Atmospheric CO₂ concentrations during ancient greenhouse climates were similar to those predicted for A.D. 2100

Abstract

Quantifying atmospheric CO₂ concentrations ([CO₂]_atm) during Earth’s ancient greenhouse episodes is essential for accurately predicting the response of future climate to elevated CO₂ levels. Empirical estimates of [CO₂]_atm during Paleozoic and Mesozoic greenhouse climates are based primarily on the carbon isotope composition of calcium carbonate in fossil soils. We report that greenhouse [CO₂]_atm have been significantly overestimated because previously assumed soil CO₂ concentrations during carbonate formation are too high. More accurate [CO₂]_atm, resulting from better constraints on soil CO₂, indicate that large (1,000s of ppmV) fluctuations in [CO₂]_atm did not characterize ancient climates and that past greenhouse climates were accompanied by concentrations similar to those projected for A.D. 2100.

The anthropogenically driven rise in [CO₂]_atm is well established (1) but its effect on future climate is less certain (2). Many recent studies indicate that [CO₂]_atm has controlled or strongly amplified Phanerozoic (542 Ma–present) climate variations (3–8) and therefore understanding the relationship between [CO₂]_atm and climate over geologic time provides crucial empirical constraints on the magnitude of future global warming (1, 9). Estimates of Paleozoic and Mesozoic [CO₂]_atm are largely based on the soil carbonate CO₂ paleobarometer (10), which is the most temporally continuous proxy (indicator) for [CO₂]_atm over the past 400 million years. The CO₂ paleobarometer is also considered the most reliable provider of [CO₂]_atm estimates for times when [CO₂]_atm was significantly above modern values (11). The CO₂ paleobarometer suggests that [CO₂]_atm values exceeded 3,000 parts per million by volume (ppmV) during Permian (289–251 Ma) and Mesozoic (251–65 Ma) greenhouse climates (5, 8, 12). However, other [CO₂]_atm proxies, which are either considered to be less reliable at high [CO₂]_atm (stomatal index) or are newly developed and therefore less widely utilized (e.g., fossil bryophytes), typically result in [CO₂]_atm estimates for greenhouse climates that are much lower than estimates from soil carbonate (5, 6). The large discrepancy among proxies can be interpreted two ways: 1) large (1,000s of ppmV) variations in [CO₂]_atm occurred over relatively short time periods (in certain cases shorter than the temporal resolution of the proxy records) throughout the Phanerozoic or 2) some of the proxy estimates are inaccurate. In this study we use data from modern soils and incorporate an improved understanding of pedogenic carbonate formation to recalibrate the CO₂ paleobarometer. We report that the most often quoted [CO₂]_atm values, those previously determined from pedogenic carbonate, are too high, and that paleo [CO₂]_atm values did not persist above 1,500 ppmV during the past 400 million years.

Pedogenic (soil) carbonate (calcite, CaCO₃) forms in soils where potential evapotranspiration exceeds precipitation, typically in arid to subhumid regions that receive less than 100 cm of rain per year. , released primarily by the dissolution of Ca-bearing minerals in dust (13) is carried to depth by downward-percolating water and eventually reprecipitates as pedogenic carbonate. The dissolution and precipitation of calcite in soils occurs by the reaction:

and the following equation illustrates changes in the soil environment that can drive calcite precipitation

[1]

where a_CaCO3 is the activity of calcite; m_Ca²⁺ is the concentration of calcium ions in aqueous solution; pCO₂ is the partial pressure of CO₂ in the soil gas; and K₁, K₂, K_cal, and K_CO2 are temperature-sensitive equilibrium constants for the dissociation of carbonic acid, the dissociation of bicarbonate, the dissociation of calcite, and the hydration of CO₂, respectively. Eq. 1 is valid for the system CaCO₃-H₂O-CO₂ assuming activities equal concentrations and pH < 9. At thermodynamic equilibrium, calcite will precipitate when a_CaCO3 reaches a value of 1. The value of a_CaCO3 is increased by the following changes: 1) an increase in the concentration of Ca²⁺ in soil solution, 2) a decrease in soil pCO₂, or 3) an increase in soil temperature (the product of the equilibrium constants increases with temperature, meaning calcite is less soluble at higher temperature). Quantification of m_Ca²⁺ required for a_CaCO3 to reach a value of 1 under seasonally varying soil conditions indicates that the simultaneous occurrence of evapotranspiration, degassing of CO₂, and heating of the soil probably drives pedogenic carbonate formation (14).

