Study On The Methods Of Snow Information Extraction Under Forest Cover Based On Multi-source Remote Sensing Data

Snow is the most widely distributed and obvious factors of seasonal change in the cryosphere. Accurate information of snow is an important parameter of snowmelt runoff model, and it is the primary foundation and guarantee to analyze the snow!s effect on the climate and snow-related hazard. Having the advantage of temporal, multi-angle, covering a wide range, remote sensing is regarded as an ideal means of extracting information snow. Seasonal snow covers 8% of the global land area in the northern hemisphere, but since the effect of canopy shadowing on snow, it makes the information of snow in the forest coverage area difficult to be extracted.This paper chooses the OLI and MODIS data as the main remote sense information source, as well as a large area of forest distribution of Manas River basin as the typical research area to carry out the method study of increasing the snow information extraction accuracy in forest coverage area.Main contents:1. Carrying out the spectral characteristics of snow in visible and near infrared bands, and analysing the spectral characteristics of typical feature in order to determine the best band to extract the snow information.2. Through the optimal selection of threshold value and combined with NDVI data, I build the methods of segmented NDSI snow index and S3 snow index, and compare it with methods of traditional NDSI and S3 in accuracy comparison and analysis. In addition, according to the difference of snow in forest coverage and non-forested area in shape and texture features aspects, I extract the snow information using object-oriented method.3. According to the requirement of the time resolution and the size of the study area and combining with the NDVI data, I construct the segment inversion model based on sub-pixel scale, and I do the precision verification at the same time.The main conclusions:1. Through the analysis of snow spectral characteristics, I find that due to the effect of forest canopy on snow, the snow in forest coverage shows partial spectral characteristics of vegetation. Therefore, traditional method of NDSI and S3 is not a good way to extract the snow-covered forest area. Comparison with NDSI, S3 is better due to the addition of the reflection characteristics of the vegetation reflecting the red band.2. Combined with NDVI segment index method, I!m able to extract the snow information from forest coverage area,and it improves the accuracy of extracting the snow information from there. So the method is suitable for extracting high-precision information in basin scale.3. When there is no satellite Landsat 8 transit, in order to meet the requirements of temporal resolution, MODIS data can be used. To improve the spatial resolution of MODIS data, and at the same time to increase the extraction accuracy in forest coverage area, this paper constructs the segment inversion model based on sub-pixel scale. This model ensure the accuracy of snow extraction not only in non-forest area, but also in forest cover area.