Regional reanalyses are now available for Europe and specific sub-domains, e.g. produced by national meteorological services. Global and regional reanalyses are an important element of the Copernicus Climate Change Services.
The interest in extracting climate information from reanalysis is rising and they are used in a wide range of applications. In recent years, it has become apparent that reanalyses are a popular basis for training in machine learning methods that enable successful AI-based weather forecasts, for example.
This session invites papers that:
• Present the status of reanalysis activities in Europe and beyond.
• Explore and demonstrate the capability of global and regional reanalysis data for climate applications, including energy applications.
• Illustrate the role of reanalysis data for machine learning and artificial intelligence.
• Compare different reanalysis (global, regional) with each other and/or observations
• Improve recovery, quality control and uncertainty estimation of related observations
• Analyse the uncertainty budget of the reanalyses and relate to user applications
Depending on the submitted contributions, the session could also provide a platform for discussions about the requirements of reanalysis producers towards data providers.
At ECMWF, reanalysis forms a key contribution to the Copernicus Services (C3S, CAMS) that are implemented by ECMWF on behalf of the European Commission. The most recent ECMWF global atmospheric reanalysis, ERA5, provides hourly snapshots of the Earth's atmosphere, land surface and ocean waves from 1940 with daily updates 5 days behind real time. It has hundreds of thousands of users in many sectors in society, ranging from scientific to commercial applications. ERA5 has become a major training dataset for data-driven weather prediction models.
The production of its successor, ERA6, is underway. It benefits from eight additional years of R&D at ECMWF, more and better observations and forcing datasets, and 14km (rather than 31km) horizontal resolution for the atmosphere, land and ocean waves. The model component includes a coupled ocean (at 0.25 degree horizontal resolution), using initial ocean conditions from the ORAS6 ocean reanalysis. ERA6 resolves several ERA5 known issues. This includes taking better care of systematic model errors in the stratosphere, to further improve the quality of climate trends in this domain.
Based on user demand, ERA6 will contain a number of new products. These include 3D ocean fields plus a number of 2D key variables, 3D Clear-Air Turbulence, several new atmospheric surface parameters and a number of key parameters on height levels up to 500 metres from the surface. All fields vary hourly, and in addition to monthly, daily precalculated statistics will be provided as well.
Like ERA5, ERA6 is produced in parallel streams of 10 years each. Such streams from 1987 are produced and made available first, while streams back to around 1950 will be produced later.
This presentation will provide a summary of ERA5, what is new in ERA6 with respect to science and products, will show some initial ERA6 results and will indicate projected timelines.
How to cite: Hersbach, H., Bell, B., Herbert, C., Kaandorp, M., Poli, P., Radu, R., Schepers, D., Simmons, A., Soci, C., Steele, L., Strube, C., Buontempo, C., and McNally, A.: From ERA5 to ERA6, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-251, https://doi.org/10.5194/ems2026-251, 2026.
Climate change is currently one of the main threats our planet is facing. Observations are playing a pivotal role in underpinning the science to understand the climate system and monitor its changes including extreme events, which have adverse effects on human lives. Information generated from measurements by Earth observation satellites contribute significantly to the development of this understanding and to the continuous monitoring of ongoing climate change and its impacts. However, the meaningful use of data from these satellites requires them to be long-term, spatially and temporally homogeneous, and uncertainty characterised. The process of preparing satellite data for climate studies is tedious and only recently being recognised as fundamental first step in preparing records of Essential Climate Variables (ECV) from these data.
EUMETSAT has generated several fundamental climate data records (FCDR) consisting of measurements from instruments operating from microwave to visible frequencies. These measurements are not only from EUMETSAT’s own satellite but also from satellites operated by other agencies such as NOAA, NASA, and JMA. This presentation outlines the basic principles of FCDR generation illustrated through a few examples. Basic steps of the FCDR generation is comprised of quality control of the raw data, recalibration of the raw data to produce physical quantities, such as radiances or reflectance, generate quality indicators, and create the outputs in user-friendly formats, e.g., NetCDF4. Furthermore, harmonisation of a suit of instruments are performed. The presentation will further domonstrate how the assimilation of these data records improve the quality of ERA6, the upcoming version of the European Reanalysis.
