Reviews of Geophysics

Prediction of atmospheric δ13CO2 using fossil plant tissues

Authors


Abstract

[1] Reconstruction of the carbon isotope composition of atmospheric CO2 is critical to the understanding of long-term global carbon cycling. We have suggested that the δ13C value of land plant carbon (δ13Cp) preserved in the geologic record should reflect the δ13CO2 at the time during which the plants grew (δ13Ca), based on a meta-analysis of modern plant data. Here we present the results of laboratory experiments designed to quantify the relationship between plant tissue δ13C and δ13CO2 values under varying environmental conditions, including differential pCO2 ranging from 1 to 3 times today's levels. As predicted, plants grown under elevated pCO2 showed increased average biomass compared to controls grown at the same temperature. Across a very large range in δ13Ca (≈24‰) and pCO2 (≈740 ppmv) we observed a consistent correlation between δ13Ca and δ13Cp (p < 0.001). We show an average isotopic depletion of −25.4‰ for aboveground tissue and −23.2‰ for belowground tissue of Raphanussativus L. relative to the composition of the atmosphere under which it formed. For aboveground and belowground tissue, grown at both ∼23°C and ∼29°C, correlation was strong and significant (r2 ≥ 0.98 and p < 0.001); variation in pCO2 level had little or no effect on this relationship. These results validate our initial conclusion that in the absence of environmental stress, plant δ13C primarily reflects atmospheric δ13CO2 linearly across pCO2 levels; the demonstrated excellent correlation in δ13Ca and δ13Cp suggests a high level of predictive power across varying environmental conditions.

1. INTRODUCTION

[2] Modeling the global carbon cycle and associated paleoclimate reconstruction in the geologic past requires an understanding of the size and composition of active carbon reservoirs and the flux among them. Of particular interest are those reservoirs, such as the atmosphere, that vary significantly in both size (pCO2) and composition (δ13CO2) through time [Berner, 1998; Strauss and Peters-Kottig, 2003; Sundquist, 1985]. These reservoirs most clearly reflect the dynamic nature of the carbon cycle, and these carbon pools may be important agents of global climate change. The atmospheric carbon reservoir is also of special interest because it links the ocean and biosphere reservoirs over short timescales [Sundquist, 1985]. Although this short-term variation may be irrelevant to long-term carbon cycling, the ocean, atmosphere, and biosphere are central players in times of catastrophic perturbation of carbon cycling, such as dramatic climate change and mass extinction [Arens and Jahren, 2000; Bralower et al., 1994; Dickens et al., 1997]. The size of the atmospheric carbon reservoir may also influence flux between reservoirs over longer geologic timescales through changes in weathering rates [Berner, 1993; Yapp and Poths, 1996].

[3] Flux changes among reservoirs can be linked to carbon sources and sinks by their isotopic composition [e.g., Dickens et al., 1997; Weissert, 1989]. This is especially true when normal carbon cycling is interrupted [e.g., D'Hondt et al., 1998]. Therefore a quantitative estimate of the isotopic composition of atmospheric CO2 would greatly facilitate efforts to describe and model carbon cycling through geologic time and the paleoclimatic consequences of carbon cycle change.

[4] Because of its central importance a variety of methods have been developed to assess the size and carbon isotopic composition of the atmospheric carbon pool through geologic time. Some workers have chosen to address the size of the reservoir through modeling [Berner, 1994, 2001; Berner and Kothavala, 2001]; others have chosen proxy methods that reconstruct pCO2 through a correlation with the density of stomata on plant leaves observed in modern plants [Beerling and Royer, 2002; Chaloner and McElwain, 1997; Greenwood et al., 2003; Hesselbo et al., 2003; McElwain et al., 1999; Retallack, 2001; Royer et al., 2001a, 2001b] or the carbon isotopic composition of paleosol carbonates [Bowen and Beerling, 2004; Cerling and Hay, 1986]. Similarly, the isotopic composition of atmospheric CO2 has been assessed by measuring the carbon isotopic value of various proxies, most notably carbonate carbon and organic matter in marine systems [e.g., Hayes et al., 1989; Rask and Schoenau, 1993] and paleosols [e.g., Bowen and Beerling, 2004; Mora et al., 1996]. Because land plants sample the atmosphere directly during photosynthesis, we contend that organic carbon derived from plants provides another substrate from which information on the isotopic composition of the atmospheric carbon pool may be sampled [Arens et al., 2000].

