Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
CONTRIBUTION OF WAVE BREAKING TO QUAD-POLARIZATION
SYNTHETIC APERTURE RADAR
Vladimir Kudryavtsev (1,2), Shengren Fan (3), Biao Zhang(3),
Alexis Mouche (3,4), and Bertrand Chapron (1,4)
(1) Russian State Hydrometeorological University, St. Petersburg, Russia
(2) Marine Hydrophysical Institute, Sebastopol, Russia
(3) Nanjing University of Information Science and Technology, China
(4) Institute Français de Recherche pour l’Exploitation de la Mer, Plouzané, France
IGARSS-2019
Yokohama, Japan, July 28 – August 2, 2019
Sentinel-1
Cosmo-SkyMed
Radarsat-2 TerraSAR-X
Operating SAR are multi-polarized instruments
Since RADARSAT-2, TerraSAR-X and Sentinel-1, quad-pol and dual-pol SAR images provided new information on surface signatures of the ocean-atmosphere processes. Polarization sensitivity has helped to advance our understanding and interpretation of SAR measurements, especially to retrieve highly resolved winds under extreme conditions.
• However, in spite of the great informational capabilities provided by multi-polarization SAR, physics of SAR imaging of the ocean is still poorly understood.
• One of the issues,- is what role of wave breaking and how to describe their impact on co- and cross-pol scattering .
• Significant contribution of wave breaking to dual co-pol radar scattering is already well-established, and confirmed experimentally via remarkable deviation of polarization ratio from Bragg scattering predictions
• Kudryavtsev et al (2013) suggested an effective method to transform dual-pol VV and HH radar signal to “polarized” and “scalar” components describing contribution of short Bragg waves and wave breaking to SAR signal.
Demonstration of efficiency of dual co-pol imagery (Kudryavtsev, Chapron, Myasoedov, Collard & Johannessen, 2013)
slicks Currents signatures
Wind field feature
VV-pol image is very similar to HH-pol
Polarization Difference, PD
Wind field feature
Current signatures
Slick
Non-Polarized Scattering, NP
PG image: slicks are dark, currents are not visible, local wind fields features
detectable
NP image: slicks not visible, currents are well visible, local wind fields features
detectable, but differently
VV and HH transformed to PD = VV – HH (Bragg waves) and NP (wave breaking) become very “different”
As demonstrated, different sensitivity of short Bragg waves and wave breaking to wind, current and surface contaminations opens new opportunity to investigate various ocean phenomena using dual-polarized SAR
In this presentation we further dwell on proposed polarized signal decompositions, to obtain more robust statistical relationships.
The main Goal
To derive empirical dependencies of wave breaking contributions to dual co- and cross-pol C-band radar signals at different wind conditions, incidence angles and azimuths;
To compare suggested empirical dependencies with the semi-empirical models;
To develop C-band quad-pol SAR model where scattering from
regular (non-breaking) surface and breaking waves are discriminated and parameterized.
0
0 0
where
is impact of breaking waves
is 2-scale Bragg scattering
wb
pp
B
pp ppB wb
General Approach: Chapron et al., 1997; Kudryavtsev et al., 2003
Based on idea of decomposition of radar backscatter on two components describing scattering from different surface patches related to: - Regular sea surface, and - Breaking waves
In application to dual VV&HH measurements, s0_wb is non-polarized, and thus polarization ratio (PR) is explicit indicator of wave breaking importance: - Observed PR significantly deviate from PR predicted by Bragg TSM. -This indicates significant contribution of non-pol radar returns to VV&HH
00
0 0
hhhhB wb
vv vvB wb
P
Notations: C-SARMOD: Mouche& Chapron, (2015). C-SARMOD2: Lu, Zhang, Perrie, Mouche, Li, & Wang. (2018). CMODH: Zhang, Mouche, Lu, Perrie, Zhang & Wang. (2019). SAD: Zhang, Perrie, & He. (2011).
