土地资源的多级网格数据结构建立与应用研究

传统的基于行政区的土地统计数据不能完全表现区域内部土地利用的空间分异特征,以武汉市为实验区,对基于网格的统计信息算法STING(Statistical Information Grid-based method)进行扩展,以景观多样性指数为定量化指标对实验区进行四叉树划分生成不均匀多级网格,建立一种拟合了行政区划界线的不均匀的多级网格结构来存储、管理和分析土地数据。并以此多级网格数据结构为平台计算和生成实验区人口密度空间分异渲染图,初步抽取了人口分布与土地利用之间的关系。实验表明,基于多级网格的统计方法能更好地表达土地利用及其相关数据的空间分异性,利于对土地资源数据的进一步挖掘以抽取所需知识。

Researches on the Multi-Grids Land Resources Data Structure

The traditional land-use statistics based on districts can not represent the spatial differences of land use structure entirely. At first, the paper extends the algorithm of Statistical Information Grid-Based (STING) and then puts forward a kind of multi-grids land resources data structure according to the spatial difference of the regional land use structure. It’s theory and method can be generalized as follows:Because different land use researches and applications have different data requests, the land spatial data and other natural environment/socioeconomic data should be mutually matched conveniently. So according to a certain method the land resources data space needs to be divided non-uniformly into more layers which can be called multi-grids land data structure to satisfy the distribution of data items. Namely the region with crowded data (land use structure is fine) should contain the massive small grids, but the region with sparse data (land use structure is extensive) only includes a few big grids. Through analysis the landscape multiplicity is selected as a quantitative index for building such a multi-grids land data structure and the expression of landscape multiple index is:H=-∑(S_i/∑S_i) ×log₂(S_i/∑S_i) (i=1,2,…,m)where H is landscape multiple; m is the number of the type of landscape in the region; and S_i is the area of one land use/cover type. RS images contain rich land use/cover information, so the land classification results using RS images can be applied to figure out the value of 'H’ index. In the paper the experiment data are TM images of Wuhan district(1998-10), with an overall precision of land use/cover classification result being 89% and KP value 0.87. Taking the entire experimental land data space as the root point the quad tree division is done. In advance a threshold named N of 'H’ should be decided (for example taking one half of the H_max as a threshold) to judge whether a grid point continues to be divided or not. If the 'H’ value of a grid point is smaller than N, then the division stops otherwise continues. In the experiment a multi-grids structure with 16 layers and 2728 leaves was obtained. The multi-grids land resources data system provides a good data index structure for further data analysis and researches, so according to different applications it can be filled with many kinds of data and it’s structure can be changed conveniently. For example, the traditional statistics of population density is spot population density based on such a premise hypothesis that population in one administrative region is even, but in fact it is uneven because of natural/social resources’ uneven distribution. Based on the multi-grids land data structure, an experiment on simulation of surface population density in Wuhan district was made. Different saturated red colour was used to indicate the different population ranks and the simulation of population density spatial distribution in Wuhan district was obtained.The simulated population density shows that the multi-grids land data structure can integrate land use data with other socioeconomic statistics conveniently and neatly to complete some simple land data mining tasks such as the effective data computation or analysis, and can reflect the land use/cover status of one region.