[5] The carbon stable isotope value (δ13C [‰]) assigned to C3 plant tissue (the dominant photosynthetic pathway) during photosynthesis was first conceptualized by G. D. Farquhar and can be described as the following:

equation image

where δ13Ca and δ13Cp represent the carbon stable isotope composition of atmospheric CO2 and of the resulting plant tissue, respectively; a and b encompass the isotopic discrimination inherent to gaseous diffusion and enzymatic carboxylation, respectively; (ci/ca) is a unitless scalar that reflects the partial pressure of CO2 within the stomatal chamber (ci) relative to the partial pressure of CO2 in the external atmosphere (ca) [Farquhar et al., 1989]. The value a = 4.4‰ was calculated from diffusivity comparisons of 13CO2 versus 12CO2 in air [Craig, 1953, 1954] and is generally accepted. Depending upon the role envisioned for the primary enzyme that facilitates carboxylation, published estimates of b range from 27.0 to 29.0‰ [Christeller et al., 1976; Farquhar, 1979; Schmidt et al., 1978; Wong et al., 1979]. Published estimates of (ci/ca) range from 0.38 to 0.89 [Arens et al., 2000] and are meant to reflect a time-averaged value resulting from moment-to-moment stomatal regulation of gas exchange.

[6] Because plants regulate stomatal conductance in response to changing environmental conditions, classic studies have showed straightforwardly that δ13Cp value can be increased by subjecting a plant to water stress [e.g., Toft et al., 1989] or increased salinity [e.g., Guy et al., 1980], both of which force plants to close stomata and result in 13C-enriched tissues. Conversely, the δ13Cp value was shown to decrease with direct sunlight [e.g., Madhavan et al., 1991] and low environmental temperatures [e.g., Körner et al., 1991]. A wealth of other potential and actual environmental effects on δ13Cp has been discussed at length elsewhere [i.e., Dawson et al., 2002]. Our 2000 survey of the primary literature revealed a spread of no less than 10‰ in δ13Cp values all derived from approximately the same atmospheric δ13Ca; this variability reflected a large number of C3 species (n = 176) under an extremely wide range of growth conditions (including pCO2 level) [Arens et al., 2000]. One striking result of our initial analysis was the strong linear correlation between δ13Cp and δ13Ca values (r2 = 0.91 and 519 measurements on 176 species [Arens et al., 2000, Figure 2]), even across a broad range of environmental conditions and ecological settings. This led us to hypothesize that the δ13Cp value of fossil plant tissue could be used to reconstruct the δ13Cp value of the atmosphere, provided that the analytical substrate included the contribution of many species across various states of ecological functioning and could be meaningfully assessed for diagenetic alteration [Arens and Jahren, 2000, 2002; Arens et al., 2000; Jahren, 2002; Jahren et al., 2001, 2005]. This approach has been questioned at several levels ranging from the possibility of nonlinearity at more extreme values of δ13Cp and δ13Ca [Gröcke, 2002] to the possibility of variable (ci/ca) under changing environmental conditions at both ecological and evolutionary timescales [Beerling and Royer, 2002]. In order to address these concerns we designed a set of experiments to answer the question: Do plants grown under conditions of controlled and constant δ13Ca values and pCO2 levels show a significant, linear relationship between δ13Ca and δ13Cp values?

2. MATERIALS AND METHODS

[7] Two identical growth chambers were built from 1/4-inch Plexiglas (Figure 1b) (91.44 cm wide × 121.92 cm high × 45.72 cm deep). One served as a control with CO2 concentrations maintained at ambient atmospheric levels (∼370 ppmv = RCO2 = 1×, i.e., equivalent to present levels, here R indicates the carbon dioxide level expressed as a ratio to today's level), while the other chamber was supplied with air enriched with CO2. In the first experiment the enriched chamber was maintained at pCO2 of ∼780 ppmv = RCO2 = 2× (i.e., twice present atmospheric pCO2). During the second experiment the enriched chamber was maintained at ∼1090 ppmv = RCO2 = 3×, and the control chamber was maintained at ambient levels as above. Ambient air was delivered to the chambers through an intake pipe (Figure 1b) via an inline fan at a rate of 49 dm3 min−1. For the elevated chamber, CO2 levels were increased to ∼780 or ∼1090 ppmv by bleeding pure cylinder CO2 into the intake pipe, where it mixed with the ambient air before reaching the growth chamber. The CO2 flow rate can be precisely controlled with an inline needle valve (Figure 1b), thus enabling enrichment of air to any desired concentration of CO2. Air is exhausted through the upper pipe (Figure 1b) and vented to a fume hood. Complete air turnover in both the control and elevated chamber was approximately once every 10.4 min. Adjustable shelving provided support for the samples and grow lights, enabling adjustment of lamp height to maintain a constant distance from plant leaves throughout the growth period (Figure 1b). Four fluorescent grow lamps (33-W GE Brightstik) per shelf provided 600 lumens (approximately 100 microeinsteins m−2 s−1 photosynthetic photon flux density, the equivalent of full shade at midday [Halling and Peters, 1987]) of light intensity as measured at the leaf surface. The lighting cycle was controlled with an electrical timer; lights turned on in the morning at 0700 LT and turned off at 1845 LT. Heat generated from the grow lamps provided the daytime temperature difference between the upper and lower sample shelves. The lower shelf provided a “daytime” growing temperature of ∼23°C; the upper shelf provided a “daytime” growing temperature of ∼30°C. Temperature measurements were taken at each sample shelf (uncertainty = ±1°C). Relative humidity was measured at the lower sample shelf (uncertainty = ±1%), and CO2 concentrations were measured at the upper sample shelf (uncertainty = ±25 ppmv).

