Explore the vast range of titles available.

The representation of nonlinear subgrid processes, especially clouds, has been a major source of uncertainty in climate models for decades. Cloud-resolving models better represent many of these processes and can now be run globally but only for short-term simulations of at most a few years because of computational limitations. Here we demonstrate that deep learning can be used to capture many advantages of cloud-resolving modeling at a fraction of the computational cost. We train a deep neural network to represent all atmospheric subgrid processes in a climate model by learning from a multiscale model in which convection is treated explicitly. The trained neural network then replaces the traditional subgrid parameterizations in a global general circulation model in which it freely interacts with the resolved dynamics and the surface-flux scheme. The prognostic multiyear simulations are stable and closely reproduce not only the mean climate of the cloud-resolving simulation but also key aspects of variability, including precipitation extremes and the equatorial wave spectrum. Furthermore, the neural network approximately conserves energy despite not being explicitly instructed to. Finally, we show that the neural network parameterization generalizes to new surface forcing patterns but struggles to cope with temperatures far outside its training manifold. Our results show the feasibility of using deep learning for climate model parameterization. In a broader context, we anticipate that data-driven Earth system model development could play a key role in reducing climate prediction uncertainty in the coming decade.

The Importance of Sentiment in Promoting Reasonableness in Children explores the contributions that eighteenth-century Scottish philosophers Thomas Reid, Adam Smith, and David Hume make to our understanding of important factors in the development of children as they gradually acquire central features of reasonableness. Smith and Reid explicitly discuss the importance of sentiment and reason in the development of children. Their views are favorably influenced by the writings of their English predecessor Joseph Butler. Hume, too, valued much of Butler's thinking. But, unlike Smith and Reid, he said little about Butler's specific reflections on sentiment and reason. Despite this, one of the aims of this little book is to show that each contributes to our understanding today of what the encouragement of the philosophical thinking of children can play in helping them to come to an appreciation of reasonableness. They advocate a social environment for children that moves them to mix sentiment and reason in ways that support the values of reasonableness.

Future changes in the position of the intertropical convergence zone (ITCZ; a narrow band of heavy precipitation in the tropics) with climate change could affect the livelihood and food security of billions of people. Although models predict a future narrowing of the ITCZ, uncertainties remain large regarding its future position, with most past work focusing on zonal-mean shifts. Here we use projections from 27 state-of-the-art climate models and document a robust zonally varying ITCZ response to the SSP3-7.0 scenario by 2100, with a northward shift over eastern Africa and the Indian Ocean and a southward shift in the eastern Pacific and Atlantic oceans. The zonally varying response is consistent with changes in the divergent atmospheric energy transport and sector-mean shifts of the energy flux equator. Our analysis provides insight about mechanisms influencing the future position of the tropical rain belt and may allow for more-robust projections of climate change impacts.The intertropical convergence zone is predicted to narrow under climate change with large uncertainties about its location. Analysis with CMIP6 models shows a zonally varying response, with northward shift over east Africa and the Indian Ocean and southward shift in east Pacific and Atlantic oceans.

The climate sensitivity of the Madden–Julian oscillation (MJO) is measured across a broad range of temperatures (1°–35°C) using a convection-permitting global climate model with homogenous sea surface temperatures. An MJO-like signal is found to be resilient in all simulations. These results are used to investigate two ideas related to the modern “moisture mode” view of MJO dynamics. The first hypothesis is that the MJO has dynamics analogous to a form of radiative convective self-aggregation in which longwave energy maintenance mechanisms shut down for SST ≪ 25°C. Inconsistent with this hypothesis, the explicitly simulated MJO survives cooling and retains leading moist static energy (MSE) budget terms associated with longwave destabilization even at SST < 10°C. Thus, if the MJO is a form of longwave-assisted self-aggregation, it is not one that is temperature critical, as is observed in some cases of radiative–convective equilibrium (RCE) self-aggregation. The second hypothesis is that the MJO is propagated by horizontal advection of column MSE. Inconsistent with this view, the simulated MJO survives reversal of meridional moisture gradients in the basic state and a striking role for horizontal MSE advection in its propagation energy budget cannot be detected. Rather, its eastward motion is balanced by vertical MSE advection reminiscent of gravity or Kelvin wave dynamics. These findings could suggest a tight relation between the MJO and classic equatorial waves, which would tend to challenge moisture mode views of MJO dynamics that assume horizontal moisture advection as the MJO’s propagator. The simulation suite provides new opportunities for testing predictions from MJO theory across a broad climate regime.

