Urban Heat Island Spatialization in an Alpine Valley: A Heatwave Case Study in Grenoble, France.

The problem of urban heat islands (UHIs) in complex topographic settings, such as alpine valleys, poses a growing challenge for climate-adaptive city planning because terrain-driven airflow modulates nocturnal temperature patterns and intensifies spatial heterogeneity. We aim to present an end-to-end pipeline for high-resolution mapping of nocturnal UHI intensity in the Grenoble alpine valley (France), combining a dense IoT sensor network (~140 street-level stations), Local Climate Zone (LCZ)-derived land-cover fractions (built-up, vegetation, water) computed at multiple spatial scales, and XGBoost regression with automated feature selection and Bayesian hyper-parameter optimisation. Applied to the heatwave period of 18-24 August 2023, the pipeline shows that LCZ-based built-up and vegetation fractions, optimised at radii of 200-600 m, are the dominant and most consistent predictors of UHI intensity, while elevation acts as a secondary predictor during the earliest nocturnal hours (02:00-03:00) and water fraction plays a minor role due to limited urban water bodies. The resulting spatial UHI patterns are qualitatively consistent with an independent reference map derived from the same sensor network, confirming the pipeline's operational viability for municipal climate-adaptation services without requiring physics-based model inputs. The approach highlights how recent technological advances, high-density IoT monitoring, LiDAR-derived elevation, standardized LCZ products, and algorithmic innovations (Spearman-based scale optimisation, hierarchical partitioning for feature importance, Bayesian optimisation via Optuna) advance UHI research, LCZ mapping, and thermal-comfort assessment. Future developments include integrating dynamic predictors such as real-time wind fields or soil moisture to improve generalisability across diverse synoptic conditions and extending the methodology to other alpine cities for scalable, operational climate services.