The raw data selects the L1B-class Earth Observation Full Resolution (EFR) product of the Ocean and Land Colour Instrument (OLCI) sensor carried by the European Space Agency (ESA)/ EUMETSAT Sentinel-3A/B satellite. The reason is that the Sentinel-3 series satellites are the core satellites of the European Copernicus project. The OLCI sensors they carry are specially designed for land and sea observations and have the advantages of high spectrum, high spatial resolution, and high coverage frequency. Suitable for long-term dynamic monitoring of large-scale snow cover. The OLCI sensor has 21 spectral bands covering the spectral range of 400-1020nm, of which the center wavelength of the 21st band is 1.02 microns. This band is highly sensitive to changes in the scattering characteristics of snow particle size and is the core band for retrieving snow particle size., provides a key spectral basis for quantitative inversion of snow physical parameters. The original spatial resolution of the sensor is about 300 meters and the width is 1270 kilometers. The two satellites, Sentinel-3A and 3B, work together to achieve full coverage observation once a day on a global scale, which can ensure the spatial coverage of the data. and time observation density to meet the monitoring needs of snow parameters in the long-term series from 2016 to 2025, Central Asia and northern China.

Use the iCOR plug-in (which supports visual operations of the SNAP module or batch calls of the icor.bat script) to perform atmospheric correction processing on L1B level data. The reason is that the L1B-level data is the original radiation calibration data of the sensor, which includes various disturbances such as atmospheric scattering, absorption, and reflection, and cannot be directly used for surface parameter inversion. Atmospheric correction is the basic prerequisite for quantitative remote sensing, and the radiation received by the sensor can be converted. The value is converted into the true reflectivity of the surface, providing accurate spectral data for subsequent inversion of snow cover parameters. The iCOR plug-in was chosen instead of the traditional atmospheric correction model because the iCOR plug-in is specially designed for remote sensing data with medium spatial resolution and large coverage such as Sentinel-3 OLCI. It can effectively process atmospheric models with high heterogeneity and have higher correction accuracy., more adaptable. During the processing process, the five bands Oa13, Oa14, Oa15, Oa19, and Oa20 affected by strong atmospheric absorption are removed, and only 16 reliable surface reflectance bands are output.(Oa01-Oa12, Oa16-Oa18, Oa21) and latitude and longitude information, the reason is that the above five bands are in the strong absorption band of atmospheric water vapor and carbon dioxide, are seriously interfered by the atmosphere, and the reliability of spectral data is low. After removal, it can improve the accuracy of subsequent data processing. At the same time, the 16 bands retained cover the core spectral range of snow monitoring, which can meet the needs of snow classification and parameter inversion. The SNAP IdePix OLCI processor is used to identify clouds, cloud shadows and snow elements in the image, and only five types of snow-related coded pixels of 1216, 1232, 1248, 1264, and 3264 are retained. Clouds and cloud shadows in remote sensing images will be confused with snow cover and become the main source of interference in snow cover identification and parameter inversion. Clouds, cloud shadows and other non-snow cover pixels must be eliminated through professional classification algorithms to accurately extract snow cover pixels. SNAP IdePix OLCI processor is a dedicated pixel classification algorithm developed for OLCI sensors. It can accurately distinguish the types of underlying surfaces such as clouds, cloud shadows, snow, and land based on spectral features and texture features, and the classification accuracy is higher than that of the general classification algorithm; The five types of codes such as 1216 and 1232 are exclusive codes for snow elements in this algorithm. Only retaining such pixels can effectively eliminate non-snow interference, ensure that subsequent processing is only targeted at snow elements, and improve the inversion of snow parameters. Targeted and accurate. Parameter inversion: Based on Kokhanovsky's physical model of snow reflectance (spherical albedo, geometric optical model, external mixed model), using key bands such as 400 nm, 560 nm, 865 nm, and 1020 nm to retrieve snow particle sizes (considering shape parameters 3.62, 4.53, and 5.8) and pollutant amounts. This model is a classic quantitative model developed for the reflectivity characteristics of snow cover. It can comprehensively consider the scattering and absorption characteristics of snow