Fossil soils constrain ancient climate sensitivity

Global temperatures have covaried with atmospheric carbon dioxide (CO₂) over the last 450 million years of Earth’s history (1). Critically, ancient greenhouse periods provide some of the most pertinent information for anticipating how the Earth will respond to the current anthropogenic loading of greenhouse gases. Paleo-CO₂ can be inferred either by proxy or by the modeling of the long-term carbon cycle. For much of the geologic past, estimates of CO₂ are consistent across methods (1). One exception is the paleosol carbonate proxy, whose CO₂ estimates are often more than twice as high as coeval estimates from other methods (1). This discrepancy has led some to question the validity of the other methods and has hindered attempts to understand the linkages between paleo-CO₂ and other parts of the Earth system. In this issue of PNAS, Breecker and colleagues (2) break important new ground for resolving this conflict.

The paleosol carbonate proxy for atmospheric CO₂ is based on the analysis of carbonate nodules that precipitate in soils in seasonally dry to dry climates. These nodules incorporate carbon from two sources: atmospheric CO₂ that diffuses directly into the soil and in situ CO₂ from biological respiration. Because the stable carbon isotopic composition of these two sources is distinct, the concentration of atmospheric CO₂ can be inferred if the concentration of soil CO₂ and the isotopic compositions of the two sources are known (3). Atmospheric CO₂ estimates scale directly with soil CO₂ concentration: If the soil term is wrong by a factor of two, the inferred atmospheric CO₂ will be off by a factor of two. Estimates of soil CO₂ concentration for fossil soils have been based on measurements taken during the growing season in equivalent living soils. However, Breecker et al. (2, 4) demonstrate convincingly that the window of active carbonate formation is restricted to the warmer and dryer parts of the growing season. Carbonate formation is simply not thermodynamically favorable during cooler and wetter seasons. Critically, biological productivity and respiration are low during these dry periods. As a result, soil CO₂ concentration during the critical window of active carbonate formation has been overestimated in most soils by a factor of two or more (2).

What does this mean? CO₂ estimates from the paleosol carbonate proxy can be cut in half (or more). Doing so snaps the paleosol-based estimates in line with most other approaches (2) (Fig. 1B) and produces the most precise view to date of Earth’s CO₂ history. We are now better equipped to answer some important, basic questions. For example, what is the quantitative relationship between CO₂ and temperature? That is, for every doubling of CO₂, what is the long-term (10³–10⁴ years) equilibrium response of global temperature (termed here climate sensitivity)? Most assessments of climate sensitivity for the present day hover around 3°C per CO₂ doubling (5), although if the long-term waxing and waning of continental ice sheets are considered it is probably closer to 6°C (6). Less is known about climate sensitivity during ancient greenhouse periods, simply because having poles draped in forest instead of ice represents a profound rearrangement of climate feedbacks.

Fig. 1.

Constraints on climate sensitivity for the globally warm Cretaceous and early Paleogene (125–40 Mya). (A) Tropical sea surface temperature (SST) records (15–29). Boxes represent studies with high sampling density. The red band corresponds to estimated tropical SST under an atmospheric CO₂ concentration of 1,000 ppm and a climate sensitivity (ΔT_2×) of 3°C per CO₂ doubling. Temperatures above this band probably represent times when climate sensitivity exceeded 3°C (see text). (B) Atmospheric CO₂ records (compilation from ref. 2). Dashed lines correspond to present-day CO₂ (387 ppm; A.D. 2009) and 1,000 ppm CO₂. Records from the boron proxy are excluded because they are likely unreliable (30); records from the goethite and nahcolite proxies are excluded because they are not yet widely applied.

Records of CO₂ and temperature are now sufficiently robust for placing firm minimum constraints on climate sensitivity during parts of the Cretaceous and early Paleogene (125–40 Mya), a well-known globally warm interval. Indeed, owing to the logarithmic relationship between CO₂ and temperature, the geologic record is ideally suited for establishing minimum thresholds. This is because, to accommodate a declining sensitivity, other boundary conditions of the Earth system need to shift exponentially, for example, unreasonable oscillations in atmospheric CO₂. Policywise, establishing a basement value for climate sensitivity is a critical step for addressing our current climate crisis (5).

With few exceptions, CO₂ during the Cretaceous and early Paleogene was <1,000 ppm (2) (Fig. 1B). Global mean surface temperature is very difficult to establish for these ancient periods. However, temperature change in the tropics today scales at roughly two-thirds the global change (5, 6). If we assume a similar relationship in the past and a climate sensitivity of 3°C per CO₂ doubling, a rise in atmospheric CO₂ to 1,000 ppm results in a 3.6°C warming in the tropics (relative to a 280-ppm baseline). Given that tropical sea surface temperatures range from 27° to 29°C today, tropical temperatures exceeding 30.6°–32.6°C (red band in Fig. 1A) during the Cretaceous and early Paleogene likely correspond to a climate sensitivity >3°C.

This threshold was commonly surpassed during the Cretaceous and early Paleogene (Fig. 1A). For times when CO₂ was <1,000 ppm, the tropical temperature threshold for a 3°C climate sensitivity would shift to correspondingly cooler values. Further, there is abundant evidence

Breecker et al. have contributed significantly toward improving one of the more popular paleo-CO₂ proxies.

for flatter latitudinal temperature gradients during greenhouse periods (7, 8), meaning, again, that the tropical temperature threshold used here is probably a maximum. Together, it is clear that during the Cretaceous and Paleogene climate sensitivity commonly exceeded 3°C per CO₂ doubling.

Although further work is needed, the geologic evidence (2) (Fig. 1) is most consistent with long-term, future climate change being more severe than presently anticipated (5). Also, global climate models tuned to ancient greenhouse periods commonly have emergent climate sensitivities of <3°C and they fail to simulate the shallow latitudinal temperature gradients (9). Thus even for times with little ice, there are important positive feedbacks that are presently not captured adequately in climate models. Processes for warming the high latitudes without a change in CO₂ include more vigorous heat transport (10, 11), more widespread stratospheric clouds in the high latitudes (12), and climate feedbacks from polar forests (13).

and their study highlights the value of a clearly resolved paleo-CO₂ record. However, a limitation is that they uniformly apply a “best guess” value of 2,500 ppm for soil CO₂ concentration. They recognize this as an oversimplification and is an area for future work. Better modeling of the term, perhaps through independent proxy (14), may result in a further tightening of the paleo-CO₂ record.