How to cite: John, V. O.: EUMETSAT's contribution towards producing climate quality reanalyses, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-519, https://doi.org/10.5194/ems2026-519, 2026.
Surface reanalyses are an important tool for examining past natural hazards, such as wildfires, sudden floods and droughts, in great detail and on a consistent scale. With the surge in deep learning, these reanalyses could contribute to the development of new tools for identifying, characterizing and anticipating such hazards. Météo-France has just conducted a 60-year reanalysis of Western Europe using the AROME numerical weather prediction model, known as ARRA. Here, we use these atmospheric conditions to reanalyse the continental surface of the same area using the Land Data Assimilation System (LDAS-Monde).
ARRA reanalyses are used to drive the Interaction Surface Biosphere Atmosphere (ISBA) land surface model via the SURFEX platform, incorporating prognostic soil moisture, leaf area index and woody biomass at kilometer scale. These modeled variables are then corrected by assimilating Copernicus Land Monitoring Service LAI satellite data and ESA-CCI above-ground biomass satellite data using a simplified extended Kalman filter. The reanalysis period will range between 2000 and 2025 to coincide with CLMS LAI observation availability. Incorporating LAI into ISBA enhances the model's representation of the carbon and water cycles, as well as addressing agricultural practices not currently modeled by ISBA. We will additionally evaluate the impact on river discharge using CTRIP river routing.
The additional focus on biomass in the reanalyses stems from the need to improve the representation of living fuel moisture content and dead biomass in our model, as these are two critical variables for providing an early warning of fire ignition. To this end, above-ground biomass will be assimilated to adjust the distribution of biomass between leaves, wood, and litter.
How to cite: Vanderbecken, P., Baehr, Y., Rojas-Munoz, O., Deferrez, M., Munier, S., and Calvet, J.-C.: Regional land surface variable reanalysis: LDAS-ARRA, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-238, https://doi.org/10.5194/ems2026-238, 2026.
The atmospheric Copernicus European Regional Reanalysis (CERRA) and its land component CERRA-Land, including the MESCAN daily precipitation analysis, were produced from September 1984 to June 2021 and are available through the Copernicus Climate Data Store (CDS). A new service contract, currently underway and funded by the Copernicus Climate Change Service (C3S) under the Copernicus program, has been established to continue producing the existing reanalysis and extend it back in time. The three-hourly output from the CERRA reanalysis, based on the Harmonie-ALADIN 3D-VAR at 5.5 km, is used to generate hourly forcing files - including precipitation analyses using MESCAN - for the CERRA-Land surface reanalysis, an offline simulation of the SURFEX land surface modeling platform. Production of CERRA and CERRA-Land has resumed, reaching near real time, and the back extension production from September 1960 to August 1984 is ongoing.
This work focuses on the MESCAN precipitation analysis, which aims to improve total precipitation estimates from CERRA by assimilating in situ observations. MESCAN assimilates around 8,000 surface observations per day for the present period and around 6,500 observations at the start of the back-extension period. These observations come from the ECA&D, MARS and national databases. Their temporal variability and spatial distribution can impact long-term studies that use this product. That is why we strive to document changes and maintain consistency wherever possible. The quality control measures and characteristics of the CERRA-Land product will be presented. Furthermore, the added value compared to CERRA and its comparison with ERA5 for precipitation will be presented.
How to cite: Martí, B., Verrelle, A., Selly, Y., Glinton, M., Geidne, A., Bazile, E., and Le Moigne, P.: MESCAN precipitation analysis as part of the CERRA reanalysis, for near real time and back extension production., EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-139, https://doi.org/10.5194/ems2026-139, 2026.
