DEM重采样误差空间分布格局及差异性分析  

Spatial point pattern of DEM re-sampling error and difference analysis of influence factor

作　　者：陈永刚[1,2] 汤国安[1] 祝士杰[1] 

机构地区：[1]南京师范大学虚拟地理环境教育部重点实验室,江苏南京210046 [2]浙江农林大学浙江省森林生态系统碳循环与固碳减排重点实验室环境科技学院,浙江临安311300

出　　处：《中国矿业大学学报》2011年第4期653-659,共7页Journal of China University of Mining & Technology

基　　金：国家自然科学基金项目(40930531);浙江省教育厅科研计划项目(Y201017891)

摘　　要：基于重采样在实现不同尺度数字高程模型(DEM)数据变换中的重要作用,分析了重采样误差的空间分布特征和采样方法及坡度对重采样误差的影响.以某地山区5 m分辨率DEM数据为基础,通过邻近点插值法(NEAR)、双线性插值法(BIL)和立方卷积插值法(CUB)3种不同方法对原始格网DEM进行重采样处理,利用随机样方法分析了大于5 m的重采样高值误差空间分布格局;同时,利用非参数Mann-Whitney,Kruskal-Wallis和Mood中位数检验方法检验了采样方法与坡度等级对重采样误差影响的差异性.结果表明:绝对值大于5 m的重采样误差成显著性空间集聚分布;BIL与CUB方法无显著差异,而NEAR方法与BIL,CUB方法差异显著;坡度等级对重采样误差具有显著影响.Based on the important role multi-resolution DEM（digital elevation model） resample plays in multi-scale DEM conversion, the spatial point pattern of re-sampling error and the in- fluence of sampling methods and the slope on the re-sampling error was studied. The original grid DEM which has a resolution of 5 meters in the mountain was resampled by three different methods of NEAR （Nearest neighbor）, BIL （Bilinear interpolation）, and CUB （Cubic convolu- tion）. The spatial point pattern of re-sampling error whose absolute value^is more than 5 me- ters was analyzed by the Quadrat Analysis method. Mean while, the different influence on resampling error between sampling methods and the slope was tested by Mann-Whitney, KruskalWallis and Mood Median Tests. The results show that the re-sampling error whose absolute value is more than 5 meters is clustered; BIL and CUB methods have no significant difference between each other, but NEAR is significantly different with BIL and CUB methods;Slope has the significant influence on DEM re-sampling error.

关 键 词：数字高程模型 重采样误差 非参数检验 空间分布格局 

分 类 号：G206.2[文化科学—传播学]