Uncertainty quantification of steady and transient source term estimation in an urban environment

The growing concern of the effects of potential releases of chemical, biological or radiological materials in populated areas has led to an increase in urban dispersion modelling over the past several decades. More recently, there has been a surge of research in the area of source term estimation (STE), in which inverse computational methods are used to predict a source (release) location and strength based on sensor readings. Many studies to date have focused on idealized, free-field scenarios estimating continuous or instantaneous gaseous releases. There have been limited efforts including geometry (e.g. terrain or urban structures) effects using computational fluid dynamics (CFD) and no efforts towards estimating highly complex, transient sources. The first contribution of this work is the development of a proposed methodology to approximate the strength and location of transient source terms, whether mobile or changing in strength. The transient prediction tool is demonstrated to accurately predict the location and strength of sources exhibiting low to moderate transient behaviour. For fast moving or rapidly changing sources, the model becomes heavily reliant on adequate sensor positioning. The second contribution of this work is to quantify the uncertainty of the STE tool given uncertain measurements in atmospheric conditions (e.g. wind speed, wind angle and surface roughness) which are often sparse and prone to variations. The uncertainty quantification study is performed on steady, instantaneous, mobile and variable-strength sources in an idealized free-field setting. The wind angle was found to have the most effect on the prediction of the release position. The true release location was within 10-90th percentiles, with standard deviations on the order of one CFD cell size, for all cases assessed indicating a robustness of the algorithm to handle uncertain inputs. The free-field analysis is used as a baseline for applying the uncertainty quantification to predictions in a full-scale urban environment using the Joint Urban 2003 experimentation. Despite uncertain atmospheric conditions in the urban setting, the predicted source location was generally in the correct vicinity, although sometimes in the adjacent upwind street. It is recommended that the uncertainty quantification be applied to a probabilistic prediction tool to quantify the uncertainty of a statistical source term representation. Further, the analysis could be applied for more complex, highly transient and multi-source scenarios to fully assess the robustness of the algorithm.