Tree Physiology 20, 49–56
© 2000 Heron Publishing—Victoria, Canada
Light spectral composition in a tropical forest: measurements and
model
FRANCISCO DE CASTRO1,2
1
Department of Biology, University of Puerto Rico, P. O. Box 23360, UPR. Río Piedras, Puerto Rico 00931-3360, USA
2
Present address: Department of Fisheries and Wildlife, University of Minnesota, 200 Hodson Hall, 1980 Folwell Avenue, St. Paul, MN 55108, USA
Received February 18, 1999
Summary I present a simple model that simulates vertical
variations in the light spectrum within a forest canopy. The
model considers only the vertical, downward transmission of
light. The light in each canopy level was assumed to consist of
non-intercepted radiation and radiation intercepted within the
level and transmitted. The spectrum of non-intercepted light in
each canopy level is the same as that of incident light above
the canopy (input parameter), whereas the spectrum of transmitted light depends on leaf area index (LAI) and the mean
transmission spectrum of leaves. The model was tested in a
forest and provided adequate predictions of measured values.
Stronger deviations were produced in the near infrared (NIR)
waveband in lower canopy levels. Multiple regression between
LAI, as the dependent variable, and spectral characteristics
(Blue, Green, Red and NIR intensities) had an r 2 of 0.926. As
a complement to other methods, I suggest light spectrum
analysis as a non-destructive technique for estimating LAI in
forest canopies.
Keywords: canopy, irradiance, radiation, leaf area index,
Puerto Rico, transmission.
Introduction
Forests exhibit much variation in light environment, not only
in light quantity, but also in light quality (spectral characteristics). Both factors affect many important processes, such
as photosynthesis, growth and morphogenesis of plants, communication among animals, microhabitat selection (Endler
1993), and seed germination (Metcalfe 1996, Lee et al. 1996).
Many canopy light interception models have been developed,
based on a variety of objectives and assumptions (e.g., Federer
1971, Charles-Edwards and Thorpe 1976, Jarvis and Leverenz
1983, Charles-Edwards et al. 1986, Cannell et al. 1987, Wang
and Baldocchi 1989, Wang and Jarvis 1990, Yang et al. 1990,
Gholz et al. 1991, Myneni 1991, Oker-Blom et al. 1991,
Pukkala et al. 1991, West and Welles 1992, Bégué 1993, Ryel
et al. 1993). However, in most models of radiation interception, only the quantitative, not the qualitative, aspect of light is
taken into account. Only a limited range of the spectrum, the
photosynthetic active radiation (PAR), is usually considered,
because it is the primary variable affecting photosynthesis and
plant growth. Some models also include the near infrared
(NIR) waveband (Wang and Jarvis 1990). In many, if not all
models, a parameter of primary importance is leaf area index
(LAI). The importance of LAI has led to the development of
several devices to estimate it that, in general, rely on the
inversion of models of light transmittance through the canopy.
There are also many models in the literature dealing with
the detailed spectral characteristics of reflected radiation, as
recorded by satellites (e.g., Kuusk 1994, 1995a, 1995b, 1996,
Lewis 1994, Baret et al. 1995, Otterman et al. 1995). These
models allow estimation of some important biophysical properties of the canopy over large areas, but they are not designed
to predict the radiation transmitted through the canopy, and so
they cannot predict the radiation environment at different
heights from the top of the canopy to the forest floor.
In this work, I present measurements of the light spectra of
a 21-m vertical profile within the canopy of a tropical forest.
The relationship between the spectral composition of light,
LAI and optical properties of leaves is used to develop a model
that simulates changes in the spectral composition at several
heights in the canopy. The model considers only one dimension by assuming that the flux of radiation is directly downward. This is justified by the fact that, in a dense tropical forest,
the gradient of light is strongly determined by the vertical axis.
Although the model assumes that the flux of radiation is from
top to bottom, this does not imply that the sun has to be at an
elevation of 90°.
