Abstract: In the GPS data processing, it always needs the high-sampling data of the precise clock errors. The general method is to fit or interpolate the 15min or 5min interval precise clock errors, which are provided by IGS, to the user needed interval. In this paper, the Gaussian processes (GP) method is used to densify the 5min interval clock errors to 30s interval with a suitable kernel function. The densification results are compared with the IGS data and the densified clock errors using the four order polynomial interpolation method. The results indicate that the GP method can provide high densification precision and is applicable to densify the clock errors of all GPS on-board atom clocks. The accuracy is up to the centimeter level.