https://arxiv.org/api/c8G5gzeOqlyZC11pVUih1Wzkdr0 2026-06-18T10:53:59Z 8381 360 15 http://arxiv.org/abs/2603.25635v1 2026-03-26T16:46:54Z

Air flow modeling at a local scale is essential for applications such as pollutant dispersion modeling or wind farm modeling. To circumvent costly Computational Fluid Dynamics (CFD) computations, deep learning surrogate models have recently emerged as promising alternatives. However, in the context of urban air flow, deep learning models struggle to adapt to the high variations of the urban geometry and to large mesh sizes. To tackle these challenges, we introduce Anchored Branched Steady-state WInd Flow Transformer (AB-SWIFT), a transformer-based model with an internal branched structure uniquely designed for atmospheric flow modeling. We train our model on a specially designed database of atmospheric simulations around randomised urban geometries and with a mixture of unstable, neutral, and stable atmospheric stratifications. Our model reaches the best accuracy on all predicted fields compared to state-of-the-art transformers and graph-based models. Our code and data is available at https://github.com/cerea-daml/abswift.

2026-03-26T16:46:54Z Armand de Villeroché Rem-Sophia Mouradi Vincent Le Guen Sibo Cheng Marc Bocquet Alban Farchi Patrick Armand Patrick Massin http://arxiv.org/abs/2604.03289v1 2026-03-26T09:13:46Z

Coupling constitutes a foundational mechanism in the Earth system, regulating the interconnected physical, chemical, and biological processes that link its spheres. This review examines how emerging artificial intelligence (AI) methods create new opportunities to enhance Earth system coupling and address long-standing limitations in multi-component models. Rather than surveying next-generation modelling efforts broadly, we focus specifically on how state-of-the-art AI techniques can strengthen cross-domain interactions, support more coherent multi-component representations, and enable progress toward unified Earth system frameworks. The scope extends beyond climate models to include any modelling system in which Earth spheres interact. We outline emerging opportunities, persistent limitations, and conceptual pathways through which AI may enhance physical consistency, interpretability, and integration across domains. In doing so, this review provides a structured foundation for understanding the role of AI in advancing coupled Earth system modelling.

2026-03-26T09:13:46Z Maria Kaselimi Anna Belehaki http://arxiv.org/abs/2410.04822v4 2026-03-26T07:21:18Z

The aim of this second part of the article is to study the absolute definition of the seawater entropy described in Part I with several concrete cases. Observed vertical profiles and polar transects, as well as analysed surface data, show that very different temperatures and salinity values can organise to create new isentropic regions. This can only be revealed by the absolute formulation of the entropy of seawater (Arctic Ocean; Bay of Bengal; Mediterranean, Black, and Caspian Seas). Existing hypotheses to explain these results include the possible impact of turbulent processes that must be applied to the entropies of the atmosphere and oceans.

2024-10-07T08:13:40Z V1: First submitted on the 7th of October 2024. V2: arXiv reference for the Part-II. V3: First Revised version. V4: Accepted in the Comptes Rendus (Geoscience) Paris, France on the 25th of March, 2026 (7 Figures; 16 pages) Pascal Marquet 10.5802/crgeos.334 http://arxiv.org/abs/2410.04821v4 2026-03-26T07:13:46Z

The first improvements concern the complex non-linear dependence of entropy on pressure, temperature and salinity, with the use of the standard TEOS10 formulation based on a fit of the oceanic Gibbs function to more recent observations. On the other hand, more recent thermodynamic tables have been used to increase the accuracy of the Millero's salinity increment to this standard formulation, to deduce the absolute version of entropy with new values for the pure-water and sea-salts absolute reference entropies. The differences between the values of the seawater entropy calculated with the Millero and TEOS10 formulations (standard and absolute) are documented, before a more complete study shown in the second part of the paper of the absolute seawater entropy computed from observed vertical profiles and analysed surface datasets.

