土地利用变化是指由于自然或人为因素引起的土地覆盖和土地利用方式的变化。现有研究多集中于人类活动的影响,而对自然环境驱动的土地利用变化的探讨较少。本研究聚焦青藏高原南部,利用Landsat卫星数据,采用最大似然法(Maximum Likelihood Classification, MLC)对1985~2005年间的遥感影像进行了监督分类。结果显示,模型在耕地、水体和冰川及永久性积雪等特征明显的类别上表现出较高的准确率、精确率、召回率和F1分数,而对于林地和湿地等光谱特性相似或内部异质性较高的类别,分类精度略有下降。总体而言,二十年间模型的分类性能稳定且可靠,显示出随着时间推移逐步优化的能力。研究表明:1) 基于MLC的方法依然在自然环境驱动区域的土地利用变化监测中具有较好的应用前景;2) 通过对1985~2005年间研究区土地利用转移矩阵的分析,揭示了该时期内土地利用类型的显著动态变化,耕地、草地、林地的变化体现了生态恢复政策的有效实施,冰川和永久积雪、湿地的变化则反映了气候变化对自然环境的影响。未来的研究可以通过引入更多样化的样本和改进模型参数来进一步提升分类精度,并探索结合多时相影像和其他遥感源以增强不同类别之间的可分性。Land use change refers to the change of land cover and land use mode caused by natural or human factors. Most existing studies focus on the effects of human activities and less on natural environment-driven land use change. This study focused on the southern Qinghai-Xizang Plateau and used Landsat satellite data and Maximum Likelihood Classification (MLC) to supervise and classify remote sensing images between 1985 and 2005. The results showed that the model showed high accuracy, precision, recall and F1 scores in the categories well characterized by cultivated land, water, glaciers and permanent snow, while the classification accuracy decreased slightly for the categories with similar spectral characteristics like woodlands and wetlands or high internal heterogeneity. Overall, the classification performance of the models was stable and reliable in the past 20 years, showing the ability to gradually optimize over time. The study shows that: 1) The method based on MLC still has a good application prospect in monitoring land use changes in the natural environment-driven area;2) The analysis of land use transfer matrix in the study area between 1985 and 2005 reveals the significant dynamic change of land use type during the period, the change of cultivated land, grassland, forest land reflects the effective implementation of ecological restoration policy, the change of glaciers and permanent snow, wetland reflects the impact of climate change on the natural
暂无评论