Unravelling indoor temperature response to summer heat through long-term crowdsourced observations in Dutch residences

City dwellers are increasingly exposed to summer heat due to climate change and urbanization. Summer heat, which causes heat stress, is intensified especially at night in urban areas and is projected to become more extreme due to climate change. City dwellers are not just increasingly exposed to this heat outdoors but mainly indoors, as they spend the majority of their time inside their homes. However, observational and modelling studies on indoor heat stress are relatively scarce, especially concerning the interconnections between indoor and outdoor climatic conditions.

This study observes, analyses, and models the evolution of indoor air temperature during Dutch summer heat using two unique crowdsourced datasets. The first dataset consists of long observational records, spanning up to 27 years, from citizen weather stations (CWS) located in seven residences across the Netherlands. This dataset provides insight into the benefits of long-term observations at residences. The second dataset consists of indoor CWS placed by us since 2022 in 100 residences across Amsterdam. This dataset offers insight into the benefits of measuring in a large number of residences.

Conventional high-resolution building energy models are commonly validated in controlled settings. In contrast, our study utilizes real-world residences inhabited by individuals, thereby capturing actual occupant behaviour. Moreover, crowdsourced indoor climate observations, just like ours, often lack supplementary data such as building characteristics and occupant behaviour. Therefore, we adopt an analysis and modelling approach only taking indoor temperature as an input parameter of the residence. We demonstrate that indoor temperature typically warms up more slowly than outdoor temperature but also cools down more slowly. The seven residences in the first dataset had, on average, a lag difference of approximately 260 minutes in the diurnal cycle during summer. Indoor temperature also remained higher than outdoor temperature for up to 5 days after a heatwave. For the 100 residences in the Amsterdam dataset, the analysis results will be presented. To model indoor temperature evolution, we simulated daily changes in indoor temperature evolution with a physics-based statistical model. The model includes outdoor conduction, indoor conduction, and solar transfer components, calculated from indoor temperature observations and outdoor temperature, solar irradiance, and wind observations. Results of this computationally-fast model for the seven residences are promising, with on average a mean absolute error of 0.43 K day^-1 during summer. Preliminary results suggest a higher model performance for modelling of the warming of the residences compared to the cooling. The model results for the 100 residences will be presented, providing insight into the variability in model performance for indoor temperature in Amsterdam.

The study's findings illustrate the high potential of the model applied to crowdsourced observations to promote understanding of the fundamental processes influencing indoor temperature response to summer heat.