It is generally believed that the GPS need be calibrated with the relative time observed with the Hafele-keating experiment. But, it is not true.
1. The altitude of the GPS satellite is less than 25000km. In less than 0.1 seconds, the clock in the satellite can be calibrated one time with the clock on the surface of the Earth. Therefore, the clocks between on the surface of the Earth and in the satellite were calibrated many times in a day. But, the time effect in the Hafele-keating experiment only can be obtained in a long time, such as a day.
2. The Hafele-keating experiment could not be credible for two reasons: First, the Newtonian gravity has a bigger effect on an atomic clock. The variation of the height on the level less than 10cm by the Newtonian gravity can be observed with an atomic clock. However, this variation cannot be observed in the Hafele-keating experiment. The force F=GMm/r^2 between the Earth (M) and the atom (m) can affect the motion of the atom in this clock. (It is noted that, it does not affect the frequency of the atom.) This effect has been used as a usual method to measure the Newtonian gravitational constant G. But, in the Hafele-keating experiment this effect was not considered. Thus, the Hafele-keating experiment is invalid theoretically. Second, an atomic clock is very precise. A little change, such as motion, temperature, magnetic field, can make very big change of the readings of an atomic clock. Figuratively, as a person walk near to an atomic clock, the readings of it are changed. Thus, the Hafele-keating is not credible under the condition that Einstein’s time effect is right.
3. The gravitational effect of time cannot be measured with a random clock. For example, in different fountain standard atomic clock, the atoms run different length L. A phase shift of the motion of the atoms caused by the force F=GMm/r^2 between the Earth (M) and the atoms (m) is determined with the length L. Thus, with different L, the gravitational effect on the clock is different.