Urban morphometric properties and their relationship to intra-urban air temperature across temporal and spatial scales

Deciphering the complex relationships between urban form and canopy-layer air temperature has the potential to guide the parameterization of urban weather and climate models at the neighborhood scale. Many neighbourhood-scale urban morphometric properties easily extractable from OpenStreetMap and digital surface models could provide a proxy for the complex physical processes controlling canopy-layer air temperatures. Our aim is to understand how urban form at horizontal scales between 10 m and 1 km both responds to and influences pedestrian-level conditions as a first step to improving local-scale urban weather/climate modelling.

We investigate buildings-street metrics and canopy-layer air temperature variability within a city, using one year of data from 41 weather stations sited 3 m above ground in Freiburg, Germany. We explore measures of size, shape, spatial distribution and connectivity of 2- and 3-dimensional urban features at different spatial radii around the stations and how they correlate with temperature at different times of the day and year. We characterise the area with 100+ morphometric parameters, computed for and aggregated to multiple spatial scales. Some parameters are highly correlated throughout the city, many can be categorised as measures of properties, such as building density and neighbourhood heterogeneity.

For 20 heat-island nights (urban-rural temperature difference > 4 K), metrics with high values at the historic center, such as building area and compactness, are strongly correlated to nighttime canopy-layer air temperature at large radii (> 250 m). At medium radii (100 m – 300 m), metrics linked to the urban canyon such as height-width ratios strongly correlate. At small radii (< 100 m), correlations show low confidence as parameter aggregations are highly sensitive to individual building/street polygons. At this scale, metrics linked to the largest nearby building (e.g., maximum height/volume) show the most potential. Daytime correlations are weaker and more variable across temporal and spatial scales.