squares regression. After derived from a geometric-optical canopy reflectance
model (Soenen et al. 2010; Chopping et al. 2011), SA is applied to calculate
biomass based on general allometric relationships in a large regional area. This
approach has been demonstrated to successfully calculate biomass in Canada using
10 m multispectral SPOT data with a plot-based accuracy of 31.7 Mg/ha, which is
superior to empirical methods (NDVI and shadow fraction) (Soenen et al. 2010).
Similarly, the canopy structure parameters have also been retrieved from MODIS
and MISR data for biomass calculation (Chopping et al. 2011).
The satellite-based generalized allometric model has advantages over others
although the model construction requires a set of spatially representative samples
acquired from field plots. Once the parameters of canopy properties are derived
from satellite data, the model could be applied to calculate forest biomass in a
broad area without regeneration and calibration of models. As a result, the biomass
can be easily updated interannually. Moreover, this approach does not require
geo-referenced plots to match satellite-derived parameters. This avoids the model
errors caused by matching samples between field measurements and satellite
pixels.
3.3.3 Forest Biomass from Radar
Radar data physically measure biomass through the interaction of the radar waves
with tree scattering elements. The widely used active radar data are from spaceborne synthetic aperture radar (SAR) sensors, such as the L-band ALOS PALSAR,
the C-band ERS/SAR, RADARSAT/SAR or ENVISAT/ASAR and the X-band
TerraSAR-X instruments, which transmit microwave energy at wavelengths from
3.0 (X-band) to 23.6 cm (L-band). The proposed ESAEarth Explorer Mission
BIOMASS is the prime candidate to be the first P-band SAR satellite (Le Toan
et al. 2011). The major advantage of all SAR systems is their weather- and daylight-independency.
The ability of radar sensors to measure biomass mainly depends on how deep
the radar signals can penetrate into the canopy. The longer the wavelength is, the
deeper the penetration is. The L- and P-band backscatter, particularly single
polarization HV (horizontal transmit and vertical receive) and HH (horizontal
transmit and horizontal receive) polarized backscatter, is strongly dependent on
biomass amount (e.g., Le Toan et al. 1992; Ranson and Sun 1994; Imhoff 1995;
Saatchi et al. 2007a, b). P-band backscatter shows stronger dependence on biomass
than L-band backscatter. The radar backscatter increases approximately linearly
with increasing biomass until it is saturated at a certain biomass level that varies
with the radar wavelength (Dobson et al. 1992). The biomass level for backscatter
saturation is about 200 Mg/ha at P-band, 100 Mg/ha at L-band, and 30–50 Mg/ha
at X- and C-bands (Le Toan et al. 2011).
The observed relationship between radar backscatter and biomass can be
physically illustrated using electromagnetic scattering models (Ulaby et al. 1990;
3 Remote Sensing of Forest Biomass
79
model (Soenen et al. 2010; Chopping et al. 2011), SA is applied to calculate
biomass based on general allometric relationships in a large regional area. This
approach has been demonstrated to successfully calculate biomass in Canada using
10 m multispectral SPOT data with a plot-based accuracy of 31.7 Mg/ha, which is
superior to empirical methods (NDVI and shadow fraction) (Soenen et al. 2010).
Similarly, the canopy structure parameters have also been retrieved from MODIS
and MISR data for biomass calculation (Chopping et al. 2011).
The satellite-based generalized allometric model has advantages over others
although the model construction requires a set of spatially representative samples
acquired from field plots. Once the parameters of canopy properties are derived
from satellite data, the model could be applied to calculate forest biomass in a
broad area without regeneration and calibration of models. As a result, the biomass
can be easily updated interannually. Moreover, this approach does not require
geo-referenced plots to match satellite-derived parameters. This avoids the model
errors caused by matching samples between field measurements and satellite
pixels.
3.3.3 Forest Biomass from Radar
Radar data physically measure biomass through the interaction of the radar waves
with tree scattering elements. The widely used active radar data are from spaceborne synthetic aperture radar (SAR) sensors, such as the L-band ALOS PALSAR,
the C-band ERS/SAR, RADARSAT/SAR or ENVISAT/ASAR and the X-band
TerraSAR-X instruments, which transmit microwave energy at wavelengths from
3.0 (X-band) to 23.6 cm (L-band). The proposed ESAEarth Explorer Mission
BIOMASS is the prime candidate to be the first P-band SAR satellite (Le Toan
et al. 2011). The major advantage of all SAR systems is their weather- and daylight-independency.
The ability of radar sensors to measure biomass mainly depends on how deep
the radar signals can penetrate into the canopy. The longer the wavelength is, the
deeper the penetration is. The L- and P-band backscatter, particularly single
polarization HV (horizontal transmit and vertical receive) and HH (horizontal
transmit and horizontal receive) polarized backscatter, is strongly dependent on
biomass amount (e.g., Le Toan et al. 1992; Ranson and Sun 1994; Imhoff 1995;
Saatchi et al. 2007a, b). P-band backscatter shows stronger dependence on biomass
than L-band backscatter. The radar backscatter increases approximately linearly
with increasing biomass until it is saturated at a certain biomass level that varies
with the radar wavelength (Dobson et al. 1992). The biomass level for backscatter
saturation is about 200 Mg/ha at P-band, 100 Mg/ha at L-band, and 30–50 Mg/ha
at X- and C-bands (Le Toan et al. 2011).
The observed relationship between radar backscatter and biomass can be
physically illustrated using electromagnetic scattering models (Ulaby et al. 1990;
3 Remote Sensing of Forest Biomass
79