Geographic models are becoming key methods for geographic environment research. In addition to the progress in model sharing and reusing, geographic simulation (geo-simulation) knowledge sharing and application are increasingly attracting much attention. This study aims to provide both models and geo-simulation knowledge simultaneously to support Web-based simulation. Firstly, we propose a geo-simulation approach consisting of three modules concerning data, model and knowledge with a conceptual framework. Secondly, taking the Storm Water Management Model as instance, a Web-based prototype system is implemented with geo-simulation knowledge graph constructed based on literatures using bilateral long shortterm memory - conditional random field (BiLSTM - CRF). Thirdly, application of the system in Nanchong, China is conducted with user experiences analysis. The results demonstrate that the proposed approach can effectively provide geo-simulation knowledge. Moreover, it can alleviate the difficulty of the geographic simulation process for diverse modelers, and further improve the use of the model simultaneously.