Predicting shifts in local extreme weather under global warming is key for climate adaptation, but climate projections lack detail. A new tool, EXSoDOS (DOwnScaling of weather EXtremes Shifts), combines ground measurements, reanalysis data, and climate models to improve estimates of extreme weather, aiding better risk planning. Tested in five regions, it accurately captures temperature, rainfall, and wind extremes including their past changes, outperforming raw model data. Results show worsening heat (stress) and precipitation by 2100.

Boreal peatlands store much of the global soil carbon and the service is closely related to nutrient cycling. This study improved a major land surface model to better represent how plants gain nitrogen and phosphorus through fine roots and mycorrhizal association. The new model more accurately captured observed carbon fluxes than the default model at an experimental site in Minnesota, and suggests shifts in nutrient uptake strategy mitigates transition from carbon sink to source under warming.

This paper presents a crop yield emulator for four major crops (maize, rice, soybean, and wheat). Trained on process-based crop model simulations, it captures yield responses to key drivers: atmospheric CO₂, temperature, water availability, and nitrogen use. The emulator closely reproduces the behavior of complex crop models and aligns well with FAO (Food and Agriculture Organization)-reported historical yields. It provides a robust and efficient method for representing agricultural outcomes in climate change impact assessments.

The Arctic may have lost its summer sea ice 127,000 years ago during a naturally warm period in Earth’s past. Climate models can be tested by recreating those conditions, with similar sunlight and greenhouse gas levels. Analysing the large sea ice changes in these simulations helps us understand how the Arctic might respond in the near future and improves how we test and trust our climate models.

Current atmospheric models are limited by the computational time required for physical processes, known as physical parameterizations. To address this, we developed neural network-based emulators to replace these parameterizations in the IPSL climate model, using a simplified aquaplanet setup and a realistic configuration. We found that incorporating some physical knowledge, such as latent variables, into the learning process can improve predictions.

The growing size of datasets used in climate science makes it difficult to store, analyze, and distribute dataset. Lossy compression algorithms can significantly reduce the disk space required to store datasets, but it can be difficult to understand and compare the behavior of different compression algorithms. ClimateBenchPress provides a benchmark to standardize comparisons between lossy compression algorithms and guide development of novel algorithms specifically targeted towards climate data.

We present a new implementation of our method for estimation of human-created carbon dioxide in the ocean. Tracer-based Rapid Anthropogenic Carbon Estimation relies on transient tracer measurements to infer gas exchange and circulation. Our work implements practical and fundamental improvements increasing accessibility, flexibility, and skill of the method. We provide an updated data product of global ocean carbon inventories spanning the industrial era and a range of future projections.

Reducing emissions of greenhouse gases such as methane and carbon dioxide is essential for addressing climate change. We developed HyperGas, an open tool that uses hyperspectral satellite images to retrieve and detect greenhouse gas plumes. It helps scientists locate emission sources, estimate their strength, and examine uncertainties through an easy workflow and visual app. Our goal is to make tracking human-made emissions more accurate and accessible, supporting better climate monitoring.

In this experiment, the Parallelized Large-Eddy Simulation Model (PALM)’s performance in simulating incoming and outgoing short-wave radiation in a densely built, highly heterogeneous urban environment was validated. In particular, we assessed whether the micro-scale model realistically resolves the effects of three-dimensional urban morphology and vegetation on short-wave radiation, including its propagation, shading, reflection, and attenuation within the simulated domain.

We developed a contrastive learning method (CLPFV - contrastive learning-based precipitation forecast verification) to improve the accuracy of precipitation forecast verification. The proposed method uses precipitation augmentation to simulate real-world forecast errors with gradients and then employs an improved loss function to reflect these errors in the contrastive learning. Experimental results show that the proposed method outperforms traditional and spatial verification methods across different error types and aligns better with expert judgment.

The ocean and sea ice are central to Earth's climate system, influencing global heat and carbon cycles, weather patterns, and sea level rise. Here the earth system model variables needed for studies of the ocean and sea ice are prioritized and requested. These requested variables were developed through an international, community-based process.  

