This demonstrator implements an operational attribution workflow based on two internally consistent worlds:
Climate input variables used in the final gridded products are:
ERA5 + ERA5-Land ingestion (step_1-2), EMO-1 preprocessing and harmonization (step_3-7), ISIMIP3BASD bias adjustment + statistical downscaling (step_8), variable reconstruction (step_9-10), and climate-map publication.
Step_8 inputs are passed to ATTRICI full-grid simulations, then rebuilt through step_8_cf, step_9_cf and step_10_cf for date-aligned factual vs counterfactual comparisons.
step_10 and step_10_cf drive crop pipelines with LISVAP preprocessing, yield prediction, and attributable yield/price indicators shown in Crop Yield Analysis.
The climate pipeline starts from ERA5 and ERA5-Land reanalysis products. Monthly GRIB files are ingested and transformed into daily fields for key variables (temperature, precipitation, wind, radiation and humidity-related quantities). During this phase we enforce temporal completeness, naming consistency and compatible units.
The operational daily-data workflow prepares analysis-ready inputs through sequential processing. This includes conversion to daily statistics, variable derivation (e.g., pd from humidity/temperature), and harmonization to a common grid and format. The workflow also manages gap-filling support datasets, grid remapping weights, and chunked NetCDF outputs required by the next stages. EMO-1 based products are used as reference information during the harmonization and remapping stages.
In step_8 we apply ISIMIP3BASD in two coupled stages: bias adjustment on coarse-scale series and statistical downscaling to the target grid. Variable-specific settings are used. The objective is to preserve realistic variability and extremes while aligning modeled and reference statistics.
Step_9 reconstructs physically interpretable daily variables and applies required time alignment and unit transformations. Step_10 merges BASD-adjusted fields with ERA5-based support where needed, producing final daily gridded datasets for pd, pr, rg, tn, tx, ws. These datasets feed map generation and the Climate Maps tab in the demonstrator.
Counterfactual climate is generated with ATTRICI framework, which simulates a full-grid counterfactual based on the step_8 bias-adjusted inputs. The resulting fields are then propagated through the dedicated CF chain (step_8_cf, step_9_cf, step_10_cf). This mirrors the factual processing structure and enables like-for-like attribution between factual and counterfactual conditions for each date and variable.
The crop workflow consumes step_10 (factual) and step_10_cf (counterfactual) meteorology through synchronized inputs. Each branch runs LISVAP preprocessing, daily feature aggregation, and the yield prediction stage. The prediction models are trained on Poland crop yield records for 1974-2025. Outputs are generated by province and crop type for both worlds.
Attribution is computed as Factual - Counterfactual at climate and crop-impact levels. For economic estimates, predicted yield is combined with cultivated area and crop price assumptions (yield × area × price). Final data are uploaded and rendered in the Climate Maps and Crop Yield Analysis tabs.
COMPASS "Compound extremes attribution of climate change: towards an operational service" is a collaborative project funded under the Horizon Europe Research and Innovation Programme. The project began in January 2024 and will run for 3 years.
The COMPASS project aims to develop a harmonised, yet flexible, methodological framework for climate and impact attribution of various hazard types, which is a crucial initial step towards the deployment of a European operational service. COMPASS will innovate by going from the attribution of single-driver extremes to the attribution of more complex extremes (including compounding, sequential and cascading hazard events) and enabling a shift from a hazard-centred analysis to an impact-centred perspective. Main novelties include event-based hazard and impact modelling using a multi-scale approach, the use of weather type analysis for better understanding the physical drivers that give rise compound extremes, and the use of contextualized storylines to communicate attribution results. Our framework will be validated and applied to a set of use cases that cover historical extremes for various hazard types and impact context, as well as extreme events happening during the project.
COMPASS will create a modular and scalable framework for on-the fly analysis, and thus transferable to other extremes and regions, and will lay the scientific foundation for the operational deployment as part of the Copernicus Climate Change Services. To promote the uptake of the project's results, our data, methods and tools will be made openly available, while a web-based demonstrator will showcase the outputs of the use cases, and clear guidelines for climate attribution will be developed.
