Bridging data and delivery: The world is evolving towards an increasingly data-centric ecosystem. Open Data and Open Source have become widespread and partly mandatory across Europe. The challenges of managing growing data volumes and diverse use cases are increasing. Using cloud technologies to distribute and process data has become a common way to address these challenges. To ensure robust operational reliability and true interoperability, appropriate choices of data-management technologies, data formats, software architectures, APIs, and governance processes are essential.
We invite your contributions for these technical topics, user-focused stories, and your thoughts and strategies. Don't spare us from technical details!
Here are some examples:
- Initiatives designed to bring users closer to the data and to enable the processing of large data sets, such as the European Weather Cloud (EWC) [1] and WEkEO [2]
- New Open Data sets, including hosting Open Data on-premises and in the cloud
- Use of clouds to distribute and process data, including cloud-native data formats
- Tools and interfaces (APIs) for distributing, accessing, and exploiting data
- Building data pipelines for open and hybrid access via APIs, in clouds and on-premises
- Operational aspects of data sharing and data-lifecycle management
- Data visualisation and user-facing systems
- Collaborative open-source software development and sustainability
- Integrating AI/ML data pipelines into operational workflows
- Community building and collaboration around data and open-source software
- Data governance topics, policies, and practices
- any other technical details you feel are important to share
[1] https://europeanweather.cloud
[2] https://wekeo.eu
The integration of AI-based models into operational weather forecasting requires not only advances in model development, but also robust, scalable, and reproducible production environments. At the Finnish Meteorological Institute (FMI), we have established an operational framework for running AI-driven forecasting systems using in-house GPU resources, combined with modern software engineering practices and cloud-native orchestration.
This presentation focuses on the operational inference pipeline of AI-based Aila model, developed within the Anemoi framework as part of the ECMWF Machine Learning Pilot Project, in collaboration with ECMWF and member states. Inference is executed on dedicated GPU infrastructure, enabling both deterministic and ensemble forecasts at high spatial and temporal resolution. We describe the computational requirements of running these models in practice, including GPU allocation strategies, resource utilization, and runtime characteristics, as well as trade-offs between resolution, ensemble size, and latency in an operational setting.
The inference pipeline runs in a containerized environment, with ecFlow handling scheduling and the jobs themselves are executed on OpenShift. This enables scalable, fault-tolerant deployment that integrates with FMI's production systems and is easily reproducible across platforms. Model execution is coupled with data ingestion, post-processing and model verification, forming an end-to-end workflow suitable for real-time forecasting.
Supportive tools will also be addressed, along with how the model and code lifecycle is managed. MLflow is used to handle model versioning, configurations, and metadata, ensuring traceability and reproducibility across operational runs. In addition, deployment workflows and operational configurations, such as YAML files defining inference setups, are maintained through GitHub-based version control and continuous integration practices.
We also discuss practical challenges encountered in operational AI inference, including dependency management, consistency between development and production environments, inconsistency between training and operation environment, and monitoring of system performance.
How to cite: Laine, O.: Operational inference of AI-based weather models at FMI , EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-357, https://doi.org/10.5194/ems2026-357, 2026.
Earth observation workflows for producing analysis-ready data products must balance three competing requirements for operational readiness: consistency, cloud-scale processing, and a simple user experience. In practice, these goals often conflict. Systems focused on operational control can become rigid, while those optimised for extensibility can become inconsistent, non-standard, or difficult to operate.
Here, we introduce Firecube, an open-source ingestion framework for building analysis-ready data cubes in cloud-native environments. Firecube is designed around a stateless CLI that integrates with external orchestrators and exposes modular runtime and plugin interfaces. It provides a storage abstraction across local disk and S3-compatible object storage, and supports cloud-native formats including Zarr and Parquet. Firecube is designed for batch-oriented workflows and provides shared ingestion semantics, configurable parallel batch execution, observability, and a write-ahead log (WAL) backed chunk-management control plane. Rather than bundling product logic itself, Firecube provides the runtime, control-plane tooling, and plugin interface that allow users to bring their own product-specific ingestors while preserving consistent behaviour for batching, appends, partial writes, resume, and write-domain coordination. This keeps semantics consistent across plugins even when they target different EO products, ingestion patterns, and cloud-native formats.
