Impact of a Permo-Carboniferous high O₂ event on the terrestrial carbon cycle

Abstract

Independent models predicting the Phanerozoic (past 600 million years) history of atmospheric O₂ partial pressure (pO₂) indicate a marked rise to approximately 35% in the Permo-Carboniferous, around 300 million years before present, with the strong potential for altering the biogeochemical cycling of carbon by terrestrial ecosystems. This potential, however, would have been modified by the prevailing atmospheric pCO₂ value. Herein, we use a process-based terrestrial carbon cycle model forced with a late Carboniferous paleoclimate simulation to evaluate the effects of a rise from 21 to 35% pO₂ on terrestrial biosphere productivity and assess how this response is modified by current uncertainties in the prevailing pCO₂ value. Our results indicate that a rise in pO₂ from 21 to 35% during the Carboniferous reduced global terrestrial primary productivity by 20% and led to a 216-Gt (1 Gt = 10¹² kg) C reduction in the vegetation and soil carbon storage, in an atmosphere with pCO₂ = 0.03%. However, in an atmosphere with pCO₂ = 0.06%, the CO₂ fertilization effect is larger than the cost of photorespiration, and ecosystem productivity increases leading to the net sequestration of 117 Gt C into the vegetation and soil carbon reservoirs. In both cases, the effects result from the strong interaction between pO₂, pCO₂, and climate in the tropics. From this analysis, we deduce that a Permo-Carboniferous rise in pO₂ was unlikely to have exerted catastrophic effects on ecosystem productivity (with pCO₂ = 0.03%), and if pCO₂ levels at this time were >0.04%, the water-use efficiency of land plants may even have improved.

Atmospheric O₂ is a key gas regulating the metabolism of the Earth's aerobic biota. A Phanerozoic (past 600 million years) history of atmospheric O₂ partial pressure (pO₂) shows that pO₂ over much of this time was relatively stable as a result of a variety of geological and biological feedbacks, with an important excursion to about 35% centered at around 300 million years B.P. during the Permo-Carboniferous (see ref. 1 by Berner for a review). This marked pO₂ increase results from the evolution of vascular land plants on the continents (2, 3) and enhanced burial of recalcitrant organic matter in swamps (4), the latter being represented by abundant and widespread coal deposits of this age. The high Permo-Carboniferous value was calculated originally from the abundance of organic carbon and pyrite sulfur (FeS₂) in sedimentary rocks, because global fluxes of reduced carbon and sulfur are the two dominant pO₂ controls on a time scale of millions of years (1–3). More recently, an independent approach to modeling Phanerozoic pO₂ evolutionary history, by using global carbon and sulfur isotope mass balance analyses and incorporating O₂-sensitive isotope fractionation by the terrestrial and marine biota (5), closely reproduced the large pO₂ peak at 300 million years B.P.; this excursion is consistent with biological data from the fossil record such as the sudden rise and fall in insect gigantism (6).

Given that two rather different approaches to modeling pO₂ history in the atmosphere point to a high Permo-Carboniferous pO₂ value, there is a need to assess its likely effects on the photosynthetic productivity of vascular land plants and the terrestrial biosphere as a whole at this time. The need is underscored, because the dual carboxylase-oxygenase function of Rubisco (ribulose-1,5-bisphosphate carboxylase/oxygenase), the primary CO₂-fixing enzyme of C₃ land plants, is influenced strongly by atmospheric pO₂ and pCO₂, with the potential to modify the biogeochemical cycling of carbon by terrestrial ecosystems. Carboxylation leads to photosynthesis via the photosynthetic carbon reduction pathway, and oxygenation leads to photorespiration via the carbon oxidation pathway, with the evolution of CO₂ (7). A high pO₂ atmosphere favors the oxygenation reaction of Rubisco, with O₂ tending to out compete CO₂ for the acceptor molecule ribulose bisphosphate, leading to increased photorespiration and decreased net photosynthetic CO₂ fixation (7). However, crucial to determining the oxygenation/carboxylation ratio of Rubisco is the prevailing pCO₂ value, and for the Permo-Carboniferous, some uncertainty exists regarding its value. Estimates of pCO₂ during the late Carboniferous from soil carbonates and organic matter (8) overlap those made from theoretical considerations of the long-term carbon cycle (9); however, these estimates encompass a range of values (between 0.03 and 0.08%). A simple consideration of Rubisco kinetics leads to the expectation that this range will be critical for determining the impact of 35% O₂ on rates of photosynthetic CO₂ uptake (10–12).

