Line source estimation of environmental pollutants using super-Gaussian geometry model and bayesian inference

Abstract

A line is a common geometry for pollution sources, e.g., outdoor traffic pollution, and is thus useful for developing a suitable source term estimation (STE) method. Most existing methods regard the source as a single point that only includes location and strength parameters; however, limited attention has been paid to the geometric information of the source. This negligence may cause errors, or even failure, in the STE. Therefore, this paper proposes a line source estimation method that combines Bayesian inference with the super-Gaussian function. This function can approximate the shape of sources with several intuitive coefficients, which are adjusted to their true value through Bayesian inference. The performance of the proposed method was evaluated through estimation of a line source in two cases: an ideal urban boundary layer, via simulation, and a complex urban square, via a wind tunnel experiment. The results demonstrate that this method is capable of identifying the source information without any prior geometric information regarding the source. Moreover, it was confirmed that the conventional point-based assumption method leads to failure in estimating the line source, which implies that geometry estimation is necessary for STE.