Hybrid physical-statistical models for air quality prediction from traffic data

点击翻译按钮获取中文摘要

In the present study, we shall develop a hybrid physical-statistical model for air quality prediction. The research is a collaborative work between AVL and the EPSRC Centre for Doctoral Training Statistical Applied Mathematics (SAMBa). The project aims to combine deterministic and statistical modelling to forecast air pollution levels within a city. The focus is on air quality change resulting from traffic patterns, vehicle types, urban layout and meteorological conditions. The aim is to move towards incorporating real time data from traffic and pollution monitors to inform the forecast. It is also an aim to develop methods that can distinguish the relative impact of traffic from other sources of pollution. First, a physical and chemical model will be used to describe automobile traffic and pollutant concentrations. A statistical model will be used to analyse data provided by traffic and pollution monitoring sensors located in a city. Subsequently, the statistical model will be used to identify and account for deviations from the assumptions of the physical model as well as to calibrate the model parameters. Finally, the goal will be to develop a hybrid system where both physical and statistical models work together to improve forecasts. The supervisory team consists of Prof. Paul Milewski who is an expert on physical modelling of fluid and continuum processes, and Dr Theresa Smith who is an expert on spatial statistics. Together they span the expertise needed for the project. The AVL contact is Gerhard Schagerl who will provide guidance on the industrial impact of the project.