German Climate Computing Centre

Large language models (LLMs) have the potential to transform how climate scientists interact with data by lowering technical barriers and enabling more intuitive analysis workflows. Building on previous demonstrations of LLM-assisted climate analysis, we present how FrevaGPT, an LLM-powered scientific assistant integrated into Freva - a climate data search and analysis platform- , supports climate scientists in their day-to-day data exploration and analysis. FrevaGPT interprets natural language queries and automatically generates traceable, editable, and reusable analysis scripts that can be executed within established scientific environments. It retrieves relevant datasets and literature, performs analyses, and visualises results, therefore allowing researchers to focus on scientific interpretation rather than coding intricacies. By leveraging a broad repository of climate observations and model output, FrevaGPT ensures transparent and reproducible workflows that adhere to best practices in climate research. It also integrates seamlessly into Jupyter-AI and, by making use of the Freva library, combines the code-generating capabilities of LLMs with contextual understanding of how to access relevant datasets on the HPC cluster. As a “co-pilot” for geoscientists, the system not only responds to explicit requests but also proactively suggests relevant climate modes, events, and next analytical steps, helping to uncover insights that might otherwise be overlooked. Practical use cases demonstrate how FrevaGPT assists with interactive exploratory analysis and hypothesis refinement across climate datasets of varying complexity. By embedding LLM-assisted natural language interaction into real-world climate research workflows, this work highlights methodological considerations and opportunities for enhancing scientific productivity, promoting broader adoption of NLP and AI tools among Earth system scientists. We provide scientific evaluation of FrevaGPT’s capability through a benchmark suite. A live demo will be presented and can be used by the audience to do real climate analysis on a high-performance computer with access to petabytes of Earth system data - starting with a simple prompt.

Climate science enterprise both produces and depends on extremely large datasets in order to meet the needs of diverse scientific and downstream user communities, especially as climate models are increasingly run at kilometre-scale resolutions, resulting in rapidly growing data volumes which increase demands on data handling infrastructures. Individual flagship simulations are no longer used by a single research group, but are routinely reused by dozens or even hundreds of researchers globally. Consequently, data findability, accessibility and reuse must be straightforward, data provenance must be transparent, and the full heritage of simulation data should be preserved in a machine-actionable manner to ensure scientific rigour, explainability and reproducibility.In this contribution, we present a conceptual infrastructure-level approach developed within the WarmWorld project based on leveraging the versatility of globally unique persistent identifiers (PIDs) to address these challenges. Specifically, we illustrate that by assigning handles to simulation datasets already at the point of production, simulation data stored locally at a HPC data center can become part of a globally interoperable data ecosystem. In our concept, handle profiles contain an URL at which the dataset can be opened. Further, machine-actionable metadata, such as the detailed provenance information describing the employed model configuration or a data reuse license and citation, would be available from the handle landing page. Thus, the motivation behind the approach we follow here is akin to that of the FDO specifications.Finalized simulation datasets would be exposed through globally accessible SpatioTemporal Asset Catalogs (STAC), where PIDs serve as the authoritative entry point for discovery and access. Data access would be handled by system libraries that resolve storage locations across heterogeneous storage tiers. Crucially, data access shall be designed to be globally open without the need for credentials, reflecting a strong demand from the climate research community, as clearly demonstrated during the WCRP kilometre-scale hackathon (May 2025).Systematic assignment and pragmatic leveraging of handles assigned to locally stored datasets can thus enable scalable and interoperable access to flagship climate datasets across infrastructures and communities, effectively integrating traditionally closed HPC data environments into the global data space and facilitating interoperability with other large-scale data holdings.

CLN3 disease, also called Juvenile Neuronal Ceroid Lipofuscinosis (JNCL), or Batten disease, is an ultra‑rare, neurodegenerative lysosomal storage disorder generally affecting individuals during the first decade of life. There can be a delay in diagnosis or misdiagnosis due to a lack of awareness, and when the most common presenting symptom of visual loss is attributed to more common conditions affecting vision. We used a previously published Expert Mapping Tool (EMT) to identify multidisciplinary professionals with diagnostic or clinical management expertise, as well as patient advocates with experience of CLN3 disease. A systematic literature review of published evidence using the Preferred Reporting Items for Systematic Reviews and Meta‑Analyses (PRISMA) guidance was conducted independently and simultaneously to develop key clinical care statements. Each statement was based on the strength of the evidence. The statements formed the basis of an international modified-Delphi consensus process using a virtual meeting platform (Within3). Experts were asked to agree or disagree with each statement and suggest any changes. Statements that reached a consensus of 75% or over are the guiding statements within this manuscript. The processes and manuscript have been independently assessed using the Appraisal of Guidelines for Research and Evaluation (AGREE II) criteria. Thirty‑nine international experts from eight specialities were identified, including a patient advocate. Fifty‑three recommendation statements were developed covering eleven domains: General statements, Diagnostics, Clinical Recommendations and Management, Assessments, Social Considerations, Ocular Management, Epilepsy/Seizures, Nutrition, Respiratory Health, Sleep and Rest, and End-of-Life Care. Consensus was reached after one round of voting for all except three statements. The overall AGREE II score for developing these recommendations was 6.4, where 1 represents the lowest and 7 is the highest quality. Currently, there are no comprehensive clinical recommendations for CLN3 disease. These recommendations provide a comprehensive, evidence- and consensus‑based tool that can be used by all healthcare professionals involved in the management of CLN3 disease and other similar neurodegenerative conditions. The goal is to address the unmet clinical need for CLN3 disease management and complement other information available.

