Vertical wind profiles and urban morphology effects on nocturnal urban heat islands from Doppler LiDAR observations and machine learning

Source abstract

Nocturnal urban heat islands (UHI) in high density cities are influenced by local morphology and vertical atmospheric structure. Most UHI assessments have relied on near-surface measurements and the interaction between vertical wind profiles and heterogeneous urban morphology is insufficiently explored. This study investigates how vertical wind pattern is associated with nocturnal UHI during summer months in Hong Kong. High-resolution Doppler LiDAR wind profiles are integrated with surface meteorological observations, ERA5 reanalysis data and urban morphology across multiple Local Climate Zones (LCZs). The Self-Organizing Map method (SOM) is used to classify hourly vertical wind profiles into four wind patterns. Extreme Gradient Boosting (XGBoost) performs best among the tested models and is used to interpret the interactions between weather conditions and localized morphology. Thermal differences are found within the same LCZ types and vary under the vertical wind structure. Pattern 2 shows the weakest ventilation and strongest nocturnal stability with wind speed of 1.97 m s -1 at 50 m at the urban core site and boundary layer height of approximately 230 m. Normalized Difference Vegetation Index (NDVI) and building density have their greatest relative predictive importance under this pattern. Pattern 4 is associated with the deepest boundary layer and lowest nocturnal UHI among all patterns, although lower-level wind speeds at the urban site remain lower than those at the exposed site. Incorporating the structural variations of the atmospheric boundary layer into spatial assessments may provide a physical foundation for heat resilience and climate-sensitive design in urban planning.