Regional downscaling for extreme events

Urban heat islands (UHIs) are areas within urban environments where air temperatures are consistently higher than those in surrounding rural areas. UHIs arise from urban morphology, anthropogenic heat emissions, and altered radiative and evapotranspiration balances. 

The local effects of urban heat islands during extreme events (e.g., heat waves) are difficult to predict, in great part due to the mismatch between large-scale atmospheric processes and small-scale urban physics, and the mechanisms involved in the overall energy transfer during the formation, persistence, and decay of UHIs under these circumstances still remain unclear. The objective of this study is to determine the triggering factors that influence these mechanisms and to better understand the onset of this phenomenon.

We use experimental observations from a network of meteorological stations with a 10-minute sampling rate, deployed between March and December 2025 in six cities near the Rouen metropolitan area in Normandy (France), along the Seine River. The dataset obtained by these stations is complemented by publicly available data from local operational stations.

We calculate urban–rural temperature differences and their temporal variability under different general and local conditions of paired urban–rural sites, using a combination of physical and statistical analyses, such as moment analysis, auto- and cross-correlation analysis, and temporal evolution of the probability density function for temperature measurements. Preliminary results indicate that urban temperatures are on average about 1°C higher than those in neighboring rural areas, with peaks reaching up to 5°C in four cities along the Seine Valley, near Rouen, between June 19 and July 4 and between August 8 and 18, 2025, periods during which heat waves were reported in France. During these periods, we found that most stations reached these peak values at night, consistent with the normal UHI behavior reported in the literature.

Future work will focus on reproducing observed UHI patterns through high-temporal- and spatial-resolution numerical simulations with the Weather Research and Forecasting (WRF) model. Subsequently, Markov processes will be explored to develop a UHI prediction model based on experimental data from the stations and results from numerical simulations.