Abstract: By comparing several typical terrain simplification algorithms, local adaptive optimization simplified terrain model (LaostModel for short) was be established, which is based on quad-tree model and supports massive 3D terrain data. With LaostModel, many quad-trees are built by the mode which the massive terrain data is organized by multi-resolution blocked processing. Also, the LaostModel algorithm dynamically split or merges quad-tree data block according to viewpoint as well as the degree of roughness of local terrain, and the different level terrain is display by real-timely adjusting the different quad-tree nodes in a data block. In addition, LaostModel can solve a lot of other problems, such as, the crack problem caused by different resolution grid mosaic of terrain which is solved by breadth-first search algorithm, and the problem that how to determine which quad-tree node needs to be drawn which is solved by "bounding box" method, and so on. Testing of prototype system shows that the model fits satisfactorily and offers a useful reference for further studying.