TC Intensity Estimation: SATellite CONsensus (SATCON) Derrick Herndon, Chris Velden, Tony Wimmers, Tim Olander University of Wisconsin - Madison Cooperative Institute for Meteorological Satellite Studies Jeff Hawkins Naval.
Download ReportTranscript TC Intensity Estimation: SATellite CONsensus (SATCON) Derrick Herndon, Chris Velden, Tony Wimmers, Tim Olander University of Wisconsin - Madison Cooperative Institute for Meteorological Satellite Studies Jeff Hawkins Naval.
TC Intensity Estimation: SATellite CONsensus (SATCON) Derrick Herndon, Chris Velden, Tony Wimmers, Tim Olander University of Wisconsin - Madison Cooperative Institute for Meteorological Satellite Studies Jeff Hawkins Naval Research Laboratory Monterey, CA International Workshop on Tropical Cyclone Analysis and Intensity Honolulu, HI 13-16 April 2011 The support of the research sponsors, the Oceanographer of the Navy through the program office at the PEO C4I&Space/PMW-120, under program element PE-0603207N and the Office of Naval Research under program element PE-0602435N is gratefully acknowledged. Motivation • Contemporary methods to estimate TC intensity can vary by more than 40 knots • Several objective TC intensity methods exist, but the goal of SATCON is to assess the current intensity by combining the confident aspects of the individual objective estimates into a single “best” estimate TCS-08 result showing the spread of 5 expert Dvorak analysts for TY Sinlaku (15W) who were blind to recon obs Recon vs Dvorak for 15W (MSW) 140 130 120 110 100 90 80 70 60 50 40 30 6 :0 0 6 :0 0 8 :0 0 1 2 :0 0 1 8 :0 0 5 :0 0 4 :0 0 1 8 :0 0 7 :0 0 9 -Sep 1 0 -Sep 1 0 -Sep 1 1 -Sep 1 2 -Sep 1 8 -Sep 1 9 -Sep 1 9 -Sep 2 0 -Sep B1 B5 B3 B4 B2 Even taking an average of five expert Dvorak intensity estimates can lead to significant differences. Recon vs Dvorak for 15W (MSW) 140 130 120 110 100 90 80 70 60 50 40 30 6 :0 0 6 :0 0 8 :0 0 1 2 :0 0 1 8 :0 0 5 :0 0 4 :0 0 1 8 :0 0 7 :0 0 9 -Sep 1 0 -Sep 1 0 -Sep 1 1 -Sep 1 2 -Sep 1 8 -Sep 1 9 -Sep 1 9 -Sep 2 0 -Sep B2 B lind M ean Rec on B1 B5 B3 B4 SATCON Members ADT (Advanced Dvorak Technique) Uses IR imagery to objectively assess storm cloud patterns and structure to infer intensity Latest version uses information from MW to make adjustments Clear Eye Pinhole Eye Large Eye Uniform Curved Band Shear SATCON Members: CIMSS AMSU Channel 8 150 mb Channel 7 250 mb Channel 6 350 mb 55 Knots AMSU TbAMSU-A Anomaly cross Chanel 8 Tb vertical Anomaly vs Recon Del-P section for Katrina 2005 TC Pressure Anomaly Magnitude 120 -1 100 70 Knots 80 60 40 20 0 0 1 2 3 4 5 6 7 AMSU Channel 8 Tb Anomaly Magnitude 8 125 knots SATCON Members: CIRA AMSU Similar to CIMSS approach, however the AMSU-A Tb are used to retrieve a temperature profile at 23 pressure levels. Estimates of Vmax are then determined from the thermal warm core structure and non-linear balance equation. IR image from NRL TC Page SATCON Strategy The strengths and weaknesses of each objective method are assessed based on statistical analysis, and that knowledge is used to assign weights to each method in the consensus algorithm based on situational performance to arrive at a single superior intensity estimate Another important component of SATCON is cross-method information sharing • Utilize relationships that exist between the output parameters of the individual member algorithms • Unique information from each of these parameters can be shared between the algorithms to improve the performance of the individual members • Situational corrections can be made to each algorithm’s intensity output, then the member weights re-derived to produce an improved consensus estimate SATCON cross-method information sharing Example: ADT to CIMSS AMSU ADT Estimate of Eye Size In clear eye scenes, IR can be used to estimate eye size Compare to AMSU-A FOV resolution Adjust AMSU pressure if needed CIMSS AMSU uses eye size information to correct resolution