ERS186: Environmental Remote Sensing

1
ERS186Environmental Remote Sensing

• Lecture 10
• Species Discrimination Using Remote Sensing

2
Keep in mind . . .
Three factors determine canopy reflectance.... and
thus our ability to discriminate plant canopies

• Factor 1 Spectral scattering/absorbing
  properties of canopy components. (leaves, stems,
  flowers, fruit, soil, etc.)
• Factor 2 Canopy architecture. (above-ground
  biomass leaf area index arrangement of foliage
  in x,y,z,q,f space for example, are all leaves
  vertical and located in one layer or perhaps
  they are arranged in space like the area on a
  sphere etc.)
• Factor 3 Directions of illumination and view.
  (Is the sun the only significant source or does
  aerosol- or Rayleigh-scattered light provide
  hemispherical illumination is direction of view
  toward the hot spot or nadir or )

3
Overview

• Applications
• Ecology
• Agriculture
• Physical Principles
• Cellular absorption and scattering
• Non-selective scattering
• BRDF
• Sensors
• Hyperspectral
• Hyperspatial

4
The Question

• What plant species are present in a remote
  sensing image?

5
Species Identification

• Not all vegetation looks the same! We can use
  this to help identify different species using RS.

6
Species Identification

• Why do the spectra of different species vary?
• Cellular differences (protein, cellulose and
  lignin, water, pigments, etc) factor 1,
  scattering/absorbing properties of canopy
  components (leaves)
• LAI, leaf angle, and leaf shape differences
  factor 2, architecture
• Trunk, stem and branch differences (size, number,
  color) factor 2, architecture
• Crown size and shape factor 2, architecture

7
Cellular Differences
Factor 1 scattering/absorbing properties of
canopy components (leaves)

• PROSPECT (Jacquemoud et al., 1996) models the
  light path through a simulated leaf with
  differing structural and chemical properties.
• Structural differences included rough, medium and
  smooth epidermis
• Chemical differences included differences in
  protein, cellulose and lignin, and water.
• The structural and chemical properties were
  derived from real leaves.
• Found differences in modeled reflectance with
  different properties, and these matched
  real-world reflectance curves.

8
Cellular Differences Pigments
Factor 1 scattering/absorbing properties of
canopy components (leaves)

• Pigments can and will vary between species, even
  closely related ones.
• Mature Valley vs. Live Oak reflectance and
  pigment contents

Ustin et al. 1998
9
Cellular Differences Water
Factor 1 scattering/absorbing properties of
canopy components (leaves)

• Water absorption features can help determine the
  amount of water in a leaf.
• Water differences can indicate different species,
  or different stress levels within a species.

Greenberg et al. 2001, healthy and water stressed
cotton spectra.
10
Canopy Level Differences, LAI
Factor 2 architecture

• All things being equal, LAI intercepts light
  according to Beers Law in the visible.
• Detection of LAI usually requires indices or
  proxy variables
• NDVI vs. LAI
• EWT vs. LAI (Roberts et al., in review)

11
Canopy BRF LAI Differences
Factor 2 architecture

• The relationship between LAI and canopy
  reflectance depends on species, age/growth, scale
  of measurement, distribution of leaves in a
  crown, leaf angle distribution, and many other
  factors.
• gtgt Key Point LAI is important, but
  differences in LAI do not necessarily mean
  differences in species nor differences in canopy
  reflectance and vice versa.

LAI vs. canopy species at WRCCF, Thomas and
Winner 2000. Shading refers to different canopy
strata.
12
Canopy BRF LAI Differences
Factor 2 architecture, hypotheical example A

• Consider pathological example A Two razor
  blade canopies...
• Factor 1, Same leaves (black), different soil
  (white/black)
• Factor 2, Same LAI in each canopy.
• Factor 3, Same view/illumination directions

• One canopy LAI value corresponds to two canopy
  reflectances
• gtgt Conclusion the relationship between BRF and
  LAI is not unique ltlt

Sun shining down the rows of razor blades
illuminates soil
Sun shining down the rows of razor blades
illuminates soil
Sensor, nadir view
Sensor, nadir view
View down the rows of razor blade leaves
View down the rows of razor blade leaves
Small BRF (black)
Large BRF (white)
Same LAI
soil, white
soil, black
13
Canopy BRF LAI Differences
Factor 2 architecture, hypotheical example B

