Biases due to widespread use of low-cost sensors for urban heat stress assessments

Sarah Berk, TC Chakraborty , Angel Hsu

Urban heat stress is an area of critical research interest due to its relevance for public health and policy. Given the lack of operational-grade weather stations within cities, different types of low-cost sensors have been used to assess urban heat stress. However, these sensors are traditionally not designed for capturing fine scale differences in temperature and humidity (i.e., the main components of moist heat), which are expected across a city, and their siting is often suboptimal due to logistical challenges. Here, through several calibration exercises over lawn and rooftop settings using 41 Kestrel sensors, we demonstrate significant issues with the use of such low-cost sensors for urban heat stress assessments. Issues stem from use of sensors without radiation shields, and even with cheaper non-aspirated shields, exposure to confounding environmental factors, and local land cover influences. Unshielded sensors overestimated temperature by up to 0.7 °C relative to shielded counterparts. Within unaspirated radiation shields, daytime measurement error was strongly correlated with diurnal temperature range (lawn r = 0.85; unshaded rooftop r = 0.98). Humidity cross-sensor variability exhibited a weaker correlation with its diurnal range. Heat index, derived from temperature and humidity, combined biases in the two measurements, magnifying resulting errors (4.7% increase compared to standard error for temperature). Finally, sensors near anthropogenic heat sources showed cross-sensor variability up to seven times higher than interference free sites. Based on these potential mismeasurements in urban heat stress gradients seen using inexpensive sensors, we provide recommendations on their appropriate deployment in urban environments.