The rapid and accurate prediction of the release source and concentration of pollutants remains a crucial issue in emergency rescue. A suitable gas diffusion model, an appropriate distribution of monitoring points, and an inverse calculation method are required to solve this problem. Optimization modeling using monitoring data is proposed in this paper. The objective function is established using the sum of the squared errors between the observed and calculated concentrations. The Gaussian plume model was improved with ground reflection coefficient and modified He and compared with AFTOX as the monitoring data to increase the accuracy of the inverse calculation of the source strength. The stochastic inertia weight particle swarm optimization algorithm is utilized to meet the needs of emergency rescue operations for the more accurate prediction of the leakage point. The results show that it is necessary to establish a gas diffusion model for each location to ensure the accuracy of the source strength estimate.