ObjectivesHepatitis‐associated aplastic anemia (HAAA) is described as acute severe hepatitis of unknown origin followed by bone marrow failure (BMF). We aimed to provide a comprehensive picture of pediatric HAAA.MethodsTwo‐center retrospective analysis was performed using data from children diagnosed with acquired BMF, including severe aplastic anemia (SAA) and myelodysplastic syndrome type refractory cytopenia of childhood (RCC). The assessment of the subcohort of HAAA included clinical features indicative of diagnosis and disease progression, with additional data from previously published case series.ResultsCohort comprised 62 children with acquired BMF and 22 children with HAAA. Median age of HAAA patients was 13.5 years. Potentially triggering viral infections were detected in 45%. The median interval from hepatitis onset to cytopenia was 3 weeks. All cases presented with severe hepatitis (median alanine transaminase 2127 U/L) and all but one with hyperbilirubinemia (median bilirubin 15.3 mg/dL). Coagulopathy was variable (median international normalized ratio 1.5). Four patients (18%) developed acute liver failure, two (9%) required liver transplantation. Hepatic parameters normalized within a median of 8.5 weeks. There was no statistically significant difference in the course of hepatitis between patients with SAA and RCC. Early lymphopenia was a key finding in patients with HAAA, progressing from a median of 905/µL at hepatitis onset to 530/µL within 4 weeks.ConclusionsHAAA occurs in both SAA and RCC. Most cases present with severe acute cholestatic hepatitis and variable coagulopathy. Hepatic recovery is common. Lymphopenia at disease onset is frequent and may serve as a diagnostic marker.

Cancer immunotherapy is defined as a type of cancer treatment that helps the immune system recognize and attack cancer cells more effectively [...].

BackgroundPersons living with a rare disease (PLWRD) often encounter burdensome and stressful events that may severely affect their psychosocial vulnerability. There is an urgent need for psychosocial support in PLWRD. We aimed to reach consensus about the most prominent psychosocial needs in rare diseases and future research directions to develop adequate psychosocial support for rare diseases.MethodsAn adapted nominal group technique (NGT) session was conducted in Nijmegen (The Netherlands) with a European expert group of professionals working in the rare disease field. Seventeen participants from seven different European countries took part in the NGT session (seven researchers, two therapists, one patient representative, four clinicians/researchers, and three clinicians), and 23 participants took part in the overall workshop and contributed to this consensus statement.ResultsAn initial list of 57 psychosocial needs was aligned with Pickers’ theoretical framework of eight principles of care: patient-centredness, emotional support, access to care, information and education, partner and family involvement, respect and autonomy, care organisation, continuity of care, and physical comfort. For future research directions, six items remained with main focal points addressing international collaborations, inclusivity in rare diseases (patient representatives, ethnic minorities, and elderly), identifying common needs across rare diseases, and translating psychosocial models from common chronic conditions.ConclusionsA consensus meeting was organised that primarily addressed the psychosocial vulnerability in rare diseases. With the outcomes of this study, we aim for better representation of psychosocial vulnerability within health care and to create future research directions to reduce psychosocial vulnerability in rare diseases.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13023-025-04017-3.

Using data from the MERRA-2 database, the study identifies a clear warming trend in August, with significant peaks in the annual average air temperature in 2010 (28.3 °C) and 2021 (27.9 °C) in the Caspian Sea. Concurrently, the Caspian Sea’s water levels have declined by approximately 16 cm annually from 1992 to 2025, totaling a drop of around 2.04 m, driven by higher evaporation rates linked to rising temperatures. Heat extremes are observed, with maximum daily air temperatures reaching 30.3 °C and minimums of 26.8 °C in August 2021, representing deviations of 2.6 °C and 2.4 °C from the climatological mean. The most intense event exceeded 2.63 °C for the maximum 2 m air temperatures in August 2021 and 2.16 °C above the climatological mean for an extended 12 days. Furthermore, the frequency of heatwaves has increased, with two major heatwave events occurring in 2010 and 2021. The study also highlights significant changes in sea surface temperature, Chlorophyll concentrations, and sea salt levels, with higher sea salt concentrations in 2021 correlating with lower Chlorophyll levels. Spatial analysis reveals regional temperature variations, with warmer temperatures in the southeast in 2010 and cooler conditions in the eastern regions in 2021. Additionally, the study observed a shift in the subtropical jet stream, which weakened and shifted eastward in 2021 compared to 2010. The jet stream split around the high-pressure system, transitioning from zonal to meridional flow. These changes in atmospheric circulation, characterized by dry conditions and reduced low-level relative humidity in 2021 (5–10% drier than in 2010), contributed to the extreme heatwave events. The findings highlight the role of atmospheric circulation patterns, including the development of high-pressure systems and shifts in the jet stream, in driving temperature extremes and heatwaves in the Caspian Sea region, with significant implications for future climate variability and ecosystem impacts. The results underline the need for further research and mitigation efforts to address the ongoing warming, heatwaves, and water level decline in the Caspian Sea, which pose significant ecological and socio-economic challenges. Marine heatwaves in the Caspian Sea have been causing significant ecological disruptions.

