1. Stand Structure in Terrestrial Ecosystems
Indirect Methods
As noted in the previous section, rangefinders can
be used to obtain vertical profiles of foliar density.
The main concern is the length of time needed to
obtain a sufficient number of distances to accurately
characterize foliar height frequency distribution. To
facilitate reconstruction of foliar profiles of low
vegetation, such as crops, Sinoquet et al. (1993)
mounted a mobile laser rangefinder in a 5 X 5 m
frame above the canopy, acquiring a large number
of sensor-plant distances to profile light interception in the canopy.
Koike (1985) describes a tomographic approach
for obtaining two-dimensional vertical profiles of
foliar distribution in forests using horizontal photographs taken from several locations in a section
plane of the canopy. One-dimensional profiles are
now routinely obtained by acquiring gap fraction
data at a series of heights below the canopy. As
mentioned previously, there are several methods for
analyzing gap fraction data to retrieve LAI, but all
depend on the assumption of random foliar distribution. This assumption appears reasonable for
many canopy types (Nilson 1971), an important exception being that of conifer canopies (e.g., Norman and Jarvis 1974). Bidlake and Black (1989),
working in Larix occidentalis forests, compared allometric estimates of LA! to indirect estimates using gap fraction analysis of hemispherical photographs and found that gap fraction procedures
underestimated LAI by nearly 50% due to clumping of leaves and branches. However, the estimates
were highly correlated and a calibration factor
could be used to bring the approaches into close
agreement. Martens et al. (1993) compared the onedimensional inversion model of Norman and
Campbell (1989) to the Beer-Lambert estimation
procedure (Marshall and Waring 1986) for estimating the LAI of needle-leaved and broad-leaved
trees, and found that results depended strongly on
both the instrument used to measure gap fraction
and the analytical technique. One-sided LA! estimates ranged from 2.93 to 8.98 for conifer forest
and 2.1 to 6.56 for walnut orchard, and there were
no consistent patterns that allowed for simple crosscalibration of methods.
The ability of lasers and microwaves to penetrate
plant canopies raises the possibility of obtaining information on vertical leaf distribution in forests
21
from profiling systems. Efforts to date have focused
on height retrieval, although Hyyppa and Hallikainen (1996) also estimated the average height of the
crown base with a precision of 1.5 m. Others (e.g.,
Hallikainen et al. 1993) show histograms of backscatter as a function of depth in the canopy, from
which estimates of vertical foliar distribution can
potentially be derived.
Stand Density
Direct Methods
Density, or the number of individuals per unit area,
is a common descriptor of stand structure, although
limited in its applicability to ecosystem studies,
where other measures of abundance, such as cover,
volume, and biomass, are usually desired. Bonham
(1989) provides a detailed treatment of distancebased, angle-order, transect, and plot methods for
direct measurement of density. Distance-based or
"plotless" methods are frequently used to estimate
tree and shrub density (e.g., Cottam and Curtis
1956), and can also be used to map individual plant
locations (Rohlf and Archie 1978). Such methods
are greatly expedited by optical and electromagnetic range finders, as are those methods that require
laying out large sample plots.
Aerial Indirect Methods
Strahler et al. (1986) distinguish two different scene
models for remote sensing, the high-resolution
(H-resolution) model in which the elements of the
scene are larger than individual pixels, and lowresolution (L-resolution) models in which the opposite is true. For density estimation, the elements
of interest are individual plants or plant canopies.
Most applications of remote sensing to density estimation have involved H-resolution approaches in
which individual plants are resolvable in the imagery, for example, the use of film photography or
aerial videography to estimate shrub or tree density
in open shrublands and woodlands. Accurate density measurement requires rectified imagery or
some other method to minimize effects of image
distortion on area estimation.
Image processing methods have been developed
to exploit spatial variance or texture in H-resolution
imagery. For example, Cohen et al. (1990) applied
variogram analysis to I-m digital aerial photogra-
Indirect Methods
As noted in the previous section, rangefinders can
be used to obtain vertical profiles of foliar density.
