Principal Component Analysis (PCA) is another widely applicable tool because it
compresses spectral data into a few bands associated with physical scene characteristics (Crist and Cicone 1984). It produces three important variables that are
brightness index (TC b ), greenness index (TC g ) and wetness index (TC w ). However, the TC coefficients are sensor-dependent, so that the coefficients have been
developed for Landsat TM (Crist and Cicone 1984), Landsat ETM/ETM+ (Huang
et al. 2002), MODIS data (Zhang et al. 2002), and IKONOS (Horne 2003), separately. Furthermore, the spatial texture of spectral bands, which is a characteristic
for identifying objects or regions of interest in an image, is also used to enhance
biomass estimates (Sarker and Nichol 2011). Although there are numerous vegetation indices developed (e.g., Foody et al. 2003; Heiskanen 2006a, b), Table 3.2
lists several commonly used vegetation indices for biomass estimates.
Alternatively, tree canopy attributes are derived from optical satellite data and
are considered to be effective proxies of AGB. Tree components are physically
correlated in allometric functions (Nelson et al. 1999; Phua and Saito 2003;
Popescu et al. 2003), so that several canopy variables that are retrieved from
satellites are frequently applied to estimate forest biomass.
(1) LAI is one half the total green leaf area per unit ground surface area and is
expressed in terms of square meters of leaf (half surface area) per square meter
of ground. This biophysical parameter can be related to photosynthesis,
evaporation and transpiration, rainfall interception, and carbon flux. LAI can
be estimated from spectral vegetation index using either various regression
models (e.g., Heiskanen 2006a, b) or radiative transfer (RT) algorithm
(Myneni et al. 1997; Knyazikhin et al. 1998; Myneni et al. 2002). The latter
simulates the surface reflectances (bidirectional reflectance factors) as a
function of biome type, view/illumination geometry, LAI/FPAR, canopy
structure, leaf optical properties, and soil patterns. The final LAI estimate is
the average of all acceptable solutions within specified uncertainties from the
algorithm outputs which are retrieved from all canopy and soil patterns
(Myneni et al. 2002).
(2) Canopy structure, such as tree crown size (area), height, and density, is
effective parameters in calculating foliage biomass and total standing biomass
(Franklin and Hiernaux 1991; Wu and Strahler 1994; Soenen et al. 2010). The
crown size and density in each satellite pixel can be estimated using
Li–Strahler geometric-optical (GO) canopy reflectance model (Li and Strahler
1985, 1986). The Li-Strahler model is a three-dimensional model in describing
individual plant canopies, which characterizes the variation in reflectance due
to different vegetation properties, illumination (solar) and view (sensor)
angles. This model treats vegetation cover as a collection of discrete objects
and the reflectance from vegetation cover is modeled as a function of the
pattern of plants, shadows, and soil visible from a given viewing position.
The GO model has been used to derived meaningful canopy structural
parameters from multiangle data (Zeng and Schaepman 2009; Chopping et al.
2008, 2009, 2012; Wang et al. 2011; and Laurent et al 2011). Using data from
70
X. Zhang and W. Ni-meister
compresses spectral data into a few bands associated with physical scene characteristics (Crist and Cicone 1984). It produces three important variables that are
brightness index (TC b ), greenness index (TC g ) and wetness index (TC w ). However, the TC coefficients are sensor-dependent, so that the coefficients have been
developed for Landsat TM (Crist and Cicone 1984), Landsat ETM/ETM+ (Huang
et al. 2002), MODIS data (Zhang et al. 2002), and IKONOS (Horne 2003), separately. Furthermore, the spatial texture of spectral bands, which is a characteristic
for identifying objects or regions of interest in an image, is also used to enhance
biomass estimates (Sarker and Nichol 2011). Although there are numerous vegetation indices developed (e.g., Foody et al. 2003; Heiskanen 2006a, b), Table 3.2
lists several commonly used vegetation indices for biomass estimates.
Alternatively, tree canopy attributes are derived from optical satellite data and
are considered to be effective proxies of AGB. Tree components are physically
correlated in allometric functions (Nelson et al. 1999; Phua and Saito 2003;
Popescu et al. 2003), so that several canopy variables that are retrieved from
satellites are frequently applied to estimate forest biomass.
(1) LAI is one half the total green leaf area per unit ground surface area and is
expressed in terms of square meters of leaf (half surface area) per square meter
of ground. This biophysical parameter can be related to photosynthesis,
evaporation and transpiration, rainfall interception, and carbon flux. LAI can
be estimated from spectral vegetation index using either various regression
models (e.g., Heiskanen 2006a, b) or radiative transfer (RT) algorithm
(Myneni et al. 1997; Knyazikhin et al. 1998; Myneni et al. 2002). The latter
simulates the surface reflectances (bidirectional reflectance factors) as a
function of biome type, view/illumination geometry, LAI/FPAR, canopy
structure, leaf optical properties, and soil patterns. The final LAI estimate is
the average of all acceptable solutions within specified uncertainties from the
algorithm outputs which are retrieved from all canopy and soil patterns
(Myneni et al. 2002).
(2) Canopy structure, such as tree crown size (area), height, and density, is
effective parameters in calculating foliage biomass and total standing biomass
(Franklin and Hiernaux 1991; Wu and Strahler 1994; Soenen et al. 2010). The
crown size and density in each satellite pixel can be estimated using
Li–Strahler geometric-optical (GO) canopy reflectance model (Li and Strahler
1985, 1986). The Li-Strahler model is a three-dimensional model in describing
individual plant canopies, which characterizes the variation in reflectance due
to different vegetation properties, illumination (solar) and view (sensor)
angles. This model treats vegetation cover as a collection of discrete objects
and the reflectance from vegetation cover is modeled as a function of the
pattern of plants, shadows, and soil visible from a given viewing position.
The GO model has been used to derived meaningful canopy structural
parameters from multiangle data (Zeng and Schaepman 2009; Chopping et al.
2008, 2009, 2012; Wang et al. 2011; and Laurent et al 2011). Using data from
70
X. Zhang and W. Ni-meister