Among the many factors that affect light quality as it traverses the canopy, filtering of light through foliage elements is
the most important. Thus, it is assumed that the composition
of light can be derived from the structure and optical characteristics of the canopy (Endler 1993), represented by LAI, leaf
angles and the optical properties of leaves. The model parameters are the vertical profile of LAI and leaf angles, mean
transmission spectrum of leaves, and the spectrum of incident
radiation. The model calculates the spectrum of light for each
of the heights at which LAI was measured.
Materials and methods
The study was conducted near El Verde Field Station (18°20′
N, 65°49′ W) in the north-western section of the Luquillo
Experimental Forest (Puerto Rico). The area (a 16-ha perma-
50
DE
CASTRO
nent plot) supports tabonuco (Dacryodes excelsa Val.) forest
(subtropical wet forest in the Holdridge system; Ewel and
Whitmore 1973). In addition to tabonuco, the forest in the area
is dominated by the palm Prestoea montana (R. Graham)
Nichols, and the trees Manilkara bidentata (A. DC.)
A. Cheval. and Sloanea berteriana (Choisy) (Odum and Pigeon 1970, Brown et al. 1983). Stocking density is approximately 800 trees ha −1 and includes 88 species. Because of
human disturbance, none of the forest in the 16-ha plot can be
considered primary forest. However, the southern part of the
plot (about 6 ha) is similar to virgin stands of tabonuco forest
(Odum 1970). A more detailed description of the plot is provided by Zimmerman et al. (1994).
Measurements of LAI, mean leaf angle, and the spectrum of
light between 0.35 and 1.15 µm were made at 2-m vertical
intervals from a 25-m tower located near a corner of the 16-ha
plot.
calculated as the mean of the ten measurements made at that
level.
The top of the tower was above the canopy, so measurements at the highest level represent conditions in the open.
Profiles of spectra were taken at midday on April 24 and 26
(one each day), when the sun was almost directly overhead at
this latitude and time of year. I also took samples of leaves of
all four of the species within reach of the tower at all levels.
The transmission spectra (between 0.35 and 1.1 µm) of three
randomly chosen leaves per species were measured with the
spectroradiometer and an external integrating sphere (Li-Cor
LI-1800-12). The leaves were maintained in high humidity and
exposed to light until they were processed, always within a day
of sampling. It was assumed that there were no significant
differences in optical properties of the leaves along the vertical
gradient (cf. Poorter et al. 1995). This assumption was justified
by the need to keep both the model manageable and the
number of parameters to a minimum.
Leaf area index and leaf angles
I measured LAI and mean leaf angles with an LAI-2000
canopy analyzer (Li-Cor Inc., Lincoln, NE), which estimates
both variables simultaneously (LAI is presented as half of
all-sided leaf area). The LAI measurements were made under
conditions considered ideal for the LAI-2000 (LAI-2000 manual (1990)): i.e., in late afternoon, with a cloudless sky and the
sun below 27°. These conditions ensured that direct light did
not hit the sensor rings, which have a minimum elevation range
of 27°. In addition, a view restriction cap of 15° was used to
avoid the effect of excessive brightness of the sky zone in the
direction of the sun. Although the Li-Cor LAI-2000 detects
branches and twigs as well as leaves (Li-Cor 1990, Welles and
Norman 1991), LAI estimates are probably affected more by
clumping of the foliage than by the inclusion of non-foliar
elements.
Spectra and PPFD
Spectral measurements (mean of three scans) were taken with
a Li-Cor spectroradiometer equipped with a remote cosine
receptor (LI-1800-11), which was held level approximately
1 m outside the tower, and was connected to the spectroradiometer with a fiber optic probe (LI-1800-2). The sensor was
placed in shade except at the two uppermost canopy levels
(where there was almost no shade). Shaded positions were
chosen to maximize the effect of light filtering by leaves. The
locations were evenly spaced within the 2-m range of the
tower. The sensor was placed at the two extremes of the
platform and at the center, always oriented in the same direction. It was impossible to measure the light spectrum over a
wider area, because the tower was the only means of canopy
access.