Temperature further exerts an important modifying influence on the efficiency of Rubisco, by altering the relative solubility of CO₂ and O₂ and the specificity of Rubisco for CO₂ (11, 12). Consequently, pCO₂, O₂, and climate will all interact to modify carbon cycling by terrestrial ecosystems, and such interactions require a global-scale approach for an adequate assessment of the Permo-Carboniferous high O₂ event on the biosphere. Herein, we take the global view to assess first the effect of a rise in pO₂ from 21 to 35% on the primary productivity of terrestrial vegetation and carbon storage in vegetation biomass and soil organic matter at a constant pCO₂ content (0.03%) by using a process-based terrestrial carbon cycle model (13, 14) forced with a global general circulation model (GCM) simulation of the late Carboniferous climate (15, 16). We next determine through a series of sensitivity experiments how this response is modified by uncertainties in the Permo-Carboniferous pCO₂. Changes in the possible functioning of the terrestrial biosphere at this time, as predicted by the model, are compared and discussed with data from plant growth experiments in which the pO₂ and pCO₂ values were manipulated.

Materials and Methods

Terrestrial Carbon Cycle Modeling.

The University of Sheffield terrestrial carbon cycle model simulates, under steady-state conditions of climate and atmospheric composition (CO₂ and O₂), the basic plant processes of photosynthesis, respiration, and transpiration (13, 14). Canopy transpiration is regulated by stomatal conductance and feeds back to influence soil water status. The aboveground productivity module is dynamically coupled to the Century biogeochemistry model (17) describing the cycling of carbon and nitrogen in soils. Surface litter inputs (leaves and roots) derived from the vegetation are decomposed through the various Century routines to compute soil nutrient status that, in turn, influences vegetation primary productivity. Equilibrium model solutions were obtained by iteration under a given climate and atmospheric composition for 500 years. The key model outputs are net primary productivity (NPP), leaf area index, canopy transpiration, and carbon storage in vegetation biomass and soil organic matter.

Paleoclimate Simulations.

In all of the modeling experiments described herein, the carbon cycle model was forced with a global paleoclimate simulation (monthly temperature, precipitation, and relative humidity) representing the late Carboniferous, which was made by using the U.K. Universities' Global Atmospheric Modeling Program (ugamp) GCM at the University of Reading (Reading, U.K.). A full description of this model is given by Valdes et al. (18). Briefly, the model has a horizontal spatial resolution of 3.75° × 3.75° with 19 levels in the vertical. Continental positions were similar to those used by Crowley and Baum (19); other major boundary conditions were 3% lower solar luminosity than the present day, a pCO₂ value of 300 ppm, and orbital configuration of the Pleistocene interglacials. The simulation used prescribed sea surface temperatures based on energy balance results that were energetically consistent with the choice of CO₂ and solar constant. The model was integrated for 5 years, and data from the last 2 years were averaged to produce the Carboniferous climate. Further details of the simulation and resulting global patterns of mean annual temperature and precipitation are given by Beerling et al. (15) and Valdes and Crowley (16).

Model Experiments.

To assess the effect of the Permo-Carboniferous rise in O₂ and the uncertainty in pCO₂ on the terrestrial carbon cycle, four simulations were performed. The first, at 21% O₂ and 0.03% CO₂, defines case 1, the “control.” Three simulations were then made at 35% O₂, each with a different pCO₂ level (0.03%, case 2; 0.045%, case 3; 0.06%, case 4); these three pCO₂ values represent, respectively, lower, “best guess,” and upper estimates from theoretical modeling (20). A Permo-Carboniferous pO₂ value of 35% was taken from two geochemical model predictions (3, 5). Each pCO₂ and pO₂ value was prescribed within the aboveground vegetation productivity module of the terrestrial carbon model. In that module, CO₂ and O₂ act on rates of photosynthesis through the well validated, biochemically based mechanistic model of leaf photosynthesis of Farquhar et al. (21). All simulations were performed with the same GCM paleoclimate data set for different prescribed pCO₂ and pO₂ values. We assume that plants with the C₃ photosynthetic pathway dominated Carboniferous floras, because there is only limited and equivocal evidence for plants operating with any other photosynthetic pathway in the Carboniferous (e.g., ref. 22).

