基于邻接关系的土地利用变化空间关联规则研究

空间关联规则可在海量空间数据中挖掘出空间对象之间相互作用、空间依赖、因果或共生的模式。将空间关联规则方法引入土地利用变化时空特征分析中,以浙江省桐庐县2004~2009年的两种典型地类变化(城镇用地的转入、耕地的转出)为例,结合GIS的空间统计和空间分析功能,首先通过土地利用转移矩阵分析土地利用面积变化的整体情况,再采用经典Apriori算法挖掘这两种典型地类变化与其周边邻接的地类变化间的关联性规律。研究表明:(1)桐庐县土地利用呈现明显的非农化趋势,这主要归因于城镇化进程以及其作为特色发展的生态建设;(2)城镇用地的发展主要依靠耕地和农村居民点的转型,表现为集中连片的扩展方式,且城镇用地与交通水利建设用地的发展具有明显的共生现象;(3)耕地的同种转变类型在空间上呈现连片集聚分布,且耕地发展易受相邻地类的驱动或制约,尤其是林地和农村居民点。空间关联规则挖掘能够定量分析在空间邻接的情况下地类间的转变倾向以及变化的关联规律,为土地利用变化时空分析提供有用信息。

RESEARCH ON SPATIAL ASSOCIATION RULES OF LAND USE CHANGE BASED ON THE SPATIAL ADJACENCY RELATIONS

The spatial association rule is a basic method of spatial data mining, and it has been used in massive spatial data mining to find mutual interaction, spatial dependence, causal or symbiotic mode between spatial objects. Since land use/land cover change research has always been the pivotal frontier subject of global change research program, this study introduced spatial association rule technology into temporal and spatial characteristics research of land use change. Tonglu County was chosen as study area and its two typical land use change (the urban land transferred in; the arable land transferred out) from 2004 to 2009 was taken as the study cases. Combining the spatial calculation and spatial analysis functions of GIS, in this paper we first analyzed the general situation of Tonglu land-use area change and the process of urban land transferring in and arable land transferring out, by means of transition matrix of land use structure. Then classic spatial association rule algorithm (Apriori algorithm) was applied to dig out the spatial association rules between two typical land type changes and their surrounding adjacent land-type changes. The conclusions are as follows:(1) in Tonglu County, land use change showed significant non-agricultural trend, mainly due to its new urbanization process and idiomatical ecological construction. (2) Urban land development mainly depended on the transformation of arable land and rural settlements, both of which had high self-adjacent probability, demonstrating a concentrated and continuous expansion mode. Besides, urban land had obvious symbiosis relationship with construction land of transportation and water conservancy. (3) Arable land demonstrated a spatially continuous distribution feature when it comes to the same change type. Also it was likely to be driven or constrained by adjacent land type, especially forested lands and rural settlements. The application case demonstrates that spatial association rule mining gives quantitative analysis on the potential tendency of land types transformations and correlation regularity between land type changes under spatial adjacency circumstance, thus it can provide useful information for spatial and temporal characteristic research of land use change.