Polarized remote Sensing of Surfaces, Aerosols and Clouds: Lessons learnt from POLDER and Parasol - François-Marie Bréon Laboratoire des Sciences ...
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Polarized remote Sensing of Surfaces,
Aerosols and Clouds:
Lessons learnt from POLDER and Parasol
François-Marie Bréon
Laboratoire des Sciences du Climat et de lEnvironnement
breon@lsce.ipsl.fr
1/38
What is polarization ?
Light is an electromagnetic wave.
Usually, the electric field is randomly oriented.
It may get preferentially oriented after a scattering or reflexion process
In such case, the light is said “polarized”
A polarizer let through the light waves that have the electric field
parallel to the polarizer direction
Obviously, my screen generates fully polarized light !
2/38
Natural processes that polarize light
Molecular scattering polarize sunlight, depending on the scattering
angle. Skylight is highly polarized !
Specular reflexion generates polarization depending on the
incidence angle. Full polarization for incidence at the Brewster
angle (≈50°)
Scattering by aerosols and clouds generate polarized light
depending on aerosol and droplet characteristics.
The measurement of polarization provides additional information to
identify the presence and type of scattering particles in the
atmosphere
3/38
Overview of the POLDER-Parasol saga
POLDER-1 onboard the ADEOS platform (1996-1997). Solar panel failure
terminated operations (8 months of continuous acquisition)
POLDER-2 onboard the ADEOS-2 platform (2003). Solar panel failure
terminated operations (7 months of continuous acquisition)
Parasol onboard a CNES microsat satellite. Fly in formation with the Aqua
Train. Near-continuous acquisition since March 2005 (8+ year worth of
data). Lack of fuel=> left the A-Train in 2010. Drift of local time
Eight spectral bands from 440 (blue) to 910 (near IR) nm
Bi-dimensional CCD matrix. ≈2400x1800 km IFOV
Spatial Resolution ≈6 km
Three “Polarized bands”: 440/490, 670, 865 nm.
4/38
POLDER Multidirectional acquisition
Multispectral measurements
acquired every 20 sec
14 POLDER views
Phase
Nadir angles
Principal Plane
View Angles
Perpendicular Plane
5/38
A microsat is fun to work with…
• Parasol pointing 15° forward
• Sees Earth limb
• Allowsto quantify directional signatures
from Nadir to Limb
• Work to be done [operational processing
of Level-0 data does not work for such
configuration…]
6/38
Parasol polarization measurements
Three successive measurements with
polarizer turned by step of 60°
Inversion of radiometric model yield linear
polarization parameters [I, Q, U]
Three spectral bands 490, 670, 865 nm
Rough correction for molecular scattering. Three colour composites 7/38
A few examples of Directional/
Spectral / Polarized signatures
3 color composites in Directional
total light signatures.
443-670-865 nm Total light Location of swath
and image
3 color composites in Directional
polarized light signatures.
