Skip to content
What's on

Feedbacks on climate in the Earth system









The Royal Society, London, 6-9 Carlton House Terrace, London, SW1Y 5AG


Scientific discussion meeting organised by Professor Eric Wolff FRS, Professor John Shepherd CBE FRS, Dr Emily Shuckburgh and Professor Andrew Watson FRS

A field of sea ice, illustrating the sea ice feedback. Credit: British Antarctic Survey/Chris Gilbert

Event details

The response of Earth’s climate system to a perturbation depends on the sign and strength of several feedback processes. This meeting will present critical assessments of major feedbacks, including those (such as ice sheets and the carbon cycle) operating over long timescales. For each, their role in past and present climate change, and their expected future effects will be discussed.

The draft programme (PDF) is available to download. Biographies of the organisers and speakers are available below, and speaker abstracts will be made available closer to the meeting date. Recorded audio of the presentations will be available on this page after the event and the papers will be published in a future issue of Philosophical Transactions A.

Attending this event

This event has already taken place. Recorded audio of the presentations can be found below.

The meeting was immediately followed by a related, two-day satellite meeting, Climate feedbacks - setting the research agenda, at the Royal Society at Chicheley Hall, home of the Kavli Royal Society Centre.

Enquiries: Contact the events team

Event organisers

Select an organiser for more information

Schedule of talks

Session 1: Water vapour-cloud-climate feedbacks

4 talks Show detail Hide detail

Mechanisms of low-latitude cloud feedback

Professor Christopher Bretherton, University of Washington, USA


For decades, cloud feedbacks have been a leading source of uncertainty in the climate sensitivity simulated by global climate models. The largest inter-model difference in net cloud feedback are in the subtropics, where they are mainly due to diverse responses of simulated marine boundary layer cloud to greenhouse warming. As a result, some climate models simulate very little net global feedback of cloud changes on greenhouse warming, while others simulate a reduction in subtropical cloud cover and strong positive feedback. It has proved challenging to constrain global cloud feedbacks using observations, due to the large natural variability of cloudiness and the ever-changing observing system.

This presentation describes the application of high-resolution (turbulence-resolving) simulations using small domains and fine grid spacings of 5-100m to study the response of marine cloud-topped boundary layer clouds to specified large-scale atmospheric changes expected to accompany greenhouse warming. Results are found to be more robust to the model formulation than are the cloud responses predicted by climate models, which are very sensitive to subgrid parameterisation choices for boundary-layer turbulence, shallow cumulus convection and clouds.   Using multiple turbulence-resolving model simulations, the response is found to combine four mechanisms, each of which has observational support: a  ‘thermodynamic’ cloudiness reduction from the warming of the atmosphere-ocean column, a ‘radiative’ cloudiness reduction from the CO2-induced increased atmospheric emissivity, a ‘stability-induced’ cloud increase due to increased stratification of the subtropical marine lower troposphere, and a ‘dynamical’ cloudiness increase due to reduced subsidence. The cloudiness reduction mechanisms dominate, leading to net positive cloud feedbacks.  

When combined with physical parameterisation uncertainties, these counteracting physical mechanisms lead to a broad spread of predicted cloud responses in global climate models.  They involve complex interaction between parameterisations that remains challenging to unravel using either single-model or multi-model studies. We are still working toward a framework that allows effective comparison of cloud feedbacks in turbulence-resolving models, climate models, and observations.

Show speakers

De-evolving climate models

Mark Webb, Met Office Hadley Centre, UK


SPOOKIE is a recent initiative building on the experimental protocol of the Cloud Feedback Model Intercomparison Project (CFMIP). Its aims are to establish the relative contributions of different areas of model physics to inter-model spread in cloud feedback, and their roles in mechanisms of robustly simulated cloud feedback. The approach is to perform "mechanism denial" sensitivity experiments where specific processes such as parametrised convection are removed or simplified. Pilot amip/amip4K "ConvOff" experiments with parametrised convection switched off are presented. We find that models are able to run without parameterised convection at current climate model resolutions. Four of the models (MRI-CGCM3, MIROC5, HadGEM2-A and MPI-ESM-LR) retain positive low-level cloud feedbacks in stable regimes in the subtropics in ConvOff experiments, indicating that processes other than convection contribute to positive subtropical feedback.  However these models' cloud feedbacks exhibit a strong convergence in character in ConvOff experiments compared to the standard model versions, with the range in the global mean net cloud feedback being reduced substantially. The net cloud feedback shows a reduced or similar spread across all tropical stability regimes, and a relatively smooth and monotonic transition from weaker net cloud feedback in unstable regimes to more positive feedback in stable regimes. Hence we conclude that inter-model differences in the details of convective parametrisations contribute substantially to inter-model spread in cloud feedback in these models. The CNRM-CM5 model behaves somewhat differently however; removing parametrised convection introduces a strong negative cloud feedback in ascending regions, largely driven by increases in cloud fraction and cloud liquid water at around the freezing level. Future plans to develop this approach, for example by simplifying convective parametrisations, will be discussed.

