vol. 155, no. 4 the american naturalist april 2000
Forest Canopy Stratification?Is It Useful?
Geoffrey G. Parker1,* and Martin J. Brown2,?
1. Smithsonian Environmental Research Center, P.O. Box 28,
Edgewater, Maryland 21037;
2. Synthesis Research and Analysis, P.O. Box 4054, Portland,
Oregon 97208
Submitted February 16, 1999; Accepted November 22, 1999
abstract: It has long been recognized that the forest canopy has
a complex structure that is significant for environmental interactions,
regeneration, growth, and biotic habitat. Not only is the structure
variously complex, but also there are many ways to conceptualize
that complexity. Yet the persistent theme when considering the struc-
ture of canopies continues to be that of stratification: whether struc-
tural units are arranged in layers above the ground. We examined
the use of the terms ?stratification,? ?layering,? and others in con-
nection with canopy structure and found they had various meanings
(often only implied) that were difficult to reconcile and to measure.
We applied the definitions to the structure of a single, well-studied
canopy located in Virginia, U.S.A., and found they failed to define
consistently and clearly the presence, number, or location of strata.
Additionally, we found the concept had limitations related to scale
dependence, point of reference, and spatial averaging. Thus, asserting
that a forest is stratified or naming the number of layers generally
provided no guide to its structure. We propose alternative ways of
conceptualizing and studying the forest canopy that avoid most of
the problems associated with stratification. Among these are direct
measurement and mapping of structural and environmental variables
that have clear potential connections with canopy functions and
viewing the distribution of structures or environmental conditions
within the canopy as ecological gradients.
Keywords: forest, canopy, layer, stratification, stratum.
There opened to my view one of the most magnificent pros-
pects that forest scenery could afford; the gigantic measure of
some of the trees was altogether surprising, but yet, on account
of their various heights, their foliage lay as if it were in strata,
and the denseness of the ramification wove the branches into
a chaos as picturesque as it was inextricable. (Schweinfurth
1874, p. 488)
* E-mail: parker@serc.si.edu.
? E-mail: LM@spiritone.com.
Am. Nat. 2000. Vol. 155, pp. 473?484. q 2000 by The University of Chicago.
0003-0147/2000/15504-0004$03.00. All rights reserved.
Traveling through Africa in 1870, G. A. Schweinfurth saw
forest canopies in what seem like modern terms, as both
?stratified? and ?chaotic.? He also seized upon the ele-
ments of much confusion and argument about forest can-
opy structure in twentieth-century ecology.
?Stratification,? ?layering,? and related terms have long
been used by ecologists to describe the simultaneous va-
riety and organization they see when looking up into the
forest canopy. With these words, they have generally as-
serted that there can be different things?be they struc-
tures, species, or environments?at different heights in the
canopy, to a degree that they might define identifiable
zones, and that canopies can differ in this organization.
For example, several recent papers have examined inter-
actions between strata (Craig 1993; Gilliam and Turrill
1993; Wilson et al. 1995). Terborgh (1985) discussed the
hypothesis that there are more strata in tropical than tem-
perate forests. Mimicking the ?stratification? of natural
forests has been one goal in designing some agricultural
(Hart 1980; Altieri et al. 1983; Ewel 1986) and agroforestry
(Unruh 1991) systems. In the wet forests of Oregon and
Washington, efforts have recently focused on recreating
?multilayered? canopies?one of the distinguishing char-
acteristics of old-growth Douglas-fir (Pseudotsuga menzie-
sii) forests (Franklin and Spies 1991). In short, stratifi-
cation is often presented as a scientific system for
classifying or measuring canopies.
The popularity of stratification has reached beyond ac-
ademic journals to appear in ?coffee-table? books (Ayensu
1980; Middleton 1992; Terborgh 1992; Moffett 1993), un-
dergraduate texts (Gerking 1974; Kimmins 1987; Smith
and Smith 1998), and encyclopedia entries on forests (En-
cyclopaedia Britannica 1982; Goulding 1997). It is one of
the first concepts many people are taught about forests.
But in the scientific literature, the popularity of the term
has produced neither clarity nor agreement. Disputes
about exactly what kind of organization is implied by strat-
ification are practically as old as observations of canopy
structure, and still remain unresolved despite an emotional
debate in some circles. Richards (1952, p. 22), in the classic
citation on the concept, describes an argument going on
back to the 1920s: ?There are also authors who state, or
imply, that any grouping of the trees according to their
474 The American Naturalist
height is arbitrary and that ?strata? have no objective reality.
This is the point of view of Mildbraed (1922).? Subse-
quently, there were occasional calls for clarification and
precision (Newman 1954; Grubb 1966). Smith (1973, p.
671) wrote in The American Naturalist, ?Vertical stratifi-
cation of plant and animal communities has been a basic
concept in forest ecology; yet it has seldom received critical
examination.? He went on to discuss several ways that
canopies could be considered stratified, and the possible
ramifications of those patterns. Despite these thoughtful
efforts, definitions have continued to multiply in recent
years.
With stratification growing in importance but not in
clarity, we believe it deserves another kind of assessment.
We will not enter into any arguments about whether strat-
ification exists (it surely does in some cases), whether can-
opy structure changes as forests develop (it does), nor
whether structure has ecological consequences (it likely
does). Instead, we will review the current definitions of
canopy stratification and related terms. To help evaluate
the tractability and utility of each definition, we will then
attempt to apply each one to the canopy of one well-
studied forest. This review and trial will be the basis for
a discussion of the many problems involved in the appli-
cation of stratification as a scientific system of classification
or measurement of canopies. Finally, we will suggest sev-
eral alternative approaches to the subject of canopy struc-
ture and environment that avoid most of the problems we
have identified.