443-670-865 nm Polarized light
Directional sampling
Simple formula for View angle
polarized reflectance Phase or Glint angle 8/38
Sunglint over Mediterranean Sea
θs=36°
'π − γ *
FP ) ,
( 2 + tan 2 β
RP (θ s,θ v ,ϕ ) = 2 4
exp(− 2
) Wave slope
4 cosθ s cosθ v σ cos (β ) σ
σ 2 = 0.003 + 5.12 10−3 WS 9/38
Vegetation (Amazonian)
(π − γ +
FP * -
) 2 ,
RP (θ s,θ v ,ϕ ) ≈
4 (cos θ s + cos θ v )
10/38
€Desert
(π − γ +
FP * -
) 2 ,
RP (θ s,θ v ,ϕ ) ≈ Overestimate at large
4 cosθ s cosθ v
phase angles 11/38Aerosols over the Ganges Valley
PP (γ )
RP (θ s,θ v ,ϕ ) ≈
4 (cos θ s + cos θ v )
(1− exp(−m τ a ))
12/38Biomass Burning Aerosol
13/38Antarctic + Cloud Bow
14/38Low Cloud and High Cloud
Goloub Ph et al., JGR, 2000
Riedi J. et al., GRL, 2000Surface BPDFs We have analyzed polarization measurements and derived typical models for the polarized reflectance of land surfaces. BPDF characteristics are very different than those of BRDFs Minimum at backscatter. Increases with phase angle Varies from ≈0 to a few percent Forest < crops < bare soil < snow Appears Spectrally neutral Generated by specular reflection Did not find any useful information about the surface that can be derived from polarization, and that cannot be obtained more easily Maignan et al., Rem. Sens. Env., 2010
Cloud Droplet Radius
from Multidirectional polarisation measurements
17/38Liquid phase clouds
Stratocumulus Scattering angle
cloud field
Baja
California
Same scene in polarized light
Sun Zenith
3-color composite 443-670-865 nm θs
Satellite
θv
In some cases, clouds fields show specific
features in polarized light for scattering Scattering Angle
angles between 140 and 170°
18/38Many such examples...
The position of color bands relative to the scattering angle is
variable !
19/38Droplet polarized phase function
2
Size distribution : σ
eff
= 0.01
= 0.02
! r $ −3 ) ! ! $ $,
σ
eff
Size distribution n(r)
1.5
= 0.05
1 # ln # r r σ
exp ++ &. eff
n( r) = ## &
& # #" r
&−
& r + 1& .
σ
eff
= 0.10
" reff % * σ eff " eff % eff %- 1 r
eff
= 9 microns
Rainbow 0.5
Spectral signature : 0
0 2 4 6 8 10 12 14
0.3 Droplet Radius (µm)
0.25 0.86 µm
0.67 µm Dampening for wider size distributions:
Polarized Phase Function
0.44 µm
0.2 0.3
0.25 σ = 0.01
0.15 λ = 0.67 µm eff
r e f f = 9 µm r eff = 9 µm
Polarized Phase Function
0.2 σ = 0.02
eff
0.1 = 0.05
σe f f = 0.02 0.15
σ
eff
σ = 0.10
0.05 0.1 eff
0.05
0
0
-0.05 -0.05
-0.1
-0.1 140 145 150 155 160 165 170
140 145 150 155 160 165 170 Scattering Angle
Scattering Angle
The polarized phase function shows oscillations that explain the observed
features. Angular position of maxima and minima depend on wavelength and
effective radius. Such feature require a narrow size distribution.
Polarized reflectance mostly generated by single scattering 20/38
Bréon FM and Ph Goloub, GRL, 1998Measurement-model fit
5.5 µm 7.5 µm
Polarized Reflectance
13 µm
443nm
670nm
10 µm 865nm
Scattering Angle [°] 21/38
Bréon FM and M. Doutriaux Boucher, IEEE TGARS, 2005Retrieved spatial distributions of CDR
April August
June October
6 10 14
Poor sampling because of limitations on viewing geometry, extended cloud field
and narrow size distribution.
Shows smaller droplets over continents, and in particular polluted areas
22/38
Bréon FM and S. Colzy, GRL, 2000Comparison with MODIS
Excellent correlation over the Oceans
Poor correlations for small droplets [in particular found over land surfaces]
Bias of 2 µm (POLDER < MODIS).
- CDR at the very cloud are smaller than deeper in the cloud ?
- Spatial heterogeneity ?
- Size distribution different than assumed ?
23/3824/56
CDR with polarization. Conclusions
Multidirectional polarization provides an alternative (to spectral)
method for the estimate of CDR
Advantage of polarization : Measures the single scattering, which
provides a near-direct measurement of the scattering phase
function
Requires specific conditions (geometry and cloud properties) so
that its statistic is poor
Measurements have shown that the CDR distribution is not as
expected, in particular over stratocumulus clouds.
