Journal of
Applied Ecology
2000, 37,
1044±1053
ADVANCES IN APPLIED ECOLOGICAL TECHNIQUES
An improved method for the rapid assessment of forest
understorey light environments
NICK BROWN * , STEVE JENNINGS * , PHIL WHEELER{ and
JACOB NABE-NIELSEN {
*
Oxford Forestry Institute, University of Oxford, South Parks Road, Oxford OX1 3RB, UK;{Department of
Biological Sciences, Stopford Building, University of Manchester, Oxford Road, Manchester M13 9PT; and
{Department of Systematic Botany, Universitetsparken bld. 137, Aarhus University, 8000 Aarhus, Denmark
Summary
1. The high spatial and temporal variability of forest understorey light environments requires lengthy and/or extensive sampling in order to characterize it by
direct measurement. As this is often impractical, a number of surrogate measures
have been developed that estimate light availability from assessments of forest
canopy structure.
2. The subjective crown illumination index developed by Clark & Clark (1992) was
compared with Garrison's (1949) moosehorn and two new methods: (i) the crown
illumination ellipses method, which compares the size of canopy gaps with a series
of standard area ellipses printed on a transparent screen; and (ii) the canopy-scope
that, like the moosehorn, uses an array of 25 dots printed on a transparent screen
to assess canopy openness, but is more robust and portable, measuring the largest
canopy gap visible from the point of measurement rather than canopy openness
overhead.
3. The new measures were more highly correlated with canopy openness in the
range 0±30%, measured from hemispherical photographs, than the crown illumination index, and showed lower levels of between-observer variability.
4. The canopy-scope has the potential to be widely used for the simple and rapid
assessment of forest understorey light environments. It has the advantage of giving
ratio scale measurements that can be used in parametric statistics. The crown illumination ellipses can be used to score the illumination of crowns that are above
head height.
Key-words: canopy openness, canopy-scope, crown illumination ellipses, crown illumination index, forest dynamics, hemispherical photography.
Journal of Applied Ecology (2000) 37, 1044±1053
Introduction
Assessments of forest understorey light environments provide important information on the growing conditions of understorey plants, including
seedlings, saplings and subdominant trees. These are
of growing importance with the increasing emphasis
that is being placed on natural regeneration in the
sustainable management of forest ecosystems.
Although a number of methods exist for the direct
# 2000 British
Ecological Society
Correspondence: Nick Brown (fax 01865 275074; e-mail
[email protected]).
measurement of forest understorey light environments, almost all are time consuming or require
considerable investment in equipment. Small absolute changes in canopy openness produce greater
changes in mean irradiance when the forest canopy
is almost complete than when the canopy is very
open (Brown 1993). Small changes in irradiance
have a greater impact on plant growth and survival
at sites with low levels of photosynthetically active
radiation (PAR) than at those with high levels, due
to the non-linear response of photosynthesis to
increasing PAR. Hence, if a measure of canopy
openness is to be of value to the plant ecologist or
1045
N. Brown et al.
# 2000 British
Ecological Society
Journal of Applied
Ecology, 37,
1044±1053
forester it needs to be able to distinguish between
low levels of canopy openness.
Direct measurement of PAR is expensive and, if
measurements are to be used to describe understorey
growing conditions, records must be made on a suitable sample of days throughout an entire year. The
light environment of a forest understorey is extremely variable, both in space and time. The range of
spatial autocorrelation of light, in the forest understorey, has been found to be very small (Becker &
Smith 1990; Baldocchi & Collineau 1994; Clark et al.
1996; Nicotra, Chazdon & Iriarte 1999). This
implies that lengthy and/or extensive sampling is
necessary in order to ensure that descriptions of this
environment are representative. Baldocchi & Collineau (1994) estimated that more than 17 sensors
would be required to de®ne the light environment
within 10% of the mean if the spatial coecient of
variation (CV) of light exceeded 25%. They stated
that in many tropical forests the spatial CV of light
may often be in excess of 100%, meaning that over
270 sensors would be required.
