Principles of Remote Sensing Using Spectral Information

1
Principles of Remote Sensing Using Spectral
Information
Some steps along the way from satellite
observations to useful geophysical content.
Richard Kleidman, SSAI
2
From Spectra to Product
The spectral response of an unknown can help us
establish the identity of what we are observing
much like a fingerprint can help us establish the
identity of the person who made the print
?
Product A set of values that is used to
describe, in a consistent way, physical
properties of an observed geophysical phenomena.
Products can consist of Images Qualitative
evaluations Cloudy or clear Quantitative
measurements Amount of aerosol, percent cloud
There are many complications but lets examine a
little bit about how we get from raw spectra to
useful product.
3
The signal reaching any space borne sensor is a
complex mixture of surface and atmospheric
components.
Lets begin simply by examining some sample
spectra of different surfaces.
One of the advantages of MODIS is its broad
spectral range. The wider the spectral range the
more information content we have when we observe
the Earth - Atmosphere system.
Aspen Leaves - very uniform
Aspen Yellow Leaf
Aspen Green Leaf
4
Leaves top - Forest bottom
Pine Leaves
Spruce Leaves
Spruce Forest
Pine Forest
5
Pine Forest
Conifer Forest Meadow
Fescue and Wheatgrass
Cheatgrass
6
Rangeland
Concrete
Melting snow 25 vegetation
Melting snow 50 vegetation
7
Gaseous Absorption
Atmospheric gases - CO2, O2, and H2O absorb the
solar radiation at specific locations in the EM
spectra causing the gaps we see at the left. In
most cases the absorption bands limit our ability
to obtain useful information There are some
cases where we can exploit the absorption bands
to obtain additional information.
8
Slant-Path Absorption of the Atmosphere
Location of Primary Atmospheric Windows
Courtesy of Michael King
0.05
O2 A-Band
O2 B-Band
0.04
0.03
Absorption
0.02
0.01
0.00
0.50
0.55
0.60
0.65
0.75
0.80
0.85
0.70
Wavelength (µm)
9
Primary Atmospheric Windows in the Visible and
near-IR Spectra
From E. Vermote et. al. 6s Manual
10
From Spectra to Product
Some steps along the way from satellite
observations to useful geophysical content.
Aviris Spectra
MODIS Band 4
MODIS Aerosol Optical Depth
MODIS Cloud Fraction
11

• We have examined spectra in the visible to near
  IR range for a few surfaces going from relatively
  uniform to mixed surface types.
• Now lets look at one wavelength at a time for an
  entire MODIS scene.
• As the film plays notice how various features
  become more or less easy to distinguish from each
  other as we change from band to band.
• The film includes the thermal bands which we have
  
• not yet discussed.

12
Movie courtesy of Steven Platnick, NASA GSFC
13
From Spectra to Product
Some steps along the way from satellite
observations to useful geophysical content.
Aviris Spectra
MODIS Band 4
MODIS Aerosol Optical Depth
MODIS Cloud Fraction
14
Examples of Using Spectral Informationfor Clouds
and Aerosols

• Choosing a single band to find cirrus clouds
• Using multiple bands to separate smoke, dust and
  surface
• Using multiple bands to separate cloud, land and
  ocean surfaces.