2024-10-07T08:13:25Z V1: First submitted on the 7th of October 2024. V2: arXiv reference for the Part-II. V3: First Revised version. V4: Accepted in the Comptes Rendus (Geoscience) Paris, France on the 23th of March, 2026 (3 Figures; 27 pages) Pascal Marquet 10.5802/crgeos.333 http://arxiv.org/abs/2409.04063v3 2026-03-26T06:30:22Z

Models of complex dynamical systems like the Earth's climate often involve large numbers of uncertain parameters. Comprehensive exploration of the parameter space is typically prohibitive due to excessive computational costs. Systematic gradient-based parameter optimization is not feasible because such models are typically not differentiable. This is especially problematic in cases where the models describe highly nonlinear and possibly abrupt dynamics, where sensitivity to parameter changes is high. Components of Earth's climate system, such as the North Atlantic Overturning Circulation or the polar ice sheets, are at risk of undergoing critical transitions in response to anthropogenic climate change. Concerns have been raised that these Earth system components are too stable in state-of-the-art models. In my presentation, we will see examples how new scenario simulations allow studying the possibility and the consequences of tipping events in Earth system models. Also, we will discuss a method for efficient, systematic, and objective calibration of dynamical complex system models, targeted at adjusting system stability. Given a number of physical or observational constraints, the method can efficiently adjust stability biases in a range of complex system models and help reveal potentially hidden instabilities, with important implications for Earth system modelling.

2024-09-06T07:14:14Z Daniel Pals Sebastian Bathiany Richard Wood Joel Kuettel Niklas Boers http://arxiv.org/abs/2603.24867v1 2026-03-25T23:13:47Z

Understanding how weather and climate influence fire risk is important for many purposes, including climate adaptation planning and decision-making in sectors such as emergency management, finance, health and infrastructure (e.g., for energy and water availability). In this study, bias-corrected 20CRv2c reanalysis data are used to investigate the climatology and long-term trends of weather conditions associated with landscape fires in Australia. The McArthur Forest Fire Danger Index (FFDI) is used here as a broad-scale representation of weather conditions known to influence fire behaviour based on wind speed, humidity, temperature and rainfall measures. In particular, using this reanalysis dataset allows analysis over a longer time period than previous studies, from 1876 to 2011. Another novel aspect is that trends are examined using several different approaches, including a method to help account for the influence of interannual drivers of climate variability not previously used for fire weather analysis. Results show increases in mean and extreme seasonal FFDI values throughout Australia in general, with all statistically significant trends being positive in sign for individual climate zones. Humidity and temperature trends, attributable to human-caused climate change, are shown to be the main cause of the increase in dangerous weather conditions for fires. These findings build on previous studies, with the novel data and methods used adding confidence to the overall understanding of fire risk factors in a changing climate.

2026-03-25T23:13:47Z 33 pages, 17 figures, 2 tables Soubhik Biswas Andrew Dowdy Savin Chand http://arxiv.org/abs/2603.24488v1 2026-03-25T16:29:48Z

The Pattern Effect describes the dependence of top-of-atmosphere radiation anomalies on changes in the pattern of sea surface temperatures. The emerging consensus in the field explains the impact of Pacific warm pool temperature on radiation using Convective Quasi-Equilibrium Weak Temperature Gradient (QE-WTG) theory: warm pool warming leads to increase in free-tropospheric temperatures across the tropics, a strengthening of inversion, increased cloud cover in the East Pacific low cloud decks, and negative radiative anomalies. Here we call on overlooked past results and new simulations from the Energy Exascale Earth System model to show that Rossby waves dominate the low-cloud response over the subtropical East Pacific low cloud decks, leading to decrease cloud cover in the low cloud decks. While the global radiative response is negative and consistent with QE-WTG, it is dominated by the response of the deep tropics, rather than the subtropical low cloud decks.

