Abstract: According to application requirement for elevation data in RDSS, this paper has presented a method which adopted terrain classified data extraction and compression for elevation data. A model with data block organization and storage in data files, quad-tree indexing and file classification management has been designed in this paper. A method of data fitting and extraction combining bilinear interpolation, bicubic interpolation and moving surface interpolation has been proposed. Experiment result illustrates that the elevation database application model referred in this paper will reduce the elevation data amount effectively, regulate the precision loss, speed up data searching and accord with the requirement of RDSS.