European Risk Managers Call for Climate Risk Collaboration
The Federation of European Risk Management Associations (FERMA) is urging the European Union to treat climate risk as a systemic challenge that is "too big for anyone to go alone." The organization is advocating for shared data, modeling frameworks, and mitigation strategies. A separate report from DWF highlighted the need for updated catastrophe models and real-time analytics, underscoring the data challenges facing actuaries and underwriters.
- The European Environment Agency's (EEA) first-ever European Climate Risk Assessment, published in March 2024, identified 36 major climate risks for Europe, creating a scientific foundation for policy. Only about 25% of climate-related economic losses in Europe are currently insured. - The EU already operates the European Climate Adaptation Platform (Climate-ADAPT), a partnership with the EEA that shares data on climate change projections, vulnerability, and adaptation case studies to support planning. Another initiative, IRISCC, is a consortium of research infrastructures offering sponsored access to climate-risk data, tools, and labs, with its third open call for projects launching in February 2026. - A significant challenge in catastrophe modeling is that historical data is becoming less predictive due to climate change, forcing a shift away from statistical stationarity. Current climate models also struggle to represent extreme, localized hazards like hail or the strongest tropical cyclone winds, potentially leading to underestimation of risk. - To address these data challenges, some insurers are using AI neural networks to create global hazard maps by combining data from climate models with existing hazard data. Advanced aerial imagery and geocoding are also becoming more common to enhance the granularity of catastrophe models down to the neighborhood or individual home level. - For data engineers building modern insurance data platforms, a common and effective architecture combines Snowflake for elastic storage and compute, dbt for SQL-based transformations and data quality testing, and Airflow for orchestration. This stack allows for clear separation of concerns: Snowflake executes queries, dbt manages transformation logic, and Airflow schedules and manages dependencies. - In managing actuarial models, MLOps best practices are crucial for governance in a regulated environment. This involves versioning not just code but also datasets and models using tools like DVC and MLflow, automating CI/CD pipelines for validation and deployment, and actively monitoring for data and concept drift post-deployment. - As a potential pivot, AI in the fashion and retail industry focuses heavily on hyper-personalization, with brands like Nike and Stitch Fix using AI to customize products and styling recommendations. These AI applications have been shown to reduce return rates by up to 25% and improve customer satisfaction by nearly 40%. - The underlying AI technology in retail often includes virtual try-on tools powered by computer vision and AR, as well as recommendation engines that analyze browsing history and purchase data to increase sales by as much as 8%. These systems help retailers optimize inventory and reduce overproduction by better forecasting demand.
