In London air pollution contributes to thousands of premature deaths a year, with many others suffering the adverse health effects of air pollution exposure.
However researchers at the University of Warwick and The Alan Turing Institute have won research funding worth £619,000 from EPSRC (The Engineering and Physical Sciences Research Council) for a project that aims to help keep London's air clean.
The project is led by the Department of Statistics, and Department Computer Science at the University of Warwick. Project partners from The Alan Turing Institute's data centric engineering programme which is supported by the Lloyds Register Foundation, have committed resources of over £300,000; with the Greater London Authority (GLA) committing further in-kind contributions, via staff time, complementary projects and new data sources. The project will also fund two postdoctoral researchers for the next three years.
The funding will enable researchers to build on their existing work on air quality and simulation-based inference to revolutionise pollution forecasting by combining modern machine learning and statistical methodology.
The project will develop and utilise computational techniques based around the simulation of large ensembles of "particles" to allow us to estimate and quantify our uncertainty. These techniques will be combined with models inspired by modern machine learning, particularly utilising deep Gaussian processes to describe the profile of atmospheric pollutants as they evolve over time.
Using these methods will allow for measurements from disparate sensors to be fused in a principled way allowing for as much information as possible to be extracted from sensor networks and combined with auxiliary measurements of related quantities while simultaneously providing an indication of how good these estimates are.
Dr Adam Johansen, from the Department of Statistics at Warwick and leader of the project comments:
"This is an exciting opportunity to develop techniques from computational statistics and advanced machine learning to make a step change in the monitoring and prediction of air quality in major urban environments. Involvement from stakeholders including the Greater London Authority will ensure that the project has genuine and lasting impact."
The researchers will introduce robust techniques for large scale inference in spatial settings which evolve over time that leverage synergistic developments in computational and robust statistics as well as machine learning and urban analytics.
Co-investigator Dr Theo Damoulas, Deputy Programme Director for Data Centric Engineering (DCE) at the Turing, added:
"I'm delighted that the Turing's DCE programme will continue to actively support this important research at the interface of several of our areas of strength that will allows us to contribute to one of the 21st century grand challenges."