At each canopy level, photosynthetic photon flux density
(PPFD) was measured with a Li-Cor quantum sensor, connected to a data logger. The quantum sensor was attached to
the sensor of the spectroradiometer, so both were located at the
same point. Ten measurements of PPFD were made in the time
that the spectroradiometer took to complete the three scans
(approximately 1 min). The PPFD value for each level was
Description of the model
The model was applied in two steps. First the total amount of
non-intercepted radiation reaching each canopy level was calculated, then the amount of radiation absorbed in each level
was calculated as the difference between non-intercepted radiation in two consecutive levels. Second, once the total radiation in each level was known, its spectral composition was
calculated according to the assumptions explained below. In
this work, the term “ total incident radiation” means the total
downward solar radiation flux density (Rt) between 0.35 and
1.1 µm, which were the limits of the waveband examined.
The canopy was divided into several levels of equal thickness. Each canopy level was characterized by its LAI and mean
leaf angle (α). It was assumed that the flux of light is vertical.
The radiation that passes through any level of the canopy
without being intercepted (Ru) can be expressed by a negative
exponential function:
Ru = Rte−(Lk),
(1)
where Rt is total incident radiation (direct plus diffuse), L is
LAI in level H, and k is the extinction factor, which depends
on the relative angles of leaves (α) to the sun. Because radiation flux was assumed to be vertical, the sun elevation was
considered to be 90°, so k was always the cosine of mean leaf
angle (α). The amount of radiation intercepted (Ri) within a
canopy level H, is the difference between the amount of nonintercepted radiation at that level and the next (H + ∆H):
Ri (H) = Ru(H+∆H)−Ru(H).
(2)
Part of the intercepted radiation is reflected, another part is
absorbed and the rest is transmitted downward and passed to
the next canopy level. The amount of radiation transmitted in
each wavelength depends on the mean transmission coefficient
of the leaves for that wavelength. So, light changes both
quantitatively (total amount) and qualitatively (spectrum) as it
TREE PHYSIOLOGY VOLUME 20, 2000
LIGHT SPECTRAL COMPOSITION IN A TROPICAL FOREST
passes through the canopy. In the application of the model, the
values of non-intercepted radiation (Ru) and intercepted radiation (Ri) were computed for each level, based on the values of
LAI and leaf angles. The Ru and Ri values were then divided
into the different wavelengths according to the following assumptions. (1) The proportion of radiation at each wavelength
in non-intercepted radiation at a given level is the same as in
the incident radiation above the canopy. (2) The proportion of
radiation intercepted that is transmitted in each wavelength
depends on the transmission coefficient of leaves in that wavelength. (3) Radiation intercepted within the level and transmitted (in each wavelength) is added to the corresponding
wavelengths in the current level. (4) Second-order reflections
are considered negligible.
Non-intercepted radiation
We can express total incident radiation (Rt ) between two
wavelengths λ1 and λ2 as:
λ2
Rt = ∫ R(λ)dλ .
(3)
λ1
Because spectroradiometers have a limited spectral resolution (1 nm in the case of the LI-1800), the measurement of Rt
is not truly continuous, as expressed in the above formula, but
is discrete, made at finite wavelength intervals (∆λ). Thus,
Equation 3 can be expressed as:
λ2
Rt = ∆λ∑ R(λ).
(4)
λ1
The proportion of radiation at wavelength λ with respect to
the total radiation at a given height H, denoted c(λ, H), can be
expressed as:
c (λ,H ) =
R(λ,H)
Rt
51
The model was applied iteratively to every canopy level,
producing a simulated spectrum between 0.35 and 1.1 µm for
each level. The simulated spectra were then compared with
measured spectra to test the model. From the simulated spectra
it was possible to calculate PPFD for further comparisons with
measured PPFD values.
