空间尺度上推方法的精度评价——以内蒙古锡林郭勒盟土地利用数据为例  

Accuracy analysis of up-scaling data: A case study with land use data in Xilin Gol of Inner Mongolia,China

作　　者：胡云锋[1] 徐芝英[1,2] 刘越[1,2] 艳燕[1] 王倩倩[1,3] 

机构地区：[1]中国科学院地理科学与资源研究所,北京100101 [2]中国科学院大学,北京100049 [3]辽宁工程技术大学测绘与地理科学学院,辽宁阜新123000

出　　处：《地理研究》2012年第11期1961-1972,共12页Geographical Research

基　　金：科技部973计划资助项目(2010CB950904);国家自然科学基金资助项目(40971223);国家科技支撑计划资助项目(2008BAH31B04)

摘　　要：不同的空间尺度上推方法会导致不同程度的信息丢失、信息歪曲等后果;但目前少有研究在较长的尺度序列上、对不同尺度上推方法所得成果开展精度分析。本研究首先提出尺度上推方法精度评价的三个准则,即:保持土地类型构成特征、保持土地面积精度、保持土地空间分布格局和斑块形态;继而使用格点中心值、最大面积斑块、最大聚合面积斑块等3种尺度上推方法,配合100m～50km土地利用数据开展尺度转化实验;最后基于上述评价准则和尺度上推实验所得的系列输出结果,分析了不同尺度上推方法的精度。研究表明:(1)格点中心值上推方法能更好地保留区域土地类型构成、土地面积精度等特征;(2)锡林郭勒盟地区土地研究的适宜尺度应小于10km,最大不应超过30km;(3)在尺度上推过程中,土地斑块的平均面积、形态以及空间分布格局对尺度上推成果精度有着重要影响。Different up-scaling methods induce information loss, misrepresentation, and other accuracy reduction effects in varying degrees. Up-scaling is fundamentally important to the researches relevant with the land science, ecological and climatic modeling, but few systematic specifications are available to evaluate the quality of up-scaling data. Thus, in this study we firstly put forward three rules on evaluating the accuracy of resulting data. （1） maintaining the landscape composition; （2） preserving the land area accuracy. and （3） keeping the spatial distribution patterns. Based on the three criteria and the following six indices, suitable models were developed. The land-use data with a map scale of 1 ： 100,000 for Xilin Gol of Inner Mongolia was used in this study, with three common UP-scaling methods applied, i. e. Cell-Center method, Maximum-Area method, and Maximum-Combined-Area method embedded in the GIS toolbox. In the procedure, based on these three methods, the original vector data was rasterized and converted to different scales, ranging from 100 m to 50 km. Then these maps were compared according to the evaluating models, and the results were drawn as follows. （1） the Cell-Center method pre- served the landscape composition better than the other two up-scaling methods, as well as the land area accuracy; （2） the appropriate scale for Xilin Gol region was less than 10,000 m, and the permittedly maximum scale should not exceed 30 km; （3） during the scalingup process, the average patch area, patch shape, and spatial distribution patterns have important effects on the output accuracy.

关 键 词：空间尺度 尺度转换 精度 信息熵 分维 土地类型 

分 类 号：P208[天文地球—地图制图学与地理信息工程] F301[天文地球—测绘科学与技术]