Through
Targeted Observations in Data Sparse Northeast Pacific Ocean in the 24-96
hour lead time range over CONUS
Adaptive approach to collection of observational data:
1) Only prior to significant Winter Weather events of Interest
2) Only in areas that influence high impact event forecasts
Results:
70+% of Targeted Numerical Weather Predictions Improve
10-20% error reduction for high impact event forecasts
2008 P-3 out of Portland, WA
12-hour gain in predicting high impact events – earlier warnings possible
WSR data has 2.7 times more impact per obs than RAOB observations
Courtesy: Yucheng Song
Concept
Take observations in key sensitive areas (target area) that will
improve the forecast in the verification region.
Target
Obs. Area
Verification
Area
Need additional observational platforms (primarily aircraft)
Need method to determine where additional observations
should be taken
To
Tv
Observation errors taken from GDAS
Ensemble forecast perturbation given
Sensitivity maps for East Coast system at
12 UTC 4 Nov 2010
Use ensemble transform Kalman filter (ETKF, Bishop et al., 2001), which combines
error covariance information from ensemble forecasts (~170 members) with error
statistics associated with the routine and adaptive observational networks to predict
the reduction in forecast error variance within a prescribed forecast verification
region, due to the assimilation of any given set of adaptive observations.
Red circle shows our
key area of interest
Black contours show
us the sensitivity –
larger the values the
more impact
observations in those
areas at the Obs time
would have on the key
east coast area.
-All maps based on ensembles initialized at 00 UTC 29 Oct
Possible flight tracks
are also overlaid.
Sensitivity Map
GFS forecast
trough
Chaba
Sensitivity to hypothetical observations taken 00 UTC 30 October
Courtesy: David Novak
to improve the forecast over Eastern US 12 UTC 4 November
Sensitivity Map
GFS forecast
trough
Chaba/trough
Sensitivity to hypothetical observations taken 00 UTC 31 October
to improve the forecast over Eastern US 12 UTC 4 November
Sensitivity Map
GFS forecast
Broad
trough
Ahead of
trough
Sensitivity to hypothetical observations taken 00 UTC 1 November
to improve the forecast over Eastern US 12 UTC 4 November
Sensitivity Map
Ahead of
trough
GFS forecast
Broad
trough
Downstream
trough
Sensitivity to hypothetical observations taken 00 UTC 2 November
to improve the forecast over Eastern US 12 UTC 4 November
Sensitivity Map
GFS forecast
Energy over
the ridge
Sensitivity to hypothetical observations taken 00 UTC 3 November
to improve the forecast over Eastern US 12 UTC 4 November
Sensitivity Map
GFS forecast
Verification region
Sensitivity to hypothetical observations taken 00 UTC 4 November
to improve the forecast over Eastern US 12 UTC 4 November
Relationship Between Sensitive Regions
and Wave Packet
300 hPa Packet Envelope
Time
300 hPa U and V wind
Asia
= sensitive
regions identified
in previous slides
USA
Courtesy: Dr. David Novak
Using Ensemble Data More Effectively in the
Forecast Office
Advanced Linux Prototype System (ALPS)
http://www-sdd.fsl.noaa.gov/~ramer/alps/ensembles/ensembles.html
Warm Season vs Fire Threat Days Surface Temperature
Mean Error (12-36h) for SBU and SREF Ensembles
NCEP SREF
(oC)
(oC)
SBU 12-km Ensemble
MYJ
MM5
WRF
SREF sub-group averages
Courtesy: Michael Erickson
Impact of Using Previous 5 Fire Threat Days
for Bias Correction of Surface Temperature (12-36 h avg)
NCEP SREF
(oC)
(oC)
SBU 12-km Ensemble
MM5
WRF
Note: MAE for SBU+SREF ens mean is 1.84 K (~0.10 K less than best
member, but comparable to SBU mean).
Impact of Using Previous 5 Fire Threat Days
for Bias Correction of Surface Temperature (12-36h avg)
Raw: SBU 12-km Ensemble
(oC)
After Bias Correction
(oC)
Impact of Using Previous 5 Fire Threat Days
for Bias Correction of Surface Temperature (12-36h avg)
SBU 12-km Ensemble T > 282 K
Improvement in BSS for Temp
25 Dec 2002 Ensemble of
12-km MM5, WRF, NCEP
SREF
(15/16 produced band)
18-21h Ensemble: 12 February 2006 Event
(LESS PREDICTABLE – WHY???
Using ensemble sensitivity analysis
from a ENKF system (Torn and Hakim
2008), one can identify where small
changes to the initial conditions can
have a significant impact on the
subsequent forecast in the banding
region. Forecasters can monitor
observations in these sensitive regions.
25 Dec 2002: IC and 18 h PV (475-250 hPa) 36-km
difference between farthest NW and SE band position
IC
-1
-2
0
1
-1
18 h
2
3
0
4 PVU
1 PVU
12 Feb 2006: IC and 18 h PV (475-250 hPa) 36-km
difference between farthest NW and SE band position
IC
-1
-2
0
1
-1
2
0
18 h
3
1
4 PVU
2 PVU
Summary
• There are still many challenges in using ensembles
more effectively in operations.
• Recent progress has been made in the automated
tracking and verification of wave packets.
• Ensemble post-processing shows promise (e.g. Jun
Du is now testing bias correction for SREF).
• On-line CSTAR tutorials have been completed for
Rossby wave packets, ALPS software, and targeted
observations.
• See our CSTAR Web page for more details, and if you
want to join our CSTAR email group.
http://dendrite.somas.stonybrook.edu/CSTAR/cstar.html