Figure 1.

(a) Raphanussativus L. after 5 weeks of growth within the elevated-CO2 growth chamber. (b) Plant growth chambers used in all experiments.

[8] Plants grown were Raphanus sativus L. (variant Cherry Belle Radishes, Brassicaceae) ordered from Harris Seeds of Rochester, New York (ID 0071, see www.harrisseeds.com) (Figure 1a). Raphanus was selected for the experiment because it grows rapidly and produces a large proportion of its biomass underground, allowing us to separate the significance of leaf versus other tissues. Eight plants were grown on each shelf, with a total of 16 plants per chamber per experiment. Water was delivered to the plants through tubing that entered the chambers from the top (Figure 1b), enabling us to minimize the opening of chambers throughout the growth experiment. Growth was started with soil at “field moist” conditions [Soil Survey Staff of the Soil Conservation Service, 1999]. Over the course of the experiment, 690 ± 1 mL of water were added to each plant; two of these water additions contained fertilizer (“Miracle-Gro” at commercially prescribed concentration). Fertilizer additions took place on day 9 and day 21 of the 37-d experiments. Addition of water sometimes resulted in an increase in relative humidity. The response was fleeting: It occurred over 5 h and was about +4% to +10% in magnitude. Experiments were designed to ensure that plant growth was not limited by water, light, or nutrient availability in order to highlight the effects of elevated pCO2 and minimize the effect of (ci/ca) on its carbon isotopic composition.

[9] Upon harvest, aboveground tissue was separated from belowground tissue, dried for 5 d at 100°C, and weighed (uncertainty ±1 mg). All aboveground and belowground tissues were separated from each plant, lyophilized for 24 h, and ground to a fine uniform powder in preparation for stable isotope analysis, thus yielding a bulk or combined tissue value. Plant tissues were analyzed for δ13C value using a Eurovector automated combustion system in conjunction with an Isoprime stable isotope mass spectrometer at the Johns Hopkins University. All samples were introduced to the combustion system in pure tin capsules; stable isotope values are reported in standard δ notation: δ = (RsampleRstandard/Rstandard)1000‰. The reporting standard is Pee Dee Formation limestone with R = 13C/12C = 0.011237. Analytical uncertainty associated with each measurement was ±0.05‰. Values presented represented the average of three replicate capsules; the average isotopic variability of replicate capsules was ±0.12‰ for all tissues. Combustion also resulted in a quantification of percent carbon content in each sample. Neither elevated pCO2 nor elevated growth chamber temperature affected the bulk carbon composition of the plants. Average composition of aboveground tissue was 34.1% C; average for belowground tissue was 35.5% C (all values ±1% C analytical uncertainty; variability was less than this uncertainty).

[10] We sampled CO2 in air (Tables 1 and 2) using a preevacuated 250-cm3 glass vessel. We placed the vessel in the environment that we wished to sample (e.g., the control or elevated growth chambers), opened the stopcock, and let it equilibrate with the surrounding air for 15 min. We then isolated the CO2 from the sampled atmosphere by slowly bleeding the air sample through a vacuum manifold, freezing any water into an ethanol/liquid nitrogen–cooled trap, freezing CO2 into a series of liquid nitrogen–cooled traps, and pumping away noncondensable gases. Because naturally occurring N2O will disrupt the isotopic analysis of CO2 (because of isobaric interference), we further purified the sample by reaction with 0.5 g of reduced copper at 450°C for 2 h. The resulting N2 was separated from CO2 cryogenically. Pure CO2 sample gas was analyzed for δ13CO2 value through the dual inlet of the mass spectrometer. All statistical analyses were performed in Aabel 1.5.8 (Gigawiz Ltd.) for Macintosh G5 running OSX.4.

Table 1. Average Stable Isotope Values of CO2 (δ13CO2) Presented With the Standard Deviation (σ) Seen in the Number of Days of Measurements (n)a
 δ13CO2, ‰σ, ‰n
  • a

    Cylinder measurements took place over one 24-h period.

  • b

    Stability of these values about a mean is presented in Figure 6.