15
Taking Advantage of An Atmospheric Absorption Band
Aqua true color image of cloud and ocean surface.
11 um temperature channel of Aqua image. Cold
high clouds are white. Warmer low clouds are
gray.
When cirrus clouds are very thin the temperature
channel will receive a mixed signal from the
surface and the clouds. Cirrus clouds may not be
detected.
We would like to be able to identify cirrus
clouds in this scene
There is a very strong water absorption band at
1.38 um This signal is attenuated as we travel
downward through the atmosphere. It will be
sensitive to signals high in the atmosphere but
not from the surface.
1.38 um channel image. Note visible cirrus.
16
Extracting Information by Specific Band Usage
Biomass burning Cuiabá, Brazil (August 25, 1995)
R 0.66 µm G 0.55 µm B 0.47 µm
R 1.6 µm G 1.2 µm B 2.1 µm
Ag (2.1 µm) lt 0.10 0.10 lt Ag (2.1 µm) lt 0.15
?0 36
20 km
12 km
AVIRIS
17
Spectral Optical Properties of Aerosol
Both dust and smoke interact with the
shorter wavelengths reflecting light back to the
sensor.
The larger dust particles interact with the
longer infrared wavelengths but not the smaller
smoke particles which remain invisible.
Dust
This distinction is made possible by the
wide spectral range of the MODIS sensor.
Smoke
from Y. Kaufman
18
Spectral Optical Properties of Aerosol
Here you can see the spectral response of the
large and small particles.
Dust /Sea Salt
Smoke/Pollution
wavelength
19
Using multiple bands to separate cloud, land and
ocean surfaces.
The analysis and images used in this example have
been generated with the Hydra tool for
multispectral data analysis. The tool was
created by the Space Sciences Engineering Center
at the University of Wisconsin and can be found
at http//www.ssec.wisc.edu/hydra/
The tool allows you to look at individual bands,
combinations of bands and to link plots of the
data to the data images.
20
Using multiple bands to separate cloud, land and
ocean surfaces.
This is an Aqua MODIS image of a storm system
off the coast of South America
We are going to use a combination of three
different bands to quantitatively draw
a distinction between clouds, land and ocean.
21
We begin by making an ocean vs. land mask
MODIS band 2 0.86 um
Band 2 / Band 1
Accentuates the differences between land and
ocean Is not sensitive to clouds. Clouds are
spectrally bright through all of the reflectance
bands. When we divide band 2 by band 1 almost all
values for cloud become very close to 1.0.
Sensitive to vegetation Clouds are bright Water
is dark
MODIS band 1 0.66 um
Land is dark Clouds are bright Water is dark
22
Separating Cloud from Non-Cloud
MODIS Band 31 11 um
This is a thermal band. Clouds are cold
when compared to land and ocean surfaces. We use
an inverted color scale so that the low values
of clouds appear bright.
23
Putting it all together
This is a scatter plot of band 31 on the
x-axis Vs. Band 2 / Band 1 on the y-axis We
have managed to separate the points into somewhat
distinct groups.
24
Putting it all together
Using hydra we can select some of the
distinctive clusters of points in the scatter
plot and color their corresponding locations in
the image. The image at bottom left now shows
our crude mask of Land, Ocean and Cloud.
25
Complications
There is much more involved in creating a product
than just finding the right combination of
spectral channels.

• Two broadly grouped sets of difficulties in using
  remote sensing information to identify
• aerosols and clouds are
• Surface based problems
• 2) Satellite based problems

26
Observe the following true color images.
See how many surface issues, satellite issues or
combinations of the two you can find from the
following three images.
09 April 2004
12 June 2004
19 December 2004
27
What are some complications as we add more
aerosol?
28
In an ideal situation with no atmosphere all of
the incoming radiation would reach the surface.
A portion of the photons would be absorbed at the
surface. The remaining photons reflect back up
into space.
Measured radiance directly depends on surface
properties
NO ATMOSPHERE
Diagrams from E. Vermote et. al, 6S manual
29
What does the satellite see? What information do
the photons contain?
1) Backscattered photons which never reach the
surface.
Signal or Noise?
and for whom?
Diagram from E. Vermote et. al, 6S manual
30
Diffuse solar radiation.
2) Scattered photons which illuminate the
ground.
Signal or Noise?
3) Photons reflected by the surface and
then scattered by the atmosphere.
Diagrams from E. Vermote et. al, 6S manual
31
Multiple scattering events
This is usually ignored after one or two
interactions.
Diagram from E. Vermote et. al, 6S manual
32
The real atmosphere complicates the signal. Only
a fraction of the photons reach the sensor so
that the target seems less reflecting.
Real Atmosphere

• Photons lost due to
• Atmospheric absorption
• Scattering

From E. Vermote et. al, 6S manual
33
Geometric Issues of the Illumination and the
Measurement
Very important for surface and atmospheric signal
Solar zenith angle
Sensor zenith angle
Sensor view angle
Solar view angle