Satellite-Derived Atmospheric Motion Vectors (AMVs): Tropical Cyclone Data Assimilation and NWP Impact Studies Howard Berger1, C. Velden1, R. Langland2, C. Reynolds2 Hui Lui3, Jeff Anderson3, and Sharan Majumdar4. 1-Cooperative Institute for

1
Satellite-Derived Atmospheric Motion Vectors
(AMVs) Tropical Cyclone Data Assimilation and
NWP Impact Studies Howard Berger1, C.
Velden1, R. Langland2, C. Reynolds2Hui Lui3,
Jeff Anderson3, and Sharan Majumdar4.1-Cooperati
ve Institute for Meteorological Satellite
Studies, Univ.-Wisconsin2-Naval Research
Laboratory, Monterey, CA3-NCAR Institute for
Mathematics Applied to Geosciences4-RSMAS/Univer
sity of Miami
2
Outline

• Brief Review of Recent Tropical Cyclone Studies
  Examining the Impact of AMVs
• NRL-CIMSS Collaborative Efforts using
  NAVDAS/NOGAPS with AMV Datasets Processed during
  TPARC
• NOPP Collaborative Efforts with NCAR and
  RSMAS/UMiami Mesoscale WRF-DART AMV Data Impact
  Experiments

3
Recent Tropical Cyclone Studies Examining the
NWP Impact of AMVs

• Goerss and Velden, 1998 MWR (NOGAPS)
• Soden and Velden, 2001 MWR (GFDL)
• Kelly, 2004 ECMWF Report (ECMWF)
• Zapotocny et al., 2005 WAF (NCEP/AVN)
• Goerss, 2009 MWR (NOGAPS)
• Langland, Velden and Berger, 2009 MWR (NOGAPS)
• Berger, Langland, Velden, Reynolds, 2011 JAMC
  (NOGAPS)

4
GFDL Direct Assimilation of AMVs (Soden and
Velden, 2001)
5
Impact of AMVs in ECMWF (Kelly, 2004) 200 hPA
Vector Wind in the Tropics
6
Impact of AMVs in NCEP/AVN (Zapotocny et al.,
2005)
7
Impact of AMVs in NCEP/AVN (Zapotocny et al.,
2005)
8
Impact of AMVs in NOGAPS (Goerss, 2009)
Mean Track Forecast Error
- Forecast Hr - of Cases
9
Impact of AMVs in NOGAPS (Goerss, 2009)
MFE Degradation After Removal
Forecast Hr
10
Katrina Case Study Impact of GOES Rapid-Scan
AMVs on NOGAPS Track Forecasts (Langland et al.,
2009)
11
Katrina Case Study Impact of GOES Rapid-Scan
AMVs on NOGAPS Track Forecasts (Langland et al.,
2009)
12
Katrina Case Study Impact of GOES Rapid-Scan
AMVs on NOGAPS Track Forecasts (Langland et al.,
2009)
NOGAPS 48hr forecast of Hurricane Katrina
positions verifying at 12 UTC 29 August 2005.
RS AMV forecast (red) and CNL forecast (blue).
Observed track (green). All positions indicated
at 12-hr intervals.
13
SPECIAL AMV DATA ANALYSIS AND NWP IMPACT STUDIES
DURING TPARC Howard Berger1, C.S. Velden1, R.
Langland2, and C. A. Reynolds21-Cooperative
Institute for Meteorological Satellite Studies,
Univ.-Wisconsin2-Naval Research Laboratory,
Monterey, CA Presented by C. Velden at the
recent WMO DAOS committee meeting, Montreal,and
paper being submitted to JAMC
14
T-PARC Thorpex - Pacific Asian Regional Campaign

• International field campaign during August
  October, 2008 with special observing periods to
  investigate the formation, structure,
  intensification and prediction of tropical
  cyclones in the western North Pacific.

