Study On Multiscale Estimation Method Of Fraction Vegetation Coverage In Hulunbeier Grassland

Grassland is one of the important land resources on earth,and posses important ecological,economic and social values.The vegetation coverage is the main indicator to monitor the quality and function of grassland,which plays an important role in regional ecological environment monitoring,ecological service function assessment and ecosystem protection research.Hulunbeier grassland is an important part of China’s northern ecological security barrier,and composed of a large number of high-quality natural pastures.To accurately continuous monitoring on vegetation coverage changes in grazing land is an important element of grassland ecological protection,so the comparative research on multi-scale extraction methods of vegetation coverage in Hulunbeier grassland have significances in theoretical research and practical application.This study takes the experimental area of beef cattle grazing gradient in Hulunbeier grassland livestock demonstration area as an example,the field photos and three kinds of satellite remote sensing imagines are used for vegetation coverage extraction,and compare the differences of vegetation coverage at different scales by use of statistical analysis,regression analysis,and least squares methods.Finally,this study establish the optimal prediction model of vegetation coverage to reveal the spatial pattern of vegetation coverage in the experimental sample area of Hulunbeier grazing.The main conclusions are as follows:1)According to the distribution of sample areas in the study area,digital photos were actually taken to obtain 617 sample squares(1*1m).The average vegetation coverage of the study area was 0.49 and 0.35 by using of two extraction methods of EXG and HSV,respectively.If the methods of EXG and HSV were combined,the average vegetation coverage of the extracted study area reached 0.72,which is closer to the actual vegetation coverage of the grazing land.2)At the confidence level of 0.5%-99.5%,the vegetation coverage extracted from five vegetation indices based on Landsat-9 images(30*30m)ranged from 0.59-0.67 while the vegetation coverages from Sentinel images(2*2m)and GF-1 images(16*16m)ranged from 0.63-0.73 and0.56-0.89.While at 5%-95% confidence level,the vegetation coverages extracted from the three data sources ranged from 0.51-0.61,0.61-0.66 and0.52-0.58,so the vegetation coverage extracted based on VARI,NDVI and EXG indices at 0.5%-99.5% confidence level was closest to the mean vegetation coverage of grazing land(0.72).3)At the 0.5%～99.5% confidence interval,the vegetation coverages from Landsat-9 satellite VARI index(RMSE=0.107)and vegetation coverage extracted based on GF-1 image Ex G index(RMSE=0.122)were better,while the vegetation coverage extracted from the RDVI index of Sentinel image had the highest fitting effect(RMSE= 0.059),so the empirical equation of grassland coverage of grazing land is proposed as FVC = 0.243 + 0.656 × RDVI,and the change of RDVI index of Sentinel image can be applied to monitor on vegetation coverage change in Hulunbeier grassland.