Targeted Observations for Improving Tropical Cyclone Predictability –

1. Targeted Observations for Improving Tropical Cyclone Predictability – DOTSTAR, TH08, TCS-08, and T-PARC Chun-Chieh Wu Acknowledging collaborators in DOTSTAR and T-PARC: Po-Hsiung Lin, Jan-Huey Chen, Shin-Guan Chen, TDRC, COOK (NTU), Chi, Yeh, Wu, Cheng, Chen (CWB), PL Lin (NCU), CH Liu, K.-H. Chou(CCU),Chia Chou (RCEC), AIDC,Sim Aberson (HRD), T. Nakazawa, M. Yamaguchi (JMA/MRI), Sharan Majumdar (U. of Miami), Melinda Peng, Simon Chang, C. Reynolds (NRL), Martin Weissmann (DLR), R. Buizza (ECMWF), Tim Li (IPRC), Mu Mu (LASG, CAS), S. Park (Ewha Univ.), H. M. Kim (Yonsei Univ.), S. Yoden (Kyoto Univ.), D. Parsons, C. Davis, W.C. Lee (NCAR) Grants: NSC, CWB, RCEC/Academia Sinica, ONR

2. Dropwindsonde Observations for Typhoon Surveillance near the Taiwan Region (DOTSTAR, 2003 – 2008) Up to present, 38 missions have been conducted in DOTSTAR for 31 typhoons, with 630 dropwindsondes deployed during the 200 flight hours. 27 typhoons affectingTaiwan 20 typhoons affecting (mainland) China 6 typhoons affecting Japan 3 typhoons affecting Korea 7 typhoons affecting Philippines • Useful real-timedata available to majoroperationalforecast centers • Positive impact to the track forecasts to models in major operation centers (NCEP/GFS, FNMOC/NOGAPS, JMA/GSM) • Targeted observation Wu et al. (2005 BAMS, 2007a JAS, 2007b WF, 2009a,b,c MWR), Chou and Wu (2008 MWR), Chen et al. (2009 MWR), Yamaguchi et al. (2009 MWR), Chou et al. (2009 JGR)

3. NCEP GFS Impact from 2003 to 2008 (All 36 cases) (27) (19) (14) (34) (33) (36) ** * ** ** * ** Paired t-test statistical examination * : statistically significant at the 90% confidence level ** : statistically significant at the 95% confidence level Wu et al. 2009d

4. Since 2003, several objective methods, have been proposed and tested for operational/research surveillance missions in the environment of Atlantic hurricanes conducted by HRD/NOAA(Aberson 2003) and NW Pacific typhoons by DOTSTAR (Wu et al. 2005). NCEP/GFS ensemble variance (collaborating with Aberson) ETKF (collaborating with Majumdar) NOGAPS Singular Vector (collaborating with Reynolds and Peng) Adjoint-Derived Sensitivity Steering Vector (ADSSV) JMA moist Singular Vector (collaborating with Yamaguchi) -- ECMWF Singular Vector Targeted observations in DOTSTAR and T-PARC (Aberson 2003) (Majumdar et al. 2006) (Peng and Reynolds 2006) (Wu et al. 2007b) (Yamaguchi et al. 2007) (Buizza et al. 2006)

5. Impact ofDOTSTAR data: Sensitive analysis result • Sensitive region shows vertically accumulated total energy by the 1st moist singular vector. • Targeted area for the SV calculation is 25N-30N, 120E-130E. × × CONSON’s center position OSE result on CONSON’s (2004) track forecast Wind & Z at 500hPa Red: (I) all dropsonde obs Blue: (II) no dropsonde obs Green: (III) Three dropsonde obs within the sensitive region Light blue: (IV) Six dropsonde obs outside the sensitive region (Yamaguchi et al. 2009, MWR)