Parent & Messier (1996) have demonstrated a
strong correlation between instantaneous measures
of the percentage photosynthetic photo ¯ux density
(%PPFD) made under totally overcast sky conditions in a forest understorey and the mean daily
%PPFD measured at the same point. The measure
%PPFD is more widely referred to as total site factor (TSF) (Anderson 1964). Under totally overcast
sky conditions TSF is equivalent to the indirect site
factor (ISF; the proportion of di€use radiation
reaching a given location). A very similar strong
correlation with ISF was reported by Whitmore
et al. (1993). However, this method of assessing
understorey illumination su€ers from the requirement for a totally overcast sky, a condition that
rarely occurs in many tropical and some temperate
areas. Gendron, Messier & Comeau (1998) present a
variation on this method using a LAI-2000 Plant
Canopy Analyser (LI-COR Inc., Lincoln, NE
68504, USA), which does not require overcast sky
conditions but the equipment is expensive.
Canopy openness is de®ned as the proportion of
the sky hemisphere that is not obscured by vegetation when viewed from a single point (Jennings,
Brown & Sheil 1999). It is highly correlated with
many aspects of forest microclimate, including total
receipts of PAR (Whitmore et al. 1993). Canopy
openness can be measured using hemispherical
photographs (Rich 1990). Recent developments in
both hardware and software for personal computerbased image analysis have made this technique both
more accessible and less costly. However, the technique is still not suciently rapid that it will be used
widely in forestry applications. The camera equipment required for taking hemispherical photographs
is expensive, delicate and cumbersome. The development of digital camera technology has speeded up
the process of acquiring digitized images signi®cantly but has not reduced costs. Hemispherical
photographs can only be taken under uniformly
overcast conditions, at certain times of day. Direct
sunlight causes bright re¯ections on foliage that are
dicult to distinguish from sky.
The spherical densiometer (Lemmon 1956) is used
widely in forestry for assessing canopy openness. It
has a large viewing angle but the small size of the
re¯ected image of the forest canopy results in poor
resolution. It is also slow to use, a signi®cant constraint when making large numbers of observations.
Vales & Bunnell (1988) found measures made with a
spherical densiometer to have a low level of consistency between observers. Comeau, Gendron &
Letchford (1998) reported a strong correlation
between measurements made with a concave spherical densiometer and TSF.
Our aim was to develop a method for the rapid
assessment of understorey light environments that is
cheap, robust, accurate and repeatable. In this paper
we report a comparison of four methods for estimating canopy openness that require only very simple
equipment or none at all. All four methods allow
estimates to be made extremely quickly, thus permitting large data sets to be collected with relative ease.
Methods
COMPARISON OF DIFFERENT METHODS
Four methods were used to estimate canopy openness at 10 points in Wytham Wood, Oxford, UK
(latitude 51 420 N, longitude 1 220 W), 13 points in
Little Wittenham Nature Reserve, Oxford, UK (latitude 51 380 N, longitude 1 100 W), ®ve points in
Cockshot Wood, Orielton, Pembrokeshire, UK (latitude 51 390 N, longitude 4 570 W), and at a further
18 points in Pataua Forest, near MarabaÂ, in ParaÂ
State, Brazil (latitude 5 220 S, longitude 49 110 W).
The forests di€er in structure and composition.
Wytham Wood is an ancient semi-natural broadleaved woodland dominated by Quercus robur L.
and Fraxinus excelsior L., with a canopy at 18±25 m
height. The woodland at Little Wittenham Nature
Reserve is secondary and is dominated by F. excelsior and Acer pseudoplatanus L., developed from
natural regeneration and coppice stools. The canopy
is generally less than 30 m in height. Cockshot
Wood is an ancient semi-natural woodland site. The
present wood is dominated by a dense stand of F.
excelsior and A. pseudoplatanus that has regenerated
after clear-felling approximately 50 years ago. The
canopy is generally less than 20 m in height. Pataua
Forest is a selectively logged evergreen tropical rain
forest with a canopy 30±40 m in height, with
Bertholletia excelsa Bonpl. being the most common
emergent.
1046
Rapid assessment
of understorey
light
At each point a vertical hemispherical photograph
was taken of the forest canopy, at approximately 1
m above the ground, using a Nikon 8-mm lens.
