Precision Source Parameter Inversion for Typical Air Pollutant Emissions at Microscale: An Integrated PSO-NM Algorithm and Gaussian Dispersion Model Approach

• • PSO-NM coupled algorithm outperforms GA, NM, and other hybrids on Shubert, Hartmann, and Shekel functions, achieving mean values closest to theoretical optima with minimal standard deviation, ensuring reliable convergence in multi-dimensional, multi-extremum optimization. • • Single-point SF6 source inversion yields source strength relative deviations of -27.1% to 38.5% (mean absolute relative deviation <30%) and positional errors <10 m, demonstrating high accuracy and repeatability for rapid emission source identification. • • Multi-point source inversion shows stability but lower accuracy (deviations typically >30%); low-release sources achieve source strength relative deviations of 37.3%-70.4%, while high-release sources range from -71.4% to 161.8%, indicating the need for source-specific constraints. • • Positional parameter inversion errors remain below 50 m in most multi-point scenarios, with low-release sources achieving x0 deviations of -1.6 to 8.2 m, yet y0, z0, and distance parameters are less accurate, highlighting the impact of source interference and meteorological variability.