The mid-Miocene (15.98 to 13.82 Ma) was characterized by substantially warmer temperatures than today and atmospheric CO2 concentrations comparable to near-future projections. Climate models have generally struggled to reproduce proxy-based reconstructions from this interval, particularly at high latitudes where model temperatures are consistently lower than observations. Here, we present new mid-Miocene simulations using a previously unpublished geography and evaluate the climate's sensitivity to several key components: paleogeography (including land-sea distribution, topography and ice sheets), atmospheric CO2 concentration, atmospheric model choice, and solar forcing. Our baseline mid-Miocene climate yields a global mean surface temperature (GMST) of 19.8 °C. In mid-Miocene sensitivity experiments of two and four times pre-industrial CO2 concentrations, consistent with estimates for the mid-Miocene, GMST varies by up to 3.2 °C between simulations. Removal of the Antarctic ice-sheet leads to expected local warming of around 25 °C at the maximum height of the ice sheet, but nevertheless records an overall global cooling of 1.3 °C. Solar forcing and subtle changes of land-sea mask each impact GMST by around 0.2 °C. The choice of atmospheric model substantially affects the simulated mid-Miocene climate through modified feedback mechanisms. We estimate an equilibrium climate sensitivity (ECS) of 2.9 °C (2.5–3.3 °C, 95 % prediction interval) for the mid-Miocene, similar to modern-based estimates from our model (2.8 °C, 2.2–3.4 °C, 95 % prediction interval), indicating the potential for the Miocene to contribute to constraining ECS. Global precipitation is tightly coupled to GMST across all our simulations. As with previous studies, all our simulations, regardless of specific configuration, underestimate high-latitude proxy-reconstructed temperatures. This highlights the need to improve our understanding on polar amplification and on the limitations affecting the proxy record.

Geological evidence of a snowball Earth state indicate persistent tropical sea ice cover during the Neoproterozoic (>635 million years ago). Current theory is that a strengthening of the positive surface albedo feedback with cooling temperatures, eventually exceeding the sum of all other feedbacks, leads to a global climate instability. Several recent high sensitivity climate models with strongly positive cloud feedbacks have not been able to simulate the much warmer Last Glacial Maximum (LGM) state, suggestive that they cool excessively in response to a modest decrease in atmospheric carbon dioxide levels and therefore enter the snowball instability by this mechanism. Using a coupled Earth system model, MPI-ESM1.2, we show that clouds accelerate the transition to a snowball Earth state and reduce the radiative forcing required to trigger the snowball instability. Positive cloud feedbacks over tropical oceans and ahead of the sea-ice edge act to cool down the oceans and promote sea ice formation. Regardless, when approached slowly, the snowball Earth transition appears to occur around a global mean temperature of zero degree Celsius, simultaneously with the sea ice edge advancing into the sub-tropics thereby strengthening the surface albedo feedback. This temperature threshold, if supported by several climate models, could be used as a novel and independent constraint on the upper bound of climate sensitivity by using the relationship of simulated LGM temperatures and the models' equilibrium climate sensitivity. The constraint depends only on the simple fact that Earth did not enter a snowball instability during the recent ice ages. Using the here estimated transition temperature, we find it is implausible that Earth's climate sensitivity exceeds 6.2 °C (3.9–8.4 °C, 5 %–95 % confidence). This upper bound estimate of climate sensitivity is only weakly sensitive to uncertainty in the transition temperature, approximately 0.3° per degree.

We evaluate five commonly-applied criteria to validate that a climate model is in so-called “quasi-equilibrium,” using a suite of five simulations with CO₂ concentrations between 1× and 16× Pre-Industrial values. We find that major changes in ocean circulation can occur after common thermal equilibrium criteria are reached, such as a small Top of Atmosphere radiative flux imbalance, or weak trends in surface air temperature, sea surface temperature, and deep ocean temperature. Ocean circulation change, in turn, impact high-latitude SAT, sea ice, and the Inter-tropical Convergence Zone position. For future modeling studies and intercomparison projects aiming for an ocean in quasi-equilibrium, we suggest that time series of key meridional overturning circulation (MOC) metrics in the Atlantic, Pacific, and Southern Ocean are saved, and that MOC trends are less than 1 Sv/1000 years, and DOT trends less than 0.1°C/century for the final 1000 years of the simulations.