Show speakers

Feedbacks and climate sensitivity

Professor Reto Knutti, ETH Zurich, Switzerland


Climate is changing rapidly due to man-made carbon dioxide emissions, and impacts are expected to increase with warming. Climate sensitivity and the transient climate response characterise the equilibrium and transient warming to changes in the atmospheric carbon dioxide concentration, and can help with the translation of a specified warming target into required atmospheric carbon dioxide levels. The most recent assessment of the Intergovernmental Panel on Climate Change indicates a likely range (>66% probability) for climate sensitivity of 1.5-4.5°C. The quest to determine climate sensitivity has now been going on for decades, with disturbingly little progress in narrowing the large uncertainty range, but substantial new insights into the climate system. However, the discrepancy between climate sensitivity estimates from the warming in the ocean and atmosphere, and those based on models simulating present day climate also point to several open questions. Some comprehensive models may overestimate feedbacks, simple energy balance models that assume constant feedbacks in response to radiative forcing likely oversimplify the problem, observations of ocean heat uptake and estimates of the aerosol forcing are still uncertain, and natural variability on decadal timescales may be more relevant than previously assumed. Nevertheless, a few robust conclusions on the transient and equilibrium response and the implications for the required carbon dioxide emission reductions to limit climate change are possible.

Show speakers

Formation of aerosol particles and cloud condensation nuclei, and interaction with clouds

Professor Urs Baltensperger, Paul Scherrer Institute, Switzerland


Atmospheric aerosol particles are either directly emitted or formed by nucleation in the atmosphere after oxidation of precursor gases. Particles that have a size of ~50-100 nm can act as cloud condensation nuclei (CCN) by which they can modify cloud properties and therefore the radiative balance of the earth. Even though nucleation is responsible for a substantial fraction of CCN and despite extensive research, many questions remain about the dominant nucleation mechanisms, and a quantitative understanding of the dependence of the nucleation rate on the concentration of the nucleating substances such as gaseous sulfuric acid, ammonia, water vapor and others as well as of the possible role of galactic cosmic rays (GCR) has not been reached. This is of relevance for climate as the atmospheric concentrations of sulfuric acid, ammonia and other nucleating agents are strongly influenced by anthropogenic emissions.
The CLOUD (Cosmics Leaving OUtdoor Droplets) collaboration has been formed to elucidate the relevant processes and has provided substantial progress in recent years. These results as well as future plans to provide the required link between particle formation and climate forcing will be the topic of this presentation.

Show speakers

Session 2: Cryosphere - climate feedbacks

4 talks Show detail Hide detail

Polar amplification in the 21st Century

Professor Cecilia Bitz, University of Washington, USA


Recent surface warming averaged across the Arctic is amplified compared to global surface warming, but no such widespread polar amplification is observed yet in the Antarctic. Polar amplification results from a combination of meridional gradients in feedbacks, forcing, ocean heat uptake, and atmospheric heat transport. Feedbacks associated with reduced effective longwave cooling of cold surfaces and the relatively dry stable polar atmosphere have been found to be at least as important as ice-albedo feedback in the Arctic. The same is true for the Antarctic, yet subtle differences in profiles of temperature and humidity and cloud cover at the poles change the order of importance of the strongest feedbacks. Antarctic is delayed strong ocean heat uptake. Moreover, climate models have increased northward ocean heat transport toward both poles under global warming, taking away heat from the Southern Ocean and adding heat to the Arctic Ocean. Climate model projections are in broad agreement with the observed strong Arctic amplification and lack of Antarctic amplification and models only begin to exhibit Antarctic amplification in the 22nd century.

