Snow Recognition In Mountain Areas Based On SAR And Optical Remote Sensing Data

This study 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" (Grant No.:95-Y40B02-9001-13/15-04) and National Natural Science Foundation of China "Retrieval of snow water equivalent based on SAR and high spatial resolution optical remote sensing" (Grant No.:41271353). According to the research requirement of the two project, a typical mountain area in Manasi River Basin of Tianshan Mountain was selected as the study area. Based on multiple field investigations and ground synchronization observations, RADARSAT-2 SAR data and GF-1 optical remote sensing data in snow-free, snow-accumulation and snow-melt period were used to analysis the characterization of remote sensing images for snow cover information and investigate the method of identifying snow cover surface type and snow wetness state. And then using the complementarity of the two types of sensors to study the theory and method for snow recognition in mountain areas.Snow cover is an important component of the cryosphere, and also one of the most active natural elements of the land surface. Seasonal snow cover is the main sources of fresh water resources in arid and semi-arid regions of Northwest China. Moreover, snow disaster is one of the main meteorological disasters. Research on snow cover extraction, especially identifying the snow surface type and snow wetness state simultaneously, is of great significance to the snowmelt process monitoring, local climate research, snow disaster assessment, and water resources management, etc. Remote sensing can be used to monitor snow cover in large-scale and high-precision, especially for the poor weather conditions and rarely meteorological observation mountain areas, remote sensing is the only effective means of obtaining snow cover information. Optical remote sensing has the advantage of high precision on snow cover extraction and the ability to identify snow surface type, but it is difficult to differentiate snow from cloud. In contrast, synthetic aperture radar (SAR) can be used to discriminate snow with other surfaces regardless of solar illumination conditions because of its high penetrability, but the backscatter signal of SAR is severely affected by topography of mountain areas. Therefore, SAR and optical remote sensing data has good complementary in recognition snow cover in the mountain areas.This study is based on the ideological line that "Field investigations and ground synchronization observations; Characterization of remote sensing images for snow cover information analysis; Snow surface type and snow wetness state recognition; Construction of snow cover recognition model based on SAR and optical remote sensing data". A model for snow cover recognition in mountain areas using combined SAR and optical remote sensing data was build, the technical difficulties of snow cover extraction in rugged mountain terrain and cloud covered areas were overcomed, and the simultaneous recognition on snow surface type and snow wetness state was realized. Main research contents and conclusions of this paper are as follows:(1) Field investigations and ground synchronization observations. According to the transit time of satellite and distribution of snow cover, respectively in December 2013, March 2014, April 2014 and April 2015, field investigations and ground synchronization observation were carried out. Snow reflectance spectra, snow wetness, snow depth, snow density, snow grain size, air temperature were observed and acquired. The analysis of measured reflectance spectra shows that: in visible and near-infrared bands, reflection spectra of the snow is similar with cloud, and new snow, aged snow and polluted snow shows different spectral curves. The analysis of snow wetness shows that:in snow-accumulation period, snow surface wetness is below 3%, in snow-melt period, snow surface wetness is between 0% and 7%, the snow surface wetness present obvious spatial and temporal distribution characteristics. Field investigation and ground synchronous observation provide the data for remote sensing image analysis, and verification for snow cover extraction results.(2) Characterization analysis of remote sensing images for snow cover information. The simultaneous field observation data in snow-accumulation and snow-melt period, combined with the optical data after comprehensive radiometric calibration were used to analysis the reflectance of snow, cloud and other kinds of surface type. Meanwhile by using SAR images acquired in snow-free, snow-accumulation and snow-melt period, the backscattering coefficient, interferometric coherence are analyzed combining with underlying surface type, local incidence angle and polarization state. The analysis for the characterization of optical remote sensing images shows that:in the four bands of WFV sense of GF-1 satellite, average reflection of the new snow is higher than the aged snow. The analysis for the characterization of SAR backscattering coefficient images shows that:There is no backscattering coefficient difference