Differences among climate models in equilibrium climate sensitivity (ECS; the equilibrium surface temperature response to a doubling of atmospheric CO₂) remain a significant barrier to the accurate assessment of societally important impacts of climate change. Relationships between ECS and observable metrics of the current climate in model ensembles, so-called emergent constraints, have been used to constrain ECS. Here a statistical method (including a backward selection process) is employed to achieve a better statistical understanding of the connections between four recently proposed emergent constraint metrics and individual feedbacks influencing ECS. The relationship between each metric and ECS is largely attributable to a statistical connection with shortwave low cloud feedback, the leading cause of intermodel ECS spread. This result bolsters confidence in some of the metrics, which had assumed such a connection in the first place. Additional analysis is conducted with a few thousand artificial metrics that are randomly generated but are well correlated with ECS. The relationships between the contrived metrics and ECS can also be linked statistically to shortwave cloud feedback. Thus, any proposed or forthcoming ECS constraint based on the current generation of climate models should be viewed as a potential constraint on shortwave cloud feedback, and physical links with that feedback should be investigated to verify that the constraint is real. In addition, any proposed ECS constraint should not be taken at face value since other factors influencing ECS besides shortwave cloud feedback could be systematically biased in the models.

Snow albedo feedback (SAF) behaves similarly in the current and future climate contexts; thus, constraining the large intermodel variance in SAF will likely reduce uncertainty in climate projections. To better understand this intermodel spread, structural and parametric biases contributing to SAF variability are investigated. We find that structurally varying snowpack, vegetation, and albedo parameterizations drive most of the spread, while differences arising from model parameters are generally smaller. Models with the largest SAF biases exhibit clear structural or parametric errors. Additionally, despite widespread intermodel similarities, model interdependency has little impact on the strength of the relationship between SAF in the current and future climate contexts. Furthermore, many models now feature a more realistic SAF than in the prior generation, but shortcomings from two models limit the reduction in ensemble spread. Lastly, preliminary signs from ongoing model development are positive and suggest a likely reduction in SAF spread among upcoming models.

A highly uncertain aspect of anthropogenic climate change is the rate at which the global hydrologic cycle intensifies. The future change in global‐mean precipitation per degree warming, or hydrologic sensitivity, exhibits a threefold spread (1–3%/K) in current global climate models. In this study, we find that the intermodel spread in this value is associated with a significant portion of variability in future projections of extreme precipitation in the tropics, extending also into subtropical atmospheric river corridors. Additionally, there is a very tight intermodel relationship between changes in extreme and nonextreme precipitation, whereby models compensate for increasing extreme precipitation events by decreasing weak‐moderate events. Another factor linked to changes in precipitation extremes is model resolution, with higher resolution models showing a larger increase in heavy extremes. These results highlight ways various aspects of hydrologic cycle intensification are linked in models and shed new light on the task of constraining precipitation extremes.

Emergent constraints use relationships between future and current climate states to constrain projections of climate response. Here we introduce a statistical, hierarchical emergent constraint (HEC) framework in order to link future and current climates with observations. Under Gaussian assumptions, the mean and variance of the future state are shown analytically to be a function of the signal‐to‐noise ratio between current climate uncertainty and observation error and the correlation between future and current climate states. We apply the HEC to the climate change, snow‐albedo feedback, which is related to the seasonal cycle in the Northern Hemisphere. We obtain a snow‐albedo feedback prediction interval of (−1.25,−0.58)%/K. The critical dependence on signal‐to‐noise ratio and correlation shows that neglecting these terms can lead to bias and underestimated uncertainty in constrained projections. The flexibility of using HEC under general assumptions throughout the Earth system is discussed.

