One of the main features of the Asian Water Tower imbalance is the massive melting of snow, necessitating enhanced snow monitoring. However, the sensors aboard polar-orbiting satellites, such as MODIS, yield only one to two valid observations daily. This, coupled with the extensive cloud cover and prolonged duration over the Asian Water Tower, results in a significant number of data gaps. China's new generation of geostationary satellites FY-4A and FY-4B have high frequency observations, making it possible to monitor snow with high precision. In this study, we systematically analyze the image pixel size stretching of FY-4A and FY-4B in the Asian Water Tower region, based on their imaging geometry. This analysis offers a theoretical foundation for fractional snow cover retrieval in the subsequent integration of these two satellites. Concurrently, this study conducts fractional snow cover (FSC) retrieval for FY-4A and FY-4B, utilizing the multiple endmember spectral mixture analysis algorithm with automatic endmember extraction (MESMA-AGE). High spatial resolution Landsat-8 imagery serves as reference data for accuracy assessment. The results indicated that FY-4A's retrieval accuracy remained unaffected by pixel size stretching, achieving an Overall Accuracy (OA) of up to 0.97 and a Root Mean Square Error (RMSE) of less than 0.13. For FY-4B, the retrieval accuracy demonstrated higher snow identification precision, with an OA of up to 0.95. However, the RMSE varied significantly due to pixel size stretching, ranging from 0.12 to 0.21. FY-4A and FY-4B fusion enables high precision and near-real-time snow monitoring.

Accurate simulation of the microwave signatures of snow using the emission models is of guiding significance to develop the snow parameters retrieval algorithm. This study based on reanalysis dataset ERA5-Land and auxiliary data to investigate the potential of the DMRT model combined with the τ –ω model for simulating passive microwave brightness temperature (TB) of snow cover at 10.65 GHz, 18.7 GHz, and 36.5 GHz. The results showed that the correlation coefficient (R) and bias between the simulations and the ground-based microwave radiometer observations at the Altay is 0.45~0.66, 8.21 k~13.3K at V polarization, and 0.44~0.63, 9.32K~14.68K at H polarization, respectively. In addition, the R and bias between the simulations and the AMSR2 and FY-3D TB is 0.61-0.81, 0.58~0.73, and 18.2K~20.75K, 18.25K~19.2K at V polarization, and 0.47~0.65, 0.51~0.69, and 19.79K~28.62K, 20.52K~26.8K at H polarization, respectively. In some forested areas, there is a significant increase in the simulation bias at 36 GHz, which could be attributed to an overestimation of vegetation influence at this frequency.

The detailed satellite mapping of seasonal snow cover, with many spectral bands from High-resolution remote sensors, is largely investigated recently. Snow cover maps can be extracted from optical data using relatively simple approaches given the difference in spectral reflectance. However, some high-resolution images of illuminated snow-covered surfaces suffered from limited detector saturation due to overexposure. Also, these detailed images involve large data volumes that prohibit complex analysis. This study developed and automated four algorithms, including Random Forest Model (RF), Maximum Likelihood Classifier (MLC), the Threshold method based on Water-resistant Snow Index (WSI), and the Blue Snow Threshold method (BST) respectively, for discriminating snow from other surface types with Gaofen-1 optical satellite images. To reduce the impact of overexposure on the extraction of snow-cover areas, a simple cross-calibration methodology has been utilized before estimation. Image pairs from the Operational Land Imager (OLI) on Landsat-8 and Wide Field of View (WFV) on Gaofen-1 were used to verify the radiometric calibration of WFV with respect to the well-calibrated OLI sensor. Different algorithms' performance was tested by utilizing Gaofen-2 Multi-Spectral (PMS) sensor data for validation. RF-derived estimates of Snow Cover Areas (SCA) did better in terms of overall accuracy. Due to simplicity and efficiency, these algorithms have the potential to be used to develop high spatiotemporal resolution maps of SCA.

Cloud detection is the first step in remote sensing surface parameter retrieval. Due to the similar spectral properties of cloud and snow, cloud products commonly used in snow monitoring sensors have a certain degree of cloud and snow misjudgment problem. With the launch of a new generation of geostationary satellite (such as China's FY-4A), its time resolution is 15 minutes, and high-frequency observations make accurate cloud and snow identification possible. This study utilizes the high-frequency and multispectral observation characteristics of FY4A, combining the multi band threshold method with deep learning algorithm, to fully explore the spectral and texture differences of cloud and snow, as well as the characteristics of rapid cloud changes, and achieve high-precision cloud detection. Then, the CALIPSO data is used to evaluate the accuracy of the new algorithm's detection results. From the results, the new algorithm for cloud and snow recognition is more accurate and consistent with CALIPSO observations. In terms of specific accuracy indicators, the cloud hit rate(CHR) increase 1.07% and the false alarm rate(FAR) decrease 5.15%. At the same time, both in terms of single scene or daily composite results, the proportion of cloud cover has decreased about 20%, and the proportion of snow cover can increase by up to about 15%. This laid a solid foundation for high-precision fractional snow cover retrieval and spatiotemporal reconstruction in the future.

