Hyperspectral Sensors and Applications
35
moved or mitigated using techniques such as spectral mixture analysis (SMA).
Furthermore, information on both the biochemical and biophysical properties of canopies can still be retrieved, despite the degradation of the 'pure'
spectral reflectance feature. Differences in these attributes can also be used to
discriminate vegetation species and/or communities using either single date
or multi-temporal imagery. Such applications are outlined in more detail in
the following sections.
1.6.2.2
Foliar Biochemistry and Biophysical Properties
A major focus of hyper spectral remote sensing has been the retrieval of foliage
chemicals, particularly those that play critical roles in ecosystem processes,
such as chlorophyll, nitrogen and carbon (Daughtry et al. 2000; Coops et al.
2001). Most retrievals have been based on simple, conventional stepwise or
multiple regression between chemical quantities (measured using wet or dry
chemistry based techniques) and reflectance (R) or absorbance (calculated as
log 10 (i/R) ) data or band ratios of spectral reflectance, which themselves are
usually correlated to measures of green vegetation amount, cover or functioning (Gitelson and Merzlyak 1997). Modified partial least squares (MPLS), neural network, statistical methods (Niemann and Goodenough 2003) and model
inversion techniques (Demarez and Gastellu-Etchegorry 2000; Jacquemoud et
al. 2000) have also been used to assist retrieval. MPLS has proved particularly
useful as the information content of hundreds of bands can be concentrated
within a few variables, although the optimal predictive bands still need to
be identified using correlograms of spectra and biochemical concentration or
regression equations from the MPLS analysis.
Although a wide range of foliar chemicals exist in the leaf, hyperspectral
remote sensing has shown greatest promise for the retrieval of chlorophyll
a and b (Clevers 1994; Gitelson and Merzlyak 1997), nitrogen (Curran 1989;
Matson et al. 1994; Gastellu-Etchegorry et al. 1995; Johnson and Billow 1996),
carbon (Ustin et al. 2001), cellulose (Zagolski et al. 1996), lignin (GastelluEtchegorry et al. 1995), anthocyanin, starch and water (Curran et al. 1992;
Serrano et al. 2000), and sideroxylonal-A (Blackburn 1999; Ebbers et al. 2002).
Other biochemicals which exhibit clearly identifiable absorption features and
are becoming increasingly important to the understanding of photosynthetic
and other leaf biochemical processes include: yellow carotenes and pale yellow
xanthophyll pigments (strong absorptions in the blue wavelengths), carotene
absorption (strong absorption at ~ 450 nm), phycocyanin pigment (absorbs
primarily in the green and red regions at ~ 620 nm) and phycoerythrin (strong
absorption at ~ 550 nm). However, chlorophyll pigments are dominant and
normally mask these pigments. During senescence or severe stress, the chlorophyll dominance may be lost, causing the other pigments to become dominant
(e. g. at leaf fall). Anthocyanin may also be produced in autumn causing leaves
to appear bright red when observed in visible wavelengths.
A number of studies have successfully retrieved foliar chemicals from airborne and spaceborne hyperspectral data. As examples, foliar nitrogen concen-
35
moved or mitigated using techniques such as spectral mixture analysis (SMA).
Furthermore, information on both the biochemical and biophysical properties of canopies can still be retrieved, despite the degradation of the 'pure'
spectral reflectance feature. Differences in these attributes can also be used to
discriminate vegetation species and/or communities using either single date
or multi-temporal imagery. Such applications are outlined in more detail in
the following sections.
1.6.2.2
Foliar Biochemistry and Biophysical Properties
A major focus of hyper spectral remote sensing has been the retrieval of foliage
chemicals, particularly those that play critical roles in ecosystem processes,
such as chlorophyll, nitrogen and carbon (Daughtry et al. 2000; Coops et al.
2001). Most retrievals have been based on simple, conventional stepwise or
multiple regression between chemical quantities (measured using wet or dry
chemistry based techniques) and reflectance (R) or absorbance (calculated as
log 10 (i/R) ) data or band ratios of spectral reflectance, which themselves are
usually correlated to measures of green vegetation amount, cover or functioning (Gitelson and Merzlyak 1997). Modified partial least squares (MPLS), neural network, statistical methods (Niemann and Goodenough 2003) and model
inversion techniques (Demarez and Gastellu-Etchegorry 2000; Jacquemoud et
al. 2000) have also been used to assist retrieval. MPLS has proved particularly
useful as the information content of hundreds of bands can be concentrated
within a few variables, although the optimal predictive bands still need to
be identified using correlograms of spectra and biochemical concentration or
regression equations from the MPLS analysis.
Although a wide range of foliar chemicals exist in the leaf, hyperspectral
remote sensing has shown greatest promise for the retrieval of chlorophyll
a and b (Clevers 1994; Gitelson and Merzlyak 1997), nitrogen (Curran 1989;
Matson et al. 1994; Gastellu-Etchegorry et al. 1995; Johnson and Billow 1996),
carbon (Ustin et al. 2001), cellulose (Zagolski et al. 1996), lignin (GastelluEtchegorry et al. 1995), anthocyanin, starch and water (Curran et al. 1992;
Serrano et al. 2000), and sideroxylonal-A (Blackburn 1999; Ebbers et al. 2002).
Other biochemicals which exhibit clearly identifiable absorption features and
are becoming increasingly important to the understanding of photosynthetic
and other leaf biochemical processes include: yellow carotenes and pale yellow
xanthophyll pigments (strong absorptions in the blue wavelengths), carotene
absorption (strong absorption at ~ 450 nm), phycocyanin pigment (absorbs
primarily in the green and red regions at ~ 620 nm) and phycoerythrin (strong
absorption at ~ 550 nm). However, chlorophyll pigments are dominant and
normally mask these pigments. During senescence or severe stress, the chlorophyll dominance may be lost, causing the other pigments to become dominant
(e. g. at leaf fall). Anthocyanin may also be produced in autumn causing leaves
to appear bright red when observed in visible wavelengths.
A number of studies have successfully retrieved foliar chemicals from airborne and spaceborne hyperspectral data. As examples, foliar nitrogen concen-