Bias with MODIS. Several hypothesis. My best hypothesis is that
evaporation at cloud top makes droplets smaller than deeper into
cloud. Polarization sensitive to the very cloud top.
Still not clear why MODIS shows large spatial variability in CDR
when Parasol retrieval indicates a more homogeneous CDR field
25/38Atmospheric Aerosols
from Multidirectional polarisation measurements
26/38Aerosols over the oceans:
Information content
0.008
0.008
0.008
0.01
0.01
0.01
MISR
POLDER
MODIS
METEOSAT
AVHRR
Reflectance
0.006
0.006
0.006
Reflectance
Reflectance
Reflectance
Polarized Reflectance
0.008
0.008
0.008
One channel
Multi-directional
One direction
Two measurements
0.004
0.004
0.004
0.006
0.006
0.006 Many channels
channels
One direction
No constrain on
0.002
0.002
0.002
Polarized
0.004
0.004
0.004
aerosol model
0.002
0.002
0.002 000
000
80
80
80 100
100
100 120
120
120 140
140
140 160
160
160 180
180
-0.002
-0.002
180
-0.002 Multi-directional
measurements
Scattering
Scattering Angle
Angle
Scattering
Scattering
ScatteringAngle
Scattering Angle
Angle
Angle
+ Polarization
27/38Aerosol Inversion over the oceans
0.012 0.01
0.014 0.01
sbrepacn5
0.01 0.012
sbreatls1
0.008
0.008
Large to
Polarized reflectance
0.01
Total reflectance
0.008
Polarized reflectance
Total reflectance
0.006
small
0.008 0.006
0.006
0.004 0.006
0.004
particles 0.004
0.004
0.002
0.002 0.002
0.002
0 0
100 110 120 130 140 150 160 170 0 0
110 120 130 140 150 160 170
Scattering Angle
Scattering Angle
0.012 0.025 0.01
0.03
sbrechin1 sbreatln1
0.01 0.008
0.025 0.02
Polarized reflectance
Polarized reflectance
0.008
Total reflectance
Total reflectance
0.006
0.02
0.015
0.006
0.004
0.015
0.01
0.004
0.01 0.002
0.002 0.005
0.005 0
0
0 0 -0.002
90 100 110 120 130 140 150 160 170 100 110 120 130 140 150 160 170 180
Spectral effect increases
Scattering Angle Scattering Angle
150° arc decreases Deuzé JL. et al., GRL, 1999
Deuzé JL. et al., JGR, 2000Main Parasol products over the oceans
Sept. 2005
Total Optical Thickness Angström Coefficient
Fine Mode Optical Thickness Coarse Mode Optical Thickness
+ Effective radii, info on scattering phase function and quality indices
29/38Identification of non spherical particles
0.014 0.01 0.025 0.01
0.012
sbreatls1 sbreatlnx
0.008
0.008 0.02
0.01
Polarized reflectance
Polarized reflectance
Total reflectance
Total reflectance
0.006
0.006 0.015
0.008
0.004
0.006
0.004 0.01
0.002
0.004
0.002 0.005
0.002 0
0 0 0 -0.002
110 120 130 140 150 160 170 100 110 120 130 140 150 160 170 180
Scattering Angle Scattering Angle
150° arc indicates the presence of Small spectral effect but no Arc:
large, spherical particles Non spherical particles.
Fraction of non-spherical
particles in coarse mode
Herman M. et al., JGR, 2005 30/38Aerosol non-sphericity (2/2)
Total Optical Thickness Coarse mode Non-Spherical Optical Thickness
Fraction of non spherical in Coarse Mode Coarse mode Spherical Optical Thickness
Non spherical particles downwind of dust sources. Spherical particles
where hydrated, sea-salts are expected. 31/38Over land…
Why is polarization so useful for aerosol remote sensing over land ?
