Cyclones. J. Horstmann, S. Falchetti, S. Maresca. NATO Undersea Research Center, Italy UNCLASSIFIED

Transcription

1 SAR Wind Field Retrieval with SAR Wind Field Retrieval Respect to Tropical Cyclones with Respect to Tropical Cyclones funded by ONR Code 32 J. Horstmann, S. Falchetti, S. Maresca NATO Undersea Research Center, Italy

2 Why SAR for Tropical Cyclones QuikSCAT SAR

3 Hurricane eye of Katrina

4 Hurricane eye of Katrina

5 Hurricane eye of Katrina

6 Hurricane eye of Katrina

7 Hurricane eye of Katrina

8 SAR Typhoon Processing System within the ITOP Project of ONR model data SAR processing ght hei n ct io Wa ve di r e Wa ve s ur e n Pre s Wi nd di r e c t io ed s pe Wi nd oc a el Ey h Lon eadi n g g Lat i t ude i t ud & e Lan dm as k Filt er ma sk NR CS bet t er i te ang Sa t el l nc e In c i de NR CS le Reader t i on Descalloping Recalibration Debanding NURC eye GD eye WiSAR APL pressure Model eye GD wind GD waves CSTARS wave

9 General Approach for Ocean SAR Wind Field Retrieval (WiSAR) Local Gradient Method (B2B 4xy )3 Sobel (B2B 4xy ) Φ Geophysical Model Function Binomial filter 2 dim. B2 Filter 2 dim. B4 Filter σ 0pol = a (θ )uγ (θ ) [1 + b(θ ) cos φ + c(θ ) cos(2φ )] Φ θ u10 Optimized Sobel-Filter GMF for C-, Xand L-band σo

10 Wind Direction Ambiguity Removal

11 Wind Direction Ambiguity Removal 1. Select grids with only one wind direction (400m)

12 Wind Direction Ambiguity Removal 2. Select nearest neighbor for the other wind directions

13 Wind Direction Ambiguity Removal Polar wind directions around hypothetic eye with 180 deg ambiguity Limit radius around the eye

14 Wind Direction Ambiguity Removal 5. Retrieve 60% quantile of simulated polar wind at 400 m grid wind

15 Wind Direction Ambiguity Removal 6. Use eye location and polar wind to remove 180 deg ambiguity and wind directions with difference above 60 deg

16 Wind Direction Ambiguity Removal Select nearest neighbor of all scales to previously selected wind directions Smooth wind directions Radarsat Aug. 2007

17 Comparison of SAR to Quikscat winds RR Q LRQ ni w CRUN LFQ RF Q ] s/ m [ dee ps dni w CRUN WiSAR direction GD direction QuikSCAT direction cor 0.69 rms 4.55m/s bias m/s QuikSCAT wind speed [m/s] cor 0.97 rms 16.6 bias 5.7 QuikSCAT wind direction [ ]

18 SAR Wind Speed Error Masks Latitude Non wind phenomenon mask Longitude

19 SAR Wind Speed Error Masks Latitude Out of Definition Mask ] Bd[ S CRN F out of definition no error defined =20, downwind wind speed [m/s] Longitude

20 SAR Wind Speed Error Masks Latitude Uncertainty Mask Longitude

21 Comparison of SAR and SFMR Wind Speeds Wind speed [m/s] Comparison of wind speeds Time delay [h] Comparison of wind speed errors Wind speed [m/s] SAR wind speeds with superimposed adjusted SFMR flight track Time delay [h]

22 Assimilation SAR Winds into HWIND Ra d arsa Typhoon Malakas t-2 S AR NUR C ima g Wind Field e

23 Wind Speed Error and Certainty Masks

24 Assimilation into HWIND HWIND with in situ and SFMR HWIND with SAR wind field solely

25 Comparison of SAR to Quikscat winds SAR wind directions NURC-GD merged cor = 0.96 rms = 17.6 bias = 6.4 HWIND wind directions assimilation of SAR wind field (solely) cor = 0.97 rms = 16.6 bias = 5.7

26 Comparison of SAR to Quikscat winds SAR wind speeds NURC-GD merged HWIND wind speeds assimilation of SAR wind field (solely)

