Abstract: Abstract: Calibrating travel times of seismic phases recorded at specific stations is an important means to improve event location and discriminating low magnitude seismic events recorded by a sparse station network. In an effort to improve sparse network event location, we firstly calculate first arrival P-wave and Lg-wave travel-time residuals for the calibration events relative to the IASPEI1991 or Mainland China velocity model at stations AAK, MAKZ, NIL, TLG and WMQ. We then use the residual data and non-stationary Bayesian Kriging approach to form continuous travel-time correction surfaces for the above mentioned stations. Lastly, the effectiveness of the Kriging interpolated travel-time correction surfaces is evaluated by locating, with and without corrections, a series of seismic events which occurred in Xinjiang Uygur Autonomous Region and have already had accurate locations. Results obtained show that the Krigged correction surfaces can improve the accuracy of event location by a sparse network and reduce the area of confidence ellipses effectively.