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

Abstract

In this thesis we propose a model to forecast air pollution levels within an area of a city that combines deterministic and statistical modelling approaches. It focuses on air pollution due to traffic patterns and meteorological conditions. For this study, we focused on an area of the City of Madrid, Spain. First, we included traffic data from sparsely located sensors in an optimisation model to infer traffic flow values at unmonitored streets within the network. These fluxes were then used to estimate traffic emission values that we input in the pollution model. Second, we modelled pollution dispersion in the area where advection was solved numerically using an explicit upwind scheme and diffusion using the implicit Crank-Nicolson method. Traffic emissions were included as a source term in the PDE. Meteorological data are also included in the pollutant dispersion model. A statistical model is used to analyse data provided by traffic and pollution monitoring sensors located in a city. Finally, the statistical model is used to identify and account for deviations from the assumptions of the physical model as well as to calibrate the model parameters.

Date of Award	15 Nov 2023
Original language	English
Awarding Institution	University of Bath
Supervisor	Paul Milewski (Supervisor) & Theresa Smith (Supervisor)