Pedogenic carbonate is thought to form in carbon isotope (¹³C/¹²C) equilibrium with soil CO₂ (CO₂ in the gas-filled soil pore spaces) because carbonate precipitates slowly and changes in the soil environment below 20–30 cm occur relatively slowly. Therefore, pedogenic carbonates provide a record of the carbon isotope composition of coexisting soil CO₂. The carbon isotope composition of soil CO₂ is, in turn, influenced by [CO₂]_atm because CO₂ in soil pore spaces is a mixture between atmospheric CO₂ and CO₂ respired by organisms in the soil (15). Thus, measured carbon isotope compositions of pedogenic carbonate can be used to calculate [CO₂]_atm at the time the carbonate formed using the following equation derived from an isotope mass balance relationship (16):

[2]

where S(z) is the soil-derived component of total CO₂ in the soil (i.e., S(z) = [CO₂]_soil - [CO₂]_atm) at depth z; δ¹³C is the carbon isotope composition in standard delta notation;* s, r, and a refer to soil CO₂, soil-derived (respired) CO₂, and atmospheric CO₂, respectively; and the coefficient 1.0044 and the constant 4.4 derive from the difference in diffusivity between and .

In order to determine ancient [CO₂]_atm, the value of used in Eq. 2 is calculated from measured δ¹³C values of carbonate preserved in paleosols (fossil soils) using the temperature-dependent carbon isotope fractionation factor between CO₂ and calcite (17). Values for carbonate formation temperature, , and are either assumed or estimated from other proxies.^† Values for S(z) were originally and somewhat arbitrarily assumed to be ∼5,000–10,000 ppmV (10). That range was primarily based on mean growing season CO₂ concentrations in modern soils that do not contain pedogenic carbonate (16, 18) and with only a few exceptions (e.g., 8) has yet to be substantially revised. Efforts to determine the range for S(z) through measurement of soil CO₂ concentrations in modern calcic soils are notably lacking. Therefore, values for S(z) that are appropriate in Eq. 2 remain poorly constrained despite the fact that calculated values of [CO₂]_atm are directly proportional to S(z). Improved estimates of S(z) are required to increase the accuracy and reduce the uncertainty in estimates of paleoatmospheric CO₂ concentrations.

Results

We have determined that pedogenic carbonate does not form under mean growing season soil conditions (14) as had previously been assumed (10, 12, 16) and therefore appropriate values for S(z) are significantly less than 5000 ppmV. Results from monitoring the concentration and stable isotope composition of soil CO₂ in central New Mexico demonstrate that pedogenic carbonate forms under seasonally warm and very dry conditions (14). As a soil warms and dries, all of the variables in Eq. 1 change in such as way as to contribute to an increase in a_CaCO3: temperature increases, m_Ca²⁺ increases as water is removed from the soil, and soil pCO₂ decreases as soil respiration rates become moisture limited. The observation that low soil CO₂ concentrations help drive the precipitation of calcite in soils suggests that the value of S(z) during pedogenic carbonate formation is significantly lower than it is during mean growing season conditions. S(z) during pedogenic carbonate formation is below 1,000 ppmV in the desert soils studied and is below 2,500 ppmV in soils forming in a Piñon-Juniper Woodland despite the fact that CO₂ concentrations in these soils increase to values of ∼3,000 and ∼7,000 ppmV, respectively, during the wet summer monsoon season (14). The lower values attending carbonate formation are valid for soils forming in semiarid regions and are the only published values of S(z) during pedogenic carbonate formation (14).

