As a representation of smart and sustainable city development, bicycle-sharing system is one of the hottest topics in the domains of transportation, public health, urban planning, and so on. In this paper, a model is proposed for analyzing the potential reduction in emissions associated with the adoption of a bicycle-sharing system. Methods are proposed for extracting human travel modes from mobile phone GPS trajectories, together with a geometry based probability model, to support particle swarm optimization. A comparison study is implemented to analyze the model's computational efficiency. Based on the resulting optimal layout for the network of bicycle docking stations, and considering demand uncertainty, a multi-scenario integer linear programming model is proposed to optimize rebalancing procedures (i.e., moving bicycles between docking stations according to demand), to determine the detailed design-scale information required. Mobile phone GPS trajectories from approximately 3.7 million local mobilities are used to construct a case study for Setagaya Ward, Tokyo. The results show that, compared with the previous methods, the optimal layout solved by the proposed method could reduce emissions by a further 6.4% and 4.4%. With an increase from 30 to 90 bicycle stations, the adoption of bicycle-sharing can reduce CO2 emissions by approximately 3.1-3.8 thousand tonnes. However, emission reduction will maximally decrease by 21.26% after offset by bicycles production and rebalancing-generated emission.