地理研究 ›› 2010, Vol. 29 ›› Issue (5): 852-862.doi: 10.11821/yj2010050009

• 地球信息科学 • 上一篇    下一篇

顾及数据特性的格网DEM分辨率计算

刘学军,王彦芳,晋蓓,马锦绢   

  1. 南京师范大学虚拟地理环境教育部重点实验室, 南京 210046
  • 收稿日期:2009-08-30 修回日期:2009-11-12 出版日期:2010-05-20 发布日期:2010-05-20
  • 作者简介:刘学军(1965-),男,陕西合阳人,博士,教授,博士生导师。主要研究方向为DEM及其地形分析、GIS空间分析、空间数据不确定性等。E-mail:liuxuejun@njnu.edu.cn
  • 基金资助:

    国家自然科学基金资助项目(40971230、40571120);国家863计划资助项目(2006AA12Z212)

Resolution analysis of grid digital elevation model based on data property

LIU Xue-jun, WANG Yan-fang, JIN Bei, MA Jin-juan   

  1. Key Laboratory of Virtual Geographic Environment, Ministry of Education, Nanjing Normal University,Nanjing 210046, China
  • Received:2009-08-30 Revised:2009-11-12 Online:2010-05-20 Published:2010-05-20

摘要:

水平分辨率是格网DEM的决定性变量之一,直接决定着DEM对地形的逼近程度和地形参数计算、地学模拟的精度。基于地统计学理论和非参数密度估计方法,提出了地形宏观变异和微观变异相结合的DEM适宜分辨率计算方法。即首先按系列支撑对采样数据进行格网划分,形成具有不同尺度的支撑域;然后利用正则化理论,对高程点数据进行正则化变换,通过不同支撑上正则化变量的半变异函数分析,探索不同支撑尺度上的地形宏观变异规律,从而确定地形宏观变异的最佳支撑尺度;第三,在所确定的宏观变异最佳支撑尺度内,借鉴非参数密度估计中直方图的理论方法,从微观角度计算DEM适宜分辨率。最后通过陕北黄土高原的实际采样数据,对本文提出的方法进行了验证。

关键词: 数字高程模型, 水平分辨率, 地统计, 正则化

Abstract:

Horizontal resolution, which directly determines the degree of the closeness for DEM representing landform, is one of key variables for grid DEM. It also has distinct effects on the accuracy of terrain parameters and geosciences simulation based on grid DEM. So many researchers focus on the study of how to choose or decide a suitable resolution. This paper puts forward a method of suitable horizontal resolution based on geostatistics and nonparametric density estimation, which combines macroscopic variance and microcosmic variance. Firstly, supports of various scales are made by dividing the sampled data with different grids. Then regularization theory in geostatistics is used to carry out regularization variation of different supports based on the elevation sampled data. In order to ascertain the optimal support size to express macroscopic spatial variability structure of terrain, semivariance at a lag of one support interval plotted against different support sizes. The support size at which the peak occurs may help to identify the predominant scale of macroscopic spatial variation of the raw data, so it is named optimal support in this paper. After that, the theory estimation of the optimal bin size that can estimate the probability density function is referred to decide the appropriate resolution in the optimal scale support. The resolution is the suitable grid size to express the microcosmic terrain variance. Finally, the method was verified in practice by taking the sampled data sited in the Loess Plateau which is in the north of Shanxi province. Anisotropy, compute efficiency, RMSE statistics and contour-matching are used to analyze the results. The paper shows that the results resolution meets the exact accuracy limits for the given quality index. It is proved that the method may serve as a guide to decide the resolution from sampled elevation point data considering variance information contents of the raw data and topographic expression. The paper did the experiments only by taking the data from the Loess Plateau for examples. Future work needs to involve other topographic data and other different scales. Also, methods for verifying the result resolution should be further considered.

Key words: Digital Elevation Model (DEM), horizontal resolution, geostatistics, regularization