The normalized difference snow index (NDSI) is a fundamental spectral indicator of snow/ice in visible and shortwave-infrared imagery. The complex grain shapes in nature have well-known significant effects on the single-scattering properties (SSPs) and subsequently the bidirectional reflectance of snow. The shape effects on snow NDSI need to be further characterized as NDSI is a nonlinear combination of two reflectance bands. Considering the common snow grain shapes represented by sphere, spheroid, hexagonal plate, and Koch snowflake, we use the ray-tracing approach to simulate the SSPs of ice particles and the discrete ordinate algorithm to solve the bidirectional reflectance function and calculate NDSI of snow. According to simulating results, the angular pattern of snow NDSI is subject to snow grain shape, whereas the shape effects can be significantly weakened by the increasing surface roughness of ice particles. The shape of Koch snowflake causes an NDSI habit different from other three shapes for large snow grain size. Moreover, snow NDSI also has complex responses to aspect ratio (AR) for spheroid and hexagonal prism. The theoretical characterization of the snow NDSI responses to various grain shapes would enrich the knowledge of NDSI variation mechanism in snow-covered area mapping applications.

The spatiotemporal distribution of snow depth (SD) has a significant impact on the energy and water balances of the Earth’s system. However, passive microwave remote sensing widely used for SD estimation has large uncertainties due to the variations in snow physical properties. In this study, we demonstrate a new method to minimize these uncertainties and to increase the accuracy of SD estimation. Our method is based on the synergy between the passive microwave AMSR-2 brightness temperature (TB) and a physical snow process model (SNTHERM) to estimate snow grain size, snow density and first-guess SD as priors. On one hand, we used TB from three frequencies and removed non-representative ground measurements at the stations to improve deep snow estimation. Then, we applied a machine learning (ML) algorithm based on both the AMSR-2 TB and the SNTHERM simulations to retrieve the global SD. The results showed that the root-mean-squared error (RMSE) of the retrieved SD was 12.4 cm at the meteorological stations. Independent validations showed that our method significantly reduced the SD and snow water equivalent (SWE) underestimation in the mountains compared to the current satellite products.

Advanced Very High Resolution Radiometer(AVHRR) has accumulated nearly 40 years of data, making it essential for long time series snow monitoring. However, the majority of AVHRR snow algorithms and products are binary forms, which can have a significant impact on hydrological process simulations and estimates of water and energy cycles. Based on AVHRR data, three fractional snow cover retrieval algorithms are implemented in this study: the multiple endmember spectral mixture analysis algorithm based on automatic endmember extraction (MEAMA-AGE), the Snow Index method and the Snow/no-Snow Two-endmember model algorithm. In order to compare and verify the accuracy of the three fractional snow cover retrieval algorithms, we use the high spatial resolution Landsat-8 image snow cover retrieval results as the “ground truth”. The results show that all three algorithms can effectively retrieve fractional snow cover from AVHRR data, among which the Snow Index method and the Two-endmember model algorithm have higher accuracy, while the MESMA-AGE algorithm has lower precision due to the influence of the endmembers representation.

The spatial extent and duration of soil freeze/thaw (F/T) control water and heat exchange, the energy cycle, and climate change. Global warming causes permafrost thawing, which increases carbon emissions and in turn exacerbates climate change. Passive microwave remote sensing has been proven to be effective in monitoring land surface F/T. However, it was found that the applicability of existing passive microwave remote sensing-retrieved F/T products in large-scale areas (such as the Qinghai-Tibetan Plateau (QTP)) was influenced by some landscape factors, such as the arid climate type and terrain elevation gradient. FY-3 series satellites have accumulated nearly 10 years of passive microwave data, but there is little work based on FY-3 passive microwave data to see its potential in land surface F/T status monitoring. In this work, we proposed a dynamic method to determine the surface F/T status by combining the edge detection method and discriminant function algorithm from FY-3B X-and Ka-band microwave radiation imager (MWRI) data. Comparing the results against three F/T products based on in situ 5 cm soil temperature, we demonstrate that this algorithm performs best over different validation areas with an overall accuracy of 86.5%. More specifically, the new algorithm improved the accuracy of current F/T products in arid and semiarid regions from 73% to 90%. Additionally, the spatial distribution of frozen days over the QTP of 2018 based on the new algorithm has good consistency with the permafrost map. However, the accuracy is influenced by snowmelt and appears to be overestimated for thaw soil during the day. This algorithm performs well in QTP areas with complex topography and climate types and holds the promise of providing users with highly accurate F/T products on larger and even global scales.

Spatiotemporal continuous fractional snow cover(FSC) dataset is needed as an important input for the study of large-scale hydrological, meteorological and climate research. But optical data often have gap due to cloud cover. The widely used Moderate Resolution Imaging Spectroradiometer (MODIS) snow cover dataset uses an eight-day composite approach to remove cloud effects, but it cannot meet the requirement of monitoring snow cover, a parameter with high temporal and spatial variability. This paper uses multiple endmember spectral mixture analysis (MESMA) algorithm to retrieve FSC from geostationary satellite (FY-4A and Himawari-8) and polar-orbiting satellite (MODIS) data; Then geostationary satellite retrieve results are used to fill the MODIS cloud cover pixels, which can reduce the cloud cover from 50% to 15%; Finally, the daily FSC product is obtained by using Piecewise Cubic Hermite Interpolating Polynomial (PCHIP) algorithm. Landsat-8/OLI data processed by the MESMA algorithm is determined as “ground truth” to validate this product. The result shows that the accuracy of the daily cloud-free FSC product is high with the root mean square error is 0.1-0.15.