sub-sampling SATCON cross-method information sharing Example: Objective estimates of eye size from CIMSS ‘ARCHER’ method (using MW imagery) Currently, CIMSS AMSU method uses IRbased eye size or values from op center if no eye in IR MW imagery (MI) often depicts eyes when IR/ADT cannot ARCHER method (Wimmers and Velden, 2010) uses objective analysis of MI and accounts for eyewall slope ARCHER eye = 33 km Information can be input to CIMSS AMSU method SATCON cross-method information sharing CIMSS AMSU position with bracketing correction can be applied to correct CIRA AMSU estimate CIRA MSW Error Compared to AMSU-B Tb Near TC Center -70 R2 = 0 .1 7 9 1 -60 -50 -30 -20 -10 20 10 0 -10 -20 -30 -40 0 10 20 30 AMS U-B 89 Gh z Tb For FO V Use d for Estim ate -50 -60 -70 MSW Error (knots) -40 SATCON Weighting Scheme Weights are based on situational analysis for each member • Separate weights for MSW and MSLP estimates • Example criteria: scene type (ADT) scan geometry/sub-sampling (AMSU) Example: ADT Scene type vs. performance CDO RMSE 14 knots EYE RMSE 12 knots SHEAR RMSE 18 knots SATCON Weighting Scheme Example: AMSU scan geometry vs. performance A CIRA RMSE 12 knots CIMSS RMSE 10 knots B CIRA RMSE 15 knots CIMSS RMSE 12 knots C CIRA RMSE 18 knots CIMSS RMSE 15 knots Additional SATCON Adjustments • Use ARCHER scores to determine how much motion component to add. A greater component is added to storms with strong well developed eyewalls, and less component for storms with poor inner core structure • Use the statistically superior SATCON MSLP estimate to estimate MSW using TC structure information. Take an average of this P-W derived MSW estimate and the SATCON MSW weighted estimate to get final MSW • ARCHER TC eye size is used to adjust MSW upward for for small eyes and downward for large eyes SATCON Examples ADT determines scene is an EYE scene B CIMSS AMSU: Good near nadir pass. Eye is well-resolved by AMSU resolution CIRA is sub-sampled by FOV offset with TC center SATCON Weighting: ADT = 28 % CIMSS AMSU =47 % CIRA AMSU = 25 % SATCON Examples Center of TS Chris ADT determines situation is a SHEAR scene CIMSS AMSU indicates no sub-sampling present CIRA AMSU: little/no subsampling error due to position offset from FOV center SATCON Weighting: ADT = 18 % CIMSS AMSU =41 % CIRA AMSU = 41 % 1999-2010 SATCON Performance (Vmax) N = 289 CIMSS AMSU CIMSS ADT CIRA AMSU SATCON Dvorak BIAS 0.6 - 2.0 -7.1 - 0.5 - 1.9 AVG ERROR 8.7 10.5 11.7 7.1 7.7 RMSE 11.1 14.3 15.6 8.9 9.9 Cases: ATL = 263 EPAC = 8 WPAC = 18 Independent sample. Values in knots. Validation is Best Track Vmax coincident with aircraft recon +/- 3 hours from estimate time. Negative bias = method was too weak. Important Note: “Dvorak” performance values are derived from a consensus of available estimates (the consensus is usually superior to individual OFC estimates 1999-2010 SATCON Performance (Vmax) WPAC N = 18 CIMSS AMSU CIMSS ADT CIRA AMSU SATCON Dvorak BIAS - 3.7 - 2.0 -7.2 - 1.5 - 4.9 AVG ERROR 7.37 10.5 14.0 8.4 10.8 RMSE 9.0 14.3 17.0 9.9 12.5 Independent sample. Values in knots. Validation is Best Track Vmax coincident with aircraft recon +/- 3 hours from estimate time. Negative bias = method was too weak. Important Note: “Dvorak” performance values are derived from a consensus of available estimates (the consensus is usually superior to individual OFC estimates 1999-2010 SATCON Performance (MSLP) N = 289 CIMSS AMSU CIMSS ADT CIRA AMSU SATCON Dvorak BIAS 0.3 - 2.5 -2.6 0.1 -2.0 AVG ERROR 5.4 8.9 6.8 4.6 6.8 RMSE 7.3 12.5 10.4 6.5 9.3 Cases: ALT = 263 EPAC = 8 WPAC = 18 Independent sample. Values in millibars. Validation is coincident with aircraft recon +/- 3 hours from estimate time. Negative bias = method was too weak. Important Note: “Dvorak” performance values are derived from a consensus of available estimates (the consensus is usually superior to individual OFC estimates 1999-2010 SATCON Performance Comparison