27 TerraSAR-X Winds Compared to ASCAT Results (Priliminary)

28 Inclusion of Doppler Shift Information for SAR Wind Retrieval

29 Summary & Outlook Research Center SAR wind directions are retrieved from wind streaks (rms of 20 ) lack of inflow SAR wind speeds are retrieved from Cband models (rms of ~2 m/s) For extreme winds the wind speed the rms is significantly larger (~5 m/s) SAR wind retrieval scheme is fully automated Validation of X-band model for high winds Inclusion of Doppler shifts for wind direction retrieval

30 Application of APLs Boundary Layer Model PBL U10 model k P 0 swath of pressure gradient vectors swath of pressure gradient vectors face r u s a fit ield f e r u press

31 Input Input Output Output Assimilation into HWIND

32 Comparison of SAR to Quikscat winds SAR wind directions NURC-GD merged cor = 0.96 rms = 17.6 bias = 6.4 HWIND wind directions assimilation of SAR wind field (solely) cor = 0.97 rms = 16.6 bias = 5.7

33 Comparison of SAR to Quikscat winds SAR wind speeds NURC-GD merged HWIND wind speeds assimilation of SAR wind field (solely)

34 Application of APLs Boundary Layer Model SAR wind field Hurricane Ike SAR out of definition mask SAR input to SLP model

35 Application of APLs Boundary Layer Model SLP pressure field Hurricane Ike SLP retrieved wind field SAR retrieved wind field

36 Application of APLs Boundary Layer Model Hurricane Ike

37 Typhoon Malakas, Radarsat-2 Retrieved Wind Fields Ra da r sat-2 S ar ima ge GD W ind Fi eld NUR C Wind Field WindSpeeds(m/s)

38 Typhoon Malakas, Wind Speed Error and Certainty Masks

39 Typhoon Malakas, Assimilation into HWIND HWIND with in situ and SFMR 1 min HWIND with in situ and SFMR 10 min. HWIND with SAR wind field 10 min.

40 Typhoon Malakas, Comparison to Numerical Model Results HWIND with in situ and SFMR 10 min. HWIND with SAR wind field 10 min. ECMWF winds

41 Typhoon Megi, Comparison to ASCAT Results

42 Typhoon Megi, 200m Resolution Wind Field

43 Typhoon Malakas, Assimilation into APLs Boundary Layer Model SAR input wind including masks Boundary Layer Model retrieved wind field Boundary Layer Model retrieved pressure field

44 Publications: Synthetic Aperture Radar Retrieved Winds Assimilated into HWIND (SLP-winds) Estimation of Wind Speed Uncertainties in Synthetic Aperture Radar Wind retrieval Automated SAR Eyefinding in SAR images Synthetic Aperture Radar Wind Retrieval of Tropical Cyclones Descalloping of Synthetic Aperture Radar Images SAR Retrieved Pressure Fields SLP iterated SAR winds The Dave s Paper s Empirical Wave Retrieval Physical Wave Retrieval Comparison or merged SAR wave retrieval Duncan Ross Comparison (Waves) SARTYPS group paper? 14.

45 NURC Summary Run eye finder on entire CSA Hurricane data set promising results Uncertainty estimator comparison to SFMR validation with in situ ongoing Comparison of SAR retrieved winds in Hurricanes to QuikSCAT show to little inflow Assimilation of SAR retrieved winds into HWIND shows improvement of wind field compared to QuikSCAT (e.g. no hourglass) GMF for X-band shows promising results (to be investigated also for moderate winds)

46 Simulated Effect of Wind Speed Ambiguities Hwind + SAR Cluster-Mask Upper solution Lower solution Rebuilt Wind

47 Comparison to Numerical Model Results HWIND with in situ and SFMR 10 min. HWIND with SAR wind field 10 min. ECMWF winds

48 Comparison of SAR to QuikSCAT winds e ps dni W ] s/ m [ dee ps dni w CRUN cor 0.68 rms 5.16 m/s bias 1.76 m/s QuikSCAT wind speed [m/s] pts radi<100km pts radi<200km pts radi<350km Radar look direction with respect to wind direction [deg]