des
A N°
avet
2 au
Coll.
aux,
s de
's de
trie.
urg,
22:
ance
tion
MSS
Atlas
967.
ation
e. In
lieux
mbre
'e de
>
ional
ngle,
se of
»ens.,
eillet
5 TM
‘mote
e "La
233.,
puter
he 5S
ESTIMATING FOLIAR CHEMICAL CONCENTRATIONS WITH THE AIRBORNE
VISIBLE/INFRARED IMAGING SPECTROMETER (AVIRIS)
Paul J. Curran
Department of Geography, University College of Swansea,
University of Wales, Singleton Park, Swansea SA2 8PP, UK.
(ISPRS Commission VII)
ABSTRACT
The reflectance spectra of a vegetation canopy contains information on the chemicals within that canopy. This paper reviews the underlying
theory and the use of the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) to estimate foliar chemical concentrations. There are several
problems associated with the use of AVIRIS data in this way. Sampling of forest canopies, wet laboratory analysis of foliar chemicals and
visualization of AVIRIS data are discussed with reference to AVIRIS data collected for a flat site in Florida and a hilly site in Wales.
KEY WORDS: AVIRIS, foliar chemical concentrations, forest ground data collection.
1. INTRODUCTION AND REVIEW
Measures of foliar chemical concentration provide indicators of plant
productivity (e.g., levels of chlorophyll), rate of litter decomposition
(e.g., levels of lignin) and availability of nutrients (e.g., nitrogen). By
making such estimates remotely we have the potential of studying the
quality of the vegetation and a portion of several nutrient cycles on a
local to a global scale (Janetos et al, 1992). A number of agencies
have stated that their research goals for the 1990s include the study
of global biochemical cycling and that remote sensing is a tool needed
to achieve this goal (Curran, 1989). For example, the US Committee
on Earth Sciences was very specific about the role that remotely
sensed information on foliar chemical concentrations could play.
“New, high spectral resolution remote sensing techniques show
promise of estimating canopy chemical composition parameters that
can be used to elucidate ecosystem properties. Basic understanding
and wider validation of this approach are needed" (Committee on
Earth Sciences, 1989, p.58).
The reflectance spectra of all types of vegetation in the 400-2400 nm
spectral region are, in general, similar (Figure 1). In near-infrared
wavelengths there is high reflectance as a result of leaf scattering and
at approximately equal distances throughout the spectrum there are
5 major absorption features. These absorption features are the result
of electron transitions in chlorophyll (400-700 nm) and of the bending
and stretching of the O-H bond in water and other chemicals (970,
1200, 1400, 1940 nm) (Osborne and Fearn, 1986; Williams and
Norris, 1987). In addition, spectroscopic measurements of dried and
ground leaves, made by researchers with the U.S. Department of
Agriculture (USDA), have revealed over 40 minor absorption features
(Curran, 1989). These minor absorption features have been
correlated with the concentration of organic compounds (e.g.,
cellulose, lignin, protein, oil, sugar, starch) in dried and ground leaves.
These organic compounds absorb radiation strongly in the ultraviolet
(«400 nm) and middle-infrared (2400 nm) spectral regions as a result
of stretching and bending vibrations of the strong molecular bonds
between hydrogen atoms and the atoms of carbon, nitrogen, and
oxygen (Banwell, 1983; Osborne and Fearn, 1986). The minor
absorption features we see in the 400-2400 nm spectral region are
relatively weak and broad, with a 30-40 nm half-depth bandwidth.
They are the result of harmonics and overtones of the stronger
absorptions and combination bands at longer and shorter wavelengths
(Peterson and Running, 1989; Barton et a/, 1992). For some years
now researchers at the USDA have been using regression
relationships between reflectance derivatives (Dixit and Ram, 1985) in
several narrow wavebands and chemical concentrations to estimate
accurately the chemical composition of dried and ground vegetation
(Williams and Norris, 1987). Today reflectance spectroscopy is a
routine procedure, where the accuracy and repeatability of reflectance
estimates of protein, lignin and starch concentrations in dried and
ground plant materials are at least comparable to those obtained by
wet laboratory methods (McLellan et al, 1991b). As a result, the
technique has been certified by the Association of Official Analytical
Chemists (AOAC, 1990) and is used throughout the American
agricultural industry (Weyer, 1985; Marten et al., 1989; Williams and
Norris, 1987; Barton and Windham, 1988). The procedures developed
705
by the USDA have been extended most successfully to the analysis
of dried and ground tree leaves in the laboratory (Wessman et al.,
1988a; Card et al, 1988, Peterson et al., 1988; McLellan, 1991a,
1991b), a forest canopy in the field (Peterson et al., 1988; Wessman
et al., 1988b; 1989; Johnson and Peterson, 1991) and fresh leaves in
the laboratory (Curran et al., 1992). In recent years the techniques of
spectral matching (Goetz et al., 1990) and spectral decomposition
(Card and Peterson, 1992) have also been used as a way of avoiding
some of the pitfalls associated with the standard regression-based
methodology that uses reflectance derivatives and a wet laboratory
analysis (Curran, 1989; Clark, 1991; Janetos et a/., 1991). This is just
an interim step, as in the future we will need to develop a physically-
based modelling approach to the remote sensing of foliar chemical
concentrations (Wessman, 1990; Peterson, 1991; Janetos et al., 1992;
Baret et al., 1992).
2.0 HE
. nm. sr
2
Radiance, L, pW/cm
pdt Ed eu re ge Len,
0 500 1000 1500 2000 2500
Wavelength, A, nm
Figure 1. A radiance spectrum of one forested pixel recorded by
AVIRIS.
There is a wide range of laboratory sensors that are capable of
recording the spectrum of a leaf or a sample of dried and ground
foliage. In addition, high-quality field and airborne sensors have been
developed which can record vegetation canopy spectra (Goetz, 1991).
Three sensors in particular have encouraged the remote-sensing
community to use reflectance spectra to estimate foliar chemical
concentration. These sensors are the Airborne Imaging
Spectrometers | and Il (AIS) (Vane and Goetz, 1988), the Airborne
Visible/Infrared Imaging Spectrometer (AVIRIS) (Vane, 1987) and the
satellite-borne High Resolution Imaging Spectrometer (HIRIS) (Goetz
and Herring, 1989) that is scheduled for launch early in the 21st
century (Dozier, 1991). While an ideal sensor for estimating foliar
chemical concentrations does not exist (Peterson and Hubbard, 1992),
an acceptable sensor would record a spectrum between 400-2400 nm
with a spectral resolution of 10 nm or less and a level of noise about
an order of magnitude smaller than the depth of the absorption feature