Control chamber atmosphere (RCO2 = 1×)−13.860.549
Elevated chamber atmosphere (RCO2 = 2×)−24.050.399
Ambient laboratory atmosphere−14.450.449
CO2 cylinder−42.57b0.10b4
Table 2. Summary of the Environmental Characteristics of the Control and Elevated Chambers, Separated According to Upper and Lower Growth Shelf, as Measured Throughout the Preexperiment and the RCO2 = 2× and RCO2 = 3× Experimentsa
 Preexperiment, 8 WeeksRCO2 = 2×, 5 WeeksRCO2 = 3×, 7 Weeks
  • a

    Average values are presented with the standard deviation (±σ) seen in the number of days of measurements (n); these data are presented graphically in Figures 2, 3, and 4. T is temperature; RH is relative humidity.

Control chamber pCO2, ppmv387 ± 25 (54)389 ± 30 (36)348 ± 16 (53)
Control chamber δ13CO2, ‰−9.11 ± 0.08 (3)−9.00 ± 0.10 (3)−9.04 ± 0.16 (3)
   Upper T,°C29.2 ± 0.8 (42)29.0 ± 0.6 (30)28.9 ± 0.7 (51)
   Upper RH, %23.2 ± 3.3 (51)35.1 ± 7.9 (36)59.5 ± 4.0 (51)
   Lower T,°C21.8 ± 1.4 (54)22.5 ± 0.9 (36)22.9 ± 0.8 (53)
   Lower RH, %21.0 ± 2.8 (51)31.6 ± 8.3 (36)52.8 ± 3.8 (51)
Elevated chamber pCO2, ppmv764 ± 25 (54)778 ± 33 (35)1088 ± 30 (52)
Elevated chamber δ13CO2, ‰−27.52 ± 0.13 (3)−26.73 ± 0.04 (3)−32.61 ± 0.01 (3)
   Upper T,°C29.5 ± 0.8 (41)29.4 ± 0.7 (30)29.3 ± 0.7 (51)
   Upper RH, %19.4 ± 2.8 (51)27.7 ± 4.7 (36)39.7 ± 2.1 (51)
   Lower T,°C21.9 ± 1.5 (54)22.7 ± 1.0 (36)23.2 ± 0.9 (53)
   Lower RH, %18.9 ± 2.9 (50)27.3 ± 5.0 (36)38.0 ± 2.4 (51)

3. RESULTS

[11] Environmental conditions were closely monitored in the chambers for an 8-week period prior to the experiments and during the two experiments that followed. Carbon dioxide levels in the control chamber were maintained with less than 8% standard variability throughout preexperiment and experimental periods. Elevated chamber levels were maintained with less than 4% standard variability, with no significant difference in performance between the preexperiment and each of the elevated experiments (Table 2 and Figure 2). Temperatures within the chambers were maintained with great consistency throughout the preexperiment and both experimental periods. Within each chamber the upper growth shelf was maintained at ∼29°C; the lower growth shelf was maintained at ∼23°C (standard variability was less than 1°C in all circumstances, see Table 2 and Figure 3). Relative humidity (RH) levels (percent) in the chambers did change in response to plant growth (Table 2 and Figure 4). Both chambers exhibited identical RH values during the preexperiment. The introduction of plants during experiment 1 (elevated RCO2 = 2×) resulted in an average increase in RH of ∼10% in both chambers. The increase in the RH of the control chambers from experiments 1 to 2 may also reflect the increasing RH of the greater environment as Maryland's spring turned to summer. In contrast, in experiment 2 (elevated RCO2 = 3×), RH differed between control and experimental chambers. The enriched pCO2 chamber experienced at least 10% lower RH than did the control chamber. This may be due to lower stomatal conductance at the highest pCO2 (Table 2 and Figure 4) [Hungate et al., 2002]. Plants were watered regularly to preclude water limitation [cf. Beerling and Woodward, 1995].

Figure 2.

Partial pressure of CO2 stability in growth chambers before and during each experiment. For reference, the following levels of pCO2 are included on the graph: Preindustrial (grey line); modern (blue line); RCO2 = 2× (red line); and RCO2 = 3× (purple line). Data from the control chamber are represented by blue circles, while data from the elevated pCO2 chamber are represented by red circles. The black bar represents plus/minus the average standard deviation in pCO2 readings calculated across all experiments.

Figure 3.

Temperature stability in growth chambers before and during each experiment. Preexperiment and experiment 1 represent control and elevated chambers at RCO2 = 1× and 2×, respectively. Experiment 2 represents control and elevated chambers at RCO2 = 1× and 3×, respectively. Data from the lower shelf of the control chamber are represented by plus signs, while data from the lower shelf of the elevated pCO2 chamber are represented by crosses. Similarly, data from the higher shelf of the control chamber are represented by blue circles, while data from the higher shelf of the elevated pCO2 chamber are represented by red circles. The black bar represents plus/minus the average standard deviation in temperature readings calculated across all experiments.

Figure 4.