Images were digitized using an Olympus ES-10 ®lm
scanner and analysed using HemiView 2.0 canopy
analysis software (Delta-T Devices, Burwell, Cambridge, UK). Total canopy openness (VisSky in
HemiView 2.0) was calculated for each image. These
measured values of canopy openness were used as a
standard against which three rapid assessment methods were compared. Total canopy openness measured from hemispherical photographs is often
found to be highly correlated with TSF when the
full range of canopy openness conditions are used
(Rich et al. 1993; Whitmore et al. 1993). However,
some authors have found measures from hemispherical photographs to have relatively poor correlation
with directly measured TSF when only understorey
sites are compared (Gendron, Messier & Comeau
1998; Machado & Reich 1999).
Crown illumination index
Clark & Clark (1992) modi®ed the widely used
method for making simple visual assessments of tree
crown illumination devised by Dawkins (1958).
Details of their index are given in Table 1. Most
points in the forest understorey fall into Dawkins'
class 2 (lateral light). Therefore, Clark & Clark
(1992) subjectively divided this class into high, medium and low lateral light (scores of 15, 2 and 25),
to give a seven-point scale of greater utility in assessing understorey illumination.
Crown illumination ellipses
A practical problem associated with using subjective
indices is that both between- and within-observer
errors can be large. These errors arise due to di€er-
ences in the interpretation of classes, especially of
the three classes of lateral light and because an individual's interpretation of classes may drift during a
large census, or change between successive monitoring events. In order to overcome this problem one
of us developed a method where the size of a hole in
the canopy is assessed by comparing it with a series
of ellipses printed on a transparent Perspex screen
(Fig. 1). A 20-cm long cord is attached to the screen,
to ensure that it is always held at the same distance
from the eye. The score is determined by the size of
the ellipse that ®ts entirely into the largest canopy
opening visible anywhere in the canopy, whilst
standing at the point of measurement. Detailed de®nitions of each class are given in Table 1. A ®le (in
portable document format) containing a template
for the ellipses can be obtained from the correspondence author. In order to test whether use of the
ellipses improved the repeatability of measurements,
we compared how consistently points in Amazonian
forest were scored, using the two methods. As a test
of within-observer repeatability we compared two
independent studies, one using the crown illumination index and the other using ellipses. In both studies 100 points were scored and then scored for a
second time, by the same observer, less than 2 weeks
later. In a separate test of between-observer repeatability, three observers independently scored 18
points in the Pataua Forest using the two di€erent
methods.
Moosehorn
A method for estimating canopy openness, commonly known as the `moosehorn' (due to its similarity in shape to an instrument used in hunting
moose), uses 25 dots in a 5 5-square array, painted
on a transparent screen. The observer looks vertically upwards through the screen and counts the
Table 1. De®nition of crown illumination index (Clark & Clark 1992) and crown illumination ellipses indices
Index value
De®nition of crown illumination classes
De®nition using crown illumination ellipses
5
2
Completely exposed to light from that part of
the sky circumscribed by a 45 zenith angle
Full overhead light. Lateral light blocked within
part of the sky circumscribed by a 45 zenith angle
Some overhead light (10±90% exposed to
vertical light)
High lateral light (from one major or multiple
medium-sized gaps)
Medium lateral light
15
Low lateral light (no large or medium-sized gaps)
1
No direct light
Completely exposed to light from that part of
the sky circumscribed by a 45 zenith angle
Light from an overhead gap large enough to
enclose an area twice the size of the largest ellipse
Light from an overhead gap large enough to
enclose the largest ellipse
Light from a lateral gap large enough to enclose
the largest ellipse or two medium ellipses
Light from a lateral gap large enough to enclose
the medium ellipse or two small ellipses
Light from a lateral gap large enough to enclose
the smallest ellipse
No canopy gap large enough to enclose the
smallest ellipse
4
3
25
# 2000 British
Ecological Society
Journal of Applied
Ecology, 37,
1044±1053
1047
N. Brown et al.
Fig. 1. Crown illumination ellipses. The ellipses are printed on Perspex and have the following sizes:
Large
Medium
Small
A
B
C
Area
183 28 cm
130 20 cm
92 14 cm
130 40 cm
92 28 cm
65 20 cm
80 65 cm
56 47 cm
40 34 cm
410 cm2
205 cm2
103 cm2
number of dots that coincide with gaps in the
canopy. A moosehorn was constructed based on the
model proposed by Garrison (1949). The moosehorn
is aligned to look exactly at the zenith using a bubble level attached to the transparent screen. A 45
mirror enables the recorder to view the canopy
whilst looking horizontally.
