MISR Remote Sensing of Cloud Properties

1
MISR Remote Sensing of Cloud Properties Larry Di
Girolamo Department of Atmospheric Sciences,
University of Illinois at Urbana-Champaign larry_at_a
tmos.uiuc.edu Contributors Roger Davies1,
David Diner1, Akos Horvath2, Catherine Moroney1,
Mike Wilson3, Yuekui Yang3, Guangyu Zhao3 1Jet
Propulsion Laboratory, Pasadena, CA 2University
of of Arizona, Tucson, AZ 3University of Illinois
at Urbana-Champaign, Urbana, IL
2
Central Italy June 12, 2002
3
MISR exploits the information content of the
angular anisotropy of the upwelling radiation
field. Algorithms for remote sensing by way
of angular signatures are in their
infancy. Plenty of pioneering work to be done
with angular signatures!
4
MISR Cloud Products Cloud Mask
Products Radiometric Camera-by-camera Cloud
Mask Angular Signature Cloud Mask Stereo
Derived Cloud Mask Stereo Products Stereo
heights (several flavors) Cloud-track wind
velocity Albedo Products Local
albedo Restrictive albedo Expansive
albedo Cloud Classifier Products Altitude-binned
cloud fractions Angle-by-angle cloud
fractions Texture indices
Details in ATBDs Level 1 Cloud Detection JPL
D-13397 Level 2 Cloud Detection and
Classification JPL D-11399 Level 2 TOA Albedo JPL
D-13401
5
Radiometric Camera-by-camera Cloud Mask Water
Tests R(865 nm) at 1.1 km resolution ?(670 nm)
of 275 m red BRFs within 1.1 km sample 720
thresholds are a function of Sun-view geometry
and are static with time. Land Tests D
NDVIb/R2(670 nm) at 1.1 km resolution DSVI
spatial variability index of D (3.3 km effective
resolution) (b 0.4 for desert and b 0.6 for
vegetation) 1,137,600 thresholds are a function
of Sun-view geometry and 1580 land surface
types. Thresholds are not static with time. They
are updated every 16 days based on data collected
over a 32 day window. A global, non-parameteric,
automated threshold selection procedure is
applied to histograms of this data.
6
Orbit 6663, Tropical Pacific Ocean, March 19,
2001, AN Camera
7
Orbit 6663, Tropical Pacific Ocean, March 19,
2001, AN Camera
8
East Texas August 25, 2001 AN camera
9
RCCM Status RCCM over water Provisional RCCM
over land Beta (Prov. by October 2002) Unique
Attributes Provides cloud cover for the same
scene as a function of view angle. Novel
global, non-parametric, scene dependent threshold
selection procedure (no manual tuning, except
over water) Major Problems Not reliable over
snow and sea ice
10
Angular Signature Cloud Mask
The Band Differenced Angular Signature (BDAS) is
the only thresholded quantity in the ASCM. Using
cameras that measure forward scattered
radiation BDAS Blue NIR70 Blue
NIR60
Antarctica April 2, 2001
Thresholds are static with time over ocean and
variable with time and location over land in the
same manner as the RCCM. The ASCM is a
terrain-projected cloud mask in the D-camera used
in computing the BDAS.
11
Beaufort Sea May 19, 2001
12
Beaufort Sea May 19, 2001
13
East Antarctic Plateau January 7, 2002
14
ASCM Status The ASCM is not yet available
through the Langley DAAC. It is anticipated to be
available as Beta Version spring 2003. Unique
Attributes Works very well over all surface
types, including snow and ice Very good
sensitivity for detecting very thin cirrus Known
Problems None found, but still early
15
Stereo Derived Cloud Mask The stereo derived
height is the only thresholded quantity in the
SDCM. Thresholds are static with time and are
equal to the DEM heights 560 m. Confidence
levels are placed on the SDCM based on the
matcher and ray skewness scores.
16
Orbit 6663, Tropical Pacific Ocean, March 19,
2001, AN Camera
17
SDCM Status SDCM Provisional Unique
Attributes Extremely simple logic with a
single global threshold Independent of surface
spectral signature, which makes it insensitive to
surface type Known Problems Very low clouds
(lt 600 m altitude) may go undetected Occasional
blockiness at 70 km scale due to stereo height
product
18
Stereoscopic parallax
Forward-viewing MISR camera
MISR flight direction
cloud-top height
apparent cloud position
19
Stereoscopic parallax
Backward-viewing MISR camera
MISR flight direction
cloud-top height
parallax
20
Clouds over Florida and Cuba 6 March 2000
21
Hurricane Eric, September 8, 2001
22
Passive Stereo Wall
23
MISR Stereo Matching common features in images
viewed at different angles is key to
stereo. MISR stereo uses a hierarchy of area
and feature matchers. MISR stereo height
uncertainty 500 m to 600 m with standard global
processing algorithm.
24
Cloud-Track Winds from MISR
By tracking cloud features over a mesoscale
domain in a triplet of MISR images, cloud
velocity and height can be unscrambled from the
measured disparities.
The cloud wind for the domain is taken as the
average velocity and assigned to the median
height.
25
Winds at standard 70.4 km resolution
North Atlantic April 9, 2001
26
High density winds at 35.2 km resolution
North Atlantic April 9, 2001
27
The MISR Wind Advantage MISR is the first polar
orbiting instrument to measure cloud track winds
everywhere within its swath. Geostationary
satellites cannot measure cloud-track winds
greater than 55 latitude MISR
can. Geostationary satellites have cloud-track
wind errors 7 m/s, where half this error comes
from poor cloud height assignment MISR can get
3 m/s with good cloud height assignment.
28
Stereo Height Products Baja, December 31, 2001
29
Stereo Status The stereo heights are
Provisional The winds are Beta Version (Prov.
October) Unique Attributes Most accurate
global cloud heights/winds from passive
instrument to date. Fills in high latitude
regions missed by geostationary satellites Known
Problems Stereo quality is sensitive to the
quality of georegistration
30
MISR Albedo The anisotropy of the upwelling
radiation field requires angular distribution
models (ADM) to convert radiance to flux. 9
views capture part of the anisotropy and better
constrain the choice of ADM compared to a single
view.
Nadir view only
MISR
forward
RMS error
Solar zenith angle ()
670 nm BRF (simulated)
(simulated 35.2 km resolution)
31
MISR Albedo The anisotropy of the upwelling
radiation field requires angular distribution
models (ADM) to convert radiance to flux. 9
views capture part of the anisotropy and better
constrain the choice of ADM compared to a single
view.
Nadir view only
MISR
forward
RMS error

