Land Cover Classification Of Northeast Asia Using MODIS/NDVI Time-series Datasets

Land cover refers to a variety of biological or physical cover types on the earth's land surface, including surface vegetation (natural or artificial), construction land (buildings, roads, etc.), lakes, glaciers, bare rock and sand, etc. The natural properties on the earth's surface are the focus object to describe. Land cover and land change is an important factor in the process of global environmental change, while the land cover/vegetation types distribution play a very important role in the Earth's ecosystem processes for the material and energy exchange, but also important content of global change and the carbon cycle simulation, climate modeling study. Since the 1970s, a series of important achievements have been gained by utilizing Remote Sensing to research macro land cover. Land cover classification using the multi-temporal and multi-spectral remote sensing observation data has been an effective technique. In this paper, using MODIS data as the main data source, based on researching of the MODIS/NDVI time-series datasets and multi-spectral reflectance datasets, we extract the land cover types of the Northeast Asian region after accessing to the characteristics of land cover classification. The main research contents included as the following:1. Based on MODIS/NDVI time-series product (MOD13Q1), MODIS multi-spectral reflectance product (MOD09A1) and DEM data, land covers types over the study area were mapped. In the process of classification, selecting characteristic bands according to differences among different land cover types and determining threshold for each land cover type is the key to construct decision tree.2. Being integrated with the concrete conditions of Northeast Asia, the land cover classification system in this paper was established on the basis of IGBP and GLC2000 classification system, and by which, the terrestrial ecosystem model and ecological process study as well as the actual situation of land cover in Northeast Asia can be better proceeded and reflected, respectively.3. Through the combination of the unsupervised classification and decision tree classification methods, the land cover classification system of study area was divided into nine first-degree types and twenty-one second-degree types. Its overall accuracy was 71.29%, Kappa was 0.5989.4. Compared with MOD 12Q1₂001 and GLC₂000, NEA₂007 was more accurate and detailed from the visual interpretation and quantitative view. The datasets, the classification methods, and the classification results have practical application value.