Development of SAR-Derived Ocean Surface Winds at NOAA/NESDIS Pablo Clemente-Colón, William G. Pichel, NOAA/NESDIS Frank M. Monaldo, Donald R. Thompson The Johns Hopkins University Applied Physics Laboratory Christopher C. Wackerman Veridian COAA Spring Meeting, MARCH 23, 2002, University of Maryland 1
Alaska SAR Demonstration Pre-Operational SAR-Derived Wind Products Routine production of RADARSAT-1 (C-HH) SAR-derived wind maps is in place. Typical data-at-satellite to wind speed product is 5 hours. Alaska SAR Facility provides data in near real-time, typically 2-2.5 hours. Data transfer to SAA and cataloguing takes 20 minutes. Wind speed inversion takes 50 minutes per frame. Multiple image frame delivery slows processing. Potential improvements. Image processing can be reduced to 1-1.5 hours w/ ASF local processing since transmission time can be eliminated. 1 GHz computer can compute a wind product in 10 minutes. Wide swath imagery covers Alaska region. Large quantity of data available (~ 5000 images or ~250 GBytes of data) to evaluate usefulness in a quasi-operational context. Comparison with other wind speed measurements possible: Models, buoys, and scatterometers. 2
Challenges NRCS dependence on wind Speed and direction V-pol well-validated for scatterometers H-pol derived from limited tower measurements Implicit calibration issues with different SAR processing facilities Wind speed inversion needs wind direction (A single NCRS can be associated with different wind speed and direction pairs.) Model directions Global models Regional models Scatterometer directions Linear features in SAR images Rational combination not yet developed Data sources Real-time / research Long term availability 3
Wind Speed and Remote Sensing Altimeter Remote Sensing Methods Linked at the Air/Sea Interface. Nadir Scattering / Emission Off-Nadir Scattering / Emission Scatterometers/ Passive Microwave / SAR /GPS Wind Roughness Scales and amplitude of clean-water roughness dependent on wind speed, the air- sea temperature profiles, underlying ocean wave spectrum. 4
Polarization, Incidence Angle and Wind Speed Dependence H-pol data points from: Unal, C. M. H., P. Snooji, and P. J. F. Stewart, The polarization-dependent relation between radar backscatter from the ocean and and wind vectors at frequencies between 1 and 18 GHz, IEEE Trans. Geosce. Remote Sens., vol. 29,, 621-626, 626, 1991. 5
NRCS vs. Wind Speed and Direction Nominal Form CMOD4 γ σ 0 = a ( θ ) U [1 + b( θ ) cosφ + c( θ ) cos2φ ] σ V γ ( θ ) 1.6 0 = a ( θ ) f ( U ) [1 + b( θ )cosφ + c( θ, U )cos 2φ ] 6
Polarization Ratio ( α tan 2 θ) ( 1 2tan 2 θ) 1+ σ H = σ V U,, 0 2 0 + 2 ( θ φ) Where a is a parameter that is still an area of research. Note that when q =0, the return from vertical and horizontal polarization is the same. α=0, Bragg Scattering α=1, Kirchhoff Scattering α=0.6, Empirical θ= = Incidence Angle φ= = Angle with respect to wind. U= = Wind speed. Polarization ratio is probably dependent on φ and U as well. 7
Wind Speed Processing Flow Read SAR Files Compute Wind Speed SAR Imagery (ASF) Model Fields (MEL) Buoy Wind Speed & Dir. (NDBC) Database at APL/NOAA Register to Geographic Coordinates Read Model Wind Dir. & Buoy Apply Calibration Resample Wind to Geographic Coordinates Create SAR Wind Image Create Images of Fields ACT/WIPE Database Web Page JHU/APL - NOAA Web Page JHU/APL -NOAA Interpolate Model Dir. Image ACT/WIPE Database 8
Merging Wind Directions Global Model Wind Directions Available over the globe Low spatial resolution Scatterometer Wind Directions Higher resolution Same underlying generation as SAR observations SAR Wind Directions Direct observable not displaced in time Errors from other features with the similar structure High Resolution Models Account for local topography May introduce high frequency errors. 9
