S32-009
Sensitivity of photosynthesis and carbon sink in tropical rainforests to projected
atmospheric CO2 and climate change
G Lin1,, J Berry2 , J Kaduk 2 , A Southern1 , J van Haren1 , B Farnsworth1 , J. Adams1 , K Griffin3 ,
W Broecker3
1
Columbia University/Biosphere 2 Center, Oracle, AZ 85623, USA(Fax: 5208965160; email:
glin@bio2.columbia.edu); 2 Dept. of Plant Biology, Carnegie Institution of Washington, Stanford, CA 94305,
USA. 3 Dept. of Earth and Environ. Sciences, Columbia University, Palisades, NY 10964, USA.
Keywords: photosynthesis, carbon sink, CO2 , climate change, rainforest
Introduction
Terrestrial ecosystems currently take up as much as one third of anthropogenic CO2 emitted
annually to the earth’s atmosphere (Schimel 1995). CO2 fertilization (i.e. the stimulation in
photosynthesis and plant growth by elevated atmospheric CO2 concentration), forest re-growth,
nitrogen deposition and climate change are believed to contribute to this C sink (Field 2001).
CO2 fertilization is of particular interest as it could provide a negative feedback on the growth
rate of the atmospheric CO2 concentration. Recent field measurements and model simulations
indicate that a significant fraction of the terrestrial sink (approximately 1 Gt C yr-1 ) may be
attributed to CO2 fertilization occurring in tropical rainforests (Malhi and Grace 1998).
However it is crucial to understand the dynamic properties of this C sink before assuming that
CO2 fertilization in tropical rainforests will play a significant role in offsetting future
anthropogenic CO2 emissions. Previous evaluations of C sink dynamics have been limited
mostly to model simulations (Tian et al. 1999; Kicklighter et al. 1999, Chamber et al. 2001),
and have not been linked with experiments at appropriate spatial scales. Here we present
results from an empirical evaluation of photosynthesis and carbon sink sensitivity to projected
CO2 and climate change, based on measurements in a large-scale rainforest mesocosm.
Materials and Methods
Description of the rainforest mesocosm. Our study takes advantage of the technical innovations at Biosphere 2, a
1.27 ha enclosed structure near Tucson, Arizona. With a ground area of 1,900 m2 , an air volume of 35,000 m3 and
a soil volume of 6,000 m3 , the rainforest mesocosm we used in this study contains more than 400 individual plants
belonging to ∼120 species as well as diverse groups of soil micro-organisms. This rainforest models after a humid
tropical rainforest in South America and has been separated from the rest of Biosphere 2 using a partitioning
curtain since January 1999. During the study period, the daily mean air temperature was set at 27 o C (Max. 35 o C,
Min. 20 o C), with relative humidity (RH) at about 85 %.
CO2 control and treatments. We controlled the CO2 concentration in the rainforest mesocosm’s atmosphere at
each of four target levels (400, 700, 1000 and 1200 ppmv) for 4 days by either injecting CO2 or using pull/push
fans. Two experimental periods, Sept. 9-Oct. 10 of 1999 and May 1-June 5 of 2000, were selected for stable light
levels. During Sept. 9-Oct. 10 of 1999, CO2 concentration was first increased and then decreased step-wise from
400 to 1200 ppmv. In contrast, CO2 was increased step-wise twice during May 1-June 5 of 2000.
Calculations of net ecosystem exchange (NEE). We calculated NEE over 15 min intervals for each day based on
the changes in CO2 concentration, the amount of CO2 injected, the CO2 exchange amounts by the pull/push fans,
and the CO2 exchange due to leakage through the curtains. NEE was then deconvolved into canopy
photosynthesis (Acanopy) and total respiration (Recosystem).
Leaf-level measurements. Leaf-level measurements of photosynthesis under different light and CO2
concentrations were measured on four canopy species (Cecropia schreberiana, Ceiba pentandra , Arenga pinnata,
Clitoria racemosa) and two understory species (Costus sp., Coffea arabica) using a LI-6400 photosynthesis
system (LICOR, Inc.). The temperature inside the chamber was 30.0±0.5 o C for the canopy species and 25.0±0.5
o
C for the understory species, while the RH was set at 85 % for all species. The light level was 1500 µmol m-2 s-1
for the canopy species and 100 µmol m-2 s-1 for the understory plants.
Response coefficient of photosynthesis. We related our experimental results to modeling approaches using the
response coefficient (Rc) of photosynthesis to Ca according to Woodrow et al. (1990):
Rc = (dP/dCa )*(Ca /P)= (dP/P)/(dCa /Ca ).