2026-03-25T16:29:48Z Cristian Proistosescu Pappu Paul Nicholas J. Lutsko Andrew I. L. Williams Malte F. Stuecker http://arxiv.org/abs/2603.24374v1 2026-03-25T14:55:25Z

Hourly rainfall extremes cause some of the most destructive weather disasters, yet numerical weather prediction models still struggle to forecast them, and a physical basis for their predictability remains unclear. Here, we identify a trivariate clustering of hourly rainfall extremes with surface temperature, phases of the Monsoon Intraseasonal Oscillation (MISO), and precipitable water vapor, establishing a physical foundation for the medium range predictability of these events. This clustering arises from multiscale interactions in which extremes organize into storm systems embedded within mesoscale convective clusters and synoptic low-pressure systems during active MISO phases. We develop an algorithm to identify, track, and monitor these storm systems. Although rapid error growth limits the prediction of isolated hourly extremes, our results provide basis for a physics informed training of deep learning, data driven models to forecast organized clusters of hourly rainfall extremes more than a week in advance, offering substantial potential to reduce losses from extreme rainfall.

2026-03-25T14:55:25Z 6 figures, 1 table, 9 supplementary figures, and 1 supplementary table Bijit Kumar Banerjee Devabrat Sharma Mahen Konwar Simanta Das Utpal Sarma B. N. Goswami http://arxiv.org/abs/2603.24233v1 2026-03-25T12:13:51Z

Italy and the surrounding seas are recognised as one of the European hotspots for tornadoes and waterspouts. In recent years, the town of Rosignano Solvay (on the Northern Tyrrhenian coast) experienced repeated waterspouts affecting the same areas, raising local concern about the possible influence of heated wastewater discharged into the sea by a nearby industrial site. We reconstruct the mesoscale meteorological conditions of four intense waterspouts near Rosignano Solvay using a limited-area weather model at high-to-very-high resolution (inner domain grid spacing 500 m; sensitivity tests at 100 m). At the reported event times, the intensity of key mesoscale precursors (low-level wind shear, 1 km storm-relative helicity, maximum updraft intensity, and lifting condensation level) is consistent with the values typically associated with EF1 (or stronger) tornadoes and waterspouts. The model systematically predicts the peak of instability indices 2-3 hours earlier than the reported event times. For one case study, we conduct two sea surface temperature sensitivity experiments to assess the potential atmospheric impact of heated wastewater discharge (temperature increases of +1.5 K and +5 K over a 10 km$^2$ area). The resulting changes in instability indices are marginal, with differences of at most 3\% relative to the control run. A simple mass-balance estimate for the modified sea patch suggests that, given the reported discharge rates, a plausible impact of the warm water released from the industrial site could lead to an increase in the local sea surface temperature of approximately +0.7 °C over two months. We conclude that synoptic and mesoscale conditions primarily govern waterspout initiation in this region, while the direct effect of the small, warm coastal plume from the industrial discharge appears to be minor.

2026-03-25T12:13:51Z Valerio Capecchi Bernardo Gozzini Mario Marcello Miglietta http://arxiv.org/abs/2601.07983v2 2026-03-23T18:14:51Z

Recent works suggest there are periods when the Sun encountered massive interstellar cold clouds which compressed the heliosphere to within Earth's orbit. This would expose Earth to increased galactic cosmic rays (GCRs) and energetic particles of heliospheric origin (HEPs). We model 10Be production in Earth's atmosphere during possible interstellar cloud encounters and supernovae, and estimate their detectability in marine and ice records of variable temporal resolution. We find that encounters with sub-parsec-scale clouds cannot be detected via 10Be measurements in iron-manganese crusts. GCR signal is not detectable in ocean sediments; only parsec-scale clouds can be detected in iron-manganese crusts. Encounters with sub-parsec-scale clouds can be detected if Earth is exposed to HEPs, or by measuring ice cores with less than or equal to 10 kyr temporal resolution. We find that the peak observed by Koll et al. (2025) 10 Ma cannot be attributed to a supernova, but could be attributed to an interstellar cloud crossing.