References to layering and stratification cross several
disciplines. We make a broad review but limit most of our
discussion to terrestrial plants. Though canopies change
with time (e.g., Aber 1979; Oliver 1981; Waring and Schles-
inger 1985; Oliver and Larson 1990), we limit our dis-
cussion to structure described at one time. There remains
a large literature that uses ?stratification? to describe the
vertical zonation of animals in terrestrial vegetation (e.g.,
de Vries 1988; Reagan 1992) and of various organisms in
aquatic and littoral environments.
Varieties of Stratification
Many things may be meant when an author writes about
a ?layer? or ?level? or ?stratum? or ?tier? or ?story? or
that a canopy is ?stratified.? Here we list some recent
usages of the terms in the ecological literature. Note that,
since the terms are often used without being explicitly
defined, we often had to deduce the meaning from the
context.
Definition 1: Synonym for Height. OneStratification = A
of the most common usages is one of the most subtle.
Sometimes ?layer,? ?level,? ?stratum,? and so forth are
used without further qualification or description and can
logically be interpreted as synonyms for the elevation
above the ground or other reference level. For example,
?black-backed Woodpeckers ... were found to excavate
mainly on logs and at the base of large-diameter tree
trunks. In contrast, Three-toed Woodpeckers ... preferred
higher strata and smaller diameter trunks? (Villard 1994,
p. 1957) and ?linear relations between foliar dry matter
and cross-sectional area of first-order branches in lower,
middle, and upper crown strata also were examined.
Branches in the lower stratum had less foliar dry matter
per unit of cross-sectional area than branches in the other
two strata? (Valentine et al. 1994, p. 576). The last sentence
could be rephrased as ?branches in the lower height class
had less foliar dry matter per unit of cross-sectional area
than branches in the other two height classes,? without
any loss of meaning.
Definition 2: Life Forms or AgeStratification = Different
Classes at Different Heights. A second very common usage
for ?layer? or ?stratum? is to indicate a plant life-form
group, for example, the tree, shrub, and herb layers (Hus-
sain et al. 1994), or an age class, such as the tree, sapling,
and seedling layers (Craig 1993), that tend to exist at a
characteristic height. Sometimes the relative coverage of
these different forms is used to describe stands (Cain and
de Oliveira Castro 1959; Spies 1991; Okuda 1994).
Definition 3: Variability in PlantStratification = General
Matter. When a large variety of sizes and types of plant
matter, usually including some especially large items, is
observed, the stand is called ?multilayered? or ?stratified,?
without specifying what the layers are. Shorter, usually
younger forests with a more consistent look are called
?monolayered? or ?unstratified.? In this definition, there
are often no special measurements or analyses: for ex-
ample, ?diversification of tree structure may begin early.
Many 90- to 130-year-old stands begin to show greater
ranges in size of trees and a multilayered canopy. Time
for development can also vary.... The broad range of sizes
and varied canopy (as opposed to the monolayer of Doug-
las-fir canopies in young-growth stands) do not generally
become well developed, however, until stands reach 200
to 250 years of age? (Franklin et al. 1981, p. 19).
Definition 4: Continuous Vertical Dis-Stratification = A
tribution of Foliar Surfaces. Franklin and Spies (1991, p.
76) later extended the previous definition for cases where
leaves are found at all heights: ?Multiple canopy layers or,
more specifically, the continuous distribution of foliar sur-
faces from the top of the crown to the ground, is another
stand-scale structural feature. Such canopy distributions
are significant in creating greater quantities and greater
diversity of animal habitat.?
Definition 5: Vertical Distribution ofStratification = The
Foliage. In this usage, the question, Is the canopy stratified?
does not have a yes/no answer. Rather, stratification refers
Is Canopy Stratification Useful? 475
to the vertical distribution of foliage; thus, a graph of leaf
area index or merely leaf presence or absence within dif-
ferent height categories, would be a picture of stratification
(e.g., Hubbell and Foster 1986; Malcolm 1994). Though
this usage often depends on height categories, like 0?2,
2?5, 5?10, 10?20, 20?30, and 30?40 m above ground
(Malcolm 1994), no particular functional or structural dif-
ference between the height categories is necessarily
presumed.
Definition 6: Set of Crown Limits.Stratification = A
Richards (1952, p. 23) defined a stratum as ?a layer of
trees whose crowns vary in height between certain limits.?
The choice of these strata often came from exquisitely
drawn ?profile diagrams? (e.g., Davis and Richards 1933),
and the subjectivity of these choices caused a lot of debate
(Halle? et al. 1978; Bru?nig 1983; Whitmore 1984). He later
acknowledged the technique was illustrative rather than
statistical (Richards 1983, 1996). The use of profile dia-
grams has been continued by some, who argue that
thoughtfully prepared illustrations are more valuable than
random sampling, given the current poor knowledge of
canopy structure (Kuiper 1988).
Definition 7: Leaf Area withStratification = Clumped
Height. Other ecologists following the lead of Richards
adopted a slightly more refined definition: a clumped dis-
tribution of leaf area, leaf area density, or crown cover in
vertical space (Koike et al. 1990; Terborgh and Petren 1991;
Ashton and Hall 1992; Koike and Syahbuddin 1993). Strat-
ification should be detectable as modes on a graph of leaf
area against height; the number of strata would be equiv-
alent to the number of modes. Koike and Syahbuddin
(1993) used ANOVA to try to locate these modes in some
subalpine forests. This usage is common in some ecosys-
tem and physiological studies (e.g., Hollinger 1989; Nobel
et al. 1993). Note that this definition is the opposite of
definition 4, which implies a somewhat even distribution
of leaf area with height.
