Ocean models are used to predict currents, temperature, and salinity of the earth's oceans, much like weather forecasting. As supercomputer hardware changes with evolving technology, models must be updated, and sometimes rewritten. Here we document Omega, a new ocean model that was designed to run on the world's fastest supercomputers. Testing shows that Omega accurately solves the model equations, and runs efficiently on many different computer architectures, including exascale computers.

Forests are shaped by how trees compete for light. We improved a widely used vegetation model by introducing spatially explicit canopies, allowing light distribution to vary realistically, especially after disturbances like tree death or harvest. This enhances the simulation of forest structure, tree coexistence, and regeneration, providing a better model for predicting forest dynamics and responses to management and environmental change.

Emissions from fires are tallied in Canada's National Forest Carbon Monitoring Accounting and Reporting System. Mapped fire extents and regional carbon stock estimates are used, but a fixed and high fire severity is assumed. This paper improves on existing methods by calculating fire emissions based on mapped fire severity. This new method compares well against independent measurements for the 2023 fires in Canada.

We evaluated Rossby waves in dynamical downscaling simulations over North America, and their connections to surface air temperature variability and heatwaves. Simulated Rossby wave propagation is distorted by flow discontinuities at lateral boundaries and by biased mean wind patterns, thereby breaking the region-specific connections between Rossby waves and surface temperature. Adjusting simulated large-scale winds to match the forcing data can reduce these biases.

We introduce a new approach based on information theory to study climate variability using observational and model data. To showcase our methodology, we study the tropical and South Atlantic by examining broad patterns in ocean and rainfall data at decadal scales in four climate models under mid‐Holocene and pre‐industrial conditions and observations. Our findings show that large-scale patterns help explain long-term climate behavior.

We present a simple method that turns coarse flood volume estimates into local flood depth maps by using the shape of the land and mapped levee zones. Used with a large-scale river model, it keeps total water volume consistent while spreading water realistically inside and outside levees. Tests show levees often confine water and cut flood volume by about 10 %–15 % for many event sizes. The method reveals place-to-place differences in protection and yields clearer hazard maps for planning.

Climate forcings, such as solar radiation or anthropogenic greenhouse gases, are required to run global climate model simulations. Stratospheric aerosols, which mostly originate from large volcanic eruptions, are a key natural forcing. In this paper, we document the stratospheric aerosol forcing dataset that will feed the next generation (CMIP7) of climate models. Our dataset is very different from its predecessor (CMIP6), which might affect simulations of the 1850–2021 climate.

This study presents a scheme, Concepteur de Réseaux Optimaux d’Observations Atmosphériques (CRO²A), for designing optimal mesoscale atmospheric monitoring networks without relying on typical inverse modeling assumptions. It leverages direct simulations of greenhouse gas concentrations to minimize the number of ground-based monitoring stations and maximize network performance through automated processing at a balanced computational cost, while being compatible with high-performance computing.

Central processing units (CPUs) and graphics processing units (GPUs) are different devices that suit different kinds of work. Using a climate modeling component, we provide a clearer way to tell which device type is faster for a given task. This matters because runs usually use only one device type. Our results are actionable: they guide device choice, report performance gains fairly, highlight code areas to improve, and show how code structure and optimization can change conclusions.

The Brazilian Regional Atmospheric Modeling System (BRAMS) was enhanced with crown fire spread in the surface fire spread (SFIRE) module and dynamic smoke emissions, enabling coupled fire–atmosphere–radiation simulations. Applied to the 2017 Sertã wildfire, the results show qualitative agreement with MERRA-2 aerosol optical depth (AOD), capturing the main plume structure, timing, and smoke-driven radiative effects, as well as interactions affecting atmospheric instability and inversion layers.

Recent advancements in regional ocean modelling allow higher resolution simulations providing improved estimates of the large-scale ocean state, while also revealing new insights into the fine-scale processes connecting the open ocean to the continental shelf seas. Our study highlights the importance of increased model resolution in regions of the ocean that are particularly turbulent while in quasi-stable circulation regions (e.g., jets), the current state-of-the-art global models do suffice.

Most climate models cannot resolve clouds and cloud-radiation interactions at coarse horizontal resolutions of about 100 km, which introduces uncertainties. High-resolution models resolve clouds better but are expensive to run. We use short high-resolution simulations and artificial intelligence to learn the cloud-radiation interactions without making any assumptions about the small scales. We propose a new method that significantly reduces cloud related errors.

As meteorological models grow in complexity, the volume of output data increases, making compression increasingly desirable. However, no specialized methods currently exist for compressing data in the Lagrangian frame. To address this gap, we developed psit, a pipeline for the lossy compression of Lagrangian flow data. In most cases, psit achieves performance that is equivalent or superior to non specialized alternatives, with compression errors behaving similar to measurement inaccuracies.

An upgraded version of a numerical solver is introduced to better capture the 3D interactions between surface water and groundwater. Built using open-source software, the custom solver adds new features to handle the complexity of real environments, including the representation of subsurface geology and the simulation of diverse dynamic processes, such as solute transport and heat transfer, in both domains. A test case and a full description of the novel features are provided in this paper.

This study presents an open-source Python solver that uses graphics processing units to simulate shallow water flows in floods and tsunamis without costly hardware or complex code. We refined numerical methods to reduce wave-spread errors and validated them on standard cases and real events, including a French dam break and a Chilean earthquake tsunami. The solver effectively reproduces wave heights and speeds, potentially enhancing early warnings in flood-prone areas.

Extreme fire events are increasing, and fires are projected to rise across most regions, posing growing risks to ecosystems and society. As a key Earth system process, fire is now modeled in most Earth System Models (ESMs). FireMIP (Fire Modeling Intercomparison Project) within CMIP7 will evaluate fire simulations in state‑of‑the‑art ESMs, project future fire changes, and provide quantitative, process‑based understanding of fire's role in the Earth system.