Cross-Polarized SAR: A New Potential Technique for Hurricanes Will Perrie and Biao Zhang Bedford Institute of Oceanography Dartmouth, Nova Scotia, Canada
SAR-wind models PR dependence on incidence angle PR vs. incidence angle, from quad-polarization data. Compared to other PR models … PR dependence on incidence angle Figure 2a. Polarization ratio (PR) as a function of incidence angle, from RADARSAT-2 quad-polarization observations and our PR models, as well as selected empirical and theoretical PR models from the literature. We develop a polarization ratio model to improve wind speed retrievals…
SAR PR ocean wind models Nonlinear Least Squares fit …. Coefficient Fitted values A 0.5474 B 0.0333 C -0.0802 PR Model I only incidence angle dependence PR Model II with additional wind speed dependence Coefficient Fitted values P1 0.0012 P2 -0.0162 P3 0.9559 Q1 -0.0006 Q2 -0.0505
Dataset Mouche et al #2 Our #1model Mouche et al #1 downwind Mouche et al #1 crosswind Johnsen et al Thompson et al Comparisons of observed NRCS in VV polarization (x-axis) from RADARSAT-2 measurements with equivalent values (y-axis) calculated from the various empirical and theoretical PR models in the literature: (a) Model 2 from Mouche et al. [2005] (M2005), (b) Model with incidence angle dependence from this study, (c) Model 1 from Mouche et al. [2005] (M2005), (d) Model with from Horstmann et al. [2000] (H2000), (e) Model from Johnsen et al. [2008] (J2008), (f) Model with from Thompson et al. [1998] (T1998), (g) Model with Eq.(5) from Elfouhaily [1996] (E1996), (h) Model with wind speed and incidence angle dependence from this study, and (i) Model with from Vachon and Dobson [2000] (V2000). Elfouhaily Our #2 model Vachon&Dobson PR PR from RADARSAT-2
SAR winds from CMOD5.N and our model#I + R2 HH pol image Dataset 14.1 m/s 12.9 m/s 12.4 m/s 11.6 m/s SAR wind map retrieved by our PR model with only incidence angle dependence and CMOD5.N, from a RADARSAT-2 SAR image as shown in Figure 6a. The colorbar denotes wind speed, in units of m/s. The black arrow indicates wind direction (with 0 to the east). 15.9 m/s 16.6 m/s Zhang, B., W. Perrie, and Y. He (2011), Wind speed from RADARSAT-2 quad-polarization images using a new polarization ratio model, J. Geophys. Res., doi:10.1029/2010JC006522. 17
C-band models Co-Polarization Model: CMOD5.N HH VV HV VH 1:wind speed, wind direction, incidence angle. 2: NRCS_VV saturated under high winds. VH Cross-Polarization Ocean model: C-2PO U10 U10 Correlation coefficient for C-2PO = 0.91 1: only wind speed dependence. 2: NRCS_VH not saturated under high winds.
Dataset R2 Quad-Polarization Ocean Backscatter Measurements NRCS_VV, NRCS_HH depend on incidence angle, wind direction NRCS_VV saturates U10 U10 NRCS_HV, NRCS_VH not sensitive to incidence angle, wind direction no NRCS_HV saturation a)-b) Normalized radar cross section (NRCS) versus wind speed, for six 5 incidence angle bins between 20 and 50 for (a) VV polarization, (b) HH polarization. In (a), include CMOD4, CMOD5, CMOD5.N. c)-d) Normalized radar cross section (NRCS) versus wind speed, for six 5 incidence angle bins between 20 and 50 for (a) HV polarization, (b) VH polarization. The solid line corresponds to a linear fit, with correlation coefficient of 0.97. C-2PO corr coeff = 0.91 U10 U10
Hurricane wind-speed retrievals with C-2PO Hurricane Bertha – 12 July 2008
Hurricane Ike – 10 Sept 2008
Hurricane Bill – 22 Aug 2009
Hurricane Danielle – 28 Aug 2010
RADARSAT-2 dual-polarization SAR image Hurricane Earl on Sep 02, 2010 at 22:59 UTC RADARSAT-2 dual-polarization SAR image acquired over Hurricane Earl at 22:59 UTC on September 2, 2010, (a) VV polarization and (b) VH polarization. VH polarization VV polarization RADARSAT-2 dual-polarization SAR image
