| In the desert of arid and semi-arid areas,biological crusts are widely distributed,and the coverage of biological crusts in some areas is as high as more than 70%.Due to the physical,chemical and biological properties of the soil on its surface,the wind erosion resistance of biological crust is significantly stronger than that of loose sandy soil.At the same time,it can also increase the soil nutrient content and improve the regional ecological environment.It is a biological protective layer on the surface,which also plays an obvious role in blocking water infiltration and water conservation.Therefore,biological crust is an important basis for vegetation succession in arid,semi-arid and desert areas.The spatial distribution of biological crusts in arid and semi-arid regions can be quickly and accurately grasped by remote sensing classification and identification.In this paper,the biological crusts in Mu Us Sandy Land were identified and extracted on multiple scales based on multiple platforms and sensors by using pixel unmixing model and spectral index methods.Monitoring and tracking the pattern and dynamic changes of biological crusts at the regional scale through remote sensing is the fundamental way to evaluate the role of biological crusts in the ecosystem of arid areas and effectively regulate and scientifically manage such resources,the basic way of these resources to areas of biological crust resources protection and rational utilization to provide basic data management support and management decision basis.The following research results were obtained:(1)Terrain Classification: Through precision comparison,the best method suitable for ground object classification of different platforms and sensors is obtained,among them,the support vector machine(SVM)classification method has the highest accuracy in distinguishing ground features of UAV remote sensing images,and the random forest(RF)classification method has the fifth highest accuracy in distinguishing Landsat8 OLI multispectral images.(2)Pixel Unmixing: For the sample plot scale,the spectral angle pixel unmixing method can better separate two kinds of biological crusts,and the average coverage of algae crusts and moss crusts in the five sample plots is 23.35% and 22.35%.For the regional scale,the spectral angle pixel unmixing method is better than the adaptive analysis pixel unmixing method.It is obtained that the algal and bryophyte skin coverage are 22.6% and 21.3% respectively in Mu Us Sandy Land.(3)Spectral index: For UAV multispectral images,the triangular vegetation index(TVI)spectral index has the highest recognition accuracy,followed by normalized vegetation index(NDVI).The crust index(CI)has the lowest recognition accuracy.For Landsat8 OLI multispectral images,TBSCI can effectively improve the recognition accuracy,and the proportion of biological crust pixels in Mu Us Sandy Land is 23.31%.For EO-1 hyperspectral images,GBSCI index can identify regional biological crust-skin well,and the proportion of biological crust pixels is 35.59%.(4)Regional scale: Algal crusts and moss crusts are widely distributed in the entire Mu Us Sandy Land,of which the coverage of algal crusts is 22.6%,the coverage of moss crusts is21.3%.The coverage of algal crust decreases from southeast to northwest.The coverage of moss crust is the most in the middle and the least in the southwest.At the same time,a more suitable multispectral identification index TBSCI and hyperspectral identification index GBSCI are proposed.Through this study,it can be found that the method of pixel unmixing model and spectral index can better identify and extract the biological crust in the region and sample plot.Use the UAV sample observation results as a bridge can solve the problem of matching between the regional scale identification results of satellite remote sensing platform and the field results of sample points to a certain extent,so as to ensure the reliability of regional scale biological crust remote sensing identification results and methods. |
PDF Full Text Request