In this study, we present a comprehensive evaluation of the Austrian Reanalysis ensemble dataset (ARA). ARA is an 11-member high resolution (2.5 km) regional reanalysis ensemble system developed to provide spatially and temporally consistent 2D & 3D essential climate variables from 2012-2022 with an hourly temporal resolution for greater Alpine region with particular focus over Austria.
ARA ensemble is based on dynamical downscaling of the European Centre for Medium-Range Weather Forecasts ERA5 reanalysis ensemble (31 km resolution) using the non-hydrostatic limited-area model AROME (Application of Research to Operations at Mesoscale). A key component of ARA is the integration of a three-dimensional variational (3DVAR) data assimilation framework within AROME, enabling the incorporation of diverse observational datasets, including satellite, radiosonde, aircraft, and wind profiler measurements. This approach allows for the generation of a dynamically consistent ensemble, capturing both large-scale forcing and small-scale atmospheric variability.
The performance of the ARA system is evaluated across a comprehensive range of synoptic conditions, including extreme precipitation, fog, freezing events, storm situations, and snowstorms. Model outputs are rigorously validated against in-situ station observations and gridded datasets, with operational model outputs used as a reference. The evaluation results show that ARA consistently matches or outperforms current operational models, demonstrating clear added value from high-resolution dynamical downscaling and data assimilation. Results indicate a significant improvement in the representation of mesoscale processes and spatial variability. Quantitative assessment using bias, root mean square error (RMSE), and spatial correlation, combined with detailed two-dimensional spatial analyses at daily timescales, confirms the system’s capability to deliver accurate and physically consistent fields. These findings establish ARA as a state-of-the-art high-resolution reanalysis ensemble, providing significant benefits for climate monitoring and impact-oriented applications in Austria.
How to cite: Awan, N. K. and Scheffknecht, P.: A high resolution regional reanalysis dataset for Austria (ARA), EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-767, https://doi.org/10.5194/ems2026-767, 2026.
Reanalyses are widely used to assess climate variability and long-term trends. In recent years, high-resolution regional and convection-permitting reanalyses have provided new opportunities to investigate climate processes and extremes at local scales. However, recent studies have highlighted potential limitations in the temporal consistency of long-term climate signals derived from ERA5 and its dynamical downscalings, raising questions about their suitability for robust trend detection.
This study assesses the temporal consistency of five state-of-the-art convection-permitting regional reanalyses developed for Italy (MORE, CHAPTER, MERIDA-HRES, VHR-REA_IT, and SPHERA), all produced through dynamical downscaling of ERA5 using different numerical models (MOLOCH, WRF, and COSMO). Long-term seasonal trends in both precipitation and temperature are evaluated over the period 1981–2020, depending on the temporal availability of each individual product, against a homogenized observational dataset (UniMI/CNR-ISAC) specifically designed for climate analysis, ensuring temporal consistency at the national scale.
This comparison enables a detailed quantification of uncertainties in trend estimates across different regions and seasons. Inhomogeneities may originate from the driving ERA5 reanalysis or be further introduced during the downscaling process, for instance through the assimilation of local observations (e.g., observational nudging). Similarities and differences among the various reanalyses are examined in the context of the specific architecture and configuration of each product, highlighting how these factors may affect temporal consistency.
By providing a comprehensive assessment of biases and uncertainties in seasonal precipitation and temperature trends over Italy, this work offers guidance to reanalysis users on the reliability of trend estimates for climate studies and impact assessments. It also provides insights for reanalysis developers aimed at improving the temporal consistency of future regional products. Ultimately, the results contribute to strengthening the observational and analytical foundations of climate services, where reliable reanalysis-based products are essential for monitoring and decision-making.