Clay-rich paleosols that formed in moister (subhumid) regions with temperate and tropical climates are commonly used for paleoatmospheric CO₂ barometry. A compilation of minimum CO₂ flux measurements from calcic, clay-rich soils in temperate grasslands, tropical grasslands, and tropical forests suggest that S(z) values in these soils decrease below 2,500 ppmV during dry, hot episodes (Table 1). Therefore, despite the fact that mean annual soil CO₂ concentrations in temperate and tropical soils are generally higher than in desert soils, the S(z) values that occur in clay-rich soils from subhumid regions are similar to those that occur in gravelly soils from semarid regions during hot, dry episodes when pedogenic carbonate forms. A compilation of soil CO₂ concentration measurements supports the idea that large seasonal variations in soil CO₂ concentration occur in calcic soils and that warm-season S(z) values can decrease below 2,500 ppmV, even in climates with mean annual precipitation in excess of half a meter per year (28–31, Fig. 1).

Table 1.

Compilation of soil surface CO₂ fluxes.

Ecosystem	Soil Order	max Jo*	min†Jo	max S(z) (ppmV)	min S(z) (ppmV)	Ref.
temperate grassland	Vertisol	55.5	3.5	25,700	1,620	(19)
temperate grassland	Mollisol	26.1	6.3	12,100	2,920	(20)
temperate grassland	Mollisol	57.6	3.6	26,700	1,670	(21)
temperate grassland	Mollisol	20.0	4.2	9,300	1,940	(22)
temperate grassland	Inceptisol slightly calcareous	28.8	3.6	13,300	1,670	(23)
tropical grassland	alluvial soil	29.2	4.6	13,500	2,130	(24)
tropical grassland	Oxisol	13.5	6.5	6,200	3,010	(25)
tropical forest	Oxisol	3.5	1.7	3,200	1,560	(26)
tropical forest	Oxisol	11.0	1.5	10,100	1,380	(27)

^*J_o is the soil surface CO₂ flux in mmol/m²/hr

^†Minimum summer values are shown for the temperate grasslands; minimum annual values are shown for tropical soils

Fig. 1.

A compilation of observed seasonal variations in soil pore space CO₂ concentrations. Data from modern calcic soils with native vegetation are shown. The soil atmosphere was sampled from soil gas wells installed below 50 cm in all studies. Mean annual precipitation (cm) for each study site is: New Mexico Piñon-Juniper woodland = 38, West Texas grassland = 40, Sonoran Desert shrubland = 30, Southern Nevada Fir-Pine = 55, Saskatchewan grassland = 35. Thin, horizontal lines highlight 5,000 ppmV, typically used for paleosol barometry, and 2,500 ppmV, a more accurate estimate of S(z) during pedogenic carbonate formation in most soils.

The large variations in soil CO₂ concentrations indicated in Table 1 and Fig. 1 may be the most important factor influencing pedogenic carbonate formation and are probably a universal characteristic of calcic soils. The high CO₂ concentrations during warm, wet periods of the year are responsible for carbonate dissolution and the mobilization of Ca²⁺, whereas decreasing concentrations associated with hot, dry periods are responsible for carbonate precipitation. Therefore, appropriate values for S(z) in Eq. 2 are not those that occur during mean growing season conditions when [CO₂]_soil is near its highest, but rather those that occur during hot, dry periods when [CO₂]_soil is near its lowest value and pedogenic carbonate is forming.