AN IMPROVED DESIGN FOR THE
MOOSEHORN
# 2000 British
Ecological Society
Journal of Applied
Ecology, 37,
1044±1053
The moosehorn, as designed by Garrison (1949),
was soon found to be cumbersome and fragile.
These are major drawbacks if the instrument is to
be in frequent use in ®eld ecology or practical forestry. The instrument was therefore redesigned as a
transparent Perspex screen with a 20-cm cord
attached to one corner. The cord is used to ensure
that the screen is always held at the same distance
from the eye. The screen was engraved with 25 dots,
approximately 1 mm in diameter, spaced 3 cm apart
(centre to centre), in a 5 5-square array. Hereafter,
this instrument is referred to as the canopy-scope, in
order to distinguish it from the moosehorn. Figure 2
illustrates the canopy-scope in use.
The original design requires the observer to point
the moosehorn vertically and make a measurement
of canopy openness centred on the zenith. Many
forest understorey sites receive large amounts of lateral light. Gaps in the canopy below a zenith angle
of approximately 23 were outside the ®eld of view
of the moosehorn and hence canopy openness was
not scored accurately. We hypothesized that a closer
correlation would be found between total canopy
openness and the canopy-scope score if, rather than
pointing the instrument at the zenith, it was pointed
at the largest gap visible anywhere in the canopy
whilst standing at the point of measurement.
In order to test this hypothesis we carried out
three tests. First, at the 18 sites in Pataua Forest,
canopy openness measurements were made with
both the zenith-centred moosehorn and the largest
gap-centred canopy-scope. These were then correlated with total canopy openness as measured from
hemisphere photographs taken at the same sites.
Secondly, 23 hemisphere photographs were analysed
to determine the proportion of the total canopy
openness that was contributed by the largest gap.
This was done by manually editing the digitized
images to black out all but the largest gap. Thirteen
hemispherical photographs from Little Wittenham
and 10 from Pataua Forest were selected for this
analysis, to give a wide range of total canopy openness values. Thirdly, the same 23 hemispherical
photographs were analysed to determine the zenith
angle at which the canopy was most open. This was
1048
Rapid assessment
of understorey
light
Fig. 2. The canopy-scope in use. The screen is held exactly 20 cm from the eye and readings are taken centred on the largest
canopy gap above the point of measurement. The observer counts the number of 25 dots that coincide with sky rather than
canopy.
# 2000 British
Ecological Society
Journal of Applied
Ecology, 37,
1044±1053
done by dividing each photograph into a series of
concentric annuli, each 5 in width. The proportion
of each annulus that was open sky was then determined.
In order to test the e€ect of distance between the
canopy-scope screen and the eye, three lengths of
cord were used, 20 cm, 35 cm and 50 cm. A shorter
cord gives a greater viewing angle but as the screen
gets progressively closer to the eye, it becomes
increasingly hard to focus, simultaneously, on both
the dots and the canopy beyond. A 20-cm cord gives
a maximum viewing angle of 46 and was found to
be as close as most people could comfortably focus.
The 35-cm and 50-cm cords give maximum viewing
angles of 272 and 1926 , respectively. Thirty-one
observers used each cord length to assess canopy
openness at ®ve forest understorey sites at Cockshot
Wood. These data were then used in a mixed-e€ects
model in order to estimate whether there was a difference in the consistency of measurements made
with di€erent lengths of cord. Ninety-eight observers
independently measured the same ®ve sites using a
20-cm cord length in order to investigate the repeatability of measurements made with the canopyscope. These estimates were then used to make a
pooled estimate of the standard error of canopyscope measurements.
