Fine-scale evaluation of the urban heat island effect drivers using citizen weather stations

The mitigation urban heat is a central planning priority for many cities, which are increasingly exposed to rising temperatures and heat waves. Such an endeavor requires understanding the relationship between air temperature and the spatial pattern of the built environment and urban green infrastructure. Nevertheless, the spatial sparsity of official monitoring stations is a major impediment towards understanding the drivers of the urban heat island effect.

To overcome such a drawback, we use air temperature measurements from citizen weather stations (CWS) to extend our previous work on urban heat island modeling, which largely improves of the spatial resolution by an order of magnitude. More precisely, the previous case study of the Swiss urban agglomeration of Lausanne was based in the data from 9 official monitoring stations, whereas 110 CWS can be found in the same area (after using state-of-the-art quality checks to filter out CWS showing suspicious pattern). We assembled hourly time series of temperature measurements collected from the five largest heatwaves in 2022 and 2023 is used to explore how the drivers of the urban heat island effect change along the daily cycle. We extend the features considered in the previous model (i.e., tree shade, albedo and evapotranspiration) by including additional fine-grained features that reflect the morphology of the urban canyon, such as building volumes, spacing between buildings and street orientations.

Unlike in our previous study using only official stations, linear models are not capable of representing the observed spatial distribution of temperatures. Therefore, we explore more advanced techniques such as generalized additive models, non-linear machine learning models and novel explainable artificial intelligence methods. The results novel insights of the urban heat island drivers that can be used to spatially and quantitatively plan mitigation strategies.