APPROACH for Dual Co-Pol Data Analysis: (Kudryavtsev, Chapron, Myasoedov, Collard & Johannessen, 2013)
Decomposition of VV and HH in two other signals describing very different scattering properties of the sea surface:
Resonant scattering from
regular (non-breaking) surface
Non-polarized (NP) radar returns
from breaking waves
SOLUTION of 2 equations for VV and HH
gives
Polarization Difference (PD) is associated with Bragg scattering
from regular surface, described by TSM
Relative Contribution of NP to VV and HH
APPROACH for Cross-Pol Data Analysis: (Kudryavtsev, Kozlov, Chapron, & Johannessen, 2013)
Similar to dual co-pol, co-pol signal (CP) is decomposed in two components describing:
Cross-pol scattering from regular (non-breaking) surface
Cross-pol scattering from breaking crest roughness
residual part is treated as CP scattering from
breaking waves
Relative contribution of breaking waves to CP-signal
Empirical Relations based on 1696 estimates of NP, CPwb, and PD supplemented with in situ data
Dependence on incidence angle and wind speed
- NP decays with incidence much faster than CPwb; - Wind exponents of CPwb and NP (at moderate incidence) corresponds to “whitecaps exponent”, - At low incidence, NP wind exponent is smaller, that probably, confirms its specular reflection origin.
NP
CP
Azimuthal distributions
NP CP
Azimuthal anisotropy of NP is much stronger than CPwb.
Distribution of NP is almost unimodal with max in up-wind direction.
Distribution of CPwb is almost isotropic.
Comparison with Semi-Empirical Models
0
2 2
0 2 4 2
tanexp
cos
wb wb
wb
wb wb
q
R
s s
0
2 2
0 4 2tan 2sin
vh vh
wb wb
vh vv hh wbwb wb
q
G G sB
k
NP scattering mechanism: (Kudryavtsev et al., 2003;2005)
quasi-specular reflection from large-scale, k<d*k_B, breaking crests roughness
CPwb scattering mechanism: (Kudryavtsev et al., 2014)
Resonant scattering from small-scale, k ~ k_B, breaking crests roughness
Sketch of breaking crest roughness spectrum
(Walker, Lyzenga, Ericson & Lund, 1996)
Empiric: red lines; Semi-empiric: black lines; Blue lines: Voronovich&Zavorotny (2001)
Empiric is fitted with Bwb=10^(-2) that by factor 10 exceeds spectrum over regular surface
Short Wind Waves Spectrum retrieval from PD measurements considered as proxy of Bragg scattering
0
4 2 2( )
tan | | (1 )pp pp i
B kG g s
0 2
Angular spreading:
( ) cos2mB B A A
Short Wind Waves Spectrum retrieval from PD measurements which are proxy of Bragg scattering.
Omni-directional spectrum
0 0
* *
4
*
Omnidirectional Spectrum:
( ) 2
Parametrization:
( )
5.7 10
0.62 0.67( )
B
m m m
n
m br br
B br R R
B B d A B
B k u c
n k k k
Notations: Red lines: RS-2 measurements; Blue lines: Hwang & Fois, (2015); Black lines: Elfouhaily et al., (1997); Star-lines: Yurovskaya et al., (2013)
Spectra derived from RS-2 PD data are: consistent with Yurovskaya et al. (2013) spectra measured over breaking-free surface areas, but about by factor 2 below Elfouhaily et al. (1997) and
Hwang&Fois (2015) spectra which probably possess wave breaking roughness contributions
Empirical C-band Quad-Pol Model
* * 0 2
2 3 2 2
10
( ) ( cos2 )
4.5 10 ln
Bn
br br
d
B k u c A A
s k U g
Wave breaking to NP Wave breaking to CP Regular Surface: Spectrum & MSS
Co-Pol TSM
Cross-Pol TSM
Full Co-Pol Model
Full Cross-Pol
Model
Summary Quantitative estimates of wave breaking contributions to dual co- and cross-pol radar signals based on 1696 RADARSAT-2 quad-pol SAR images, co-located with 65 NDBC) buoys are presented.
Contribution of breaking waves to dual co- and cross-pol signals is significant: for VV it varies from 60% to 20% with increasing incidence, whereas for HH and cross-pol it is about 60%-70% for all incidence.
Robust empirical dependencies on impact of breaking waves on co- and cross-pol signals as functions of wind speeds, incidence angles and azimuth are derived.
Polarization difference is good proxy of Bragg scattering, and it is used here to derive short wave spectrum and to suggest its parametrization.
Combination of suggested empirical relations provides development of C-band quad-pol SAR model where scattering from regular (non-breaking) surface and breaking waves are discriminated and parameterized.
Thank you!
Acknowledgements:
V. Kudryavtsev and B. Chapron acknowledge support of the Russian Science Foundation under Project 17-77-30019.
The work of S. Fan and B. Zhang was supported by the National Key Research and
Development Program of China under Grant 2016YFC1401001.