基于信息损失量的坡度精度分析
作者简介:陈楠(1975-),男,博士,副研究员,主要从事数字地形分析方面研究。E-mail:fjcn99@163.com
收稿日期: 2014-02-10
要求修回日期: 2014-04-02
网络出版日期: 2014-11-01
基金资助
国家自然科学基金重点项目“基于DEM的黄土高原地貌形态空间格局研究”(40930531)
国家自然科学基金项目“基于信息论的地形因子精度研究”(41001254)
国家支撑计划子课题“南方红壤水土保持评估与决策”(2013BAC08B02)
福建省自然科学基金项目“信号处理理论在地形因子精度研究中的应用”(2014J01155)
Study on Slope Accuracy Based on Information Loss
*The author: CHEN Nan, E-mail:fjcn99@163.com
Received date: 2014-02-10
Request revised date: 2014-04-02
Online published: 2014-11-01
Copyright
从DEM中可获取众多地形因子。其中,坡度作为基本的地形因子,其精度对于退耕还林、水土流失评估等研究具有重要影响。但是,由于DEM只能是真实地表形态的近似表达。如何模拟坡度误差的规律成为研究的热点。本文在黄土高原地区选取代表不同地貌类型的神木、绥德、延川、富县和宜君5个研究样区,以1:1万地形图为基本数据源,建立不同分辨率的DEM,用以分析坡度信息损失量随DEM分辨率变化的规律。故此,提出了基于单个栅格方式的坡度信息量损失指标,并在每一样区随机选取的20个子实验样区得到了其与DEM分辨率的函数关系,以及相关的经验公式,据此求得所需的最适宜DEM分辨率,并在每一样区随机地选取16个子样区进行了检验。
陈楠 . 基于信息损失量的坡度精度分析[J]. 地球信息科学学报, 2014 , 16(6) : 852 -858 . DOI: 10.3724/SP.J.1047.2014.00852
As a fundamental form of spatial data, Digital Elevation Model (DEM) plays a considerable role in many fields, such as surveying and mapping, territorial planning, and national defense. DEM is usually discrete in data structure and is an approximation of the real terrain. Though the technology of obtaining the elevations of sampling points has been improved, the real surface can only be represented by discrete grids. This means that the slope can only be computed by using the elevations of discrete grids. Therefore, the derived slope will have an unavoidable error. However, it is the slope that determines the transfer of matters and energy on the terrain surface and its accuracy exerts a great influence on the related researches and applications. 5 areas of Shenmu, Suide, Yanchuan, Fuxian and Yijun in Loess Plateau that represent the typical geomorphology are chosen as the sample areas, in which the DEMs with resolutions of 5 m, 15 m, 25 m, 35 m, 45 m, 55 m, 65 m and 75m are established according to the topographic map scaled at 1:10000. The slope is derived by using the algorithm of the third-order finite difference weighted by the reciprocal of squared distance. The slope derived from DEM with a resolution of 5 m is assumed to be the true value, and the slopes derived from other DEMs are taken as the investigated subjects. Each area is divided into 36 subareas. The relationship between the index and the resolutions of DEMs is investigated in 20 subareas selected randomly from the 36 subareas. We proposed an index for the slope information loss based on the single grid, and obtained empirical equations to describe the relationship between the resolutions of DEM and the values of the index. The equations are tested in the other16 subareas and proved to be effective. Using the equations, we can get the most suitable resolution for DEM when the index is known. The users, who want to select proper resolutions for DEM to minimize the amount of data and reduce the cost while ensuring the accuracy of the slope information of DEM, will benefit from the empirical equations.
