Presentation is loading. Please wait.

Presentation is loading. Please wait.

1 Improvements in Real-Time Tropical Cyclone Products Mark DeMaria NESDIS/StAR/RAMM Branch Presented at.

Similar presentations


Presentation on theme: "1 Improvements in Real-Time Tropical Cyclone Products Mark DeMaria NESDIS/StAR/RAMM Branch Presented at."— Presentation transcript:

1 1 Improvements in Real-Time Tropical Cyclone Products Mark DeMaria NESDIS/StAR/RAMM Branch Presented at

2 2 Outline RAMMB Experimental Tropical Cyclone Web Page Experimental Extended/Updated NESDIS Tropical Cyclone Formation Probability Product Annular Hurricane Index Improvements in the Statistical Hurricane Intensity Prediction Scheme (SHIPS) for the 2007 Hurricane Season

3 3 The RAMMB Experimental Tropical Cyclone Web Page http://rammb.cira.colostate.edu/products/tc_realtime/

4 4 Purpose To display real-time tropical cyclone products created and/or developed by RAMMB and CIRA scientists and products not available elsewhere. Serve these data to the public via a web integrated database designed to accommodate future product development.

5 5 Information Content Globally occurring tropical cyclones Track history and current forecasts from NOAA (NHC or CPHC) or DOD/JTWC Earth relative IR loops (4-km Mercator) Storm relative LEO IR and Vis (1-km Mercator) AMSU-derived radius vs. height cross sections (T and V grad ) and intensity estimate time series Earth relative oceanic heat content + forecast track Multi-platform satellite only tropical cyclone surface wind analysis Digital Dvorak intensity estimate time series Storm relative TPW and water vapor imagery

6 6 Experimental Product Examples (TC wind analysis)

7 7

8 8 Experimental Product Example (oceanic heat content) http://www.aoml.noaa.gov/phod/cyclone/data/ kJ/cm^2 Data provided by G. Goni at NOAA/AOML

9 9 Operational Product Example (AMSU intensity) Available for all active tropical cyclones at ftp://ftpprd.ncep.noaa.gov/pub/data1/amsu

10 10 Example: Subtropical Storm Andrea

11 11 Digital Dvorak Intensity Estimates

12 12 Example Water Vapor Products Storm Relative Total Precipitable Water Storm Relative Water Vapor Imagery Blending of AMSU and SSM/I TPW is explained in Kidder and Jones (2007, JAOT)

13 13 RAMMB Experimental TC Web Page - Future Add New products –STIPS and SHIPS forecast time series –1-km forecast track relative visible imagery –Kinetic energy time series derived from the surface wind analyses Transition (try to) the experimental products –Multi-platform surface wind analysis –Ocean heat content Use in STIPS based consensus WMO website (maybe) Respond to user requests

14 14 Extension of the NESDIS Tropical Cyclone Formation Probability Product to the Central and Western Pacific

15 15 CURRENT TCFP PRODUCT - OVERVIEW Determines the 24-hr probability of TC formation within each 5 ° x 5 ° lat/lon grid box within the product domain (0 ° to 45 ° N, 140 ° to 350 W ° ) Uses the following data sources: –NCEP GFS operational analyses –GOES-East water vapor imagery –NHC Best tracks Running operationally at SSD since 2004 hurricane season. Shows plots of domain-wide formation probabilities, predictor values, and sub-basin time series. Updates every 6 hours. Available at : http://www.ssd.noaa.gov/PS/TROP/genesis.htmhttp://www.ssd.noaa.gov/PS/TROP/genesis.htm

16 16 CURRENT TCFP PRODUCT – OVERVIEW (CONT…) CLIMATOLOGICAL FORMATION PROBABILITY (1949 – 2003) PERCENT LAND DISTANCE TO PREXISTING STORM CLIMATOLOGICAL SST VERTICAL SHEAR (800 – 250 hPa) 850 hPa CIRCULATION VERTICAL INSTABILITY GOES COLD PIXEL COUNT CLOUD-CLEARED WATER VAPOR BRIGHTNESS TEMP. NCEP GFS + GOES-E Water Vapor + Current TC Positions Screening Step Determine Discriminant Value Probability of TC formation

17 17 CURRENT TCFP PRODUCT EXAMPLE – FLORENCE 2006 - 18 hr- 12 hr- 6 hr - 0 hr = Time of TC Genesis

18 18 CURRENT PRODUCT - VERIFICATION Dependent probabilities are skillful by two measures –Brier Skill Score –Relative Operating Characteristic (ROC) score Interannual Variability Expected Yearly # TCs = ∫∫∫ P(x,y,t)dxdydt

19 19 NEW TCFP PRODUCT – EXPANDED DOMAIN Expanded domain longitude eastward to 100 ° E (covers Central & Western N. Pacific Basins) Redefined basins based on satellite coverage Performed separate screening steps and discriminant analysis on each of the 3 basins.

