3.2 General Principles
Remote sensing measures the amount of radiation energy in the electromagnetic
spectrum that is emitted or reflected by the object or surrounding area being
observed. Passive optical remote sensing is particularly sensitive to forest foliage
which provides a synoptic view of the area of interest that enables the estimation
of biomass values over a large area. Satellite observations represent the topof-atmosphere (TOA) radiance which is a combination of top-of-canopy (TOC)
and atmospheric radiance. TOC reflectance reflects forest properties, particularly,
leaf area index (Jarvis and Leverenz 1983), which can be retrieved using vegetation radiative transfer models from satellite data by minimizing atmospheric
effects. Canopy green leaves scatter strongly solar radiation in both near-infrared
wavelength (0.7–1.3 lm) with a value of about 40–50 % of incident light and
green wavelength while the leaves absorb radiation in blue and red wavelengths by
chlorophyll and foliage water (Hofer 1978; Ripple 1986). The radiation measured
in spectral bands helps us to distinguish the forest properties. A combination of
two or more spectral bands produces a vegetation index (VI), which can be calculated by rationing, differencing, rationing differences, and linear combinations of
spectral bands. Vegetation index generally enhances vegetation signal while it
minimizes the influences from solar irradiance, solar angle, sensor view angle,
atmospheric and soil background effects.
Spectral reflectance and vegetation index characterize forest properties and
forest biomass. Biomass is basically calculated using the density of unit biomass
and the area of forest growth. The unit biomass of AGB (foliage, branch, and stem)
can be estimated from optical remote sensing in two different ways. First, forest
biomass is generally estimated using models that are statistically established in a
relationship between spectral responses and field samples of biomass measurements. The models are generated using either regression analyses or non-parametric
imputation approaches. The model parameters or coefficients are affected by
various factors that include the atmosphere, sun angle, satellite view angle,
phenological state of vegetation growth at the time of image acquisition, topography, and imperfections in radiometric calibration and geo-metric registration.
Second, forest biomass is calculated using satellite-based allometric models. Such
models are physically meaningful because biomass is associated with forest components (attributes) which include leaf area index (LAI) and canopy structure
(crown closure and height). These components can be directly estimated from
optical remotely sensed data.
Radar wavelengths range from less than 1 mm to 1 m. They are sensitive to
dielectric and geometrical properties of forests, and are thus more closely related to
measurements of AGB than optical data, which mainly respond to chemical
properties of the vegetation constituents. Theory and observations show that the
radar backscattering coefficient (i.e. the normalized backscattered power) varies
with increasing forest biomass for lower levels of biomass, but saturates (remains
approximately constant) for higher levels.
3 Remote Sensing of Forest Biomass
65
Remote sensing measures the amount of radiation energy in the electromagnetic
spectrum that is emitted or reflected by the object or surrounding area being
observed. Passive optical remote sensing is particularly sensitive to forest foliage
which provides a synoptic view of the area of interest that enables the estimation
of biomass values over a large area. Satellite observations represent the topof-atmosphere (TOA) radiance which is a combination of top-of-canopy (TOC)
and atmospheric radiance. TOC reflectance reflects forest properties, particularly,
leaf area index (Jarvis and Leverenz 1983), which can be retrieved using vegetation radiative transfer models from satellite data by minimizing atmospheric
effects. Canopy green leaves scatter strongly solar radiation in both near-infrared
wavelength (0.7–1.3 lm) with a value of about 40–50 % of incident light and
green wavelength while the leaves absorb radiation in blue and red wavelengths by
chlorophyll and foliage water (Hofer 1978; Ripple 1986). The radiation measured
in spectral bands helps us to distinguish the forest properties. A combination of
two or more spectral bands produces a vegetation index (VI), which can be calculated by rationing, differencing, rationing differences, and linear combinations of
spectral bands. Vegetation index generally enhances vegetation signal while it
minimizes the influences from solar irradiance, solar angle, sensor view angle,
atmospheric and soil background effects.
Spectral reflectance and vegetation index characterize forest properties and
forest biomass. Biomass is basically calculated using the density of unit biomass
and the area of forest growth. The unit biomass of AGB (foliage, branch, and stem)
can be estimated from optical remote sensing in two different ways. First, forest
biomass is generally estimated using models that are statistically established in a
relationship between spectral responses and field samples of biomass measurements. The models are generated using either regression analyses or non-parametric
imputation approaches. The model parameters or coefficients are affected by
various factors that include the atmosphere, sun angle, satellite view angle,
phenological state of vegetation growth at the time of image acquisition, topography, and imperfections in radiometric calibration and geo-metric registration.
Second, forest biomass is calculated using satellite-based allometric models. Such
models are physically meaningful because biomass is associated with forest components (attributes) which include leaf area index (LAI) and canopy structure
(crown closure and height). These components can be directly estimated from
optical remotely sensed data.
Radar wavelengths range from less than 1 mm to 1 m. They are sensitive to
dielectric and geometrical properties of forests, and are thus more closely related to
measurements of AGB than optical data, which mainly respond to chemical
properties of the vegetation constituents. Theory and observations show that the
radar backscattering coefficient (i.e. the normalized backscattered power) varies
with increasing forest biomass for lower levels of biomass, but saturates (remains
approximately constant) for higher levels.
3 Remote Sensing of Forest Biomass
65