most important process includes obtaining as pure as possible signals from the target,
which can contain a mixture of scene elements if proper height adjustments of the FoV
are not performed. The FoV should be broad enough to cover a homogeneous species
composition. At the same time, the FoV should minimize the contribution of the
underlying soil to the spectral response of the target or should represent a pixel on the
ground (Ben-Dor et al., 2009). Multiple relative reflectance readings on different
positions in close proximity over the canopy can also be recorded, following a white
panel reading, in order to account for the variability contained in a single pixel of the
currently offered image products (Manakos et al., 2010; Manevski et al., 2012).
Emphasis is also placed on the collection of metadata that describe all the
instrumentation and environmental factors. Such data provide an evaluation of
the suitability of the collected spectral data sets for potentially new applications in
the future (Hueni et al., 2009). Digital photos of the vegetation offer benefits in the
interpretation stage, such as screening the vegetation state and composition (Milton
et al., 2006, 2009; Zomer et al., 1999). Details regarding the measurement process
also have to be documented in specific measurement field logs (Salisbury, 1998). If
platforms are reflective, a common practice is the application of low reflective
materials to the wider spectral range possible on it to suppress the scattered light
(Zomer et al., 2009). The self-shadow effect is usually prevented by positioning the
platform and the sensor toward the south for the North Hemisphere and vice versa for
the South Hemisphere.
15.2.2 Developing Field Spectral Library
At the field scale, fundamental to the spectral discrimination between land cover is the
success in extracting pure spectra of the targets. That can be achieved if parameters
that affect the measurement environment are considered, including the time of the
measurement which reflects the phenological status of vegetation land cover. Such
information, organized and stored together with the spectral signatures, is called a
spectral library (Hueni et al., 2009). The field scale spectral signatures are related to
different geo–bio–chemo–physical parameters of the land cover analyzed, such as the
soil or understory vegetation cover in the background surface, the phenology of the
vegetation, shades, pigments, and water content (Williams, 1991). A special difficulty
is that land cover parameters vary in time and space and are difficult to be controlled
on the field (Chang et al., 2005; Schaaf, 2009). That increases the variability of the
field spectra reflectance data and additionally contributes to the difficulties in
transferability of the spectral libraries and the spectral discrimination of different
land cover. Therefore, it is understandable that a commonly accepted and clear
methodology for construction of a spectral library is a challenging task and of crucial
importance for ensuring adequate data quality and relevance to the application
considered each time (Milton et al., 2009; Salvaggio et al., 2005). Most efforts
are concentrated on obtaining unaltered field reflectance spectra in order to better
characterize the spectral signatures of diferent land cover. The most common types of
spectral libraries recorded at the field scale are presented next, providing also
reference to examples of spectral libraries available today.
292
HYPERSPECTRAL REMOTE SENSING WITH EMPHASIS ON LAND COVER MAPPING
which can contain a mixture of scene elements if proper height adjustments of the FoV
are not performed. The FoV should be broad enough to cover a homogeneous species
composition. At the same time, the FoV should minimize the contribution of the
underlying soil to the spectral response of the target or should represent a pixel on the
ground (Ben-Dor et al., 2009). Multiple relative reflectance readings on different
positions in close proximity over the canopy can also be recorded, following a white
panel reading, in order to account for the variability contained in a single pixel of the
currently offered image products (Manakos et al., 2010; Manevski et al., 2012).
Emphasis is also placed on the collection of metadata that describe all the
instrumentation and environmental factors. Such data provide an evaluation of
the suitability of the collected spectral data sets for potentially new applications in
the future (Hueni et al., 2009). Digital photos of the vegetation offer benefits in the
interpretation stage, such as screening the vegetation state and composition (Milton
et al., 2006, 2009; Zomer et al., 1999). Details regarding the measurement process
also have to be documented in specific measurement field logs (Salisbury, 1998). If
platforms are reflective, a common practice is the application of low reflective
materials to the wider spectral range possible on it to suppress the scattered light
(Zomer et al., 2009). The self-shadow effect is usually prevented by positioning the
platform and the sensor toward the south for the North Hemisphere and vice versa for
the South Hemisphere.
15.2.2 Developing Field Spectral Library
At the field scale, fundamental to the spectral discrimination between land cover is the
success in extracting pure spectra of the targets. That can be achieved if parameters
that affect the measurement environment are considered, including the time of the
measurement which reflects the phenological status of vegetation land cover. Such
information, organized and stored together with the spectral signatures, is called a
spectral library (Hueni et al., 2009). The field scale spectral signatures are related to
different geo–bio–chemo–physical parameters of the land cover analyzed, such as the
soil or understory vegetation cover in the background surface, the phenology of the
vegetation, shades, pigments, and water content (Williams, 1991). A special difficulty
is that land cover parameters vary in time and space and are difficult to be controlled
on the field (Chang et al., 2005; Schaaf, 2009). That increases the variability of the
field spectra reflectance data and additionally contributes to the difficulties in
transferability of the spectral libraries and the spectral discrimination of different
land cover. Therefore, it is understandable that a commonly accepted and clear
methodology for construction of a spectral library is a challenging task and of crucial
importance for ensuring adequate data quality and relevance to the application
considered each time (Milton et al., 2009; Salvaggio et al., 2005). Most efforts
are concentrated on obtaining unaltered field reflectance spectra in order to better
characterize the spectral signatures of diferent land cover. The most common types of
spectral libraries recorded at the field scale are presented next, providing also
reference to examples of spectral libraries available today.
292
HYPERSPECTRAL REMOTE SENSING WITH EMPHASIS ON LAND COVER MAPPING