Subtropical gyre (STG) depth and strength are controlled by wind stress curl and surface buoyancy forcing1,2. Modern hydrographic data reveal that the STG extends to a depth of about 1 km in the Northwest Atlantic, with its maximum depth defined by the base of the subtropical thermocline. Despite the likelihood of greater wind stress curl and surface buoyancy loss during the Last Glacial Maximum (LGM)3, previous work suggests minimal change in the depth of the glacial STG4. Here we show a sharp glacial water mass boundary between 33° N and 36° N extending down to between 2.0 and 2.5 km—approximately 1 km deeper than today. Our findings arise from benthic foraminiferal δ18O profiles from sediment cores in two depth transects at Cape Hatteras (36–39° N) and Blake Outer Ridge (29–34° N) in the Northwest Atlantic. This result suggests that the STG, including the Gulf Stream, was deeper and stronger during the LGM than at present, which we attribute to increased glacial wind stress curl, as supported by climate model simulations, as well as greater glacial production of denser subtropical mode waters (STMWs). Our data suggest (1) that subtropical waters probably contributed to the geochemical signature of what is conventionally identified as Glacial North Atlantic Intermediate Water (GNAIW)5–7 and (2) the STG helped sustain continued buoyancy loss, water mass conversion and northwards meridional heat transport (MHT) in the glacial North Atlantic.

Recent observation-based studies suggest that the Amazon rainforest has lost substantial resilience since 1990, indicating that the forest might undergo a critical transition in the near future due to global warming and deforestation. The idea is to use trends in a lag-1 auto-correlation of leaf density as an early-warning signal of an imminent critical threshold for rainforest dieback. Here we test whether the observed change in auto-correlations could arise from internal variability using historical and control simulations of nine sixth-generation Earth system model ensembles (Phase 6 of the Coupled Model Intercomparison Project, CMIP6). We quantify trends in the leaf area index auto-correlation from both models and satellite-observed vegetation optical depth from 1990 to 2017. Four models reproduce the observed trend with at least one historical realization whereby the observations lie at the upper limit of model variability. Three out of these four models exhibit similar behavior in control runs, suggesting that historical forcing is not necessary for simulating the observed trends. Furthermore, we do not observe a critical transition in any future runs under the strongest greenhouse gas emission scenario (SSP5-8.5) until 2100 in the four models that best reproduce the past observed trends. Hence, the currently observed trends could be caused simply by internal variability and, unless the data records are extended, have limited applicability as an early-warning signal. Our results suggest that the current rapid decline in the Amazon rainforest coverage is not foremost caused by global warming.

The use of paleoclimates to constrain the equilibrium climate sensitivity (ECS) has seen a growing interest. In particular, the Last Glacial Maximum (LGM) and the mid-Pliocene warm period have been used in emergent-constraint approaches using simulations from the Paleoclimate Modelling Intercomparison Project (PMIP). Despite lower uncertainties regarding geological proxy data for the LGM in comparison with the Pliocene, the robustness of the emergent constraint between LGM temperature and ECS is weaker at both global and regional scales. Here, we investigate the climate of the LGM in models through different PMIP generations and how various factors in the atmosphere, ocean, land surface and cryosphere contribute to the spread of the model ensemble. Certain factors have a large impact on an emergent constraint, such as state dependency in climate feedbacks or model dependency on ice sheet forcing. Other factors, such as models being out of energetic balance and sea surface temperature not responding below −1.8^∘C in polar regions, have a limited influence. We quantify some of the contributions and find that they mostly have extratropical origins. Contrary to what has previously been suggested, from a statistical point of view, the PMIP model generations do not differ substantially. Moreover, we show that the lack of high- or low-ECS models in the ensembles critically limits the strength and reliability of the emergent constraints. Single-model ensembles may be promising tools for the future of LGM emergent constraint, as they permit a large range of ECS and reduce the noise from inter-model structural issues. Finally, we provide recommendations for a paleo-based emergent constraint and notably which paleoclimate is ideal for such an approach.

The addition of carbon dioxide (CO₂) in the atmosphere due to human activities is the main driver of global warming. How much the Earth will warm in the future is often represented by the Earth's equilibrium climate sensitivity (ECS), the long-term temperature response considering the effect of climate feedbacks after an abrupt and sustained doubling of atmospheric CO₂ from pre-industrial concentration. Assessing ECS is critical as it is one of the most relevant metric to evaluate global temperature change by 2100 in a fast warming climate. However, there have been considerable difficulties in constraining ECS for more than a century. In recent years, there has been a focus on alternative lines of evidence to elicit ECS, such as the study of past climates.

The work of this thesis investigates the evidence on ECS and climate feedbacks obtained from paleoclimates. In our studies, we use past climate reconstructions and climate modelling to estimate ECS out of the cold Last Glacial Maximum (LGM) and the warm Pliocene. Our work focuses on the statistical relationship existing between simulated past temperatures and ECS following the emergent constraint theory, and how the physics of modelled paleoclimates can affect such relationship. We explore further how climate feedbacks behave and depart from a linear behaviour in extreme cold conditions by performing simulations of snowball Earth states.

