Tree Physiology 22, 393–401
© 2002 Heron Publishing—Victoria, Canada
Short-term light and leaf photosynthetic dynamics affect estimates of
daily understory photosynthesis in four tree species
ELKE NAUMBURG1–3 and DAVID S. ELLSWORTH1,4
1
Nicholas School of the Environment, Duke University, Durham, NC 27708-0328, USA
2
Present address: Desert Research Institute, 755 E. Flamingo Rd, Las Vegas, NV 89119, USA
3
Author to whom correspondence should be addressed ([email protected])
4
School of Natural Resources and Environment, 430 E. University Ave., University of Michigan, Ann Arbor, MI 48109-1115, USA
Received May 29, 2000; accepted October 4, 2001; published online March 1, 2002
Summary Instantaneous measurements of photosynthesis
are often implicitly or explicitly scaled to longer time frames to
provide an understanding of plant performance in a given environment. For plants growing in a forest understory, results
from photosynthetic light response curves in conjunction with
diurnal light data are frequently extrapolated to daily photosynthesis (A day), ignoring dynamic photosynthetic responses to
light. In this study, we evaluated the importance of two factors
on A day estimates: dynamic physiological responses to photosynthetic photon flux density (PPFD); and time-resolution of
the PPFD data used for modeling. We used a dynamic photosynthesis model to investigate how these factors interact with
species-specific photosynthetic traits, forest type, and sky conditions to affect the accuracy of A day predictions. Increasing
time-averaging of PPFD significantly increased the relative
overestimation of A day similarly for all study species because of
the nonlinear response of photosynthesis to PPFD (15% with
5-min PPFD means). Depending on the light environment
characteristics and species-specific dynamic responses to
PPFD, understory tree A day can be overestimated by 6–42% for
the study species by ignoring these dynamics. Although these
overestimates decrease under cloudy conditions where direct
sunlight and consequently understory sunfleck radiation is reduced, they are still significant. Within a species, overestimation of A day as a result of ignoring dynamic responses was
highly dependent on daily sunfleck PPFD and the frequency
and irradiance of sunflecks. Overall, large overestimates of
A day in understory trees may cause misleading inferences concerning species growth and competition in forest understories
with < 2% full sunlight. We conclude that comparisons of A day
among co-occurring understory species in deep shade will be
enhanced by consideration of sunflecks by using high-resolution PPFD data and understanding the physiological responses
to sunfleck variation.
Keywords: Acer, Cornus, daily photosynthesis, Liquidambar,
Liriodendron, modeling, shade, sunflecks.
Introduction
A central goal of tree physiological ecology is to compare the
performance of species in different environments. In forest
understories, photosynthetic CO2 assimilation (Anet) is
strongly limited by light availability. Because portable photosynthesis systems are widely available, instantaneous leaflevel Anet is commonly measured in understory trees (e.g.,
Abrams and Mostoller 1995, Barker et al. 1997, Walters and
Reich 1999, DeLucia and Thomas 2000). These and other
studies in the past three decades have focused on describing
leaf photosynthetic characteristics as a first step in understanding how assimilation affects carbon uptake of species (Körner
1991), and hence growth and competitiveness for scarce resources in the forest understory. However, a detailed understanding of the linkages between assimilation and plant carbon
uptake requires longer-term estimates of photosynthesis than
readily measured instantaneous rates. Because of the high spatial and temporal variability of photosynthetic photon flux
density (PPFD) in forest understories, intensive all-day measurements in the field or modeling are required to estimate
daily leaf assimilation (Aday) in these environments. Relatively
few studies have directly measured Aday in understory plants
(Björkman et al. 1972, Pearcy and Calkin 1983, Weber et al.
1985, Pearcy 1987, Stickan and Zhang 1992, Pearcy et al.
1994), and most attempts to estimate assimilation at time
scales of days or weeks have been based on instantaneous
measurements of leaf Anet.
Because of the large influence of light on understory assimilation, the simplest and most frequently applied modeling approach to estimate Aday is based on the steady-state relationship between Anet and PPFD (e.g., the light response curve) in
conjunction with daily courses of PPFD (e.g., Chazdon 1986,
Ellsworth and Reich 1992, Zipperlen and Press 1996, Herrick
and Thomas 1999, Tang et al. 1999, Walters and Reich 1999,
Beaudet et al. 2000). However, there are several qualitative
limitations to this approach. Most importantly, these estimates
of Aday ignore both stomatal and biochemical limitations to
sunfleck photosynthesis as a result of stomatal closure and
394
NAUMBURG AND ELLSWORTH
photosynthetic enzyme deactivation during shade (Pearcy et
al. 1994). Both of these limitations require the use of more
elaborate modeling procedures than the simple light response
curve to predict Aday in understory plants (Gross et al. 1991,
Pearcy et al. 1994, Naumburg et al. 2001). Furthermore, overestimation of Aday can result from low-resolution PPFD data
that do not resolve sunflecks of 1 min or less. Because understory light environments are generally characterized by mostly
brief sunflecks with intermittent shade periods (e.g., Chazdon
1988), this overestimation of Aday is associated with the nonlinear response of photosynthesis to PPFD (Norman 1993).
