Source term estimation for continuous plume dispersion in Fusion Field Trial-07: Bayesian inference probability...

Area of Science:

• Environmental Science
• Atmospheric Chemistry
• Computational Modeling

Background:

• Sudden gas releases and pollution emergencies necessitate effective source control for residential air quality.
• Accurate inverse source tracking methods are crucial for promptly identifying pollutant source parameters using diverse measurement data.

Purpose of the Study:

• To propose and evaluate a source term estimation (STE) method combining probability adjoint and Bayesian inference.
• To identify pollution source information from continuous point releases under unsteady wind conditions.
• To compare the performance of Metropolis-Hastings Markov Chain Monte Carlo (MH-MCMC) and Differential Evolution Markov Chain Monte Carlo (DE-MCMC) sampling methods.

Main Methods:

• Established the general form of the pollutant inverse transport equation.
• Employed a Bayesian inference probability adjoint inverse method for source identification.
• Utilized Fusion Field Trials 2007 data for single continuous point releases under unsteady wind.

Main Results:

• The Differential Evolution Markov Chain Monte Carlo (DE-MCMC) algorithm demonstrated superior convergence and higher accuracy compared to MH-MCMC.
• DE-MCMC proved particularly effective for highly nonlinear and multi-modal distribution systems.
• Integrating Union standard Adjoint Location Probability (UALP) as prior information enhanced accuracy and robustness by narrowing the sampling range.

Conclusions:

• The proposed Bayesian inference inversion technique, especially with DE-MCMC and UALP integration, offers a robust approach for point source identification.
• This research provides valuable insights into the practical application of Bayesian inference for real-world pollution source tracking.
• Further exploration of covariance matrix impact on inverse identification accuracy is recommended.