| Snow area and snow distribution are the important input parameters of the hydrological and the ecological model, which are also the foundation stone for snow damage forecasting. In the rugged terrain mountain with the harsh climatic conditions and lacking of meteorological data, satellite derived snow cover information is the only source to monitor snow cover changes in large area.12scenes of Landsat-5TM images are used in this paper to study the snow extraction method in the rugged terrain mountain of Manasi River Basin, which is in the northern slope of Tianshan Mountains, Xinjiang. First, snow extraction algorithm SNOMAP is used to get the preliminary snow extraction information. Effect factors of satellite snow cover extraction are then analyzed based on the preliminary snow extraction result. Quantitative analysis of topographical effects on snow extraction is stated in detail in the paper. Considering the total radiation received by the slope, DEM and6S atmospheric radiation transmission model are combined to eliminate the atmospheric and topographic effects. After combined correction, the reflective characteristics of different features are regained. In the meanwhile, snow information is retrieved in the deep shadows. After that, different snow extraction method and thresholds are emplemented according to the different snow underlying surface conditions. The main contents and results of the research are presented as follows:(1) Altitude has influence on the upward and downward atmospheric transmittance of the target point, resulting in different reflectance values of the snow in high and low altitude. Incidence angle on the slope directly determine the direct solar irradiance received by the target point. Deep shadows emerge in the areas receiving little or none solar radiation, where snow are mixed with other land cover features and could not be correctly identified. Topographic effects can’t be eliminated by the ratio of band2and band5because these two bands have different correlation coefficients with incidence angle on the slope, which means NDSI does not have the adaptability to the rugged terrain. Snow cover area will be underestimated when directly using SNOMAP algorithm.(2) Combined radiometric correction based on DEM and6S radiation transmission model eliminates the atmospheric and topographic effects. In the process of combined radiometric correction, two types of terrain shadows (self-shadows and cast-shadows) are taken into account, and several models are compared. Finally, Perez model is selected to compuate diffuse irradiance on the inclined slopes, and Dozier model is selected to compute reflected irradiance by adjacent surface, which are both approciate for the study area. After the combined correction, reflectance characteristics of different features can be distinguished easily and snow information in the deep terrain shadows are retrieved.(3) Vegetation beneath snow will have influence on the precision of snow extraction. In this paper, different land cover types in the study area are categorized into three major underlying surfaces based on the influences they bring to the snow extraction. They are named as non-vegetated area, cropland and grassland, forest. Snow extraction on different underlying surfaces is using different methods and thresholds. Images after combined correction are used to extract snow information. The precision of snow mapping based on different underlying surfaces is increased compared to using SNOMAP algorithm with the same NDSI threshold. The following research will be considered to use snow and vegetation radiation transmission model to simulate the reflectance characteristics of none-snow-covered forest and snow-covered forest, and to improve the accuracy of snow identification in forest. Some efforts should be made to eliminate the DEM and remotely sensed image registration errors. Ground measured data and high spatial resolution images are desired to better determine the thresholds in the process of snow extraction. |
PDF Full Text Request