Firecube relies on two key design choices. The first is ChunkManager, the public facade for chunk state. Rather than relying on a shared mutable manifest, Firecube uses a product-local control-plane root in which the write-ahead log is authoritative, while snapshots provide an efficient current-state view, and claims represent write-domain coordination state. This supports safe re-ingests, cleanup, lifecycle tracking, and operational tooling. The second is that Firecube remains orchestrator-agnostic. Firecube acts as an ingestion layer rather than a workflow engine and relies on external cloud-native orchestrators, such as Kubeflow and Argo Workflows, for scheduling, workflow orchestration, dependencies, retries, and resource allocation. This separation creates a portable ingestion layer without coupling product logic to a single scheduler.
We illustrate this with practical examples showing how Firecube can transform Earth observation products into analysis-ready datasets and manage their lifecycle in operational environments.
How to cite: Karatosun, A. and Erdmann, A.-L.: Firecube: Plugin-driven framework for turning Earth Observation products into analysis-ready data cubes., EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-421, https://doi.org/10.5194/ems2026-421, 2026.
Combining 2-D map data with 3-D satellite image timeseries and 4-D atmospheric timeseries naturall lead to datacubes as a common paradigm allowing for a unified treatment and fusion. Therefore, datacube services, pioneered by the rasdaman team, have received increasing attention recently. In comparison to the zillions of NetCDF or GRIB2 files, such datacubes offer more suitable abstractions, intuitive for humans and with APIs convenient for tools.
Suitability of datacubes for value-adding services has been demonstrated in the Cube4EnvSec (Datacubes for Environment and Security) project, an exemplary service for aviation safety has been established. Based on pilot-relevant weather forecasts obtained periodically every 6 hours from German Weather Service (DWD), forecast datacubes are built up and continually extended which have a spatial footprint covering Europe and surroundings and a temporal resolution of 1 hour at a length of meanwhile 3 years. New incoming 72h forecasts overwrite existing timeslices, extend with new timeslices, and keep the not overwritten timeslices, so the nowcasts, which constitute the long-tail timeseries. Variables accommodated include wind speed, temperature, icing, tropopause, volcanic ash and dust, and more. Additionally, via federation the lightning strike datacubes fed from observations by the World Wide Lightning Location Network (WWLLN) - a service built under the Colin Price research group at Tel Aviv University - are available, likewise with a temporal resolution of 1h. Altogether, the service currently holds 28 TB of datacubes, linked into the global EarthServer federation where ad-hoc fusion with, e.g., the Sentinel satellite timeseries of CoperniCUBE is possible.
Recently, in the FAIRgeo project advanced AI support is being added, making AI on datacubes simpler (e.g., 2 lines of query instead of 100 lines of Python code), safer (by enabling the datacube server to estimate accuracy of applying a model to the datacube region and time chosen), and faster (by integrating processing with the highly effective query optimization techniques of the datacube engine).
Access is based on OGC standards, thus enabling direct access through a wide spectrum of third-party clients. The most powerful way is the OGC Web Coverage Processing Service (WCPS) geo datacube query language which is on a high semantics level, thereby abstracting away all the coding complexities incurred, e.g., with Python.
Supported by live demos, we give a technical deep-dive showing key aspects of the aviation datacube service based on geo datacube queries based on the OGC/ISO WCPS standard, including: 4-D corridor datacube queries; zero-coding datacube interaction; dashboard-based datacube visualization; inference with 3rd party ML models on datacubes; cloud/edge integration; and more.
How to cite: Baumann, P.: Weather Datacubes in Action: An Aviation Safety Use Case, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-753, https://doi.org/10.5194/ems2026-753, 2026.
Delivering hyper-local weather forecasts on mobile typically places a heavy burden on backend infrastructure: point-based post-processing services must either pre-compute forecasts for large sets of locations or respond to on-demand requests at scale. Both approaches carry growing infrastructure costs and require users to share their location with a server. We propose an alternative architecture that inverts this model: a lean backend paired with an intelligent client that runs post-processing entirely on device.
European weather data is anything but uniform: providers publish in different file formats or APIs, on different schedules, and in different map projections. Our backend ingests data from 15 services and normalizes them into a single, unified archive, stored in a cloud-optimized Zarr format for fast timeseries access at scale. Map products are pre-rendered as compressed raster tiles.
