| 11 Mar 2026
Forecasting European temperature-related mortality in Summer 2024: data-driven vs physics-based forecast approaches
Abstract. Heat has emerged as a major public health concern. Over 62,000 heat-related deaths were estimated to have occurred during the European summer of 2024, exemplifying the pressing need to develop effective early warning systems. Such systems depend critically on the quality of the underlying forecasts, and recent work has focused on developing impact-based forecasts for heat-related mortality, which provide impact-oriented information. To date, heat-related mortality forecasts have been based on the output of numerical weather prediction models, or physics-based forecasts. The field of weather forecasting is undergoing a rapid transformation with the advent of skillful data-driven forecasts. This study compares European temperature-related mortality forecasts for summer 2024 based on physics-based weather forecasts with those based on data-driven weather forecasts. Our results highlight that both the physics-based and data-driven forecasts systematically underestimate temperature-related mortality, more pronouncedly so in the latter. Both types of forecasts appear sensitive to forecast errors at hot temperatures, due to the non-linear relationship between temperature and mortality. Nevertheless, temperature-related mortality forecasts based on data-driven weather forecasts appear to be a promising alternative to traditional physics-based weather forecasts, and targeted improvement of the representation of hot temperatures through bias correction or adjustment of the loss function to give greater weighting to hot temperatures would be beneficial for temperature-related mortality forecasting. We suggest the application of this approach to both data-driven and physics-based forecast ensembles as an important next step in the continued development of informative, impact-oriented forecasts.
The manuscript addresses an important and timely topic, namely the development of impact-based forecasts for heat-related mortality. Overall, the manuscript clearly achieves its stated objectives and is well written, with a logical structure that makes the analysis easy to follow. I suggest some minor revisions, mainly related to clarification of certain aspects, as outlined below.
"We also use 2m temperature from archived forecasts from two different types of weather prediction models, one physical model (IFS HRES cycle 48r1), and one data-driven model (AIFS single v1).” It would be helpful if the authors briefly explained why these specific models were selected.
The epidemiological framework relies on exposure–response functions from the multi-city analysis of Masselot et al. (2023), which is based on all-cause mortality data. It would be helpful to state this explicitly, as the estimated impacts therefore correspond to temperature-related mortality across all causes rather than specific cause-of-death categories (e.g., cardiovascular or respiratory mortality which are the most common causes that are used and studied in health imapct studies). A brief clarification regarding this would help.