Hurricane Earl on Sep 02, 2010 at 22:59 UTC Winds -buoy #41001 is 18.1 (m/s) C-2PO is 16.0 CMOD5.N is 17.4 H*Wind is 16.8 RADARSAT-2 dual-polarization SAR image acquired over Hurricane Earl at 22:59 UTC on September 2, 2010, (a) VV polarization and (b) VH polarization. Colorbar shows sigma-naught in VV polarization () and in VH polarization () in dB, respectively. SAR-retrieved wind speeds from (c) the CMOD5.N model and , with external wind directions from NOAA HRD H*Wind are overlaid, and (d) from the C-2PO model and . Colorbar shows wind speeds at 10-m height () in m/s. H*Wind (m/s) CMOD5.N (m/s) C-2PO model (m/s)
Comparison of C-2PO and CMOD5. SAR wind retrievals eyewall? rain? gradients? Along track-SFMR (hr) time series Comparisons of C-2PO and CMOD5.N SAR-retrieved wind speeds (at 22:59 UTC, on September 2, 2010) with collocated SFMR-measured along-track 10s averaged surface winds during 22:30~23:30 UTC, on September 2, 2010, (a) time series plot, (d) gives SFMR-measured 10s path-integrated rain rates (mm/hr). SFMR-measured 10s rain rates (mm/hr) time series (hr)
Hurricane Earl C-2PO CMOD5.N Comparisons of C-2PO and CMOD5.N SAR-retrieved wind speeds (at 22:59 UTC, on September 2, 2010) with collocated SFMR-measured along-track 10s averaged surface winds during 22:30~23:30 UTC, on September 2, 2010, (b) and (c) scatter plots Comparisons of C-2PO and CMOD5.N SAR-retrieved winds U10 (Sep 02, 2010 at 22:59 UTC) with collocated H*Wind
Hurricane Ike dual-polarization SAR image at 23:56 UTC on Sep 10, 2008 VV polarization VH polarization CMOD5.N + wind directions via H*Wind A RADARSAT-2 dual-polarization SAR image acquired over Hurricane Ike at 23:56 UTC on September 10, 2008, (a) VV polarization and (b) VH polarization. Colorbar shows sigma-naught in VV polarization () and in VH polarization () in dB, respectively. SAR-retrieved wind speeds from (c) the CMOD5.N model and , with external wind directions from NOAA HRD H*Wind are overlaid, and (d) from the C-2PO model and . C-2PO model U10
Hurricane Ike C-2PO CMOD5.N Comparisons of C-2PO and CMOD5.N SAR-retrieved wind speeds (at 23:56 UTC, on September 10, 2008) with collocated H*Wind derived wind speeds at 01:30 UTC, on September 11, 2008. Comparisons of C-2PO and CMOD5.N SAR-retrieved winds U10 (23:56 UTC, on September 10, 2008) with collocated H*Wind CMOD5.N bias of -4.89 m/s and RMS error of 6.51 m/s C-2PO bias of -0.88 m/s and RMS error of 4.47 m/s
Summary C-2PO model presented insensitive to wind direction, radar incidence angle easy mapping of observed cross-pol NRCS to wind speed avoids errors in wind speed retrievals that occur in CMOD5.N in quad-pol data, C-2PO does not seem to saturate potential for hurricane wind retrievals dual-pol Earl: high wind comparisons R2 SAR – airborne SFMR Earl: C-2PO bias= -0.89 m/s, RMSE= 3.23 m/s, CMOD5.N bias= -4.14 m/s, RMSE= 6.24 m/s. Reasons for under-estimates: CMOD5.N is saturated in high winds co- and cross-polarized NRCS calibration error CMOD5.N and C-2PO do not account for rain, high sea states, eye gradients, e.g. dampen the NRCS biases in wind speeds inaccurate wind directions for CMOD5.N.
Some other sources… Paul Hwang et al. TU2.T10.4 (11:20) Wind retrieval with cross-polarized SAR returns Vladimir Zabeline et al. TH2.T02.2 RADARSAT Application in ocean wind measurements Hwang, P. A., B. Zhang, and W. Perrie, 2010: Depolarized radar return for breaking wave measurements and hurricane wind retrieval. Geophys. Res. Lett., 37, L01604, doi:10.1029/2009GL041780. Vachon, P.W. and J. Wolfe, (2010), C-band cross-polarization wind speed retrieval, IEEE Geosci. Remote Sens. Lett., 7, 456-459. Zhang, B., and W. Perrie (2010), C-band Quad-Polarization ocean backscatter measurements: A new polarization ratio model. SEASAR2010, January 25-29, Frascati, Italy. Zhang B., W. Perrie, and Y. He, 2011: Wind speed retrieval from RADARSAT-2 quad-polarization images using a new polarization ratio model. J. Geophys. Res., 116, doi:10.1029/2010JC006522. Zhang, B., W. Perrie (2011), Cross-polarized synthetic aperture radar: a new measurement technique for Hurricanes, under review - Bulletin of the American Meteorological Society (BAMS).