How to cite: Cavalleri, F., Stocchi, P., Brunetti, M., Manara, V., Maugeri, M., and Davolio, S.: Validating and understanding seasonal precipitation and temperature trends in convection-permitting reanalyses over Italy, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-77, https://doi.org/10.5194/ems2026-77, 2026.
Reanalysis data sets are becoming increasingly popular for a broad spectrum of applications such as climate adaptation and mitigation, renewable energy, agriculture, or hydrology, including the assessment of meteorological hazards and extremes. Recently, the high relevance of reanalysis data sets has increased further due to their value as a basis for training AI-based NWP model emulators such as AICON or AIFS. Most of these applications require much finer grid spacing compared to ERA5 (31 km), ERA6 (14 km), or even ICON-DREAM (dual resolution reanalysis for emulators, applications, and monitoring, 13 and 6.5 km). Therefore, regional reanalysis is designed and produced for limited geographical regions, allowing for the provision of high-quality and high-resolution data sets.
The 2 km ICON-reanalysis based on conventional observations for climate applications in Central Europe (ICON-FORCE-c) has been produced adapting the operational 2 km ICON-RUC (Rapid Update Cycle) numerical weather prediction (NWP) model framework, which uses the 2-moment microphysics parameterization to better represent precipitation processes. Its operational data assimilation cycle comprises the KENDA LETKF-based data assimilation scheme at hourly intervals, complemented by a snow analysis every 6 hours, and T2M, SST and soil moisture analysis every 24 hours (at 00 UTC). The background error covariances are provided by a 20 member ensemble at the same 2.1 km resolution of the deterministic run. The boundary conditions come from the ICON-DREAM European nest domain. The next version of ICON-FORCE-c will also include SEVIRI radiometer satellite data to better constrain the upper atmosphere.
This presentation is a companion work with colleagues from the research and development department of DWD, jointly comparing and evaluating our product with their reanalysis products ICON-FORCE and ICON-DREAM. While the full-input ICON-FORCE regional reanalysis aims to provide the best description of the Earth system, with the modern regional observational network, the sparse input ICON-FORCE-c reanalysis aims to provide the best possible climate state for the area, thus focusing on climate trends and its consistency. We aim to present first evaluation results by comparing ICON-FORCE-c with ICON-FORCE, ICON-DREAM and ERA5.
How to cite: Kelbch, A., Valmassoi, A., Külheim, F., Borsche, M., Spangehl, T., Schomburg, A., and Wahl, S.: Validation results for the 2 km ICON-reanalysis based on conventional observations for climate applications in Central Europe, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-540, https://doi.org/10.5194/ems2026-540, 2026.
High-resolution meteorological data are essential for the energy sector because they are central for the design and operation of renewable‑energy systems. Existing data sets, however, exhibit shortcomings. Within the MEDAILLON project, the German Meteorological Service (DWD), in cooperation with Fraunhofer Institute for Energy Economics and Energy System Technology (IEE), the University of Kassel, and menzio GmbH, uses the ICON-DREAM reanalysis to develop a meteorological data set tailored for Germany’s energy sector. Potential stakeholders were consulted during the development phase to ensure the dataset meets their needs.
The meteorological dataset builds upon the ICON‑DREAM reanalysis – “ICON‑Dual‑resolution Reanalysis for Emulators, Applications and Monitoring” [1] – developed and produced by DWD based on the ICON numerical-weather-prediction (NWP) framework. ICON‑DREAM also includes a 20‑member ensemble for uncertainty assessment and is statistically processed to produce ensemble statistics such as mean, median, standard deviation, and min/max for each variable. Current data cover 2010 to July 2025 at hourly temporal resolution [2]. We aim to update the data continuously and extend the archive back to the 1980s.
