1 October 2003
flectance curves, and reflectance is theoretically an unvarying property of a material, the
spectra in the spectral libraries should match those of the same materials at other times or
(a) If data in spectral libraries are not appropriate, reflectance curves can be ac-
quired using a spectrometer. The instrument is aimed at a known target and records the
radiance reflected from the target over a fixed range of the spectrum (the 0.4- to 2.5-m
range is relatively common). The instrument must also measure the radiance coming in to
the target, so that the reflected radiance can be divided by incoming radiance at each
wavelength to determine spectral reflectance of the target. Given the time and expense of
gathering spectra data, it is best to check spectral libraries first.
(b) Two major spectral libraries available on the internet (where spectra can be
downloaded and processed locally if needed) include:
US Geological Survey Digital Spectral Library (Clark et al. 1993)
"Researchers at the Spectroscopy lab have measured the spectral reflectance of hundreds
of materials in the lab and have compiled a spectral library. The libraries are used as ref-
erences for material identification in remote sensing images."
ASTER Spectral Library (Jet Propulsion Laboratory, 1999)
"Welcome to the ASTER spectral library, a compilation of almost 2000 spectra of natural
and man made materials."
(c) The ASTER spectral library includes data from three other spectral libraries:
the Johns Hopkins University (JHU) Spectral Library, the Jet Propulsion Laboratory
(JPL) Spectral Library, and the United States Geological Survey (USGS--Reston) Spec-
(8) Real Life and Spectral Signatures. Knowledge of spectral reflectance curves is
useful if you are searching a remote sensing image for a particular material, or if you
want to identify what material a particular pixel represents. Before comparing image data
with spectral library reflectance curves, however, you must be aware of several things.
(a) Image data, which often measure radiance above the atmosphere, may have
to be corrected for atmospheric effects and converted to reflectance.
(b) Spectral reflectance curves, which typically have hundreds or thousands of
spectral bands, may have to be resampled to match the spectral bands of the remote
sensing image (typically a few to a couple of hundred).
(c) There is spectral variance within a surface type that a single spectral library
reflectance curve does not show. For instance, the Figure 2-25 below shows spectra for a
number of different soil types. Before depending on small spectral distinctions to separate