the 2004 operational synoptic surveillance missions: an...
TRANSCRIPT
The 2004 Operational Synoptic Surveillance Missions: An Unprecedented Dataset for
Tropical Cyclone Targeting Research
Sim AbersonNOAA/AOML/Hurricane Research Division
First THORPEX International Science Symposium 8 December, 2004Montreal, Quebec
Special thanks to the AOC, HRD, and AF crews of the G-IV and C-130s, especially Jack Parrish; CARCAH; NCEP/GMB and CCO, and TPC, especially James Franklin.
Brian EthertonUniversity of North Carolina-Charlotte
Sharanya MajumdarUniversity of Miami/RSMAS/MPO
2004 Summary31 total cases > 1400 dropwindsondes
30 NOAA G-IV missions NOAA aircraft alone5 AF C-130 missions
26 one-plane G-IV missions1 one-plane C-130 mission4 G-IV/C-130 two-plane missions
19 landfalls during forecast periodsMore than US$43 billion in damageMore than 3500 deaths
This is the highest number of missions during one year since operational surveillance began in 1997
Targeting and Sampling Strategies1982 - 1998: Uniform spread1998 - : NCEP ensemble spread2004 - : Ensemble Transform Kalman Filter
Previous Results
2002-2003GFS: 10 - 32%
improvementGFDL: -18 - 20%
improvement(mixed results)
2004 Results
Mixed results from both GFS and GFDL
Average errors by
storm
Average improvement
by storm
Most GFS degradation from Ivan
GFDL continues to notrespond well toadded observations,possibly a result of vortex removal technique removingdata
Some tests
An operational parallel cycle was run from the beginning of Frances through the end of Jeanne (2004082900 through 2004092600) since only a few days separated each case.
Some testsA second nodrop cycle (AVAL) was started at the beginning of Ivan (2004090600). All Frances sondes were included to see the impact of the Frances missions on the later Ivan forecasts.
A second alldrop cycle (AVN7) was begun for the second Ivan mission (2004090800), the first with a large degradation, by removing all previous sondes but using those from that mission.
AVNN - nodropAVNO - operational alldrop
The Frances data 4 to 9 days earlier still had an impact (in this case positive) on this Ivan forecast (AVAL). This difference became larger in time.
The surveillance data from the previous day (leading to a large forecast improvement)had a huge negative impact on the subsequent forecast (AVN7).
The only difference between the orange and green is the removal of surveillance data from 2004090700, 9 days earlier.
The only difference between the orange and black is the removal of surveillance data through 2004090412, 11.5 days earlier.
Impact of 2004090800 sondesSonde impact maxima are never co-located with sonde locations.
The data create an asymmetry in the vortex that sends the storm northward. This may be a residual from the previous day's data.
QuestionsAberson (2003) and the current work have shown that additional data, especially at the edges of regions of data coverage and from serial targeting, can degrade forecasts, sometimes significantly.
1. Can we identify the times and regions when this might occur?2. Does this imply that "negative" targeting - removal of some regular observations such as rawinsondes - might be advantangeous?
Future Field ProgramsOperational reconnaissance and surveillance began in 1945 and will continue annually in the Atlantic.
2005 - TCSP, RAINEX, IFEX, operationsNOAA P3s, NOAA G-IV, NASA DC-8, NASA ER-2, NRL P3, AF C-130-Js, aerosonde, driftsonde
2006 - AMMA, IFEX, operationsNOAA P3s, NOAA G-IV,AF C-130-Js, aerosonde, etc.
DataAll data are available:
http://www.aoml.noaa.gov/hrd/[email protected]
Includes dropwindsonde, flight-level, radar reflectivity, airborne Doppler data, upper ocean data, SFMR, SRA, cloud physics, electrification, etc. See website for details.
Summary
2004 was a record-breaking year in terms of numbers of surveillance missions and dropwindsondes released.
Results were mixed, and more research into data assimilation the optimal mix of observations in both time and space is recommended.