Transcript Lecture 3&4: Spectral Signature, VNIR Radiation Models

Lecture 3: Remote Sensing
Spectral signatures, VNIR/SWIR, MWIR/LWIR
Radiation models
Video
http://www.met.sjsu.edu/metr112videos/MET%20112%20Video%20Li
brary-MP4/energy%20balance/
• Solar Balance.mp4
Jin: We failed to show this one on class, you can access it from the link above
Spectral signature
Much of the previous discussion centered around the
selection of the specific spectral bands for a given theme
 In the solar reflective part of the spectrum (350-2500 nm), the shape of the
spectral reflectance of a material of interest drives the band selection
Recall the spectral reflectance of vegetation
Select bands based on an absorbing or reflecting feature in the material
In the TIR it will be the emissivity that is studied
The key will be that
different materials have
different spectral reflectances
 As an example, consider
the spectral reflectance
curves of three different
materials shown in the graph
These divisions are not precise and can vary depending on the publication
1)
2)
3)
4)
Visible-Near IR (0.4 - 2.5);
Mid-IR (3 - 5);
Thermal IR (8 - 14);
Microwave (1 - 30 centimeters)
VNIR - visible and near-infrared ~0.4 and 1.4 micrometer (µm)
Near-infrared (NIR, IR-A DIN): 0.75-1.4 µm in wavelength,
defined by the water absorption
Short-wavelength infrared (SWIR, IR-B DIN): 1.4-3 µm,
water absorption increases significantly at 1,450 nm.
The 1,530 to 1,560 nm range is the dominant spectral region
for long-distance telecommunications.
Mid-wavelength infrared (MWIR, IR-C DIN) also called intermediate
infrared (IIR): 3-8 µm
Long-wavelength infrared (LWIR, IR-C DIN): 8–15 µm
Far infrared (FIR): 15-1,000 µm
Spectral Signature
Spectral signature is the idea that a given material has a
spectral reflectance/emissivity which distinguishes it from
other materials
Spectral reflectance is the efficiency by which a material reflects
energy as a function of wavelength
The success of our differentiation depends heavily on the sensor we
use and the materials we are distinguishing
Unfortunately, the problem is not as simple as it may appear since
other factors beside the sensor play a role, such as
•Solar angle
•View angle
•Surface wetness
•Background and surrounding material
Also have to deal with the fact that often the energy measured by
the sensor will be from a mixture of many different materials
This discussion will focus on the solar reflective for the time being
Spectral Signature - geologic
Minerals and rocks can have distinctive spectral shapes based
on their chemical makeup and water content
 For example, chemically bound water can cause a similar feature to
show up in several diverse sample types
 However, the specific spectral location of the features and their shape
depends on the actual sample 1
Spectral signature - Vegetation
Samples shown here are for a variety of vegetation types
 All samples are of the leaves only
 That is, no effects due to the branches and stems is included
Vegetation spectral reflectance
Note that many of the themes for Landsat TM were based on
the spectral reflectance of vegetation
 Show a typical vegetation spectra - KNOW THIS CURVE
 Also show the spectral bands of TM in the VNIR and SWIR as well
as some of the basic physical process in each part of the spectrum
Spectral signature - Atmosphere
Recall the graph presented earlier showing the transmittance
of the atmosphere
 Can see that there are absorption features in the atmosphere that could be
used for atmospheric remote sensing
 Also clues us in to portions of the spectrum to avoid so that the ground is
visible
A signature is not enough
Have to keep in mind that a spectral signature is not always enough
 Signature of a water absorption feature in vegetation may not indicate the
desired parameter
 Vegetation stress and health
 Vegetation amount
 Signatures are typically derived in the laboratory
 Field measurements can verify the laboratory data
 Laboratory measurements may not simulate what the satellite sensor
would see
 Good example is the difficult nature of measuring the relationship between
water content and plant health
 Once the plant material is removed from the plant to allow measurement
it begins to dry out
 Using field-based measurements only is limited by the quality of the
sensors
 The next question then becomes how many samples are needed to
determine what signatures allow for a thematic measurement
This is a black spruce forest in the BOREAS experimental region in Canada.
Left: backscattering (sun behind observer), note the bright region (hotspot)
where all shadows are hidden. Right: forwardscattering (sun opposite observer),
note the shadowed centers of trees and transmission of light through
the edges of the canopies. Photograph by Don Deering.
http://www-modis.bu.edu/brdf/brdfexpl.html
A soybean field. Left: backscattering (sun behind observer). Right: forwardscattering
(sun opposite observer), note the specular reflection of the leaves.