Results and discussion
The profile and the accumulated profile of LAI at the tower are
shown in Figure 1. The profile shows an approximately bimodal distribution, with a maximum accumulated LAI of 7 at
ground level.
It should be noted that we do not know the true LAI, only
the estimate provided by the LAI-2000. Several studies have
reported that the LAI-2000 may underestimate the true LAI by
as much as 35–40% (Gower and Norman 1991) or 45% (Chason et al. 1991) because of foliage clumping, which is not
considered in the underlying theory of the device. However,
Chason et al. (1991) found that elimination of the three outer
rings of the sensor in the calculations of LAI yielded an
estimate that was not significantly different from the true value
(4.17 ± 0.73 versus 4.89 ± 0.95), even when a random distribution of foliage was assumed. I compared the results obtained
with different numbers of rings with those obtained with all
five rings (Figure 2). The relationships were always quite
good, with the slopes close to one, the intercepts close to zero
and high r 2 values, which decreased as rings were eliminated
in the calculations. It was concluded, therefore, that all of the
estimates were similar in this particular location whatever the
number of rings (but see de Castro and Fetcher 1998). When
fewer than five rings were used in the calculation of LAI, the
field of view of the device was also reduced: the outmost ring
points to an elevation of 27° above the horizon, whereas the
(5)
.
Thus, non-intercepted radiation in wavelength λ, at level H,
Ru(λ, H), can be calculated as the product of the coefficient
c(λ, H) for incident radiation, and total non-intercepted radiation at level H.
Transmitted radiation
Part of the radiation intercepted in each level is reflected, part
is absorbed and part is transmitted. The transmitted proportion
depends on the leaves’ transmission coefficient at each wavelength (τ). Following the same procedure as described for
non-intercepted radiation, we can express the radiation transmitted within level H, at wavelength λ, Rt(λ, H) as:
Rt(λ,H ) = Ri(H)c(λ,H)τ(λ).
(6)
So, the final expression of the model is:
R(λ,H ) = Ru(H)c(λ,0) + Ri(H)c(λ,H)τ(λ).
(7)
Figure 1. Vertical profile of leaf area index (LAI, half of all-side leaf
area) per canopy level (bars) and accumulated LAI (line) against
height of measurement.
TREE PHYSIOLOGY ON-LINE at http://www.heronpublishing.com
52
DE
CASTRO
Figure 2. Comparison of LAI estimates made with different numbers of
sensor rings of the LAI-2000 (vertical
axis), with that obtained with five
rings (horizontal axis).
innermost ring points vertically. On the other hand, the sensor
of the spectroradiometer always receives radiation from all
directions. For these reasons, I used the estimate of LAI obtained with all five rings in this work.
Because the transmittance spectra of the leaves of species
found around the tower were similar (Figure 3), the mean of
the four species was used as the transmission coefficient in the
calculations.
The percentage of PPFD at ground level with respect to total
PPFD at the top of the canopy was 0.53 and 1.35% on the two
days of sampling, respectively. This value is similar to values
found in the understory of other tropical forests (cf. Chazdon
and Fetcher 1984) and is consistent with the heavy shading at
the study site as a result of the relatively high LAI.
The measured vertical variations in the spectrum of light are
shown in Figures 4a and 4c. In the visible portion of the
spectrum (0.4 to 0.7 µm), the radiant flux density decreased
uniformly with decreasing height, but the form of the spectrum
was mostly unchanged. In contrast, the proportion of the near
infrared waveband (0.7 to 1.1 µm) relative to the total radiation
increased with decreasing height. As a result, the radiation at
the lower levels of the canopy was enriched in NIR wavelengths.
The vertical variation in the ratio of red to far-red irradiance
is shown in Figure 5. The red and far-red wavelengths are
defined as centered on 0.66 and 0.73 µm, respectively, with a
bandwidth of 10 nm (Smith 1994). At the top of the canopy,
the R:FR ratio was 1.35, decreasing to 0.36 at ground level.
