Snow Recongnition From High Resolution Remote Sensing Image In Manasi River Basin

Snow cover is an important component of the Cryosphere,and one of the most active natural elements of the earth surface.The seasonal snow is an important supply source for regional river in arid and semiarid area of Northwest China,and at the same time a precious fresh water resource which has the great significance for ecological environment and regional harmonious development in arid and semiarid area.Conducting the research on snow recognition has an important significance on snowmelt runoff process monitor in spring and summer,local climate change,river basin water resources management and snow disaster assessment.As a principal mean of earth observation,remote sensing makes it possible on monitoring snow cover in large-scale and high-speed.Especially in nearly a decade,remote sensing makes snow recognition subtly and accurately possible with the rapid increase on spatial resolution and time resolution of satellite.This present paper is funded by National Science and Technology Major Project"Snow and ice monitoring and its evaluation based on high-resolution remote sensing data in central Tianshan mountains in Xinjiang"(95-Y40B02-9001-13/15-04)and National Natural Science Foundation Project of China "Retrieval of snow water equivalent based on SAR and high spatial resolution optical remote sensing"(41271353).According to the needs of the projects on snow recognition in mountain areas,we choose the typical area in Manas River Basin of Tianshan Mountain as the study area,exploring a method on snow recognition in rugged terrain to distinguish snow surface types,based on domestic high resolution remote sensing satellite GF-1 WFV data,the snow simultaneous field observation data and DEM as so on.The main research contents and conclusions include:(1)The influences of rugged terrain mountain on snow recognition.The concrete analysis about the influence of rugged terrain mountain on snow recognition showed that altitude and topographic shadows were the main factors which influenced the snow extraction in mountain areas.The wide difference on altitude would result in different reflectance in high altitude and low altitude area,and then impacting the snow recognition with an appropriate threshold of GFSI;The existence of topographic shadow also would result in a similar spectral characteristics among different land cover types,therefore it was unable to identify snow cover in shadow areas accurately.Meanwhile,in consideration of snow in different slope and aspect should have different reflectance with a forward scattering characteristic obviously,the non-Lambert characteristic of snow surface must be in view.In this present paper,a method combined with anisotropy correction and topographic correction was proposed to covert the snow reflectance in different slope to the nadir direction on the flat surface,to eliminate the influence caused by rugged terrain and atmosphere.The results showed that the land cover types in shadow area could be distinguished,and the snow surface reflectance tended to be uniform whether in shadow or non-shadow area after correction.(2)Snow information of remote sensing.According the simultaneous field spectral measurement constant with GF-1 WFV data,the spectral reflection curve of different types of snow was achieved.The results of simultaneous field spectral measurement showed that:the spectral reflection of new snow and old snow had an obvious difference,which was the new snow reflectance was the above 0.9 and the old snow was between 0.5?0.8 in visible band,and the reflectance of new snow and old snow began to reduce when entering the near infrared wave band.Meanwhile the response characteristics of the new snow and old snow on GF-1 WFV1 data showed that:it was an similar characteristics between the response characteristics and measured spectral of new snow and old snow,that illustrated the GF-1 WFV sensor can reflect the old snow reflection characteristics in visible and near infrared band accurately.The analysis about snow measured spectral and response characteristics will provide a theoretical support on GFSI snow index based on GF-1 WFV data.(3)The establishment of GFSI snow index.Making the NDSII as a prototype,the fundamental form of GFSI was established;The comparative analysis between snow and non-snow on inter-class separability in each band showed that:snow and non-snow had a largest inter-class separability in blue band,and at the same time the GFSI combined with blue and near-infrared band had a largest inter-class separability as well.Therefore the blue band of GF-1 WFV was the best remote sensing band for snow reconginition,and the determine form of GFSI was established by the blue and near-infrared band,to identify snow and non-snow types.(4)Snow reconginition based on GFSI.The snow recongnition threshold was confirmed as-0.01 by bimodal threshold segmentation method,and the snow surface types recongnition threshold was confirmed as 0.02 by threshold selection experiments.The snow recongnition with GFSI showed that:the snow in shadow areas could be identified accurately through the atmosphere and topographic correction,and the snow results in non-shadow area was improved;The overall accuracy of snow in the study area was 95%;There was a certain spatial distribution difference of snow surface with the changes of altitude,aspect and slope.The final snow recognition results could provide data support to snownelt runoff process simulation and river basin water resources management in study area.Aiming at the domestic high resolution remote sensing satellite GF-1 WFV sensor,a snow index named GFSI was established to identify snow quickly and accurately.The study proved that the GFSI combined with blue band and near-infrared band would reflect the reflection characteristics of snow factually,and distinguished snow and other non-snow land cover types.It had a certain theoretical innovation.