Relative humidity stability in growth chambers before and during each experiment. Preexperiment and experiment 1 represent control and elevated chambers at RCO2 = 1× and 2×, respectively. Experiment 2 represents control and elevated chambers at RCO2 = 1× and 3×, respectively. Data from the lower shelf of the control chamber are represented by plus signs, while data from the lower shelf of the elevated pCO2 chamber are represented by crosses. Similarly, data from the higher shelf of the control chamber are represented by blue circles, while data from the higher shelf of the elevated pCO2 chamber are represented by red circles. The black bar represents plus/minus the average standard deviation in relative humidity readings calculated across all experiments.

[12] Plants grown under elevated pCO2 showed increased average biomass compared to controls grown at the same temperature. Average aboveground biomass of plants grown under RCO2 = 2× at ∼23°C was ∼36% greater than that grown at ∼23°C under RCO2 = 1× (p < 0.001, see Table 3 and Figure 5a). Average aboveground biomass of plants grown under RCO2 = 3× at ∼23°C was only ∼24% greater than those from the ∼23°C control (p < 0.001). Belowground biomass showed even greater increase under elevated pCO2. Average belowground biomass for plants grown under RCO2 = 2× at ∼23°C was ∼190% greater than that grown at ∼23°C under RCO2 = 1× (p < 0.001, see Table 3 and Figure 5c). Similarly, under RCO2 = 3× at ∼23°C, belowground biomass was ∼62% greater than that of the ∼23°C control (p < 0.001).

Figure 5.

Biomass increases resulting from growth under elevated pCO2. Significant increases were seen between the mean values (squares) of individual plants (circles) (bars represent the standard deviation of individual measurements) both for (a and b) aboveground and (c and d) belowground tissues grown at all temperatures.

Table 3. Summary of the Differences in Biomass Between the Control and Elevated Chambers, Separated According to Upper and Lower Growth Shelf, Resulting From the RCO2 = 2× and RCO2 = 3× Experimentsa
 Upper Growth Shelf
pCO2 = 389 ppmv Biomass, gpCO2 = 778 ppmv Biomass, gRCO2 = 2× Δ Biomass, %
  • a

    Average biomass values are presented with the standard deviation (±σ) seen in the number of plants (n); change in biomass values are presented plus/minus the uncertainty associated with the propagation of σ; these data are presented graphically in Figure 5.

  • b

    This value represents a statistically significant difference (p < 0.02).

Aboveground Biomass
∼22°C0.24 ± 0.06 (8)b0.32 ± 0.07 (8)36 ± 55
∼29°C0.18 ± 0.06 (8)b0.31 ± 0.07 (8)75 ± 74
 
Belowground Biomass
∼22°C0.14 ± 0.07 (8)0.39 ± 0.07 (8)b190 ± 104
∼29°C0.05 ± 0.04 (8)0.21 ± 0.09 (8)b359 ± 283
 Lower Growth Shelf
pCO2= 347 ppmv Biomass, gpCO2= 1088 ppmv Biomass, gRCO2 = 3 × ΔBiomass, %
Aboveground Biomass
∼22°C0.23 ± 0.04 (8)0.28 ± 0.06 (8)24 ± 44
∼29°C0.17 ± 0.03 (8)0.31 ± 0.07 (8)69 ± 54
 
Belowground Biomass
∼22°C0.19 ± 0.08 (8)0.31 ± 0.15 (8)b62± 120
∼29°C0.07 ± 0.04 (8)0.14 ± 0.08 (8)b118 ± 185

[13] Within a given pCO2 level, mean biomass measurements did show a tendency for elevated temperature to result in decreased biomass. However, this trend was not uniform across pCO2 levels, nor was it statistically significant. In the first experiment at RCO2 = 1×, aboveground biomass was significantly lower at higher temperature (p < 0.001, see Table 3), but belowground biomass was not (p = 0.16). However, this result was not replicated in the second experiment; no statistically significant variation in biomass was observed with temperature (Table 3). At RCO2 = 2× and 3× the situation was reversed. Aboveground biomass showed no statistically significant variation with temperature (p = 0.60 and 0.36, respectively), while belowground biomass decreased significantly (p < 0.02, see Table 3 and Figures 5b and 5d).

[14] Measured carbon isotope values for the control and experimental atmospheres (δ13Ca) ranged from −32.6 to −9.0‰ based on the progressive enrichment of laboratory air with isotopically light carbon dioxide from a pressurized CO2 cylinder (Figure 1 and Table 2). One of the CO2 cylinders used was characterized to have δ13CO2 = −42.6‰ and exhibited very low variability over several consecutive measurements (Table 1); however, the values of δ13Ca against which values of δ13Cp were ultimately regressed were measured upon atmosphere directly sampled from the chambers (Table 2). In addition to carbon dioxide concentration the carbon isotope composition of carbon dioxide in the chambers was also measured (Figure 6). The standard deviation of δ13Ca through time is presented as uncertainty when comparing δ13Cp against δ13Ca (Figures 7 and 8).