Since the weather is chaotic, it is necessary to forecast an ensemble of future states. Recently, multiple AI weather models have emerged claiming breakthroughs in deterministic skill. Unfortunately, it is hard to fairly compare ensembles of AI forecasts because variations in ensembling methodology become confounding and the baseline data volume is immense. We address this by scoring lagged initial condition ensembles—whereby an ensemble can be constructed from a library of deterministic hindcasts. This allows the first parameter‐free intercomparison of leading AI weather models' probabilistic skill against an operational baseline. Lagged ensembles of the two leading AI weather models, GraphCast and Pangu, perform similarly even though the former outperforms the latter in deterministic scoring. These results are elaborated upon by sensitivity tests showing that commonly used multiple time‐step loss functions damage ensemble calibration. Plain Language Summary 2023 was a seminal year for data‐driven weather forecasts with several high‐profile publications claiming that AI outperformed traditional physics‐based approaches to weather forecasts. These claims are mostly supported by scoring deterministic forecasts, even though it is widely known that forecasting is a probabilistic problem. Probabilistic intercomparisons have proved challenging because of the data volumes involved and because they are confounded by particulars of how probabilistic forecasts are built. As a workaround, we propose benchmarking weather forecasts using lagged ensemble forecasting where forecasts initialized at different times are considered independent samples. When benchmarked in this way, we show that some AI models achieve better deterministic scores by reducing the variance of their forecasts at the cost of physical realism. Key Points Lagged ensembling is a practical, quantitative, and parameter‐free framework for benchmarking weather models Lagged ensembles of some recent data‐driven forecasts are under‐dispersive despite claims of “state of the art” deterministic skill Training data‐driven models with multi‐step loss functions damages ensemble calibration

Localized tropical rainfall changes commonly occur on 500–1,000 km scales under various climate forcings, but understanding their causality remains challenging. One helpful process‐oriented diagnostic (POD) decomposes the effects of undilute buoyancy and lower free‐tropospheric moisture through a precipitation‐buoyancy relationship, but its applicability at subregional scales is uncertain. We examine month‐to‐month rainfall changes in five South Asian monsoon subregions. The POD accurately characterizes the precipitation‐buoyancy relationship across all subregions and successfully predicts the sign of rainfall changes in four out of five subregions. However, the POD's ability to predict rainfall change magnitudes and identify causal mechanisms varies, providing confident explanations in only two subregions, where lower free‐tropospheric moisture emerges as the dominant driver of change. While these findings demonstrate the POD's utility in specific contexts, they also reveal limitations. We caution against using the POD as a standalone tool at these scales for predicting rainfall changes or decomposing their drivers.

