Data assimilation-based reconstruction of urban pollutant release characteristics

Abstract

The development of efficient methodologies for the reconstruction of pollutant source characteristics in urban areas is a challenging issue due to the complexity of the flow in such environments, requiring the use of expensive computational fluid dynamics (CFD) solvers for the numerical estimation of the flow. In the present study, it is proposed to apply data assimilation (DA) techniques to the identification of both pollutant source characteristics and wind parameters from observations of the gas concentration at discrete locations in an urban environment. A non-intrusive ensemble-based variational DA scheme enhanced by proper orthogonal decomposition (POD) is deployed in conjunction with a Large Eddy Simulation (LES) solver, which provides high-fidelity estimation of the urban flow. The proposed methodology is proved to be able to accurately identify both pollutant source and wind characteristics with a computational cost of O(10) LES simulations. In addition, a sensor placement procedure based on global sensitivity analysis techniques is proposed in order to improve the performances of the assimilation process, which appear to be very sensitive to the choice of the measurement locations, in terms of computational cost and quality of the reconstructed parameters. Despite of several limitations in the use of the present approach for real time applications due to computational costs, this study demonstrates the potentiality of deploying sensitivity analysis and DA methodologies to solve inverse problems for urban flows in conjunction with high-fidelity CFD solvers.