2. NASA UAVSAR GULF OIL SPILL CAMPAIGN
22-23 JUNE 2010 DEPLOYMENT
o 2 days, 21 flight hours
o ~5500 km of flight lines with 22 km swath width
o imaged an area of 120,000 km2
L-Band 1217.5 to 1297.5 MHz
Frequency
(23.8 cm wavelength)
Intrinsic Resolution
1.7 m Slant Range, 0.8 m Azimuth
DWH rig site, photographed from NASA G3
Polarization
HH, HV, VH, VV
Repeat Track
± 5 meters
Accuracy
Transmit Power
> 3.1 kW
Radiometric
1.2 dB absolute, 0.5 dB relative
Calibration
Noise Floor
-47 dB average
IGARSS 2011, 25-29 July 2011, Vancouver, Canada Cathleen Jones (Jet Propulsion Laboratory) 2
3. UAVSAR FLIGHT LINES
COVERING MAIN SLICK OF THE DEEPWATER HORIZON SPILL
Two UAVSAR lines viewing the main
slick from opposite directions were
using in our analysis of the
polarimetric response of the oil from
the DWH spill.
gulfco_32010_10054_101_100623
collected 23-June-2010 21:08 UTC
gulfco_14010_10054_100_100623
collected 23-June-2010 20:42 UTC
Because of the vast extent of the spill,
we are able to measure the radar cross
section as a function of incidence
angle between 26°-65° using this data
set. By looking at different portions
of the slick, we could also quantify
the variability of the returns from oil.
UAVSAR data available at:
Polarimetric L-band Radar Signatures of Oil from the Deepwater Horizon Spill www.asf.alaska.edu
Brent Minchew, Cathleen Jones, Ben Holt, submitted to TGARS uavsar.jpl.nasa.gov
IGARSS 2011, 25-29 July 2011, Vancouver, Canada Cathleen Jones (Jet Propulsion Laboratory) 3
4. SURFACE CONDITIONS
JUNE 23, 2010
Photographs of the DWH spill site on 6/23/2010
Surface conditions:
sea state 1.0-1.3 m SWH
winds 2.5-5 m/s from 115°-126°
Photos from NOAA RAT-Helo, EPA ASPECT, USF
IGARSS 2011, 25-29 July 2011, Vancouver, Canada Cathleen Jones (Jet Propulsion Laboratory) 4
5. BRAGG SCATTERING THEORY
WAVE FACET MODEL
Radar backscatter from the ocean surface is dominated by scattering from small scale capillary
and gravity-capillary waves that roughen the surface. In Bragg scattering theory, the dominant
mechanism is resonant backscatter from surface waves of wave number kBragg where
As the incidence angle increases, the wavelength of
kBragg = 2k sin(" inc ) the Bragg surface wave decreases to a minimum of
k = 2# λradar/2 at grazing angles.
$radar
L-band (λradar=23.8 cm) : λBragg = 23.8 cm (30°), 13.7 cm (60°)
2
' sin($ + % )cos & *2 ' sin & *2
"HH = 4 #k 4 cos!($ i )W(2k sin($ + % ),2k cos($ + % )sin &) )
4
, RHH +) , R
( sin $ i + ( sin $ i + VV
Ocean wave spectral density at Bragg wavelength
2
' sin($ + % )cos & *2 ' sin & *2
"VV = 4 #k 4 cos 4 ($ i )W(2k sin($ + % ),2k cos($ + % )sin &) ) , RVV + ) , RHH
( sin $ i + ( sin $ i +
2
"HV = 4 #k 4 cos 4 ($ i )W(2k sin($ + % ),2k cos($ + % )sin &) RVV - RHH
! i = cos!1[cos(! + " )cos(# )]
! (! r !1)(! r (1+ sin 2 (! i )) ! sin 2 (! i )) ! r !1
RVV = 2
RHH = 2
(! cos(! ) +
r i
2
! r ! sin (! i )) ) (cos(! ) +
i
2
! r ! sin (! i )) ) " =out-of-plane tilt angle
IGARSS 2011, 25-29 July 2011, Vancouver, Canada Cathleen Jones (Jet Propulsion Laboratory) 5
!