cover and the mixing effect of pollutants. Compared with the empirical model, the physical model has stronger mechanism and higher inversion accuracy. It is suitable for parameter inversion in different regions and different snow cover conditions; Integrating the spherical albedo, geometric optical model, and external mixing model can respectively simulate the spherical scattering characteristics, geometric optical scattering characteristics, and external mixing state of ice and pollutants, which is more consistent with the actual physical and optical characteristics of snow cover. In the process of calculating the particle size, it is also necessary to calculate the imaginary part of the negative refractive index of ice, which can be interpolated here. At the same time, it is necessary to calculate the equivalent azimuth angle, which is directly calculated by using the observation azimuth angle using the sensor and the azimuth angle of the sun. If their absolute values are greater than 180°, subtract their absolute values by 360°. Rasterization and interpolation: Kriging space interpolation is performed on discrete snow pixel point cloud data to generate iso-latitude and longitude projection raster images with a resolution of 200 meters. Point cloud data is discrete point features, which cannot intuitively reflect the spatial distribution characteristics of snow cover parameters, and it is difficult to carry out spatial analysis and visual expression on a regional scale. The discrete point data is converted into continuous area grid data through rasterization and interpolation, which can realize spatial continuous expression of snow cover parameters and meet the needs of regional snow cover spatio-temporal distribution analysis. Set the grid resolution to 200 meters to improve the spatial fineness of the data; use equal latitude and longitude projection, which can maintain the uniqueness and consistency of geographical coordinates, facilitate spatial splicing, overlay analysis and integration with other geographical data. Monthly synthesis: Only the maximum synthesis method is used for NDSI parameters to generate cloud-free grid products for about 4 scenes per month. The snow particle size and pollutant parameters are retained for daily single-scene products and are not synthesized. Finally, all data is output as ENVI .img grid file. The reason for synthesizing the monthly maximum value of NDSI is that NDSI is mainly used for snow cover monitoring and is greatly interfered by clouds and cloud shadows. There are a large number of missing areas of cloud coverage in daily images. The maximum value synthesis method can select the maximum NDSI value of the same pixel in the month, effectively eliminates the interference of clouds and cloud shadows, improves the cloud-free coverage and spatial continuity of data, and meets the needs of macro monitoring of snow cover; Snow particle size and pollutant parameters are sensitive to short-term changes in snow cover, and the impact of cloud cover on the retrieval results can be effectively eliminated through previous snow cover classification. Therefore, daily single-scene products are retained to ensure the temporal precision of the data and meet the needs of short-term dynamic changes in snow cover. Output the data into ENVI .img format, which is a standard raster data format in the field of remote sensing. It supports direct reading and processing by mainstream remote sensing and geographical information software such as ENVI, ArcGIS, and QGIS. No additional format conversion is required and data usage is reduced. Threshold, at the same time, this format can completely retain metadata such as spatial reference and band information of raster data to ensure data integrity.

Atmospheric correction quality: iCOR has performed well in Sentinel-3 OLCI land applications. The corrected surface reflectance data can truly reflect the actual spectral characteristics of the surface, providing accurate basic data for subsequent snow cover parameter inversion; the difference between blue light and near-infrared individual bands is slightly high, mainly related to the "excess of brightness" phenomenon of OLCI sensor. This phenomenon is caused by the sensor's own system characteristics and belongs to the controllable error range and has little impact on the overall snow cover parameter inversion. Inversion uncertainty: Affected by observation geometric angle, atmospheric residual error, cloud pollution and terrain. Composite monthly NDSI maximum values can significantly improve data continuity and cloud-free coverage. Overall applicability: The data quality is suitable for quantitative remote sensing analysis of snow cover areas in mid-to-high latitudes, but it is recommended to conduct further verification in conjunction with specific application scenarios when using it.