TROPICAL CYCLONE PROBABILITY PRODUCTS LECTURE 1C: WIND PROBABILITY

Transcription

1 TROPICAL CYCLONE PROBABILITY PRODUCTS LECTURE 1C: WIND PROBABILITY Russell L. Elsberry Materials provided by Mark DeMaria and John Knaff Outline What contributes most uncertainty to wind at a point? CIRA Monte Carlo wind probability model Proposed application for setting Conditions of Readiness Bonus Mini-Lecture Preliminary results from TCS08 satellite intensity validation 1st TRCG Technical Forum, Jeju, Korea May 2009

2 NEED TO IMPROVE TRACK FORECASTS FIRST PRIORITY OF PARTICIPANTS AT INTERNATIONAL WORKSHOP ON TROPICAL CYCLONE LANDFALL RELATIVE CONTRIBUTIONS OF TRACK ERRORS, INTENSITY ERRORS, AND STRUCTURE ERRORS TO DeMARIA and KNAFF TRACK PROBABILITY DISTRIBUTIONS TRACK PERTURBATIONS ONLY IWTC-VI November 2006

3 TROPICAL CYCLONE MOTION SIZE PERTURBATIONS ONLY INTENSITY PERTURBATIONS ONLY IWTC-VI November 2006

4 An Improved Wind Probability Program: A Joint Hurricane Testbed Project Update Mark DeMaria and John Knaff, NOAA/NESDIS, Fort Collins, CO Stan Kidder, CIRA/CSU, Fort Collins, CO Patrick Harr, Naval Postgraduate School, Monterey, CA Chris Lauer, NCEP/TPC, Miami, FL Presented at the Interdepartmental Hurricane Conference March 5, 2008

5 Monte Carlo Wind Probability Model Estimates probability of 34, 50 and 64 kt wind to 5 days Implemented at NHC/JTWC for 2006 hurricane season Replaced Hurricane Strike Probabilities 1000 track realizations from random sampling NHC track error distributions Intensity of realizations from random sampling NHC intensity error distributions Special treatment near land Wind radii of realizations from radii CLIPER model and its radii error distributions Serial correlation of errors included Probability at a point from counting number of realizations passing within the wind radii of interest TCC 2009

6 MC Probability Example Hurricane Ike 7 Sept UTC 1000 Track Realizations 64 kt h Cumulative Probabilities TCC 2009

7 Project Tasks 1. Improved Monte Carlo wind probability program by using situation-depending track error distributions Track error depends on Goerss Predicted Consensus Error (GPCE) 2. Improve timeliness by optimization of MC code 3. Update NHC wind speed probability table product Extend from 3 to 5 days Update probability distributions (currently based on )

8 Wind Speed Probability Table

9 Wind Speed Probability Table Developed by E. Rappaport and M. DeMaria as part of original NHC graphical products Limitations addressed by JHT project Based on NHC error statistics Extends only to 3 days Other limitations Does not directly account for land interaction Inconsistent with other probability products from MC model Rick Knabb and Dan Brown suggestion*: Use output from MC model as table input Addresses all of the above limitations Will automatically update when MC model updates *via Dave Thomas

10 Wind Speed Probability Table Evaluation Procedure Examine MC model intensity probability distributions for idealized storms Compare MC intensity probabilities with WSPT values for real forecasts Frances 29 Aug UTC Katrina 24 Aug UTC Katrina 27 Aug UTC Ernesto 29 Aug UTC Ernesto 29 Aug UTC Humberto 12 Sep UTC Humberto 12 Sep UTC Ingrid 13 Sep UTC

11 Wind Speed Probability Table Idealized Storm Cases Straight west track far from land Three cases: Constant max wind of 30, 90 and 150 kt Straight north track close to land Three cases: Constant max wind of 30, 90 and 150 kt

12 MC and Wind Speed Table Probability Comparison Hurricane Frances 29 Aug UTC Distribution of MC and WSPT Table Differences 80 Probability C1 WT C1 MC C2 WT C2 MC C3 WT C3 MC C4-5 WT C4-5 MC Frequency Forecast Interval C1 WT C4-5 MC C4-5 WT C3 MC C3 WT C2 M C C2 WT C1 M C to 5 6 to to to to to to 35 Difference Range Hurricane France Example Nine Forecast Totals

13 Forecast-Dependent Probabilities Operational MC model uses basin-wide track error distributions Can situation-dependent track distributions be utilized? Track plots courtesy of J. Vigh, CSU

14 Goerss Predicted Consensus Error (GPCE) Predicts error of CONU track forecast Consensus of GFDI, AVNI, NGPI, UKMI, GFNI GPCE Input Spread of CONU member track forecasts Initial latitude Initial and forecasted intensity Explains 15-50% of CONU track error variance GPCE estimates radius that contains ~70% of CONU verifying positions at each time

15 Use of GPCE in the MC Model database of GPCE values created by NRL* Are GPCE radii correlated with NHC and JTWC track errors? GPCE designed to predict CONU error How can GPCE values be used in the MC model? MC model uses along/cross track error distributions *Buck: Domestic or Imported?

