From Google Maps to Air Quality: a big data approach to modelling real-time NO2 concentrations in an urban street canyon from road traffic data

In most European cities, the key air pollutants driving adverse health outcomes are nitrogen dioxide (NO2) and fine particulate matter (PM2.5), with 64% of new paediatric asthma cases in urban centres attributed to elevated NO2 levels (Achakulwisut et al., 2019). In the complex landscape of a city, a synthesis of techniques to quantify air pollution is required to account for variations in traffic, meteorology, and urban geometry.

Here, we present the results from a comparison study between measured air pollutant data collected at Marylebone Road, London and the output from a three-stage modelling chain. This site was chosen due to the availability of road-side air quality data collected within a street canyon (aspect ratio approximately equal to 1) and daily traffic flow in excess of 70,000 motor vehicles. The modelling chain consists of: 1) real-time traffic information of vehicle journey times, 2) speed-related emission calculations, and 3) air quality box-model to simulate the interaction of pollutants within the environment.

While the transport sector accounts for much of the outdoor air pollution in UK cities, a limiting factor of current techniques is that traffic is approximated at coarse temporal and spatial resolutions. In this study, we present a novel technique that helps to ‘fill in’ the gaps in our traffic data by harnessing the power of real-time queries to Google Maps to obtain travel times between fixed locations, enabling the derivation of average vehicle speeds. This dataset can then be used to determine more accurate emission factors for NOx. Total emissions are then calculated with the aid of traffic flow data and vehicle fleet characteristics. The air quality box model simulates photochemical reactions that form NO2, the exchange of pollutants with the background air aloft, and advection of pollutants along the street.

Hourly travel times and total vehicle flow data were collected between July and October 2019, totalling 905 observations and calculated emissions values. Meteorological data from Heathrow airport and background air quality from the Kensington AURN site were used as supporting inputs to the air quality box model. Each observation was treated as a starting point of the box model, and the simulation was run for 1 hour, with mixing due to advection occurring every 60 seconds. Results are promising; when using the full model chain modelled and measured NO2 concentrations are significantly correlated (r = 0.467, p < 0.000). In comparison, when a constant speed of 30 mph is used to calculate total emissions, therefore excluding the impact of congestion, the strength of the correlation decreases (r = 0.362, p < 0.000) and the model underestimates pollutant concentrations.

The applications of this model chain are vast. For any street that is covered by a suitable mapping platform and has available data on vehicle numbers, it would be possible to provide a real-time estimation of pollutant concentrations at a high temporal resolution. This could be utilised in several ways, such as: assessing policy implementation, and providing a high resolution input for air quality modelling and health exposure studies.