Figure 6.

Stability of δ13C value in the CO2 of various sources over the course of 10 d of measurements. The data are presented as deviation from the 10-d mean (mean value set at zero).

Figure 7.

High correlation observed between aboveground tissue δ13C and growth chamber δ13CO2 at which the plants grew. R2 values in all cases are ≥0.98; the linear regression for tissues grown at 23°C (blue circles and line) and those grown at 29°C (red circles and line) are compared to the regression seen for the entire aboveground data set (black line). The black bar represents plus/minus the average standard deviation in measured chamber δ13CO2 values and the maximum standard deviation seen in δ13C measured across all tissues.

Figure 8.

High correlation observed between belowground tissue δ13C and growth chamber δ13CO2 at which the plants grew. R2 values in all cases are ≥0.98; the linear regression for tissues grown at 23°C (blue squares and line) and those grown at 29°C (red squares and line) are compared to the regression seen for the entire belowground data set (black line). The black bar represents plus/minus the average standard deviation in measured chamber δ13CO2 values and the maximum standard deviation seen in δ13C measured across all tissues.

[15] Across a very large range in δ13Ca (≈24‰) and pCO2 (≈740 ppmv) we observed a striking and consistent correlation between δ13Ca and δ13Cp (p < 0.001, Figures 7 and 8). For both aboveground and belowground tissue, grown at both ∼23°C and ∼29°C, correlation was strong and significant (r2 ≥ 0.98 and p < 0.001). Figures 7 and 8 illustrate an average isotopic depletion of −25.4‰ for aboveground tissue and −23.2‰ for belowground tissue of Raphanussativus L. relative to the composition of the atmosphere under which it formed. Furthermore, variation in temperature and pCO2 level had little or no effect on this relationship (Figures 7, 8, and 9).

Figure 9.

No correlation observed between Δδ13CPs and the pCO2 level at which the plants grew either for (a) aboveground or (b) belowground tissue. R2 values in all cases are ≤0.1 with no systematic difference between tissues grown at 23°C (blue symbols) and those grown at 29°C (red symbols). The black bar represents plus/minus the average standard deviation in pCO2 readings and the maximum standard deviation seen in δ13C measured across all tissues.

[16] Values of ci/ca calculated for the plants in this study ranged from 0.70 to 0.92 (average equals 0.85). Relative to the full range of values reported previously in the literature (0.38 to 0.89 [Arens et al., 2000]), these values are high. At high ci/ca values the influence of this ecological term on δ13Cp is minimal. Moreover, ci/ca values did not vary systematically with either temperature (p = 0.76), which paralleled relative humidity and thus water potential, or with pCO2 (p = 0.17) level except for the pCO2 experiment run at 778 ppmv, which had a significantly lower average value ((ci/ca) = 0.76).

[17] This experiment was not designed to test the influence of ci/ca on δ13Cp, which is well defined (see equation (1)). Instead, this experiment sought to consider the relationship between plant and atmosphere under the environmental conditions of low water and nutrient stress necessary when seeking to use the carbon isotope composition of plants to reconstruct the isotopic composition of atmospheric CO2. High and mostly invariant values of ci/ca in this experiment reflect effective control of water and nutrient stress over the various experimental and control groups.

[18] We note that ci/ca did not vary at the high and low extremes of pCO2 evaluated in this study. It has been suggested that ci/ca may serve as a proxy for variable atmospheric mixing ratios in the geologic past. This suggestion is an important basis for the hypothesis that the relationship between plant and atmospheric carbon isotope composition should be nonlinear over a wide range of atmospheric compositions. Our data demonstrate that under conditions of adequate water and nutrients (the only conditions under which we advocate the use of δ13Cp to reconstruct δ13Ca), ci/ca does not covary systematically with changing O2:CO2 mixing ratio. Thus the ecological set point was maintained even under conditions where only acclimation, not adaptation, to changing atmospheric composition was possible.

