Fast Estimation of Pollution Sources in Urban Areas using a 3D LS-RBF-FD Approach

Source term estimation (STE) is a field of growing interest in the context of air pollution, both for people living in urban areas and for decision makers. Thus retrieving maps of sources of pollution in an urban context is a necessity. Since urban pollution mainly depends on car traffic conditions, it is important to develop fast estimation methods to quickly and enough accurately identify highly-polluting vehicles. The challenge is high since the problem requires the inversion of distributed models defined on a 3D heterogeneous domain including complex obstacles. This paper proposes an estimation method based on a flexible least squares-radial basis function-finite differences (LS-RBF-FD) reduced model of an advection-diffusion PDE on 3D heterogeneous domains representing complex urban areas. The STE problem is solved by using an adjoint-based method relying on the reduced model to effectively estimate pollutant sources given a limited number of measurements. The paper provides preliminary results demonstrating the potential of the proposed approach.