Definition 8: with Different LeafStratification = Species
Heights. While the previous definition lumps all species
together, some ecologists have measured the leaf heights
or crown cover of different species by height class. Strata
are composed of species cohorts or dominance categories
with significantly different mean leaf heights or cover
heights (e.g., Wierman and Oliver 1979; Bicknell 1982;
Guldin and Lorimer 1985; Oliver and Larson 1990).
Definition 9: with Different TopStratification = Species
Heights. Another group of ecologists more concerned with
interspecific competition has defined stratification as sig-
nificant differentiation in tree heights between species (e.g.,
Palik and Pregitzer 1993). Sometimes this assessment is
made without absolute height measurements; here, strat-
ification refers to a trend in the relative positions of ad-
jacent crowns of different species (e.g., Smith 1962; Oliver
and Larson 1990; Whitney 1990).
Definition 10: Index of Vertical Struc-Stratification = An
ture. Some have combined attributes of canopy structure
to create indices to represent structural variability. Mills
et al. (1993) calculated a ?canopy layering index,? a sum
of canopy cover in height intervals 2?10, 10?20, 20?30 m,
to predict the suitability of a site as a habitat for spotted
owl. Terborgh and Petren (1991) suggest that the number
of superimposed crowns is the correct measure of strati-
fication. Spies and Cohen (1992) defined an index of ?can-
opy height diversity,? which increased with both stature
and variance in crown size and position. Ashton and Hall
(1992) devised a ?stratification index? that increased with
the disparity in crown coverage between the most and least
dense parts of the canopy. Though not labeled as a measure
of stratification, the ?foliage height diversity index? (Mac-
Arthur and MacArthur 1961) increases when there are
many layers or great differences in foliage density among
layers (e.g., Unruh 1991).
Trial Application of These Definitions
We applied each of these definitions to a forest stand that
we have studied in detail, the Twin Springs permanent plot
(TSP), in a successional, mixed-oak forest in southwestern
Virginia (described in Parker et al. 1993). For each defi-
nition, we attempted to answer, Is the TSP canopy strat-
ified? And if so, is there a numerical summary of the
stratification, such as the number of strata?
Most of the definitions could be evaluated using the
results in figure 1, which shows Twin Springs leaf area
by height, or figure 2, which shows stem density by tree
height. We supplemented those results with other field
observations (e.g., Parker et al. 1993).
The results, presented in table 1, differed in their eval-
uations of the definitions of stratification. Definitions 1,
3, and 6 could not be meaningfully applied (see comments
below). The Twin Springs stand was not stratified by def-
inition 4 but was stratified according to definitions 2, 7,
8, and 9. Furthermore, definitions 7, 8, and 9 provided
answers ranging from 3 to 6 regarding the number of
strata. In one application of definition 10, the ?stratifi-
cation index? (Ashton and Hall 1992) was 2.5. Definition
5 did not produce a yes/no evaluation of stratification or
a number of layers; however, the relevant distribution of
foliage by height is shown in figure 1, left.
Immediate Problems with the Definitions
of Stratification
Beyond finding a lack of consistency in results (table 1),
the exercise of compiling and attempting to apply the def-
476 The American Naturalist
Figure 1: Foliage height profile for the Twin Springs plot, giving the percentage of the total leaf area index (LAI) in each vertical meter band
(obtained using the method of optical point quadrats; MacArthur and Horn 1969). The left panel gives the pattern of foliage density with height
when species are lumped together, and the right group of panels shows the corresponding distribution for each of the major species.
initions revealed some immediate difficulties in using
stratification to compare and to contrast canopies or the
conditions within them.
Exaggeration of Terms
Definitions 1 and 5 use ?strata? or ?stratification? to pro-
vide an extravagant name for otherwise straightforward
measurements: height or the distribution of leaves by
height. Like some other jargon, this kind of special vo-
cabulary muddies clear communication by implying that
something very important changes with height but without
naming what it is.
Vague or Unrepeatable Methods
Definitions 3 and 6 bring forward potentially more dis-
criminating concepts: that some canopies can be composed
of a wider variety of objects than others and that crowns
can be grouped by height class. However, the evaluation
of those characteristics is based on no specific reproducible
measurements or observations.
Highly Abstract Definitions Leading to Vague Predictions
Definitions 4 and 7 look for a rather conceptual kind of
vertical organization: a continuous or clumped distribu-
tion of foliage against height. These definitions are eval-
uated objectively, on the basis of detailed measurements
(such as fig. 1, left), and the judgment of stratification is
clear (sometimes the result of a statistical test). Nonethe-
less, these definitions have little power to describe how a
canopy is organized. There are many diverse ways that a
canopy might have a clumped or continuous distribution
of foliage against height, but the definitions cannot dis-
criminate between them.
Some indices of stratification (definition 10) suffer the
same kind of problem: they produce the same index value
Is Canopy Stratification Useful? 477
Figure 2: Height distribution of the tops of live stems in the 3.0-ha Twin Springs plot, estimated from species-specific allometric equations relating
top height to stem diameter. At left is the distribution for all stems in the plot and at right is the corresponding distribution for each the major
species. The major species differ from those in figure 1 because the foliage height distribution was measured at a subset of the plot locations.
for a wide variety of structures. Computed for our stand
in the southern Appalachians, Ashton and Hall?s (1992)
?stratification index? is 2.5, similar to values those authors
computed for tropical stands in northwestern Borneo.
