Ryan J. Kramer

Aerosols offset greenhouse-gas warming, but their radiative forcing is often treated as a property of the aerosol perturbation, without accounting for the climate state in which it occurs. Here we show that this assumption is incomplete using atmosphere-only simulations with two GFDL atmospheric model configurations, GFDL-AM4.0 and GFDL-AM4-MG2, across climates from 8 K colder to 8 K warmer than present day. In both models, the same present-day-minus-preindustrial aerosol-emission perturbation produces weaker cooling as climate warms. This weakening arises from changes in both aerosol–radiation and aerosol–cloud interactions. Direct aerosol forcing weakens as aerosol optical depth decreases, reflecting changes in aerosol burden and the optical effectiveness of aerosol mass. Aerosol–cloud forcing weakens through model-dependent mechanisms: shortwave liquid-cloud adjustments dominate in GFDL-AM4.0, consistent with reduced cloud susceptibility in warmer climates, whereas GFDL-AM4-MG2 shows an additional longwave pathway involving aerosol–cloud–ice interactions and high-cloud responses. Aerosol forcing also weakens as climate feedbacks become less negative, revealing consistent within-model co-variation of forcing and feedback. These results highlight the importance of accounting for background climate state when interpreting the role of aerosol forcing in historical attribution, climate model tuning, and future climate projections.

Aerosol effective radiative forcing (ERF), particularly its aerosol–cloud interaction component (ERFaci), is commonly diagnosed using decomposition methods such as online double-call, approximate partial radiative perturbation (APRP), radiative kernels, and related proxy diagnostics. These approaches differ in whether they rely on aerosol-free or aerosol-inclusive formulations and in whether they account for cloud masking of non-cloud adjustments and, where relevant, direct aerosol effects. Here we identify a systematic bias in ERFaci estimation arising from aerosol vertical structure. When aerosols reside above clouds, they attenuate incoming shortwave radiation before it reaches cloud top, thereby reducing the radiative impact of a given cloud perturbation. As a result, methods based on aerosol-free radiative sensitivities overestimate the shortwave component of ERFaci, with the bias amplified in colder climate states where aerosol-over-cloud configurations are more prevalent. Thus, an apparent weakening of shortwave ERFaci with warming may reflect not only a physical change in cloud response but also a reduction in aerosol-over-cloud bias as the background climate changes. In addition, neglect of cloud masking can misattribute direct aerosol effects and non-cloud adjustments. Because aerosol forcing is partitioned into direct and cloud-mediated components, this bias also affects the inferred aerosol–radiation interaction component. Simulations across multiple climate states show that diagnosed ERF partitions exhibit apparent state-dependence arising from both methodological biases and physical changes. These results highlight a broader limitation of decomposition frameworks that neglect aerosol vertical structure and cloud masking, and they underscore the need for cautious interpretation of ERF partitioning.

Aerosols offset greenhouse-gas warming, but their radiative forcing is often treated as a property of the aerosol perturbation, without accounting for the climate state in which it occurs. Here we show that this assumption is incomplete using atmosphere-only simulations with two GFDL atmospheric model configurations, GFDL-AM4.0 and GFDL-AM4-MG2, across climates from 8 K colder to 8 K warmer than present day. In both models, the same present-day-minus-preindustrial aerosol-emission perturbation produces weaker cooling as climate warms. This weakening arises from changes in both aerosol–radiation and aerosol–cloud interactions. Direct aerosol forcing weakens as aerosol optical depth decreases, reflecting changes in aerosol burden and the optical effectiveness of aerosol mass. Aerosol–cloud forcing weakens through model-dependent mechanisms: shortwave liquid-cloud adjustments dominate in GFDL-AM4.0, consistent with reduced cloud susceptibility in warmer climates, whereas GFDL-AM4-MG2 shows an additional longwave pathway involving aerosol–cloud–ice interactions and high-cloud responses. Aerosol forcing also weakens as climate feedbacks become less negative, revealing consistent within-model co-variation of forcing and feedback. These results highlight the importance of accounting for background climate state when interpreting the role of aerosol forcing in historical attribution, climate model tuning, and future climate projections.

Calculating equilibrium climate sensitivity (ECS) requires simulations of a thousand years or more for the climate to re‐establish equilibrium, but most ECS estimates are derived from shorter model integrations. The widely used “Gregory method” for such estimates systematically underestimates ECS, a bias previously attributed to the time‐dependence of climate feedbacks. Here we demonstrate this underestimate instead arises primarily from an underestimate of radiative forcing. We present a modified method that corrects for the underestimate in forcing, producing ECS estimates that closely align with results from “long‐run” millennium‐scale simulations. This method is robust across models and offers an easily obtainable estimate of ECS with significantly higher accuracy. It enables reliable ECS estimates for full CMIP model ensembles, including models without millennium‐scale simulations. Using this method, we show one of every three CMIP6 climate models has an ECS greater than 5K—the upper bound of the “very likely” range in prior assessments. To further verify this method, we introduce an “integrated feedback parameter” which provides a more physically‐meaningful way to average a time‐evolving climate sensitivity by weighting the differential feedback in consecutive, non‐overlapping time segments by the fraction of energy imbalance that each segment restores. This parameter agrees well with that derived from our modified method, lending further credibility to this simplified approach.

Forcing together with radiative responses and natural variability has sharply increased Earth’s energy imbalance (EEI), a key driver of climate change, between 2003 and 2023. Declining anthropogenic aerosol emissions and increasing greenhouse gas concentrations are the leading contributors to the forcing. However, observation-based and model-based forcing estimates disagree on the magnitude of aerosol forcing, particularly for effective radiative forcing from aerosol-cloud interactions (ERF _ACI ). Here, we show that two independent observation-based estimates of forcing trends achieve unexpectedly good closure between 2003 and 2023. The corresponding model-based trend is within the uncertainty range of observations but substantially smaller, with most of the difference due to models’ underestimation of cloud fraction adjustment to aerosols. Our results suggest a prominent role of aerosols in the recent EEI increase and stronger total historical ERF _ACI and higher equilibrium climate sensitivity than the current consensus, with a likely range between 2.8° and 4.8°C.