How tropical low clouds change with climate remains the dominant source of uncertainty in global warming projections. An analysis of an ensemble of CMIP5 climate models reveals that a significant part of the spread in the models’ climate sensitivity can be accounted by differences in the climatological shallowness of tropical low clouds in weak-subsidence regimes: models with shallower low clouds in weak-subsidence regimes tend to have a higher climate sensitivity than models with deeper low clouds. The dynamical mechanisms responsible for the model differences are analyzed. Competing effects of parameterized boundary-layer turbulence and shallow convection are found to be essential. Boundary-layer turbulence and shallow convection are typically represented by distinct parameterization schemes in current models—parameterization schemes that often produce opposing effects on low clouds. Convective drying of the boundary layer tends to deepen low clouds and reduce the cloud fraction at the lowest levels; turbulent moistening tends to make low clouds more shallow but affects the low-cloud fraction less. The relative importance different models assign to these opposing mechanisms contributes to the spread of the climatological shallowness of low clouds and thus to the spread of low-cloud changes under global warming.

In the current generation of climate models, the projected increase in global precipitation over the 21st century ranges from 2% to 10% under a high‐emission scenario. Some of this uncertainty can be traced to the rapid response to carbon dioxide (CO₂) forcing. We analyze an ensemble of simulations to better understand model spread in this rapid response. A substantial amount is linked to how the land surface partitions a change in latent versus sensible heat flux in response to the CO₂‐induced radiative perturbation; a larger increase in sensible heat results in a larger decrease in global precipitation. Model differences in the land surface response appear to be strongly related to the vegetation response to increased CO₂, specifically, the closure of leaf stomata. Future research should thus focus on evaluation of the vegetation physiological response, including stomatal conductance parameterizations, for the purpose of constraining the fast response of Earth's hydrologic cycle to CO₂ forcing.

We examine the tropical inversion strength, measured by the estimated inversion strength (EIS), and its response to climate change in 18 models associated with phase 5 of the coupled model intercomparison project (CMIP5). While CMIP5 models generally capture the geographic distribution of observed EIS, they systematically underestimate it off the west coasts of continents, due to a warm bias in sea surface temperature. The negative EIS bias may contribute to the low bias in tropical low-cloud cover in the same models. Idealized perturbation experiments reveal that anthropogenic forcing leads directly to EIS increases, independent of “temperature-mediated” EIS increases associated with long-term oceanic warming. This fast EIS response to anthropogenic forcing is strongly impacted by nearly instantaneous continental warming. The temperature-mediated EIS change has contributions from both uniform and non-uniform oceanic warming. The substantial EIS increases in uniform oceanic warming simulations are due to warming with height exceeding the moist adiabatic lapse rate in tropical warm pools. EIS also increases in fully-coupled ocean–atmosphere simulations where CO2CO2 concentration is instantaneously quadrupled, due to both fast and temperature-mediated changes. The temperature-mediated EIS change varies with tropical warming in a nonlinear fashion: The EIS change per degree tropical warming is much larger in the early stage of the simulations than in the late stage, due to delayed warming in the eastern parts of the subtropical oceans. Given the importance of EIS in regulating tropical low-cloud cover, this suggests that the tropical low-cloud feedback may also be nonlinear.

Differences in simulations of tropical marine low‐cloud cover (LCC) feedback are sources of significant spread in temperature responses of climate models to anthropogenic forcing. Here we show that in models the feedback is mainly driven by three large‐scale changes—a strengthening tropical inversion, increasing surface latent heat flux, and an increasing vertical moisture gradient. Variations in the LCC response to these changes alone account for most of the spread in model‐projected 21st century LCC changes. A methodology is devised to constrain the LCC response observationally using sea surface temperature (SST) as a surrogate for the latent heat flux and moisture gradient. In models where the current climate's LCC sensitivities to inversion strength and SST variations are consistent with observed, LCC decreases systematically, which would increase absorption of solar radiation. These results support a positive LCC feedback. Correcting biases in the sensitivities will be an important step toward more credible simulation of cloud feedbacks.

