中文摘要：

以1995年和2000年Landsat TM影像的土地利用空间数据为基础，计算出土地利用类型之间相互转化的概率信息，分析了亚热带典型区域桃源县各土地利用类型变化的动态、幅度、速度、空间分布特征和方向．应用马尔柯夫概率模型进行土地景观要素的情景模拟，预测出了2050年土地利用类型的变化格局．结果表明：7种土地利用类型的面积均有不同程度的变化，主要是耕地、建设用地和水域的不断扩张，林地的不断减小．变化方向为耕地面积变化主要是毁林草开荒，建设用地扩展则表现为大量侵占耕地，水域变化是由耕地和林地转变而来．毁林草和农村扩展乱占耕地现象还未从根本上得到遏制．

英文摘要：

Land use/land cover change is a complex system, which is affected by many comprehensive factors, including socio-economic and natural and environmental factors. With the rapid development of the modern globe information science, RS and GIS are increasingly becoming the most powerful means to solve such problems as the land use changes and sustainable development decision-making ,which once were very arduous to be disposed traditional research methods. Based on the land-use data of the interpretation of Landsat by the TM in 1995 and 2000, transition matrix of land use type area was computed and land-use changes in Taoyuan County in Subtropical Typical Region were studied, including developmental trend, range, velocity, spatial distribution pattern and direction. Percentages and developmental trend of land use types in 2050 were predicted by Markov model. The results showed that, with the rapid growth of population and expansion of economy, seven kinds of land use types changed at different degree：an increase in farmland, urban and rural resident lands, and water lands; a decrease in woodlands. After 50 years, urban and rural resident lands expand quickly and increase 3.34%. Farmland changes resulted mainly from woodlands and grasslands; Urban and rural resident land increasing was mainly from farmlands; Changes of water-land areas were because of farmland and woodland conversion. It showed that in the process of development of Taoyuan County economy, paddy land and forest land were consumed greatly,in which the most was forest land. The reason was related to increasing population, developing economy and some land management policies. In a word, the phenomena of destroying woodland or grassland and occupying farmland at random were not eliminated thoroughly. Furthermore, using the Markov model as a basic tool to predicate land use/land cover may not well indicate the real reasons and it still needs other methodology, such as Canonical Correlation Analysis, to identify the driving forces of land use/land cover cha