To meet the specific needs of the German energy sector, we extract a national subset from the ICON‑DREAM‑EU domain. This subset is delivered at a spatial resolution of 6.5 × 6.5 km for the deterministic fields and at 20 × 20 km for the ensemble members. This contribution presents an evaluation of the reanalysis against parameters relevant to the renewable‑energy sector – such as wind speed and wind direction at hub height, radiation and temperature. The dataset will be released as an optimized, user‑friendly product, including uncertainty estimation from the ensemble. In accordance with the DWDs Open Data Policy the dataset will be provided free of charge. The evaluation draws on observational data and accounts for extreme events.
[1] Valmassoi, A., J. D. Keller, R. Potthast, H. Anlauf, A. Cress, F. Kaspar, and A. Becker: ICON-DREAM: the new dual resolution reanalysis from DWD. ICCARUS Book of Abstracts 2025. https://dx.doi.org/10.5676/DWD_pub/nwv/iccarus_2025
[2] Valmassoi, Arianna; Anlauf, Harald; Becker, Andreas; Keller, Jan D.; Krebber, Sibylle; Zängl, Günther; Potthast, Roland; Cress, Alexander; Fundel, Felix; Hanisch, Thomas; Lange, Martin; Steinert, Thorsten; Kaspar, Frank, "ICON-DREAM: ICON-Dual resolution Reanalysis for Emulators, Applications and Monitoring The ICON-Dual resolution reanalysis version v1.0." Deutscher Wetterdienst, 2025, doi:10.5676/dwd/icon-dream_v1.
Acknowledgement: This work was supported by the Federal Ministry for Economic Affairs and Energy through the Medaillon project (CNr: 03EI1059D).
How to cite: Schiller, R., Drücke, J., Mikalsen, A. C., Valmassoi, A., Kaspar, F., Hollmann, R., Geiger, D., Callies, D., and Pauscher, L.: Enhancing the ICON reanalysis dataset for energy‑sector applications within the MEDAILLON project, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-104, https://doi.org/10.5194/ems2026-104, 2026.
High-resolution numerical weather prediction datasets, such as the New European Wind Atlas (NEWA), are widely used in meteorological applications including forecasting, wind resource assessment and observation modelling. Despite their high spatial and temporal resolution, these datasets often exhibit systematic biases when compared to site-specific observations, particularly in complex terrain and at hub heights relevant for modern wind energy applications. Understanding and addressing these biases is important for improving the consistency between modelled and observed wind conditions.
This study focuses on a sensitivity analysis of different interpolation approaches used in the representation of wind speed profiles. The aim is to investigate how methodological choices influence wind speed estimates, with particular attention to biases arising from varying heights and terrain characteristics.
The analysis is based on mesoscale wind data in the height range of 150 m to 300 m above ground level. In a first step, wind speeds at relevant hub heights are derived at observation locations using a range of vertical interpolation approaches, including polynomial, logarithmic, power-law and piecewise cubic Hermite interpolation (PCHIP). The sensitivity of the resulting wind profiles to interpolation method, height and terrain characteristics is evaluated. In a second step, the study explores how these differences can be represented spatially by applying observationally informed adjustments to mesoscale data. The focus is on assessing general spatial interpolation strategies that incorporate terrain-related information, with the aim of examining how local variations may influence the transferability of corrections across locations and heights.
Overall, the work is intended to provide a structured assessment of how vertical and spatial interpolation choices affect wind speed profile representation in the context of observational data. The findings are expected to contribute to a better understanding of bias characteristics in mesoscale datasets and to inform the selection of appropriate methods for applications involving wind profile estimation.
How to cite: Drieling, J. and Lakdawala, Z.: Exploring the sensitivity of vertical and spatial interpolations for varying heights and terrain for bias correction of NEWA wind data using observations, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-625, https://doi.org/10.5194/ems2026-625, 2026.
It is now widely recognized that assimilating consolidated satellite-derived vegetation products improves the representation of land surface variables and fluxes. This raises the question of their suitability for both reanalysis products and near-real-time (NRT) applications. Among these variables, Leaf Area Index (LAI) plays a key role in controlling exchanges of water, energy, and carbon between the land surface and the atmosphere. While reprocessed and consolidated products are well suited for retrospective reanalyses, NRT applications require observational streams with low latency and sufficient consistency with delayed, higher-quality products.