Photograph by Don Deering. http://www-modis.bu.edu/brdf/brdfexpl.html
Signature and resolution
The next thing to be concerned about is the fact that we will not
fully sample the entire spectrum but rather use fewer bands
In this case, all four
bands will allow us to
differentiate clay and
grass
Using bands 1, 3, and
4 would also be
sufficient to do this
Even using just bands 3
and 4 would allow us to
separate clay and grass
Signature and resolution
Band selection and resolution for spectral signatures should
be chosen first based on the shapes of the spectra
 That is, it is not recommended to rely on the absolute difference between
two reflectance spectra for discrimination
 Numerous factors can alter the brightness of the sample while not
impacting the spectral shape
Shadow effects and illumination conditions
Absolute calibration
Sample purity
 Bands showngive
 Gypsum
- Low, high, lower
 Montmorillonite
- High, high, low
 Quartz
- high, high,
not so high
Quantifying radiation
It is necessary to understand the energy quantities that are typically
used in remote sensing
 Radiant energy (Q in joules) is a measure of the capacity of an EM
wave to do work by moving an object, heating, or changing its state.
 Radiant flux (Φ in watts) is the time rate (flow) of energy passing
through a certain location.
 Radiant flux density (watts/m2) is the flux intercepted by a planar
surface of unit area.
 Irradiance (E) is flux density incident upon a surface.
 Exitance (M) or emittance is flux density leaving a surface.
 The solid angle (Ω in steradians)
subtended by an area A on a
2
spherical surface of radius r is A/r
 Radiant intensity (I in watts/sr) is the flux per unit solid angle in a
given direction.
 Radiance (L in watts/m2/sr) is the intensity per unit projected area.
 Radiance from source to object is conserved
Radiometric Definitions/Relationships
Radiant flux, irradiance (radiant exitance), radiance
The three major energy quantities are related to
each other logically by examining their units
In this course, we will deal with the special case
Object of interest is located far from the
sensor (factor of five)
Change in radiance from object is small
over the view of the sensor
Then
Φdetector = L object × Areacollector × ΩGIFOV
Φdetector = E object × Areacollector
E detector = L object × ΩGIFOV
ΩGIFOV= AreaGIFOV/H2
ΩGIFOV= Areadetector/f2
Electromagnetic Spectrum: Transmittance,
Absorptance, and Reflectance
Radiometric Definitions/Relationships
Emissivity, absorptance, and reflectance
All three of these quantities are unitless ratios of energy quanities
Emissivity, ε, is the ratio of the amount of energy emitted by an object
to the maximum that could possibly emitted at that temperature
Absorptance, α, is the ratio of the amount of energy absorbed by an
object to the amount that is incident on it
Reflectance, ρ, is the ratio of the amount of energy reflected by an
object to the is incident on it
All three can be written in terms of the emitted, reflected, incident, and
absorbed radiance, irradiance, radiant exitance, or radiant flux (but since
above three quantities are unitless, numerator and denominator must be
identical units)
In terms of radiant flux we would have
Radiometric Laws - Cosine Law
Cosine Law - Irradiance on surface is proportional to cosine of the angle
between normal to the surface and incident radiance
E = E0cosθ
In figures below, if E0 (or L0 converted to irradiance using the solid angle) is
normal to the surface, we have a maximum incident irradiance
For E0 that is tangent to surface, the incident irradiance is zero
Cosine effect example
Graph on this page shows the downwelling total irradiance as a function of time for
a single day as measured from a pyranometer
Radiometric Laws - 1/R2
Distance Squared Law or 1/R2 states that the irradiance from a point
source is inversely proportional to the square of the distance from the source
Only true for a point source, but for cases when the distance from the
source is large relative to the size of the source (factor of five gives
accuracy of 1%)
Sun can be considered a point source at the earth
Satellite in terrestrial orbit does not see the earth as a point source
Can understand how this law works by remembering that irradiance has a
1/area unit and looking at the cases below
In both cases, the radiant flux
through the entire circle is same
Area of larger sphere is 4
times that of the smaller
sphere and irradiance for
a point on the sphere is ¼
that of the smaller sphere
Radiometric Laws - Lambertian Surface
Lambertian surface is one for which the surface-leaving radiance is constant with angle
 It is the angle leaving the surface for which the radiance is invariant
 Lambertian surface says nothing about the dependence of the surfaceleaving radiance on the angle of incidence
 In fact, from the cosine law, we know that the incident irradiance will
decrease with sun angle
 If the incident irradiance decreases, the reflected radiance decreases
as well
 The radiance can decrease, as long as it does so in all directions
equally
2
Radiometric Laws - Lambertian Surface
Using the integral form of the relationship between radiance and irradiance we can
show that
Elambertian=¶Llambertian
To obtain the irradiance we have to consider the radiance through an entire Hemisphere
Because of the large range of angles, we cannot simply use E=LΩ
Radiometric Laws - Planck’s Law
States that the spectral radiant exitance from a blackbody depends only on