SAR-NOGAPS Model Comparison ScanSAR Wide near range ADC saturation 10
SAR-Buoy Comparison Include angles > 25 and assume α=0.6. The standard deviation is 1.76 m/s. 11
SAR-QuikSCAT Comparison Incidence > 25 α=0.4 Time diff < 15 min. QuikSCAT directions Std Dev= 1.54 m/s Bias= -0.18 m/s 12
Observation of Coastal/Topographic Wind Effects with QuikSCAT 1999 Dec 22 0611 UTC 13
Observation of Coastal/Topographic Wind Effects with RADARSAT-1 Small arrows represent QuikSCAT wind speed and direction. 1999 Dec 22 0441 UTC 14
Veridian Automated Estimation of Wind Direction From SAR Region to use to estimate a wind vector Region to use to generate a spectrum from the image Form a smoothed spectrum by calculating a spectrum over multiple placements of the smaller region, then averaging the spectra Calculate the elongation direction of the spectral energy over large scales (3-20 km), wind direction is rotated 90 degrees from this direction (red=estimate, white=actual) 15
Wind Direction From SAR Small arrows represent Veridian SAR-derived wind directions with 180º ambiguity. 2001 Dec 13 0343 UTC 16
SAR Coastal Winds SAR has a unique ability to measure the spatial structure of coastal wind fields. Identify important physical processes to be modeled. Identify high-resolution features correlated with common synoptic wind conditions. Plans underway to use SAR winds in a 1-km resolution circulation model in Prince William Sound. SAR-derived winds are beginning to be used in the validation of very high resolution (2 km) atmospheric model in the Gulf of Alaska. Applicable to other environmentally sensitive or otherwise important areas impossible to measure with buoys or conventional spaceborne instrumentation. 17
Atmospheric Gravity Waves 2002 Jan 25 0428 UTC 18
Low Pressure Front 2000 Feb 2 0557 UTC 19
Barrier Jets 2000 Feb 18 0318 UTC 20
West Coast Storm System 2001 Nov 7 0251 UTC 21
Hurricane Danielle 1998 Aug 31 1053 UTC 22
Prince William Sound Wind Forcing 2002 Jan 24 0318 UTC 23
Prince William Sound Wind Direction 2002 Jan 13 0318 UTC 24
Validation of High-Resolution Meteorological Models RadarSAT SAR Wind Map; Prince William Sound, Alaska; 2002 Jan 1 0318 UTC RAMS Model Winds; 2002 Jan 1 0300 UTC (Peter Olsson, Univ. of Alaska) 25
Assimilation into High-Resolution Ocean Circulation Models RadarSAT SAR Wind Map; Prince William Sound, Alaska; 2002 Jan 1 0318 UTC POM Simulation (Mooers et al., 2002) 26
ENVISAT SAR HH/VV 27
Conclusions SAR can provide high resolution surface winds virtually up to the shoreline SAR scatterometry represents a potentially significant new technique for the measurement of ocean-surface wind fields at resolutions more than an order of magnitude finer than possible with any other existing spaceborne, in-situ, or model techniques Is is expected that SAR-derived winds will make an impact on the validation and/or development of both atmospheric and ocean high resolution coastal models. 28
Next Steps Fusion of JHU/APL and Veridian algorithms and wind direction approaches Test applicability of SAR high-resolution winds for validation and assimilation into atmospheric and ocean coastal models High wind speeds improvements with CMOD4HW/CMOD5 ENVISAT ASAR polarimetric analysis to resolve the H/V polarization ratio question Extend automatic wind processing to ENVISAT Validation/Fusion with QuikSCAT, ADEOS II, and WindSAT (Nov/2002) Upcoming Upcoming SAR SAR Missions Missions ENVISAT: ENVISAT: Launched Launched 3/1/2002 3/1/2002 RADARSAT-2: RADARSAT-2: 2003 2003 ALOS: ALOS: Summer Summer 2004 2004 RADARSAT-3: RADARSAT-3: 2006 2006 (?) (?) 29
Development of SAR-Derived Ocean Surface Winds at NOAA/NESDIS Pablo Clemente-Colón, William G. Pichel, NOAA/NESDIS Frank M. Monaldo, Donald R. Thompson The Johns Hopkins University Applied Physics Laboratory Christopher C. Wackerman Veridian MARCH 23, 2002 COAA Spring Meeting, University of Maryland 30