Vegetation Coverage Change Analysis Research Based On Remote Sensing Data In Loess Plateau

As a sensitive indicator of earth-atmosphere system, vegetation has a vital impact onclimate and human environment. The dynamic long-time alteration of vegetation can reflectthe trend of transformable climate. Simultaneously, dynamic monitoring of vegetationcoverage is becoming a hot issue and important content in the global change research.The distribution of vegetation type in the Loess plateau has zonal distributioncharacteristics from southeast to northwest. As an important component of continentalecosystem of the Loess Plateau, the vegetation type plays an irreplaceable role in maintainingthe balance between the energy, material circulation and the ecological environmentalimprovement in the local ecological system. Therefore, how to extract the vegetationinformation accurately based on remote sensing data, and explore kinds of vegetation typeswith their variability has an important academic and practical significance in understandingthe present situation and programming the future construction schema of the vegetation in thefield of land use/cover and management.In this paper, the Chinese Loess Plateau is taken as the research area, the Landsat TMmulti-spectral image and the dataset of standard normalized difference vegetation index(NDVI) are adopted to discuss the distribution and variability of the vegetation type in theLoess Plateau from the year of1987to2006. The main research contents and results are asfollows:(1) According to the problems about the limitation of classification andinsufficient-precision using traditional classification methods, which only depend upon thespectrum characteristics of RS images, a comprehensive model of support vector machine(SVM) integrated texture features with NDVI is developed aiming to improve the overallclassification accuracy about the vegetation in the studied area. Vegetation Atlas of China isreferenced, based on the main vegetation characteristics, the vegetation coverage types of theLoess Plateau is divided into8categories, such as the coniferous forest, broad-leaved forest,brushwood, mixed coniferous and broad-leaved forest, desert, grassland, meadow andcultural vegetation. The results show that: the total classification accuracy and Kappacoefficient reach82%and0.7981separately. The classification result can resolve theproblem of target features with similar spectral features effectively and has a higher accuracy. It proves that the new classification model developed in this paper is suitable and rational,which can be safely adopted in the vegetation classification in the Chinese Loess Plateau.Simultaneously, it is scientific and rational for the8-categories method in the vegetationcoverage classification of the Loess Plateau.(2) Spatio-temporal change of vegetation cover in the Loess Plateau is analyzed basedon AVHRR/NDVI dataset from the year of1987to2006. The result shows that: thevegetation coverage increases slightly from the past20years in general. Whilst there areobvious differences in the spatial distribution; vegetation coverage shows a strong seasonalfeature within a year. Vegetation activities increase in spring and autumn, while it is notobvious in winter and summer. Vegetation NDVI value increase from Jan. to Aug. anddecrease from Aug. to next year. The NDVI value of vegetation cover in Aug. should be thebest representative.(3) Based on the dataset of AVHRR/NDVI (8km×8km) combined with meteorologicaldata, the response of different vegetation types on meteorological factors is conducted. Theresult shows that: the NDVI value is very sensitive to precipitation in the annual growingseason. The average NDVI value has a strong positive correlation relationship with differentprecipitation, and the related coefficient is0.01which is belonging to a significant level. Atthe same time, the average NDVI value has a negative correlation relationship with thetemperature, and the relevance is rather smaller than the precipitation does. NDVI value ofeach month in the growing season has significant correlation with precipitation, andsignificant negative correlation with temperature. NDVI value does not exist significantlydelay and cumulative effect.