土地利用变化模型是理解土地变化科学研究的重要组成部分,由Markov转化过程组成的随机模型因其能用少量的数据而实现对土地利用格局的模拟,是土地利用变化模型研究中较合适的一个选择。本文构建了基于GIS技术的随机变化模型,以自然条件约束较大、受人为因素影响较小的贵州省石阡县为例,应用1988年、2001年两期土地利用数据,结合其他辅助数据模拟其土地利用变化格局。文章采用选取样点方式,计算包含土地利用变化在不同影响因子(如海拔、坡度、距道路距离、距河流距离等)的土地利用转移矩阵。再运用该矩阵,以1988年土地利用为基础,模拟2001年的土地利用变化,并与2001年实际土地利用比较,模拟结果较为理想,证明了该模型的可行性。最后模拟了在未来3种不同约束条件下(现状发展、经济优先和生态优先发展模型)的土地利用变化情景,该研究结果能更好的服务于地方土地利用规划编制和政策分析,也为摆脱土地利用变化复杂系统,仅利用少量数据就进行土地利用空间格局模拟研究提供了一个新的方法。
Land use change modeling is an important technique for the projection of alternative future pathways, and for conducting experiments that test our understanding of key processes in land use change. The stochastic model is a powerful tool to quantify and model land use change in a simple way with limited data requirements. It is based on Markov processes and transition probabilities, and describes a stochastic process that moves in a sequence of steps through a set of states. For land use change, the states of the system are defined as the amount of land covered by various land uses. The transition probabilities can be statistically estimated from a sample of transitions occurring during a particular time interval. This study integrates GIS and stochastic modeling to simulate land use change in Shiqian County, located in a karst area in southwest China, during the period 1988- 2014. In the GIS-based stochastic model, Markov process and transition probabilities are used to simulate land use change. This approach treats state transitions as stochastic processes that are conditional only on the initial state. The factor of identical land use within a 3 by 3 grid-cell neighborhood receives special consideration. Neighborhood interactions between land use types are often included in spatially explicit analysis of land use change and they play an important role. The application consists of two parts. First, transition probabilities of land use change with biophysical factors such as elevation, solope, distance to stream, distance to road and adjacent cell with the same land use are calculated and integrated into a stochastic model, together with the 1988 land use map interpreted from Landsat TM imagery. Then, using the stochastic model in ARC-GIS environment, the 2001 land use pattern is simulated based on the 1988 land use pattern. The model is verified by a statistical cell-by-cell comparison with the actual 2001 land use pattern interpreted from Landsat ETM imagery. Second, three hypothetical and plausible land use