We developed fast machine learning models to predict forest regrowth and carbon dynamics under climate change. These models mimic the outputs of a complex vegetation model but run 95 % faster, enabling global analyses and supporting climate solutions in large modeling frameworks such as LandSyMM.

We introduce a new version of WAM2layers (Water Accounting Model – 2 layers), a computer program that tracks how the weather brings water from one place to another. It uses data from weather and climate models, whose resolution is steadily increasing. Processing the latest data had become a challenge, and the updates presented here ensure that WAM2layers runs smoothly again. We also made it easier to use the program and to understand its source code. This makes it more transparent, reliable, and easier to maintain.

Non-thermal emission from chemical reactions in the Earth's middle und upper atmosphere strongly contributes to the brightness of the night sky below about 2.3 µm. The new Paranal Airglow Line And Continuum Emission model calculates the emission spectrum and its variability with an unprecedented accuracy. Relying on a large spectroscopic data set from astronomical spectrographs and theoretical molecular/atomic data, this model is valuable for airglow research and astronomical observatories.

Climate model simulations of the response to human and natural influences together, natural climate influences alone and greenhouse gases alone are key to quantifying human influence on the climate. The last set of such coordinated simulations underpinned key findings in the last Intergovernmental Panel on Climate Change (IPCC) report. Here we propose a new set of such simulations to be used in the next generation of attribution studies and to underpin the next IPCC report.

A numerical model that simulates the measurement processes behind the ground-based radars used to detect volcanic ash clouds is introduced. Using weather radars to detect volcanic clouds is not ideal, as fine ash particles are smaller than raindrops and remain undetected. We evaluate the performance of weather radars to study ash clouds and to identify optimal frequencies that balance the trade-off between a higher return signal and the higher path attenuation that comes at these higher frequencies.

Photochemical models describe how the composition of the atmosphere changes due to chemical reactions, transport, and other processes. These models are useful for studying the composition of the Earth's and other planets' atmospheres. Understanding the results of these models can be difficult. Here, we build on previous work to develop open-source code that can identify the reaction chains (pathways) that produce the results of these models, facilitating the understanding of these results.

This paper describes an R package as the machine interface to the open data of Mindat.org, one of the world's most widely used databases of mineral species and their distribution. In the past decades, many geoscientists have been using Mindat data, but an open data service has never been fully established. The machine interface described in this paper will be an efficient way to meet the overwhelming data needs.

Using geometric features of synthetic triangulated models of subsurface homoclinal interfaces, we applied machine learning to detect faults. Testing on real borehole data validated its effectiveness across various fault orientations. The supervised approach represents a significant improvement over older methods that relied on simpler clustering techniques which were capable of identifying fewer orientations of potential faults.

This study presents FINAM (FINAM is not a model), a new coupling framework written in Python to dynamically connect independently developed models. Python, as the ultimate glue language, enables the use of codes from nearly any programming language like Fortran, C++, Rust, and others. FINAM is designed to simplify the integration of various models with minimal effort, as demonstrated through various examples ranging from simple to complex systems.

Low-level jets (LLJs) are strong winds in the lower atmosphere that are important for wind energy as turbines get taller. This study compares a weather model (WRF) with real data across five North and Baltic Sea sites. Adjusting the model improved accuracy over the widely used ERA5. In key offshore regions, LLJs occur 10–15 % of the time and significantly boost wind power, especially in spring and summer, contributing up to 30 % of total capacity in some areas.

Effectively using modern supercomputers requires massively parallel algorithms. Time-parallel algorithms calculate the system state (e.g. the atmosphere) at multiple times simultaneously and have exciting potential but are tricky to implement and still require development. We have developed software to simplify implementing and testing the ParaDiag algorithm on supercomputers. We show that for some atmospheric problems it can enable faster or more accurate solutions than traditional techniques.

We introduce a new simulation platform based on the Dutch Atmospheric Large-Eddy Simulation (DALES) to simulate carbon dioxide (CO₂) emissions and their dispersion in turbulent environments at a hectometer resolution. This model incorporates both anthropogenic emission inventories and online ecosystem fluxes. Simulation results for the main urban area in the Netherlands demonstrate the strong potential of DALES to improve CO₂ emission modeling and to support mitigation strategies.

Understanding horizontal groundwater flow is important for understanding how water moves through the ground. Current climate models often simplify this process because they do not have information about the land surface that is detailed enough. Our study developed a new model that divides the land surface into hillslopes to better represent how groundwater flows. This model can help improve predictions of water availability and how it affects ecosystems.

This study evaluates various machine learning and statistical methods for interpolating turbulent heat flux data over the Tibetan Plateau. The Transformer model showed the best performance, leading to the development of the Transformer_CNN model, which combines global and local attention mechanisms. Results show that Transformer_CNN outperforms the other models and was successfully applied to interpolate heat flux data from 2007 to 2016.

Climate change has increased the likelihood of drought events across Europe, potentially threatening the European forest carbon sink. Dynamic vegetation models with mechanistic plant hydraulic architecture are needed to model these developments. We evaluate the plant hydraulic architecture version of LPJ-GUESS and show its ability to capture species-specific evapotranspiration responses to drought and to reproduce flux observations of both gross primary production and evapotranspiration.

We develop the data-driven method of dynamic mode decomposition for producing a robust and stable surrogate reduced-order model of atmospheric chemistry dynamics. The model is computationally efficient, provides interpretable patterns of activity, and produces uncertainty quantification metrics. It is ideal for the forecasting of atmospheric chemistry in a computationally tractable manner.

We developed a hierarchy of idealized deterministic–stochastic models to simulate sea surface currents in the Gulf of Trieste. They include tide- and wind-driven sea surface current components, resolving the slowly varying part of the flow, and a stochastic signal, representing the fast-varying small-scale dynamics. The comparison with high-frequency radar observations shows that the non-Gaussian stochastic model captures key dynamics and mimics the observed fat-tailed probability distribution.

The Vegetation Photosynthesis and Respiration Model (VPRM) estimates carbon exchange between the atmosphere and biosphere by modeling gross primary production and respiration using satellite data and weather variables. Our new version, pyVPRM, supports diverse satellite products like Sentinel-2, MODIS, VIIRS, and new land cover maps, enabling high spatial and temporal resolution. This improves flux estimates, especially in complex landscapes, and ensures continuity as MODIS nears decommissioning.