4. DISCUSSION

[19] Elevated atmospheric pCO2 and its potential and actual effects on photosynthesis has been the subject of no less than 15 years of study and controversy [e.g. Bazzaz and Garbutt, 1988; Schlesinger et al., 2006]. An increase of 30% in modern atmospheric CO2 over preindustrial levels due to human activities has been shown [e.g. Keeling et al., 1984]. Within a geological context, pCO2 levels may have been as high as RCO2 = 10× [Berner and Kothavala, 2001] since land plants evolved (∼490 Ma). The first comprehensive studies of the effect of CO2 on plant growth hypothesized that increased pCO2 should bolster the entire process of plant carbon acquisition, resulting in faster plant growth and greater total biomass [e.g. Bazzaz, 1996]. However, many studies have highlighted individualistic responses of plants to elevated atmospheric CO2 levels [e.g., Bazzaz and Miao, 1993; Bazzaz and Williams, 1991; Herrick and Thomas, 2001; Körner, 2000, 2004; Pataki et al., 1998] and rapid acclimation of the photosynthetic mechanism to elevated CO2 [e.g. Ainsworth and Long, 2005; Isebrands et al., 2000; Vu, 2005]. These and studies like them call into question any simple model that predicts increased terrestrial productivity from elevated CO2. Nevertheless, large-scale environmental experiments have generally revealed enhanced production of photosynthetic biomass upon exposure to elevated pCO2 levels [Ellsworth et al., 2004; Rogers and Ellsworth, 2002]. For example, crop plants grown at RCO2 = 2× have yielded an average biomass increase of 30% across a variety of environmental conditions [Mauney et al., 1994; Pinter et al., 1996; Reilly, 1996]. However, responses in natural systems are likely to be considerably more complex [Karnosky, 2003]. For these reasons most discussions of past or future scenarios of RCO2 > 1× include a vision of enhanced terrestrial primary productivity.

[20] The “CO2 fertilization” we documented in Raphanussativus L. is consistent with general descriptions of photosynthetic biomass response to elevated pCO2 [e.g. Ward and Strain, 1999]. Our results are commensurate with a small-scale study performed to specifically compare the effect of elevated pCO2 on the growth of Raphanussativus L. versus Daucus carota L. [Idso and Kimball, 1989]. Idso and Kimball [1989] documented “significant” though variable increase in dry matter production of Raphanussativus L. under RCO2 = ∼2× when compared to plants grown under RCO2 = 1×; they also noted that this increase was enhanced when plants grown under elevated temperature (∼25°C) were compared. Interestingly, the greatest net biomass increases in our experiment were observed at RCO2 = 2×, with further increases in pCO2 yielding biomass increase over control but at levels lower than that of the RCO2 = 2× treatment. We also noted lower average relative humidity in the RCO2 = 3× treatment. Together, these observations suggest reduced stomatal conductance at RCO2 = 3× as part of the plants' short-term acclimation to elevated pCO2 [Knapp et al., 1996; Isebrands et al., 2000; Vu, 2005]. However, the above is speculative: We did not perform the physiological measurements needed to confirm this hypothesis.

[21] One of the strengths of this study, with respect to the relationship between the isotopic composition of atmospheric CO2 and plant tissue derived from it, is that δ13Ca was measured directly from within the chambers (Table 2, see section 2 for details); plant growth studies have included determination of δ13Ca [e.g., Sternberg and DeAngelis, 2002] but have not included systematic determinations of resultant tissue δ13Cp. In addition, δ13Ca variability was quantified across 10 d of consecutive measurements during the experimental period (Table 1 and Figure 6). This approach straightforwardly quantifies the carbon isotopic composition of the actual “raw material” from which plants synthesize their tissue. Our method contrasts with many previous studies that estimate δ13Ca value based on the isotopic composition of the pressurized CO2 tank used for enrichment [e.g., Guy and Reid, 1986] or assume δ13Ca based on plant tissue composition [e.g., Beerling and Woodward, 1995; Polley et al., 1993]. Our work involving previously published values of 176 species [Arens et al., 2000] indicated an average isotopic depletion ≈19‰ in plant tissue relative to the atmosphere under which it formed. We attribute the difference between the isotopic depletion seen in our 2000 study and that recorded here to reflect the individual response of Raphanussativus L. (a species which did not appear in the 2000 study) compared to that of a diverse mix of C3 species used in our 2000 meta-analysis. It has been claimed that single- or few-species isotopic depletion is too variable and represents too poor a correlation to support the idea that δ13Ca can be effectively predicted from δ13Cp [Beerling and Royer, 2002; Royer et al., 2001a, 2001b]. We have made the same assertion with respect to applying such results to the geologic record [Arens and Jahren, 2000; Arens et al., 2000], and the results here seem to bear out that conclusion.

[22] Our experiments further demonstrate an excellent correlation in δ13Ca and δ13Cp that suggests a high level of predictive power across a range of pCO2 levels. As explicitly claimed by Arens et al. [2000], prediction of δ13Ca from δ13Cp is only legitimately performed upon a substrate that effectively mixes the biomass contributions of many individuals of diverse species, ideally grown under stress-free environmental conditions. Arens et al. [2000] asserted that although the δ13Cp value of individual species would parallel that of changing δ13Ca, individual physiological response would emerge when individual plants of a single species were analyzed. Raphanussativus L. conforms to the general model described by Arens et al. [2000], indicating that the relationship is valid and linear across large ranges of pCO2 [cf. Gröcke, 2002]. The lack of dependence of net isotopic depletion on pCO2 level is worth mention (Figure 9). We demonstrated a similar lack of correlation in meta-analysis of the large data set of Arens et al. [2000], and the absence of correlation between fossil plant δ13C and paleo-pCO2 has been reported by many authors [e.g., Bocherens et al., 1993; Ehleringer and Cerling, 1995; Raven and Sprent, 1989; Tu et al., 2004].