wavelength and the temperature of the blackbody
A blackbody is an object that absorbs all energy incident on it, α=1
Corrollary is that a blackbody emits the maximum of energy possible for an
object a given temperature and wavelength
Radiometric Laws - Planck’s Law
Once you are given the temperature and wavelength you can develop
a Planck curve
Planck curves never cross
Curves of warmer bodies are above those of cooler bodies
Radiometric Laws - Wien’s Law
Peaks of Planck Curves get lower and move to longer wavelengths as
temperature decreases
Maximum wavelength of emission is defined by Wien’s Law
λmax=2898/T [μm]
Solar Radiation
Sun is the primary source of energy in the VNIR and SWIR
Peak of solar curve at approximately 0.45 μm
Distance to sun varies from 0.983 to 1.0167 AU
Irradiance (not spectral irradiance) at the top of the earth’s atmosphere for
normal incidence is 1367 W/m2 at 1 AU
Terrestrial Radiation
Energy radiated by the earth peaks in the TIR
Effective temperature of the earth-atmosphere system is 255 K
Planck curves below relate to typical terrestrial temperatures
Solar-Terrestrial Comparison
When taking into account the earth-sun distance it can be shown that solar energy
dominates in VNIR/SWIR and emitted terrestrial dominates in the TIR
Sun emits more
energy than the earth
at ALL wavelengths
It is a geometry effect
that allows us to treat
the wavelength
regions separately
Solar-Terrestrial Comparison
Plots here show the energy from the sun at the sun and at the top of the earth’s atmosphere
Also show the emitted
energy from the earth
Vertical Profile of the Atmosphere
Understanding the vertical
structure of the atmosphere
allows one to understand better
the effects of the atmosphere
 Atmosphere is divided into layers
based on the change in
temperature with height in that
layer
 Troposphere is nearest the
surface with temperature
decreasing with height
 Stratosphere is next layer and
temperature increases with height
 Mesosphere has decreasing
temperatures
Atmospheric composition
Atmosphere is composed of dust and molecules which vary spatially
and in concentration
 Dust also referred to as aerosols
 Also applies to liquid water, particulate matter, airplanes, etc.
 Primary source of aerosols is the earth's surface
 Size of most aerosols is between 0.2 and 5.0 micrometers
 Larger aerosols fall out due to gravity
 Smaller aerosols coagulate with other aerosols to make larger
particles
 Both aerosols and molecules scatter light more efficiently at short
wavelengths
 Molecules scatter very strongly with wavelength (blue4sky)
 Molecular scattering is proportional to 1/(wavelength)
 Aerosols typically scatter with 1/(wavelength)
 Both aerosols and molecules absorb
 Molecular (or gaseous absorption is more wavelength dependent
 Depends on concentration of material
Absorption
MODTRAN3 output for US Standard Atmosphere, 2.54 cm column water vapor,
default ozone 60-degree zenith angle and no scattering
Absorption
Same curve as previous page but includes molecular scatter
Angular effect
Changing the angle of the path through the atmosphere effectively changes the concentration
 More material, lower transmittance
 Longer path, lower transmittance
Absorption
At longer wavelengths, absorption plays a stronger role with some spectral regions
having complete absorption
Absorption
Absorption
The MWIR is dominated by water vapor and carbon dioxide absorption
Absorption
In the TIR there is the “atmospheric window” from 8-12 μm
with a strong ozone band to consider
Radiative Transfer
Easier to consider the specific problem of the radiance at
a sensor at the top of the atmosphere viewing the surface
Radiation components
There will be three components of greatest interest in the
solar reflective part of the spectrum
 Unscattered, surface
reflected radiation Lλsu
 Down scattered,
surface reflected Lλsd
skylight
 Up scattered path Lλsp
radiance
 Radiance at the sensor
is the sum of these three
Radiative Transfer
Radiative transfer is basis for understanding how sunlight
and emitted surface radiation interact with the atmosphere
 For the atmospheric scientist, radiative transfer is critical for
understanding the atmosphere itself
 For everyone else, it is what atmospheric scientists use to allow others
to get rid of atmospheric effects
 Discussion here will be to understand the effects the atmosphere will
have on remote sensing data
 Start with some definitions
 Zenith Angle
 Elevation Angle
 Nadir Angle
 Airmass is 1/cos(zenith)
 Azimuth angle describes
the angle about the vertical
similar to cardinal directions
Optical Depth
Optical depth describes the attenuation along a path in the atmosphere
 Depends on the amount of material in the atmosphere and the type of
material and wavelength of interest
 Soot is a stronger absorber (higher optical depth) than salt
 Molecules scatter better (higher optical depth) at shorter wavelengths
 Aerosol optical depth is typically higher in Los Angeles than Tucson
 Total optical depth is less on Mt. Lemmon than Tucson due to fewer
molecules and lower aerosol loading
 Optical depth can be divided into absorption and scattering components
which sum together to give the total optical depth
δtotal = δ scatter + δabsorption
 Scattering optical depth can be broken into molecular and aerosol
δscatter = δmolec + δaerosol
 Absorption can be written as sum of individual gaseous components
δabsorption = δ H2 O + δO3 + δCO2 + .........