In this study, we sought to determine if overestimates of Aday
caused by ignoring photosynthetic dynamics during sunflecks
and using long PPFD-averaging intervals result in large errors
in estimating longer-term assimilation in understory leaves.
We also determined if species-specific photosynthetic characteristics, overstory forest type and weather conditions affect
the degree of overestimation similarly. To assess these effects,
we used the dynamic sunfleck photosynthesis model of Pearcy
et al. (1997), which has been parameterized for four hardwood
species that vary in their dynamic sunfleck responses (Naumburg et al. 2001). Comparing the outputs from the dynamic
photosynthesis model to those from a steady-state model allowed us to answer the overall question in different scenarios.
Understanding the relative merits of different approaches for
predicting Aday is important because often the goal of modeling
Aday is to gain a better understanding of the effects of understory light characteristics and plant traits on plant carbon gain
and ultimately growth and survival in forest understories (e.g.,
Chazdon 1986, Zipperlen and Press 1996, Gill et al. 1998,
Tang et al. 1999, Walters and Reich 1999, Beaudet et al.
2000).
Methods
Study sites and PPFD measurements
To assess the effect of forest structure on modeling outcomes,
PPFD was measured in a 16-year-old Pinus taeda L. plantation with an abundant hardwood understory and in an adjacent
70–80-year-old oak–hickory forest with a sparse Cornus
florida L. understory, in Duke Forest near Durham, NC
(35°52′ N, 79°59′ W). Mean overstory tree height was 14 m in
the pine forest and 26 m in the hardwood forest. Both forests
had a closed canopy with a mean leaf area index of 4.58 for the
pine forest and ranging between 4.7 and 6.0 for the hardwood
forest. Understory PPFD was measured with gallium arsenide
photodiodes (GA-1118, Hamamatsu, Bridgewater, NJ) as described by Pointailler (1990). Photodiodes were calibrated
against a commercial quantum sensor (Model LI-190, Li-Cor,
Lincoln, NE) under a range of shade and sky conditions. Calibrated photodiodes were horizontally attached to leaves on
understory sapling branches 1–2 m above ground. Twentyfour locations in the pine forest and eight locations in the hardwood forest were randomly chosen. The four study species are
common in the understory of the pine stand where the PPFD
data were collected and photosynthetic dynamics were mea-
sured. In the hardwood stand, two of the study species (shade
intolerant) were rare. Data loggers (Campbell 23X, Campbell
Scientific, Logan, UT) read PPFD at 0.5-s intervals and recorded 5-s PPFD means for each photodiode over 11 h per day
centered around solar noon. Outside this time frame sunflecks
are rare. Data were collected for about 1 week in early June
1999, which included both a sunny day as well as several partially cloudy days. For the modeling, data from one sunny day
and one partially cloudy day were chosen. For the sunny day,
this resulted in a data set of 32 diurnal PPFD courses: 24 from
the pine and eight from the hardwood forest. Sixteen diurnal
PPFD courses were available in the pine forest for the partly
cloudy day.
To estimate the effect of increasing PPFD averaging intervals on photosynthetic estimates, the PPFD data were subsequently averaged over 1 and 5 min. For the purpose of analyzing sunfleck distributions and effects on leaf-level A day, we
defined sunflecks as readings where PPFD > 60 µmol m –2 s –1,
which is about twice the ambient diffuse PPFD observed during midday on sunny days. This value has no physiological
relevance but marked a clear cutoff for distinguishing sunflecks from shade. Histograms of sunfleck duration and mean
sunfleck irradiance as well as daily sums of PPFD and sunfleck PPFD were generated from the 5-s PPFD measurements.
Photosynthetic model and parameterization
To estimate the effects of PPFD averaging intervals and the
characteristic stomatal and biochemical limitations of each
species, we used the sunfleck model of Pearcy et al. (1997) implemented according to Naumburg et al. (2001). The model is
based on the Farquhar and von Caemmerer (1982) model of
photosynthesis, but also includes time constants for the activation and deactivation of photosynthetic enzymes as well as
stomatal opening and closing in response to changes in
irradiance (Pearcy et al. 1997). The model is implemented by
first calculating the equilibrium values for parameters at a
given PPFD. Based on the calculated parameter values, predicted values are determined in exponential equations, e.g.:
Rub (t ) =
Rub eq − Rub (t −1 )
τ
,
(1)
where Rubeq is the equilibrium Rubisco activation state for a
given light value, Rub (t) and Rub (t – 1) are the activation states
at time t and t – 1, respectively, and τ is the time constant for
Rubisco activation or deactivation.
Thus, the model predicts Anet as a function of PPFD and
takes into account both enzymatic and stomatal limitations to
photosynthesis as a function of previous PPFD. The model requires inputs of species-specific photosynthetic and stomatal
parameters as well as PPFD values with associated time intervals. The model was parameterized as described by Pearcy et
al. (1997), Kirschbaum et al. (1998) and Naumburg et al.
(2001) based on measurements presented in Naumburg and
Ellsworth (2000). The model was tested by Pearcy et al.