5. CONCLUSIONS

[23] Reconstruction of paleo-δ13CO2 using δ13C measurements of plant tissues has been a central part of many carbon cycling and paleoclimate reconstructions [e.g., Arens and Jahren, 2000, 2002; Hasegawa et al., 2003; Jahren, 2002; Jahren et al., 2001, 2005; Strauss and Peters-Kottig, 2003]. Perhaps more widely used is the practice of inferring soil δ13CO2 from soil organic matter δ13C in order to ultimately solve for atmospheric pCO2 level, temperature, or other fundamental climate attribute [e.g., Cerling, 1991, 1992; Ekart et al., 1999; Jahren et al., 2004a; Krull and Retallack, 2000; Ludvigson et al., 1998; Mora et al., 1991, 1996; Nordt et al., 2002; Tabor et al., 2004; Yapp, 2004; Yapp and Poths, 1996]. The current study further validates our initial conclusion that in the absence of environmental stress, plant δ13C primarily reflects atmospheric δ13CO2 linearly across pCO2 levels.

[24] It is worth discussing several caveats, both old and new, that arise during the application of our measured isotopic offset to δ13Cp in order to solve for δ13Ca. First, the results presented here, while reinforcing the idea that such an offset is real and constant in Raphanussativus L., highlight the importance of including many and mixed plant species into the isotopic substrate. The isotopic offset we document for Raphanussativus L. differs from that seen for a mixed-species group [Arens et al., 2000] by no less than 4‰. Similar offsets can be observed in other single-species trials [e.g., Beerling and Woodward, 1995; Gröcke, 2002; Guy et al., 1989; Holtum et al., 1983]. Arens et al. [2000] further reported that single-species variation is systematic, with conifers displaying slightly lower discrimination. We contend that a substrate comprising the contribution of multiple species and many individual plants is available within a few cubic centimeters of many terrestrial rocks and comprises the optimal substrate for δ13Ca reconstruction. In order to explore substrates that reflect little or no diagenesis, researchers have also explored the isolation and analysis of many types of specific structures and compounds exclusively associated with terrestrial land plants, including wood [Gröcke et al., 1999; van Bergen and Poole, 2002], cuticle [Arens and Jahren, 2000, 2002; Jahren, 2002; Jahren et al., 2001, 2005; van Bergen et al., 1995], pollen [Jahren, 2004; Amundson et al., 1997; Loader and Hemming, 2000, 2001], rbcL gene DNA [Jahren et al., 2004b, 2006], and general “biomarkers” [Briggs et al., 2000; Pancost and Boot, 2004]. In each case, discrimination imparted during synthesis of specific plant components must be considered. As with any method of paleoreconstruction, diagenetic alteration must be considered, assessed, and ideally minimized during the selection of isotopic substrate. To facilitate this, the literature contains extensive exploration of the persistence and potential diagenetic alteration of plant tissues and compounds [e.g., Briggs et al., 2000; Poole et al., 2002, 2004; Tu et al., 2004].

[25] Recent studies have highlighted additional and subtle complications in the measurement of δ13Cp needed to reconstruct δ13Ca. For example, the fluctuation of pO2 through geologic time [Berner, 2001] may have affected plant metabolic processes [Beerling et al., 2002]. Incubations of isolated plant mitochondria indicate that rates of photorespiration may also affect overall carbon isotope fractionation during photosynthesis more than previously assumed [Igamberdiev et al., 2001]. Both of these issues could be approached using a controlled growth chamber with variable gaseous inputs, such as the experiments described here. The particularity of these results to the single species Raphanussativus L. is counterbalanced by the large ranges of tightly controlled pCO2 levels and δ13CO2 values afforded by the experimental design. Additional experiments at several more values of δ13Ca would strengthen the conclusions presented here and allow for an exhaustive assessment of nonlinearity. Our study illustrates the utility of controlled growth experiments in the investigation of the consistent offset hypothesized between δ13Cp and δ13Ca and its application to paleoenvironmental reconstructions.

Acknowledgments

[26] This work benefited from the laboratory expertise of W. M. Hagopian; it was supported by the Howard Hughes Foundation, by NSF/DEB-0107520 and NSF/EAR-0106171, and by DOE-BES:DE-FG02-06ER15831.

[27] The Editor responsible for this paper was Michael Manga. He thanks technical reviewers Isabel Montanez and Gregory J. Retallack and one anonymous cross-disciplinary reviewer.

Ancillary