Transcript R. Langland

DAOS Working Group
1
The Assessment of the DAOS WG
on Observation Targeting
Talk presented by
Rolf Langland (NRL-Monterey)
THIRD THORPEX International Science Symposium
Monterey, CA, 16 Sept 2009
Data Assimilation and Observing Strategies
THORPEX Working Group
2
Members:
Pierre Gauthier (Univ. of Quebec, co-chair)
Florence Rabier (Metéo France, co-chair)
Carla Cardinali (ECMWF)
Ron Gelaro (NASA-GMAO)
Ko Koizumi (JMA)
Rolf Langland (NRL-Monterey)
Andrew Lorenc (UK Met Office)
Peter Steinle (Australian Bureau of Met)
Michael Tsyroulnikov (Hydromet, Russia)
THORPEX DAOS Working Group
OBJECTIVES
3
• Promote research activities that lead to more-
optimal use of observations and understanding
the sources of errors in analyses and forecasts
• Contribute to development of a strategy for
evolution of the global observing system to
support NWP
• Provide guidance and evaluation for observation
deployment (including targeting) in THORPEX
regional campaigns
DAOS-WG Overview Talk - by Pierre Gauthier
17:30 Thurs Sept 17
Targeted Observing Goals
4
Deploy sets of additional observations to improve
the forecast skill of weather events anticipated to
have large societal impact
• Strong extratropical cyclones and frontal waves
• Tropical cyclones and typhoons
Identification of target areas: adjoint and ensemblebased methods
Observational resources: dropsondes, radiosondes,
ship and land-surface observations, aircraft-based
lidar, satellite observations
Field Programs for Targeted Observing
5
Winter storm targeting
• North Atlantic (FASTEX-1997, ATREC-2003)
• Eastern North Pacific (NORPEX-1998, WSR-1999-2009)
• Entire North Pacific (Winter T-PARC 2009)
Hurricane / tropical cyclone targeting
• North Atlantic (NOAA-HRD, 2000-2009)
• Western Pacific (DOTSTAR, 2003-2009)
T-PARC (TCS-08) 2008
Other Programs: AMMA, IPY
Participants: Meteo France, ECMWF, UKMO, NRL, NCEP, NCAR, JMA,
Taiwan, NOAA-AOC, NOAA-HRD, USAF Hurricane Hunters, NASA,
CIMSS, MIT, Univ. of Miami, Penn State Univ., others
Does “dropsonde targeting work” ?
6
Yes -
however ……
• Dropsonde deployments provide only partial
and intermittent surveys of target areas, so “full
impact” of targeting has not been realized
”Targeted observing” may be more effective with
methods other than dropsonde deployment
(e.g., use of satellite observations)
Singular Vector Target in FASTEX IOP-17
7
Forecast
Verification Region
NOGAPS
Dropsondes from
NOAA G-IV
SV-based Target Area
Targeting to improve 42-hour forecast of intense
cyclone over Ireland and Great Britain
ETKF Target in NOAA Winter Storm
Reconnaissance Program
8
ETKF Target Area
Dropsondes from NOAA G-IV –
only in localized area of maximum
sensitivity
Predicted Signal Propagation from
Dropsonde Assimilation
70% of WSRP cases improved by up to 12 hours
Song,Toth & Majumdar 2008
Impact of NORPEX targeted dropsondes
16 January – 27 February 1998 (NRL-NCEP)
9
45 forecast cases, ~ 10% mean error
reduction over western North
America, using NOGAPS forecast
model
700 dropsondes
RMSE 500mb ht
of 2-day forecasts
IMPROVED
FORECASTS
(n=35)
DEGRADED
FORECASTS
(n=10)
Error (m) with targeted dropsondes
Langland et al. 1999 (BAMS)
THE LAW OF LARGE (and small) NUMBERS OF OBSERVATIONS
10
Assimilating a very large number of observations is (almost) guaranteed to
improve forecast skill. Not true for smaller sub-sets of observations ….