Though the numbers are close, it is difficult to see in what
sense the stands are similar.
Weak Null Hypotheses
Definitions 2, 8, and 9 evaluate stratification depending
on how canopy objects are sorted in relation to height.
The heights of different life forms or species attributes are
tested, and stratification is ?confirmed? when one dis-
proves the (often unlikely) null hypothesis that everything
is the same at every height. Thus, by such criteria, most
canopies will turn out to be stratified in some sense.
In short, to say that a forest is ?stratified? or to say that
a forest canopy is composed of a certain number of strata
rarely provides a consistent or testable indication of its
structure. Though stratification is referred to as if it were
a system of classifying or measuring forest structure, the
definitions are often incapable of clearly distinguishing
places within a canopy, distinguishing canopies from one
another, or both.
The number and poor quality of definitions creates a
miasma of meaning where interesting, disprovable state-
ments?the kind required for the testing of hypotheses
(Peters 1991)?are difficult to make. In this sense, canopy
stratification, and all the argument over it, really is not
useful.
Further Conceptual Problems with
Stratification
Evaluations of stratification are also constrained by their
perspective and methods.
Scale Dependence
The perception of stratification depends strongly on the
scale of measurement. Consider our trial evaluation of
definition 4, where a stand is stratified if there is a con-
478 The American Naturalist
Table 1: Results from attempts to apply 10 definitions of canopy stratification to the Twin Springs plot (TSP)
canopy, Virginia
Definition of stratification
Is the TSP
canopy
stratified?
Number of strata or other
summary at TSP
1. Synonym for height NA NA
2. Different life forms at different heights Yes Tree cover 91%, shrubs 10%, herbs 30%
3. General variability in plant matter Unknowna NA
4. Continuous vertical distribution of foliar surfaces Nob NA
5. Vertical distribution of foliage NA NA
6. Set of crown limits Unknownc NA
7. Nonrandom (clumped) leaf area against height Yesd 3 stratae
8. Species with different leaf heights Yesf 4 stratag
9. Species with different top heights. Yesh 6 stratai
10. Index of vertical structure NA SI = 2.5j
a Definition does not state a required level of variability.
b Figure 1, left, shows that there is a zone between 18?19 m height where no leaves were detected.
c Definition is subjective; results depend on what crown limits were chosen.
d Figure 1, left, shows strong peaks and valleys of leaf area. No formal analysis was performed.
e Based on peaks in figure 1, left, at 2.5, 8.5, and 16.5 m.
f Figure 1, right, shows the vertical distribution of leaf height by species. ANOVA showed a significant effect of species on leaf
height (P ! .001).
g ANOVA showed that the six major species had leaf heights in four significantly different groups (P ! .05).
h Figure 2, right, shows the top heights of trees at Twin Springs. ANOVA showed that species as a significant predictor of top height
(P ! .001).
i ANOVA showed the mean heights of the six major species were all significantly different (P ! .05).
j Several indices are used to evaluate stratification in the literature. Here we give Ashton and Hall?s (1992) ?stratification index?
(SI).
tinuous vertical distribution of foliar surfaces. In the Twin
Springs stand, we measured leaf density within 1-m height
intervals. Because there were no leaves between 18 and 19
m, the stand is unstratified. However, if we had used larger
intervals, such as 2 or 5 m, the stand would be considered
stratified by definition 4. In general, the strata apparent
at one resolution may change or disappear at another
resolution.
Loss of Valuable Information to Averaging
Efforts to evaluate stratification often involve the laborious
collection of very interesting data, which are then averaged
to produce a less interesting, and sometimes deceptive,
result. A set of simple observations from the Twin Springs
stand illustrates these effects. We established a large num-
ber of vertical transects, along which we noted the presence
or absence of foliage in each of three vertical zones: 1?6,
6?12, and 112 m above the ground. The results from these
transects could be summarized in two ways: by averaging
the presence or absence within each zone across all tran-
sects to produce an average picture of leaf density by
height, and by defining eight structural types (based on
the presence or absence of leaves at each of three heights,
) and counting how many of the transects fell into32 = 8
each of those eight types (Connell et al. 1997).
Figure 3, left, shows the average presence of leaves in
each zone across the entire stand. Note that the 5?7-m
bands, though not unusually broad for studies of forest
structure, are too wide to show the detail and ?stratifi-
cation? apparent in similar graphs with 1-m bands (figs.
1, 2). At the standwide level, this view of canopy structure
is vertically rather even and, according to an informal
application of definition 7, unstratified.
Figure 3, right, shows a vastly different picture: the va-
riety of vertical structures that make up the average in
figure 3, left. Only 28% of the transects had the ?average?
structure (some foliage in all zones), with its suggestion
of a relatively even density of foliage against height. The
remaining 72% had foliage in two or fewer zones, which
usually suggested a more clumped or stratified arrange-
ment. While only 3% of transects had the classic gap struc-
ture (no foliage in any zone), these areas may be dis-
proportionately important, for example, for forest
regeneration (Runkle 1985; Denslow 1987). Averaging may
hide ecologically interesting diversity.
