8
further described by position and condition (e.g.,
dead vs. living, standing vs. fallen, photosynthetic
vs. nonphotosynthetic, etc.). The distribution of
these elements can be summarized at many levels
from individual plants to whole canopies, stands,
landscapes, and regions.
An exact reconstruction of three-dimensional
stand structure is currently not practical. Instead,
summary descriptors must be estimated in one, two,
or three dimensions. Common vertical descriptors
include canopy height, canopy depth, stratification,
and foliar profile. Common horizontal descriptors
include cover, canopy closure, canopy or stem density, leaf area, stem basal area, and phytomass per
unit area. Three-dimensional descriptions incorporate additional information on element size distribution within the stand volume, element position,
and orientation. For some applications, the void between biomass elements (e.g., canopy gaps) is the
variable of interest, although the methods will tend
to be the same for estimating either phytomass distribution or its complement.
Structural descriptors can be estimated visually
or from point, line, image, or volumetric samples.
These samples consist of one or more measurements taken using a particular set of instruments
and procedures (Bonham 1989). Vegetation structure generally has high fractal dimensionality,
meaning that structural estimates depend strongly
on the measurement scale (e.g., probe diameter, line
interval, pixel area, etc.). Some methods, in fact,
exploit the scaling features of vegetation to estimate
structural properties such as canopy gap fraction
(Chen et al. 1994) or three-dimensional distribution
of biomass (Kruijt 1989).
Norman and Campbell (1989) distinguish
ground-based direct methods that involve manual
(contact) measurement of individual plants or plant
organs from indirect methods involving remote
sensing of stand properties. We adopt this terminology and further subdivide indirect methods into
ground-based versus aerial (i.e., by aircraft or orbiting satellite) remote sensing.
Any investigation of stand structure must ultimately employ direct methods, at least for calibration and/or validation of indirect techniques. Most
direct methods are labor-intensive and rely on relatively simple estimation procedures, such as statistical summarization (e.g., mean height) or application of allometric equations, to indirectly
Frank W. Davis and Dar Roberts
estimate one structural variable from another that
can be measured directly (e.g., leaf area from directly measured sapwood area). Direct methods
have severe practical limits over large areas or heterogeneous vegetation and have the added disadvantage that they usually require disturbing the
vegetation. Direct investigation of the structure of
tall forest canopies poses a special challenge, and
may require felling trees or use of elaborate rope
systems, cranes, or suspended walkways (Koike
and Syahbuddin 1993; Parker et al. 1992; Lowman
and Nadkarni 1995).
Indirect methods are based on canopy-light interactions and range from very simple optical
tools, such as gridded concave mirrors, to sophisticated hardware and software, such as radar scatterometers or satellite-borne radiometers. Data acquisition is generally much faster than with direct
methods and can be conducted over larger areas.
However, estimation procedures can be considerably more complex, involving calibration of indirect measures (e.g., regression modeling to retrieve
biomass from radar backscatter power) or the inversion of empirical or physically based models
(e.g., inversion of gap fraction data obtained from
hemispherical photography to estimate total leaf
area). Rapid advances in remote sensing theory
over the past 20 years have led to general and
robust indirect methods that, for many structural
parameters, can provide more reliable estimates
than the traditional direct methods. Lack of familiarity, equipment cost, and the complexity of data
reduction methods are probably the main factors
preventing even wider use of indirect approaches
by ecosystem ecologists.
Models of Canopy Architecture
Many modem methods for characterizing plant
structure at the plant or stand level are based on
underlying models of plant geometry and canopylight interactions. These canopy models have been
developed for various purposes, for example, to explore the adaptive significance of canopy geometries (Hom 1971; Oker-Blom et al. 1989) or the
effects of canopy shape on crown shadowing and
light penetration in gaps (Kuuluvainen and Pukkala, 1989; Van Pelt and North 1996). Early efforts
focused primarily on modeling the distribution of
further described by position and condition (e.g.,
dead vs. living, standing vs. fallen, photosynthetic
vs. nonphotosynthetic, etc.). The distribution of
these elements can be summarized at many levels
from individual plants to whole canopies, stands,
landscapes, and regions.