Fcst Error
Increase
Degraded Forecast
targeting deployment #1
Fcst Error
Reduction
Improved Forecast
targeting deployment #2
Mid-latitudeTargeting Program Results
FASTEX, NORPEX, ATREC, WSRP, Winter-TPARC
11
Forecast Impact of Targeted Data – (adding 10-50 dropsondes at
single assimilation times)
• Targeted data improves the average skill of short-range
forecasts*, by ~ 10–20% over localized verification regions –
maximum improvements up to 50% forecast error reduction in
localized areas
• In all analysis / forecast systems*, and for all targeting
methodologies, it is found that ~ 20-30% of forecast cases are
neutral or degraded by the addition of targeted data
• Impact “per-observation” of targeted (dropsonde) data is about
3x larger than random observations, but total impact is generally
limited by the relatively small number of targeted data
* Based on published forecast impact studies performed at
NCEP, ECMWF, Meteo France, UKMO, NRL
Data Targeting System
• Results from up to 6 different centres displayed in common format
• > 500 individual cases during Aug- Sept 2008
12
JMA SVs
ECMWF SVs
NRL SVs
U. Yonsei SVs
UKMO ETKF
U. Miami/NCEP ETKF
Super Typhoon Jangmi: Targeting Time 28 Sept. 2008
TCS-08: Typhoon Sinlaku
NCEP GFS initialized 00 UTC 10th Sept
13
500 hPa ASYMMETRIC WIND DIFF
+18 h
JMA BEST
TRACK
WITH
DROPS
WITHOUT
DROPS
Effect of drops:
Strengthened vortex and subtropical ridge,
inducing northwestward flow
Provided by Sharan Majumdar (U. Miami)
Tropical Cyclone Targeting Program Results
NOAA-HRD, DOTSTAR, TCS-08 (summer-TPARC)
14
• Targeted observing produces significant
improvements to some TC track forecasts – some
cases not improved
• Different methods (ETKF, adjoint are in general
agreement on identification of target areas
• Satellite observations (including rapid-scan winds)
can provide more-complete and more-frequent
coverage of target areas and may produce larger
improvements in TC track forecasts
Improvement of Katrina track forecasts with assimilation of
Rapid-Scan wind observations
15
CNL
350
RS1
Control forecasts – no rapid-scan winds
300
Track Error
250
(n mi)
200
Forecasts with rapid-scan winds
150
100
Track forecast error
significantly reduced
50
0
108
96
84
72
60
48
36
NOGAPS forecast length (hr)
24
DAOS-WG recommendations for
Targeted Observing Field Programs
16
1. Expensive observation campaigns should not be
justified based only on previous methods of
targeting
2. Develop and test new targeting approaches –
consider use of targeted satellite observations
3. Carefully consider data assimilation issues
(impacts of small vs. large sets of observations,
frequency of special observations, etc. )
4. Consider pre-campaign tests with OSEs or OSSEs
Targeted Observing Research Issues
17
1. Impact of targeted observations from previous
field programs (esp. WSRP, TPARC)
2. OSSEs and predictability experiments with
synthetic observations
3. Adaptive selection and assimilation of satellite
observations (less than 10% of available data currently used)
4. Potential for targeted observations to improve
medium-range forecasts
What is the potential benefit from observing larger sections of the targeting
subspace, instead of attempting to survey the smaller-scale areas of maximum
sensitivity which have been the primary focus of previous field programs?
Targeting Strategies –
18
How much benefit can we obtain by “tuning” the
network of existing regular satellite and in-situ
observations in a targeted sense?
- On-request rapid-scan wind data
- Targeted satellite channel selection and data-thinning
- Increase observations from commercial aircraft
- Request radiosondes at non-standard times
Adaptive tuning of the regular observing system
19
Siberian Raobs – Winter TPARC
Petropavlosk
(32540)- very
large impact
Circled stations provided
ten or more profiles
1x10-3 J kg-1 (24h Moist Total Energy Norm)
Error Reduction
Error Increase
Other Data Assimilation Priorities – DAOS WG
20
1. Diagnosis, understanding and reduction of model
errors
2. Improved characterization of observation and
background error, esp. for satellite observations
and oceanic regions
3. Development of advanced, computationally
efficient data assimilation systems, including
4D-Var and the Ensemble Kalman Filter
Advances in these areas are likely to improve
forecasts as much or more than dropsonde-based
targeted observing, by itself
Uncertainty in Atmospheric Analyses
highly correlated with observing resource patterns
21
UKMET-GFS (200-500 hPa thickness – RMSDiff – 6 months)
SMALL
LARGE
UNCERTAINTY UNCERTAINTY
SATELLITE OBS
RAOBS
AIRCRAFT
What
implications
does this have
for future
targeting
programs?
Thickness (m)
Based on analyses at 0000 UTC and 1200 UTC from 1 Jan to 30 June 2007
Langland et al. (Tellus, 2008)
Day 1
Impact of removing ALL
winter Pacific obs in ECMWF
4D-Var: Normalized 500hPa
geopotential height rmse
differences between SEAIN
forecast and SEAOUT. Bluepurple show the negative
impact and yellow-black
positive impact of SEAOUT.
Panels (a)–(d) show
forecast errors for days 1,
2, 5 and 7.
Kelly et al. 2007
22
Day 2
Day 5
Day 7