2.1. Roadway Type Classification

This study was conducted in Shanghai, which has an area of 6341 square kilometres and a population of nearly 28 million. The mobile air monitoring campaign began in December 2019 following one month of trial operation in November 2019.

The roadway classification was based on OpenStreetMap (OSM), an open-source geographic information database with comprehensive roadway coverage and classification [32,33]. Studies have shown that OSM data for prefecture-level cities in China are almost complete [33], and the positioning accuracy is reportedly high [32,34,35]. OSM has predefined 23 road types, including trunk, primary, secondary, and tertiary roads, along with motorways, bike paths etc. This study has adopted OSM definitions to facilitate data analysis and includes four road types to balance coverage and complexity: motorways (limited-access highways for long-distance, heavy, or fast traffic), trunk roads (the most important roads in a system after motorways; largely pre-existing roads), primary roads (important roads that often link towns or main roads within cities), and secondary roads (not part of a major route, but nevertheless forming a link in the national route network). Figure 1 shows the coverage by road types and they cover approximately 57% (7432 km) of the total road length in Shanghai (~13,000 km in 2019). The selected road types have varying width with double and multiple lanes. To simplify the data analysis, we set 10 m to represent the roadway width in each direction. This provides a sufficient buffer to include the position-tagged pollutant concentration data from the mobile sensor network for analysis.

Map of major study roads in Shanghai—Map credit: OpenStreetMap.

The mobile sensor used here reports the average pollution concentration every 5 s, typically spanning a travelling distance of 28–167 m with average traffic speed ranging between 20–120 km h⁻¹, the same speed range found in this study. We set 200 m as the unit length of roadway segment for analysis to accommodate the different traffic conditions and the mobile monitoring data at a specific time is assigned to the pre-defined segments as input for spatial data analysis. Finally, the four road types were divided into a total of 51,118 segments, each with a buffer of 200 × 20 m covering two directional traffic.

ArcMap (Version 10.8.1, Esri, Redlands, CA, USA) was used to split the road polygons into segments, and the data was exported to Rstudio (Version 1.4.1103) afterwards for further analysis, such as GPS point positioning and pollutant concentration calculation.