Low Earth orbit (LEO) satellites are being considered for expanding legacy terrestrial cellular networks. The end users may not be able to optimize satellite orbits and constellations, however, they can optimize locations of ground stations which aggregate terrestrial traffic and inter-connect with over-passing satellites. Such optimization requires a model of satellite visibility to decide when the satellite becomes visible to the ground station in a given geographical location. Our model assumes ideal Kepler orbits parameterized by six orbital elements to describe the satellite movements. The steps of Kepler orbit modeling are presented in detail to enable other studies requiring geometric representation of satellite trajectories in the 3D space or their 2D footprint on the Earth surface. The long-term mean time satellite visibility (MTSV) metric is defined to measure satellite visibility at a given ground station. Numerical results reveal that efficiency of optimizing the ground station locations is dependent on the satellite orbit and other satellite visibility constraints. The ground station location optimization is especially important when MTSV is improved by orthogonal time sharing schemes using multiple satellites on the same or different orbits. Similar conclusions can be drawn assuming other performance metrics such as the capacity of links between the ground station and the satellites.