Improving Air Quality in Cities
Using data to forecast air pollution across London and inform policy interventions
Modern cities face constant challenges to public health, including air pollution and disease outbreaks such as COVID-19. In London alone, an estimated 9,000+ people die each year due to breathing in pollutants, and there were 700,000+ cases of coronavirus between January 2020 and April 2021. Dr Theo Damoulas’s research aims to give the Greater London Authority (GLA) and other local authorities dynamic information on busyness (activity levels) and air quality in the capital’s streets, in order to aid public health measures. Through having access to the latest busyness data and patterns, Transport for London (TfL) has been able to widen pavements and close roads to vehicles in order to support social distancing measures. Meanwhile, the availability of air pollution data allows cyclists and people on foot to avoid problem areas, and guides targeted intervention strategies.
The GLA, TfL and public health bodies have access to a wide range of tools and data sources for monitoring mobility, transportation, traffic activity, and air pollution. Networks of ground and mobile sensors, satellites, JamCam cameras and CCTV footage, as well as public transit activity metrics, produce large and diverse data sets with the potential to improve monitoring and understanding. Machine learning algorithms can be used to identify patterns in London’s busyness and air pollution. However, these methods perform poorly for complex data, particularly with the strong spatial and temporal (space and time) factors seen in mobility, transportation, traffic activity, and air pollution data. The challenge lies in adapting statistical tools to address large scale complex problems of this nature. Once successfully adapted, such tools can be applied to an extensive variety of data-driven challenges which present themselves in any large metropolitan city.
Two approaches are combined to break down obstacles in the underlying data. The first relies on computer modelling techniques which can adapt both to unexpected data and to initial models that may not have been specified correctly. The second uses statistical sampling methods to break down large problems into those that are smaller and more manageable.
When combined, these approaches address both the unpredictability of the data and its sheer size. Further tailoring is then used to create a data science platform to tackle the real-world scenarios of London’s busyness and air pollution. Results can then be widely communicated through an application programming interface which simplifies the development of software and apps. These allow for both live-updates and detailed results to be communicated to citizens, TfL and the GLA.
Dr Damoulas and the team at the Alan Turing Institute have already developed an application programming interface to monitor London’s activity using the transport and road networks, and footfall in London. This interface is used by both TfL and the GLA to closely monitor busyness of the city’s streets and enable comparisons to normal levels of activity, supporting recovery from the COVID-19 pandemic. TfL uses the interface directly in their control room dashboard to understand busyness and dynamically manage roads, cycle lanes and pavements, accommodating social distancing measures.
The same statistical approach has also begun to provide a detailed image of London’s air quality. An app in development will produce up-to-date and local air pollution forecasts for people walking, running or cycling through London. This will allow them to change their route and avoid the worst polluted areas. Local authorities in London now have accurate means to incorporate new sources of data in their air quality and activity level analyses, to compare different strategies which combat air pollution and COVID-19 transmission. Not only will this enable improved protection of public health, it will give better value for money in public spending and provide a better, safer course of action for the coronavirus pandemic. Warwick researchers are exploring the extension of the data-driven approaches system into further cities, including Newcastle and Sydney.