10.3 Sources of Remotely Sensed Data
collective brightness, texture, pattern, size, shape,
shadow, and association (Avery and Berlin, 1992).
Deriving information about vegetation and land
cover is one of the primary applications of remotely
sensed data. Several attributes can be derived from
remotely sensed data that describe vegetation.
Some characteristics are described in the following, followed by a discussion of accuracy assessment, without which a map of vegetation characteristics has limited utility.
Canopy Structure
Canopy structure describes the vertical structure of
the vegetation, either single story or multistory. The
vegetation of a single-story canopy is generally a
uniform height, whereas a multistory canopy has
overstory and understory components. A singlestory canopy tends to have a smooth velvet texture
in remotely sensed data, such as Landsat TM imagery, whereas a multistory canopy may have a
rougher, irregular appearance.
Tree Size Class
Tree size class may be defined by the diameter of
the trees that constitute the dominant or overs tory
canopy. To derive information on tree size class
from remotely sensed data, we must assume that
tree diameters are correlated with tree height, because diameters usually cannot be measured
through remote sensing due to blockage by the
overstory canopy. Typically, tree diameters are
grouped into size classes, such as seedling, sapling,
pole, and sawtimber, based on ground measurements. Size class determination requires substantially more field data than other attributes because
of the complex relation between canopy reflectance
and size class. Field data necessary to derive this
attribute would include sampling of the diameters
of the trees in the canopy across the full range of
forest cover types. Problems associated with estimating tree size are discussed by Woodcock et al.
(1994).
Percent Crown Closure
Percent crown closure refers to the area covered by
the forest canopy. Determining percent crown closure can be complicated in multistory forest
canopies because of shadows visible on remotely
sensed data. In addition, low crown closures (less
than 15%) are hard to estimate because they are
spectrally very similar to bare ground. Percent
crown closure is a characteristic that is more consistently estimated from remote sensing than on the
141
ground due to the synoptic perspective provided by
airborne and satellite sensors. Cohen and Spies
(1992) assessed the utility of satellite data for
analysis and inventory of stand attributes of Douglas fir-western hemlock forests in western Oregon and Washington. The authors concluded that
of all the stand attributes evaluated the following
were most reliably estimated using satellite data:
standard deviation of tree sizes, mean tree size, tree
density, a structural complexity index, and stand
age. These attributes were for trees in the upper
canopy layer of the sample stands.
Vegetation Composition
Species groups or associations are often used to define vegetation composition. The level of detail
possible in determining vegetation composition is
highly correlated to the spatial and spectral resolution of the remotely sensed imagery and the types
being distinguished. Low spatial resolution data
such as A VHRR can be used to identify very broad
species groups. Landsat TM and aerial photos can
supply more detailed species groups. Congalton et
al. (1993) used satellite imagery and aerial photographs to produce a GIS database and map
old-growth forest lands in western Oregon and
Washington. GIS data layers derived from remotely
sensed data included crown closure, size classstand structure, species, and current-vegetationtype polygons.
Accuracy Assessments
Accuracy assessments are an essential part of all
remote sensing projects. First, they enable the user
to compare different methods and sensors. Second,
they provide information regarding the reliability
and usefulness of remote sensing techniques for a
particular application. Finally, and most importantly, accuracy assessments support spatial data
used in decision-making processes. Stehman and
Czaplewski (1998) discuss the basic components
and options available for conducting accuracy assessment of thematic maps constructed from remotely sensed data. Edwards et al. (1998) present
an accuracy assessment methodology applied to a
cover-type map developed for the Utah Gap Analysis. Further information on accuracy assessment
can also be found in Congalton and Green (1998).
National Vegetation Classification System
The Federal Geographic Data Committee (1997a)
approved a classification system to support the production of uniform statistics on vegetation resources
collective brightness, texture, pattern, size, shape,
shadow, and association (Avery and Berlin, 1992).
Deriving information about vegetation and land
cover is one of the primary applications of remotely
sensed data. Several attributes can be derived from
remotely sensed data that describe vegetation.
Some characteristics are described in the following, followed by a discussion of accuracy assessment, without which a map of vegetation characteristics has limited utility.
Canopy Structure
Canopy structure describes the vertical structure of
the vegetation, either single story or multistory. The
vegetation of a single-story canopy is generally a
uniform height, whereas a multistory canopy has
overstory and understory components. A singlestory canopy tends to have a smooth velvet texture
in remotely sensed data, such as Landsat TM imagery, whereas a multistory canopy may have a
rougher, irregular appearance.
Tree Size Class
Tree size class may be defined by the diameter of
the trees that constitute the dominant or overs tory
canopy. To derive information on tree size class
from remotely sensed data, we must assume that
tree diameters are correlated with tree height, because diameters usually cannot be measured
through remote sensing due to blockage by the
overstory canopy. Typically, tree diameters are
grouped into size classes, such as seedling, sapling,
pole, and sawtimber, based on ground measurements. Size class determination requires substantially more field data than other attributes because
of the complex relation between canopy reflectance
and size class. Field data necessary to derive this
attribute would include sampling of the diameters
of the trees in the canopy across the full range of
forest cover types. Problems associated with estimating tree size are discussed by Woodcock et al.
(1994).
Percent Crown Closure
Percent crown closure refers to the area covered by
the forest canopy. Determining percent crown closure can be complicated in multistory forest
canopies because of shadows visible on remotely
sensed data. In addition, low crown closures (less
than 15%) are hard to estimate because they are
spectrally very similar to bare ground. Percent
crown closure is a characteristic that is more consistently estimated from remote sensing than on the
141
ground due to the synoptic perspective provided by
airborne and satellite sensors. Cohen and Spies
(1992) assessed the utility of satellite data for
analysis and inventory of stand attributes of Douglas fir-western hemlock forests in western Oregon and Washington. The authors concluded that
of all the stand attributes evaluated the following
were most reliably estimated using satellite data:
standard deviation of tree sizes, mean tree size, tree
density, a structural complexity index, and stand
age. These attributes were for trees in the upper
canopy layer of the sample stands.
Vegetation Composition
Species groups or associations are often used to define vegetation composition. The level of detail
possible in determining vegetation composition is
highly correlated to the spatial and spectral resolution of the remotely sensed imagery and the types
being distinguished. Low spatial resolution data
such as A VHRR can be used to identify very broad
species groups. Landsat TM and aerial photos can
supply more detailed species groups. Congalton et
al. (1993) used satellite imagery and aerial photographs to produce a GIS database and map
old-growth forest lands in western Oregon and
Washington. GIS data layers derived from remotely
sensed data included crown closure, size classstand structure, species, and current-vegetationtype polygons.
Accuracy Assessments
Accuracy assessments are an essential part of all
remote sensing projects. First, they enable the user
to compare different methods and sensors. Second,
they provide information regarding the reliability
and usefulness of remote sensing techniques for a
particular application. Finally, and most importantly, accuracy assessments support spatial data
used in decision-making processes. Stehman and
Czaplewski (1998) discuss the basic components
and options available for conducting accuracy assessment of thematic maps constructed from remotely sensed data. Edwards et al. (1998) present
an accuracy assessment methodology applied to a
cover-type map developed for the Utah Gap Analysis. Further information on accuracy assessment
can also be found in Congalton and Green (1998).
National Vegetation Classification System
The Federal Geographic Data Committee (1997a)
approved a classification system to support the production of uniform statistics on vegetation resources