Key words: DEM; slope; entropy; resolution
Tab. 1 The basic information of the sample testing areas表1 实验样区基本情况表 |
样区编号 | 样区所在县名称 | 地貌类型区 | 平均高度(m) | 平均坡度(°) |
---|---|---|---|---|
1 | 神木县 | 长城沿线风沙-黄土过渡区 | 1198 | 9 |
2 | 绥德县 | 黄土丘陵沟壑区 | 995 | 29 |
3 | 延川县 | 黄土梁峁状丘陵沟壑区 | 1089 | 31 |
4 | 富县 | 黄土高原沟壑区和丘陵沟壑区交错过渡地带 | 1299 | 27 |
5 | 宜君县 | 黄土长梁残塬沟壑区 | 986 | 19 |
Fig. 1 The sketch map of the sample areas图1 研究样区示意图 |
Fig. 2 The grids of DEMs with different resolutions图2 不同分辨率的DEM对应的栅格图 |
Tab. 2 Statistical parameters of coefficients between the index q and RMSE表2 指标q与中误差相关系数统计表 |
样区名称 | 相关系数最小值 | 相关系数最大值 |
---|---|---|
富县 | -0.99970 | -0.99213 |
神木 | -0.99985 | -0.98091 |
绥德 | -0.99902 | -0.99218 |
延川 | -0.99830 | -0.98826 |
宜君 | -0.99990 | -0.99578 |
Fig. 3 The relation of the index q and the resolutions in 20 experimental sample subareas of the Shenmu sample area图3 神木样区中20个实验样区分辨率与指标q关系图 |
Fig. 4 The relation of the index q and the resolutions in 20 experimental sample subareas of the Fuxian sample area图4 富县样区中20个实验样区分辨率与指标q关系图 |
Tab. 3 The parameters of the regression equations between q and r in experimental sample subareas of the five sample areas表3 5个样区实验子区q关于r回归方程的参数表 |
样区名称 | s值范围 | k值范围 | 相关系数范围 | sig.f范围 |
---|---|---|---|---|
神木 | -1.714-0.762 | -1.683-0.686 | 0.995-1.000 | 0.000-0.001 |
绥德 | -2.040-0.896 | -2.069-0.960 | 0.989-0.998 | 0.000-0.002 |
延川 | -1.838-0.906 | -1.874-0.897 | 0.989-0.995 | 0.001-0.002 |
富县 | -1.654-0.667 | -1.631-0.675 | 0.997-1.000 | 0.000-0.000 |
宜君 | -1.786-0.809 | -1.697-0.754 | 0.996-1.000 | 0.000-0.001 |
Tab. 4 The calculation formulas of the resolution according to the q in experimental sample subareas of the five sample areas表4 5个样区实验子区中根据q计算分辨率公式表 |
样区名称 | 分辨率的计算公式 | 相关系数 | sig.f |
---|---|---|---|
神木 | exp(-1.7048q+1.4643) | 0.732 | 0.036 |
绥德 | exp(-1.4635q+1.2608) | 0.978 | 0.003 |
延川 | exp(-1.6521q+1.1118) | 0.980 | 0.002 |
富县 | exp(-1.6694q+1.5935) | 0.974 | 0.003 |
宜君 | exp(-1.8079q+1.4786) | 0.737 | 0.031 |
3.2.2 5个样区检验子区中指标q与分辨率关系 |
Fig. 5 The resolutions computed by using q in the testing areas of the Shenmu sample area图5 神木样区的检验样区中由q求分辨率的结果图 |
Fig. 6 The ratio of the average q and its theoretical extreme value in the five sample areas图6 5样区中指标q平均值与理论极值的比值曲线图 |
Fig. 7 The absolute values of variation coefficient in the five areas图7 各样区指标q变异系数绝对值变化曲线图 |
The authors have declared that no competing interests exist.
[1] |
|
[2] |
|
[3] |
|
[4] |
|
[5] |
|
[6] |
|
[7] |
|
[8] |
|
[9] |
|
[10] |
|
[11] |
|
[12] |
|
[13] |
|
[14] |
|
[15] |
|
[16] |
|
[17] |
|
[18] |
|
[19] |
|
[20] |
|
[21] |
|
[22] |
|
[23] |
|
[24] |
|
[25] |
|
[26] |
|
[27] |
|
/
〈 | 〉 |