Comparison of climate records from the Pliocene and Pleistocene geological epochs of the past five million years suggests that positive climate feedbacks are not strengthened during warm climate intervals. See Article p.49
A major concern in projecting future climate change using models is that positive climate feedbacks might become enhanced in a warm climate, accelerating future warming in response to rising greenhouse-gas levels. Climate feedbacks are changes in atmospheric or surface properties induced by climate change that magnify or diminish the overall temperature response. Their aggregate strength is represented by the climate sensitivity, which is the ratio of observed warming to climate forcing, such as increasing atmospheric carbon dioxide levels. Warm intervals of Earth's recent geological past, which can be studied through climate proxies, provide a basis for testing the response of climate sensitivity to warming. On page 49 of this issue, Martínez-Botí et al.1 use improved proxy atmospheric CO2 data to compare climate-sensitivity determinations from the warm Pliocene epoch, 5.3 million to 2.6 million years (Myr) ago, to those from the cold, extensively glaciated Pleistocene epoch, 2.6 to 0.012 Myr ago. They find that climate sensitivity differs little between these vastly dissimilar times, once the influence of ice sheets is removed.
Why should climate sensitivity be stronger in a warm world? A warmer world is likely to have less snow and ice, thereby reducing their amplifying effect on climate change2,3. But how other feedbacks, such as water vapour and clouds, respond to warming is less certain. Simulations with climate models suggest that the positive feedback due to water vapour may strengthen in warmer climates4, but uncertainties about how cloud feedbacks respond to warming confuse our understanding of the overall dependence of climate sensitivity on climate state.
Ancient climate records provide an alternative approach to assessing climate sensitivity, through the analysis of proxies, which reveal both the forcing (for example, atmospheric CO2 levels or ice extent) and response (the temperature change). This approach offers the tremendous advantage of relying on natural equilibrium climate states rather than on synthetic ones simulated in models. Past climates were also influenced by various slow feedbacks such as ice sheets, vegetation and dust — factors typically not included in climate simulations. However, the method hinges on proxy reconstructions that have associated uncertainties, especially for marine-based atmospheric CO2 reconstructions used in studies reaching beyond 0.8 Myr ago, the age of the oldest ice cores5.
Palaeoclimate researchers have targeted the Pliocene epoch because it is the most recent time interval in which conditions were substantially warmer, about 2–3 °C warmer globally, than pre-industrial conditions6. Proxy reconstructions indicate that the Arctic climate during the Pliocene was much warmer than it is today, about 8–19 °C warmer, depending on location and season7 (Fig. 1). But this extreme Arctic warmth seems to have coexisted, paradoxically, with atmospheric CO2 levels that are similar to the present ones, implying an extreme amplification of positive climate feedbacks in the Pliocene8.
Martínez-Botí and colleagues challenge this existing hypothesis using a well-validated technique to reconstruct Pliocene atmospheric CO2 between 3.3 and 2.3 Myr ago at higher temporal resolution and with less variability than previous proxy reconstructions5. Their reconstruction clearly indicates for the first time that mid-Pliocene atmospheric CO2 was up to 60% higher than pre-industrial values (450 parts per million (p.p.m.), compared with 280 p.p.m. before the Industrial Revolution and 400 p.p.m. today). The new record also indicates clear transitions in atmospheric CO2 that are coherent with known climate events, including a drop in atmospheric CO2 between 2.9 and 2.7 Myr ago that precedes global cooling and the onset of Northern Hemisphere glaciation 2.6 Myr ago — remarkable findings in themselves.
The researchers go a step further, applying their results to climate sensitivity for the warm Pliocene state by developing averaged reconstructions of land and ocean temperature and comparing them directly to their atmospheric CO2 reconstruction. The slope relating the forcing (atmospheric CO2) and response (temperature) at each time slice yields a tight constraint on climate sensitivity that is specific to the Pliocene. The authors find Pliocene climate sensitivity to be half as strong as that found for the cold Pleistocene. A repeat of the analysis after removing the forcing associated with glacial–interglacial changes in ice sheets reveals that Pliocene and Pleistocene climate sensitivities to atmospheric CO2 changes alone were essentially the same.
In a broader context, these results also relate to attempts to use the instrumental temperature record to narrow the range of equilibrium climate sensitivity, which is the equilibrium temperature change caused by a doubling of atmospheric CO2 allowing for 'fast' feedback processes only. Some studies have argued9 that the slightly weaker rate of global warming since 2001 reduces the lower boundary of equilibrium climate sensitivity to well below 2 °C. Although Martínez-Botí and colleagues' derived Earth-system sensitivity10 includes slow feedbacks, which complicates direct comparison to results from climate models, their results are likely to translate10 to an equilibrium climate sensitivity of between 2 and 3 °C, well within the generally accepted range.
Despite the significant advance Martínez-Botí and co-workers' study represents, several challenges remain. First, given the wide range of proxy atmospheric CO2 data for the Pliocene5, it will be essential to validate the new results and assess why earlier reconstructions and methodologies differ from this one. Second, the extreme Pliocene warming in the terrestrial Arctic (Fig. 1) still requires enhanced polar climate feedbacks that remain unexplained7. Finally, for climate modellers, there remains the substantial challenge of reconciling emerging palaeoclimate-based sensitivity results with simulations of both past and future climate states.