Development of a Smart Algorithm for Air Pollution Sources Identification using Physical Dispersion Modeling and Bayesian Inference

Abstract

Air pollution plumes are commonly observed in the atmosphere above many cities and residential areas. These plumes may be the result of either a normal operation or an accidental release from certain sources. In both cases, it is of great importance to identify and characterize these sources for the assessment of the harmful effects of their resulting pollution fields and for the proper construction of an emergency response plan in case of accidental releases. This involves the inverse problem, from destination of pollution back to its source, and the inference of the different parameters characterizing this source given certain known or measured sets of observations. The aim of this thesis work is to introduce and develop a smart algorithm that is able to identify and characterize an air pollution source that is responsible for an observed concentration field of pollutants in a specific urban location. As an application, we will infer several parameters of an active source that is releasing air contaminants into the atmosphere of a selected domain around KAUST (King Abdullah University for Science and Technology) in the region of Thuwal, KSA. These parameters include the source geographic location, emission strength and emission duration. A stochastic approach using Bayesian inference and Monte Carlo sampling will be implemented to solve the ill-posed inverse problem and characterize the emitting source. In this scope, the forward Lagrangian model will be adopted to study the atmospheric dispersion of pollutants and resolve the urban characteristics of the domain. The implementation of this model will be done while considering the prevailing wind field as the main driving source and based on the well-known urban configuration of buildings and the natural topographic features of the location