Weakness of Height as a Meaningful Proxy Axis
Many of the definitions of stratification use height as an
essential measurement. But distance from the ground has
little ecological meaning on its own; rather, it is a con-
Is Canopy Stratification Useful? 479
Figure 3: Foliage presence in three vertical layers at the Twin Springs plot, from observations at 517 locations (following the method of Connell
et al. 1997). The left panel summarizes the frequency of foliage presence in each of three layers over all observations. The panels to the right depict
the eight types of vertical structure (defined by the presence or absence of leaves in three layers). The widths of the individual columns at the right
are proportional to the representation of that type of structure in the stand?note that the classical ?gap? structure, with no leaves at any level (type
8), is the least common.
venient proxy for a complex gradient that spans conditions
from the ?outside? of the canopy (its interface with the
atmosphere) to those of a more protected interior (Parker
1995). This gradient does not always correspond well with
the vertical axis.
McCune (1993), working in western conifer forests,
found that, at lower elevations in tree crowns, lichens could
be found either both on the bole and far from it on ex-
tended branches. Higher up in the crown, they were found
only on the bole. The suggestion is that there is a practical
equivalence in habitat (probably conditions of tempera-
ture, humidity, etc. brought on by the protective influence
of branches, leaves, etc.) between the two kinds of loca-
tions. This equivalence can explain the observed distri-
bution of lichens, whereas height alone does not.
Similarly, Reagan (1991, 1992) found that anoles (liz-
ards) were predominantly detected at high elevations in a
Puerto Rican canopy, perched on small diameter branches.
However, when a hurricane brought many small diameter
branches to the ground, the anole distribution shifted to-
ward the ground as well. The lizards preferred or depended
on the small diameter perches, not on a particular height.
In short, height can be deceptive because it is only a
rough substitute for structures or environments that have
more ecological importance. A researcher who rigidly as-
sociated height and stratification may fail to understand
the basic biological relationships suggested by the results.
Alternatives to Stratification
Many of the problems we have discovered in our review
could be ameliorated. Methods and definitions could be
standardized, weak null hypotheses could be strengthened,
and inevitable limitations such as dependence on scale and
problems with averaging could be explicitly recognized and
accommodated.
However, such efforts do not change the powerful un-
derlying presumption that canopies are, or in some sense
should be, stratified. It may be more fruitful to discard
480 The American Naturalist
Figure 4: Transmittance of photosynthetically active radiation (PAR) in a vertical cross section of a large forest gap on the Maryland coastal plain,
obtained with vertical transects with a balloon-mounted quantum sensor. The percentage of transmittance is presented as a contour graph with
heavier shading indicating darker conditions.
that presumption entirely. In this section, we suggest some
ways that might be done.
Use Existing Stand-Level Variables
Stratification is an attribute of a whole stand or canopy.
For the purpose of comparing whole stands to one another,
one simple alternative to stratification is to fall back on
commonly accepted and understood attributes of whole
stands, such as tree diameter, age, leaf area index, and
variants thereof.
Such simple measurements rarely depend on the height
axis as a proxy or presume some sort of stratification.
Nonetheless, they may still be valuable for advancing eco-
logical understanding. For example, the density of large-
diameter standing dead trees helps indicate the habitats
for many vertebrate and invertebrate animals (Franklin
and Spies 1991); the mean diameter of standing dead trees
can be a predictor of bird populations (Mills et al. 1993).
The simple presence or absence of a tree-fall gap (and its
unique environment) affects the composition of the bat
community because the ?clutter? of leaves and branches
in the canopy affects how effectively bats of different mor-
phology can fly (Crome and Richards 1988).
Map the Distributions of Relevant Structures,
Environments, and Functions
Even when giving up the goal of identifying strata, basic
efforts to illustrate the distribution of relevant structures
across space are still essential. Though the vertical distri-
bution of foliage (fig. 1) is one of the most common il-
lustrations of this kind, other canopy structures are largely
unexplored. The example of the Puerto Rican lizards, dis-
cussed earlier, suggests that the density of fine diameter
perches may be an interesting structural attribute to map.
Similarly, in a tropical forest, epiphyte seedling abundance
may depend on particular sorts of crotches, where survival
is based on water or nutrient availability (T. Laman, per-
Is Canopy Stratification Useful? 481
Figure 5: Vertical zonation of light environments in the canopy of an old-growth Douglas-fir/western hemlock forest in southern Washington
(adapted from Parker 1997). The left and middle panels give the mean and spatial variation in the transmittance of photosynthetically active radiation
(PAR) estimated from measurements in 16 vertical transects. The right panel shows the height limits of three light environment zones based on
these.
sonal communication); this suggests crotches may be an
interesting structural attribute to map.
The literature about canopy stratification has tradi-
tionally emphasized studies of structure. However, this
article?s examples concerning lizards, lichens, and epi-
phytes strongly suggest that our comprehension of the
canopy can be greatly increased if we also map (or at
least consider) the arrangement in the canopy of related
functions (such as providing habitat for lizards, photo-
synthetic capacity, and others) and environments (such
as the microclimate conditions, which support lichens,
the intensity and quality of light, and others).
These maps of structure, function, and environment are
illustrations of ecological gradients within the canopy sys-
tem. Though these gradients will often be summarized
along a vertical axis, they are not necessarily dependent
on or restricted to the vertical. Figure 4 displays one such
gradient in a two-dimensional cross section: the extinction
of light (photosynthetically active radiation) in a forest
gap. Light is obviously a critical resource for plants, and
yet in this case the gradient has a complex and variable
relationship with height.
Such gradients have characteristics that provide a way
to compare and to distinguish canopies from one another.
