矢量数据在多尺度栅格化中的精度损失模型探讨

Models of accuracy loss during rasterizing landuse vector data with multi scale grid size

在进行空间分析时 ,由于栅格数据便于空间分析 ,因而通常将矢量数据转化成栅格数据进行空间分析。在转化过程中 ,采用不同尺度的栅格大小会造成怎样的精度损失 ?精度损失、栅格大小和平均斑块大小之间的关系是否可以用模型来表达 ?本文正是针对这样的问题 ,以重庆市 1∶10万的土地利用矢量数据为例 ,探讨了在不同尺度栅格大小 (从 30m× 30m到10 0 0m× 10 0 0m)的情况下 ,各种土地利用类型在转化过程中的精度损失情况 ,对精度损失与栅格大小和平均斑块大小之间的关系进行了定量分析。研究表明 ,精度损失与地类的平均斑块大小和栅格尺度之间的关系可以用模型来表达。该模型为 :Y =5 366- 0 179X -0 978ln (S) + 0 0 34 8Xln (S)。模型中Y为精度损失 ,S为地类的平均斑块大小 ,X表示栅格的大小。该模型的复相关系数为 0 93。

In the process of spatial analysis, the vector data are always rasterized into the raster data because of the convenience for spatial analysis. How the area accuracy loses during rasterizing needs to be answered when different grid sizes are adopted in order to choose a suitable grid size. The question is answered here by rasterizing the landuse vector data of Chongqing City with grid of different sizes. Firstly, the landuse vector data are rasterized with grid of different resolutions such as 30m,60m,100m,200m,300m,400m,500m,600m,700m,800m and 1000m. The value of each grid takes the landuse type code as the dominant area. Secondly, the area for each landuse type is calculated in grid. Thirdly, the area for each landuse type is also calculated in vector and used as the actual area. Both the number and average size of polygons for each landuse type are calculated. Fourthly, the relationship among the accuracy loss, grid sizes and the average block size of each landuse type is analyzed. Finally, models for describing the relationships are formulated. It is shown that: (1)The relationship between the accuracy loss(Y) of each landuse type except the high canopy grassland and grid sizes(X) which can be described by the linear regression models such as, Y = A + BX . Here, A and B are coefficients of the models. The correlation coefficients for each model of each landuse type are all above 0.82?The correlations of the absolute value of accuracy losses and grid sizes are positive. The accuracy losses denote that the area reduces compared to the true area when the area of the average block size of each landuse type is above 101 hectare, which is opposite when the area of the average block size of each landuse type is below 82.(2)The correlations between the accuracy loss and the area of the average block size (S) of each landuse type for each grid size can be described by the regression models such as, Y = A +B ln (S) . Here, A and B are coefficients of the models. The correlation coefficients for each model of each grid size are all above 0.72 except grid size of 30 M. The correlations are positive. (3) Generally, the relationship among the accuracy loss(Y), grid size(X) and the average block size(S) of each landuse type can be described by such model as, Y = 5.366 -0.179 X -0.978ln (S) + 0.0348 X ln (S) ?The correlation coefficient of the model is 0.93.