In this study, we assess the potential of the Copernicus Land Monitoring Service (CLMS) 300 m LAI products for land surface analysis and NRT applications within the LDAS-Monde system coupled to the ISBA land surface model for the year 2021 in a global scale. We investigate the assimilation of the different CLMS processing streams, namely RT0 and RT1, available in NRT, and RT6, which is provided as a consolidated product with a 60-day latency. A set of global offline experiments forced by ERA5 is conducted to evaluate their consistency and suitability for land data assimilation.
The results show that RT1 captures the main temporal variability of the consolidated RT6 product while benefiting from a substantially reduced latency. Compared with RT0, RT1 provides a more stable and consistent signal, offering a robust compromise between timeliness and data quality. Assimilation experiments within LDAS-Monde indicate that analyses driven by RT1 are consistent with those obtained using RT6, supporting its use in systems targeting NRT land monitoring. In addition, an in situ evaluation based on observations from the ICOS FR-Tou site is conducted for selected surface variables, providing an independent benchmark to further assess whether the use of RT1 leads to comparable performance to RT6.
How to cite: Rojas Muñoz, O. J., Calvet, J.-C., Vanderbecken, P., and Vural, J.: Towards near-real-time Leaf Area Index assimilation in LDAS-Monde, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-239, https://doi.org/10.5194/ems2026-239, 2026.
Accurately representing wind in complex terrain remains challenging for atmospheric reanalyses, but is an important prerequisite for estimating wind energy potential. In this study, wind measurements from two meteorological masts at the WINSENT ("Wind Science and Engineering Test Site in Complex Terrain") on the Swabian Alb are analyzed with respect to orographic effects and local surface influences. The site is located on top of the Swabian Alb plateau, in the vicinity of the forested escarpment, creating a heterogeneous environment. The performance of the high-resolution reanalysis ICON-DREAM ("ICON-Dual resolution Reanalysis for Emulators, Applications and Monitoring", operated by Germany's national meteorological service DWD (Deutscher Wetterdienst)) and the global reanalysis ERA5 ("ECMWF ReAnalysis 5", operated by European Centre for Medium-Range Weather Forecasts (ECMWF)) is evaluated for the location of WINSENT by comparing the reanalysis data with continuous mast observations for 2019-2024 at hub-height (100 m above ground level).
The comparison of the measurements from the two masts, although only about 200 m apart, shows pronounced differences in wind characteristics caused by small-scale orographic effects such as speed-up over exposed terrain as well as reduced wind speeds due to the influence of the nearby forest. These small-scale features are not resolved by reanalysis datasets and therefore cannot be reproduced adequately.
Both ICON-DREAM and ERA5 show a systematic underestimation of wind speeds compared to the observations. The magnitude of the bias varies over the course of the day, with larger underestimations occurring during nighttime conditions for both reanalysis datasets. Despite its higher spatial resolution, ICON-DREAM shows particular limitations in reproducing the observed diurnal cycle of wind speed, with especially low wind speeds at night. While observations show the occurrence of a nocturnal low-level jet at 100 m, ICON-DREAM only shows this behaviour for higher model levels. This behaviour may indicate reduced model performance of ICON-DREAM under stable atmospheric conditions.
How to cite: Schulte, J., Büchau, Y., Bange, J., and Platis, A.: Evaluation of Reanalysis Models in Complex Terrain: Comparing ICON-DREAM and ERA5 Against Met-Mast Observations at the WINSENT Test Site, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-356, https://doi.org/10.5194/ems2026-356, 2026.
