Abstract: A three-dimensional water vapor tomography algorithm is presented.The precise point positioning method is used to obtain the tropospheric slant water vapor.Then COSMIC occultation data is exploited to calculate the water vapor density to be used as initial values in the tomography iteration procedure.Afterwards,the inversion of three-dimensional water vapor is conducted using a fitting method of the Gaussian weighted distance.GPS datasets collected in Hong Kong on three days and the corresponding COSMIC occultation data are used for the algorithm test.RMS statistical value for differences between SWV derived from PPP and SWV derived by tomography is better than 0.87 mm.Comparing the water vapor density by tomography with that derived by the Radiosonde sounding data,their differences are typically below 1.1 g/m3,which demonstrates the feasibility of the presented three-dimensional tomography algorithm.