6. Inter-comparison of Targeted Observation Guidance for Tropical Cyclones in the Western North Pacific • To highlight the unique dynamic features in affecting the TC tracks, we compare six different targeted techniques based on 84 cases of two-day forecasts of the Northwest Pacific tropical cyclones in 2006. • The six targeted methods: • TESVs form ECMWF, NOGAPS, and EPS of JMA • ETKF • DLM wind variance • ADSSV Chun-Chieh Wu1, Jan-Huey Chen1, Melinda Peng2, Sharan Majumdar3, Carolyn Reynolds2, Sim Aberson4, Munehiko Yamaguchi5, Roberto Buizza6, Shin-Gan Chen1, Tetsuo Nakazawa7 and, Kun-Husan Chou1 Wu et al. (2009b, MWR)

7. Inter-comparison of targeted guidance • Common target locations • Modified Equitable Threat Scores (METS) (Majumdar et al. 2006) • Provides the quantitative measure of how the leading targets of two sets of guidance are similar to each other. Leading targets = 2 % of total grid points in the domain number of common grid points between each pairs of guidance number of leading targets expected number of common grid points between all feasible realizations of guidance Wu et al. (2009b, MWR) # 10 WP04 Ewiniar Ti=20060702 To=20060704

8. Common features: sub-tropical high Wu et al. (2009b, MWR)

9. Common features: mid-latitude trough Wu et al. (2009b, MWR)

10. 11 September, 2009, Typhoon Sinlaku DOTSTAR + Falcon + P3 + C130 Flight tracks Falcon T-PARC P3 DOTSTAR First time with four aircrafts observing typhoons over NW Pacific ocean P3 C130

11. Impact of dropwindsondes to NCEP GFS forecasts of Sinlaku 00 UTC Sept. 10, 2008; 00 UTC Sept. 11, 2008 Degradation due to the inner-core dropsonde data (Aberson 2008) 12 UTC Sept. 11, 2008 (JMA/GSM, from Nakazawa) (Wu et al. 2009e)

12. 50 kts 34 kts After Willoughby et al. (2006) EnKF data assimilation Observations: position, motion vector, axisymmetric structure • 3 hour besttrack data, interpolated to 30 minutes interval by cubic-spline method. • TC radius (34, 50 kts) data from JTWC. • DOTSTAR (Wu et al. 2005, 2007) surface wind data (MBL150, Franklin 2003) on 26 July. 1200 UTC (final time of the initialization period). DOTSTAR flights (Wu et al. 2009f)

13. 3. Experiments on initialization Experiment “TK-MS” Lowest sigma level (Wu et al. 2009f)

14. 4. Experiments on update cycle analysis Experiment “TK-MS-TP-ALL” time Outer eyewall Outer eyewall Inner eyewall Inner eyewall (Wu et al. 2009f)

15. Validation and OSE studies: added value and data assimilation (cost-effective?) Understanding and physical interpretation of the structure of the targeted guidance products (ADSSV, SV, ETKF), along with the PV dynamics Targeted observations of other data (especially the satellite data: satellite thinning) EnKF data assimilation and dynamical analyses Intercomparison of targeted schemes - to gain more insights into the physics of targeted observations Intercomparison of the data impact to different model systems in T-PARC Future perspectives

16. Q: the small reduction in the forecast errors do not justify the cost of the targeted observations. A: I believe this is not the case, at least for the targeting of tropical cyclones. Q: Based on observation sensitivity calculations, although the positive analysis and forecast impact per observation is larger for the targeted observations than for the other types of observations, because of the relatively low number of targeted observations, their overall impact on analysis and forecast quality is very small. A: That may well be the case for several reasons. The volume of lower-quality satellite observations that are assimilated into models has been increasing over the past decade, and they may overwhelm the small number of higher-quality dropwindsonde observations. Also, some data assimilation schemes such as NCEP's GSI produce conservative analysis increments, i.e. they don't let the observation have a significant large-scale influence on the model atmospheric flow. Overall -Sure that we need to better assess the added value of targeted observation for TCs. For TCs, may need to assess the added values of track improvements due to G-IV observations, as well as the statistics for DOTSTAR missions. Given that the benefit of even a small improvement in TC track forecasts is so high, the targeted observations are very cost-effective. For example, in the USA, it is informally reported that it costs $1M to evacuate a mile of coastal population.