These values are similar to those in other tropical forests (Lee
et al. 1996).
I also examined the relationship between LAI measured by
the LAI-2000 and the spectral composition of solar radiation.
First, the irradiance in the blue (0.4–0.5 µm), green (0.5–0.6
µm), red (0.6–0.7 µm), and NIR (0.7–1.1 µm) wavebands was
calculated. Then I calculated the ratio of each waveband color
Figure 3. Transmission spectra of leaves of the species around the
tower. Values plotted are the means of three leaves per species.
TREE PHYSIOLOGY VOLUME 20, 2000
LIGHT SPECTRAL COMPOSITION IN A TROPICAL FOREST
53
Figure 4. Spectral irradiance at several
heights within the canopy. The graphs
on the left-hand side (a, c) are observed values, and those on the righthand side (b, d) are simulated by the
model. The two upper graphs (a, b)
show the values of the upper part of
the canopy (from 23 to 11 m). The
two lower graphs (c, d) show the
lower canopy levels (from 9 to 1 m).
Note that the spectrum at 23 m in b, is
not simulated, but measured, because
it is an input parameter for the model.
at each canopy level to its equivalent in incident radiation
above the canopy. These ratios were used as independent
variables in a multiple linear regression with LAI as the dependent variable (Table 1). The r 2 was 0.926, indicating the
potential value of the spectral composition of light as a basis
for predicting LAI. It should be emphasized that the aim of this
procedure is not to use the model to predict LAI; this is not
possible because the model uses LAI as an input parameter.
The irradiances used as independent variables in the regression
were those measured with the spectroradiometer, not those
predicted by the model, whereas the dependent variable (LAI)
was measured with the LAI-2000.
To test the model, values of total irradiance and irradiance
in different wavebands (blue, green, red and NIR), were calculated from both observed and predicted spectra, and then
compared by linear regressions. The shape of the spectra can
also be used as a subjective test of the adequacy of the model.
In a further test, the actual PPFD measured with the quantum
sensor was compared with predicted values of PPFD calculated from the simulated spectra with the Li-Cor software for
the LI-1800.
The predicted values of total irradiance and PPFD were in
Table 1. Multiple linear regression between the spectral irradiance at
different wavebands and leaf area index. The wavebands considered
are: blue (0.4–0.5 µm), green (0.5–0.6 µm), red (0.6–0.7 µm) and NIR
(0.7–1.1 µm). The multiple r 2 = 0.926, P < 0.0001. Note that the only
positive coefficient is that of the green, indicating enrichment in green
wavelengths as light passes through the canopy.
Effect
Coefficient
Standard
error
Standard
coefficient
t-Value
Constant
Blue
Green
Red
NIR
9.006
−644.038
876.628
−119.462
−122.122
1.787
548.164
888.325
265.955
78.582
0.0
−71.560
96.993
−13.188
−13.014
5.038
−1.175
0.987
−0.449
1.554
Analysis of Variance
Figure 5. Vertical profile of the red:far-red ratio. The values shown are
the ratio between irradiances at red and far-red wavelengths, defined
as centered on 0.66 µm and 0.73 µm respectively, with a bandwidth of
10 nm.
Source
Sum
of squares
df
Mean
square
F
P
Regression
Residual
64.057
5.153
4
7
16.014
0.736
21.754
< 0.001
TREE PHYSIOLOGY ON-LINE at http://www.heronpublishing.com
54
DE
CASTRO
close agreement with measured values (Figure 6). The comparison of simulated irradiance in the different wavebands
with the measured values also had high r 2 values (Figure 7).
The form of the predicted spectra, on the other hand, showed
an overestimation in the NIR region at the upper canopy levels,
whereas it showed an underestimation at the lower canopy
levels (Figure 4).