Show speakers

Ice sheet / climate interactions in the deep past

Professor Dan Lunt, University of Bristol, UK


This talk will focus on past climate and ice sheet (and associated sea level) changes in Earth’s ‘deep past’ (i.e. over timescales of millions of years), and the feedbacks and processes that operate between them. Exploring the links between climate and ice sheets through Earth’s history is important because (a) as inquisitive humans we have a desire to understand the world we live in and how it has changed through time, and (b) understanding of these links can give qualitative and quantitative information about possible future changes in Earth’s climate and sea level, and can therefore have societal benefit. 

We will review current knowledge of how climate and ice sheets have changed over the last 100 million years, using data derived from a variety of sources, including ocean sediments, coastal geological formations, and terrestrial fossil records. We will then discuss the various drivers (changes in tectonics, carbon-cycle, and sunlight) which have caused these changes, and their relative importance. Finally, we will explore the feedbacks and processes that operate between climate and sea level on these timescales, including presenting new modelling work on the influence of ice sheets on climate during the formation and expansion of the Antarctic ice sheet, ~35 million years ago.       

Show speakers

Interactions between continental ice sheets and the global ocean/atmosphere – exploring instability, drivers and feedbacks

Professor David Vaughan, British Antarctic Survey, UK


Our uncertainty surrounding future ice-sheet changes in Greenland and Antarctica remains a significant issue for future projections of sea-level and ocean circulation. Ice sheets, like glaciers, are systems that can achieve equilibrium states on surprisingly short timescales in stable ocean and atmospheric climates, but are also highly-sensitive to forcing by external changes. And because the annual fluxes of water and heat in and out of the ice sheets are globally significant, significant changes in these fluxes could impact the global climate system. Understanding the amplification or attenuation of the forcing on ice sheets that will occur through feedbacks between ice, atmosphere and ocean is vital if we are to project future ice-loss and its impact with confidence. In this talk, I examine the long-discussed potential instability in marine ice sheets which remains significant and controversial. This instability is a postulated internal positive feedback within the ice sheet, which might be initiated by external forcing, and which could lead to complete collapse of the portion of an ice sheet resting on rock below sea level. Recent observations suggest that such forcing is now present in some parts of Antarctica and Greenland; others indicate accelerating ice-loss from precisely those regions where the instability might be expected. Such observations may have clarified the question, neither these nor our models have yet provided a clear answer. I will examine the potential for positive and negative feedback loops between oceans and ice sheet, which have not yet been identified or explored. Feedbacks that could provide rate-limiting processes, or even avenues to stabilisation, of future ice-sheet change.

Show speakers

The inconstancy of transient climate sensitivity

Professor Jonathan Gregory, University of Reading and Met Office Hadley Centre, UK


Thanks to the large number of contributing models, and the inclusion of some new experiments, CMIP5 allows many questions relating to climate change projections to be addressed with greater clarity, including the quantification of CO2 radiative forcing, climate feedback and transient climate sensitivity (TCS, global mean surface warming per unit of radiative forcing in a time-dependent scenario). TCS depends on both climate feedback and ocean heat uptake; on the AOGCM average, about 1/3 of the radiative forcing is taken up by the subsurface ocean, thus mitigating the surface warming. Under constant CO2 concentration, the climate feedback parameter is remarkably constant, but decreases somewhat with time (i.e. effective climate sensitivity increases), due to changing SST patterns as warming proceeds. TCS under rising CO2 concentration also increases, because both the climate feedback parameter and the ocean heat uptake efficiency decline. The former could be due to a dependence on CO2 concentration, or to CO2 forcing increasing more rapidly than logarithmically with CO2 concentration. The latter is due to the penetration of heat to greater depths in the ocean. Relatively little is known in terms of physical processes about the reasons for the model spread, by a factor of two, in ocean heat uptake efficiency. Explosive volcanic eruptions cause much less global cooling than expected from the TCS for CO2 increase. This is mainly because ocean heat uptake is much more effective on short timescales. However climate feedback may also be larger for volcanic forcing (i.e. smaller climate sensitivity) than for CO2.