A calibration of the CO₂ paleobarometer using Holocene (11.7 ka–present) pedogenic carbonate supports the idea that S(z) values during carbonate formation in soils are lower than previously thought. Table 2 shows values of S(z) that were calculated by solving Eq. 2 for S(z) and then evaluating the resulting expression using measured values of all of the other variables. Previously compiled δ¹³C values of pedogenic carbonate from Holocene soils in different climates (12) were used with soil temperatures 5 °C higher than the mean growing season soil temperatures reported by (12) (characteristic of hot, dry periods) (14) to calculate . [CO₂]_atm and were set to their preindustrial values of 280 ppmV and -6.5‰, respectively (32). values were taken as the δ¹³C values of organic matter in the soils, which were reported previously (12). The resulting mean S(z) value is 2,800 ppmV, approximately half the typically assumed value of 5,000 ppmV and similar to S(z) values determined from surface efflux measurements and the stable isotope-based study of soil carbonate formation (14). The calculated S(z) values in Table 2 range from 1,000 to 6,000 ppmV, suggesting that the uncertainty in S(z) for any one soil is currently quite large. Calculating S(z) from measurements of the depth to the soil carbonate horizon (33) may eventually reduce uncertainty in S(z). However, this technique would benefit from the consideration of all of the factors that influence the depth of carbonate in soils (34) and a calibration in which the S(z) values during carbonate formation are determined and the depth to carbonate is measured in the same soils.

Table 2.

Values of S(z) calculated from Holocene soils.

	*	*	*	Soil T (°C)†		atm CO2 (ppmV)	S(z) (ppmV)
Bolivia	-23.3	-6.5	-8.5	16	-18.4	280	5,300
France	-25.0	-6.5	-10.0	17	-19.7	280	3,838
Greece	-25.7	-6.5	-9.3	19	-18.8	280	1,329
Greece	-23.7	-6.5	-7.5	19	-17.0	280	1,245
Nevada	-23.7	-6.5	-8.5	12	-18.8	280	6,139
Nevada	-23.9	-6.5	-8.5	11	-19.0	280	5,389
Nevada	-23.4	-6.5	-6.8	13	-17.0	280	1,433
New York	-25.6	-6.5	-9.4	21	-18.7	280	1,297
Saskatchewan	-24.2	-6.5	-7.9	20	-17.3	280	1,168
Saskatchewan	-24.1	-6.5	-8.4	20	-17.8	280	1,586
Turkey	-24.5	-6.5	-10.0	23	-19.0	280	3,019
Turkey	-24.5	-6.5	-10.3	23	-19.3	280	4,151
Utah	-24.5	-6.5	-7.4	16	-17.3	280	1,034
Utah	-24.4	-6.5	-7.5	16	-17.4	280	1,120
Utah	-23.8	-6.5	-8.8	16	-18.7	280	4,094
mean							2,809
1σ							1,842

^*δ¹³C values are reported in ‰ vs PDB

^†5 °C higher than temperature used by Ekart et al 1999.

Discussion

Atmospheric CO₂ concentrations through the past 400 million years were recalculated in this study using an S(z) value of 2,500 ppmV, which is our best estimate for S(z) based on the results discussed above. The recalculated [CO₂]_atm values are compared with previously published CO₂ records in Fig. 2. The recalculated values are up to 2,500 ppmV lower than those previously estimated from pedogenic carbonate (5, 12) and fall into agreement with estimates from other proxies (e.g., 6, 35–37) (Fig. 2B), thereby reconciling much of the discrepancy that had existed among different proxies. The recalculated proxy-based [CO₂]_atm values are also in good agreement with [CO₂]_atm values from the most recent version of the GEOCARB carbon cycle model (38) (Fig. 2C). The use of a single S(z) value for all paleosols is certainly a simplification and is being refined currently by further study of modern soils. However, S(z) = 2,500 ppmV is appropriate for averages of large numbers of paleosols, and we argue that it improves [CO₂]_atm estimates for much of the Phanerozoic.

Fig. 2.