(1997) and Naumburg et al. (2001), and predicts sunfleck
TREE PHYSIOLOGY VOLUME 22, 2002
PHOTOSYNTHETIC DYNAMICS AND DAILY UNDERSTORY ASSIMILATION
photosynthesis with a coefficient of determination of 0.96 and
a slope of 0.93 when compared with measured sunfleck photosynthesis for the four species in this study (Naumburg et al.
2001, see also Figure 1). To predict Anet equivalent to Anet
based on steady-state photosynthetic light response curves,
time constants are set to zero and equations containing the
time constants bypassed. Thus, light-specific equilibrium values are used for model calculations in this scenario.
The sunfleck model was parameterized for four species:
Acer rubrum L. (red maple), Cornus florida (flowering dogwood), Liquidambar styraciflua L. (sweetgum) and Liriodendron tulipifera L. (tulip-poplar). Both steady-state and dynamic gas exchange data for model parameterization were
collected for each species in June 1998 when soil water content was high. The species differed significantly in their induction responses to changes in PPFD (enzyme deactivation/ activation and stomatal closing/opening rates) (Naumburg and
Ellsworth 2000). The species also differed in steady-state gas
exchange characteristics at low PPFD, whereas steady-state
light-saturated A net was similar among species (Table 1). The
principal difference in dynamic PPFD responses was a rapid
induction response in sunflecks and maintenance of relatively
high induction in shade by Liriodendron, whereas Acer had
lower steady-state stomatal conductance and lost induction
more rapidly in shade (Naumburg and Ellsworth 2000). Thus,
the four species provide a range of dynamic characteristics to
evaluate the effect of dynamic responses on Aday estimates.
The model was run in the dynamic mode for the 56 diurnal
5-s PPFD courses for each of the four species-specific parameterizations. In addition, the model was run in the steadystate mode for the 5-s, 1-min and 5-min averaged PPFD
courses for each species-specific parameterization.
395
Data analyses
For the sunny day model, results from two photodiodes in the
hardwood forest resulted in negative Aday estimates. Two additional diurnal PPFD courses during the partly cloudy day resulted in negative Aday. These data were excluded from further
analysis because they yielded negative ratios (see below).
We used two analyses to assess the overestimate of Aday
caused by omitting dynamic responses and using averaged
PPFD data. First, we calculated ratios by dividing the 5-s
5 -s
steady-state model predictions (Aday
− s ) by dynamic model pre5 -s
dictions (Aday − d ). Thus, assuming that the dynamic model outputs more closely approximate the field behavior of understory trees in variable light (Figure 1; Pfitsch and Pearcy 1989,
Pearcy et al. 1994, Naumburg et al. 2001), ratios greater than
1.0 give the relative over-prediction of Aday resulting from ignoring dynamic light responses. To assess the effect of using
1- or 5-min average PPFD data, ratios of Aday estimates at time
x min
interval × Aday
− s relative to the 5-s Aday were calculated, e.g.,
x min
x min
5 -s
5 -s
A day − s / A day − d for dynamic model results and A day
− s / A day − s for
x min
5 -s
steady-state model results. A day − s / Aday −d evaluated the combined effect of using averaged PPFD data and omitting inducx min
5 -s
tion dynamics, whereas A day
− s / A day − s evaluates the effect of
using averaged PPFD data on Aday alone. We used one-sample
t-tests (one-tailed) to determine if the ratios were significantly
greater than 1.0.
Second, the data were subjected to analyses of covariance
with tests for separate slopes and intercepts for the four species. For these analyses, which were only conducted for the
sunny day data, Aday estimates derived from the steady-state
model (5 s, 1 min, 5 min) were dependent variables. Independent variables were the 5-s dynamic model outputs and species
coded as binary variables. The analyses were conducted as
stepwise multiple regressions with α < 0.01 as the entry crite-
Figure 1. Comparison of net photosynthesis (Anet) predicted by either the
steady-state or dynamic model to measured Anet over the course of 1 day.
Measurements were aborted in the afternoon because of thunder storms. Total PPFD intercepted by the leaf over
the 6.75 h measurement period was
1.84 mol m –2, measured Anet was
35.2 mmol m –2, steady-state model
Anet predictions were 40.3 mmol m –2,
and dynamic model Anet predictions
were 38.0 mmol m –2. The model predictions are equivalent to a 14 and 8%
overestimation of the measured value.
TREE PHYSIOLOGY ONLINE at http://heronpublishing.com
396
NAUMBURG AND ELLSWORTH
Table 1. Steady-state and dynamic leaf photosynthetic characteristics of four contrasting tree species that were incorporated in the dynamic
model. A qualitative comparison of the species with respect to stomatal dynamics and induction highlighted differences in dynamic behavior represented by time constants in the dynamic model.
Light saturated Anet (µmol m –2 s –1)
Light compensation point (µmol m –2 s –1)
Dark respiration (µmol m –2 s –1)
Rate of stomatal opening/induction gain1
Rate of stomatal closure/induction loss1
1
Acer
Cornus
Liquidambar
Liriodendron
6.6
16
0.50
rapid
rapid
4.7
15
0.50
slow
slow
6.4
18
0.47
very slow
very slow
5.8
14
0.47
rapid
slow
From Naumburg and Ellsworth (2000).
rion. In this analysis, species-specific slope estimates are
equivalent to the ratios calculated above if the regression lines
have a zero intercept. We used this approach rather than an
ANOVA because there was a possibility that a significant intercept would strongly affect ratio estimates at low Aday values.