• Consider pathological example B Two razor
  blade canopies...
• Factor 1, Same leaf color (black), same soil
  (white)
• Factor 2, Different LAI in each canopy.
• Factor 3, Same view/illumination directions

• One canopy BRF corresponds to two canopy LAI
  values
• gtgt Conclusion the relationship between BRF and
  LAI is not unique ltlt

Sun shining down the rows of razor blades
illuminates soil
Sun shining down the rows of razor blades
illuminates soil
Sensor, nadir view
Sensor, nadir view
View down the rows of razor blade leaves
Large LAI
View down the rows of razor blade leaves
Small LAI
Same BRF
soil, white
soil, white
14
Canopy BRF LAI Differences
Factor 2 architecture, hypotheical example C

• Consider pathological example C Two razor
  blade canopies...
• Factor 1, Same leaves (black), same soil (white)
• Factor 2, Same LAI 1.0 in each canopy but
  different leaf angle distribution
• Factor 3, Same view/illumination directions

• One canopy LAI value corresponds to two canopy
  reflectances
• gtgt Conclusion the relationship between BRF and
  LAI is not unique ltlt

Sun shining down the rows of razor blades
illuminates soil
Sun shining down the rows of razor blades
illuminates soil
Sensor, nadir view
Sensor, nadir view
View down the rows of razor blade leaves
Razor blade leaves form contiguous horizontal
layer above soil
Different BRF (White/black)
LAI1.0
LAI1.0
soil, white
soil, white
15
LAI and Ecosystems
Factor 2 architecture, examples
16
LAI and Ecosystems
Factor 2 architecture, examples
17
Definition of Leaf Area Index, LAI
Factor 2 architecture

• One sided green leaf area per unit ground area
• Example Total square meters of one side of green
  leaves above 1.0 square meter of soil
• LAI units m2 of leaf area/m2 of ground e.g.
  dimensionless

Green leaves
1.0 m2
soil
18
Leaf Angles Distribution
Factor 2 architecture

• Plants can dynamically change the angle of their
  leaves to increase or decrease the amount of EMR
  (and increase or decrease the heat loading).
• Leaves range from planophile (horizontally
  oriented) to erectophile (vertically oriented).
• Leaf angle probability density function is
  approximately spherical in many canopies i.e.
  canopy leaf area is distributed in angle like the
  area on a sphere.
• The angle of incident solar radiation and the
  angle of the leaf affect the at-sensor
  reflectance.

19
LAI/Leaf Angle and Spectra
Factor 2 architecture
MLA is Mean Leaf Angle
Asner, 1998
20
Leaf Angle Differences
Factor 2 architecture, examples
Asner, 1998
21
Leaf Shape
Factor 2 architecture
P. Cull
D. Tortosa

• Conclusion The shape of leaves can also affect
  reflectance.

22
Woody Matter
Factor 2 architecture

• Amount of woody matter can influence spectra,
  albeit slightly (Asner, 1998).
• SAI Stem area index.

23
Crown Shape
Factor 2 architecture

• Gerard and North 1997 modeled forests to look at
  red and NIR reflectance under different canopy
  conditions.
• Found wide, flat crowns (typical of tropical
  trees) were more reflective across all
  wavelengths than tall, skinny crowns (typical of
  northern conifers).

24
Crown Shape
Factor 2 architecture

• The shape of crowns is diagnostic of certain
  species.
• Example coniferous (conical) vs. deciduous
  (spherical)
• Hyperspatial imagery can be used to assess the
  actual shape.

25
Mapping Invasive Species
Putting it all together....example 1
DiPietro, 2002
26
Mapping Crop Types
Putting it all together....example 2

• Clark et al. 1995 used AVIRIS, Tricorder and
  reference spectrum to differentiate different CO
  crops.

27
Scaling in remote sensing
Putting it all together....

• Atomic/molecular properties (Factor 1)
• Absorption, transmission, molecular scattering
• Microscopic and small particle properties (Factor
  1)
• Scattering (cellular and particulate)
• Macroscopic structure properties (Factor 2)
• BRDF, geometric optics
• Landscape properties (Factor 2)
• Mixed pixels, ecosystem structure

Large small
28
Putting it all together....

• Most of the fundamental work on the mechanisms of
  remote sensing involves lab or field
  spectrometers, or modeling approaches and a good
  understanding of physics.
• Most of the work on the applications of remote
  sensing involves aerial or satellite sensors and
  a good understanding of statistics.
• The connection between the two scales is
  important, but is not well understood.