Many industrial cities in China continue to face severe particulate matter (PM) pollution. Strategically selecting green space locations within urban areas is an effective approach to mitigate PM pollution. However, existing studies predominantly focus on the mechanisms by which different pollution sources influence PM dispersion and distribution, often overlooking the heterogeneity caused by individual variations within the same type of pollution source. This oversight leads to inaccurate estimations of PM dispersion ranges, thereby hindering precise guidance for green space site selection. To address this gap, this study employs a Gaussian dispersion model under the principles of PM dispersion dynamics. We calculate the two-dimensional kernel density distribution of potential PM2.5 and PM10 dispersion for each influencing factor under varying Gaussian standard deviations (i.e., at different scales). This serves as a comprehensive representation of their spatial influence intensity on surrounding areas, forming a Gaussian Kernel Density Dataset that incorporates multi-scale impact characteristics of all factors. Subsequently, a dual-precision nested algorithm is designed to screen for the most effective and representative factors, constructing an explanatory variable dataset. Finally, a geographically weighted regression model is established to quantify the spatial characteristics and influence intensity of each factor on PM dispersion, enabling precise guidance for optimal green space allocation.