Comparison of fractional vegetation cover derived from digital camera and MODIS NDVI in Mongolia

Abstract:

Satellite remote sensing can continuously observe the land surface vegetation with repetitive error over large area, though it requires complex processes to correct errors occurred from atmosphere and topography. On the other hand, the imageries captured by digital camera provide several benefits such as high spatial resolution, simple shooting method, and relatively low-priced instrument. Furthermore, digital camera has less of atmospheric effect such as path radiance than satellite imagery, and have advantage of the shooting with actual land cover. The objective of this study is the comparison of fractional vegetation cover derived from digital camera and MODIS Normalized Difference Vegetation Index (NDVI) in Mongolia. 670 imageries for the above ground including green leaves and soil surface captured by digital camera at the 134 sites in Mongolia from 2011 to 2014 were used to classify the vegetation cover fraction. Thirteen imageries captured by Mongolia and South Korea were selected to determine the best classification method. Various classification methods including the 4 supervised classifications, 2 unsupervised classifications, and histogram methods were used to separate the green vegetation in camera imageries that were converted to two color spaces such as Red-Green-Blue (RGB) and Hue-Intensity-Saturation (HIS). Those results were validated using the manually counted dataset from the local plant experts. The maximum likelihood classification (MLC) with HIS color space among classification methods showed a good agreement with manually counted dataset. The correlation coefficient and the root mean square error were 1.008 and 7.88%, respectively. Our preliminary result indicates that the MLC with HIS color space has a potential to classify the green vegetation in Mongolia.