Intensification of the hydrologic cycle is a key dimension of climate change, with substantial impacts on human and natural systems^1,2. A basic measure of hydrologic cycle intensification is the increase in global-mean precipitation per unit surface warming, which varies by a factor of three in current-generation climate models (about 1–3 per cent per kelvin)^3,4,5. Part of the uncertainty may originate from atmosphere–radiation interactions. As the climate warms, increases in shortwave absorption from atmospheric moistening will suppress the precipitation increase. This occurs through a reduction of the latent heating increase required to maintain a balanced atmospheric energy budget^6,7. Using an ensemble of climate models, here we show that such models tend to underestimate the sensitivity of solar absorption to variations in atmospheric water vapour, leading to an underestimation in the shortwave absorption increase and an overestimation in the precipitation increase. This sensitivity also varies considerably among models due to differences in radiative transfer parameterizations, explaining a substantial portion of model spread in the precipitation response. Consequently, attaining accurate shortwave absorption responses through improvements to the radiative transfer schemes could reduce the spread in the predicted global precipitation increase per degree warming for the end of the twenty-first century by about 35 per cent, and reduce the estimated ensemble-mean increase in this quantity by almost 40 per cent.

Snow-albedo feedback (SAF) is examined in 25 climate change simulations participating in the Coupled Model Intercomparison Project version 5 (CMIP5). SAF behavior is compared to the feedback’s behavior in the previous (CMIP3) generation of global models. SAF strength exhibits a fivefold spread across CMIP5 models, ranging from 0.03 to 0.16 W m⁻² K⁻¹ (ensemble-mean = 0.08 W m⁻² K⁻¹). This accounts for much of the spread in 21st century warming of Northern Hemisphere land masses, and is very similar to the spread found in CMIP3 models. As with the CMIP3 models, there is a high degree of correspondence between the magnitudes of seasonal cycle and climate change versions of the feedback. Here we also show that their geographical footprint is similar. The ensemble-mean SAF strength is close to an observed estimate of the real climate’s seasonal cycle feedback strength. SAF strength is strongly correlated with the climatological surface albedo when the ground is covered by snow. The inter-model variation in this quantity is surprisingly large, ranging from 0.39 to 0.75. Models with large surface albedo when these regions are snow-covered will also have a large surface albedo contrast between snow-covered and snow-free regions, and therefore a correspondingly large SAF. Widely-varying treatments of vegetation masking of snow-covered surfaces are probably responsible for the spread in surface albedo where snow occurs, and the persistent spread in SAF in global climate models.

In 36 climate change simulations associated with phases 3 and 5 of the Coupled Model Intercomparison Project (CMIP3 and CMIP5), changes in marine low cloud cover (LCC) exhibit a large spread, and may be either positive or negative. Here we develop a heuristic model to understand the source of the spread. The model’s premise is that simulated LCC changes can be interpreted as a linear combination of contributions from factors shaping the clouds’ large-scale environment. We focus primarily on two factors—the strength of the inversion capping the atmospheric boundary layer (measured by the estimated inversion strength, EIS) and sea surface temperature (SST). For a given global model, the respective contributions of EIS and SST are computed. This is done by multiplying (1) the current-climate’s sensitivity of LCC to EIS or SST variations, by (2) the climate-change signal in EIS or SST. The remaining LCC changes are then attributed to changes in greenhouse gas and aerosol concentrations, and other environmental factors. The heuristic model is remarkably skillful. Its SST term dominates, accounting for nearly two-thirds of the intermodel variance of LCC changes in CMIP3 models, and about half in CMIP5 models. Of the two factors governing the SST term (the SST increase and the sensitivity of LCC to SST perturbations), the SST sensitivity drives the spread in the SST term and hence the spread in the overall LCC changes. This sensitivity varies a great deal from model to model and is strongly linked to the types of cloud and boundary layer parameterizations used in the models. EIS and SST sensitivities are also estimated using observational cloud and meteorological data. The observed sensitivities are generally consistent with the majority of models as well as expectations from prior research. Based on the observed sensitivities and the relative magnitudes of simulated EIS and SST changes (which we argue are also physically reasonable), the heuristic model predicts LCC will decrease over the 21st-century. However, to place a strong constraint, for example on the magnitude of the LCC decrease, will require longer observational records and a careful assessment of other environmental factors producing LCC changes. Meanwhile, addressing biases in simulated EIS and SST sensitivities will clearly be an important step towards reducing intermodel spread in simulated LCC changes.