16 72 hr Atlantic NHC Along Track Error Distributions Stratified by GPCE ( ) Lower GPCE Tercile Upper GPCE Tercile Frequency (%) Along Track Error (nmi)

17 NHC Along and Cross Track Error Standard Deviations Stratified by GPCE ( Atlantic Sample) NHC Along Track Error Standard Deviation NHC Cross Track Error Standard Deviation Standard Deviation (nmi) Lower GPCE Tercile Middle GPCE Tercile Upper GPCE Tercile Forecast Interval (hr) Standard Deviation (nmi) Lower GPCE Tercile Middle GPCE Tercile Upper GPCE Tercile Forecast Interval (hr)

18 MC Model with Track Errors from Upper and Lower GPCE Terciles Lower Tercile Distributions Upper Tercile Distributions Hurricane Frances Sept 00 UTC Example 120 hr Cumulative Probabilities for 64 kt

19 34-kt, 120-h Cumulative Probabilities Current GPCE Differences High Uncertainty Group Low Uncertainty Group Tropical Storm Hanna 5 Sept UTC Hurricane Gustav 30 Aug UTC TCC

20 Brier Score Improvements 2008 GPCE MC Model Test for the Atlantic Brier Score Improvement (%) kt Cumulative 50 kt Cumulative 34 kt Cumulative Brier Score Improvement (%) kt Incremental 50 kt Incremental 34 kt Incremental Forecast Period Forecast Period Cumulative Incremental TCC 2009

21 Summary Code optimization is complete Factor of 6 speed up Wind speed table product input from MC model is a reasonable approach Implementation in 2008 GPCE-dependent MC model is promising Further evaluation needed Real time parallel runs in Aug 2008?

22 Monte Carlo Wind Probability Application: Objective Warning/TC-COR Guidance Goal: Develop an objective hurricane warning scheme based on wind probabilities (Atlantic) Approach: land-threatening Atlantic TCs as development sample Examined 64-kt, 36-h cumulative MC wind probabilities versus NHC hurricane warnings over sample Choose probability thresholds P up = when hurricane warnings issued P down = when hurricane warnings dropped Thresholds chosen by maximizing the fit (by R 2, MAE, averages) of the total distance warned and the total duration of warnings per storm between the scheme and NHC official warnings Imposed condition that scheme could not miss any official warnings TCC

23 Experimental TC-COR Guidance For Atlantic, p up = 8.0%, p down = 0.0% Objective warning scheme verified well with NHC warnings MCP NHC Average Distance Warned per TC (mi) Average Warning Duration per TC (hr) MCP Objective vs. NHC MAE, Distance (mi) / Duration (hr) 65 / 5 E.g. NHC (top) and objective scheme (bottom) warnings for Hurricane Gustav, R 2, Distance 0.94 / 0.74 Used similar methodology to develop similar schemes for TC- COR (64-kt winds at t=24, 36, 60, and 84 h) TCC

24 EXPERIMENTAL TC-COR SETTINGS SITE TC-COR Atsugi 4 Camp Fuji 3 Camp Zama 4 Iwakuni 3 Kadena AB 1 Narita Airport 4 Pusan 3 Sasebo 2 Tokyo 4 Yokosuka 4 Yokota AB 4 Yokohama 4 *** BASED ON JTWC WARNING NR 020 FOR TYPHOON 88W (CORTEST) *** NOTES: TC-COR SETTINGS ARE BASED ON RELATIONSHIP BETWEEN HURRICANE WATCHES/WARNINGS AND 64 KT CUMULATIVE PROBABILITIES IN THE ATLANTIC AND GULF OF MEXICO. THEY ARE OBJECTIVE GUIDANCE FOR ONSET OF 50 KT WINDS AT NAVY INSTALLATIONS. EACH SITE HAS ITS OWN SENSITIVITIES, WHICH THESE TC-COR SETTINGS DO NOT ADDRESS. THE FOLLOWING CUMULATIVE PROBABILITIES ARE USED FOR THE TC-CORR THRESHOLDS: TC-COR4 5% PROBABILITY OF 50 KT AT 72 H TC-COR3 6% PROBABILITY OF 50 KT AT 48 H TC-COR2 8% PROBABILITY OF 50 KT AT 24 H TC-COR1 12% PROBABILITY OF 50 KT AT 12 H TCC 2009 END OF EXPERIMENTAL TC-COR SETTINGS TC-COR2 Threshold same as for NHC Hurricane Warning