6. EFFECT OF SURFACE LAYER OF OIL
ON RADAR BACKSCATTER FROM WATER
Oil damps the small-scale capillary and gravity-capillary waves on the ocean surface mainly
through a reduction in the surface tension at the gas-liquid interface.
gravity is the restoring force
Dispersion relationship for waves at the interface
between air and a liquid of density ρ with surface tension σ: " 2 = gk + (# $ )k 3
surface tension and inertia are
ρoil/ρwater ≈ 0.8 - 0.9 the restoring forces
σoil/σwater ≈ 0.25 - 0.5 g "
! v phase = + k for a given velocity, k
k # increases when the
surface tension decreases
Ocean waves are excited by resonant forcing in a
turbulent wind field. The wavelength of capillary waves
!
resonantly excited in the presence of oil is smaller than
for a clean water-air interface, hence the damping of the
smaller wavelengths. This affects the roughness scale of
the water surface. In a real slick, the surface
characteristics will vary between pure H20 and pure oil,
depending upon layer thickness, oil type, and areal
coverage.
Also, in viscoelastic fluids gravity waves with short
wavelength are damped by restoring forces arising from
gradients in the surface tension (Marongoni effect).
IGARSS 2011, 25-29 July 2011, Vancouver, Canada Cathleen Jones (Jet Propulsion Laboratory) 6
7. POLARIMETRIC DECOMPOSITION
ENTROPY/ANISOTROPY/ALPHA
We have applied two polarimetric decompositions to the oil spill data, the Cloude-Pottier
decomposition and the Shannon decomposition.
Cloude-Pottier:
Scattering Matrix is expressed in the Pauli basis:
" SHH SHV % Pauli 1 T
$ ' ( ( () k = [SHH + SVV SHH * SVV 2SHV ]
# SVH SVV & 2
Diagonalization of the coherency matrix T=kk* gives 3 eigenvalues, , and eigenvectors, u.
Consider 4 variables derived from the eigenvalues and eigenvectors:
3# & # &
"i "i
Entropy: H = )% (Log3% ( 0 * H *1
$ "1 + "2 + "3 ' $ "1 + "2 + "3 '
i=1
" + "3
Anisotropy: A = 2 0 * A *1
"2 + "3
Mean angle: , (u)
3# "2i
&
Averaged intensity: - = )% (
% "1 + "2 + "3 (
i=1$ '
IGARSS 2011, 25-29 July 2011, Vancouver, Canada Cathleen Jones (Jet Propulsion Laboratory) 7
!
8. POLARIMETRIC DECOMPOSITION
SHANNON ENTROPY
Shannon Entropy:
We derived the Shannon entropy parameters from the Pauli basis vectors and the coherency
matrix:
1 T
k=
2
[SHH + SVV SHH " SVV 2SHV ]
T = kk *
!
SE = # PDF(k) = log(PDF(k)dk $ SE intensity + SE polarization
& %eTrace(T) )
SE intensity = 3log( +
' 3 *
& 27Det(T) )
SE polarization = log( +
' Trace(T) *
This decomposition is significantly less computationally intensive than the H/A/ /
decomposition since it requires calculation of only the trace and determinant of the coherency
matrix.
!
C.E. Shannon, Bell Systems Technical Journal, 27, 1948; Refregier and Morio, J. Opt. Soc. of America A, 23(12), 2006
IGARSS 2011, 25-29 July 2011, Vancouver, Canada Cathleen Jones (Jet Propulsion Laboratory) 8
9. AVERAGED INTENSITY
OVER THE DWH SLICK
Averaged Intensity Images
320° Heading 140° Heading
In the following slides, for each UAVSAR
line the parameters are averaged in the
Clean Water
along track direction and plotted as a
wind
function of incidence angle for a clean water
region and for three strips within the slick.
scene overlap
common point
Oil 1
MAIN SLICK
Oil 4
Oil 2
Oil 5
Oil 6
Oil 3
Clean Water + rig site
5 km
IGARSS 2011, 25-29 July 2011, Vancouver, Canada Cathleen Jones (Jet Propulsion Laboratory) 9
13. DETAILS OF THE OIL SLICK
VARIATIONS IN THE AVERAGED INTENSITY
NOT ONLY IS THE OIL SLICK CLEARLY DIFFERENTIATED FROM THE SURROUNDING WATER (DARK BLUE IN THE UAVSAR
IMAGE), BUT THE LOW NOISE UAVSAR RADAR BACKSCATTER CAN DIFFERENTIATE SOME OIL CHARACTERISTICS WITHIN
THE SLICK.
