Statistical Calibration of NO2${\rm NO}_2$ Low‐Cost Sensors

ABSTRACT In recent years, air quality sensors have become key tools for assessing the concentration of pollutants in the atmosphere. The aim of this article is to investigate the methods used to calibrate these sensors. For this reason, two datasets of pollutant measurements in the city of Rouen, France, have been considered. One was constructed during the colocation period and the other during the deployment of the sensors, and both consist of micro‐sensor measurements of , CO, , and meteorological parameters, in addition to measurements from two reference stations. For low‐cost sensors, two issues are of interest: first, the conversion of the micro‐sensor measurements from microvolts () to micrograms per cubic meter (), and second, the calibration of the micro‐sensors. Therefore, this paper explores how to convert a set of low‐cost sensor measurements and understand the variables involved in the conversion of voltage to concentration model. It compares blindly calibrated models and models calibrated at monitoring stations during collocation. Finally, it demonstrates a homogenization of the converted measurements, both in terms of mean and variance, thanks to the conversion model.