In this work we examined how different assumptions regarding soil carbon model calibration affect the resulting model performance. We found that how the litter inputs are set have a meaningful impact on the calibrated model parameters. Furthermore, two calibration methods produced parameter sets that differed meaningfully from each other but fit the validation dataset equally well. These results raise meaningful questions how we evaluate soil carbon model performance.

Cloud microphysics is a leading source of error in both regional and global climate models and this limits our ability to understand the Earth’s climate and how it is changing. However a fairly new type of model called a Superdroplet Model (SDM) may improve both regional and global models if it can be made cost-efficient enough. Hence we are introducing a novel version of SDM, called CLEO, and it's key features that make it efficient, especially on very high performance, exascale, computers.

We are developing a model called CLEO, a type of “Superdroplet Model” (SDM) for cloud microphysics, to try to overcome some of the issues faced by climate models which are caused by errors in cloud modelling. Here we describe the equations for cloud microphysics CLEO uses and how we solve them, such as to see how water-droplets move around and grow/shrink in the atmosphere. We also provide some demonstrations of the microphysical processes we model to show that CLEO works as intended.

The early Eocene, about 50 million years ago, was a super-warm period of Earth's history, with high concentrations of carbon dioxide in the atmosphere. Here, we provide a framework and experimental design for climate modellers to carry out a coordinated project, simulating this period. This is the second phase of this project, and here we provide updated maps of the Earth's mountains and ocean floor, and vegetation, to enable more accurate modelling.

Understanding climate effects of airborne particles requires computer models that can represent small-scale dynamics. We introduce a new regional configuration of a kilometre-scale aerosol–climate model and apply it to air pollution in Central Europe, sea spray in the Atlantic Arctic, and smoke and dust over Australia. The simulations reproduce key particle patterns and variability, but also reveal limitations due to simplified sources and chemistry. The results guide future model improvements.

As climate risks grow, society needs reliable predictions for the coming years and decades. We developed a framework to collectively compare climate prediction systems and examine their performances on global temperature, rainfall, and sea ice. As a complementary to traditional analyses, our new framework offers tracking evolution of model performance in simulation time, helping scientists and stakeholders better understand strengths and limits of decadal climate prediction.

Coral reefs and limestones hold crucial records of Earth's climate history, but scientists have lacked accessible tools to simulate how these systems form over thousands to millions of years. We developed CarboKitten, free software that models how tropical sediments and associated organisms grow under changing sea levels and environmental conditions. The program runs fast on standard computers and can test scientific theories about how these geological features preserve the Earth's history.

We developed a new cloud microphysics scheme that links simulated cloud and precipitation particles with radar signals. Idealized squall line and a real thunderstorm case in Taiwan show that the scheme produces physically meaningful storm structures and encouraging radar comparisons, while further refinement is still needed for ice particles, hydrometeor orientation, and radar-viewing assumptions.

Understanding how greenhouse gases and pollutants move through the atmosphere is crucial. A new model, the Boundary Layer Dispersion and Footprint Model (BLDFM), tracks their movement. Unlike previous models, BLDFM uses a numerical approach without simplifying assumptions. It is flexible and can be used for climate impact studies and industrial emissions monitoring. Our testing and comparison results show BLDFM's potential as a valuable research tool.

This article is a comprehensive description of the 3.0.2 stable release of the Crocus snowpack model. It describes various new implementations since the last reference article in 2012 and a review of the available scientific evaluations and applications of the model. This provides guidance for the future of numerical snow modelling.

This study develops a fast, data‑driven tool to virtually test millions of ways to manage winter wheat fields in Denmark, without running slow process-based crop models each time. It finds fertilizer, residue, manure, catch crop and irrigation strategies that cut nitrogen pollution and greenhouse gases while increasing yields and soil carbon, all without using more fertilizer overall.

We introduce the new plant functional diversity model JeDi-BACH, a novel tool that integrates the Jena Diversity Model (JeDi) within the land component of the ICON Earth System Model. JeDi-BACH captures a richer set of plant trait variations based on environmental filtering and functional tradeoffs without a priori knowledge of the vegetation types. JeDi-BACH represents a significant advancement in modeling the complex interactions between plant functional diversity and climate.