Maria Rugenstein’s research while affiliated with Colorado State University and other places

Plain Language Summary Ideally, we could use information from how radiation depends on global‐mean surface temperature and its spatial patterns over the last decades to predict radiation in the future. Radiation and surface temperature together result in radiative feedbacks which set the final response of the climate system to any external forcing, such as CO2 ${\text{CO}}_{2}$ or aerosols. Attempts over previous decades to link internal variability to the forced response of radiation have been only mildly successful. We develop a new approach, using convolutional neural networks, which are a data‐driven, nonlinear statistical tool, to predict global‐mean top of the atmosphere radiation from spatial patterns of surface temperature. This method can indeed predict radiation far into a future which has not been seen during training. The results are robust across climate models and pass physical plausibility tests.

The sensitivity of the radiative flux at the top of the atmosphere to surface temperature perturbations cannot be directly observed. The relationship between sea surface temperature and top-of-atmosphere radiation can be estimated with Green's function simulations by locally perturbing the sea surface temperature boundary conditions in atmospheric climate models. We perform such simulations with the Ai2 Climate Emulator (ACE), a machine learning-based emulator trained on ERA5 reanalysis data (ACE2-ERA5). This produces a sensitivity map of the top-of-atmosphere radiative response to surface warming that aligns with our physical understanding of radiative feedbacks. However, ACE2-ERA5 likely underestimates the radiative response to historical warming. We argue that Green's function experiments can be used to evaluate the performance and limitations of machine learning-based climate emulators by examining if causal physical relationships are correctly represented and testing their capability for out-of-distribution predictions.

Human emissions continue to influence Earth's climate. Effective radiative forcing quantifies the effect of such anthropogenic emissions together with natural factors on Earth's energy balance (Soden et al. 2018; Gregory et al. 2020; Forster et al. 2021, 2024). Evaluating the exact rate of effective radiative forcing is challenging, because it can not be directly observed. Therefore, estimating the effective forcing usually relies heavily on climate models (Forster et al. 2024). Here, we present an estimate of effective radiative forcing that makes optimal use of observations. We use artificial intelligence to learn the relationship between surface temperature and radiation caused by internal variability in a multi-model ensemble. Combining this with observations of surface temperature and the Earth's net radiative imbalance (Loeb et al. 2018, 2021; NASA/LARC/SD/ASDC 2023), we predict an effective forcing trend of 0.72+-0.20 Wm^{-2} per decade for 2001-2023. Our method enables a new and independent assessment of the observed effective radiative forcing since 1985, that can be updated simultaneously with available observations. We make advances to close the Earth's energy budget on annual timescales, separating the influence of forcing versus the radiative response to surface temperature variations. Effective radiative forcing has substantially increased since 2021 and has not been countered by a strongly negative radiative response, consistent with an exceptionally warm year of 2023 and 2024.

Understanding the relationships between internal variability and forced climate feedbacks is key for using observations to constrain future climate change. Here we probe and interpret the differences in these relationships between the climate change projections provided by the CMIP5 and CMIP6 experiment ensembles. We find that internal variability feedbacks better predict forced feedbacks in CMIP6 relative to CMIP5 by over 50%, and that the increased predictability derives primarily from the slow (>20 years) response to climate change. A key novel result is that the increased predictability is consistent with the higher resemblance between the patterns of internal and forced temperature changes in CMIP6, which suggests temperature pattern effects play a key role in predicting forced climate feedbacks. Despite the increased predictability, emergent constraints provided by observed internal variability are weak and largely unchanged from CMIP5 to CMIP6 due to the shortness of the observational record.

Plain Language Summary Precipitation trends in the southwestern United States (SWUS) are sensitive to the pattern of sea surface temperature (SST) trends in the Tropical Pacific. Since the turn of the century, a decrease in SWUS precipitation has been linked to a cooling of the Central and Eastern Pacific (1990–2020). Notably, climate models are unable to simulate this observed cooling SST trend. In this study, we answer how SWUS precipitation projections may be impacted by potential error in the simulation of future SST trends by climate models. We first demonstrate that slight changes in the pattern of SST trends leads to either a wetting or drying of the SWUS. Second, if the current 30‐year cooling trend in the Central and East Pacific switches to a warming trend, the SWUS could experience a near‐term increase in precipitation. While climate models are the main tool to predict the global response to anthropogenic climate change, we must consider and account for their error in projections of global warming.

