This paper proposes a data-driven method to simulate pedestrian flow within a target area by integrating low-granularity GPS location data with traffic data measured by sensors at specific spots. The primary challenge addressed by this method is the combination of different data sources to enhance simulation accuracy. Our approach aggregates GPS data points into Points of Interest (PoIs) within the area and employs Gibbs sampling to estimate a two-dimensional Gaussian Mixture Model for departure and travel times between PoIs. This distribution enables the simulation to replicate realistic pedestrian movements between PoIs. Additionally, we incorporate spot traffic data measured by LiDAR sensors to adjust the simulated pedestrian counts, further improving the accuracy of the simulation. We evaluated the proposed method using GPS logs from a large public park, known for being a popular tourist destination, and spot traffic data measured by LiDAR at five locations within the park. The simulation effectively reproduced pedestrian flow patterns, achieving an average cosine similarity of approximately 0.8 between the actual and simulated time-dependent population density distributions. Furthermore, the method predicted visitor numbers at park facilities with a prediction error of 13.6%.