25 Future Plans for MC Model Test GPCE version in all basins in 2009 Results on password protected web page Operational transition of GPCE version in 2010 if recommended by NHC Automated coastal watch/warnings (JHT project) Provide landfall intensity and timing distributions (JHT project) TCC 2009

26 References Bessho, K., M. DeMaria, and J.A. Knaff, 2006: Tropical Cyclone Wind Retrievals from the Advanced Microwave Sounder Unit (AMSU): Application to Surface Wind Analysis. J. of Applied Meteorology. 45:3, DeMaria, M., 2009: A simplified dynamical system for tropical cyclone intensity prediction. Mon. Wea. Rev., 137, DeMaria, M., J. A. Knaff, R. Knaff, C. Lauer, C. R. Sampson, and R. T. DeMaria, 2009: A New Method for Estimating Tropical Cyclone Wind Speed Probabilities. Wea. Forecasting, Submitted. Mueller, K.J., M. DeMaria, J.A. Knaff, J.P. Kossin, T.H. Vonder Haar:, 2006: Objective Estimation of Tropical Cyclone Wind Structure from Infrared Satellite Data. Wea Forecasting, 21:6, Schumacher, A.B., M. DeMaria and J.A. Knaff, 2009: Objective Estimation of the 24-Hour Probability of Tropical Cyclone Formation, Wea. Forecasting, 24, Published papers are available at TCC

27 BONUS: TCS08/T-PARC SATELLITE VALIDATION Objective is to validate satellite-based estimates of minimum sea-level pressure and maximum surface winds in western North Pacific Expert satellite analysts selected to do blind satellite estimates during selected periods with aircraft reconnaissance, i.e., they had access only to satellite observations with no knowledge of aircraft or other observations STORM VALIDATION OBSERVATIONS Number Name C-130 P-3 Buoy 13W Nuri W Sinlaku W Hagapit W Jiangmi Best-track team was formed to evaluate all in situ validation observations during these periods Chris Velden research team will then validate satellite estimates 2009 Tropical Cyclone Conference, Ford Island, Hawaii, 29 April-1 May 2009

28 TCS-08 Satellite Cal/Val WC-130J Penetrations: TC Intensity (MSLP & Max Winds): Single Double Triple 13W Nuri 15 W Sinlaku 19 W Jangmi 03 Center Fixes 11 Center Fixes 8 Center Fixes August Time Line September

29 TCS-08 3rd flight into the Pre-TY Nuri (13W) [Harr] Area of ELDORA radar coverage in the next slide 18 August 2008 NRL P-3 flight track WC-130J flight track Planned WC-130 flight track Screen capture of realtime display during aircraft operations

30 TCS-08 Satellite Cal-Val 3 Engines: One WC-130J Nuri penetration then home (Guam)

31 Satellite-Based TC Intensity Estimates Validation Campaign using TCS-08 Recon } } TCS-08 Recon Objective Methods Subjective Methods Typhoon Nuri Max Wind (Kts)

32 Analysis of Sat-Based TC Intensity Estimation in the WNP WC-130J storm center fixes within +/- ~4 hours of corresponding AMSU overpasses Storm yyyymmddhhmm lat lon mslp msw amsu_pass(ddhhmm) 13W N E W N E W N E W N E W N E W N E W N E W N E W N E W N E W N E W N E W N E W N E TCS-08 satellite validation cases were limited!

33 Positive Bias indicates method estimates are too strong Analysis of Sat-Based TC Intensity Estimation in the WNP Comparison of All Satellite-based Estimates Vmax (Kts) N=13 Blind Dvorak Consensus Oper Dvorak Consensus (w/koba) ADT w/mw CIMSS AMSU SATCON Bias Abs Error RMSE

34 Satellite-Based TC Intensity Estimates Validation Campaign using TCS-08 Recon Preliminary Findings (Based on limited sample of 15 recon validation points) Ave Vmax estimate errors (kts): subj Dvorak ~ 11 (blind*), 13 (oper*), 14 (obj-adt*) Subj Dvorak ave error spread (kts): 8-17 (blind-5 analysts), (oper-3 agencies) [JMA (incl their Koba et al. Tnum>Vmax adjustment) superior to other 2 agencies] AMSU* and SATCON* ave errors (kts): Both ~ 9 (subset of 13 validation pts) Blind = No access to real time recon or oper estimates of intensity Oper = Operational fix agencies (JTWC, NESDIS-SAB, JMA) ADT = Advanced Dvorak Technique (UW-CIMSS obj method) AMSU = Advanced Microwave Sounding Unit (UW-CIMSS obj method) SATCON = SATellite CONsensus (UW-CIMSS weighted con. of ADT and AMSU) General Preliminary Conclusions Objective methods are very competitive Significant spread in subjective Dvorak estimates Consensus improves accuracies for all methods