Report on integration of datasets and models

Deliverable D5.1 presents a modular and scalable workflow framework designed to support the impact attribution of climate-related compound events. Developed within the COMPASS project, the workflow will be demonstrated using UC5 (Poland – heatwaves and droughts), and conceptually extended to other use cases (UC1–UC4). This provides a foundation for generalising the approach while also acknowledging its current practical scope and limitations. It shows how to integrate publicly available datasets and open-source modelling tools into a unified, script-based environment for generating both factual and counterfactual climate scenarios.

The proposed solution is not limited to a specific case study or hazard type. Instead, it offers a flexible foundation that can be adapted to diverse geographical contexts and climate risks, including but not limited to heatwaves, droughts, floods, tropical cyclones, and winter storms. It does so by leveraging:

•  global climate datasets (e.g. ERA5, ERA5-Land),
•  regional high-resolution observations (e.g. EMO-1),
•  and impact and statistical models (e.g. ISIMIP3BASD, LISVAP, ATTRICI, CatBoost),

enabling consistent integration of hazard, exposure, and vulnerability data for comprehensive attribution assessments.

The framework supports both probabilistic (unconditioned) and storyline (conditioned) attribution approaches, making it suitable for a wide range of scientific and policy-relevant applications. Its open and modular design ensures reproducibility, transparency, and ease of use for researchers, decision-makers, and other stakeholders engaged in climate risk assessment.

Ultimately, this deliverable provides a conceptual and operational foundation for integrating climate and impact datasets through a transferable workflow architecture. Beyond its standalone value, the conceptual workflow presented in D5.1 (Report on integration of datasets and models) serves as the technical basis for follow-up deliverables—particularly D4.5 (Prototype of the operational use case), which focuses on the application of attribution results, D5.2 (Online web-based demonstrator) with presentation of workflow implementation in different use cases and D5.3 (Open-source software development), which documents continued development, refinement, and scaling of the system.

At the same time, the development of an automated event attribution workflow also revealed several key challenges. These include ensuring the robustness of attribution results across diverse data sources, managing the computational demands of high-resolution modeling, and harmonizing heterogeneous datasets across use cases. Lessons learned from each use case have been instrumental in identifying these barriers and proposing pragmatic solutions.

The developed approach contributes to the overarching goals of climate resilience, knowledge-based adaptation strategies, and improved understanding of the human influence on extreme weather and climate events.
To summarize, the presented COMPASS workflow for climate and impact attribution is:

• Modular and open-source: The workflow is composed of interoperable components that can be tailored to different hazards, regions, and impact models, using only open-source tools and datasets.
• Globally applicable: The workflow is demonstrated for heatwaves and droughts in Poland (UC5). However, the proposed workflow architecture is scalable to other hazards (e.g. floods, storms) and locations, including the full range of COMPASS use cases.
• Supporting both factual and counterfactual scenarios: Through integration with the ATTRICI model, the workflow enables comparative attribution to isolate the influence of climate change on extreme events.
• Towards operational implementation and integration: Scaling the workflow requires activities such as calibration, validation, incorporation of local data, and expert judgement on model configuration—steps that currently cannot be fully automated and different challenges are faced depending on the characteristics of the compound extreme. However, the workflow is already being implemented in an operational demonstrator for UC5 (heatwaves and droughts in Poland), supporting its future integration into regional climate services, adaptation assessments, and decision-support platforms.
• Designed for co-development: The workflow is built to evolve with stakeholder needs, the architecture encourages collaboration across research, policy, and practice communities.