Abstract: Taking Beijing mountainous areas as a study case, this paper analyzes and simulates the cultivated land change between 1996 and 2004 using ARCGIS and Logistic regression model with the data of land use map, DEM, population, GDP, and so on. The results show that: 1) during the year of 1996-2004, one third of cultivated land lost and mainly transited to built-up land and forest land; 2) by spatial Logistic regression, sople, elevation, GDP density, population density, and distance to the nearest road were the most influential factors in controling cultivated land distribution in the Beijing mountainous areas; 3) the total correction percentage of the results simulated by the spatial Logistic regression model was 78%, the correction percentage of cultivated land was 70%, and the non-cultivated land was 79%.

Key words: cultivated land change, GIS, Logistic regression model, mountainous areas around Beijing

摘要： 利用1996年和2004年两期北京市土地利用图、北京市DEM数据、北京市各区的总人口、城镇人口、人口密度、GDP、第二产业总值、第三产业总值和固定资产投资等数据,运用ARCGIS分析了北京山区1996-2004年间的耕地变化时空特征。结果发现：1996-2004年间，北京山区的耕地面积减少了1/3，减少的耕地主要流向林地和建设用地。同时还利用SPSS中的Logistic回归模型，在GIS平台上以1996年的数据为基础，模拟了北京山区2004年的耕地空间分布,并与实际状况进行了对比。模拟结果较为理想，其中模拟耕地的准确率达到70%，非耕地达到79%，总体准确率达到78%。

关键词: 耕地变化, GIS, 空间Logistic回归, 北京山区