Experimental methodology for analyzing uncertainties linked to outdoor thermal comfort in hot weather: Sensitivity analysis and clustering used to the ESTP campus in Cachan

This study uses a case study on the ESTP campus in Cachan, France, to investigate urban overheating and its effects on outdoor thermal comfort. An experimental measurement effort was carried out during the July 2022 heatwave, focused on daytime conditions. In order to evaluate the generalizability of the proposed approach, a monitoring system that combined a weather station and microclimatic sensors was set in place to gather the data needed for the computation of thermal comfort indices at various locations and periods. Developing an experimental approach to evaluate the propagation of uncertainties is the aim of this work. After grouping similar data points using K-means clustering, each cluster centroid is subjected to a three-level factorial sensitivity analysis to measure the uncertainty within each group. The results indicate that splitting the data into six clusters provides an adequate compromise between computational efficiency and accuracy of the model. Uncertainties associated with the mean radiant temperature are ensured by low (±1.28 to ±1.39 ∘ C ). The UTCI uncertainties are less affected by fluctuations in radiation and wind speed, ranging from ±0.94 to ±1.48 ∘ C . Conversely, PET shows significantly larger uncertainty (±1.85 to ±5.09 ∘ C ), which is indicative of its strong sensitivity to air velocity, especially in low wind circumstances (0.6 to 0.9 m / s ). These uncertainties might be significantly reduced by using an ultrasonic anemometer rather than a 3-cup anemometer, particularly for PET and to a lesser degree for UTCI . This highlights the important role that sensor selection plays in the accurate evaluation of outdoor thermal comfort indicators.