Results
COMPARISON OF CROWN ILLUMINATION
INDEX WITH CROWN ILLUMINATION
ELLIPSES
The within-observer repeatability of measurements
made using the crown illumination index and the
crown illumination ellipses was compared by assessing the proportion of 100 Amazonian forest sites
that, when scored by the same observer for a second
time less than 2 weeks after an initial assessment,
were given the same score. No signi®cant di€erence
was found in the frequency distribution of remeasurement di€erences (Fig. 3) using the two methods
(Kolmogorov±Smirnov test, D ˆ 0242, 2 d.f.). However, the frequency with which remeasurements differed by two classes was substantially reduced by
the ellipse method. Twelve per cent of remeasurements made using the crown illumination index differed by two classes from the original measurement,
whereas only 1% of ellipse measurements did so.
Our repeatability ®gure for the crown illumination
index was higher than that found by the authors
themselves, using an identical methodology. They
found 68% of scores to be identical in a second enumeration of 97 points, but measured sapling and
pole-sized trees, which are more dicult than seed-
1049
N. Brown et al.
Fig. 3. Number of categories di€erence between ®rst and
second measurements of 100 forest understorey sites in
Amazonia, scored using the crown illumination index and
the crown illumination ellipses.
score forest understorey light environments, this limitation was of little importance.
Of much greater importance was the ability of the
three measures to distinguish between low levels of
canopy openness. In order to assess this, scores for
the three measures were correlated with canopy
openness measured from hemispherical photographs, using sites with a canopy openness of less
than 30% and omitting more open sites. The Spearman's rank correlation coecients for all three
methods were highly signi®cant (a ˆ 001 n ˆ 25).
However, the correlation coecients for the crown
illumination ellipses (0824) and the moosehorn
(0861) were greater than that of the crown illumination index (0749).
COMPARISON OF THE MOOSEHORN WITH
THE CANOPY-SCOPE
lings to score accurately (Clark & Clark 1992). The
use of ellipses reduces subjectivity in the categorization of large, medium and small gaps, thus enhancing the repeatability of measurements.
In a second test of between-observer repeatability
we found that three observers gave a mean of 161
di€erent scores to each of 18 di€erent sites in Pataua
Forest, using the crown illumination index, but a
mean of only 144 di€erent scores to the same sites
using the crown illumination ellipses.
COMPARISON OF THE CROWN
ILLUMINATION INDEX, CROWN
ILLUMINATION ELLIPSES AND MOOSEHORN
# 2000 British
Ecological Society
Journal of Applied
Ecology, 37,
1044±1053
Measured values of canopy openness from hemispherical photographs were compared with estimates
made using the three rapid assessment methods, the
crown illumination index, the crown illumination
ellipses and the moosehorn (Fig. 4). None of the
methods appeared to be a€ected by forest stature.
At any given level of canopy openness (as measured
from hemisphere photographs) similar scores were
given in Amazon rain forest and British broadleaved woodland. Figure 4 shows that all three
methods were insensitive measures of high levels of
canopy openness. All sites fully exposed to overhead
light (where that part of the sky circumscribed by a
45 zenith angle is completely free from obstruction)
were scored in category 5 by both the crown illumination index and crown illumination ellipses. However, calculations from geometry showed that such
sites may vary in openness from 30% to 100%.
Similarly, all gaps where that part of the sky circumscribed by the 23 zenith angle was completely free
from obstruction had a moosehorn score of 25,
although they may have varied in openness from
8% to 100%. In practice, as the objective was to
Pearson's product moment correlation coecients
were calculated for data collected in Pataua Forest
in order to assess whether a zenith-centred moosehorn score or largest gap-centred canopy-scope
score was more closely correlated with canopy openness. Once again, only scores for sites with a canopy
openness of less than 30% were used. Both measures
were highly signi®cantly correlated but lateral gapcentred canopy-scope scores were more strongly correlated with canopy openness (r ˆ 0954, P < 001, n
ˆ 16) than zenith-centred moosehorn scores (r ˆ
0907, P < 001, n ˆ 16). Support for the conclusion
that the canopy-scope method gave an improved
estimate of canopy openness was provided by analysis of 23 hemispherical photographs. In each photograph, all canopy gaps, except the largest one, were
manually blacked out, using image-editing software.
