基于Logistic回归模型的土地利用格局模拟分析——以新疆吐鲁番市为例

土地利用/土地覆被变化(LUCC)是当前研究全球变化的重要内容,而区域土地利用格局模型模拟是LUCC研究的核心内容之一.本文采用2005年吐鲁番市土地利用现状图,将其转化为grid格式,将土地利用类型分为:耕地、园地、林地、草地、城镇用地、农村居民点用地、独立工矿用地、未利用地、其他用地.运用Logistic回归分析的方法,选择了对该地区土地利用/覆盖变化有重要贡献的12种驱动因子,在ArcView中进行空间分析,生成驱动力文件,利用SPSS 13.0的Logistic regression功能分析每种驱动力的权重(β值),建立logistic回归模型,并使用ROC方法对所有回归模型的拟合优度进行了检验,采用GIS软件绘制研究区土地利用格局卒间分布的经验统计概率图.ROC检验结果显示,各种土地类型的拟合度分别为:耕地0.934,园地0.915,林地0.858,草地0.793,城镇用地0.999,农村居民点用地0.929,独立工矿用地0.874,未利用地0.816,其他用地0.762.拟和度均大于0.75,拟和度较好.研究结果揭示了居民点、道路、渠系和地形分布等因素对研究地区土地利用格局形成与演变的重要决定作用,这为进一步研究吐鲁番市未来的土地利用动态变化情景奠定了基础,同时也可为该市的土地利用管理提供科学依据.

Simulation and Analysis of Land Use Patterns Based on Logistic Regression Model——A Case Study of Turpan

Land use/land cover change (LUCC) is an important content of geographical research on global change today, while spatial simulation on regional land use patterns is one of the key contents of LUCC research. Adopting the current situation picture of the land utilization of Turpan in 2005, turning it into grid form and merging pattern of the land use into: cultivated land,field,forest land,the meadow,land used for cities and towns ,area of residential area of countryside, independent industrial and mining land, unutilized area,others. Using the method analyzing, it has chosen 12 kinds of drive factors that has made important contribution to land use/coverage change of this area, set up driving force file with the method ArcView space analyses,analyzed the weight(β value)of each kind of driving force with logistic regression function of SPSS 13.0, set up logistic equation. ROC inspection result shows that fit degree of various kinds of land including: cultivated land is 0. 934, field is 0. 915, forest land is 0. 858, meadow is 0. 793, urban area is0. 999, area of residential area of countryside is 0. 929, independent industrial and mining area is 0. 874,unutilized land is 0. 816, other area is 0.762, degree is greater than 0. 75, degree is rather good. It is argued that these types of the analysis can provide the valuable information for modeling future land use change seenarios that need to consider local and regional conditions of actual land use. On the other hand,the probability maps of the land use types obtained from this study can also support the local government on making land management decisions.