Reliable local earthquake locations depend on many factors including the quality of earthquake data, the spatial distribution of network stations, the location algorithms, and the local and regional seismic velocity models. The uncertainties of P- and S-wave arrival times at various seismic stations are, in large part, responsible for the uncertainty in the determination of crustal velocity models and earthquake locations. We took the advantage of the high-quality and high-dynamic range PANDA data from a 33-month seismic array experiment in the central United States to investigate the importance of crustal velocity models, especially the S-wave model, on earthquake locations. Several crustal models are used to relocate a group of selected earthquakes from the PANDA database. It is clear that earthquake depths are determined less reliably than the epicenters when the shear-wave velocity model is not properly determined. The mislocations in some examples are significant that the importance of determining reliable P and S velocity models independently cannot be overlooked. If the crustal P- and S-wave velocity structures in the region can be well determined, reliable earthquake locations are possible even with only a few accurate P and S readings.