This thesis demonstrates that both LGM and Pliocene are relevant candidates to elicit ECS and highlights the contribution of paleoclimates in understanding modern and future climate change. In particular, we show that the Pliocene is a robust constraint on ECS under the emergent constraint theory despite large observational uncertainties. Our estimate of ECS using the most recent generation of climate models is 4.8 K, which lies in the high end of previous assessment from Pliocene evidence. On the contrary, the LGM constraint is weak due to substantial differences in ice sheet forcing as well as differences in the behaviour of climate feedbacks in cold temperatures in climate models. Our results suggest that LGM simulated temperatures are challenging to use in emergent constraint framework on ECS. An alternative difficulty in using the LGM arises from the lack of high ECS models in the ensemble. Our results indicate that the minimum global temperature for the LGM state is around 0°C, where the strengthening of the sea-ice albedo feedback with cooling temperatures and a substantial contribution of cloud feedbacks will then move the climate towards a snowball state. Highly sensitivity models are most likely to fail at simulating the LGM when approaching these low temperatures.

The thesis highlights the importance of using a variety of models with different sensitivity to simulate paleoclimates and use them in estimating ECS and feedbacks. Warm paleoclimates such as the Pliocene are likely the best candidates to infer ECS. These estimates of ECS are dependent on geological reconstructions which are continuously improving. Assessments on the role of past climates in constraining ECS and feedbacks are therefore key elements in understanding both paleoclimates and future climate change and should be considered with great interest.

Circum-Arctic glacial ice is melting in an unprecedented mode, and release of currently trapped geological methane may act as a positive feedback on ice-sheet retreat during global warming. Evidence for methane release during the penultimate (Eemian, ca. 125 ka) interglacial, a period with less glacial sea ice and higher temperatures than today, is currently absent. Here, we argue that based on foraminiferal isotope studies on drill holes from offshore Svalbard, Norway, methane leakage occurred upon the abrupt Eurasian ice-sheet wastage during terminations of the last (Weichselian) and penultimate (Saalian) glaciations. Progressive increase of methane emissions seems to be first recorded by depleted benthic foraminiferal δ¹³C. This is quickly followed by the precipitation of methane-derived authigenic carbonate as overgrowth inside and outside foraminiferal shells, characterized by heavy δ¹⁸O and depleted δ¹³C of both benthic and planktonic foraminifera. The similarities between the events observed over both terminations advocate for a common driver for the episodic release of geological methane stocks. Our favored model is recurrent leakage of shallow gas reservoirs below the gas hydrate stability zone along the margin of western Svalbard that can be reactivated upon initial instability of the grounded, marine-based ice sheets. Analogous to this model, with the current acceleration of the Greenland ice melt, instabilities of existing methane reservoirs below and nearby the ice sheet are likely.

In this paper we introduce a Bayesian framework, which is explicit about prior assumptions, for using model ensembles and observations together to constrain future climate change. The emergent constraint approach has seen broad application in recent years, including studies constraining the equilibrium climate sensitivity (ECS) using the Last Glacial Maximum (LGM) and the mid-Pliocene Warm Period (mPWP). Most of these studies were based on ordinary least squares (OLS) fits between a variable of the climate state, such as tropical temperature, and climate sensitivity. Using our Bayesian method, and considering the LGM and mPWP separately, we obtain values of ECS of 2.7K (0.6-5.2, 5th-95th percentiles) using the PMIP2, PMIP3, and PMIP4 datasets for the LGM and 2.3K (0.5-4.4) with the PlioMIP1 and PlioMIP2 datasets for the mPWP. Restricting the ensembles to include only the most recent version of each model, we obtain 2.7K (0.7-5.2) using the LGM and 2.3K (0.4-4.5) using the mPWP. An advantage of the Bayesian framework is that it is possible to combine the two periods assuming they are independent, whereby we obtain a tighter constraint of 2.5K (0.8-4.0) using the restricted ensemble. We have explored the sensitivity to our assumptions in the method, including considering structural uncertainty, and in the choice of models, and this leads to 95% probability of climate sensitivity mostly below 5K and only exceeding 6K in a single and most uncertain case assuming a large structural uncertainty. The approach is compared with other approaches based on OLS, a Kalman filter method, and an alternative Bayesian method. An interesting implication of this work is that OLS-based emergent constraints on ECS generate tighter uncertainty estimates, in particular at the lower end, an artefact due to a flatter regression line in the case of lack of correlation. Although some fundamental challenges related to the use of emergent constraints remain, this paper provides a step towards a better foundation for their potential use in future probabilistic estimations of climate sensitivity.