Conclusions
The model predictions were in relatively good agreement with
measured values, especially in the visible part of the spectrum.
The main differences were an overestimate of the amount of
near-infrared radiation at the upper canopy levels, and an
underestimate at the lower canopy levels. These discrepancies
could be caused by several factors. First, the predicted spectra
were based on the spectrum of incident light. This was measured only once, at the beginning of the sampling period, but it
can change quite rapidly because of clouds, and thus, when the
lower canopy levels were measured, the incident light may
have had a different spectrum. This problem could be solved
by using a second spectroradiometer to measure incident light
at regular intervals. The spectrum of incident light also
changes with the sun’s azimuth (McFarland and Munz 1975).
Because the measurement process takes some time, this spectral change could affect the accuracy of the prediction. Second,
Figure 6. Comparison of observed (points) and predicted (lines) values
of total irradiance (above) and photosynthetic photon flux density
(PPFD, below) in the vertical profile within the canopy. The measured
values of total irradiance were obtained with a Li-Cor LI-1800 spectroradiometer. The measured values of PPFD were obtained with a
quantum sensor.
Figure 7. Regressions between observed (abscissa, independent variable) and predicted (ordinate, dependent variable) values for irradiance
at different wavebands: blue (0.4–0.5 µm), green (0.5–0.6 µm), red
(0.6–0.7 µm) and near infrared (NIR, 0.7–1.1 µm).
and probably more importantly, the reflection of radiation was
not included in the model. Because the reflection of foliage in
the infrared is high, incorporating reflection would increase
the predicted values in this range of the spectrum producing a
better prediction.
The good agreement between measured and simulated values suggests that the assumptions of the model, in particular
the assumption that all the radiation is directly downward, are
not seriously in error. Direct radiation comes mostly from the
sun with varying angles, and diffuse radiation come from all
the sky hemisphere, but the filtering effect of the canopy
makes the downward gradient the most important. The omission of reflected radiation from the model is probably the cause
of some of the observed discrepancies. The model presented
here could be incorporated in more complex models of radiation interception, to enhance the detail of the predictions. The
wavelength interval used in this work (2 nm) is probably too
small for most purposes, and could be increased to simplify the
calculations. It could be useful to use irregular wavelength
intervals matching the main characteristics of the absorption
spectra of leaves.
The variation in light through the canopy, both the total
amount and its spectral composition, is mostly caused by
interception by leaves. This raises the possibility of using the
spectral properties of measured transmitted radiation to estimate LAI. The multiple regression between LAI (dependent
variable), as estimated by the LAI-2000, and the spectral
characteristics measured with the spectroradiometer (blue,
green, red and NIR values) had a high r 2 value, suggesting that
the method could provide a non-destructive estimation of LAI
in this type of forest. However, it will first be necessary to test
this technique for predicting LAI from spectral characteristics
against direct measurements of LAI. It should be emphasized
that the LAI data used in this paper are not measurements, but
the results of the inversion of a radiation model that is incor-
TREE PHYSIOLOGY VOLUME 20, 2000
LIGHT SPECTRAL COMPOSITION IN A TROPICAL FOREST
porated in the manufacturer’s software for the LAI-2000.
Changes in spectral composition of the light at intermediate
and large scales should be less affected by the spatial distribution of foliage than other methods for estimating LAI based on
the inversion of radiation transmission models (Chason et al.
1991). It should be noted that the coefficients of the regression
will probably change for different vegetation types because of
differences in the transmission spectrum of the species and
different characteristics of incident light. It is also possible that
the same method will not work equally well in other types of
forests. In the study site, a tropical rain forest, the canopy is
relatively homogeneous and closed, and most of the radiation
reaching a given level in the canopy is filtered through the
foliage, which increases the effect of leaves on the spectral
composition of the light. This might not be the case in a sparser
forest, where a significant amount of radiation can reach the
ground unfiltered by vegetation, which could weaken the relationship between LAI and light spectral composition.
