Retrieval of Unknown Number of Source Terms in Dispersion Events Involving Multiple Point Sources

Abstract

In dispersion events involving multiple sources, estimation of correct number of sources is imperative prior to their localization or source term estimation. The study pertains to a dispersion scenario of multiple point releases simultaneously emitting a nonreactive tracer that results into a mixture of concentrations, as measurements, sampled at the receptors. The objective is to estimate correct number of sources along with their release parameters (mainly locations and strengths) using a limited merged set of concentration measurements. A new methodology, based on five criteria coupled with a multiple release identification algorithm, is proposed here. Criterion helps in discriminating the releases and accounts for detection and false‐alarm probability associated with each retrieved source. The number of source terms are estimated based on sum of scores achieved by each predicted source under each criterion. The methodology is evaluated with synthetic, noisy synthetic, and pseudo‐real data prepared from India Institute of Technology Delhi diffusion experiment conducted in 1991 at Delhi, India. With synthetic data, the number of point sources and their parameters are retrieved exactly as their true values. With pseudo‐real data, true number of sources are retrieved. The source locations and strengths are retrieved within 30 m and a factor of 3, respectively. Synthetic simulations with controlled measurement noise shows that the proposed methodology is robust in presence of model uncertainties and can be applicable in real scenarios.

Keywords

• detection
• false alarm
• inverse modeling
• multiple point releases
• source estimation