Quadtree　is　widely　used　in　large-scale　terrain　visualization　system,　to　manage　and　store　terrain　data.　Thus,　the　structure　of　quadtree　is　the　key　factor　which　affects　system’s　IO,　data　structure　and　the　design　visualization　algorithms.　Rather　than　emphasizing　how　to　design　a　ingenious　rendering　algorithm　to　accommodate　a　complex　quadtree,　we　focus　on　the　structure　and　data　layout　of　quadtree　in　which　not　only　surface　approximation　with　a　multiple　levels　of　detail　hierarchy　but　also　getting　next　data　could　be　made　easy.　We　have　two　achievements　in　this　paper.　Firstly,　we　simplify　the　design　of　quadtree.　Secondly,　we　design　a　bottom-up　algorithm　to　calculate　the　roughness　of　terrain　with　the　result　of　sub-node,　the　algorithm　significantly　improves　the　efficiency　of　the　terrain　node　evaluation.　Experiment　results　show　that　the　algorithms　have　been　implemented　for　sampling　large-scale　terrain　with　high　quality.　©　2014　WIT　Press.