In the model tests, I found relatively good agreement between observed and predicted spectra, and the trends of decreasing total irradiance and PPFD down the canopy were
correctly predicted by the model. The model can be applied to
any range of wavelengths, provided that the total incident
radiation matches the spectrum of incident light and the spectral characteristics of leaves in all the required wavebands are
known.
Acknowledgments
Thanks to Adisel Montaña for her help in the field and valuable
criticism of earlier versions of this manuscript. This work was supported by a NASA/EPSCoR grant to the University of Puerto Rico.
References
Baret, F., J.G.P.W. Clevers and M.D. Steven. 1995. The robustness of
canopy gap fraction estimates from red and near-infrared reflectances—A comparison of approaches. Remote Sens. Environ.
54:141–151.
Bégué, A. 1993. Leaf area index, intercepted photosynthetically radiation, and spectral vegetation indices: A sensitivity analysis for
regular clumped canopies. Remote Sens. Environ. 46:45–69.
Brown, S., A.E. Lugo, S. Silander and L. Liegel. 1983. Research
history and opportunities in the Luquillo Experimental Forest. General Technical Report No. SO-44, USDA Forest Service, Southern
Experimental Station, New Orleans.
Cannell, M.G.R., R. Milne, L.J. Sheppard and M.H. Unsworth. 1987.
Radiation interception and productivity of willow. J. Appl. Ecol.
24:261–278.
Charles-Edwards, D. A. and M.R. Thorpe. 1976. Interception of diffuse and direct beam radiation by a hedgerow apple orchard. Ann.
Bot. 40:603–613.
Charles-Edwards, D.A., D. Doley and G.M. Rimmington. 1986. Modelling plant growth and development. Academic Press, Australia,
235 p.
Chason, J.W., D.B. Baldocchi and M.A. Huston. 1991. A comparison
of direct and indirect methods for estimating forest canopy leaf
area. Agric. For. Meteorol. 57:107–128.
Chazdon, R. and N. Fetcher. 1984. Photosynthetic light environments
in a lowland tropical forest in Costa Rica. J. Ecol. 72:553–564.
55
de Castro, F. and N. Fetcher. 1998. Three dimensional model of the
interception of light by a canopy. Agric. For. Meteorol. 90:215–
233.
Endler, J.A. 1993. The color of light and its implications. Ecol.
Monogr. 63:1–27.
Ewel, J.J. and J.L. Whitmore. 1973. The ecological life zones of
Puerto Rico and the U.S. Virgin Islands. US Forest Service Research Paper ITF-18. Institute of Tropical Forestry, Río Piedras,
Puerto Rico, 72 p.
Federer, C.A. 1971. Solar radiation absorption by leafless hardwood
forests. Agric. Meteorol. 9:3–20.
Gholz, H.L., S.A. Vogel, W.P. Cropper, Jr., K. McKelvey, K.C. Ewel,
R.O. Teskey and P.J. Curran. 1991. Dynamics of canopy structure
and light interception in Pinus elliottii stands, north Florida. Ecol.
Monogr. 61:33–51.
Gower, S.T. and J.M. Norman. 1991. Rapid estimation of leaf area
index in conifer and broad leaf plantations. Ecology 72:1896–1900.
Jarvis P.G. and J.W. Leverenz. 1983. Productivity of temperate deciduous and evergreen forests. In Encyclopedia of Plant Physiology, Vol. 12D. Eds. O.L. Lange, P.S. Nobel, C.B. Osmond and H.
Ziegler. Springer-Verlag, Berlin, pp 233–280.
Kuusk, A. 1994. A multispectral canopy reflectance model. Remote
Sens. Environ. 51:342–350.
Kuusk, A. 1995a. A markov-chain model of canopy reflectance.
Agric. For. Meteorol. 76:221–236.
Kuusk, A. 1995b. A fast invertible canopy reflectance model. Remote
Sens. Environ 51:342–350.