Optical Depth and Beer’s Law
Beer’s Law relates optical depth to transmittance
 Increase in optical depth means decrease in transmittance
 Assuming that optical depth does not vary horizontally in the
atmosphere allows us to write Beer’s Law in terms of the vertical
optical depth
 1/cosθ=m for airmass is valid up to about θ=60 (at larger values must include refractive
corrections)
 Recalling that optical depth is the sum of component optical depths
Beer’s Law also relates an incident energy to the transmitted energy
Directly-transmitted solar term
First consider the directly transmitted solar beam,
reflected from the ground, and transmitted to the sensor the unscattered surface-reflected radiation, Lλsu
Solar irradiance at the ground
Can also write the transmittance as an exponential in terms of optical dept
 Beer’s law
 Need to account for the path length of the sun due to solar zenith
angle of the sun in computing transmittance
 Account for the cosine incident term to get the irradiance on the
surface
Recall m=1/cosθsolar
Eλground, solar is
the solar irradiance at
the bottom of the
atmosphere normal
to the ground surface
(shown here to
be horizontal)
Requires a 1/r2 to account
for earth-sun distance
Incident solar irradiance
The surface topography will play a critical role in
determining the incident irradiance
 Two effects to consider
 Slope of the surface
 Lower optical depth because of higher elevation
 Good example of the usefulness of a digital elevation model (DEM) and
assumption of a vertical atmospheric model
Example: Shaded Relief
Surface elevation
model can be used to predict
energy at sensor
 Given
 Solar elevation angle
 local topography
(slope, aspect) from DEM
 Simulate incident angle
effect on irradiance
 Calculate incident
angle for every pixel
Determine cos[θ(x,y)]
 Creates a “shadedrelief” image
TM: Landsat thematic mapper
Directly-transmitted solar term
Reflect the transmitted solar energy from the surface
within the field of view of the sensor
 Once the solar irradiance is determined at the ground in the direction
normal to the surface it is reflected by the surface
 The irradiance is converted to a radiance
 Conversion from irradiance to radiance is needed because we want to
use the nice features of radiance
 Recall the relationship between irradiance and radiance derived
earlier for a lambertian surface - E=¶L
 There is a similar relationship between incident irradiance and reflected
radiance from a Lambertian surface
Directly-transmitted solar term
Last step is to transmit the radiance from the surface to
the sensor along the view path
 Simply Beer’s law again, except now we use the view path instead of
the solar path
Reflected downwelling atmospheric
Atmosphere scatters light towards the surface and this
scattered light is reflected at the surface to the sensor
 Compute an incident irradiance from the incident radiance due to
atmospheric scattering
 This incident irradiance is reflected from our lambertian surface to give
 Still need to transmit
this through the atmosphere
to get the at-sensor radiance
In the shadows
Image below is three-band mix of ETM+ bands 1, 4, and 7
 Note that there is still energy coming from the shadows
 Scattered skylight - which will have a blue dominance to it
Path Radiance Term
Path radiance describes the amount of energy scattered by the
atmosphere into the sensor’s view
 Basically, any photon for which the last photon scattering event
occurred in the atmosphere is a path radiance term
Can include or exclude an interaction with the ground
If it includes a surface interaction then this can be affected by
atmospheric adjacency effects
 The intrinsic path radiance is the radiance at the sensor that would
be measured if there were zero surface reflectance
 Contribution only from the atmosphere
 Depends only on atmospheric parameters
 No simple formulation
 Requires radiative transfer code
 Use Lλsp
Over water
A similar effect can be seen over water
Images here are also bands 3, 4, and 7 of ETM+ (LANDSAT)
Water is highly absorbing at these wavelengths thus almost all
of the signal is due to atmospheric scattering
At-sensor radiance in solar reflective
Summing the previous three at-sensor radiances will give
the total radiance at the sensor
There is a huge amount of buried information in the above
 This is a simplified way of looking at the problem
 Phase function effects from scattering and single scatter albedo are contained in
Edown and the path radiance
 Optical depths due to scattering and absorption are combined in the
transmittance terms
Also assumes lambertian surface!!!