Rc is a dimensionless parameter that gives the fractional change in photosynthesis (either measured or modeled) to
a fractional change in Ca , and P is photosynthetic rate at specific scale. Here, we compare Rc values calculated for
commonly used approaches for modeling the response of GPP to Ca (the linear and logarithmic β-factor,
Kicklighter et al. 1999) and a theoretically derived Rc for a leaf photosynthesis model based on enzyme kinetics
(Farquhar et al. 1980) with our observed leaf and canopy level responses.
Description of carbon sink model. To simply illustrate the dynamics of the interaction of GPP and respiration
and the resulting sink, we constructed a model with one pool carbon (M). Respiration, R, depends linearly on M,
R=k*M, where k is a first order rate constant. This rate constant is the inverse of the mean residence time of C in
M, or turnover time, τ (yr). The time evolution of the C mass, M(t) (Gt C), is then given by the following
differential equation:
dM(t)/dt = – M(t)/ τ + GPP(t)
where t denotes time in years and GPP(t) (Gt C yr-1 ) is gross primary productivity at time t. The net C sink, S (Gt
C yr-1 ), during one year can then be approximated by:
St = GPPt –Rt = GPPt - M t –1 / τ
where St is the C sink in the year [t-1, t), Mt –1 is the biomass at time t-1. We evaluated the dynamics of the CO2
fertilization carbon sink in global rainforests for two scenarios of change in Ca. Both scenarios are based on
observed Ca from 1860 to 1990; after 1990, Ca was assumed to increase linearly with time at 1.5 ppmv yr-1
(L1200), or to increase more rapidly at first but eventually stabilize at 750 ppmv (S750). Further we assumed
either no temperature increase associated with a Ca increase or a 2.5 o C increase with a doubling of Ca from 350 to
700 ppmv (ca. 0.007 o C ppmv -1 ). To calibrate the turnover time and the initial biomass we fitted the model to the
biomass in 1985 reported by Amthor et al. (1995).
Results and Discussion
Photosynthetic responses at different scales. Responses of photosynthetic rates to light
and CO2 at the whole mesocosm scale were similar to that observed at smaller scales in
experiments with leaves (Fig. 1). Photosynthetic CO2 uptake is proportional to light for a
PPFD lower than 300 µmol m-2 s-1 at both scales. Although the canopy response is the daily
sum but the leaf level responses are instantaneous responses, all curves show similar
saturation responses to increasing Ca .
Figure 1. Responses of photosynthetic
CO2 uptake by leaves (Aleaf, a-b) and
by the whole mesocosm (NEE, c-d) to
increasing photosynthetic proton flux
density (PPFD, left panels) and
atmospheric CO2 concentration (right
panels) within the tropical rainforest
mesocosm of Biosphere 2. NEE values
in Fig. 1c were 15-min means while
those in Fig. 1d were the daily sum
over the photo period. Relatively
higher daily sum NEE values observed
in 2000 than in 1999 (Fig. 1d) were
likely the results of much higher daily
integrated light.
Response coefficient of photosynthesis. Our measurements suggest that the sensitivity of
photosynthesis decreases with increasing CO2 concentration and biological organization level
(data not shown). The linear β-factor does not represent the saturation of this response with
increasing Ca (decreasing values of Rc). The logarithmic β-factor represents this saturation
response better, and while the modeled sensitivity declines with Ca , it cannot be parameterized
to match the observed response over the range of the observations used in this study. The Rc
values of the leaf and enzyme responses are nearly identical below about 600 ppmv CO2
indicating that the enzyme step is in full control of CO2 flux in this region. The canopy
response shows a lower sensitivity to CO2 than the leaf or the modeled biochemical
formulations for all Ca, and this is interpreted as indicating that other processes (probably light
availability at different points in the canopy) are co-limiting the rate of CO2 fixation. These
differences in response with the scale of observations provide information on subtle changes in
the controlling processes that should be addressed in models that scale from the enzyme to
ecosystem.
Dynamics of C sink in tropical rainforests. Our model simulations indicate that the response
coefficient of photosynthesis to Ca (Rc), the future trajectory of Ca and the associated
temperature change all have profound effects on the temporal dynamics of the C sink in global
rainforests (Fig. 2).
