基于二元逻辑斯蒂模型的桐梓河流域土地利用变化及模拟

为探明典型喀斯特流域土地利用时空变化过程,揭示流域土地利用变化主控因素,本文以桐梓河流域1980~2015年五期的遥感影像解译数据以及研究区的自然社会经济等数据为基础,综合运用土地利用转移概率矩阵以及二元逻辑斯蒂回归模型对流域的土地利用格局进行了最佳模拟尺度选择,在此基础上构建了流域不同土地利用类型空间分布概率图。研究结果表明：桐梓河流域的景观类型异质性较强,土地利用类型在研究期间内转移过程比较复杂,各种土地类型流转中“转出与转入”现象普遍存在;桐梓河流域土地利用分析及其模拟的最佳空间模拟尺度为120m×120m;喀斯特流域林地与灌木林地的发生概率与降水量的大小并没有直接的关系,降水量越大的地方流域的林地分布概率也不一定越高,其发生概率一般与气温呈现出正相关关系,温度越高,林地和灌木林地的出现概率越大;喀斯特流域草地的分布与岩性有着极大的关系,草地发生概率一般出现在喀斯特地区以及降水量高的地区,而气温越高的地区草地出现的概率越低;从逐点对比的精度来看,建设用地的模拟准确率最高(89. 62%),水体次之(86. 15%),而草地的模拟准确率最低,其值仅为62.68%。由此可见在对喀斯特流域土地利用时空数据变化及其驱动力识别时,二元逻辑斯蒂回归模型具有很好的模拟能力。

Land Use Change and Its Simulation in Tongzi River Basin Based on Binary Logistic

In order to explore the temporal and spatial change process of land use in typical karst basins and reveal the main controlling factors of land use change in the basins, this paper combines the probability matrix of land use transfer and the binary logistic logistic regression model to select the optimal simulation scale for the land use pattern based on the remote sensing image interpretation data from 1980 to 2015 and the natural and socio-economic data in the Tongzi river basin, and then a probability map of spatial distribution of different land types in the basin is constructed. The results show that the heterogeneity of landscape type in the basin is strong, and the transfer process of the land use type is complex during the study period. The phenomenon of "turning out and transferring" is common in the study area. The best spatial simulation scale for land use analysis and land use simulation is 120m * 120m in Tongzi river basin. The probability of the occurrence of woodland and shrub land in Karst is not directly related to amount of precipitation, and the probability of distribution of woodland and shrub land in watershed with higher precipitation is not necessarily higher. The occurrence probability of woodland and shrub land is generally positively correlated with temperature. The higher the temperature, the greater the probability of occurrence of woodland and shrub land. There is a great relationship between the distribution of grassland and the lithology in Karst. The grassland usually appears in the Karst region and the region with high precipitation, and the higher the temperature. From the raster of point by point, the accuracy of the construction land is the highest (89.62%) and the water body is in second (86.15%), while the simulation accuracy of the grassland is the lowest, and its value is only 62.68%. It is concluded that the binary logistic regression model has a good simulation ability when identifying the spatial and temporal data changes and driving forces of land use in the karst basin.