Gradient characteristics like length, shape, and slope can
be objectively evaluated and compared. For example,
Brown and Parker (1994) found that a collection of can-
opies varying greatly in stature and leaf area index had
very similar light conditions at the forest floor. Accord-
ingly, these stands must vary in the shape or slope of their
gradients of light extinction. Exploring this kind of vari-
ation can lead to new hypotheses about forest processes.
If Necessary, Define Strata as Segments of Gradients
One of the goals of stratification analysis is to identify
distinct, meaningful zones within the canopy (e.g., Smith
1973). Gradients of structure, function, and environment
offer a clear basis for defining such zones when they are
desirable or necessary. Much as we arbitrarily identify col-
ors by marking off a particular region in a continuous
gradient of wavelengths, a ecological zone or stratum
might be objectively defined as all the locations that hap-
pen to be within a range of specific structural, environ-
mental, or functional conditions (see Whittaker 1975, p.
157).
482 The American Naturalist
For example, in oceanography and limnology, the ?eu-
photic? zone is sometimes defined as the region where
light transmittance is greater than the ?compensation
point,? where phytoplankton respiratory losses match
photosynthetic gains (e.g., Parsons et al. 1984). One could
make similar definitions for canopy environments, such
as the overstory is the zone with 150% light transmittance,
the understory is the zone with !10% light transmittance,
and the midcanopy is the zone with light levels between
these.
Strata defined this way have appealing properties. They
can be based on variables that are standardized, commonly
understood, and easily comparable between sites. They
may be simple or complex in conception, relying on single
or multiple variables. They are not fixed to the vertical
axis, as figure 4 shows. They can also vary in time, shifting
as both the forest and the external environment change.
Such strata could reveal unexpected equivalences between
disparate times and locations.
Use, Do Not Discard, Information about Variability
The variability inherent in observations of canopy struc-
ture, environment, and function need not be discarded
through averaging. The variation itself can be useful in
a description of canopy environments; differences of var-
iation between situations may reveal distinct ecological
features. For example, figure 5 defines three potentially
relevant zones for an old-growth Douglas-fir/western
hemlock (Pseudotsuga menziesii/Tsuga heterophylla)
stand, based on both the mean and standard deviation
of the transmittance of light. The ?bright? zone is a region
of high transmittance, with little variation; the ?dim?
zone is a region of low transmittance, with some vari-
ation. In between, the ?transition? zone is a region of
great variability in transmittance.
These ideas and examples offer just a few ways that the
forest canopy might be conceptualized and studied without
reference to the traditional notion of stratification. There
are surely many more.
Conclusion
The forest canopy is still a little-known environment and
still engenders the same impressions of simultaneous chaos
and order that Schweinfurth (1874) described so well. The
various concepts of stratification aimed, admirably, to spell
out that order but were too ill defined to be of much help.
The drive to number the strata in the forest was probably
premature. It helped us ignore significant aspects of how
forests are organized and how structure can relate to eco-
logically interesting environments and functions. In a
sense, stratification has not provided an approach for un-
derstanding Schweinfurth?s ?chaos? but rather a means to
avoid it.
Acknowledgments
We thank M. Moffet for bringing Schweinfurth?s work to
our attention. R. Hansing gave helpful background infor-
mation about oceanography. B. Boelts, J. Connell, R.
Dueser, and S. Dunbar made useful comments on an ear-
lier version of this manuscript. Support came from the
James Smithson Society and the Smithsonian Environ-
mental Science Program. This is a contribution to the
Global Change program of the Smithsonian Environmen-
tal Research Center.
Literature Cited
Aber, J. D. 1979. Foliage-height profiles and succession in
northern hardwood forests. Ecology 60:18?23.
Altieri, M. A., D. K. Letourneau, and J. R. Davis. 1983.
Developing sustainable agroecosystems. BioScience 33:
45?49.
Ashton, P. S., and P. Hall. 1992. Comparisons of structure
among mixed dipterocarp forests of north-western Bor-
neo. Journal of Ecology 80:459?481.
Ayensu, E. S. 1980. Jungles. Crown, New York.
Bicknell, S. H. 1982. Development of canopy stratification
during early succession in northern hardwoods. Forest
Ecology and Management 4:41?51.
Brown, M. J., and G. G. Parker 1994. Canopy light trans-
mittance in a chronosequence of mixed-species decid-
uous forests. Canadian Journal of Forest Research 24:
1691?1703.
Bru?nig, E. F. 1983. Vegetation structure and growth. Pages
49?75 in F. B. Golley, ed. Tropical rain forest ecosystems:
structure and function. Elsevier Scientific, Amsterdam.
Cain, S. A., and G. M. de Oliveira Castro. 1959. Manual
of vegetation analysis. Harper, New York.
Connell, J. H., M. D. Lowman, and I. R. Noble. 1997.
Subcanopy gaps in temperate and tropical forests. Aus-
tralian Journal of Ecology 22:163?168.
Craig, R. J. 1993. Regeneration of native Mariana Island
forest in disturbed habitats. Micronesica 26:99?108.
Crome, F. H. J., and G. C. Richards. 1988. Bats and gaps:
microchiropteran community structure in a Queensland
rain forest. Ecology 69:1960?1969.
Davis, T. A. W., and P. W. Richards. 1933. The vegetation
of Moraballi Creek, British Guiana: an ecological study
of a limited area of tropical rain forest. Journal of Ecol-
ogy 21:350?384.
Denslow, J. S. 1987. Tropical rainforest gaps and tree spe-
cies diversity. Annual Review of Ecology and Systematics
18:431?452.
de Vries, P. J. 1988. Stratification of fruit-feeding nym-
Is Canopy Stratification Useful? 483
phalid butterflies in a Costa Rican rainforest. Journal
of Research in Lepidoptera 26:90?108.