15
AMV Processing for TPARC

• Generated at CIMSS (essentially the operational
  NESDIS algorithm) by objectively targeting and
  tracking clouds and WV structures in sequential
  JMA MTSAT multi-spectral geostationary satellite
  images
• AMV heights are assigned using multispectral and
  semi-transparency techniques
• Apply objective quality control and assign
  quality indicators (QI)

16
Special AMV Datasets for TPARC

• 1) Hourly datasets generated from routinely
  available MTSAT imagery (30-min hemispheric
  images), for the entire duration of the
  experiment
• 2) Datasets generated from special MTSAT-2
  rapid-scan 15-minute images over the western
  North Pacific for limited periods during selected
  TCs

17
MTSAT AMVs produced hourly (by UW-CIMSS) during
TPARCExample Typhoon Sinlaku -- 11th Sep. 2008
18
Example of AMVs from MTSAT-2 Rapid Scan images
Left AMV (IR-only) field produced from routinely
available 30-min sequence of MTSAT-1 images
during Typhoon Sinlaku Bottom Left Same as
above, but using a 15-min rapid scan sequence
from MTSAT-2 (better AMV coverage and
coherence) Bottom Right Same as above, but
using a 4-min rapid scan sequence (improved
coverage/detail of typhoon flow fields)
19
NRL/FNMOC Analysis System(Naval Research
Lab/Fleet Numeric Meteorology and Oceanography
Center)
NAVDAS-AR NRL Atmospheric Variational Data
Assimilation System-Accelerated Representer

• Full 4D-VAR algorithm solved in observation space
  using representer approach
• Weak constraint formulation allows inclusion of
  model error
• T239L42, model top at 0.04 hPa
• More effective use of asynoptic and single-level
  data
• More computationally efficient than NAVDAS for
  large of obs
• Adjoint developed for observation impact with
  real-time web monitoring capability

20

• NRL/FNMOC Analysis System(Naval Research
  Lab/Fleet Numeric Meteorology and Oceanography
  Center)
• Superobbing strategy for AMVs
• First remove any duplicates and obs from
  deselected levels, channels
• Superob only like obs in a 2lat/lon prism in a
  50 mb layer
• Obs from same satellite, same channel, same
  time (or nearly so)
• At least two consistent observations required
• Require all winds to agree within specified
  criteria
• Speed, u and v criteria vary as a function of
  windspeed
• from 7 m/s for mean speeds less than 25 m/s
• to 14 m/s for mean speeds greater than 75 m/s
• u and v criterion sqrt(((speed
  criterion)2)/2) to ensure
  consistency with speed criterion
• Alternate direction criterion specified to be
  lt20
• Innovations (superob background) are calculated
  and used in NAVDAS to produce the analysis.
  Observation errors assigned to the superobs are
  assumed to be the same as for operational geo
  AMVs.

21
AMV Data Assimilation Experiments Collaboration
with Rolf Langland and Carolyn Reynolds at the US
Naval Research Lab (NRL) in Monterey

• Continuously assimilate all hourly MTSAT AMV
  datasets using NRL 4DVAR during the 2-month TPARC
  period
• Assess impact on NRL/FNMOC NOGAPS TC forecasts
• CTL All conventional and available special
  TPARC observations (except for dropsondes),
  including hourly AMV datasets from MTSAT-1 (but
  no rapid-scan AMVs)
• EX1 (No-CIMSS AMV) CTL with hourly AMVs removed
• Rapid-Scan CTL with Rapid-Scan AMVs included

22
NOGAPS track forecasts (nm) for TPARC
NOGAPS run with hourly and Rapid-Scan AMVs
reduces TC track forecast errors notably at
longer forecast times
23
AMVs reduce the larger track forecast busts at
120-hours
24
Example Typhoon Sinlaku 120-h forecast on Sept.
11, 2008 12UTC MSLP (hPa)
Control w/ AMVs
Best-Track
NO-AMVs
Influence of transient mid-latitude troughs??
25
Example Typhoon Sinlaku 120-h forecast on Sept.
11, 2008 12UTC MSLP (hPa)
Control w/ AMVs
Best-Track
Rapid-Scan
Influence of transient mid-latitude troughs??
26
500 hPa analyses in the Mid-Lats during TC
Sinlaku
Hourly MTSAT AMVs have positive impact,
particularly during the period of large NOGAPS
track forecast errors (NOAMV exp.) for Sinlaku
27
Summary

