Urban Heat Health Forecasting with Destination Earth: Leveraging Digital Twins and Machine Learning for Scalable Urban Climate Services

Extreme heat is increasingly recognized as one of the most severe climate-related risks affecting urban populations, with disproportionate impacts on public health, energy systems, and vulnerable communities. As heatwaves intensify under climate change, cities require near-real time, high-resolution and actionable information to support early warning systems, preparedness, and long-term adaptation. Addressing this challenge at urban-local scale demands not only methodological innovation, but also robust digital infrastructures capable of delivering consistent and interoperable climate intelligence across regions.

Destination Earth (DestinE), a strategic initiative of the European Union, represents a transformative step in this direction by providing global, high-resolution climate and weather simulations, through Digital Twins of the Earth System. By coupling advanced numerical models, Earth Observation (EO) data, and high-performance computing, DestinE establishes a common backbone for next-generation climate services. However, translating these powerful datasets into locally relevant, operational products for cities remains a critical challenge.

DE_395-Urban Heat Health Forecasting (UHHF) project addresses this gap by demonstrating how DestinE Extremes Digital-Twin outputs can be transformed into urban-scale user-oriented heat-health indicators through the operational use of Machine Learning (ML). The project applies ML-based downscaling techniques to near-surface air temperature (T2m) and relative humidity (RH) forecasts, enhancing spatial resolution from kilometre-scale to approximately 200 m. These downscaled fields are subsequently used to derive human-biometeorological indicators such as the Universal Thermal Climate Index (UTCI) and Thermal Stress Duration (TSD), supporting health-oriented risk assessment.

The UHHF framework integrates DestinE atmospheric drivers with EO-derived and geospatial predictors describing urban form, land cover, vegetation, and topography, including Local Climate Zones. Quality-controlled crowdsourced observations from citizen weather stations are combined with WMO reference data to constrain and validate the ML models, ensuring robustness under both average and extreme conditions. The approach is being implemented across four climatically and socio-environmentally diverse Functional Urban Areas, e.g. Naples, Chicago, Santiago, and Cape Town, enabling a systematic evaluation of models across continents.

By building directly on DestinE and complementary European programmes led by ECMWF, ESA, and Copernicus, drawing on both their data assets and operational services, UHHF aims to illustrate how these can be leveraged to develop affordable, scalable, and reproducible urban-scale climate information and services. The project highlights the strategic importance of climate data platforms in bridging the gap between global simulations and local decision-making, contributing to the development of interoperable urban climate and health services aligned with European and international resilience frameworks.