ESM 266: Hyperspectral remote sensing

1
ESM 266 Hyperspectral remote sensing
2
Hyperspectral a terrible term

• Imaging spectrometry, imaging spectroscopy
• hyperspectral (too many, excessive) 100s of
  bands
• Ultraspectral 1000s of bands
• Generally we mean measuring the reflected or
  emitted radiation at a fine enough spectral
  resolution to identify materials

3
(No Transcript)
4
Example the mineral alunite
from Roger Clark, Spectroscopy of Rocks and
Minerals
5
Alunite at various spectral resolutions
from Roger Clark, Spectroscopy of Rocks and
Minerals
6
Atmospheric transmissivity
7
Nnik, Index of refraction (complex)
?i ?r
dl
I0
I
8
Optical properties of ice and water
9
Snow is a collection of scattering grains
10
Basic scattering properties of a single grain
(ice, mineral, etc)

• Mie theory, based on N and x2?r/?
• ? single-scattering albedo
• g asymmetry parameter
• Qext extinction efficiency

11
Snow spectral reflectance and absorption
coefficient of ice
12
Dust and algae
13
Spectral reflectance of dirty snow and snow with
red algae (Chlamydomonas nivalis)
14
Conventional approach to estimating albedo

• Satellite radiance (5 error)

Surface reflectance (gt5)
Narrowband albedo (5-10)
Broadband albedo (5-10)
15
Different paradigm, based on physical properties
and radiative transfer theory

• Measure 1.03mm absorption feature
• Estimate grain size
• Model spectral reflectance over all wavelengths
• Convolve with solar irradiance to estimate
  broadband albedo

ALBEDO
RT Model
Snow Grain Size
16
Estimate grain size from the1.03mm absorption
feature
17
Measured vs remotely sensed grain size
ETH/CU Camp12 June 2001q46.6
500
550
600
18
Snow-covered area in the Tokopah Basin (Kaweah
River drainage), Sierra Nevada
AVIRIS (20m pixels, 224 spectral bands)
21 May 1997
05 May 1997
18 June 1997
20 km
19
Grain size in the Tokopah Basin (Kaweah River
drainage), Sierra Nevada
21 May 1997
05 May 1997
18 June 1997
20 km
20
Absorption by three phases of water
21
Wet snow
(Green et al, WRR, forthcoming)
22
Surface wetness with AVIRIS, Mt. Rainier,14 June
1996
AVIRIS image, 409, 1324, 2269 nm
precipitable water, 1-8 mm
liquid water, 0-5 mm path absorption
vapor, liquid, ice (BGR)
23
Progression of snow wetness throughout morning
N
70 km
24
Optical properties of quartz
25
Causes of absorption

• Electronic
• Isolated atoms and ions have discrete energy
  states
• Absorption of photons of a specific wavelength
  causes a change from one energy state to a higher
  one
• Vibrational
• Bonds in crystals are like springs, can vibrate
• Frequency depends on strength of bonds and mass
  of molecules
• Rotational and translational
• Limited in solids, occurs in liquids and gases

26
Reflectance spectra showing vibrational bands
27
Reflectance spectra of different ices
28
Reflectance of green and dry vegetation
29
Concept of an imaging spectrometer
30
Airborne Visible Infrared Imaging Spectrometer
(AVIRIS)

• 400 2500nm at 10nm res.
• 224 spectral bands
• 20m spatial resolution
• 4 spectrometers VIS, NIR, SWIR1, SWIR2

Link to AVIRIS home page
31
Mineral map showing acid mine drainage
32
Most pixels are mixtures of different components

• Linear mixture
• Materials are optically separated, so no multiple
  scattering between components
• Intimate mixture
• Different materials are in intimate contact
  (e.g., minerals in a rock) so multiple scattering
  occurs
• Coatings
• One material coats another, each coating is a
  scattering/transmitting layer
• Molecular mixture
• Liquid/liquid or liquid/solid

33
Snow area and grain size validation
34
Spectral characteristics of different corals
Goodman Ustin, UC Davis
35
Currently operational

• AVIRIS
• 224 bands 0.4-2.5µm, flies on ER-2 or
  low-altitude Twin Otter
• EO-1
• Technology demonstration mission, includes
  Hyperion instrument
• 220 bands 0.4-2.5µm, 7.5x100km swath

36
Hyperion Imaging Spectrometer

• On-board NASA EO-1 satellite (demonstrating new
  sensor technologies)
• Pushbroom sensor at 705 km altitude (7.6 km swath
  width)
• Near-polar orbit (98o inclination)
• Flying in formation w/Landsat 7 (1 minute apart)
• spectral range 0.43 - 2.4 mm, 10 nm bandwidths
• 220 spectral bands
• 30m spatial resolution
• 12-bit quantization

37
(No Transcript)
38
Multispectral vs. hyperspectral remote sensing

• Multispectral
• separated spectral bands
• wider bandwidths
• coarse representation of the spectral signature
• not able to discern small differences between
  reflectance spectra
• smaller data volumes
• fewer problems with calibration
• multi-decadal history of continuous image
  acquisition

• Hyperspectral
• no spectral gaps
• narrow bandwidths (10nm)
• complete representation of the spectral signature
• ability to detect subtle spectral features
• large data volumes
• radiometric and spectral calibration are
  time-consuming
• currently no functioning sensors in orbit