Analysis methods

Routinely collected air quality data will be analysed by the DEFRA national evaluation team in parallel to the current evaluation and will allow an analysis of trends in air quality across the district. The approach estimates the time-varying background levels of air quality from the Automatic Urban and Rural Network (AURN) of continuous monitoring stations for any given time point, and subtracts these from locally measured air quality data at roadside locations so that the contribution and trends associated with local traffic and the implementation of the B-CAP can be identified. As air pollution levels vary with fluctuations in emissions and dispersing air-flows, the underlying changes can only be established by ‘de-weathering’ and ‘de-seasoning’ local data [55]. Any abrupt or gradual changes in these trends around the intervention date of the B-CAP will then be estimated using change point detection methods [56].

The DEFRA evaluation methods for processing data will be adapted to deal with the additional data collected across our 12 schools within our intensive observation period (mobile sensing with n = 240 citizen scientists) and extended observation periods (continuous monitoring from static sensors).

The personal exposure of school citizen scientists (n = 240) before and after the introduction of the B-CAP will be mapped onto a 500 m × 500 m grid centred on each school. Any difference in levels, over the under-lying trends (as determined by the DEFRA evaluation project), will be attributed as impacts of the B-CAP. If the intervention is successful, it is expected that any differences would be greatest for the schools located within the CAZ boundary; however, as the CAZ may influence overall traffic volume and traffic flow, the benefits may extend beyond the CAZ boundary. Our study design allows us to explore any dose-response relationships with schools located on CAZ boundaries and distal to the CAZ boundary. We will explore differences using a 2 (pre/post) × 3 (location: inside, bordering or distal to CAZ) design using the Friedman test. Data from personal sensors will also be used to create an ‘exposure index’ along key transit routes to and from schools, allowing us to explore impact of the B-CAP on exposure during the school commute.

For the extended observational period (static monitoring) we will generate an understanding of pollutant distribution in the area, we will use data interpolation between the sites alongside the annual pollutant profiles close to the project schools.