In this presentation, we will present the Copernicus European Regional ReAnalysis (CERRA, https://climate.copernicus.eu/regional-reanalysis-europe). This dataset is produced by the Swedish Meteorological and Hydrological Institute (SMHI) within the Copernicus Climate Change Service (C3S). The system consists of three main components: the ensemble data assimilation component, CERRA TU, and the CERRA-Land component. The data assimilation system is based on the HARMONIE-ALADIN numerical weather prediction system, using 3DVAR for upper-air observations and an optimal interpolation approach for surface observations. The system uses ERA5 as boundary conditions, additional local observations, and provides reanalysis at 5.5 km horizontal resolution, with 3-hourly analysis fields and hourly forecast fields. The CERRA dataset includes more than 50 parameters, covering both surface and atmospheric variables, and is archived on the CDS from 1984 to the latest month with a latency of 3.5 months. The ensemble data assimilation component of the system consists of a 10-member ensemble for generating the B-matrix and uncertainty estimates at 11 km horizontal resolution.
Besides near-real-time production, climate records extending from 1961 to 1984 are also being produced using six parallel streams for the back extension. These data are planned to be made available to users in 2027.
In the current project, daily and monthly means from CERRA, along with ensemble statistics from EDA, are being produced and archived on the CDS.
The presentation will describe the CERRA system and the performance of CERRA in comparison to ERA5. The presentation will also address the challenges related to real-time CERRA production, ensuring data quality, and carrying out the back-extension runs.
How to cite: Iqbal, W., Geidne, A., Isaksson, L., Edvinsson, L., Jones, J., Ridal, M., Andrea, U., Bazile, E., Le Moigne, P., and Dahlgren, P.: Copernicus European Regional ReAnalysis (CERRA): Towards Near Real-Time Production and Extended Climate Records, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-590, https://doi.org/10.5194/ems2026-590, 2026.
High-resolution near-surface wind fields are essential for analyzing atmospheric circulation, diagnosing extreme weather events, and conducting regional climate research. However, the spatial resolution of global reanalysis datasets is often limited by computational constraints, which hinders their ability to accurately represent mesoscale and fine-scale atmospheric structures. Conventional interpolation methods, such as bilinear interpolation, typically fail to restore realistic fine-scale wind variability.
To address this, we develop a conditional diffusion-based super-resolution framework to downscale NCEP wind fields from a 2° resolution to 0.25° over the East Asia region (15°–55°N, 65°–135°E). The model leverages a denoising diffusion probabilistic process with a U-Net backbone, where low-resolution wind fields serve as conditioning information to guide the generation of high-resolution wind structures. The model is trained using daily NCEP 10-m wind components from 1981 to 2015, validated during 2016–2019, and evaluated over an independent test period from 2020 to 2023.
Model performance is assessed against bilinear interpolation using multiple statistical and spectral metrics, including root-mean-square error (RMSE), spatial correlation coefficient, and isotropic power spectrum analysis. Results show that the proposed diffusion model substantially outperforms bilinear interpolation across the entire domain. The spatially averaged correlation coefficient between the reconstructed and ERA5 reference wind fields reaches approximately 0.94, while the mean RMSE is reduced to about 0.86. Furthermore, power spectrum analysis demonstrates that the diffusion-based results closely match the high-resolution ERA5 reference across a wide range of spatial scales, effectively recovering high-frequency energy that is largely absent in interpolated fields.
These results indicate that conditional diffusion models provide a promising data-driven approach for wind field downscaling, offering improved spatial realism and scale-consistent representations. The proposed framework has potential applications in high-resolution reanalysis reconstruction, regional climate analysis, and ensemble-based uncertainty quantification.
How to cite: yuewei, F. and yi, L.: Super-Resolution of NCEP Reanalysis Wind Fields to ERA5 Resolution via Conditional Diffusion Models, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-369, https://doi.org/10.5194/ems2026-369, 2026.