Abstract. Numerical weather prediction and climate models require continuous adaptation to take advantage of advances in high-performance computing hardware. This paper presents the port of the ICON model to GPUs using OpenACC compiler directives for numerical weather prediction applications. In the context of an end-to-end operational forecast application, we adopted a full-port strategy: the entire workflow, from physical parameterizations to data assimilation, was analyzed and ported to GPUs as needed. Performance tuning and mixed-precision optimization yield a 5.6x speed-up compared to the CPU baseline in a socket-to-socket comparison. The ported ICON model meets strict requirements for time-to-solution and meteorological quality, in order for MeteoSwiss to be the first national weather service to run ICON operationally on GPUs with its ICON-CH1-EPS and ICON-CH2-EPS ensemble forecasting systems. We discuss key performance strategies, operational challenges, and the broader implications of transitioning community models to GPU-based platforms.

<strong class="journal-contentHeaderColor">Abstract.</strong> In a rapidly changing climate, evidence-based decision-making benefits from up-to-date and timely information. Here we compile monitoring datasets (published here <a href="https://doi.org/10.5281/zenodo.15327155" target="_blank" rel="noopener">https://doi.org/10.5281/zenodo.15327155</a> Smith et al., 2025a) to produce updated estimates for key indicators of the state of the climate system: net emissions of greenhouse gases and short-lived climate forcers, greenhouse gas concentrations, radiative forcing, the Earth's energy imbalance, surface temperature changes, warming attributed to human activities, the remaining carbon budget, and estimates of global temperature extremes. This year, we additionally include indicators for sea-level rise and land precipitation change. We follow methods as closely as possible to those used in the IPCC Sixth Assessment Report (AR6) Working Group One (WGI) report. The indicators show that human activities are increasing the Earth&rsquo;s energy imbalance and driving faster sea-level rise compared to the AR6 assessment. For the 2015&ndash;2024 decade average, observed warming relative to 1850&ndash;1900 was 1.24 [1.11 to 1.35] &deg;C, of which 1.23 [1.0 to 1.5]&thinsp;&deg;C was human-induced. The 2024 observed record in global surface temperature (1.52&deg;C best estimate) is well above the best estimate of human-caused warming (1.36&deg;C). However, the 2024 observed warming can still be regarded as a typical year, considering the human induced warming level and the state of internal variability associated with the phase of El Ni&ntilde;o and Atlantic variability. Human-induced warming has been increasing at a rate that is unprecedented in the instrumental record, reaching 0.27 [0.2&ndash;0.4] &deg;C per decade over 2015&ndash;2024. This high rate of warming is caused by a combination of greenhouse gas emissions being at an all-time high of 53.6&thinsp;&plusmn;&thinsp;5.2&thinsp;GtCO₂e per year over the last decade (2014&ndash;2023), as well as reductions in the strength of aerosol cooling. Despite this, there is evidence that the rate of increase in CO₂ emissions over the last decade has slowed compared to the 2000s, and depending on societal choices, a continued series of these annual updates over the critical 2020s decade could track decreases or increases in the rate of the climatic changes presented here.

Abstract. The microscale, obstacle-resolving meteorological transport and stream model MITRAS has been extended with a snow cover and precipitation scheme. The performance of the model extension is assessed by comparing the results of different model versions using a method based on hit rates originally developed for assessing wind performance. For temperature, radiation and precipitation, estimates for the threshold values were derived based on computational accuracy; these are used in the hit rate calculation for these variables. The threshold values for the deviations are 0.02 ms−1 (5 %) for the wind components, 0.05 K (0.02 %) for temperature, 0.5 Wm−2 (0.5 %) and 0.5 Wm−2 (0.2 %) for the net long and short wave radiation, and 0.001 mm (1 %) for precipitation on ground. The model extensions produce plausible results and better represent winter precipitation. This opens the opportunity to study with higher accuracy the influence of obstacles on precipitation heterogeneities.