The photographs were then re-analysed in order to
calculate the proportion of total canopy openness
contributed by the largest gap alone. The zenith
angle of the geometric centre of the largest gap was
also measured. Figure 5a shows that an almost constant proportion of total canopy openness was contributed by the single largest gap. This implies that
canopy-scope scores centred on the single largest
gap are likely to show a very close correlation with
total canopy openness. It is possible that, in some
forest sites, multiple gaps of similar size may be visible from the point of measurement. A canopy-scope
measure of the single largest gap would, under such
circumstances, be a poor predictor of total canopy
openness. Figure 5b indicates that the more closed
the forest canopy, the greater the chance that the
most open part of the canopy is not centred on the
zenith. This implies that zenith-centred moosehorn
scores may misrepresent less open sites.
Thirty-one independent observers used three cord
lengths to measure ®ve points in the understorey of
Cockshot Wood to compare the e€ects of cord
length on canopy-scope scores. These trials showed
1050
Rapid assessment
of understorey
light
Fig. 4. Total canopy openness (measured from hemispherical photographs) compared with rapid assessments made using
(a) crown illumination (Clark & Clark 1992) index, (b) crown illumination ellipses and (c) a moosehorn (Garrison 1949) at
two sites, Pataua Forest, Brazil (closed symbol) and Wytham Wood, UK (open symbol).
# 2000 British
Ecological Society
Journal of Applied
Ecology, 37,
1044±1053
that, with a ®xed screen size, the longer the cord the
greater the mean canopy-scope score recorded by 31
observers at any point. This was because as the
screen is moved further from the observer's eye, the
angle between any two dots on the screen and the
eye becomes smaller. Hence the longer the cord, the
greater the likelihood that a larger number of dots
will be observed to coincide with a gap of ®xed size.
The advantages of this increased resolution must be
o€set against a decrease in the viewing angle as the
screen moves further from the eye. The smaller the
viewing angle, the smaller the proportion of the sky
hemisphere that is sampled and therefore the less
representative any canopy-scope score is likely to
become. As the standard deviation of scores
increased with the mean score, it could not be used
as an unbiased measure of the consistency with
which measures could be made using di€erent length
cords. In order to overcome this problem, an index
of consistency (Y) was derived:
Y ˆ
X
1
…xi ÿ x†2 2
s
where xi is the ith observation at one point; xis the
mean of all observations at that point; and s2 is the
variance of all observations at that point.
Y was then used as the response variable in a linear mixed-e€ects model in which observers were
1051
N. Brown et al.
Fig. 5. Analyses of 23 hemispherical photographs taken in Little Wittenham Nature Reserve, UK, and Pataua Forest, Brazil. (a) Plot of total canopy openness, against the contribution of the largest canopy gap above the point of measurement.
(b) Plot of total canopy openness against the zenith angle of the geometric centre of the largest canopy gap.
considered to be random e€ects and cord length was
a ®xed e€ect covariable. Cord length was found to
have no signi®cant e€ect on the consistency of
observations.
Ninety-eight observers made canopy-scope scores
at ®ve di€erent sites using a 20-cm cord length.
Although a small increase in score variance was
detected at sites with intermediate levels of canopy
openness, the di€erence was not statistically signi®cant at the 95% level for any pair of sites (Fisher's
variance ratio test). A pooled estimate of the standard error of canopy-scope measurements was
therefore made from these scores and was calculated
to be 21. This meant that there was a 95% probability that two sites di€ered in canopy openness
when their canopy-scope scores di€ered by more
than 42.
# 2000 British
Ecological Society
Journal of Applied
Ecology, 37,
1044±1053
Discussion
In this paper we have sought to compare methods
that may be used for rapid assessment of forest
understorey light environments. With the exception
of hemispherical photography, all these methods
lose resolution when canopy openness exceeds 30%
(Fig. 4). They are unlikely to be of value for assessing light environments in open vegetation types
such as savanna woodland, where the majority of
sites are likely to be more open than this.
