Grassland is an important vegetation type in China and plays an important role in human subsistence and development. Remote sensing technology is an efficient means of grassland ecological monitoring, but at present we lack of systemic study on spectral characteristics of dominant grassland types in the west of China, which restricts the application research of grassland remote sensing. The key is in-depth and systemic study of ground spectrum of grasslands. So we collected ground spectrum of dominant natural grasslands every 10 days from May to October in 2003 using GER 1500 spectrometer in the region around Qinghai lake, an important pasturing area in the north of China. Using these data the paper systemically studies the change rule of green peak, red edge, and other feature parameters in relation to spectrum waveform, as well as that of commonly used vegetation index within grasslands' growth period. Then the paper elementarily analyzes the ability to classify grassland using spectrum data at different scale, and delivers corresponding methods of classification on the basis of feature extraction and feature transformation experiments.
The paper quantitatively analyzed the relationship between suspension sediment (soil) content and water spectral reflectance with the data tested with FieldSpec FR spectrometer. Then discussed the factors influencing water reflectance. After comparing the results from regressing analysis of reflectance and contents of soil at each wavelength from 350nm to 2500nm, the optimum band (wavelength) was found. The results were proved by another group of data. The conclusion would be helpful in estimating soil content of sea, river or flood with hyperspectral remote sensing, and evaluating soil erosion within water system.