between snow-free areas and snow-covered areas in snow-accumulation period, the backscattering coefficient of snow-covered areas is 5-10 dB smaller than snow-free areas in snow-melt period, while the surface wetness of snow cover changed from 0% to the 3%, the backscattering coefficient of the snow cover decreased obviously in HH, HV, VH, and W polarization. The analysis of characterization for SAR interferometric coherence images shows that:the average coherence value of HH and W polarization is higher than HV and VH polarization, the average coherence value of snow-covered areas was found to be significantly smaller than that for snow-free areas, especially in HH and VV polarizations. Moreover, the average coherence values are clearly correlated to the local incidence angle in HH and VV polarizations, such that the coherence increases from 0 ° to 30 ° and decreases from 30 ° to 90 °. The analysis results provided the foundations for snow surface type recognition by optical remote sensing data and snow wetness state identification by SAR remote sensing data.(3) Snow cover information recognition using remote sensing. Based on the characterization analysis results of optical remote sensing images, using optical remote sensing data after comprehensive radiometric calibration and SVM classification method, to identify the snow surface type. The identification results shows that:it is difficult to differentiate snow from cloud by using optical remote sensing data but the optical data can be used to identify snow surface type, the identification accuracy is related to the precision of comprehensive radiometric calibration. Based on the characterization analysis results of SAR remote sensing images, the single thresholding algorithm was used in the interferometric coherence images to extract snow cover, then using dry and wet snow sample obtained from Nagler algorithm and the optical polarimetric feature combination obtained from polarimetric feature decomposition to build the model for snow wetness state identification. The identification results shows that:snow cover extraction using SAR remote sensing data is unaffected by the cloud and the SAR data can be used to identify snow wetness state, the snow cover extraction accuracy using SAR remote sensing data is severely affected by topography of mountain areas. The snow cover recognition results provided the foundations for the combination of SAR and optical remote sensing data.(4) Construction of snow cover recognition model based on SAR and optical remote sensing data. Using the snow surface type and snow wetness state recognition results to analysis the characterization of SAR and optical remote sensing data. The analysis results shows that:the snow cover extraction results obtained from optical remote sensing data can be used to weaken the effects of topography in snow cover extraction using SAR remote sensing data, the snow cover extraction results obtained from SAR remote sensing data can be used to overcome the problem of snow cover extraction in cloud coverage using optical remote sensing data, snow surface type and snow wetness state can be identified simultaneously by using combined SAR and optical remote sensing data. Based on the characterization analysis of SAR and optical remote sensing sensors, a model for snow cover recognition in mountain areas was build based on SAR and optical remote sensing data, and snow surface type and snow wetness state in snow-accumulation and snow-melt period were identified. The overall accuracy is 75.7% in snow-accumulation period and 93.5% in snow-melt period. The identification results shows that:from snow-accumulation period to snow-melt period, the snow surface type and snow wetness state shows different spatial distribution. The results can be used to monitor the snowmelt process, study the local climate and manage the water resources.This study focus on the problem of snow cover recognition in mountainous areas, by combining SAR and optical remote sensing data, a model for snow cover recognition in mountainous areas based on SAR and optical remote sensing data was build, the model realized the complementary of multi sensor, it has the theoretical and application innovation. At the same time, using the complementary of SAR and optical remote sensing aata, tnis stuay overcome me problem of snow cover extraction in cloud coverage using optical remote sensing data, weakened the effects of topography in snow cover extraction using SAR remote sensing data, and finally realized the simultaneous recognition on snow surface type and snow wetness state, it has significant technical innovation. However, for the restriction of ground synchronization observation condition and the difficulty in obtaining complete synchronization SAR and optical remote sensing data, the accuracy is part affected. With the launch of the domestic GF-3 satellite, it is expected to combine GF-3 SAR satellite and GF-1, GF-2 optical satellite in time encryption, to realize the change process monitoring on snow surface type and snow wetness state in mountain areas.