Results and Discussion

We first assessed the effect of a marked Permo-Carboniferous rise in pO₂ on global terrestrial biosphere productivity under the late Carboniferous GCM paleoclimate, assuming a constant 0.03% pCO₂ value. Comparison of the two appropriate simulations (cases 1 and 2) indicate a rise from 21 to 35% O₂ reduced global vegetation productivity by 20% (Table 1 and Fig. 1). Under these circumstances, this reduction implies high O₂ impacts on the oxygenation/carboxylation reactions of Rubisco within individual leaves, which influence the operation of the terrestrial biosphere, even under the relatively cool Carboniferous paleoclimate. However, sensitivity analyses indicate that if the same rise in pO₂ occurred in an atmosphere with pCO₂ = 0.06% (case 4), then the CO₂ fertilization effect on photosynthetic productivity is larger than the cost of photorespiration, such that terrestrial NPP actually increases by 3 Gt C per yr⁻¹ (Table 1). The intermediate case 3, with pCO₂ = 0.045%, results in the inhibitory effects of high O₂ on vegetation productivity being largely cancelled out (Table 1). In this scenario, therefore, severe suppression of terrestrial productivity by plants with the C₃ photosynthetic pathway and their interactions with resource availability owing to a Permo-Carboniferous pO₂ excursion would seem unlikely.

Table 1.

Global terrestrial NPP and carbon storage in vegetation biomass and soil carbon reservoirs during the late Carboniferous

Simulation	NPP, Gt C yr−1	Vegetation biomass, Gt C	Soil organic matter, Gt C
Case 1. 21% O2, 0.03% CO2 (control case)	47	473	1,004
Case 2. 35% O2, 0.03% CO2	38 (−9)	373 (−100)	888 (−116)
Case 3. 35% O2, 0.045% CO2	44 (−3)	463 (−10)	984 (−20)
Case 4. 35% O2, 0.06% CO2	50 (+3)	554 (+81)	1,084 (+80)

Results were obtained by forcing the terrestrial carbon cycle with the same climate with different prescribed pO₂ and pCO₂ values. Values in parentheses denote the differences relative to the control case. Note that 1 Gt = 10¹² kg. 

Figure 1.

Global distribution of NPP (t C per ha⁻¹⋅yr⁻¹), vegetation biomass (kg C per m⁻²), and soil carbon concentrations (kg C per m⁻²) in the late Carboniferous for the atmospheric composition represented by the control case (21% O₂, 0.03% CO₂; a, d, and g, respectively), and the modeled effects of the Permo-Carboniferous pO₂ increase at two different pCO₂ levels, given as the difference between case 2 (35% O₂, 0.03% CO₂) and the control (b, e, and h) and between case 4 (35% O₂, 0.06% CO₂) and the control (c, f, and i). Note that maps b, c, and h as well as c, f, and i are difference maps and are plotted on different scales. Each individual map shows the three major landmasses of the land-sea mask used in the climate and vegetation modeling. These were Gondwana, dominating the Southern hemisphere; Eurasia, the largest Northern hemisphere landmass; and adjacent to this landmass, Kazakhstan.

Plant growth experiments with different pO₂/CO₂ values support the direction of change in NPP revealed by these model simulations. For example, a 20% reduction in global NPP through a rise in pO₂ to 35% is mirrored by an observed 30% reduction in the photosynthetic rates of Betula pubescens leaves when measured at 21 and 35% pO₂ (15). At the whole plant scale, changes in leaf photosynthesis translate to lowered plant biomass (10, 23, 24) with the vegetative biomass of Pancium bisulcatum being reduced by 30% after 24 days of growth in 40% O₂ compared with 21% (24). The compensatory effects of increasing pCO₂ during a high pO₂ event, as indicated by the model results (Table 1 and Fig. 1), is also supported by growth experiments (23, 25). The growth of soybean (Glycine max) was reduced from 28 g per plant (dry mass) at 21% O₂ and 0.03% CO₂ to 15 g per plant when the pO₂ value was increased to 40% (23); however, as with the terrestrial biosphere response, this effect was strongly diminished when plant growth pCO₂ was increased to 0.07%.

