Statistical prediction of the nocturnal urban heat island intensity based on urban morphology and geographical factors - An investigation based on numerical model results for a large ensemble of French cities

Abstract

Taking into account meteorological data in urban planning increases in relevance in the context of changing climate and enhanced urbanisation. The present article focusses on the nocturnal urban heat island intensity (UHII) simulated with a physically based atmospheric model for >200,000 Reference Spatial Units (RSU), which correspond to building patches delimited by roads or water bodies in 42 French urban agglomerations. First are investigated the statistical relationships between the UHII and six predictors: Local Climate Zone, distance to the agglomeration centre, population, distance to the coast, climatic region, and elevation differences. It is found that the maximum UHII of an agglomeration increases proportional to the logarithm of its population, decreases for cities closer than 10 km to the coast, and is shaped by the regional climate. Secondly, a Random Forest model and a regression-based model are developed to predict the UHII based on the predictors. The advantage of the regression-based model is that it is easier to understand than the black box Random Forest model. The Random Forest model is able to predict the UHII with <0.5 K absolute error for 54% of the RSU. The regression-based model performs slightly worse than the Random Forest model and predicts the UHII with <0.5 K absolute error for 52% of the RSU. A future challenge is to conduct a similar investigation at global scale, which is to date limited by the availability of a robust description of urban form and functioning.

Keywords: Local Climate Zones; Random Forest; Regression-based models; Urban heat island intensity; Urban morphology.