Rayleigh Aerosols Surface
Lmeas
p ( λ, θ s ,θ v , ϕ ,z pixel ) = Laer + mol
p (mod,δ a ) + Lsol
p × e − m( c mδ m + c aδ a )
• is small compared to atmospheric contribution
sol • is spectrally neutral
€ L p • is rather uniform (varies little with surface type)
• can be roughly estimated from surface classification
32/38Variability of surf. polarized reflectance
Surf. Pol. Reflectance @865 nm
Scattering angle [°]
Measurements show that the surface polarized reflectance is little
variable, even over rather heterogeneous surfaces
Fan et al., 2007
33/38Aerosol reflectance is highly polarized
AEROSOL
Clear atmosphere (AOT=0.03) : the Hazy atmosphere : large aerosol
reflectance at TOA is close to the contribution, 1.0×10-2 at 110°-120° for
surface values AOT=0.31
Illustration for Biomass Burning Aerosols
Deuzé et al., 2001
34/38Aerosol inversion over land
Based on the hypothesis that surface polarized reflectance is small and varies
little
Parameterization of the surface polarized reflectance (semi empirical model as a
function of surface type and NDVI; Nadal&Bréon, 1998)
Model + optical thickness estimate based on measured polarized reflectance at
670 and 865 nm.
Works well for “small” aerosol (sulfates, biomass burning) over vegetated areas
BUT…
Does not work for coarse aerosols (desert dust)
Does not work over desert or snow due to their larger polarized reflectance
reflectance
Polarized
Deuzé JL et al., JGR, 2001 35/56Fine mode optical thickness
Fine mode optical Thickness 550 nm
Aerosol load by sub-micronic particles (fine mode)
Over land: based on multi-directional polarized measurements
Over the ocean : Uses both reflectance and polarized meas.
Note annual cycle of biomass burning activity, pollution over China,
Galapagos volcanic eruption late October 2005 36/38Parasol-MODIS comparison. 2: JJA 2006
MODIS Parasol
Fine mode AOD
Tanre et al. 2008
Total AOD
37/38Parasol-MODIS comparison. 3: DJF 2007
MODIS Parasol
Tanre et al. 2008 38/38Validation against Aeronet
POLDER MODIS MERIS SEVIRI CALIPSO
Ocean
Total
Fine Mode
Ocean
Total
Land
Fine Mode
Land
Bréon et al., RSE, 2011 39/38Parasol concept for aerosol : Pros and con Multispectral + multidirectional + Polarization measurements provide a lot of constrains for the aerosol model retrieval Depending on the target location with respect to the satellite, the range of scattering angle varies Multidirectional acquisition reduces the glint issue. Spatial resolution of POLDER. Limits daily coverage in the presence of broken clouds (a problem in the tropics, in particular for Parasol afternoon views). Surface contribution to the measured polarized reflectance. Although small, the surface contribution is not fully negligible A longer wavelength polarized channel would help constraining the surface polarization contribution. Airborne measurements indicate that the surface polarized reflectance is spectrally neutral. => Was to be done by Glory mission 40/38
Aerosol Effect on cloud droplet size
Optical Thickness
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
12
Cloud Droplet Radius [µm]
Droplet radius
11
10
9
8
Land
7
Ocean
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
Aerosol Index
Aerosol load
POLDER measurements of CDR and
aerosol load
Smaller droplets in polluted areas
Statistics on cloud albedo is lacking
Demonstration of the first step of the
so-called Twomey hypothesis, i.e. the
indirect aerosol effect on the Earth
radiative budget
Bréon FM et al, Science, 2002Conclusions and Prospects
The series of POLDER instruments, onboard ADEOS-1, ADEOS-2
and PARASOL, provided the only measurements from space of
the polarization state of the reflected sunlight
Polarization provides unique information to retrieve information
about aerosol and clouds
The GLORY mission has been developed by NASA to measured
aerosols and CDR, on the basis developed using POLDER
observations, but the launch failed and the satellite was lost
Thanks to the extended spectral range, the 3MI instruments will
provide additional capabilities for the retrieval of aerosol
characteristics over land.
42/38You can also read