Our post-processing pipeline runs on the edge: We leverage the power of modern mobile devices to combine dynamic weather data with a static, high-resolution digital elevation model (DEM). This approach has several benefits:
1. Small data transfers. Very high resolution weather map animations can be achieved through on-device upsampling and interpolation without requiring prohibitively large data transfers. For example, we demonstrate how temperature maps with kilometer-scale resolution can be upsampled in real-time by on-device GPU’s to create a more realistic, interactive temperature field with a resolution of tens of meters.
2. Distributed compute. Hyper-local weather forecasts are usually achieved through post-processing forecasts for a large set of pre-defined points of interest or running post-processing on demand as a point-based backend service. Running post-processing on device achieves the hyper-local quality while keeping infrastructure costs managable.
3. Privacy. Being fully aware of all data ingested into our Zarr archive, our client chooses which data to load and autonomously runs post-processing for their exact location without it ever leaving the device.
Together, these properties enable a class of weather application that was previously out of reach: one that is simultaneously hyper-local, privacy-preserving, and visually rich — without demanding heavy server-side compute. We present Fluid Meteo as a working implementation of this approach, and discuss the practical trade-offs encountered in building and operating it.
How to cite: Dietlicher, R., Alig, M., Crezee, B., Sparber, D., Thomas, S., and Wellig, M.: Weather on the Edge: Client-Side Post-Processing For Mobile Weather Applications, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-813, https://doi.org/10.5194/ems2026-813, 2026.
Germany’s new natural hazards portal for the general public, “Naturgefahrenportal” (NGP) was released in April 2025. The NGP aims to consolidate various sources of hazard-related data onto a single accessible platform, promoting public understanding and preparedness. To this end, the NGP website presents current warnings for Germany, location-based hazards and risks as well as concrete measures for preparedness and behavior in a case of emergency.
The NGP serves as a central hub of information for the general public, aggregating the most relevant information in a uniform way, while linking to regional or single-hazard portals for further interest. It does not provide push notifications but complements existing public alert systems with a situational overview for Germany and background information on natural hazards.
This presentation will provide a technical overview of the NGP web app developed at the German National Meteorological Service Deutscher Wetterdienst (DWD) - from data sources and code to deployment pipelines and cloud services. The data sources for natural hazards information in Germany are distributed over several federal institutions at different administrative levels. The data has to be collected via APIs or received via SFTP, and the contents is prepared for display in a portal for the general public in a backend in python. Local repositories at DWD define Infrastructure as Code for the web app, which is deployed in a public cloud. The NUXT/Vue.js Frontend delivers the static website via a Content Delivery Network (CDN).
A brief section provides information on why a public cloud was chosen as platform and how the team work is organized.
How to cite: Riede, H., Brendel, C., Erhardt, B., Hafer, M., Haller, M., Hueser, I., Koziar, C., Lengfeld, K., Rode, D. K., Rauthe-Schöch, A., Reetz, B., and Walawender, E.: The Making Of Germany's Natural Hazards Portal - code, pipelines, and cloud services, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-796, https://doi.org/10.5194/ems2026-796, 2026.
Extreme value analysis (EVA) is widely used to derive thresholds for flood and drought monitoring and anticipatory action, but its application at high spatial and temporal resolution is constrained by data access and computational limitations. This work presents a cloud-native approach for performing pixel-wise EVA on large meteorological datasets, highlighting the role of cloud-optimised data zarr formats, chunking strategies, and parallel computation.
Two datasets are used: NOAA CDR CMORPH rainfall at 30-minute resolution (1998–2024) and ECMWF ERA5-derived SPI and SPEI indices (1940–2025). Both provide long time series at fine grid resolution (0.08° and 0.25°), which are required to estimate return periods and derive actionable thresholds. The primary challenge lies in efficient data access and processing rather than in the statistical methods themselves.
Traditional HPC workflows rely on domain decomposition distributed across processors via MPI or OpenMP. The cloud-native approach presented here follows a similar pattern but shifts the optimisation focus from compute parallelism to data layout and streaming efficiency. By aligning chunking strategies with the temporal requirements of EVA, the workflow enables efficient access to long time series without repeated data loading or complex file orchestration.
The workflow uses cloud-based object storage and Dask distributed computing. Datasets are accessed through cloud-native formats: Kerchunk-based references generated via VirtualiZarr, Parquet catalogs, and Zarr-compatible Icechunk stores. Initial ingestion uses spatial-first ("pan") chunking within an Icechunk data store, which is efficient for spatial subsetting. This layout is rechunked into time-contiguous ("pencil") chunks using Dask's peer-to-peer shuffle-based rechunking, which redistributes data directly between workers with stable memory usage — eliminating separate preprocessing workflows.