16 km
N
(iii)
(i)
(i) (iii)
(ii)
DWH rig site wind
(iv)
(ii) (iv)
C. Jones, B. Holt, S. Hensley (JPL/Caltech) Photos taken over the slick on 6/23/2010
B. Minchew (Caltech), Studies of the Deepwater N between 16:00 and 20:00 UTC (NOAA
Horizon Oil Spill with the UAVSAR Radar, RAT-Helo and EPA/ASPECT)
submitted to AGU monographs wind
IGARSS 2011, 25-29 July 2011, Vancouver, Canada Cathleen Jones (Jet Propulsion Laboratory) 13
14. CLOUDE-POTTIER DECOMPOSITION
WEATHERED OIL IN BARATARIA BAY
22 km Large amounts of oil
moved far into Barataria
Bay in SE Louisiana on
16-17 June 2010, with oil
remaining in the area until
after the UAVSAR over-
flight.
Weathered oil in the
interior of Barataria Bay
shows a significantly
higher entropy than oil
around the rig site or in
the Gulf of Mexico
approaching the Louisiana
shoreline.
C. Jones, B. Holt, S. Hensley (JPL/Caltech), B. Minchew (Caltech), Studies of the Deepwater Horizon Oil Spill with the UAVSAR Radar, submitted to AGU monographs
IGARSS 2011, 25-29 July 2011, Vancouver, Canada Cathleen Jones (Jet Propulsion Laboratory) 14
15. CONCLUSIONS
• Radar returns of all polarizations discriminate oil in the DWH slick from clean water in
an adjacent area, for incidence angles from 26° to 65°.
• The HV returns showed greatest sensitivity to variations in the oil within the slick,
although low signal level limited its usefulness to incidence angles <~60° for UAVSAR.
• Both the CP and Shannon polarimetric decompositions can be used to discriminate oil
from clean water.
• The CP averaged intensity and Shannon intensity are the best discriminators across the
entire incidence angle range, but the other polarimetric parameters also have regions
where they show a statistically significant difference between oil and clean water and
between different areas of oil with the slick itself.
• The Shannon entropy decomposition is a useful, computationally efficient algorithm for
oil slick detection.
• The entropy of relatively fresh oil-on-water within the main DWH slick is significantly
lower than for the weathered oil in Barataria Bay.
• UAVSAR results indicate that it could be possible to discriminate varying oil properties
within slicks under all-weather conditions given a sufficiently low noise radar
instrument.
IGARSS 2011, 25-29 July 2011, Vancouver, Canada Cathleen Jones (Jet Propulsion Laboratory) 15
16. Back-up Slides
Cathleen Jones1, Brent Minchew2,
Benjamin Holt1
1JetPropulsion Laboratory/California
Institute of Technology
2California Institute of Technology
17. UAVSAR INSTRUMENT
NOISE FLOOR
Comparison with other RADAR instruments
UAVSAR NOISE FLOOR
noise equivalent σ0 (dB)
The low noise floor of the UAVSAR
instrument makes it possible to
measure the radar cross section from
water with an L-band radar, even with
oil damping the surface waves. We
find that the instrument noise floor is
reached only at the far edge of the
swath for the HV returns from oil.
C. Jones, B. Holt, S. Hensley (JPL/Caltech), B. Minchew (Caltech), Studies of the Deepwater Horizon Oil Spill with the UAVSAR Radar, submitted to AGU monographs
IGARSS 2011, 25-29 July 2011, Vancouver, Canada Cathleen Jones (Jet Propulsion Laboratory) 17