Path radiance
Model output shows the spectral dependence of the atsensor radiance for path radiance and reflected radiance
TOA radiance, VNIR/SWIR
MISR data showing the effect of view angle on TOA radiance
with brightening and blue dominance at large views
Model versus measured
Comparison between measured spectra of RRV Playa
using AVIRIS and predicted radiance based on ground
measurements
The airborne visible/infrared imaging spectrometer (AVIRIS)
Model versus measured
Results below model the at-sensor radiance compared to
the sensor output
A raw AVIRIS spectrum (measured in digital numbers or. DN's)
TIR paths
There will also be three components of greatest interest in
the emissive part of the spectrum (or TIR)




Unattenuated, surface emitted radiation
Lλeued
Downward emitted, surface reflected skylight L λ
Upward emitted path radiance
Lλep
Radiance at the sensor is the sum of these three
Lλe = Lλeu + Lλed + Lλep
Thermal infrared problem
In the TIR, the problem is similar in philosophy as the
reflective
 Still have a path radiance, and reflected downwelling
 Direct reflected term in reflective is analogous to the surface emitted
term in the TIR
 Difference is that we are now dealing primarily with emission and
absorption rather than scattering
 Reflective we are most concerned with how much stuff is in the
atmosphere and what it is
 Aerosol loading (Gives aerosol optical depth)
 Atmospheric pressure (Gives molecular optical depth)
 Types of aerosols (Phase function and absorption properties)
 Amount of gaseous absorbers (Water vapor, ozone, carbon dioxide)
 In the TIR we must also worry about where these things are vertically
 Temperature depends on altitude
 Emission depends on temperature
 Need vertical profile of termperature, pressure, and amounts of
absorbers
Surface-emitted term
Surface emitted term will depend upon the emissivity and
temperature of the surface attenuated along the view path
 Easiest assumption is to assume that the surface is a blackbody but
then the temperature obtained will not correspond to the actual
temperature
 Better assumption is to assume the emissivity and temperature are
known and use Planck’s law to obtain the emitted radiance
 Transmitting this through the
atmosphere gives
Reflected downwelling and path radiance
Here, the equations are identical to the reflective case
 The downwelling radiance depends on atmospheric temperature and
composition
 Equations are the same
 Path radiance term is same as in reflective
 Must be computed from radiative transfer
 Depends heavily on atmospheric ,
 Use
Lλsp
 Sum is same approach as reflective
TOA Radiance, TIR
Concepts work in the other direction as well
 Radiance at the sensor will depend mostly upon where the layer is that
is emitting the energy seen by the sensor
 Location of the layer affects the temperature
 The warmer the layer, the higher
the radiance that is emitted
TIR Imagery examples
ETM+ Band 6 of Tucson showing temperature effects
 This image is from July
 Note the hot roads and cool vegetation
Bright and dark water
Water is dark in
reflective bands
but can be bright
in LWIR
 Warm water relative to
surround
 Water is also high
emissivity (nearly unity)
Bright and dark land
Example of
New Orleans
shown here
points out the
High temperatures
of the urban area
 Water in this
case is much colder
than the land
 Little contrast in
the reflective
the LANSSAT TM consists of 7 bands that have these characteristics:
Band No.
Wavelength
Interval (µm)
Spectral
Response
Resolution (m)
1
0.45 - 0.52
Blue-Green
30
2
0.52 - 0.60
Green
30
3
0.63 - 0.69
Red
30
4
0.76 - 0.90
Near IR
30
5
1.55 - 1.75
Mid-IR
30
6
7
10.40 - 12.50 Thermal IR
2.08 - 2.35
Mid-IR
120
30
TIR Imagery
CLASS Part.: WHY?
Clouds seen in the TIR (band 6 left) and visible (band 3
right) of ETM+ from July
TIR Imagery
TIR “Shadows” seen in the ETM+ band 6 image left are of
far different nature than those of the band 3 shadows
TIR Imagery
Canyons act as blackbody as well as have higher
temperatures due to lower elevations
 GOES image here
shows low radiance
as bright
 Note the Grand
Canyon is plainly
Visible
 Also evident are landwater boundaries (and
not just because of
the lines drawn to
show them)