This study explores the impact of the “topography height dependent horizontal localization scale scheme” in the GRAPES regional ensemble three-dimensional variational hybrid data assimilation system on the forecast performance of typhoons.Taking Typhoons Sudelor (2015) and Dujuan (2015) as examples, an application experiment of ensemble variational hybrid data assimilation system in these typhoons was carried out.Next, the forecast results of the typhoon track, intensity and precipitation caused by the typhoons were compared and analyzed by observing the “topography height dependent horizontal localization scale scheme” and the “non-topography height dependent horizontal localization scale scheme” of the ensemble variational hybrid data assimilation experiment.The results show that the typhoon track under the “topography height dependent horizontal localization scale scheme” is more similar to the actual situation, which can effectively reduce the track error of the typhoon, but the effect of the intensity is not obvious.In addition, compared with the “non-topography height dependent horizontal localization scale scheme, ” the “topography height dependent horizontal localization scale scheme” can reduce the underreporting phenomenon of typhoon precipitation forecast to a certain extent, and improve the TS score of each precipitation level.In general, the “topography height dependent horizontal localization scale scheme” can be used to improve typhoon precipitation forecasting skills and reduce typhoon track errors to a certain extent.