For both CMORPH and ERA5 datasets, threshold estimation completes within hours using modest cloud resources. The workflow minimises repeated data scanning and limits data movement by collocating computation with storage. Results demonstrate that chunking strategy — particularly the transition to time-contiguous pencil chunks — is a critical enabler of scalable extreme value analysis for operational climate services in East Africa.
How to cite: Kalladath, N., Mwanthi, A., Koros, H., Sikolia, C., Gudoshava, M., Otieno, G., Ouma, J., and Amdihun, A.: Pencil-chunking for cloud-native extreme value analysis of rainfall and drought thresholds, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-820, https://doi.org/10.5194/ems2026-820, 2026.
Reliable precipitation information underpins flood risk management, water resources planning, and climate risk assessment, yet remains highly uncertain. Precipitation information is derived from gauges and satellite or reanalysis products, each with complementary strengths. Gauge-independent merging provides a rigorous theoretical framework for combining multiple gridded products but cannot directly incorporate point gauges because of scale mismatches. Consequently, no unified framework exists that (i) integrates multiple satellite and reanalysis datasets in ungauged regions, (ii) incorporates gauge information respecting spatial scale differences, and (iii) delivers user-oriented outputs at a catchment scale without requiring detailed expertise in merging methodologies.
Here we introduce the RainMerge framework, which advances multi-source precipitation merging through three conceptual advances. First, RainMerge adopts a two-stage strategy in which rain occurrence is merged as a binary rain or no-rain product and subsequently used to condition the merging of precipitation magnitudes. Second, where ground observations exist, rain gauge uncertainty is quantified and spatially transported to nearby gauge-sparse and ungauged regions using geostatistical interpolation, enabling uncertainty-informed merging across all grid cells. Third, RainMerge is implemented as an automated framework that minimises subjective methodological decision making, allowing users to define catchment boundaries and obtain a state-of-the-art spatially distributed sub-daily precipitation product.
The framework is evaluated using multiple precipitation datasets and hydrological simulations. Results show improved precipitation realism, and more consistent hydrological simulations across contrasting gauge-density regimes. By integrating merging theories, uncertainty estimation, and reproducible practice, RainMerge supports transparent and transferable precipitation information for hydrological modelling, climate analysis, and decision making especially for developing countries, which are impacted by data-sacrcity.
RainMerge can be accessed from https://www.rainmerge.tech.
How to cite: Shah, S., Liu, Y., Kim, S., and Sharma, A.: RainMerge: An uncertainty-aware global framework for merging gauge, satellite, and reanalysis sub-daily precipitation , EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-322, https://doi.org/10.5194/ems2026-322, 2026.
The Finnish Meteorological Institute (FMI) is modernizing its data processing workflows to address increasing data volumes and evolving user requirements. Current post-processing relies on a well-established C++-based system, where limited development capacity has become a bottleneck for extending functionality and supporting modern high-resolution and ensemble datasets. At the same time, the current setup restricts development of data processing workflows to a small group of specialists, limiting the ability of meteorologists, scientists, and other domain experts to directly contribute to algorithm development and data products.
To support scalable data processing and delivery, we are developing a Python-based platform built on the Dask ecosystem. Using tools such as Xarray and Zarr, the system enables parallel, distributed processing of large gridded datasets while remaining accessible to both developers and domain experts. The platform is designed to support diverse use cases, including operational product generation, algorithm development, and future machine learning workflows.
The architecture is designed around cloud-native principles, with initial deployment on FMI's OpenShift platform. Data workflows emphasize cloud-friendly formats, with Zarr as a candidate internal format for scalable processing. In addition to georeferenced outputs such as Cloud Optimized GeoTIFF for visualization and distribution, the system must also support efficient generation of time series products for downstream applications and users. Support for additional input formats, as well as integration with FMI’s existing data servers and APIs (e.g. OGC EDR), is under active development.
A prototype workflow using selected FMI datasets demonstrates how distributed processing and modern data formats can improve scalability, streamline data pipelines, and improve integration between data production and downstream services and users, while lowering the barrier for domain experts to directly contribute to data processing and product generation.