Natural modes of variability on many timescales influence aerosol particle distributions and cloud properties such that isolating statistically significant differences in cloud radiative forcing due to anthropogenic aerosol perturbations (indirect effects) typically requires integrating over long simulations. For state‐of‐the‐art global climate models (GCM), especially those in which embedded cloud‐resolving models replace conventional statistical parameterizations (i.e., multiscale modeling framework, MMF), the required long integrations can be prohibitively expensive. Here an alternative approach is explored, which implements Newtonian relaxation (nudging) to constrain simulations with both pre‐industrial and present‐day aerosol emissions toward identical meteorological conditions, thus reducing differences in natural variability and dampening feedback responses in order to isolate radiative forcing. Ten‐year GCM simulations with nudging provide a more stable estimate of the global‐annual mean net aerosol indirect radiative forcing than do conventional free‐running simulations. The estimates have mean values and 95% confidence intervals of −1.19 ± 0.02 W/m2 and −1.37 ± 0.13 W/m2for nudged and free‐running simulations, respectively. Nudging also substantially increases the fraction of the world's area in which a statistically significant aerosol indirect effect can be detected (66% and 28% of the Earth's surface for nudged and free‐running simulations, respectively). One‐year MMF simulations with and without nudging provide global‐annual mean net aerosol indirect radiative forcing estimates of −0.81 W/m2 and −0.82 W/m2, respectively. These results compare well with previous estimates from three‐year free‐running MMF simulations (−0.83 W/m2), which showed the aerosol‐cloud relationship to be in better agreement with observations and high‐resolution models than in the results obtained with conventional cloud parameterizations. Key Points Nudged simulations provide more stable estimates of aerosol indirect effects Nudging increases the area a statistically significant signal can be detected Nudging enables computation‐expensive GCMs to estimate aerosol indirect effects

Modern climate projections often suffer from inadequate spatial and temporal resolution due to computational limitations, resulting in inaccurate representations of sub‐grid processes. A promising technique to address this is the multiscale modeling framework (MMF), which embeds a kilometer‐resolution cloud‐resolving model (CRM) within each atmospheric column of a host climate model to replace traditional convection and cloud parameterizations. Machine learning offers a unique opportunity to make MMF more accessible by emulating the embedded CRM and reducing its substantial computational cost. Although many studies have demonstrated proof‐of‐concept success of achieving stable hybrid simulations, it remains a challenge to achieve near operational‐level success with real geography and comprehensive variable emulation that includes, for example, explicit cloud condensate coupling. In this study, we present a stable hybrid model capable of integrating for at least 5 years with near operational‐level complexity, including coarse‐grid geography, seasonality, explicit cloud condensate and wind predictions, and land coupling. Our model demonstrates skillful online performance, achieving a 5‐year zonal mean tropospheric temperature bias within 2 K, water vapor bias within 1 g/kg, and a precipitation root mean square error of 0.96 mm/day. Key factors contributing to our online performance include an expressive U‐Net architecture and physical thermodynamic constraints for microphysics. With microphysical constraints mitigating unrealistic cloud formation, our work is the first to demonstrate realistic multi‐year cloud condensate climatology under the MMF framework. Despite these advances, online diagnostics reveal persistent biases in certain regions, highlighting the need for innovative strategies to further optimize online performance. Plain Language Summary Traditional climate models often struggle to accurately simulate small‐scale processes like thunderstorms due to compute limitations, leading to less reliable climate predictions. Machine learning (ML) offers a promising solution by efficiently modeling these processes and integrating them into hybrid ML‐physics simulations within a host climate model. While previous studies have shown success in simplified setups, such as all‐ocean planets, achieving accurate and stable hybrid simulations in complex, real‐world settings remains challenging. In this study, we developed a stable hybrid model capable of simulating the climate for 5 years using real geographic features and explicitly predicting the time evolution of temperature, moisture, cloud, and wind. Our model achieves skillful accuracy in long‐term mean atmospheric states. This success is due to several key improvements: an advanced architecture and the incorporation of cloud physics constraints. Key Points Stable hybrid climate simulations are achieved with a data‐driven emulator of subgrid physics coupled with a comprehensive atmosphere model Online performance benefits from a U‐Net architecture and microphysical constraints A realistic cloud climatology with explicit cloud condensate coupling is achieved in a hybrid multi‐scale modeling framework