2026-01-12T20:30:48Z 17 pages, 4 figures, submitted to GRL Anna Nica Merav Opher Jesse A. Miller Jennifer L. Middleton http://arxiv.org/abs/2603.22037v1 2026-03-23T14:39:13Z

Satellite observations play a critical role in numerical weather prediction where they are assimilated through an observation operator that maps model states to radiances. In the traditional Ensemble Kalman Filter, these observations are used to update the state by weighting their associated errors against model uncertainties to produce an optimal estimate. This process requires radiative transfer simulations for passive, downward-viewing satellite radiometers operating in the visible, infrared, and microwave spectra. Typically, such simulations rely on numerically integrating physical laws via models like RTTOV. In this paper, we introduce two machine learning surrogate observation operators inspired by modern computer-vision architectures: First, a fully data-driven emulator of radiative transfer, and second, a hybrid incremental correction model that learns only the residual relative to RTTOV, thereby retaining established physics while enabling data-driven refinement in complex conditions such as cloud-affected situations. The residual formulation improves radiance accuracy (lower Root Mean Squared Error (RMSE) than the fully data-driven emulator and RTTOV) and adds only moderate computational costs to the assimilation step. Both models combine 3D convolutions for vertical profile encoding with a 2D U-Net operating on latitude-longitude grids, allowing joint learning of vertical structure, spatial correlations, and inter-channel dependencies. We further provide a theoretical justification for deploying the hybrid surrogate as an observation operator in data assimilation.

2026-03-23T14:39:13Z 29 pages (incl. appendix & bibliography), 9 figures (incl. appendix) Gian Luca Buono Stefanie Hollborn Roland Potthast Jörg Schäfer Martin Simon http://arxiv.org/abs/2603.16272v2 2026-03-23T13:59:24Z

Accurate reconstruction of global Sea surface temperature (SST), which dominates the air-sea coupling and global climate variability, underpins climate monitoring and prediction. Existing SST reconstruction products primarily provide one deterministic field derived from heterogeneous satellite data and in situ observations, limiting their ability to represent observation uncertainty and to support probabilistic forecasting. Here, we introduce Satellite and in situ Adaptive Guided Estimation (SAGE), a diffusion-based uncertainty-aware generative framework for probabilistic SST reconstruction. SAGE learns a physically consistent prior from historical SST data and performs observation-conditioned posterior sampling without requiring satellite or in situ data during training, enabling flexible state inference from heterogeneous observations. Through a progressive data-fusion strategy, observations from two FengYun-3D polar-orbiting satellites constrain basin-scale structures, while sparse in situ measurements serve to refine local anomalies and extremes. The resulting ensemble SST fields well capture observational uncertainty and scale-dependent variability. Validation against independent in situ observations shows that SAGE substantially reduces reconstruction errors compared with widely used operational products. When used to initialize forecasting systems, SAGE-generated SST fields substantially reduce 10-day SST forecast errors relative to current operational analyses. At the climate scale, SAGE-driven forecasts of the 2023-2024 El Nino event show added value in capturing its onset and intensity evolution compared to conventional approaches. Our results demonstrate that SAGE represents a step toward a new paradigm for ocean state estimation and climate prediction.

2026-03-17T09:04:27Z Haijie Li Ya Wang Kai Yang Gang Huang Xiangao Xia Ziming Chen Weichen Tao Chenglin Lyu Lin Chen Miao Zhang Kaiming Hu Hainan Gong Disong Fu Lin Wang http://arxiv.org/abs/2603.21801v1 2026-03-23T10:42:58Z

Solitons are localized nonlinear wave packets that propagate without spreading because nonlinearity balances dispersion. Their robustness is well understood in effectively one-dimensional systems, but introducing additional spatial dimensions is generally expected to destabilize them or destroy their coherent character. Here we experimentally investigate how deep-water gravity-wave solitons behave when a controlled transverse degree of freedom is introduced through diffraction. Using a large-scale water-wave facility, we generate solitonic wave packets whose transverse structure is imposed across a segmented wavemaker through either a sharp slit or a smooth Gaussian apodization. The resulting two-dimensional wave fields are measured with high spatial resolution. Diffraction reshapes the transverse profile of the wave packet while its longitudinal dynamics retain the characteristic features of a soliton. Nonlinear spectral analysis confirms that the solitonic content is preserved along the direction of propagation, whereas the transverse evolution follows the linear Fresnel laws of diffraction. These observations reveal an unexpected coexistence of nonlinear soliton dynamics and classical wave diffraction.