Annual changes in NPP caused by a Permo-Carboniferous pO₂ rise feed through to influence the resulting size of the carbon reservoirs in vegetation biomass and soils in a manner that mirrors the NPP responses (Table 1). When the pO₂ rise occurred with pCO₂ = 0.03%, shown by the difference between case 1 and 2, reductions in NPP occurred throughout each of the major landmasses (Fig. 1b). However, the geographical extent of reductions in the vegetation carbon pool is more localized than that of NPP, being restricted to the northern margins and throughout central Gondwana (Fig. 1e). This result reflects the loss of accumulated carbon from forests (26). Similarly, it is only in these same forested regions where vegetation biomass increases when the pO₂ rise is simulated together with an associated rise in pCO₂ to 0.06% (Fig. 1f), because these are the only dominant plant functional types able to store significant additional carbon in stem biomass gained from increased NPP.

The response of the soil carbon pool to the prescribed pO₂ and pCO₂ values is geographically more widespread than the response of vegetation biomass (Fig. 1), because the soil surface carbon pool has a faster turnover time (27) and is therefore quite responsive to O₂-induced changes in litter production. In consequence, the pattern and direction of change in the size of the soil carbon reservoir is similar to that of NPP (Fig. 1). A rise in pO₂ to 35% with pCO₂ = 0.03% (case 2), for example, reduces soil carbon concentrations (relative to the control, case 1; Fig. 1h); this rise occurs, because the soils receive reduced surface litter input (leaves and roots) from the vegetation but, under the given GCM climate, are subjected to the same proportion of CO₂ being lost through plant and soil respiration. A rise in pO₂ with pCO₂ = 0.06% (the difference between cases 4 and 1) increases the soil carbon reservoir (Fig. 1i) reflecting higher production rates of surface litter organic matter by the vegetation and a high C:N ratio caused by plant growth in a high pCO₂ environment. Globally, the total differences in the vegetation carbon reservoir resulting from a rise in pO₂ with various pCO₂ values were similar to those in the soil carbon reservoir (Table 1).

The global-scale terrestrial carbon cycle simulations allow us to asses how climate modifies the pCO₂ and pO₂ interactions on Rubisco efficiency and, in turn, how this modification influences annual NPP. These interactions are brought out clearly by the latitudinally averaged NPP responses for each case, when expressed relative to the control (case 1; Fig. 2). This plot indicates that the largest modification of the high O₂ response by different pCO₂ values occurs in the warm equatorial regions. In case 2 (pO₂ = 35% and pCO₂ = 0.03%), the oxygenation reaction dominates, leading to increased photorespiratory CO₂ losses and a net reduction in NPP (Fig. 2). However, when the O₂ rise is simulated with pCO₂ = 0.06% (case 4), the carboxylation reaction of Rubisco dominates, allowing NPP to increase in those regions, whereas the intermediate case (Table 1) shows that the effects of a Permo-Carboniferous pO₂ of 35% are largely cancelled by an atmosphere with pCO₂ = 0.045% (Fig. 2).

Figure 2.

Effects of the Permo-Carboniferous pO₂ rise at three different pCO₂ levels on the latitudinal gradient of terrestrial NPP during the late Carboniferous. The gradients represent the difference between the case 2 and the control simulation (21% O₂, 0.03% CO₂; dashed line), case 3 and the control simulation (dot-dash line), and case 4 and the control simulation (solid line).