20 20 NEW TCFP PRODUCT - MODIFICATIONS Added 2004 & 2005 to GOES-E sample set Defined screening values and discriminant weights on each of the 3 basins, separately Used a Finer Screening Process Updated screening eliminates 5% of genesis cases and 75% of non-genesis cases (increase in eliminated non-genesis cases of ~5%  improve discriminant analysis) Added of 850 hPa Horizontal Divergence as a Predictor

21 21 NEW TCFP PRODUCT – WEB PAGE Updated/extended TCFP product currently running at CIRA, available at http://rammb.cira.colostate.edu/projects/gparm/ http://rammb.cira.colostate.edu/projects/gparm/ Main Page: points out areas of interest for TC genesis formation, links to web page for each basin Eg. Main Page graphic from 7/8/07 18Z, 18h prior to the formation of WP04 at 11.2N, 140.2 E

22 22 UPDATED TCFP PRODUCT – FUTURE The updated NESDIS TCFP product will be run locally at CIRA and compared to the current product (running at SSD) throughout the 2007 hurricane seasons. Extended product will be evaluated/verified at end of 2007 season. After evaluation, the extended product will be transitioned to operations with the help of SSD

23 SHIPS Annular Hurricane Index

24 Annular Hurricane Characteristics STRUCTURE:  Distinctly axisymmetric  Large circular eyes  Greatly reduced rainband activity  Lasts at least 3 hours  Rare, occur ~4% of the time Isabel (2003) 24 ENVIRONMENTAL CONDITIONS:  Weak Easterly/Southeasterly Wind Shear  Weak Relative Eddy Flux Convergence  200 hPa Easterlies  SSTs in a range 25.4 to 28.6 ° C steady or decreasing. OR  Weak Easterly shear, under an upper ridge, over SST <28.6 ° C ALSO  Intensity > 85 kt (Knaff et al. 2003)

25 Annular Hurricane Intensity Characteristics Do not weaken rapidly after max intensity Intensity is very close to 85% MPI wrt SST Have large intensity biases & larger than normal intensity errors 25

26 Determining Annular Structure Subjective DiagnosticObjective ParameterSource Large eyewall radiusRcRc IR Imagery Warm Eye Δ T eye IR Imagery Vertical shearSHRDNCEP analysis 200 hPa windsU200NCEP analysis Eddy flux convergenceREFCNCEP analysis SST Reynolds SST Intensity NHC Best Track Yellow = Structure Blue = Environment 26

27 Annular Hurricane Index (AHI) 27

28 AHI Step 1: Screening ParameterPrescreening Criterion Intensity< 84 kt RcRc <50 km Δ T eye < 15 o C SHRD>11.3 ms -1 U200 1.5 ms -1 REFC 11 ms -1 day -1 SST 29.1 o C +/- 3 standard deviations from means of AH’s (1995-2003) 976 (54 AH) cases (6h) > 84 kt intensity  241 cases after screening (53 AH) Hit Rate = 100%, False Alarm Rate = 19% 28

29 AHI Step 2: Linear Discriminant Analysis (Overview) Graphical Interpretation of LDA for Case With 2 Predictors (x,y) and 2 Groups DF=0 DF = c 0 + c 1 x + c 2 y Coordinate transformation that provides maximum separation of groups (From www.doe-mbi.ucla.edu) Refs: Wilks (2006), Hennon & Hobgood (MWR, 2003) 29

30 AHI Step 2: Linear Discriminant Analysis (cont…) Only predictors with significant annular vs. non- annular differences in means were used** –SST –U200 – 200 hPa zonal winds –σ c – azimuthal standard deviation of BTs at R c –VAR – variance of azimuthally-averaged BTs from TC center to 600 km –ΔT eye - max difference between R c and any azimuthally-averaged BT at smaller radius ** exceeds 95% confidence level using Student’s T test 30

31 Verification Dependent Years (1995-2003) Independent Years (2004-2006) 13256 746 AH NAH LDA “Y”LDA “N” 6114 07 AH NAH LDA “Y”LDA “N” Hit Rate ~ 87 % FA Rate ~ 6 % Hit Rate ~ 100 % FA Rate ~ 4 % STEP 1: Screening Reduced 941 (54) cases to 241 (53) FAR = 19% STEP 2: STEP 1: Screening Reduced 387 (7) cases to 82 (7) FAR = 19% STEP 2: 31

32 AHI Output & Interpretation If case doesn’t pass screening, AHI set to 0. If screening is passed, LDA function value is linearly scaled to obtain the Annular Hurricane Index, which ranges between 1 & 100. AHI is displayed at the end of the SHIPS model output file. AHI = 0  No annular structure AHI = 1  Worst match to annular structure AHI = 100  Best match to annular structure 32

