T PARC: Impact of dropsonde (and other) aircraft data

Transcription

1 T PARC: Impact of dropsonde (and other) aircraft data Sharanya J. Majumdar (RSMAS/U. Miami) Martin, Florian Harnisch, Chun Chieh Wu, Tetsuo Nakazawa, Sim Aberson, Hyun Mee Kim, Munehiko Yamaguchi Session 5.3, THORPEX/DAOS WG Third Meeting 8 9 July Objective: To improve 1 3 day forecasts of tropical cyclone motion (track) by obtaining targeted observations in regions with high sensitivity. During the field phase, a team identified potential opportunities to collect targeted observations: Cases selected 2 3 days prior to observation time. Common verification regions, Guam, Taiwan, and Japan Individually selected verification regions: calculations performed through ECMWF/Met Office PREVIEW DTS Final flight paths chosen one day prior, based on targeted observation guidance and team consensus. Post field phase: Data denial w/ ECMWF, JMA and NCEP global models; WRF nested in KMA global model 1

2 PREVIEW Data Targeting System Sensitive Area Calculations (SAC) Automatic submission of 5 fixed areas Up to 5 additional areas chosen interactively Flexible choice of targeting time (t + 18 to 102 h) and verification time (t + 36 to 120 h) Courtesy of Sarah Jones, David Richardson, Cristina Prates and Cihan Sahin. Target area calculations: Obs. Time 00Z 11 th Sept. ECMWF SV NOGAPS SV JMA SV UM ETKF UK ETKF COAMPS ADJ 2

3 TCS 08 / T PARC 2008: Dropsondes DLR Falcon 20 US Air Force WC-130 US NRL P-3 DOTSTAR Astra jet F. Harnisch Data denial: two key papers (MWR 2010, in press) The influence of assimilating dropsonde data on typhoon track and mid latitude forecasts Martin, Florian Harnisch, Chun Chieh Wu, Po Hsiung Lin, Yoichiro Ohta, Koji Yamashita, Yeon Hee Kim, Eun Hee Jeon, Tetsuo Nakazawa, Sim Aberson Sensitivity of typhoon forecasts to different subsets of targeted dropsonde observations Florian Harnisch, Martin 3

4 Typhoons Sinlaku and Jangmi M. Data denial intercomparison ( et al. 2010) ECMWF IFS DLR Germany JMA GSM Japan KMA WRF NIMR KOREA TL799L91 TL959L60 T382L64 resolution (~25 km) (~20 km) 30 km (~38 km) DA-method 12h 4D-VAR 6h 4D-VAR 6h 3D-Var 6h 3D-Var domain Globe Globe bogus NO NO (YES in oper. version) 190*190 grid points use of TC core observations YES YES YES NO denied observations Pacific dropsondes driftsondes JMA ship SYNOP JMA ship TEMP Pacific dropsondes JMA ship TEMP JMA special TEMP NO Atlantic dropsondes NCEP GFS NTU Taiwan, NOAA Globe vortex relocation, bogus if no vortex in first guess (rare) Atlantic and Pacific dropsondes driftsondes 4

5 GFS 10 Sept 00 UTC JMA 11 Sept 12 UTC ECMWF 10 Sept 00 UTC ECMWF 11 Sept 12 UTC The influence strongly depends on the modeling and DA system --> need for comparison of models to draw conclusions on targeting! M. Period: and M. 5

6 Influence on ECMWF midlatitude forecasts 2 ) diff. forecast error (m 2 /s Pacific; lead time:96 h deterio oration improvement date only 2 Sinlaku flights improved track forecast --> improved first-guess for subsequent days --> improved mid-latitude forecast overall neutral influence of observations during ET, although these were partly guided by SV calculations optimized for the Pacific M. The influence of cycling, i.e. modification of the first guess DROP UnCy NODROP Error Pacific error (m 2 /s 2 ) lead time (h) The assimilation of dropsondes without the modification of the first guess shows little difference to NoDrop Through the modification of the first guess the influence can "accumulate" over several assimilation windows This "accumulation" seems essential to get a notable influence in a complex 4D Var system M. 6

7 Separation of dropsondes into three different subsets (Harnisch et al. 2010) remote sensitive areas (ReObs) typhoon vicinity (ViObs) typhoon center and core (CeObs) What is the influence on the typhoon track forecast? F. Harnisch ECMWF data denial experiments ECMWF Integrated Forecast System (IFS) resolution T799 L91 forecast time 240 hours (00 UTC, 12 UTC) ECMWF global data assimilation method 4D-Var assimilation window 12 hourly (-3 h to +9 h) resolution inner loop T95 L91, T159 L91, T255 L91 resolution outer loop T799 L91 control experiment with denying all dropsonde observations (NoObs) experiments with different subsets of dropsonde observations at 7 cases for Sinlaku and 5 cases for Jangmi F. Harnisch 7

8 Which subset of dropsondes improves the typhoon track forecast the most? typhoon vicinity remote sensitive regions typhoon center and core degradation degradation degradation improvement improvement improvement F. Harnisch Example of Falcon Doppler Wind lidar observations IR Satellite Image 09/11/ LT 16W Okinawa Tokyo Sinlaku IR Satellite Image 09/11/ LT Doppler wind lidar: --> more representative (accurate) wind information --> assimilation ongoing in collaboration with Rolf Langland (NRL) and Carla Cardinali (ECMWF) M. 8

9 Conclusions 1 All models show an improving tendency of typhoon track forecasts Improvement from 20 40% in NCEP and WRF to a comparably low influence in ECMWF and JMA. Likely related to lower track forecast errors without dropsondes in ECMWF & JMA more extensive use of satellite data and 4d Var? The different behaviour of the models emphasizes that the benefit depends strongly on the quality of the first guess field and the assimilation system Conclusions 2 Some improvements of ECMWF mid latitude forecasts BUT improvement results more from improved typhoon track prediction, than observations in SV sensitive areas optimizedfor verificationareasin mid latitudes during ET Cycling necessary for ECMWF forecast improvements. Studies using adjoint forecast sensitivity to observations and lidar observations are ongoing 9

10 Conclusions 3 Separation of dropsonde observations into 3 subsets: typhoon vicinity: iiit largest timprovements of track kforecast remote sensitive regions: small positive to neutral influence typhoon center and core: overall neutral influence Differences in the influence on the typhoon track in the pre and post recurvature stage Small but positive influence of all subsets on the typhoon intensity Future directions More scope for satellite targeting / thinning? Investigate improvement via dropsondes in inner core, as opposed to bogus? Other observation types: HALE and LALE UAS, additional rawinsonde launches? Explore targeting for TC intensity change with mesoscale models? Very large dataset in Atlantic basin. Can conclusions be generalized to other basins? 10

11 Other studies Aberson Global effects of dropwindsondes (see 7.1 tomorrow) Wu, Kim, Hakim, Torn, NCAR MMM EnKF assimilation of T PARC data Reynolds, Langland, Doyle Data denial, observation sensitivity, moist adjoint Majumdar, Wu,, Harnisch ETKF signal variance versus 4d Var / EnKF data impact 11