Show speakers

Session 3: Carbon cycle and greenhouse gas climate feedbacks

4 talks Show detail Hide detail

Carbon is for ever (almost): Regulation of CO2 and climate on geological time-scales

Professor Andy Ridgwell, University of Bristol, UK


Regulation of atmospheric CO2 (pCO2) and global temperature on million year time-scales is widely assumed to be driven primarily by global rates of rock weathering. If this were not the case and no long-term negative feedback on pCO2 existed, even relatively small changes in volcanic out-gassing would drive unbounded swings in pCO2 and climate – a situation that is not observed in the geological record. Yet proxy evidence for the link between pCO2, climate, and weathering is somewhat ambiguous. In this talk I’ll start by outlining the cascade of processes that act to remove excess CO2 in the atmosphere, focusing on the ultimate geological regulator of atmospheric CO2 – silicate weathering. I’ll introduce an interval of prominent and progressive warming that occurred during the early Cenozoic (ca. 58 to 49 Ma) alongside what existing evidence we have for how the global carbon cycle responded. I’ll then present a new global sediment dataset that on face value appears to suggest no change in the burial and geologic removal of the products of weathering and hence argues against a link between climate change and a weathering-mediated regulation of pCO2. I’ll finish by resolving the apparent contradiction between increases in pCO2 and weathering versus its sedimentary expression, through the use of a numerical representation of the global carbon cycle and climate – a tool for picking apart how often non-linear and opposing influences can combine to produce the phenomena we observe.

Show speakers

Permafrost thaw and its role as a carbon cycle feedback to global warming

Dr Charles Koven, Lawrence Berkeley National Lab, USA


The northern high latitudes contain the largest quantity of decomposable carbon in the terrestrial system, and are uniquely vulnerable to climate feedbacks because of arctic amplification and the abrupt nature of the freeze/thaw transition on physical and biogeochemical processes. These permafrost carbon stocks have built up over long periods of time due to a set of complex ecosystem and soil processes, and are likely to change dramatically with warming. This talk will focus on approaches to quantifying the magnitude and timing of the carbon cycle feedback arising from permafrost soils. We will begin by describing the observed distributions of carbon in permafrost soils and where the greatest uncertainties in the amount of permafrost carbon stocks lie. Next the talk will examine climate model projections of physical permafrost thaw, and discuss approaches to using these projections of changing soil temperatures in combination with observations of soil carbon distributions and laboratory-measured soil carbon losses upon thaw to estimate the likely carbon cycle responses to this warming. Building on these approaches, the talk will discuss projections of the amount and timing of permafrost C emissions as predicted by terrestrial carbon cycle models that include a more complex set of interactions, such as increased vegetation productivity due to both warming and the release of nutrients that are bound up in permafrost soils. Together, these approaches point to a potentially large feedback from permafrost on a centennial timescale, which must be accounted for in long-term carbon planning.

Show speakers

PMIP3: paleo view on feedbacks

Dr Pascale Braconnot, Laboratoire des Sciences du Climat et de l'Environnement, France


The climate response to anthropogenic activities involves physical processes, feedbacks and mechanisms that are the same as the ones that contributed to modulate the Earth’s climate in the past in response to external perturbations induced by solar variations, volcanic eruptions or tectonics. Past climates therefore offer a large set of natural experiences that can be used to better understand the relative role of different climate feedbacks arising from changes in the Earth’s global energetics and Earth’s hydrological cycle or from the coupling between climate and biogeochemical cycles. In addition, the numerous climate reconstructions from different and independent ice, marine and terrestrial climate archives allow us to test how climate models reproduce past changes, and to assess their credibility when used for future climate projections. The Paleoclimate Intercomparison Projects ( was settled in 1991 with these objectives in mind. Using the results of simulations of the mid-Holocene and of the Last Glacial maximum, I will discuss the evolution in the vision these simulations provide on climate sensitivity, cryosphere feedbacks, and the role of ocean and vegetation feedbacks in monsoon regions. These two periods represent key references for model evaluation in PMIP and were included as part of the large CMIP5 set of simulations ( used as reference in last IPCC assessment (IPCC AR5, 2013). I will also highlight some of the current difficulties in the analyses resulting from model biases, and point out new possibilities resulting from the online simulation of the carbon cycle in climate models or from the comparison with other climatic periods considered in PMIP3.

Show speakers

Carbon cycle feedbacks and future climate change

Professor Pierre Friedlingstein, University of Exeter, UK


Climate and the carbon cycle are interacting on every timescale. On short, inter-annual timescales, there are numerous observational evidences revealing the strong response of the carbon cycle to climate variability. During warm El Niño years, atmospheric CO2 shows larger than average growth rate, indicating reduced storage in land and/or oceans; the opposite being observed during La Niña years. Multiple lines of evidence point towards tropical land ecosystems as main drivers of this variability. On such short time scales, the ocean shows much lower variability in its carbon exchange with the atmosphere.

On multi-millennial timescales, such as over glacial-interglacial cycles, ice core data clearly shows a strong correlation between climate and atmospheric CO2, with again, warm/cold climate being associated to higher/lower atmospheric CO2, i.e. lower/larger storage in ocean and land. Here, the ocean, probably the Southern ocean, is the main culprit for these changes.