Encyclopaedia Britannica. 1982. The Encyclopaedia Bri-
tannica. 15th ed. Encyclopaedia Britannica, Chicago.
Ewel, J. J. 1986. Designing agricultural ecosystems for the
humid Tropics. Annual Review of Ecology and Syste-
matics 17:245?271.
Franklin, J. F., and T. A. Spies. 1991. Composition, func-
tion, and structure of old-growth Douglas-fir forests.
Pages 71?81 in L. F. Ruggiero, K. B. Aubry, A. B. Carey,
and M. H. Huff, eds. Wildlife and vegetation of un-
managed Douglas-fir forests. USDA Forest Service Gen-
eral Technical Report PNW-GTR-285. Pacific Northwest
Research Station, Portland, Oreg.
Franklin, J. F., K. Cromack, W. Denison, A. McKee, C.
Maser, J. Sedell, F. Swanson, and G. Juday. 1981. Eco-
logical characteristics of old-growth Douglas-fir forests.
USDA Forest Service General Technical Report PNW-
118. Pacific Northwest Forest and Range Experiment
Station, Portland, Oreg.
Gerking, S. D. 1974. Biological systems. 2d ed. Saunders,
Philadelphia.
Gilliam, F. S., and N. L. Turrill. 1993. Herbaceous layer
cover and biomass in a young versus a mature stand of
central Appalachian hardwood forest. Bulletin of the
Torrey Botanical Club 120:445?450.
Goulding, M. 1997. Rain forest. Microsoft encarta ency-
clopedia 1998. CD-ROM.
Grubb, P. J. 1966. Stratification. Journal of Ecology 54:
281.
Guldin, J. M., and C. G. Lorimer. 1985. Crown differen-
tiation in even-aged northern hardwood forests of the
Great Lakes region, U.S.A. Forest Ecology and Man-
agement 10:65?86.
Halle?, F., R., R. A. A. Oldeman, and P. B. Tomlinson. 1978.
Tropical trees and forests: an architectural analysis.
Springer, Berlin.
Hart, R. D. 1980. A natural ecosystem analog for the design
of successional crop systems for tropical forest environ-
ments. Biotropica 12(suppl.):78?82.
Hollinger, D. Y. 1989. Canopy organization and foliage
photosynthetic capacity in a broad-leaved evergreen
montane forest. Functional Ecology 3:53?62.
Hubbell, S. P., and R. B. Foster. 1986. Canopy gaps and
the dynamics of a Neotropical forest. Pages 77?96 in
M. J. Crawley, ed. Plant ecology. Blackwell Scientific,
Oxford.
Hussain, F., M. Ahmed, G. Shaheen, and M. J. Durrant.
1994. Phytosociology of the vanishing tropical decidu-
ous forest in district Swabi, Pakistan. II. Ordination.
Pakistan Journal of Botany 26:149?160.
Kimmins, J. P. 1987. Forest ecology. Macmillan, New York.
Koike, F., and Syahbuddin. 1993. Canopy structure of a
tropical rain forest and the nature of an unstratified
upper layer. Functional Ecology 7:230?235.
Koike, F., T. Tabata, and S. B. Malla. 1990. Canopy struc-
tures and its effect on shoot growth and flowering in
subalpine forests. Vegetatio 86:101?113.
Kuiper, L. C. 1988. The structure of natural Douglas-fir
forests in western Washington and western Oregon. Ag-
ricultural University Wageningen Papers 88-5:1?47.
MacArthur, R. H., and H. S. Horn. 1969. Foliage profiles
by vertical measurements. Ecology 50:802?804.
MacArthur, R. H., and J. W. MacArthur. 1961. On bird
species diversity. Ecology 42:594?598.
Malcolm, J. R. 1994. Edge effects of central Amazonian
forest fragments. Ecology 75:2438?2445.
McCune, B. 1993. Gradients in epiphyte biomass in three
Pseudotsuga-Tsuga forests of different ages in western
Oregon and Washington. Bryologist 96:405?411.
Middleton, D. 1992. Ancient forests. Chronicle, San
Francisco.
Mildbraed, J. 1922. Wissenschaftliche Ergebnisse der
zweiten deutschen Zentral-Afrika-Expedition
1910?1911 unter Fu?hrung Adolf Friedrichs, Herzogs zu
Mecklenburg. Leipzig.
Mills, L. S., R. J. Fredrickson, and B. B. Moorhead. 1993.
Characteristics of old-growth forests associated with
northern spotted owls in Olympic National Park. Jour-
nal of Wildlife Management 57:315?321.
Moffett, M. M. 1993. The high frontier: exploring the trop-
ical rainforest canopy. Harvard University Press, Cam-
bridge, Mass.
Newman, I. V. 1954. Locating strata in tropical rain forests.
Journal of Ecology 42:218?219.
Nobel, P. S., I. N. Forseth, and S. P. Long. 1993. Canopy
structure and light interception. Pages 79?90 in D. O.
Hall, J. M. O. Scurlock, H. R. Bolhar-Nordenkampf,
R. C. Leegood, and S. P. Long, eds. Photosynthesis and
production in a changing environment: a field and lab-
oratory manual. Chapman & Hall, London.
Okuda, S. 1994. Deciduous hardwood forest communities
in eastern North America, including some conifer forest
and shrub communities. Pages 155?201 in A. Miyawaki,
K. Iwatsuki, and M. M. Grandtner, eds. Vegetation in
eastern North America: vegetation system and dynamics
under human activity in the eastern North American
cultural region in comparison with Japan. University of
Tokyo Press, Tokyo.