We investigate a set of Energy Exascale Earth System Model Multi‐scale modeling framework (MMF) (E3SM‐MMF) simulations that vary the dimensionality and momentum transport configurations of the embedded cloud‐resolving models (CRMs), including unusually ambitious 3D configurations. Issues endemic to all MMF simulations include too much Intertropical Convergence Zone rainfall and too little over the Amazon. Systematic MMF improvements include more on‐equatorial rainfall across the Warm Pool. Interesting sensitivities to the CRM domain are found in the regional time‐mean precipitation pattern over the tropics. The 2D E3SM‐MMF produces an unrealistically rainy region over the northwestern tropical Pacific; this is reduced in computationally ambitious 3D configurations that use 1,024 embedded CRM grid columns per host cell. Trajectory analysis indicates that these regional improvements are associated with desirably fewer tropical cyclones and less extreme precipitation rates. To understand why and how the representation of precipitation improved in 3D, we propose a framework that dilution is stronger in 3D. This viewpoint is supported by multiple indirect lines of evidence, including a delayed moisture‐precipitation pickup, smaller precipitation efficiency, and amplified convective mass flux profiles and more high clouds. We also demonstrate that the effects of varying embedded CRM dimensionality and momentum transport on precipitation can be identified during the first few simulated days, providing an opportunity for rapid model tuning without high computational cost. Meanwhile the results imply that other less computationally intensive ways to enhance dilution within MMF CRMs may also be strategic tuning targets. Plain Language Summary The resolution of current climate models is not sufficient to resolve cloud and convective processes. Global cloud‐resolving models (CRMs) have resolutions fine enough to represent individual cloud events but require too much computing power to be practical for large ensemble multi‐decadal climate projection. Multi‐scale modeling framework (MMF) is used to simulate climate by embedding thousands of small CRMs interactively in each grid column of a planetary model. Trade‐offs in CRM configurations can influence the resulting emergent behavior. To examine this issue, we explore the effects of unusually ambitious 3D CRM configurations. Results show some interesting differences in the regional precipitation over the tropics. The 2D MMF produces an unrealistically rainy region over the northwestern tropical Pacific. Such biases are significantly reduced in 3D due to fewer tropical cyclones. To understand why and how the representation of precipitation improved in 3D, we propose a framework in which mixing being stronger in 3D is a major part of the story. This favored explanation is hard to prove directly but a few lines of circumstantial evidence support the case. Additionally, rapid effects of mixing that is seen in the first few days of global cloud resolving simulations provide opportunities for optimizing longer‐term statistics. Key Points Dimensionality of cloud‐resolving models in the Energy Exascale Earth System Model Multi‐scale modeling framework (MMF) exhibits a striking effect on mean state precipitation patterns in subregions of the tropics MMFs tend to produce too many precipitating events but the use of 3D leads to fewer and is associated with an enhanced dilution in 3D Fast precursors of these climatological sensitivities are found that point to calibration targets for convection permitting global models

Ocean circulation responses to interhemispheric radiative imbalance can damp north–south migrations of the intertropical convergence zone (ITCZ) by reducing the burden on atmospheric energy transport. The role of the Atlantic meridional overturning circulation (AMOC) in such dynamics has not received much attention. Here, we present coupled climate modeling results that suggest AMOC responses are of first-order importance to muting ITCZ shift magnitudes as a pair of hemispherically asymmetric solar forcing bands is moved from equatorial to polar latitudes. The cross-equatorial energy transport response to the same amount of interhemispheric forcing becomes systematically more ocean-centric when higher latitudes are perturbed in association with strengthening AMOC responses. In contrast, the responses of the Pacific subtropical cell are not monotonic and cannot predict this variance in the ITCZ’s equilibrium position. Overall, these results highlight the importance of the meridional distribution of interhemispheric radiative imbalance and the rich buffering of internal feedbacks that occurs in dynamic versus thermodynamic (slab) ocean modeling experiments. Mostly, the results imply that the problem of developing a theory of ITCZ migration is entangled with that of understanding the AMOC’s response to hemispherically asymmetric radiative forcing—a difficult topic deserving of focused analysis across more climate models.