Effective climate sensitivity (E _ff CS) is a key metric for quantifying climate change, yet its uncertainty remains large, primarily due to intermodel differences in cloud feedback. Here, we assess E _ff CS and climate feedback to an abrupt doubling (2xCO2) and quadrupling (4xCO2) of CO ₂ over 450 years in GFDL CM4.0. Both runs have comparable E _ff CS estimates during the first 50 years (2xCO2: 2.99 K; 4xCO2: 3.25 K) with the 4xCO2 E _ff CS estimate still slightly stronger for Years 51–150. Extending the analysis to 450 years increases the E _ff CS estimate, but more so in 2xCO2 (5.8 K) than in 4xCO2 (4.79 K). In both runs, the E _ff CS increases due to a weakening of the net feedback with warming, largely through shortwave cloud feedback increase. A weaker E _ff CS during the slow response for 4xCO2 relative to 2xCO2 is associated with a weaker surface albedo feedback due to an earlier sea‐ice loss in both hemispheres.

An external perturbation to the climate system from anthropogenic or natural activity first impacts the climate by inducing an alteration to Earth's energy budget, known as a radiative forcing. The characteristics of the radiative forcing, such as its global-mean magnitude and spatial pattern, determine the subsequent climate response. Therefore, understanding projections of climate change first requires diagnosing radiative forcing and evaluating its persistent uncertainty in Global Climate Models. As part of the Coupled Model Intercomparison Project phase 7 (CMIP7), the second iteration of the Radiative Forcing Model Intercomparison Project (RFMIP2.0) will enable the systematic characterization of effective radiative forcing and its components across state-of-the-art climate models, through a set of fixed-sea-surface-temperature timeslice and transient experiments. The protocol for RFMIP2.0, introduced here, will in part serve as a continuity and an expansion of core RFMIP experiments first introduced in CMIP6, some of which have now been incorporated into the overarching CMIP7 DECK and Assessment Fast Track protocols given their broad utility. This will allow for a consistent estimate for radiative forcing across multiple model generations, which is valuable for model evaluation and future development. RFMIP2.0 also includes new experiments that will address open questions about the definition of radiative forcing, such as its sensitivity to evolving surface conditions, and will further enhance an ever-growing swath of science applications that rely on an understanding of Earth's energy budget.

The availability of up-to-date forcing datasets is critical for reconciling Earth system model simulations with contemporary observations. With the newly released CMIP7 forcing dataset, which extends the historical record through 2021, this study provides the first systematic evaluation of how using updated forcings over the “middle-of-the-road” SSP2-4.5 scenario forcings influences the simulated radiation budget. The updated forcing dataset used in the atmosphere model does not improve the agreement between the simulated global-mean top-of-the-atmosphere radiation trends and the satellite-observed trends. Regionally, we identify pronounced shifts in radiative energy partitioning between atmosphere and surface. Specifically, during 2001 ~ 2021, accelerated reductions in sulfate and black carbon aerosols over East Asia drive a significant decline in atmospheric solar absorption that matches observations, while a substantial increase in organic aerosols at higher latitudes induces a localized increase in atmospheric absorption that does not match observations. The substantial influence of localized forcing changes on simulated regional climate underscores the necessity for routine updates of forcing datasets.

Earth’s energy imbalance (EEI), a key driver of climate change, has risen markedly over the last two decades. Greenhouse gas forcing, aerosol forcing, and cloud feedback all contribute to this increase. However, the role of aerosol forcing, particularly effective radiative forcing through aerosol-cloud interactions (ERF_ACI), remains highly uncertain and closely intertwined with cloud feedback. Here, we estimate shortwave ERF_ACI using satellite observations and show its importance as a contributor to the EEI increase between 2003 and 2023. ERF_ACI exhibits a significant increasing trend of 0.33 Wm⁻²/decade averaged over oceans between 60^∘S and 60^∘N. The increasing trend of ERF_ACI stems from a global decline in cloud droplet number concentration driven by decreasing anthropogenic aerosol emissions. It is of similar magnitude as the combined instantaneous forcing trend from greenhouse gases and aerosol-radiation interactions. Our results may reduce the gap between the simulated and observed shortwave contribution to EEI trends while implying a weak shortwave cloud feedback. They also suggest a stronger aerosol forcing than the multi-model mean.

This paper presents a comprehensive overview of the Coupled Model Intercomparison Project Phase 7 (CMIP7) request for data unlocking key research avenues in atmospheric science and provides justification for the resources needed to produce this data. Topics within the CMIP7 Atmosphere Theme centre around processes and feedbacks in atmospheric science such as clouds, aerosols and atmospheric chemistry, atmospheric circulation, temperature variability and extremes, radiative forcings, and Earth system model evaluation. These topics are summarised in this paper as scientific “opportunities” which will be realised through CMIP7 experiments and Earth system model outputs. These opportunities were submitted by a thematic group of atmospheric science community representatives combined with an extended consultation process. The production of these variables will close key gaps and uncertainties identified during previous rounds of CMIP, and will be broadly used by scientific, policy, governmental, industry, and other communities that rely on climate model projections for research and decision making, including supporting the 7th Intergovernmental Panel on Climate Change Assessment Report (AR7). As an author group, we also reflect on the process used to collate this data request and make recommendations to future CMIP governance on implementing a consultation on this scale in the future.