Oliver, C. D. 1981. Forest development in North America
following major disturbances. Forest Ecology and Man-
agement 3:153?168.
Oliver, C. D., and B. C. Larson. 1990. Forest stand dy-
namics. McGraw-Hill, New York.
Palik, B. J., and K. S. Pregitzer. 1993. The vertical devel-
484 The American Naturalist
opment of early successional forests in northern Mich-
igan, USA. Journal of Ecology 81:271?285.
Parker, G. G. 1995. Structure and microclimate of forest
canopies. Pages 73?106 in M. Lowman and N. Nad-
karni, eds. Forest canopies: a review of research on a
biological frontier. Academic Press, San Diego, Calif.
???. 1997. Canopy structure and light environment of
an old-growth Douglas-fir/western hemlock forest.
Northwest Science 71:261?270.
Parker, G. G., S. M. Hill, and L. A. Kuehnel. 1993. Decline
of understory American chestnut (Castanea dentata
[Marsh.] Borkh.) in a southern Appalachian forest. Ca-
nadian Journal of Forest Research 23:259?266.
Parsons, T. R., M. Takahashi, and B. Hargrave. 1984. Bio-
logical oceanographic processes. Pergamon, Oxford.
Peters, R. H. 1991. A critique for ecology. Cambridge Uni-
versity Press, Cambridge.
Reagan, D. P. 1991. The response of anolis lizards to hur-
ricane-induced habitat changes in a Puerto Rican rain
forest. Biotropica 23:468?474.
???. 1992. Congeneric species distribution and abun-
dance in a three-dimensional habitat: the rain forest
anoles of Puerto Rico. Copeia 1992:392?403.
Richards, P. W. 1952. The tropical rain forest: an ecological
study. Cambridge University Press, Cambridge.
???. 1983. The three-dimensional structure of tropical
rain forest. Pages 3?10 in S. L. Sutton, T. C. Whitmore,
and A. C. Chadwick, eds. Tropical rain forest: ecology
and management. Special Publication of the British Ec-
ological Society 2. Blackwell Scientific, Oxford.
???. 1996. The tropical rain forest. 2d ed. Cambridge
University Press, Cambridge.
Runkle, J. R. 1985. Disturbance regimes in temperate for-
est. Pages 17?33 in S. T. A. Pickett and P. S. White, eds.
The ecology of natural disturbances and patch dynam-
ics. Academic Press, Orlando, Fla.
Schweinfurth, G. A. 1874. The heart of Africa. Translated
by E. E. Frewer. Harper, New York.
Smith, A. P. 1973. Stratification of temperate and tropical
forests. American Naturalist 107:671?682.
Smith, D. M. 1962. The practice of silviculture. Wiley, New
York.
Smith, R. L., and T. M. Smith. 1998. Elements of ecology.
Addison Wesley Longman, Menlo Park, Calif.
Spies, T. 1991. Plant species diversity and occurrence in
young, mature, and old-growth Douglas-fir stands in
western Oregon and Washington, USA. USDA Forest
Service General Technical Report PNW-285. Pacific
Northwest Forest and Range Experiment Station, Port-
land, Oreg.
Spies, T., and W. B. Cohen. 1992. An index of canopy
height diversity. Coastal Oregon Productivity Enhance-
ment Report 5:5?7.
Terborgh, J. 1985. The vertical component of plant species
diversity in temperate and tropical forests. American
Naturalist 126:760?777.
???. 1992. Sunlight and stratification. Pages 105?129
in Diversity and the tropical rain forest. Scientific Amer-
ican Library, New York.
Terborgh, J., and K. Petren. 1991. Development of habitat
structure through succession in an Amazonian flood-
plain forest. Pages 28?46 in S. S. Bell, E. D. McCoy, and
H. R. Mushinsky, eds. Habitat structure: the physical
arrangement of objects in space. Chapman & Hall,
London.
Unruh, J. D. 1991. Canopy structure in natural and agro-
forest successions in Amazonia. Tropical Ecology 32:
168?181.
Valentine, H. T., V. C. Baldwin, Jr., T. G. Gregoire, and H.
E. Burkhart. 1994. Surrogates for foliar dry matter in
loblolly pine. Forest Science 40:576?585.
Villard, P. 1994. Foraging behavior of black-backed and
three-toed woodpeckers during spring and summer in
a Canadian boreal forest. Canadian Journal of Zoology
72:1957?1959.
Waring, R. H., and W. H. Schlesinger. 1985. Forest eco-
systems: concepts and management. Academic Press,
Orlando, Fla.
Whitmore, T. C. 1984. Tropical rain forests of the Far East.
Oxford University Press, Oxford.
Whitney, G. G. 1990. Multiple pattern analysis of an old-
growth hemlock?white pine?northern hardwood stand.
Bulletin of the Torrey Botanical Club 117:39?47.
Whittaker, R. H. 1975. Communities and ecosystems. 2d
ed. Macmillan, New York.
Wierman, C. A., and C. D. Oliver. 1979. Crown stratifi-
cation by species in even aged mixed stands of Douglas-
fir : western hemlock. Canadian Journal of Forest Re-
search 9:1?9.
Wilson, J. B., R. B. Allen, and W. G. Lee. 1995. An assembly
rule in the ground and herbaceous strata of a New Zea-
land rainforest. Functional Ecology 9:61?64.
Associate Editor: David A. Wedin