Geoscience Reference
In-Depth Information
3.5.2 Biomass modeling using remote sensing data
Total biomass includes aboveground (AGB) and belowground living mass,
but due to difficulty in measuring belowground biomass, most research has
focused on the former. Approaches to measuring AGB are based on (1) field
data, (2) remote sensing, and (3) geographic information systems (GIS).
Traditional field measurement techniques are time-consuming, labor inten-
sive, and difficult to implement. Representative field samples are required as
input in developing AGB estimation models using remote sensing and/or GIS
approaches. GIS-based methods using ancillary data are also difficult to
implement because of problems in appropriate data acquisition, knowledge
about relationships between AGB and ancillary data, and impacts of environ-
mental conditions on AGB accumulation. The advantages of remote sensing
(e.g. repetitive nature of data collection, synoptic view, digital format, and
high correlations between spectral bands and vegetation parameters) make it
the primary data source for large-area AGB estimation (Foody et al. 2003 ).
Few studies have used low spatial resolution data or have used high spatial
resolution data in AGB estimation. Medium spatial resolution optical sensor
data, such as Landsat TM (Nelson et al. 2000 ; Foody et al. 2003 ) and SAR
(radar) data (Santos et al. 2003 ) are most often used for AGB estimation,
which is based on statistical relationships between AGB and TM or SAR
responses. AGB can be (1) directly estimated using remotely sensed data
implementing analytical approaches, such as multiple regression, K nearest-
neighbor (Fazakas et al. 1999 ), and neural network (Foody et al. 2003 ), and
(2) indirectly estimated from canopy parameters, such as crown diameter,
which are derived from remote sensed data using multiple regression analysis
or canopy reflectance models (Popescu et al. 2003 ).
TM spectral data used for AGB estimation have been explored in tropical
regions (Foody et al. 2003 ). Spectral responses alone often are not sufficient
to accurately estimate AGB because of the complexity of forest stand struc-
ture and environmental conditions. A combination of spectral responses and
textures is usually helpful in improving AGB estimation. The variables used
for AGB estimation vary, depending on the characteristics of the study areas.
Different forest stand structures related to soil conditions and land use history
significantly influence AGB estimation. Large differences in biomass density
and associated forest stand structure exist between succession and mature
forests that result in significantly different performance in AGB estimation.
Use of Landsat TM imagery is more successful for AGB estimation in
succession forests than in mature forests because of a less complex stand
structure than found in mature forests. Examples of biomass modeling
research conducted in eastern Amazonia study sites using TM data showed
that models developed for succession forests had R 2
values ranging from
 
Search WWH ::




Custom Search