Kuusk, A. 1996. A computer-efficient plant canopy reflectance model.
Comput. Geosci. 22:149–163.
Lee, D.W., B. Krishnapillay, M. Marzalina, M. Haris and K.Y. Son.
1996. Irradiance and spectral quality affect Asian tropical rain
forest tree seedling development. Ecology 77:568–580.
Lewis, M.M. 1994. Species composition related to spectral classification in an Australian spiniflex hummock grassland. Int. J. Remote
Sens. 15:3223–3239.
Metcalfe, D.J. 1996. Germination of small-seeded tropical rainforest
plants exposed to different spectral compositions. Can. J. Bot.
74:516–520.
McFarland, W.N. and F.W. Munz. 1975. The photic environment in
clear tropical seas during the day. Vision Res. 15:1053–1070.
Myneni, R.B. 1991. Modeling radiative transfer and photosynthesis in
three-dimensional vegetation canopies. Agric. For. Meteorol.
55:323–344.
Odum, H.T., and R.F. Pigeon. 1970. A tropical rain forest. NTIS,
Springfield, VA.
Odum, H.T. 1970. The El Verde study area and the rain forest systems
of Puerto Rico. In A Tropical Rain Forest. Eds. H.T. Odum and R.F.
Pigeon. NTIS, Springfield, VA.
Oker-Blom, P., J. Lappi and H. Smolander. 1991. Radiation regime
and photosynthesis of coniferous stands. In Photon–Vegetation
Interactions. Eds. R.B. Myneni, and J. Ross. Springer Verlag,
Berlin, pp 469–499.
Otterman, J., T. Brakke and J. Smith. 1995. Effects of leaf-transmittance versus leaf-reflectance on bidirectional scattering from canopy soil surface: an analytical study. Remote Sens. Environ.
51:49–60.
Poorter, L., S.F. Oberbauer and D.B Clark. 1995. Leaf optical properties along a vertical gradient in a tropical rain forest canopy in Costa
Rica. Am. J. Bot. 82:1257–1263.
Pukkala, T., P. Becker, T. Kuuluvainen and P. Oker-Blom. 1991.
Predicting spatial distribution of direct radiation below forest canopies. Agric. For. Meteorol. 55:295–307.
TREE PHYSIOLOGY ON-LINE at http://www.heronpublishing.com
56
DE
CASTRO
Ryel, R.J., W. Beyschlag and M.M. Caldwell. 1993. Light field heterogeneity among tussock grasses: Theoretical considerations of
light harvesting and seedling establishment in tussocks and uniform
tiller distributions. Oecologia 98:241–246.
Smith, H. 1994. Sensing the light environment: the functions of the
phytochrome family. In Photomorphogenesis in Plants, 2nd Edn.
Eds. R.E. Kendrick and G.H.N. Kronenberg. Kluwer Academic,
Dordrecht, The Netherlands, 828 p.
Wang, H. and D.D. Baldocchi. 1989. A numerical model for simulating the radiation regime within a deciduous forest canopy. Agric.
For. Meteorol. 46:313–337.
Wang, Y.P. and P.G. Jarvis. 1990. Description and validation of an
array model MAESTRO. Agric. For. Meteorol. 51:257–280.
Welles, J.M. and J.M. Norman. 1991. Instrument for indirect measurement of canopy architecture. Agron. J. 83:818–825.
West, P.W. and K.F. Welles. 1992. Method of application of a model
to predict the light environment of individual tree crowns and its use
in a eucalypt forest. Ecol. Model. 60:199–231.
Zimmerman, J.K., E.M. Everham, R.B. Waide, D.J. Lodge, C.M.
Taylor and V.L. Brokaw. 1994. Responses of tree species to hurricane winds in subtropical wet forest in Puerto Rico: implications for
tropical tree life histories. J. Ecol. 82:911–922.
TREE PHYSIOLOGY VOLUME 20, 2000