33 Example: Daniel 2006 7/20/06 0Z, vmax=95 kt AHI = 0 7/21/06 0Z, vmax=120 kt AHI = 50 7/22/06 0Z vmax = 130 kt AHI = 100 33

34 Sample Product Output EP05 (Daniel 2006), 7/22/06 0Z ** *ANNULAR HURRICANE INDEX* *EP EP052006 07/22/2006 00 UTC* ** SCREENING STORM INTENSITY=130kt>84kt ?--->PASSED SCREENING SST=27.3C>24.3°C ?--->PASSED SCREENING SST=27.3C<29.1°C ?--->PASSED SCREENING VERTICAL SHEAR=5.7kt<21.97kt ?--->PASSED SCREENING 200 hPa ZONAL WIND=-13.2kt>-22.94kt ?--->PASSED SCREENING 200 hPa ZONAL WIND=-13.2kt<2.92kt ?--->PASSED SCREENING 200 hPa MOM FLUX CONV=-0.1m/s-day>-9m/s-day ?--->PASSED SCREENING 200 hPa MOM FLUX CONV=-0.1m/s-day<11m/s-day ?--->PASSED SCREENING GOES RAD COLD BR TEMP=106km>50km ?--->PASSED SCREENING GOES EYE-RING BR TEMP=66.2C>15°C ?--->PASSED STORM MAY BE ANNULAR, PASSED SCREENING CALCULATE AHI FROM DISCRIMINANT ANALYSIS ************************************************** * ANNULAR HURRICANE INDEX (AHI) VALUE = 100.* * (AHI = 100. IS BEST MATCH TO ANNULAR STRUCTURE)* * (AHI = 1. IS WORST MATCH TO ANNULAR STRUCTURE)* * (AHI = 0. FOR NO ANNULAR STRUCTURE)* ************************************************** 34

35 Annular Hurricane Index - Future For the period 1995-2006, the AHI algorithm had a hit rate of 96% and a false alarm rate of 4% The AHI will be tested in a real-time operational setting, running concurrently with the SHIPS model, at the National Hurricane Center during the 2007 hurricane season. After the 2007 season, if evaluation of the algorithm is favorable the transition to an operational product will be pursued 35

36 36 2006 Operational SHIPS Intensity Model Statistical-dynamical TC intensity prediction model 16 basic predictors –atmospheric from GFS forecast fields –oceanic from Reynold’s weekly SST –climatology and persistence from ATCF input Satellite input –Ocean heat content from satellite altimetry –Convective parameters from GOES channel 4 Empirical decay equation over land Experimental Logistic Growth Equation (LGE) version also run that uses time stepping procedure

37 37 Operational SHIPS Includes OHC from Satellite Altimetry Averaged Along the Storm Track as a Predictor OHC SST

38 38 OHC Impact on SHIPS Forecasts Atlantic Sample: 3072 SHIPS Model Forecasts 1995-2006

39 39 OHC Impact Much Larger for Intense Atlantic Hurricanes Improvements in Operational SHIPS Forecasts from OHC for all 2003-2005 Cat 5 Hurricanes (from Mainelli et al 2007)

40 40 SHIPS Improvements for 2007 Re-tuning of coefficients based on 2006 cases Refinement of method for calculation of vertical wind shear Add new vortex predictor from NCEP GFS model New code installed on the NCEP IBMs (Mist/Dew) on May 9th

41 41 Improved Shear Calculation SHIPS uses NHC official track for center of shear calculation GFS vortex track can differ from NHC track Shear calculation uses large annulus to compensate (200-800 km)

42 42 Example of NHC and GFS Track Mismatch 96 h Forecast for Frances from 27 Aug 2004 12 UTC 850 hPa 200 hPa G G O O G = GFS position O = NHC Position

43 43 New Shear Algorithm Track location of GFS vortex at 850 hPa –Tracker finds location that maximizes 0 to 600 km symmetric tangential wind –Checks for reasonable translational speeds –Requires minimum cyclonic wind Symmetric circulation subtracted –Starts from outer radius where symmetric circulation is cyclonic –Subtraction radius decreases with height Shear calculation at NHC track position after vortex removed 0-500 km radius rather than 200-800 km annulus

44 44 Example of Vortex Removal 850 hPa 200 hPa Total Wind Symmetric Flow Residual

45 45 A New Predictor for SHIPS GFS vortex removal procedure provides 850 tangential wind 0-600 km average GFS tangential winds found to be significant intensity change predictor Added to 2007 SHIPS

46 46 2007 Model Compared to Operational SHIPS (Re-runs of 2004-2006 seasons) Atlantic East Pacific


Download ppt "1 Improvements in Real-Time Tropical Cyclone Products Mark DeMaria NESDIS/StAR/RAMM Branch Presented at."

Similar presentations


Ads by Google