The ocean surface wind plays a key role in the exchange of heat, gases and momentum at the atmosphere-ocean interface. High-quality wind records from scatterometers are available from 1991 onwards, with care taken to account for changes in scatterometer instrument types and spatial coverage over time. On the other hand, modelled surface winds from global climate reanalyses suffer from changes in the density and coverage of observational timeseries used in the data assimilation process. Still, global numerical weather prediction (NWP) model wind fields are widely used in ocean research.
A comparison of scatterometer observations and global NWP model wind fields reveals substantial, persistent local systematic errors in wind vector components and spatial derivatives. Temporally-averaged gridded differences between geolocated scatterometer wind data and NWP wind fields can be used to correct for persistent local model wind vector biases. By combining these bias corrections with global, hourly NWP wind fields, high-resolution wind forcing products can be created for the ocean modelling community and other users.
In 2022, new hourly and monthly Level-4 (L4) surface wind products were introduced in the Copernicus Marine Service catalogue. These products include global bias-corrected 10-m stress-equivalent wind, surface wind stress fields and spatial derivatives. The bias corrections are calculated from Copernicus Marine Service Level-3 wind products for a combination of scatterometers and their collocated European Centre for Medium-range Weather Forecasts (ECMWF) ERA5 reanalysis model winds.
We used the multi-year L4 products to identify long-term changes in ocean surface wind differences over the period 1995-2024. The spatial distribution of differences between scatterometer observations and collocated ECMWF ERA5 reanalysis are found to be highly consistent between different scatterometers and over time. Bias corrections for a single instrument display long-term variations of comparable magnitude to the scatterometer-model differences, pointing to artificial changes in the ERA5 winds over time. Furthermore, regional local bias anomalies are found for climate phenomena like the El Niño Southern Oscillation. These artificial features should be taken into account in multi-decadal ocean studies.
How to cite: Giesen, R. and Stoffelen, A.: Multi-decadal variability in ocean surface wind differences between scatterometer observations and reanalysis model fields, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-556, https://doi.org/10.5194/ems2026-556, 2026.
Droughts are intensifying under human-induced climate change, posing increasing risks to ecosystems, water resources, and food security worldwide. Improving the monitoring and characterization of drought occurrence and severity is therefore essential for both scientific understanding and operational decision-making. In this study, we present the ERA5–Drought dataset, a new global, long-term record of meteorological drought indicators derived from the ERA5 reanalysis. The dataset provides both deterministic and probabilistic representations of drought conditions from 1940 to present and is openly accessible through ECMWF’s data infrastructure under a CC-BY 4.0 license.
ERA5–Drought includes two widely used indices: the Standardized Precipitation Index (SPI) and the Standardized Precipitation Evapotranspiration Index (SPEI), computed across accumulation periods ranging from 1 month to 48 months. The dataset is complemented by validation metrics to support informed usage and interpretation. Its global coverage and temporal depth enable consistent monitoring of drought variability and trends across diverse climatic regions.
We demonstrate the applicability of the dataset through illustrative case studies in Catalonia (Spain) and Kenya. In Catalonia, long-term time series highlight the evolution and spatial extent of major drought events since 1940, including recent extreme episodes. In Kenya, we show a strong temporal correspondence between meteorological drought conditions and documented socio-economic impacts, such as food insecurity events recorded in the EM-DAT. These examples illustrate how ERA5–Drought can be used not only to characterise hazard, but also to support impact-based analyses when combined with socio-economic data.
The ERA5–Drought dataset provides a flexible and accessible tool for a wide range of applications, from local drought monitoring to global risk assessments. By enabling consistent, data-driven analyses of drought occurrence, intensity, and impacts, it supports environmental agencies, water managers, and agricultural stakeholders in addressing the growing challenges of climate variability and change.
How to cite: Keune, J., Di Giuseppe, F., Wetterhall, F., and Barnard, C.: A Global ERA5-Based Dataset for Monitoring Drought Occurrence and Impacts, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-325, https://doi.org/10.5194/ems2026-325, 2026.
Supporters & sponsors