How to cite: Oksman, S., Visa, M., and Rauhala, M.: Building Scalable Meteorological Data Pipelines with Dask, Zarr, and Cloud-Native Infrastructure, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-544, https://doi.org/10.5194/ems2026-544, 2026.
Providing efficient, web-friendly access to meteorological data is an ongoing priority for national weather services. With the Environmental Data Retrieval (EDR) API standard, the Open Geospatial Consortium (OGC) offers a practical approach to modernizing data delivery, allowing users to query precise spatial and temporal subsets without the need to download large, monolithic files or familiarize themselves with handling specialized meteorological file formats. To leverage these benefits specifically for Numerical Weather Prediction (NWP) data, the German Meteorological Service (Deutscher Wetterdienst, DWD) is actively developing an OGC EDR API implementation to deliver its Open Data in a more modern way.
We outline the recent progress, architectural choices, and technical challenges encountered during the development of the API. To build a minimum viable product, we focused on DWD's high-frequency forecast model ICON-D2-RUC (Rapid Update Cycle) and its ensemble counterpart, ICON-D2-RUC-EPS. As the API response format, we provide CoverageJSON. However, working within this standard required us to address the specific structural complexities of the ICON models. This involved solving the challenge of presenting parameters that vary vastly in temporal resolution, ranging from 5-minute intervals to hourly steps, within a single response. Beyond these challenges related to temporal resolution, we also had to address the spatial representation. By utilizing a MultiPolygonSeries domain within our CoverageJSON output, we maintain the native ICON triangular grid, providing a standardized method to represent the model cells without requiring interpolation.
This project is a collaboration between DWD's open data team and the team for digital transformation of data storage and data delivery at DWD. Together, we built a solution relying on our new data lakehouse architecture. It features Zarr, which is a storage format providing highly efficient access to multidimensional gridded data, alongside Icechunk for safe transactions in our distributed architecture. This cloud-optimized storage layer uses object storage hosted within a sovereign cloud and is continuously populated with operational data, forming the foundation of our current internal minimum viable product. Drawing from this experience, in this contribution we will share practical insights gained from building this service and discuss our ongoing roadmap, including the status of a public alpha release.
How to cite: Riedl, K., Guntrum, M., and Reetz, B.: Modernizing Open Data Access at DWD: Implementing an OGC EDR API for Spatiotemporal Subsetting of NWP Data, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-637, https://doi.org/10.5194/ems2026-637, 2026.
Accessibility, interoperability, and sustained availability of hydrological and climatological data are critical for effective climate services. This is especially true in regions where data scarcity and operational challenges limit the development of sustainable early warning systems.
Building on this need, this work presents an open-source, interoperable hydrological web platform designed to support National Meteorological and Hydrological Services (NMHSs) in the Sahel. The SLAPIS (Système Local d’Alerte Précoce contre les Inondations de la Sirba) web platform, first developed in Niger, has evolved into SLAPIS Sahel, a transboundary hydrological climate service jointly implemented in Niger and Burkina Faso. The system provides the access, processing, and dissemination of near-real-time observations and hydrological forecasts through standardised web services and APIs, fostering data reuse and integration into operational Flood Early Warning Systems (FEWS). This evolution enables authorities in both countries to address cross-border flood risks, coordinate responses, and access consistent data for regional planning and flood management.
The use of interoperable services and automated data pipelines allows seamless integration of multiple data sources and facilitates the delivery of user-oriented hydrological services.
The system integrates ground observations and interoperates via APIs with existing large-scale hydrological forecasting platforms, including GloFAS 4.0, HYPE-based models, and others whose integration is currently being finalised. Automated pipelines ingest these diverse data sources by standardising their formats and timing into a single, locally useful layer. Data pipelines are written in Python and run without manual intervention. The web front-end uses Leaflet.js for interactive mapping and vanilla JavaScript for real-time chart rendering. Processed data sets are available through a customised CKAN-based catalogue and exposed via RESTful APIs in CSV, JSON, GeoJSON, and OGC-compliant WFS formats. The lightweight architecture enables low-cost deployment and maintenance, which is vital for sustainability in resource-constrained settings. The platform uses only open-source libraries, with no proprietary components.
The platform's core feature is impact-based flood forecasting. It compares forecasted discharge with locally defined thresholds—derived from hydraulic simulations and fieldwork—to generate flood scenarios and hydrological information on 11 localities. Next, hydrometeorological data are linked to these thresholds, flood-prone areas, and associated impacts. As a result, NMHSs can assess the hydrological situation, view forecasts and historical data sets, and download relevant datasets. This improves the relevance of information for decision-making and risk management.