Solar zenith angle ()
670 nm BRF (simulated)
(simulated 35.2 km resolution)
32
MISR Albedo The anisotropy of the upwelling
radiation field requires angular distribution
models (ADM) to convert radiance to flux. 9
views capture part of the anisotropy and better
constrain the choice of ADM compared to a single
view.
Nadir view only
MISR




forward

RMS error




Solar zenith angle ()
670 nm BRF (simulated)
(simulated 35.2 km resolution)
33
Restrictive vs Expansive Albedo
Expansive Albedo
30 km altitude TOA
Restrictive Albedo
34
Albedo Products Nan Hai Sea, January 30, 2002
35
Albedo Status Local Albedo Provisional Restric
tive and Expansive albedo will be implemented in
mid-October. Unique Attributes High resolution
local albedo (2.2 km) 9 views provide strict
constraint on ADM choices Allows a comparison
in albedo definition Allows the possibility to
examine albedo as a function of altitude Known
problems None found, but still early
36
Some unique things we are currently doing
with the MISR cloud data
37
3-D radiative transfer effects on retrievals of
cloud microphysics
38
3-D radiative transfer effects on retrievals of
cloud microphysics
Off the coast of Baja June 30, 2001
39
Caribbean Sea July 9, 2002 CRYSTAL-FACE Red
Channel, AN
40
Caribbean Sea July 9, 2002 CRYSTAL-FACE Red
Channel
41
Cloud Fraction vs View Angle
Potential use 1) Correction to cloud
climatologies derived from wide swath
instruments 2) Scene ID correction for Earth
radiation budget instruments (e.g., CERES) 3)
Remote sensing of cloud geometrical thickness
All clouds 1547 orbits December 2001 April
2002 Ocean only Between 55 latitude Needs to
be stratified by cloud type
42
Other Ideas Cloud phase/ice habit
discrimination Cloud optical properties that
include 3-D radiative effects Narrow-to-broadband
conversion for albedo products Improved cloud
climatologies, especially in polar
regions Assimilate MISR cloud heights/winds into
forecast models Comparison between MISR and CERES
albedo Comparison between MISR and MODIS cloud
masks Comparison between MISR and MODIS cloud top
heights MISR-MODIS data synergy to better
understand 3-D effects MISR-MODIS data synergy to
derive improved cloud emissivity
43
Postdoc needed! Department of Atmospheric
Sciences University of Illinois at
Urbana-Champaign
44
Conclusions MISR is cool Future
Work Validation Science New Algorithm
Development
45
Conclusions MISR is cool Future
Work Validation Science New Algorithm
Development We invite you to participate!