Mapping urban temperature in Zürich

Extreme heat has a range of adverse impacts on humans, e.g., by affecting cognitive and physical capacities, mental health and sleep quality. It further increases mortality rates, the risk of accidents and disruptions in transport, information and communication technology and energy infrastructure. An increase in mean and extreme temperature can already be observed due to climate change, and this trend will most likely continue into the future. Cities and urban areas are especially vulnerable to extreme heat because of the urban heat island effect and are probably disproportionately affected in a changing climate. Since increasing parts of the world’s population live in cities, understanding the temperature distribution in urban areas is important for domains ranging from architecture and city planning to public health.

Affordable sensors

Generating high resolution temperature maps via data-driven modelling requires large datasets, which can be challenging to obtain, specifically because official in-situ measurement stations are sparse in urban areas. The increasing digitalization has led to the production and deployment of affordable sensors for smart home applications. Such low-cost sensors have become widespread for example in the form of citizen weather stations that are increasingly bought by individuals to measure meteorological parameters such as temperature, humidity and precipitation. The measurements are crowd-sensed by commercial or non-profit organizations. Using information from such low-cost sensors has a great potential to increase the spatial coverage of temperature measurements in urban areas. Hence, thanks to citizen weather station data, the challenges of in-situ measurements and statistical interpolation for urban heat maps can be overcome.

Data from heat wave in 2019

Marius Zumwald, Benedikt Knüsel, David N. Bresch and Reto Knutti from D-USYS present an approach to model urban temperature in high spatiotemporal resolution. The approach combines open government and remote sensing data, citizen weather station measurements and machine learning. The analysis is based on data from 691 sensors in the city of Zurich (Switzerland) during a heat wave using data from for 25-30th June 2019. In addition, in the study, several kinds of uncertainties are also addressed. For example contextual uncertainty which arises from biases due to the unknown exact sensor position (e.g., the height above ground). Another example is the prediction uncertainty which arises because of the use of machine learning methods.
The approach only needs few reference stations for model validation. Thus, citizen weather stations provide a cost-efficient means to gain insights into spatiotemporal temperature distribution in urban areas. Climate adaptation and specifically urban heat mitigation requires local information since the spatial temperature distribution is highly specific to the characteristics of a city. For example such heat maps of the study can be used as input for vulnerability and risk assessments, e.g., to assess how different vulnerable groups are affected by heat stress or to assess labor productivity loss on high spatiotemporal resolution.