基于MODIS的西藏高原土地覆盖分类研究

嵌的MODIS遥感图像,用最大似然估计分类器对西藏高原的土地覆盖类型进行了分类,并采用西藏高原数字高程模型(DEM)对MODIS分类结果进行了分析和改进,最后分别用混淆矩阵和图集中西藏高原植被分类汇总面积对MODIS土地覆盖分类结果进行了分类精度评价。结果表明:①MODIS遥感图像能够有效地分类出西藏高原的主要土地覆盖类型;②根据西藏高原植被的垂直带性分布特点,由DEM改进的MODIS土地覆盖分类精度明显提高,主要土地覆盖类型的面积绝对误差都小于2%,其中,河流与湖泊、森林、高寒荒漠、作物和山地草原的绝对误差都在1%以下;③混淆矩阵分析的平均分类精度为87.68%。

Land Cover Mapping in the Tibet Plateau Using MODIS Imagery

Land cover is a fundamental variable impacting and linking many parts of the human and physical environments and is a critical biophysical parameter determining the functioning of terrestrial ecosystems in biogeochemical cycling, hydrological processes, and the interaction between the surface and the atmosphere. Of various sources of land cover data, satellite remote sensing is particularly attractive. The first MODIS receiving station in the Tibetan Plateau was established in Tibet Institute of Plateau Atmospheric and Environmental Sciences in year 2002. Taking the China vegetation classification map with a scale of 1:1,000,000 as ground truth, land cover classification in the Tibetan Plateau was performed using the maximum likelihood classification algorithm based on MODIS band 6-2-1 campsite acquired on September 13 and 14, 2004. The results of the land cover classification were improved by integrating a digital elevation model (DEM) into the classification procedure. Two methods were used to carry out accuracy assessment. One was estimation from training data using confusion matrixes. The other was comparison between the total areas of land cover classification from DEM and from the digitized vegetation classification map of the Tibetan Plateau. Twelve land cover types in the Tibetan Plateau were classified. The representation of the derived land cover types was improved obviously using DEM. The absolute classification errors for alpine sparse vegetation, alpine frigid steppe, shrub and meadows, and bare land was found to be around 4%-8%, and for the rest of land cover types smaller than 2%, in which, the errors for lakes and rivers, forest, alpine frigid deserts, crop land, and mountain steppe were generally less than 1% . The overall accuracy of the land cover classification for the Tibet Plateau in terms of the confusion matrix was approximately 87.68%. Classification results showed that the alpine frigid steppe and alpine sparse vegetation are primary land cover types in Tibet, accounting for 32.32% and 20.57% of the total area, respectively. The former is overwhelmingly distributed in the northern and northwestern Tibetan plateau and some high mountain areas in the central Tibet in a high altitude and frigid climate. The latter is primarily distributed in certain high mountains above 5000 m. Affected by complexity of terrain, climate, and other factors in the Tibetan plateau, making accurate, comprehensive, and geospatially consistent land cover maps remains a challenge. With increasing availability of improved spatial, spectral, geometric, and radiometric resolution satellite imageries, detailed ground-truth data, and improved classification algorithms, it is possible to produce more accurate, comprehensive, and geospatially consistent land cover maps for the Tibet Plateau in the future.