to simple ave/consensus (un-weighted) N = 289 SIMPLE MSW SATCON MSW SIMPLE MSLP SATCON MSLP BIAS - 3.0 - 0.3 -1.6 0.1 AVG ERROR 8.1 7.6 5.0 4.6 RMSE 10.5 9.6 7.5 6.5 Cases: ALT = 263 EPAC = 8 WPAC = 18 Independent sample. Validation is coincident with aircraft recon +/- 3 hours from estimate time. Negative bias = method was too weak. SATCON and Dvorak MSW Stats for 1999-2010 Independent SATCON Sample Binned by Category 12 10 8 Knots 6 4 2 N =6 7 N =1 3 9 N =8 3 0 -2 -4 -6 T D-T S C A T 1-2 C A T 3-5 SA T C O N BI A S D V K BI A S SA T C O N A BS E RR D V K A BS E RR SA T C O N RM SE D V K RM SE HURRICANE PALOMA 2008 130 120 110 100 MSW (knots) 90 80 70 60 50 40 30 110607 110618 110622 110707 110710 110718 110721 110807 110810 110818 Date (MMDDHH) s atc on new rec on dvk s atc on_old C I M SS A M SU A DT C I R A A M SU HURRICANE KATRINA 2005 170 160 150 140 130 MSW (knots) 120 110 100 90 80 70 60 50 40 30 082408 082419 082511 082611 082720 082808 082812 082820 082908 082912 Date (MMDDHH) s atc on new rec on dvk s atc on_old C I M SS A M SU A DT C I R A A M SU HURRICANE RITA 2005 160 150 140 130 120 MSW (knots) 110 100 90 80 70 60 50 40 30 20 092320 092312 092220 092209 092120 092011 091908 091819 091810 091808 Date (MMDDHH) s atc on new rec on dvk s atc on_old C I M SS A M SU A DT C I R A A M SU HURRICANE BILL 2009 140 130 120 110 MSW (knots) 100 90 80 70 60 50 40 30 082310 082306 082207 082120 082109 082021 081917 081908 081820 081817 Date (MMDDHH) s atc on new rec on dvk s atc on_old C I M SS A M SU A DT C I R A A M SU HURRICANE RITA 2005 160 150 140 130 120 MSW (knots) 110 100 90 80 70 60 50 40 30 20 092320 092312 092220 092209 092120 092011 091908 091819 091810 091808 Date (MMDDHH) s atc on new rec on dvk s atc on_old C I M SS A M SU A DT C I R A A M SU SATCON Web Site http://cimss.ssec.wisc.edu/tropic2/real-time/satcon References Brueske K. and C. Velden 2003: Satellite-Based Tropical Cyclone Intensity Estimation Using the NOAA-KLM Series Advanced Microwave Sounding Unit (AMSU). Monthly Weather Review Volume 131, Issue 4 (April 2003) pp. 687–697 Demuth J. and M. DeMaria, 2004: Evaluation of Advanced Microwave Sounding Unit TropicalCyclone Intensity and Size Estimation Algorithms. Journal of Applied Meteorology Volume 43, Issue 2 (February 2004) pp. 282–296 Herndon D. and C. Velden, 2004: Upgrades to the UW-CIMSS AMSU-based TC intensity algorithm. Preprints, 26th Conference on Hurricanes and Tropical Meteorology, Miami, FL, Amer. Meteor. Soc., 118-119 Olander T. and C. Velden 2007: The Advanced Dvorak Technique: Continued Development of an Objective Scheme to Estimate Tropical Cyclone Intensity Using Geostationary Infrared Satellite Imagery. Wea. and Forecasting Volume 22, Issue 2 (April 2007) pp. 287–298 Velden C. et al., 2006: The Dvorak Tropical Cyclone Intensity Estimation Technique: A SatelliteBased Method that Has Endured for over 30 Years. Bulletin of the American Meteorological Society Volume 87, Issue 9 (September 2006) pp. 1195–1210 Wimmers, A., and C. Velden, 2010: Objectively Determining the Rotational Center of Tropical Cyclones in Passive Microwave Satellite Imagery. Submitted to JAMC. Analysis of Sat-Based TC Intensity Estimation in the WNP During TCS-08 Comparison of All Satellite-based Estimates – Vmax (Kts) N=14 ‘Blind’ Dvorak Consensus Oper Dvorak Consensus ADT CIMSS SATCON AMSU w/MW (w/Koba) Bias 3.6 2.0 -3.6 2.9 -0.1 Abs Error 9.3 12.0 13.6 8.6 9.0 RMSE 11.9 14.9 17.4 10.1 10.6 Positive Bias indicates method estimates are too strong Analysis of Sat-Based TC Intensity Estimation in the WNP During TCS-08 Comparison of All Satellite-based Estimates – MSLP (mb) N=14 ‘Blind’ Dvorak Consensus Oper Dvorak Consensus ADT CIMSS SATCON AMSU w/MW (w/Koba) Bias 0.7 0.1 -1.0 -1.9 -1.3 Abs Error 5.2 7.5 10.7 4.9 6.0 RMSE 6.6 8.9 12.8 6.3 7.2 Positive Bias indicates method estimates are too strong. 2mem SATCON RMSE= 4.7 Blind and Oper Dvorak conversion is Knaff/Zehr