(plot) basis (Brown et al. 1989; Brown and Iverson 1992; Brown and Lugo 1992;
Gillespie et al. 1992). This method entails harvesting plants, drying them, and then
weighing the biomass. The destructive measurement is most accurate but it is very
expensive and time consuming.
A non-destructive method is first to measure tree variables, such as canopy
crown size, crown depth, tree height, and stem diameter. These components on
randomly selected sample trees are then converted to tree biomass using allometric
models (Brown et al. 1989; Brown and Iverson 1992; Brown and Lugo 1992;
Gillespie et al. 1992; Paster et al. 1984; Ter-Mikaelian and Korzukin 1997).
The allomatric models are developed in various forms for biomass estimates (e.g.,
Ter-Mikaelian and Korzukin 1997; Paster et al. 1984), but the simplest and most
commonly used model is:
AGB ¼ aD
c
ð3:1Þ
where AGB is above-ground biomass (kg), D is the diameter of tree at breast height
(DBH) (m), a and c are coefficients.
Field data of biomass are commonly measured based on sample plots designed
for a specified study. Such field data can be aggregated to generate National Forest
Inventories (NFI) (Brown et al. 1999; Jenkins et al. 2001; Chojnacky et al. 2004).
The NFI dataset currently contains the most accurate biomass estimates in various
countries, such as in Finland and Sweden (Tomppo 1991; Reese et al. 2003),
Norway (Gjersten 2005), Austria (Koukal et al. 2005), New Zealand (Tomppo
et al. 1999), China (Tomppo et al. 2001), Germany (Diemer et al. 2000), Italy
(Maselli et al. 2005), and the United States (Franco-Lopez et al. 2001; McRoberts
2006; McRoberts et al. 2002, 2007). However, it is challenging to extrapolate plot
estimates to unit ground area of high-quality geo-referenced ground-truth (Gibbs
et al. 2007; Goetz et al. 2009).
3.3.2 Forest Biomass from Passive Optical Remote Sensing
Forest biomass is a function of remote sensing metrics that are closely related to
vegetation function (leaf area, volume, photosynthetic activity) and horizontal
structure (crown cover). Particularly, forest biomass is estimated using spectral
reflectance and vegetation indices from various satellite instruments of passive
optical remote sensing. Both empirical regression techniques (Hall et al. 2006;
Jakubauskas and Price 1997; Lefsky et al. 2001; Rahman et al. 2008; Zheng et al.
2004; Labrecque et al. 2006; Powell et al. 2010) and nonlinear nonparametric
approaches (Baccini et al. 2004; Fraser and Li 2002) are developed to estimate
forest biomass following the basic strategy as presented in Fig. 3.1. In contrast,
satellite-based allometric models calculate forest biomass using forest attributes
derived from satellite data, which is described in Fig. 3.2 (Zhang and Kondragunta
2006; Soenen et al. 2010).
3 Remote Sensing of Forest Biomass
67
Gillespie et al. 1992). This method entails harvesting plants, drying them, and then
weighing the biomass. The destructive measurement is most accurate but it is very
expensive and time consuming.
A non-destructive method is first to measure tree variables, such as canopy
crown size, crown depth, tree height, and stem diameter. These components on
randomly selected sample trees are then converted to tree biomass using allometric
models (Brown et al. 1989; Brown and Iverson 1992; Brown and Lugo 1992;
Gillespie et al. 1992; Paster et al. 1984; Ter-Mikaelian and Korzukin 1997).
The allomatric models are developed in various forms for biomass estimates (e.g.,
Ter-Mikaelian and Korzukin 1997; Paster et al. 1984), but the simplest and most
commonly used model is:
AGB ¼ aD
c
ð3:1Þ
where AGB is above-ground biomass (kg), D is the diameter of tree at breast height
(DBH) (m), a and c are coefficients.
Field data of biomass are commonly measured based on sample plots designed
for a specified study. Such field data can be aggregated to generate National Forest
Inventories (NFI) (Brown et al. 1999; Jenkins et al. 2001; Chojnacky et al. 2004).
The NFI dataset currently contains the most accurate biomass estimates in various
countries, such as in Finland and Sweden (Tomppo 1991; Reese et al. 2003),
Norway (Gjersten 2005), Austria (Koukal et al. 2005), New Zealand (Tomppo
et al. 1999), China (Tomppo et al. 2001), Germany (Diemer et al. 2000), Italy
(Maselli et al. 2005), and the United States (Franco-Lopez et al. 2001; McRoberts
2006; McRoberts et al. 2002, 2007). However, it is challenging to extrapolate plot
estimates to unit ground area of high-quality geo-referenced ground-truth (Gibbs
et al. 2007; Goetz et al. 2009).
3.3.2 Forest Biomass from Passive Optical Remote Sensing
Forest biomass is a function of remote sensing metrics that are closely related to
vegetation function (leaf area, volume, photosynthetic activity) and horizontal
structure (crown cover). Particularly, forest biomass is estimated using spectral
reflectance and vegetation indices from various satellite instruments of passive
optical remote sensing. Both empirical regression techniques (Hall et al. 2006;
Jakubauskas and Price 1997; Lefsky et al. 2001; Rahman et al. 2008; Zheng et al.
2004; Labrecque et al. 2006; Powell et al. 2010) and nonlinear nonparametric
approaches (Baccini et al. 2004; Fraser and Li 2002) are developed to estimate
forest biomass following the basic strategy as presented in Fig. 3.1. In contrast,
satellite-based allometric models calculate forest biomass using forest attributes
derived from satellite data, which is described in Fig. 3.2 (Zhang and Kondragunta
2006; Soenen et al. 2010).
3 Remote Sensing of Forest Biomass
67