Although the index developed by Clark & Clark
(1992) is a signi®cant improvement on Dawkins'
(1958), it has three major drawbacks. The ®rst is the
poor level of repeatability between observers. It has
been shown that, with experience, any one observer
is able to achieve an acceptable level of consistency.
However, the scoring of the low, medium and high
lateral light categories (15, 20 and 25) is subject to
individual interpretation and can vary considerably
between observers. This reduces the utility of the
scale applications where an objective and consistent
measure is needed to inform management decisions
or for performing repeatable observations. We have
shown that the use of the crown illumination ellipses
1052
Rapid assessment
of understorey
light
# 2000 British
Ecological Society
Journal of Applied
Ecology, 37,
1044±1053
can improve both within- and between-observer
repeatability.
Clark & Clark (1992) maintain that their index is
an ordinal scale in that categories are ranked but
the numerical di€erence between any two classes
does not have meaning. The types of analysis that
can be performed on this index are therefore constrained to ranking statistics (Siegel & Castellan
1988). In fact, in using the scale, we have found it
not to be strictly ordinal with respect to canopy
openness or understorey illumination. A large lateral
gap (that would score 25) often permits greater
penetration of light than a small overhead gap (that
would score 3). This means that the scale is, in fact,
only a partially ordered scale (Siegel & Castellan
1988). The modi®cation of this index for use with
crown illumination ellipses (Table 1) does not alleviate the problem. In contrast, the canopy-scope gives
a ratio scale measurement of the largest visible gap
in a canopy that can be used in parametric statistics.
This measure is highly correlated with total canopy
openness.
Another problem is that although Clark & Clark
(1992) modi®ed Dawkins' (1958) original scale in
order to make it more applicable for the scoring of
understorey light environments, the scale lacks precision at low levels of canopy closure (see Fig. 4a). It
should be noted, however, that both the crown illumination index and the crown illumination ellipses
can be used to score the light environment above all
sizes of plants, not just those in the understorey.
The crown illumination ellipses are the best method
for scoring the light environment of plants that are
above head-height.
We have shown that the canopy-scope has lower
within- and between-observer errors than the other
methods used. It had the highest correlation with
canopy openness measured with hemispherical
photographs. Moreover, unlike the crown illumination index, measurements are on a ratio scale, which
makes it possible to analyse data using parametric
statistics. The canopy-scope is cheap, small enough
to be carried in a jacket pocket, and suciently
robust that it has survived constant use during a
long ®eld season in the Amazon. This is important
when measurements are to be made in harsh ®eld
conditions. Although hemispherical photography is
undoubtedly the most precise method, it is dicult
to maintain the equipment in good working order
during prolonged ®eld use. The camera, lens, computer and image analysis software that are required
are also expensive. At present, image processing and
analysis are time consuming and rapid assessments
of understorey illumination at a large number of
points are not possible.
We propose that the relationship described
between canopy openness and canopy-scope score is
independent of forest type and canopy structure.
This implies that there will be no need for site-speci-
®c calibration of the instrument. Unlike almost all
other methods for assessment of canopy openness,
we have been able to attach con®dence limits to
measures made with the canopy-scope. The narrowness of the 95% con®dence interval (42 dots) means
that the method is of considerable utility for distinguishing between forest understorey sites where
levels of canopy openness are relatively low.
We conclude that for rapid assessment of canopy
openness in the forest understorey, the canopyscope is the best option available, on account of its
cheapness, robustness, speed and accuracy.
Acknowledgements
We are very grateful to Peter Savill and Tim Whitmore for helpful ideas and assistance with ®eldwork.
Mario Cortina Borja provided invaluable statistical
advice. David B. Clark, Christian Messier and Fidji
Gendron gave helpful feedback on the manuscript.
We thank NORDISK Timber Ltda. for their hospitality during our stay at Pataua Forest. The Northmoor Trust is thanked for permission to work in
Little Wittenham Nature Reserve. The entire First
Year Honours Moderations Biology class at Oxford
University assisted with ®eld work at Orielton.
References
Anderson, M.C. (1964) Studies of the woodland light climate. I. The photographic computation of light conditions. Journal of Ecology, 52, 27±41.