The deployment of SLAPIS Sahel in the Sirba River basin exemplifies how open, interoperable data services and collaborative development directly reinforce sustainable, cross-border hydrological service delivery. Furthermore, sharing ground observations strengthens model calibration and provides the local data needed to refine flood thresholds. This approach is easily replicable: the same pipeline architecture can integrate forecasts from various global hydrological models, making it applicable even in basins where local models are not available.
How to cite: De Filippis, T., Gioia, A., Rocchi, L., and Tarchiani, V.: Leveraging Open Data, Open-Source Tools, and Interoperable Web Services for Transboundary Hydrological Climate Services in the Sahel, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-704, https://doi.org/10.5194/ems2026-704, 2026.
Visualization plays a central role in the analysis of numerical weather prediction data for both research and operational forecasting. In recent years, the open-source visualization framework Met.3D (https://met3d.readthedocs.io) has been developed to enable interactive, three-dimensional, and feature-based exploration of gridded atmospheric data. Since its first public release in 2015, Met.3D has evolved into a feature-rich visual analysis tool facilitating rapid exploration of atmospheric datasets. It is designed to bridge established 2-D visualization techniques commonly used in meteorology with interactive 3-D analysis by combining both within a unified visual context. Following the introduction of Met.3D version 2.0, which improved usability and functionality, we now focus on advancing the software towards a mature research software system and demonstrating its applicability in real-world forecasting environments.
We present recent developments that enhance the robustness, portability, and reproducibility of Met.3D. A central contribution is the introduction of a reproducible CI/CD workflow. It generates test datasets (synthetic and from open sources), executes Met.3D in batch mode to produce visualization outputs, and compares the resulting images against reference baselines using image-based similarity metrics (e.g. SSIM). This enables automated evaluation of visualization results and ensures consistent behaviour across heterogeneous environments. Further developments include extended cross-platform support (Linux and Windows), improved memory management, enhancements to the rendering pipeline, and refinements to usability and user interfaces.
Many of these developments were driven by user feedback from real-world applications. For example, Met.3D was used during the North Atlantic Waveguide and Downstream Impact Campaign (NAWDIC) in a forecasting context for research flight planning under various constraints.
The presented developments lower the barrier for adoption in operational and educational settings and contribute to sustainable open-source software development in the meteorological community.
How to cite: Fischer, C., Vogt, T., Radke, T., Hartz, M., and Rautenhaus, M.: From research software to operational forecasting: Advancing Met.3D for interactive 3-D meteorological analysis in field campaigns, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-375, https://doi.org/10.5194/ems2026-375, 2026.
Downscaled multi-model ensemble mean and standard deviation through empirical-statistical downscaling have been gridded to a 5 arc-minute spatial resolution (~5km) for key seasonal temperature and precipitation statistics for the Barents Sea region, the Nordic countries, Europe, and other parts of the world. The downscaling has made use of a common EOF basis for ensuring a good representation of the predictors in terms of their spatio-temporal covariance structure, and for facilitating a quality control of the predictors. The downscaled key statistical parameters provide information about daily variables, such as the number of days with heavy precipitation. One objective has been to derive robust and reliable data, and ensemble sizes of hundreds of runs based on multiple global climate models provide a basis for robust estimates, since the addition of exclusion of a few runs don’t change the results significantly. The downscaled results also provide reliable information on temperature and precipitation, and the reliability of the results is tested against ensemble spread properties as well as the consistency between the ensemble members and the observed evolution for a common historical period. Three diagnostics are included in the netCDF files which indicate how close the ensemble spread is to being normal, how similar the historical trends are, and how the observed interannual variability compares with the ensemble range. The data include CMIP6 SSPs 1-26, 2-45, 3-70, 5-85 as well as CMIP5 RCPs 45, 26 and 85. Details about the predictors and predictands are provided in the netCDFfile header. The data can be openly accessed through OpenDap/thredds as well as FigShare, and it provides a basis for regional climate information for society (RIfS) as well as climate change impact studies. This data represents a new contribution from the CORDEX community.
How to cite: Benestad, R.: New gridded large multi-model ensembles of empirical-statistical downscaling results , EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-443, https://doi.org/10.5194/ems2026-443, 2026.
Supporters & sponsors
