Fog is the main cause of low visibility.Therefore,improving the accuracy of fog prediction is an important measure to increase transportation security.The main progress in fog prediction in recent decades is reviewed from statistical model and numerical simulation,and the advantages and disadvantages for each method are summarized.Under some conditions,the application of new statistical methods could improve the accuracy of fog forecasting.However,the statistical methods could not get rid of the inherent deficiencies of statistics.Comparatively,numerical modeling shows more extensive applications and more potential in fog prediction.Under the current computer condition,coupling of high-resolution one-dimensional fog model with three-dimensional mesoscale weather forecasting model shows better forecasting capacity,compared with the mesoscale model alone.The coupled method is much feasible for fog prediction in airport and along expressway,which are prone to heavy fog and the observations are already available.With the ever-increasing computing power,the high-resolution three dimensional models covering wide aspect of fog processes,as well as multi-model ensemble prediction system will be the future direction in accurate fog prediction.