Pithan and Mauritsen argue that the 2009 results of Boé et al. are not consistent with current understanding of the lapse-rate feedback in the Arctic. They also argue that these results arise to an important extent from self-correlation issues. In this response, the authors argue that their results are not inconsistent with current understanding of lapse-rate feedback and demonstrate that the conclusions remain unchanged when all possibilities of self-correlation are excluded.

In this study, we examine marine low cloud cover variability in the Southeast Pacific and its association with lower-tropospheric stability (LTS) across a spectrum of timescales. On both daily and interannual timescales, LTS and low cloud amount are very well correlated in austral summer (DJF). Meanwhile in winter (JJA), when ambient LTS increases, the LTS–low cloud relationship substantially weakens. The DJF LTS–low cloud relationship also weakens in years with unusually large ambient LTS values. These are generally strong El Niño years, in which DJF LTS values are comparable to those typically found in JJA. Thus the LTS–low cloud relationship is strongly modulated by the seasonal cycle and the ENSO phenomenon. We also investigate the origin of LTS anomalies closely associated with low cloud variability during austral summer. We find that the ocean and atmosphere are independently involved in generating anomalies in LTS and hence variability in the Southeast Pacific low cloud deck. This highlights the importance of the physical (as opposed to chemical) component of the climate system in generating internal variability in low cloud cover. It also illustrates the coupled nature of the climate system in this region, and raises the possibility of cloud feedbacks related to LTS. We conclude by addressing the implications of the LTS–low cloud relationship in the Southeast Pacific for low cloud feedbacks in anthropogenic climate change.

We show that intermodel variations in the anthropogenically-forced evolution of September sea ice extent (SSIE) in the Arctic stem mainly from two factors: the baseline climatological sea ice thickness (SIT) distribution, and the local climate feedback parameter. The roles of these two factors evolve over the course of the twenty-first century. The SIT distribution is the most important factor in current trends and those of coming decades, accounting for roughly half the intermodel variations in SSIE trends. Then, its role progressively decreases, so that around the middle of the twenty-first century the local climate feedback parameter becomes the dominant factor. Through this analysis, we identify the investments in improved simulation of Arctic climate necessary to reduce uncertainties both in projections of sea ice loss over the coming decades and in the ultimate fate of the ice pack.

Observational studies have shown that mesoscale variations in sea surface temperature may induce mesoscale variations in wind. In eastern subtropical upwelling regions such as the California coast, this mechanism could be of great importance for the mean state and variability of the climate system. In coastal regions orography also creates mesoscale variations in wind, and the orographic effect may extend more than 100 km offshore. The respective roles of SST/wind links and coastal orography in shaping mesoscale wind variations in nearshore regions is not clear. We address this question in the context of the California Upwelling System, using a high-resolution regional numerical modeling system coupling the WRF atmospheric model to the ROMS oceanic model, as well as additional uncoupled experiments to quantify and separate the effects of SST/wind links and coastal orography on mesoscale wind variations. After taking into account potential biases in the representation of the strength of SST/wind links by the model, our results suggest that the magnitude of mesoscale wind variations arising from the orographic effects is roughly twice that of wind variations associated with mesoscale SST anomalies. This indicates that even in this region where coastal orography is complex and leaves a strong imprint on coastal winds, the role of SST/winds links in shaping coastal circulation and climate cannot be neglected.