The regression analyses were conducted with the SAS statistical software package (Version 8, SAS Institute, Cary, NC).
Results and discussion
Understory light environment
Irradiance was relatively low in the understories. Daily PPFD
during the sunny day (11-h measurement period) ranged between 1.1 and 9.3 mol m –2 day –1 for the 24 sampling locations
in the pine forest understory. Daily PPFD in the closed-canopy
hardwood forest ranged between 0.5 and 2.4 mol m –2 day –1.
These ranges are equivalent to 2–17% of the above-canopy radiation for the pine forest and 1–4% for the hardwood forest.
Daily PPFD was 20% lower for the partly cloudy day than for
the sunny day both above the overstory and in the understory.
Reported relative solar irradiances for sunny days range between 7 and 10% for conifer stands younger than 30 years
(Washitani and Tang 1991, Chen and Klinka 1997) and 1 and
6% for mature conifer and hardwood forests (Pfitsch and
Pearcy 1989, Canham et al. 1994). Thus, PPFD data from our
stands are comparable with values reported for summertime in
other temperate forests.
During the sunny day, contributions of sunflecks to daily
PPFD tended to be lower in the hardwood forest than in the
pine forest even when the forest types were compared over the
same range of daily PPFD (Table 2). Despite a lower sunfleck
contribution, the hardwood forest had a greater number of
short-duration (< 30 s) low-irradiance sunflecks (Figure 2)
than the pine forest at comparable daily PPFD. During the
partly cloudy day, sunfleck contribution to overall daily PPFD
decreased, whereas the number of sunflecks increased. This
was largely a result of an increase in short-duration and lowirradiance sunflecks relative to the sunny day (Figures 2c and
2d) that was caused both by clouds and generally windier conditions. Sunfleck contributions to daily PPFD ranging between 1 and 90% have been reported (Weber et al. 1985,
Pfitsch and Pearcy 1989, Washitani and Tang 1991, Gildner
and Larson 1992, Roden and Pearcy 1993, Canham et al. 1994,
Chen and Klinka 1997), but no clear tendency of differences
related to forest type emerge from these data.
Estimates of daily leaf-level photosynthesis
In support of our hypothesis, we found that omitting dynamic
responses to variable light and using averaged PPFD data for
modeling A day resulted in higher Aday estimates than using a
dynamic model or high-resolution PPFD data (Figure 3). In
addition, the effect of omitting dynamic responses on Aday
overestimation depended on both species-specific dynamic
photosynthetic characteristics and sunfleck distribution
throughout the day.
PPFD data resolution Steady-state model results using the
1-min and 5-min PPFD data showed mean overestimates of 5
and 15%, respectively, relative to the 5-s steady-state model results for the sunny day and 5 and 10%, respectively, for the
partly cloudy day (open bars, Figure 3). Lower overestimates
for the partly cloudy day data were expected, given the lower
daily sunfleck PPFD (Table 2). Regression analysis of the
sunny day data confirmed that overestimates increased with increasing averaging intervals and the increases were similar for
the four species. Only Liriodendron had a significantly lower
slope than the other species for the 5-min PPFD data (F1,117 =
11, P = 0.001). Thus, the relative overestimation of Aday by
steady-state models was mostly a function of the PPFD averaging interval and sunfleck PPFD, and was largely independent
of species-specific photosynthetic traits.
Species-specific dynamic photosynthetic characteristics
For Aday estimates evaluated against the 5-s dynamic model as
denominator, species-specific PPFD responses affected how
much photosynthesis was overestimated by the steady-state
model (solid bars, Figure 3). Regressions of the sunny day data
x min
5 -s
A day
− s versus A day − d showed that Liriodendron consistently
had the lowest slope of the study species (F1,115 > 190, P <
0.001) followed by Cornus (F1,115 > 30, P < 0.001). This means
that Aday estimates based on steady-state parameters for Liriodendron were closer to the 5-s dynamic model results than
those of the other species. For Liriodendron, the species expected to have the lowest limitations to sunfleck photosynthesis based on its rapid induction gain and slow induction loss
(Table 1), the errors from the dynamic model and steady-state
TREE PHYSIOLOGY VOLUME 22, 2002
PHOTOSYNTHETIC DYNAMICS AND DAILY UNDERSTORY ASSIMILATION
397
Table 2. Mean and standard error of understory PPFD characteristics in a loblolly pine and an oak–hickory hardwood forest. The PPFD data were
derived from 11-h-long photodiode measurements at 5-s intervals during a sunny day. Dense shade is defined as PPFD of 1–2 µmol m –2 day –1.