A compilation of Phanerozoic atmospheric CO₂ records. (A) A compilation of atmospheric CO₂ records from different proxies. Paleosol carbonate-based estimates are from original papers in which S(z) values between 4,000 and 10,000 ppmV were used in Eq. 2. Compilation from (5) with additional proxy-based CO₂ estimates (6, 8, 37, 39–48, D. Royer, pers. comm.). 10 million year bin means of the paleosol carbonate-based estimates and of all the other proxy-based estimates are shown by the black and green lines, respectively. (B) Same as Fig. 2A with [CO₂]_atm estimates from paleosol carbonate recalculated using S(z) = 2500 ppmV (see Methods). The estimates from paleosol carbonate (black curve) and all other proxies (green curve) are in far better agreement than they are in Fig. 2A (two goethite data points at 6,300 ppmV (174 and 188 Ma) were not included in the 10 million year bin means shown in Fig. 2A and Fig. 2B because these outliers have a strong influence on the averages and may represent transient fluctuations rather than typical Mesozoic atmospheric conditions). (C) Comparison of the [CO₂]_atm estimates from proxies with [CO₂]_atm estimates from GEOCARBSULF (38). The region shown for GEOCARBSULF output (the blue region in front) incorporates the full range of temporal variability in granite ⁸⁷Sr/⁸⁶Sr values considered in the model. The most recently published compilation of proxy-based [CO₂]_atm estimates (9) is shown in yellow (10 million year bins means ± 1σ). The revised proxy-based estimates (10 million year bin means ± 1σ shown in red) are generally in good agreement with the GEOCARBSULF model results and are substantially lower than the most recent compilation of proxy-based estimates. (D) Paleolatitudinal extent of Phanerozoic glaciations (49) showing the strong correlation between [CO₂]_atm and glaciation. CO₂ concentrations predicted for the year A.D. 2100, assuming a heterogeneous world in which technological advancements spread slowly, population grows continuously, and economic development is primarily regional rather than global (2) (SRES anthropogenic emissions scenario A2), are shown as the horizontal gray bar in A, B, and C. Figure modified from (5).

Until now, the apparent disagreement among proxy estimates has obscured our understanding of paleo [CO₂]_atm. Reconciling the discrepancy between CO₂ estimates from paleosol carbonate and other techniques indicates that large (1,000s of ppmV) variations in [CO₂]_atm did not characterize the Mesozoic Era. Moreover, the agreement between multiple proxies strongly supports the conclusion that the warmest episodes of the Mesozoic were associated with [CO₂]_atm equal to ∼1,000 ppmV rather than 2,000–3,000 ppmV (5, 8) (compare Fig. 2A and B). The relatively low [CO₂]_atm of 1,000 ppmV during greenhouse episodes suggest that either Mesozoic warmth was partially caused by a factor unrelated to CO₂ or that the Earth’s climate is much more sensitive to atmospheric CO₂ than previously thought.

Comparison of projected future [CO₂]_atm (2) with results from the recalibrated CO₂ paleobarometer (Fig. 2B) indicate atmospheric CO₂ may reach levels similar to those prevailing during the vegetated Earth’s hottest greenhouse episodes by A.D. 2100. The abrupt increase in [CO₂]_atm during the Early Permian is similar in magnitude to that possible for the next century in the absence of CO₂ mitigation (Fig. 2B). Given that the Early Permian CO₂ increase may have caused the termination of the Late Paleozoic Ice Age (7, 8, Fig. 2D), the only known icehouse-greenhouse transition on a vegetated Earth, the effects that unmitigated CO₂ increases may have on future climate warrant careful consideration.

Materials and Methods

Individual [CO₂]_atm estimates based on paleosol carbonate were recalculated in this study using:

where [CO₂]_atm,old is the previously reported value, S(z)_old is the S(z) value originally used to calculate [CO₂]_atm,old (4,000–10,000 ppmV), and S(z)_new is equal to 2,500 ppmV. S(z) values in modern clay-rich soils were calculated from soil surface CO₂ flux measurements using a steady state production-diffusion model (16). The average depth of soil respiration used in the model was 10 cm for grassland ecosystems and 20 cm for forest ecosystems. Soil porosity, tortuosity, and temperature were assumed to be 0.4, 0.6, and 25 °C, respectively.