Baldocchi, D. & Collineau, S. (1994) The physical nature
of solar radiation in heterogeneous canopies: spatial
and temporal attributes. Exploitation of Environmental
Heterogeneity by Plants (eds M.M. Caldwell &
R.W. Pearcy), pp. 21±71. Academic Press, San Diego,
CA.
Becker, P. & Smith, A.P. (1990) Spatial autocorrelation of
solar radiation in a tropical moist forest understorey.
Agricultural and Forest Meteorology, 52, 373±379.
Brown, N. (1993) The implications of climate and gap
microclimate for seedling growth in a Bornean lowland
rain forest. Journal of Tropical Ecology, 9, 153±168.
Clark, D.A. & Clark, D.B. (1992) Life history of canopy
and emergent trees in a neotropical rain forest. Ecological Monographs, 62, 315±344.
Clark, D.B., Clark, D.A., Rich, P.M., Weiss, S. & Oberbauer, S.F. (1996) Landscape scale evaluation of
understorey light and canopy structure: methods and
application in a neotropical lowland rain forest. Canadian Journal of Forest Research, 26, 747±757.
Comeau, P.G., Gendron, F. & Letchford, T. (1998) A
comparison of several methods for estimating light
under a paper birch mixed wood stand. Canadian Journal of Forest Research, 28, 1843±1850.
Dawkins, H.C. (1958) The Management of Tropical High
Forest with Special Reference to Uganda. Tropical Forestry Paper 34. Imperial Forestry Institute, Oxford,
UK.
Garrison, G.A. (1949) Uses and modi®cations for the
`Moosehorn' crown closure estimator. Journal of Forestry, 47, 733±735.
Gendron, F., Messier, C. & Comeau, P.G. (1998) Comparison of various methods for estimating the mean grow-
1053
N. Brown et al.
# 2000 British
Ecological Society
Journal of Applied
Ecology, 37,
1044±1053
ing season percent photosynthetic photon ¯ux density
in forests. Agricultural and Forest Meteorology, 92, 55±
70.
Jennings, S.B., Brown, N.D. & Sheil, D. (1999) Assessing
forest canopies and understorey illumination: canopy
closure, canopy cover and other measures. Forestry
Oxford, 72, 59±73.
Lemmon, P.E. (1956) A spherical densiometer for estimating overstorey density. Forest Science, 2, 314±320.
Machado, J.L. & Reich, P.B. (1999) Evaluation of several
measures of canopy openness as predictors of photosynthetic photon ¯ux density in deeply shaded coniferdominated forest understorey. Canadian Journal of
Forest Research, 29, 1438±1444.
Nicotra, A.B., Chazdon, R.L. & Iriarte, S.V.B. (1999)
Spatial heterogeneity of light and woody seedling
regeneration in tropical wet forests. Ecology, 80, 1908±
1926.
Parent, S. & Messier, C. (1996) A simple and ecient
method to estimate microsite light availability under a
forest canopy. Canadian Journal of Forest Research,
26, 151±154.
Rich, P.M., Clark, D.B., Clark, D.A. & Oberbauer, S.F.
(1993) Long-term study of solar-radiation regimes in a
tropical wet forest using quantum sensors and hemispherical photography. Agricultural and Forest Meteorology, 65, 107±127.
Rich, P.W. (1990) Characterizing plant canopies with
hemispherical photographs. Remote Sensing Reviews,
5, 13±29.
Siegel, S. & Castellan, N.J. (1988) Nonparametric Statistics
for the Behavioural Sciences, 2nd edn. McGraw-Hill,
New York, NY.
Vales, D.J. & Bunnell, F.L. (1988) Comparison of methods
for estimating forest overstorey cover. I. Observer
e€ects. Canadian Journal of Forestry Research, 18,
606±609.
Whitmore, T.C., Brown, N.D., Swaine, M.D., Kennedy,
D.K., Goodwin-Bailey, C.I. & Gong, W.K. (1993) Use
of hemisphere photographs in forest ecology: measurement of gap size and radiation totals in a Bornean tropical rain forest. Journal of Tropical Ecology, 9, 131±151.
Received 12 November 1999; revision received 29 May 2000