An exact reconstruction of three-dimensional
stand structure is currently not practical. Instead,
summary descriptors must be estimated in one, two,
or three dimensions. Common vertical descriptors
include canopy height, canopy depth, stratification,
and foliar profile. Common horizontal descriptors
include cover, canopy closure, canopy or stem density, leaf area, stem basal area, and phytomass per
unit area. Three-dimensional descriptions incorporate additional information on element size distribution within the stand volume, element position,
and orientation. For some applications, the void between biomass elements (e.g., canopy gaps) is the
variable of interest, although the methods will tend
to be the same for estimating either phytomass distribution or its complement.
Structural descriptors can be estimated visually
or from point, line, image, or volumetric samples.
These samples consist of one or more measurements taken using a particular set of instruments
and procedures (Bonham 1989). Vegetation structure generally has high fractal dimensionality,
meaning that structural estimates depend strongly
on the measurement scale (e.g., probe diameter, line
interval, pixel area, etc.). Some methods, in fact,
exploit the scaling features of vegetation to estimate
structural properties such as canopy gap fraction
(Chen et al. 1994) or three-dimensional distribution
of biomass (Kruijt 1989).
Norman and Campbell (1989) distinguish
ground-based direct methods that involve manual
(contact) measurement of individual plants or plant
organs from indirect methods involving remote
sensing of stand properties. We adopt this terminology and further subdivide indirect methods into
ground-based versus aerial (i.e., by aircraft or orbiting satellite) remote sensing.
Any investigation of stand structure must ultimately employ direct methods, at least for calibration and/or validation of indirect techniques. Most
direct methods are labor-intensive and rely on relatively simple estimation procedures, such as statistical summarization (e.g., mean height) or application of allometric equations, to indirectly
Frank W. Davis and Dar Roberts
estimate one structural variable from another that
can be measured directly (e.g., leaf area from directly measured sapwood area). Direct methods
have severe practical limits over large areas or heterogeneous vegetation and have the added disadvantage that they usually require disturbing the
vegetation. Direct investigation of the structure of
tall forest canopies poses a special challenge, and
may require felling trees or use of elaborate rope
systems, cranes, or suspended walkways (Koike
and Syahbuddin 1993; Parker et al. 1992; Lowman
and Nadkarni 1995).
Indirect methods are based on canopy-light interactions and range from very simple optical
tools, such as gridded concave mirrors, to sophisticated hardware and software, such as radar scatterometers or satellite-borne radiometers. Data acquisition is generally much faster than with direct
methods and can be conducted over larger areas.
However, estimation procedures can be considerably more complex, involving calibration of indirect measures (e.g., regression modeling to retrieve
biomass from radar backscatter power) or the inversion of empirical or physically based models
(e.g., inversion of gap fraction data obtained from
hemispherical photography to estimate total leaf
area). Rapid advances in remote sensing theory
over the past 20 years have led to general and
robust indirect methods that, for many structural
parameters, can provide more reliable estimates
than the traditional direct methods. Lack of familiarity, equipment cost, and the complexity of data
reduction methods are probably the main factors
preventing even wider use of indirect approaches
by ecosystem ecologists.
Models of Canopy Architecture
Many modem methods for characterizing plant
structure at the plant or stand level are based on
underlying models of plant geometry and canopylight interactions. These canopy models have been
developed for various purposes, for example, to explore the adaptive significance of canopy geometries (Hom 1971; Oker-Blom et al. 1989) or the
effects of canopy shape on crown shadowing and
light penetration in gaps (Kuuluvainen and Pukkala, 1989; Van Pelt and North 1996). Early efforts
focused primarily on modeling the distribution of