All sites
All pine
All hardwood
Pine: dense shade
Hardwood: dense shade
Sunny day
Number of photodiode sites
Daily PPFD (mol m –2 day –1)
Sunfleck PPFD (mol m –2 day –1)
% Of above canopy PPFD
% Sunfleck PPFD (day –1)
Number of sunflecks (day –1)
32
2.80 (0.37)
2.21 (0.37)
5.11 (0.68)
68.3 (3.9)
79.8 (4.6)
24
3.25 (0.09)
2.67 (0.09)
5.92 (0.17)
75.8 (0.7)
74.3 (1.1)
8
1.41 (0.24)
0.77 (0.21)
2.60 (0.44)
45.9 (7.8)
103.5 (13.5)
11
1.59 (0.09)
1.10 (0.13)
2.90 (0.16)
67.3 (5.4)
53.64 (8.9)
5
1.56 (0.16)
0.86 (0.18)
2.88 (0.29)
52.1 (7.6)
114.6 (13.8)
Partly cloudy day
Number of photodiode sites
Daily PPFD (mol m –2 day –1)
Sunfleck PPFD (mol m –2 day –1)
% Of above canopy PPFD
% Sunfleck PPFD (day –1)
Number of sunflecks (day –1)
24
1.69 (0.19)
1.00 (0.19)
3.82 (0.44)
48.7 (4.5)
151 (22)
16
1.78 (0.26)
1.12 (0.26)
4.03 (0.58)
52.8 (5.5)
112 (21)
8
1.52 (0.28)
0.76 (0.26)
3.39 (0.63)
40.6 (7.6)
72 (14)
10
1.10 (0.09)
0.46 (0.09)
2.52 (0.20)
40.2 (5.1)
228 (41)
5
1.54 (0.12)
0.69 (0.11)
3.45 (0.26)
43.9 (5.1)
284 (41)
model ratios were similar (solid versus open bars, Figure 3).
The other species, which had either slow induction gain or
rapid induction loss and thus were expected to have greater
limitations to sunfleck photosynthesis, had significantly
x min
5 -s
higher slopes of A day
− s versus A day − d in the regression analyses,
and therefore had larger errors as a result of ignoring dynamic
photosynthetic responses than Liriodendron. These findings
suggest that the use of a simple light response curve in conjunction with a long PPFD averaging interval can result in substantial overestimates of daily photosynthesis if the species have
appreciable limitations to sunfleck photosynthesis. Such limitations can be readily evaluated by artificial light-fleck experiments (e.g., Chazdon and Pearcy 1986, Poorter and Oberbauer
1993, Valladares et al. 1997, Naumburg and Ellsworth 2000).
Figure 2. Number of sunflecks for
loblolly pine and hardwood forests by
sunfleck duration (a, c) and mean
PPFD classes (b, d) for a sunny day (a,
b) and a partly cloudy day (c, d) in
early June. Data were classified in the
shade categories based on the sunny
day data only. Bars represent means
+ 1 SEM with the following sample
sizes (photodiode locations): n = 13 on
the sunny day and n = 6 on the partly
cloudy day for pine sites with moderate shade (> 2 mol m –2 day –1), n = 11
on the sunny day and n = 10 on the
partly cloudy day for pine sites with
dense shade (1–2 mol m –2 day –1), and
n = 5 on both sunny and partly cloudy
days for hardwood sites with dense
shade (1–2 mol m –2 day –1).
TREE PHYSIOLOGY ONLINE at http://heronpublishing.com
398
NAUMBURG AND ELLSWORTH
day data, overestimates were lower (17% for Acer and Liquidambar), but showed similar trends for the different species.
The estimates for Acer and Liquidambar fell within the range
of the estimated 30% over-prediction for two understory
plants for which similar scenarios have previously been evaluated (Pfitsch and Pearcy 1989, Pearcy et al. 1994). In contrast,
for species with asymmetrical induction behavior like Liriodendron (e.g., rapid induction gain and slow induction loss;
Table 1), comparatively little error would be introduced on either sunny or partly cloudy days by ignoring dynamic responses. These data demonstrate that species comparisons of
Aday inferred from light response curves can be misleading if
the species differ in dynamic light behavior. That is, even
though two species may have similar Aday estimates based on
light response curve data, real Aday may differ between the species because of species-specific differences in induction dynamics. For partly cloudy days, the relative magnitude of
overestimation was lower, but the effect of species-specific induction dynamics persisted.
Figure 3. Relative overestimation of Aday as a function of the PPFD
averaging time interval and ignoring dynamic induction responses for
the four species. Solid bars are ratios expressed relative to the 5-s dyx min
5 -s
namic model results (A day
− s / A day − d ) and open bars are relative to the
x min
5 -s
5-s steady-state model results (A day
− s / Aday − s ). Bars represent the
mean ratios of 20 daily PPFD courses + 1 SEM. Where bars are not
visible, the ratios did not exceed unity. All ratios that were significantly greater than one are indicated by an asterisk (1-tailed, 1-sample
t-test, t1,19 < 1.729, P < 0.05).
Further, in species with moderate to strong induction limitations to sunfleck photosynthesis, the magnitude of overestimation as a result of using a steady-state model rather than a
dynamic model can be high even at short averaging intervals.