• Hourly satellite-derived AMVs allow for more
  consistent temporal coverage of the evolving
  atmospheric flow. The NRL 4DVAR DA can
  effectively utilize this frequently available
  information, resulting in improved NOGAPS TC
  track forecasts (e.g. TY Sinlaku), particularly
  at longer ranges (3-5 days).
• Rapid-Scan AMVs can better capture mesoscale flow
  features such as present in rapidly evolving TCs,
  leading to more precise kinematic diagnostics.
  They also show positive impact in NOGAPS TC track
  forecasts, and have promising applications in
  mesoscale data assimilation.

28
NOPP Collaborative Efforts with NCAR and
RSMAS/UMiami Mesoscale WRF-DART AMV Data Impact
Experiments (Hui Liu and Jeff Anderson)

• Initial Case Studies Typhoon Sinlaku (western
  North Pacific during TPARC), and Hurricane IKE
  (Atlantic in 2008)
• Experiments with 6- and 3-hourly
  assimilation/analyses.
• EnKF - 32 ensemble members are used in the
  assimilations.
• Assimilations started one week before TC genesis.
• 9km moving nest grid with feedback to 27km grid
  in the 6-hourly (or 3-hourly) forecast when a TC
  is present.
• Assimilation and analyses on 27km grid only.

29
Analysis Experiments - Hurricane Ike

• Control (CTL) 6-hourly analysis cycle. All
  routine operational data (Radiosonde, AMVs,
  surface, aircraft) and NHC/JTWC advisory TC
  positions.
• CIMSS-RS6h CIMSS rapid scan AMVs replace
  operational AMVs, 6-hourly analyses.
• CIMSS-RS3h As above, but 3-hourly analyses
  (3-hour cycle may be needed to increase benefits
  of the rapid scan obs).
• __________________________________________________
  
• Only the AMVs at the analysis times are used (no
  off-time assimilation attempts yet).
• Only analyses are finished at this point (no
  forecast results yet).

30
Example of Operational AMVs for Ike
31
Example of CIMSS Rapid-Scan AMVs for Ike

• Radiosondes,

32
Wind Analysis Increment at 300 hPa (12Z Sep 02,
2008)
CIMSS-RS6h
CTL
33
Track and Intensity Analyses for Hurricane Ike
34
Summary

• Recent studies regarding the impact of
  satellite-derived AMV observations on NWP
  tropical cyclone forecasts show positive results.
  AMVs are still very much relevant in the
  tropics!
• Efforts to optimize the assimilation of AMVs
  continue on two fronts
• 1) Global assimilation/models
  (thinning, superobbing, better utilization of AMV
  quality indicators, hourly assimilation).
• 2) Mesoscale assimilation/models (use
  of rapid-scan AMVs, EnKF methods to optimally
  assimilate high density space/time AMV obs, use
  of AMV data in lieu of or to augment TC bogus
  vortex pseudo-obs, focus on TC intensity/structure
  improvements).

35
Extra Slides
36
NOGAPS track forecasts (nm) for TPARC
Forecast Time (hrs) 0 12 24 36 48 60 72 84 96 108 120
Control w/ AMVs 22 39 70 93 114 151 213 195 167 248 317
No-AMV 22 40 67 91 108 154 227 248 245 365 450
Rapid-Scan 25 45 78 111 122 158 210 174 135 215 260
CASES 22 20 18 16 14 13 12 11 9 8 7

• NOGAPS run with hourly and Rapid-Scan AMVs
  reduces TC track forecast errors notably at
  longer forecast times

37
Impact of AMVs in ECMWF (Kelly, 2004)
NH
TP
SH