Figure 2. The simulated temporal
change in the strength of CO2
fertilization carbon sink in global
rainforests from 1860 to 2300 based on
the canopy-level response function and
two type β-factor functions assuming a
linear CO2 increase of 1.5 ppmv yr -1
until 1200 ppmv (L1200 scenario)(a)
and the simulated C sink based on the
canopy response assuming either a
stabilizing CO2 at 750 ppmv (S750
scenario) or a linear CO2 increase of
1.5 ppmv yr -1 until 1200 ppmv (L1200
scenario) under either no temperature
change or a 2.5 o C increase with a
doubling CO2 (b).
These trajectories can be understood by noting that approximately St ~ Rc *GPP *(dCa /Ca )* τ.
From 1860 to 1960 the modeled CO2 sink is quite small -- because the rate of increase in Ca
was slow. The strong increase in C sink since 1960 resulted from the significant growth rate of
atmospheric CO2 from the 60's to the present. Thus, the present sink may be a comparatively
recent phenomenon. Future behavior of this sink depends strongly on the rate of growth in Ca
and the consequent change in Rc. Imposing a linear rate of growth beginning in 1990 causes the
sink to peak and begin to decline slowly as Rc declines with increasing Ca. This decline is more
pronounced when the canopy response rather than a β-factor is used, reaching about 50% of its
current value by 2300 (Fig. 2a). Assuming that growth of Ca carries an associated temperature
change results in a lower present estimate of the sink and a more rapid decline, reaching 50%
of its current value before the end of this century (Fig. 2b). When CO2 is assumed to stabilize
at 750 ppmv, the sink peaks then declines to zero as Ca reaches a stable value. The areas under
the curves in Fig. 2 could be used to estimate total amount of C stored in global rainforests,
which translates directly to anthropogenic CO2 that could be released before Ca reaches the
ceiling of 750 ppmv. C storage using the linear β-formulation is larger (227 Gt C) than that
simulated assuming the observed canopy response (146 Gt C) and much larger than that
simulated assuming both CO2 saturation and climate change (48 Gt C). The latter corresponds
to less than 6 years of fossil fuel use at current rates or to a 27 ppmv lower final value of Ca .
Although our model is highly simplified, its simple structure is useful for exploring the effects of
different assumptions on the CO2 response of photosynthesis and the consequence of global
warming on carbon sink dynamics. Further, while our short-term measurements of canopy
response to CO2 may not capture important second order effects such as photosynthetic
acclimation (Sage et al. 1989) or chronic nutrient stress (Oren et al. 2001), the values for GPP,
biomass, and the carbon sink derived from our simulations are within accepted ranges. Therefore,
we suggest that our estimates of the future carbon sink are plausible and conservative. Based on
our analyses, we anticipate that the current C sink strength in tropical rainforests will be likely to
increase over next few decades, reach a peak, start to decrease and eventually vanish (Fig. 2). If
global warming occurs as predicted, the sink strength will be reduced, and the year when the sink
strength starts to decrease will be reached much earlier.
In conclusion, our study illustrates the key importance of a correctly representing ecosystem
carbon dynamics and respiration (our parameter τ) in carbon cycle models (Cox et al. 2000). Of
the current carbon sink, direct stimulation of GPP in that year accounts for only about 1/10th of
the total, while the remainder is related to stimulation of GPP in previous years and to delays in
the flow of respired carbon through ecosystems. To date, most experimental work has focused
on the supply side (responses of GPP), while the slower, but equally important dynamics of
carbon cycling within ecosystems and their response to temperature change have received little
attention. We propose that large-scale mesocosms inside Biosphere 2 with mass balance
capabilities are ideally suited for conducting such studies of ecosystem carbon dynamics.
Acknowledgements
We thank all technical staff at Biosphere 2 Center for their assistance in implementation of this study. We also
thank Barry Osmond for his constructive comments on the manuscript. The financial support for this study was
provided by Mr. Edward Bass through a contract to Columbia University.
References
Amthor JS et al. (1995). Global Change Biology 1: 243-278.
Chambers JQ et al. (2001). Nature 410, 429.
Cox PM et al. (2000). Nature 408, 184-187.
Farquhar GD et al. (1980). Planta, 149, 78-90.
Field CD (2001). Plant Physiol. 125, 25-28.
Grace J et al. (1995). Science 270, 778-780.
Kicklighter DW et al. (1999). Tellus B, 51, 343-366.
Malhi Y, Grace J (1998). TREE 15, 332-337.
Oren R et al. (2001). Nature 411, 469-472.
Sage RF et al. (1889). Plant Physiol. 89, 590-596.
Schimel DS (1995). Global Change biology 1, 77-91.
Tian H et al. (1999). Nature 396, 664-667.
Woodrow IE et al. (1990). Plant Cell and Environ. 13, 339-347.