Measurement report: A complex street-level air quality observation campaign in a heavy-traffic area utilizing the multivariate adaptive regression splines method for field calibration of low-cost sensors

Abstract. As part of the TURBAN project, the “Legerova campaign” investigated air quality and meteorology in a traffic-dense area of Prague, Czech Republic, from 30 May 2022 to 28 March 2023. The study deployed a network of 20 low-cost sensor (LCS) stations to measure NO2, O3, PM10 and PM2.5 concentrations, complemented by advanced meteorological instruments such as a microwave radiometer and Doppler lidar. Ensuring data quality from LCS measurements presented significant challenges. Initial field tests at a reference monitoring station revealed strong correlations between raw LCS and reference data (r > 0.90 for NO2 and PM2.5, r > 0.80 for O3 and PM10). However, individual biases were observed. Applying the multivariate adaptive regression splines (MARS) method effectively reduced biases and enhanced alignment with reference measurements for all pollutants (R2 0.88–0.97). During the campaign, sensor ageing and technical issues were identified through double mass curve analysis and final field testing. The highest NO2 concentrations were recorded in streets with dense building blocks and traffic lights, corresponding to peak traffic patterns (with medians of concentrations 20–34 ppb). Aerosol concentrations were generally low (medians of PM10 < 25 µg m−3 at all sites), with less temporal and spatial variability than NO2. Elevated PM10 and PM2.5 levels occurred primarily during temperature inversions, often linked to local sources, and during a short, non-local episode. This study highlights the MARS method as a reliable tool for field calibration of LCS networks and provides valuable data on urban air quality and its dynamics with high spatiotemporal resolution.