An urban scale inverse modelling for retrieving unknown elevated emissions with building-resolving simulations

International audienceThis study illustrates an atmospheric source reconstruction methodology for identification of an unknown continuous point release in the geometrically complex urban environments. The methodology is based on the renormalization inversion theory coupled with a building resolving Computational Fluid Dynamics (CFD) modelling approach which estimates the release height along with the projected location on the ground surface and the intensity of an unknown continuous point source in an urban area. An estimation of the release height in a three-dimensional urban environment is relatively more difficult from both technical and computational point of view. Thus, a salient feature of the methodology is to address the problem of vertical structure (i.e. height of a source) in atmospheric source reconstruction in three-dimensional space of an urban region. The inversion methodology presents a way to utilize a CFD model fluidyn-PANACHE in source reconstruction in the urban regions. The described methodology is evaluated with 20 trials of the Mock Urban Field Setting Test (MUST) field experiment in various atmospheric stability conditions varying from neutral to stable and very stable conditions. The retrieved source parameters in all the 20 trials are estimated close to their true source. The source height is retrieved within a factor of two and four in 55% and 75% of the MUST trials, respectively. The averaged location error for all 20 trials is obtained 14.54 m with a minimum of 3.58 m and maximum of 34.55 m. The averaged estimated release rate for all trials is overpredicted within a factor of 1.48 of the true source intensity and in 85% of the trials, it is retrieved within in factor of two. In source reconstruction with non-zero measurements, it was observed that the use of all concentration measurements instead of only non-zero essentially makes only the small differences in quality of the source reconstruction and gives a little additional information for better constraining the source parameters. A posteriori uncertainty analysis in the retrieved source parameters is also performed by adding a controlled noise in the concentration measurements and the source reconstruction results were also compared with an earlier study based on the stochastic Bayesian approach for identical 14 MUST trials. The study is useful for emergency regulators to detect an unknown accidental or deliberated continuous point releases in urban regions