For example, comparison of the 5-s model steady-state and
dynamic model results yielded ratios that were greater than 1.0
for all species but Liriodendron (Figure 3). The highest mean
ratios were 1.44 for Acer and Liquidambar on the sunny day,
indicating that, for species with comparatively rapid or very
slow induction responses (Table 1), understory Aday could be
overestimated by an average of 44% as a result of ignoring dynamic responses alone. This is because, with infrequent sunflecks, Acer would lose induction in the intervening shade periods, whereas Liquidambar would never reach high induction
because of its slow induction gain. Based on the partly cloudy
Forest type and sunfleck distribution In addition to the
PPFD averaging interval and species-specific photosynthetic
characteristics, the amount and distribution of sunflecks
throughout the day also impacted the degree to which the
steady-state models over-predicted Aday. Although mean absolute overestimates ranged between 4.2 and 5.7 mmol m –2 day –1
for all pine shade environments during the sunny day and for
all species but Liriodendron (Figures 4a and 4b), this translated
into different relative values because of the much lower Aday in
dense shade. Overall, for dense shade sites, Aday ratios were
generally larger and more variable than for moderate shade
sites (Figure 4c). This was especially true for model runs based
on longer light averaging intervals (data not shown). Again,
Acer and Liquidambar showed the highest and most variable
Aday overestimates of the four species (Figure 4c). Relative
overestimates were lower during the partly cloudy day than
during the sunny day (Figure 4d), but species and forest type
trends were similar. Although overall Aday values were lower
under dense pine shade, overestimates were still 29 and 35%
for Acer and Liquidambar, respectively. This suggests that,
with increasing daily PPFD resulting from both longer and
higher sunfleck PPFD, the relative overestimation of Aday becomes smaller and less variable. This result is similar to the
small discrepancies between dynamic and steady-state model
predictions found by Pearcy et al. (1994) in their modeling exercises with sunny day data for Alocasia growing in a rainforest and model runs based on gap PPFD data, which are
characterized by long sunflecks.
In contrast to this trend of small over-predictions of Aday under moderate PPFD conditions, sunny day model predictions
based on dense shade hardwood data also resulted in significantly smaller (ANOVA, F1,56 = 14, P < 0.001) and less variable ratios of steady-state to dynamic Aday than predictions
based on dense shade pine data (Figure 4c). This indicates that
the amount of sunfleck PPFD alone was not the only determinant of photosynthetic overestimation. Therefore, we compared daily light courses that had essentially the same total
TREE PHYSIOLOGY VOLUME 22, 2002
PHOTOSYNTHETIC DYNAMICS AND DAILY UNDERSTORY ASSIMILATION
399
Figure 4. Model predictions (a, b) and
relative overestimation (c, d) of Aday
given by the ratio of 1-min steady-state
versus the 5-s dynamic model predictions for the four species on a sunny
(a–c) and a partly cloudy day (d).
Sample sizes and shade classifications
are as in Figure 2. Bars represent mean
ratios + 1 SEM and, where not apparent, did not exceed 1.
PPFD but resulted in different overestimates of daily photosynthesis values (Table 3). These time courses differed both in
the relative amount of sunfleck PPFD (Table 3) and the distribution of sunfleck PPFD throughout the day (Figure 5). The
hardwood PPFD course had a greater number of sunflecks but
a smaller contribution of sunfleck PPFD to the daily total compared with the two pine PPFD courses. The hardwood forest
data also showed the least overestimation of Aday (Table 3). Although the two pine forest examples had similar daily PPFD
characteristics (Table 3), the distribution of the sunflecks differed during the day (Figures 5a and 5d). The pine PPFD
course with the greatest overestimation of Aday had sunflecks
distributed fairly evenly throughout the day (Figure 5d). As a
consequence, averaging the PPFD data resulted in many light
values that were higher than the diffuse shade background
(Figure 5f). In contrast, the second pine PPFD course had sun-
flecks occurring in four clusters (Figure 5a). With this clustered sunfleck distribution, photosynthetic enzymes should
activate and stomata should open with the first sunfleck of the
sequence (Pearcy et al. 1994), thus decreasing the effect of induction dynamics and decreasing the overestimation of Aday
compared with the more uniform sunfleck distribution. The
grouping of sunflecks also resulted in fewer light values that
were higher than the background shade PPFD (Figure 5c) and,
as a consequence, longer PPFD averages resulted in smaller
overestimates of Aday compared with the more uniform sunfleck distribution.
Conclusions
The use of long averaging intervals for PPFD data or ignoring
photosynthetic and stomatal dynamics under variable light
Table 3. Characteristics of three diurnal PPFD courses with similar daily PPFD and overestimates of daily photosynthesis (Aday) for Acer and
Liriodendron. Two of the PPFD courses were collected in the pine forest and one was collected in the hardwood forest during a sunny day in early
June. Although the two pine diurnal PPFD courses were similar with respect to three daily PPFD characteristics, the first had a clumped distribution of sunflecks, whereas the second had a more even sunfleck distribution throughout the day (see Figure 5).
Acer
Liriodendron
Hardwood
Pine 1
Pine 2
Hardwood
Pine 1
Pine 2
PPFD characteristics
Daily PPFD (mol m –2 day –1)
% Sunfleck PPFD (day –1)
Number of sunflecks (day –1)
1.52
48.0
105
1.47
74.3
45
1.53
76.8
57
1.52
48.0
105
1.47
74.3
45
1.53
76.8
57
Modeled Aday ratio1
5s
5s
A day
− s / Aday − d
5 min
5s
A day − s / Aday
−d
5 min
5s
A day − s / Aday
−s
1.25
1.42
1.14
1.50
1.70
1.13
2.54
3.51
1.37
1.06
1.16
1.10
1.07
1.29
1.21
1.33
1.85
1.39
1
Ratios expressed relative to the 5-s dynamic (1st and 2nd ratio) and 5-s steady-state (3rd ratio) model results.