Migrations of the intertropical convergence zone (ITCZ) have significant impacts on tropical climate and society. Here we examine the ITCZ migration caused by CO2 increase using climate model simulations. During the first one to two decades, we find a northward ITCZ displacement primarily related to an anomalous southward atmospheric cross-equatorial energy transport. Over the next hundreds or thousands of years, the ITCZ moves south. This long-term migration is linked to delayed surface warming and reduced ocean heat uptake in the Southern Ocean, which alters the interhemispheric asymmetry of ocean heat uptake and creates a northward atmospheric cross-equatorial energy transport anomaly. The southward ITCZ shift, however, is reduced by changes in the net energy input to the atmosphere at the equator by about two-fifths. Our findings highlight the importance of Southern Ocean heat uptake to long-term ITCZ evolution by showing that the (quasi-)equilibrium ITCZ response is opposite to the transient ITCZ response.

Plain Language Summary A realistic representation of the radiation balance at the top of the Earth’s atmosphere (TOA) by coupled climate models is essential for trust in future climate projections. Despite that relevance, it is still not clear whether the models correctly simulate the coupling between surface warming and TOA radiation because of the short observational record. We show that climate models systematically underestimate the observed increase in global TOA radiation during 2001–2022 in 552 simulations. Locally, even if a simulation reproduces observed changes in surface temperature, changes in TOA radiation are more likely under‐ than overestimated. This response bias stems from the models' inability to reproduce the observed large‐scale patterns of surface warming and from errors in the atmospheric physics which suppress the communication of the surface information to the TOA. Models that better represent the TOA radiation response to local surface warming have a relatively low equilibrium climate sensitivity, that is, a weak global‐mean surface warming in response to a doubling of the atmospheric CO2 concentration above pre‐industrial levels. Our new bias metric links a model’s current response to climate change to its behavior in the future.

The atmospheric Green's function method is a technique for modeling the response of the atmosphere to changes in the spatial field of surface temperature. While early studies applied this method to changes in atmospheric circulation, it has also become an important tool to understand changes in radiative feedbacks due to evolving patterns of warming, a phenomenon called the “pattern effect.” To better study this method, this paper presents a protocol for creating atmospheric Green's functions to serve as the basis for a model intercomparison project, GFMIP. The protocol has been developed using a series of sensitivity tests performed with the HadAM3 atmosphere‐only general circulation model, along with existing and new simulations from other models. Our preliminary results have uncovered nonlinearities in the response of the atmosphere to surface temperature changes, including an asymmetrical response to warming versus cooling patch perturbations, and a change in the dependence of the response on the magnitude and size of the patches. These nonlinearities suggest that the pattern effect may depend on the heterogeneity of warming as well as its location. These experiments have also revealed tradeoffs in experimental design between patch size, perturbation strength, and the length of control and patch simulations. The protocol chosen on the basis of these experiments balances scientific utility with the simulation time and setup required by the Green's function approach. Running these simulations will further our understanding of many aspects of atmospheric response, from the pattern effect and radiative feedbacks to changes in circulation, cloudiness, and precipitation.

Atmosphere‐ocean general circulation models (AOGCMs) struggle to reproduce recently observed sea surface temperature (SST) trend patterns. Here, we quantify the relevance of this SST pattern uncertainty to global‐mean temperature projections through convolving Green's functions with SST pattern scenarios that differ from the ones AOGCMs produce by themselves. We find that future SST pattern uncertainty has a significant impact on projections, such as increasing total model uncertainty by 40% in a high‐emissions scenario by 2085. A reversal of the current cooling trend in the East Pacific over the next few decades could lead to a period of global‐mean warming with a 60% higher rate than currently projected. SST pattern uncertainty works through a destabilization of the shortwave cloud feedback to affect temperature projections. It is critical for climate change impact, adaptation, and mitigation assessments to incorporate this previously unaccounted for uncertainty until we trust the evolution of SST patterns in AOGCMs.

In the equatorial and subtropical east Pacific Ocean, strong ocean‐atmosphere coupling results in large‐amplitude interannual variability. Recent literature debates whether climate models reproduce observed short and long‐term surface temperature trends in this region. We reconcile the debate by reevaluating a large range of trends in initial condition ensembles of 15 climate models. We confirm that models fail to reproduce long‐term trends, but also find that many models do not reproduce the observed decadal‐scale swings in the East to West gradient of the equatorial Pacific. Models with high climate sensitivity are less likely to reproduce observed decadal‐scale swings than models with a modest climate sensitivity, possibly due to an incorrect balance of cloud feedbacks driven by changing inversion strength versus surface warming. Our findings suggest that two not well understood problems of the current generation of climate models are connected and we highlight the need to increase understanding of decadal‐scale variability.