combination of optical and microwave imageries (1-km resolution) of leaf area
index, vegetation cover, vegetation types and topography. The uncertainty of
biomass estimate ranges from 6 to 53 % at the pixel level (100 ha) and 5 and 1 % at
the typical project (10,000 ha) and national ([1,000,000 ha) scales, respectively.
Moreover, a multisensory system has been developed to fully probe forest
structure, function and composition of ecosystems at the macroscale (Asner et al.
2012a). For example, the Carnegie Airborne Observatory (CAO) Airborne Taxonomic Mapping System (AToMS) was developed, which includes a high fidelity
visible-to-shortwave infrared (VSWIR) imaging spectrometer (280–2,510 nm),
dual-laser waveform lidar scanner, and high spatial resolution visible-to-near
infrared (VNIR) imaging spectrometer (365–1,052 nm). CAO-2 AToMS is a
newer version of CAO Alpha system (Asner et al. 2007), which can measure not
only high spatial resolution AGB (Asner et al. 2012b), but also ecosystem physiology, biogeochemistry, species and biodiversity.
There are several other fusion studies in biomass estimates using empirical and
physical methods. Hudak et al. (2002) developed an empirical relationship
between VIR and lidar data based on kriging and cokriging, which concluded that
the spacing of the lidar data should be \250 m for accurate extrapolation. Kellendorfer et al. (2004) extrapolated lidar heights by regressing with SRTM (InSAR), Landsat (tasseled-cap), and a canopy density layer, which resulted in a
RMSE of 3 meters. Hyde et al. (2007) and Nelson et al. (2007) developed linear
regression models to relate biomass with lidar height metrics, low frequency, low
wavelength (VHF), GOESAR (a dual-frequency, dual-polarimetic interferometric
airborne SAR instrument), and SAR data, which proved that lidar is most useful
for predicting forest biomass and radar adds little improvement in biomass estimation. Slatton et al. (2001) used a physical modeling with Kalman Filter based
multiscale estimation to retrieve surface topography and vegetation height from
lidar and InSAR data, which demonstrated significant improvement of bare surface
topography and vegetation height estimates obtained from InSAR alone. Kimes
et al. (2006) studied fusion of lidar with multi-angle data using an optical model to
exploit both spectral information and tree structure.
3.4 Validation Efforts Using In-Situ Measurements
The assessment and validation of forest biomass obtained from remote sensing is a
critical but challenging task. This requires a large set of reliable in situ data or
other estimates of biomass. In-situ measurements are generally obtained using
either destructive method or species-specific allometric models in field plots as
described in Sect. 3.1. The plots should be designed in homogeneous areas located
with GPS (Sarker and Nichol 2011; Soenen et al. 2010; Heiskanen 2006a).
Because field measurements are time consuming and expensive, the plot size is
generally small (\30 m) and the plot number is limited. These types of data are
84
X. Zhang and W. Ni-meister
index, vegetation cover, vegetation types and topography. The uncertainty of
biomass estimate ranges from 6 to 53 % at the pixel level (100 ha) and 5 and 1 % at
the typical project (10,000 ha) and national ([1,000,000 ha) scales, respectively.
Moreover, a multisensory system has been developed to fully probe forest
structure, function and composition of ecosystems at the macroscale (Asner et al.
2012a). For example, the Carnegie Airborne Observatory (CAO) Airborne Taxonomic Mapping System (AToMS) was developed, which includes a high fidelity
visible-to-shortwave infrared (VSWIR) imaging spectrometer (280–2,510 nm),
dual-laser waveform lidar scanner, and high spatial resolution visible-to-near
infrared (VNIR) imaging spectrometer (365–1,052 nm). CAO-2 AToMS is a
newer version of CAO Alpha system (Asner et al. 2007), which can measure not
only high spatial resolution AGB (Asner et al. 2012b), but also ecosystem physiology, biogeochemistry, species and biodiversity.
There are several other fusion studies in biomass estimates using empirical and
physical methods. Hudak et al. (2002) developed an empirical relationship
between VIR and lidar data based on kriging and cokriging, which concluded that
the spacing of the lidar data should be \250 m for accurate extrapolation. Kellendorfer et al. (2004) extrapolated lidar heights by regressing with SRTM (InSAR), Landsat (tasseled-cap), and a canopy density layer, which resulted in a
RMSE of 3 meters. Hyde et al. (2007) and Nelson et al. (2007) developed linear
regression models to relate biomass with lidar height metrics, low frequency, low
wavelength (VHF), GOESAR (a dual-frequency, dual-polarimetic interferometric
airborne SAR instrument), and SAR data, which proved that lidar is most useful
for predicting forest biomass and radar adds little improvement in biomass estimation. Slatton et al. (2001) used a physical modeling with Kalman Filter based
multiscale estimation to retrieve surface topography and vegetation height from
lidar and InSAR data, which demonstrated significant improvement of bare surface
topography and vegetation height estimates obtained from InSAR alone. Kimes
et al. (2006) studied fusion of lidar with multi-angle data using an optical model to
exploit both spectral information and tree structure.
3.4 Validation Efforts Using In-Situ Measurements
The assessment and validation of forest biomass obtained from remote sensing is a
critical but challenging task. This requires a large set of reliable in situ data or
other estimates of biomass. In-situ measurements are generally obtained using
either destructive method or species-specific allometric models in field plots as
described in Sect. 3.1. The plots should be designed in homogeneous areas located
with GPS (Sarker and Nichol 2011; Soenen et al. 2010; Heiskanen 2006a).
Because field measurements are time consuming and expensive, the plot size is
generally small (\30 m) and the plot number is limited. These types of data are
84
X. Zhang and W. Ni-meister