2026-03-23T10:42:58Z Filip Novkoski FAU, MSC Loïc Fache MSC, PhLAM Félicien Bonnefoy LHEEA Guillaume Ducrozet LHEEA, Nantes Univ - ECN, CNRS Jason Barckicke MSC François Copie DYSCO, PhLAM Pierre Suret PhLAM, DYSCO Eric Falcon MSC Stéphane Randoux PhLAM, DYSCO http://arxiv.org/abs/2411.14769v2 2026-03-23T08:36:38Z

We present a generalized linear response theory for mixed jump-diffusion models -- combining Gaussian and Lévy noise interacting with nonlinear dynamics -- by deriving comprehensive response formulas accounting for perturbations to both the drift term and the jumps law. This class of models is particularly relevant for parameterizing the effects of unresolved scales in complex systems. Our formulas thus quantify uncertainties in parameterized components (e.g., jump laws) or measure dynamical changes due to drift term perturbations (e.g., parameter variations). By generalizing the concepts of Kolmogorov operators and Green's functions, we obtain new forms of fluctuation-dissipation relations. The resulting response is decomposed into contributions from the eigenmodes of the Kolmogorov operator, revealing the intimate relationship between a system's natural and forced variability. We demonstrate the theory's predictive power with two distinct climate-centric applications. First, we apply our framework to a paradigmatic ENSO model subject to state-dependent jumps and additive white noise, showing how the theory accurately predicts the system's response to perturbations and how Kolmogorov modes can be used to diagnose its complex time variability. In a second, more challenging application, we use our linear response theory to perform accurate climate change projections in the Ghil-Sellers energy balance climate model, a spatially-extended model forced by a spatio-temporal $α$-stable process. This work provides a comprehensive approach to climate modeling and prediction that enriches Hasselmann's program, with implications for understanding climate sensitivity, detection and attribution of climate change, and assessing climate tipping points. Our results may find applications beyond climate, and are relevant for epidemiology, biology, finance, and quantitative social sciences.

2024-11-22T07:20:13Z 31 pages, 10 figures Mickael D Chekroun et al. 2025, Reports on Progress in Physics, 88, 127601 Mickaël D. Chekroun Niccolò Zagli Valerio Lucarini 10.1088/1361-6633/ae2206 http://arxiv.org/abs/2510.09896v2 2026-03-22T23:24:32Z

Earth's energy imbalance at the top of the atmosphere is a key climate system metric, but its natural variability is poorly constrained by the short observational record and large uncertainty in coupled climate models. While existing ocean heat content reconstructions offer a longer perspective, they cannot separate the contributions of shortwave and longwave radiation, obscuring the underlying processes. We extend the energy budget record into the pre-industrial period by reconstructing the top-of-atmosphere radiation and related surface variables over the last millennium (850-2000 CE) by using data assimilation to combine proxy data and dynamics from a coupled climate emulator. Validation reveals skill in the reconstructed radiation fields, especially in the tropics. Results show a familiar last-millennium cooling trend, which coincides with persistent heat loss and a reduction in upper-ocean heat content. The cooling trend differs by season and latitude, and is associated with radiative anomalies suggestive of an eastward shift in Indo-Pacific convection. Following large volcanic eruptions, ocean heat content anomalies persist for 10-20 years on average, supporting previous evidence that the cooling trend was forced by decadally-paced eruptions. The reconstruction also reveals that the current rate of energy gain is unprecedented relative to the period before 1850.

2025-10-10T22:21:26Z Dominik Stiller Gregory J. Hakim