The main concern is the length of time needed to
obtain a sufficient number of distances to accurately
characterize foliar height frequency distribution. To
facilitate reconstruction of foliar profiles of low
vegetation, such as crops, Sinoquet et al. (1993)
mounted a mobile laser rangefinder in a 5 X 5 m
frame above the canopy, acquiring a large number
of sensor-plant distances to profile light interception in the canopy.
Koike (1985) describes a tomographic approach
for obtaining two-dimensional vertical profiles of
foliar distribution in forests using horizontal photographs taken from several locations in a section
plane of the canopy. One-dimensional profiles are
now routinely obtained by acquiring gap fraction
data at a series of heights below the canopy. As
mentioned previously, there are several methods for
analyzing gap fraction data to retrieve LAI, but all
depend on the assumption of random foliar distribution. This assumption appears reasonable for
many canopy types (Nilson 1971), an important exception being that of conifer canopies (e.g., Norman and Jarvis 1974). Bidlake and Black (1989),
working in Larix occidentalis forests, compared allometric estimates of LA! to indirect estimates using gap fraction analysis of hemispherical photographs and found that gap fraction procedures
underestimated LAI by nearly 50% due to clumping of leaves and branches. However, the estimates
were highly correlated and a calibration factor
could be used to bring the approaches into close
agreement. Martens et al. (1993) compared the onedimensional inversion model of Norman and
Campbell (1989) to the Beer-Lambert estimation
procedure (Marshall and Waring 1986) for estimating the LAI of needle-leaved and broad-leaved
trees, and found that results depended strongly on
both the instrument used to measure gap fraction
and the analytical technique. One-sided LA! estimates ranged from 2.93 to 8.98 for conifer forest
and 2.1 to 6.56 for walnut orchard, and there were
no consistent patterns that allowed for simple crosscalibration of methods.
The ability of lasers and microwaves to penetrate
plant canopies raises the possibility of obtaining information on vertical leaf distribution in forests
21
from profiling systems. Efforts to date have focused
on height retrieval, although Hyyppa and Hallikainen (1996) also estimated the average height of the
crown base with a precision of 1.5 m. Others (e.g.,
Hallikainen et al. 1993) show histograms of backscatter as a function of depth in the canopy, from
which estimates of vertical foliar distribution can
potentially be derived.
Stand Density
Direct Methods
Density, or the number of individuals per unit area,
is a common descriptor of stand structure, although
limited in its applicability to ecosystem studies,
where other measures of abundance, such as cover,
volume, and biomass, are usually desired. Bonham
(1989) provides a detailed treatment of distancebased, angle-order, transect, and plot methods for
direct measurement of density. Distance-based or
"plotless" methods are frequently used to estimate
tree and shrub density (e.g., Cottam and Curtis
1956), and can also be used to map individual plant
locations (Rohlf and Archie 1978). Such methods
are greatly expedited by optical and electromagnetic range finders, as are those methods that require
laying out large sample plots.
Aerial Indirect Methods
Strahler et al. (1986) distinguish two different scene
models for remote sensing, the high-resolution
(H-resolution) model in which the elements of the
scene are larger than individual pixels, and lowresolution (L-resolution) models in which the opposite is true. For density estimation, the elements
of interest are individual plants or plant canopies.
Most applications of remote sensing to density estimation have involved H-resolution approaches in
which individual plants are resolvable in the imagery, for example, the use of film photography or
aerial videography to estimate shrub or tree density
in open shrublands and woodlands. Accurate density measurement requires rectified imagery or
some other method to minimize effects of image
distortion on area estimation.
Image processing methods have been developed
to exploit spatial variance or texture in H-resolution
imagery. For example, Cohen et al. (1990) applied
variogram analysis to I-m digital aerial photogra-