The research of the regional ecological environment becomes more important to regional Sustainable Development in order to achieve the harmonious relationship between the person and the nature. The advent of spatial information technologies, such as GIS, GPS and RS, have great enhanced our capabilities to collect and capture spatial data. How to discover potentially useful information and knowledge from massive amounts of spatial data is becoming a crucial project for spatial analysis and spatial decision making. Particle Swarm Optimization has a powerful ability for reasoning and semantic representation, which combined with qualitative analysis and quantitative analysis, with prior knowledge and observed data, and provides an effective way to spatial data mining. This paper focuses on construction and learning a Particle Swarm Optimization model for spatial data mining. Firstly, the theory of spatial data mining is introduced and the characteristics of Particle Swarm Optimization are discussed. A framework and process of spatial data mining is proposed. Then we construct a Particle Swarm Optimization model for spatial data mining with the given dataset. The research area is focused on the distribution of pollution sources in Wuhan City. The experimental results demonstrate the feasibility and practical of the proposed approach to spatial data mining. Finally, draw a conclusion and show further avenues for research. Through the empirical study, it has been proved that Particle Swarm Optimization algorithm is feasible and the conclusion can provide instruction for local environmental planning.