Evaluation of spatial interpolation techniques for urban heat island monitoring in small and medium sized cities

Abstract

The growth of a city is typically accompanied by densification and sprawl, the former through verticalization, urban renewal, and the filling in of empty spaces. All of these activities extend and intensify the urban heat island (UHI), which is quantified in this study as the difference in daily minimum temperature between urban and rural areas. Here, we investigate this phenomenon in the area of Rennes (France) and 17 surrounding cities using the Rennes Urban Network which comprises 93 weather stations. This study aims to 1) determine the optimal method for spatializing UHI in Rennes, France, 2) estimate and spatialize the UHI in the small peri-urban cities surrounding Rennes. For this, we model mean UHI and intense UHI using three methods of interpolation—multi-linear regression (MLR), ordinary kriging (OK), and regression kriging (RK)—based on data from 2022. We find that the RK method is the most suitable overall, with an RMSE of 0.11°C for mean UHI and 0.25°C for intense UHI. This approach allows stochasticity to be taken into account, and thus provides a better representation of UHI variation within Rennes and its peri-urban cities.