On the centennial timescale of interest for the anthropogenic perturbation, there are indications of similar behaviour during warm/cold periods over the last millennium but no direct observations over the historical period. This is primarily due to the unprecedented input of CO2 in the atmosphere due to fossil fuel burning and land use change that dwarves any natural response of the land and ocean to the current warming. However, modelling studies unanimously show, again, a reduction of carbon storage both on land and ocean due to global warming. This induces, as during glacial cycles, a positive feedback in the climate system, a warmer world leading to larger atmospheric CO2 concentration.

The talk will review the observational and modelling evidence of a positive feedback between the climate system and the global carbon cycle, highlighting the implications for 21st century warming and cumulative emissions compatible with a given climate target.

Show speakers

Session 4: Feedbacks, uncertainty and risk

3 talks Show detail Hide detail

Global impacts of arctic feedbacks, and the value of better information

Dr Chris Hope, University of Cambridge, UK


The arctic is warming roughly twice as fast as the globe as a whole. If greenhouse gas emissions continue to increase at current rates, this warming will lead to the release into the atmosphere of hundreds of billions of tonnes of CO2 and billions of tonnes of CH4 from the widespread thawing of permafrost on land and under the sea bed. Results from the PAGE09 integrated assessment model show that the CO2 and CH4 released from thawing permafrost will increase the mean net present value (NPV) of the global impacts of climate change by tens of trillions of dollars. Much effort has been devoted to improving our understanding of climate, and how it might change with increasing concentrations of greenhouse gases in the atmosphere. Runs with PAGE09 show that approximately halving the uncertainty in the transient climate response has a mean NPV of about $10 trillion, provided climate policies respond appropriately to this better information. These results suggest that urgent action to reduce the risks of thawing permafrost in the Arctic should be considered, and justify substantial further research efforts into many aspects of the climate.

Show speakers

Observational constraints on the net effect of climate feedbacks on future climate change

Dr Nathan Gillett, Canadian Centre for Climate Modelling and Analysis, Canada


Interest in climate feedbacks in both the physical climate system and in the carbon cycle is primarily motivated by their role in controlling future climate change, and by concerns over whether these feedbacks and their uncertainties are well-represented in ensembles of climate models used to make climate projections. An alternative approach to considering feedbacks individually is to use observations of global temperature changes to constrain the overall response to climate forcings, scaling model-simulated responses to greenhouse gas and aerosol changes up or down in order to best match observations of historical climate changes, and applying the same scaling to projected future changes. Such an approach can also be used to derive quantified uncertainty estimates in projections. Perfect model studies indicate that this approach works well for near-term projections under conditions of progressively increasing radiative forcing, but works less well under conditions of stabilised or declining radiative forcing. In addition, carbon cycle feedbacks, not considered in such an approach, become more important for longer-term projections. These limitations may be addressed by making use of an emergent proportionality between cumulative carbon dioxide emissions and the resultant global mean warming found across a range of earth system models and for a wide range of emissions pathways. By constraining the ratio of global warming to cumulative carbon emissions from observations, thereby constraining the net effects of both physical climate and carbon cycle feedbacks, observational constraints may be used to inform long-term climate projections, even under conditions of very high or low cumulative emissions.

Show speakers

What does global mean temperature tell us about local climate?

Professor Rowan Sutton, University of Reading, UK


The subject of climate feedbacks focusses attention on global mean surface air temperature as the key metric of climate change. But what does knowledge of global mean temperature tell us about the climate of specific regions? In the context of the ongoing UNFCCC process, this is an important question for policy makers as well as for scientists. The answer depends on many factors including: the mechanisms causing changes, the timescale of the changes, and the variables and regions of interest. In this talk we will focus especially on decadal timescales, which are of particular interest in relation to recent and near-future anthropogenic climate change. By analysing simulations with multiple climate models we will examine the respective roles of natural internal variability and forced changes in shaping the relationship between local climate (surface air temperature and precipitation) and global mean surface temperature, a relationship which is strongly influenced by climate feedbacks. The extent to which different models give consistent or differing results will be a particular focus. We will also discuss how the results from climate models inform the interpretation of observed changes in climate over the instrumental period.

Show speakers
Feedbacks on climate in the Earth system The Royal Society, London 6-9 Carlton House Terrace London SW1Y 5AG UK