A Short-Term Predictive Meta-Model for Neighborhood-Scale Surface and Air Temperature

In the context of global climate extremization, DIAMS project aims to develop a diagnostic tool to assess urban districts vulnerable to heatwaves, using thermal infrared (TIR) satellite data and microclimate modeling. TIR satellites offer cost-effective and broad spatial coverage, though current limitations in revisit frequency and spatial resolution are expected to improve with upcoming missions like TRISHNA, scheduled for launch in 2026.

As part of DIAMS, a predictive meta-model has been developed to forecast district-scale surface and air temperatures over the next three days. The meta-model integrates spatial indicators (morphological, material, optical) and temporal indicators (air temperature, relative humidity, global horizontal radiation, wind speed) to forecast equivalent air temperature (Teq) and mean surface temperature (Tse). It is based on urban archetypes classified by the GENIUS tool [1], which identifies seven archetypes by morphological properties.

In practice, morphological and spatial information is derived from BD TOPO maps provided by the French National Institute of Geographic and Forest Information (IGN), while TIR satellite data helps to parameterize materials’ thermal (effusivity) properties. Temporal data is sourced from weather stations. Currently, the meta-model has been developed for the "Continuous Blocks" archetype.

In initial tests, weather data from July 11 to August 10, 2003 was applied to this archetype, with Solene-microclimate simulations generating Teq and Tse for this period. Meta-models were then developed based on these simulations and validated using weather data from August 9 to 13, 2009, with predictions showing good accuracy compared with numerical simulation results, achieving an MAE of 0.11°C for Teq and 1.08°C for Tse.

Future work will expand the meta-model to more archetypes.