Analysis on a site-by-site basis (i.e., individual grid squares) for cases 2 and 4 clearly separates these divergent climate interactions on Rubisco, where the mean annual temperature is >5°C (Fig. 3a). Vegetation operating in an atmosphere defined by case 2 (pO₂ = 35%, pCO₂ = 0.03%) shows a progressive reduction with temperature in NPP relative to the control at sites where the mean annual temperature is between 5 and 30°C (Fig. 3), because increasing temperature favors the oxygenation of ribulose bisphosphate by decreasing, relative to O₂, both the solubility of CO₂ and specificity of Rubisco for CO₂ (11, 12). Between temperatures of 7 and 35°C, the specificity effect accounts for two-thirds of the reduction in NPP, and the solubility effect accounts for the remaining third (12). In a high CO₂ environment, the opposite effect occurs, with NPP increasing as the mean annual temperature rises, despite the high pO₂ value (Fig. 3a), because the key effect of elevated pCO₂ is increased competitive inhibition of oxygenation and hence photorespiration. The strong temperature-dependent oxygenation therefore is suppressed, and net photosynthetic rates rise proportionally. Such an effect is similar to that observed in gas exchange measurements on C₃ plants species under different pO₂ and pCO₂ conditions (28).

Figure 3.

Effects of the Permo-Carboniferous pO₂ rise at two different pCO₂ levels on annual (a) terrestrial NPP, (b) canopy transpiration (E_t), and (c) water-use efficiency (WUE, defined as NPP divided by E_t) during the late Carboniferous, analyzed on a site-by-site basis. Open symbols represent results from case 2 (35% O₂, 0.03% CO₂); solid symbols represent results from case 4 (35% O₂, 0.06% CO₂). All values are expressed as the differences between results from the specified atmospheric composition and the control (case 1).

A consequence of lower photosynthetic rates in a high O₂ environment is that pCO₂ within the leaf rises, leading to partial stomatal closure (29–31), although this effect is not always the case at pO₂ values <35% (e.g., ref. 32). Such considerations lead to the suggestion that high O₂-induced shifts in stomatal conductance might alter canopy transpiration rates which, together with changes in photosynthetic productivity documented above, may have influenced vegetation water-use efficiency during the Permo-Carboniferous. To examine this suggestion, we first tested for the potential of leaf-scale stomatal pO₂ effects to upscale to whole canopies by comparing, on a site-by-site basis, canopy transpiration rates for cases 2 and 4 relative to the control. As intimated by leaf gas exchange measurements (29–31), the model data show that, at the majority of sites, annual canopy transpiration (E_t) was reduced relative to the control by a rise in pO₂ to 35% (Fig. 3b). For case 2 (pO₂ = 35%, pCO₂ = 0.03%), the reduction is brought about through lowered photosynthetic productivity reducing canopy conductance to water vapor and to a lesser extent leaf area index. For case 4, there is an overriding effect of pCO₂ = 0.06% that strongly reduces stomatal conductance, despite a small increase in leaf area index, and this reduction results in a much stronger decline in E_t (Fig. 3b). Combining the NPP and E_t responses for the two cases reveals a clear, previously unrealized difference in the response of vegetation water-use efficiency (Fig. 3c), suggesting that an elevated pO₂ episode during the Permo-Carboniferous could have influenced the water economy of vegetation.

Our results and those from growth experiments (10, 23–25) indicate that a Permo-Carboniferous high O₂ event, sustained for several million years with pCO₂ ≈ 0.03%, could have exerted strong selection pressures on the functioning of Rubisco, through favoring the oxygenation over carboxylation reaction. In this respect, it is intriguing to note that the timing of the pO₂ excursion predicted from geochemical models is similar to the date obtained from molecular clocks for the split between conifer-cycad and angiosperm lineages (33). In fact, the latter group of plants have stomatal characteristics that tend to maximize CO₂ diffusion into the leaf (34), thereby raising intercellular CO₂ concentrations and reducing CO₂ evolution by photorespiration (15). The timing and functional significance issues therefore suggest that a Permo-Carboniferous high O₂ episode might have triggered this split between major plant groups. Regardless, patterns of plant evolution seem to be at least circumstantially linked with the predictions of current models of Phanerozoic atmospheric O₂ history.

Abbreviations

pO₂

O₂ partial pressure

pCO₂

CO₂ partial pressure

Rubisco

ribulose-1,5-bisphosphate carboxylase/oxygenase

GCM

general circulation model

NPP

net primary productivity