Model calibration is the cornerstone of spatial interaction models in many geographic, transportation and marketing analysis. Conventional questionnaire approaches that collect data for model calibration are both labor-intensive and time-consuming, and generally show a poor response rate. This study takes advantage of increasingly available vehicle GPS trajectory data to conduct spatial interaction model calibration. A Huff model for retail trading area analysis was used as an example. Model calibration and parameter validation were conducted based on over 63,000 taxi GPS trajectories for seven major shopping centers in Wuhan, a large city in China. The results were positive and in general showed satisfactory descriptive and predictive capability. This study demonstrated the feasibility of using the emerging technology to calibrate spatial interaction models (and also showed the potential for use in other related studies). The main advantage of using these new data sources is that they allow efficient use of increasingly available positioning data, which is easier to collect than conventional customer surveys, and usually with larger data sizes. It also allows inferences to be made about distance-decay rates based on accurate computation of travel time and distance. This could save both time and expense in many related areas of research, while achieving high quality model calibration results.