The Antarctic Peninsula has warmed significantly since the 1950s. This pronounced and isolated warming trend is collectively captured by 29 twentieth-century climate hindcasts participating in the version 3 Coupled Model Intercomparison Project. To understand the factors driving warming trends in the hindcasts, we examine trends in Peninsula region’s atmospheric heat budget in every simulation. We find that atmospheric latent heat release increases in nearly all hindcasts. These increases are generally anthropogenic in origin, and account for about 60% of the ensemble-mean warming trend in the Peninsula. They are driven primarily by well-understood features of the anthropogenic intensification of global hydrological cycle. As sea surface temperature increases, moisture contained in atmospheric flows increases. When such flows are forced to ascend the Peninsula’s topography, enhanced local latent heat release results. The mechanism driving the warming of the Antarctic Peninsula is therefore clear in the models. Evidence for a similar mechanism operating in the real world is seen in the increasing snow accumulation rates inferred from ice cores drilled in the Peninsula. However, the relative importance of this mechanism and other processes previously identified as potentially causing the observed warming, such as the recent sea ice retreat in the Bellingshausen Sea, is difficult to assess. Thus the relevance of the simulated warming mechanism to the observed warming is unclear, in spite of its robustness in the models.

The large spread of the response to anthropogenic forcing simulated by state-of-the-art climate models in the Arctic is investigated. A feedback analysis framework specific to the Arctic is developed to address this issue. The feedback analysis shows that a large part of the spread of Arctic climate change is explained by the longwave feedback parameter. The large spread of the negative longwave feedback parameter is in turn mainly due to variations in temperature feedback. The vertical temperature structure of the atmosphere in the Arctic, characterized by a surface inversion during wintertime, exerts a strong control on the temperature feedback and consequently on simulated Arctic climate change. Most current climate models likely overestimate the climatological strength of the inversion, leading to excessive negative longwave feedback. The authors conclude that the models’ near-equilibrium response to anthropogenic forcing is generally too small.

Climate change is expected to cause a reduction in the spatial extent of snow cover on land. Recent work suggests that this will exert a local influence on the atmosphere and the hydrology of snow‐margin areas through the snow‐albedo feedback (SAF) mechanism. A significant fraction of variability among IPCC AR4 general circulation model (GCM) predictions for future summertime climate change over these areas is related to the models' representation of springtime SAF. In this study, we demonstrate a nonlocal influence of SAF on the summertime circulation in the extratropical Northern Hemisphere. Increased land surface warming in models with stronger SAF is associated with large‐scale sea‐level pressure anomalies over the northern oceans and a poleward intensified subtropical jet. We find that up to 25–30% and, on average, 5–10% of the inter‐model spread in projections of the circulation response to climate change is linearly related to SAF strength.

The Arctic climate is changing rapidly¹. From 1979 to 2006, September sea-ice extent decreased by almost 25% or about 100,000 km² per year (ref. 2). In September 2007, Arctic sea-ice extent reached its lowest level since satellite observations began³and in September 2008, sea-ice cover was still low. This development has raised concerns that the Arctic Ocean could be ice-free in late summer in only a few decades, with important economic and geopolitical implications. Unfortunately, most current climate models underestimate significantly the observed trend in Arctic sea-ice decline⁴, leading to doubts regarding their projections for the timing of ice-free conditions. Here we analyse the simulated trends in past sea-ice cover in 18 state-of-art-climate models and find a direct relationship between the simulated evolution of September sea-ice cover over the twenty-first century and the magnitude of past trends in sea-ice cover. Using this relationship together with observed trends, we project the evolution of September sea-ice cover over the twenty-first century. We find that under a scenario with medium future greenhouse-gas emissions, the Arctic Ocean will probably be ice-free in September before the end of the twenty-first century.

Two main mechanisms can potentially explain the spread in the magnitude of global warming simulated by climate models: deep ocean heat uptake and climate feedbacks. Here, we show that deep oceanic heat uptake is a major source of spread in simulations of 21st century climate change. Models with deeper baseline polar mixed layers are associated with larger deep ocean warming and smaller global surface warming. Based on this result, we set forth an observational constraint on polar vertical oceanic mixing. This constraint suggests that many models may overestimate the efficiency of polar oceanic mixing and therefore may underestimate future surface warming. Thus to reduce climate change uncertainties at time‐scales relevant for policy‐making, improved understanding and modelling of oceanic mixing at high latitudes is crucial.