TREE PHYSIOLOGY ONLINE at http://heronpublishing.com
400
NAUMBURG AND ELLSWORTH
Figure 5. Daily PPFD courses from Table 3 averaged at 5-s, 5-min and 30-min
time intervals. The courses are similar
in daily PPFD but differ in their sunfleck characteristics and Aday predictions (Table 3).
conditions can both lead to potentially large errors in estimates
of Aday. The greatest overestimates occur in forests where total
PPFD and sunfleck frequency are low but sunflecks are a major component of total PPFD received. With respect to species,
the greatest errors in estimating Aday are expected for species
with either very rapid or very slow biochemical and stomatal
responses to PPFD, behaviors that have been observed in other
studies (cf. Table 2 in Naumburg and Ellsworth 2000). Although overestimates were smaller under partly cloudy conditions than under sunny conditions, most studies use sunny day
data for modeling (e.g., Ellsworth and Reich 1992, Zipperlen
and Press 1996, Herrick and Thomas 1999, Beaudet et al.
2000) and thus risk potentially large errors in Aday estimates. In
understory environments where photosynthesis is low, small
absolute differences in carbon uptake can result in large relative differences in species performance, so overestimates in
photosynthesis may inflate the carbon uptake of a species,
thereby invalidating comparative interpretation of its performance during forest succession. We conclude that efforts to
determine the performance of species and the mechanisms underlying tree regeneration and successful establishment in
dense understories (< 2% full sun) will be greatly enhanced by
measuring high-resolution PPFD data, and considering species-specific differences in dynamic behavior to sunflecks.
Acknowledgments
This research is part of the Forest–Atmosphere Carbon Transfer and
Storage (FACTS-1) project at Duke Forest, and was supported
through a subcontract from Brookhaven National Laboratory. We
thank G. Hendrey for enabling this work to be carried out at
FACTS-1. We are grateful to G. Katul for help in coding the dynamic
model, and numerous reviewers whose comments significantly improved the manuscript. The FACTS-1 project is supported by the U.S.
Dept. of Energy, Office of Health and Environmental Research, under
DOE contract DE-FG05-95ER62083 at Duke University and DEAC02-98CH10886 at Brookhaven National Laboratory.
References
Abrams, M.D. and S.A. Mostoller. 1995. Gas exchange, leaf structure
and nitrogen in contrasting successional tree species growing in
open and understory sites during a drought. Tree Physiol. 15:
361–370.
Barker, M.G., M.C. Press and N.D. Brown. 1997. Photosynthetic
characteristics of dipterocarp seedlings in three tropical rainforest
light environments: a basis for niche partitioning. Oecologia 112:
453–463.
Beaudet M., C. Messier, D.W. Hilbert, E. Lo, Z.M. Wang and M.J.
Lechowicz. 2000. Leaf- and plant-level carbon gain in yellow
birch, sugar maple, and beech seedlings from contrasting forest
light environments. Can. J. For. Res. 30:390–404.
Björkman, O., M.M. Ludlow and P.A. Morrow. 1972. Photosynthetic
performance of two rainforest species in their native habitat and
analysis of their gas exchange. Carnegie Inst. Wash. Year Book
71:94–102.
Canham, C.D., A.C. Finzi, S.W. Pacala and D.H. Burbank. 1994.
Causes and consequences of resource heterogeneity in forests:
interspecific variation in light transmission by canopy trees. Can. J.
For. Res. 24:337–349.
Chazdon, R.L. 1986. Light variation and carbon gain in rain forest
understorey palms. J. Ecol. 74:995–1012.
Chazdon, R.L. 1988. Sunflecks and their importance to forest understorey plants. Adv. Ecol. Res. 18:1–63.
Chazdon, R.L. and R.W. Pearcy. 1986. Photosynthetic responses to
light variation in rainforest species. I. Induction under constant and
fluctuating light conditions. Oecologia 69:517–523.
Chen, H.Y.H. and K. Klinka. 1997. Light availability and photosynthesis of Pseudotsuga menziesii seedlings grown in the open
and in the forest understory. Tree Physiol. 17:23–29.
TREE PHYSIOLOGY VOLUME 22, 2002
PHOTOSYNTHETIC DYNAMICS AND DAILY UNDERSTORY ASSIMILATION
DeLucia, E. and R. Thomas. 2000. Photosynthetic responses to CO2
enrichment of four hardwood species in a forest understory. Oecologia 122:11–19.
Ellsworth, D.S. and P.B. Reich. 1992. Leaf mass per area, nitrogen
content and photosynthetic carbon gain in Acer saccharum seedlings in contrasting forest light environments. Funct. Ecol. 6:
423–435.
Farquhar, G.D. and S. von Caemmerer. 1982. Modelling of photosynthetic response to environmental conditions. In Water Relations
and Carbon Assimilation. Physiological Plant Ecology II. Eds.
O.L. Lange, P.S. Nobel, C.B. Osmond and H. Ziegler.
Springer-Verlag, Berlin, pp 549–587.
Gildner, B.S. and D.W. Larson. 1992. Photosynthetic response to
sunflecks in the desiccation-tolerant fern Polypodium virginianum.
Oecologia 89:390–396.
Gill, D.S., J.S. Amthor and F.H. Bormann. 1998. Leaf phenology,
photosynthesis, and the persistence of saplings and shrubs in a mature northern hardwood forest. Tree Physiol. 18:281–289.
Gross, L.J., M.U.F. Kirschbaum and R.W. Pearcy. 1991. A dynamic
model of photosynthesis in varying light taking account of stomatal
conductance, C3-cycle intermediates, photorespiration and Rubisco activation. Plant Cell Environ. 14:881–893.
Herrick, J.D. and R.B. Thomas. 1999. Effects of CO2 enrichment on
the photosynthetic light response of sun and shade leaves of canopy sweetgum trees (Liquidambar styraciflua) in a forest ecosystem. Tree Physiol. 19:779–786.
Kirschbaum, M.U.F., M. Küppers, H.G.C. Schneider and S. Noe.
1998. Modelling photosynthesis in fluctuating light with inclusion
of stomatal conductance, biochemical activation and pools of key
photosynthetic intermediates. Planta 204:16–26.
Körner, C. 1991. Some often overlooked plant characteristics as determinants of plant growth: a reconsideration. Funct. Ecol. 5:
162–173.
Naumburg, E. and D.S. Ellsworth. 2000. Photosynthetic sunfleck utilization potential of understory saplings growing under elevated
CO2. Oecologia 122:163–174.
Naumburg, E., D.S. Ellsworth and G.G. Katul. 2001. Modeling dynamic understory photosynthesis of contrasting species in ambient
and elevated CO2. Oecologia 126:487–499.
Norman, J.M. 1993. Scaling processes between leaf and canopy levels. In Scaling Physiological Processes: Leaf to Globe. Eds. J.R.
Ehleringer and C.B. Field. Academic Press, San Diego, pp 41–76.
Pearcy, R.W. 1987. Photosynthetic gas exchange responses of Australian tropical forest trees in canopy, gap and understory microenvironments. Funct. Ecol. 1:169–178.
401
Pearcy, R.W. and H.W. Calkin. 1983. Carbon dioxide exchange of C3
and C4 tree species in the understory of a Hawaiian forest. Oecologia 58:26–32.
Pearcy, R.W., R.L. Chazdon, L.J. Gross and K.A. Mott. 1994. Photosynthetic utilization of sunflecks: a temporally patchy resource on
a time scale of seconds to minutes. In Exploitation of Environmental Heterogeneity by Plants. Eds. M.M. Caldwell and R.W. Pearcy.
Academic Press, San Diego, pp 175–208.
Pearcy, R.W., L.J. Gross and D. He. 1997. An improved dynamic
model of photosynthesis for estimation of carbon gain in sunfleck
light regimes. Plant Cell Environ. 20:411–424.
Pfitsch, W.A. and R.W. Pearcy. 1989. Daily carbon gain by Adenocaulon bicolor, a redwood forest understory herb, in relation to its
light environment. Oecologia 80:465–470.
Pontailler, J.Y. 1990. A cheap quantum sensor using a gallium-arsenide photodiode. Funct. Ecol. 4:591–596.
Poorter, L. and S.F. Oberbauer. 1993. Photosynthetic induction responses of two rainforest tree species in relation to light environment. Oecologia 96:193–199.
Roden, J.S. and R.W. Pearcy. 1993. Effect of leaf flutter on the light
environment of poplars. Oecologia 93:201–207.
Stickan, W. and X. Zhang. 1992. Seasonal changes in CO2 and H2O
gas exchange of young European beech (Fagus sylvatica L.). Trees
6:96–102.
Tang, Y., N. Kachi, A. Furukawa and M.B. Awang. 1999. Heterogeneity of light availability and its effects on simulated carbon gain of
tree leaves in a small gap and the understory in a tropical rain forest. Biotropica 31:268–278.
Valladares, F., M.T. Allen and R.W. Pearcy. 1997. Photosynthetic response to dynamic light under field conditions in six tropical rainforest shrubs occurring along a light gradient. Oecologia 111:
505–514.
Walters, M.B. and P.B. Reich. 1999. Low-light carbon balance and
shade tolerance in the seedlings of woody plants: do winter deciduous and broad-leaved evergreen species differ? New Phytol. 143:
143–154.
Washitani, I. and Y. Tang. 1991. Microsite variation in light availability and seedling growth of Quercus serrata in a temperate pine forest. Ecol. Res. 6:305–316.
Weber, J.A., T.W. Jurik, J.D. Tenhunen and D.M. Gates. 1985. Analysis of gas exchange in seedlings of Acer saccharum: integration of
field and laboratory studies. Oecologia 65:338–347.
Zipperlen, S.W. and M.C. Press. 1996. Photosynthesis in relation to
growth and seedling ecology of two dipterocarp rain forest tree
species. J. Ecol. 84:863–876.
TREE PHYSIOLOGY ONLINE at http://heronpublishing.com