Consistency and Accuracy Assessment of Snow Cover Products from Terra, Aqua, SNPP and JPSS-1 Satellites

The Visible Infrared Imaging Radiometer Suite (VIIRS) sensor onboard Joint Polar Satellite System (JPSS) satellites will replace the Moderate-Resolution Imaging Spectroradiometer (MODIS) to prolong data recording in the future. 
Therefore, it is a fundamental task to analyze the consistency and assess the accuracy of the snow cover products retrieved from the two sensors. 
In this study, snow cover products from MODIS/Terra, MODIS/Aqua, VIIRS/SNPP and VIIRS/JPSS-1, were evaluated in terms of Normalized Difference Snow Index (NDSI) consistency and accuracy assessment using higher resolution images of Landsat and Sentinel-2 snow cover products. Paired comparisons were performed among the four products in five major snow distribution regions over the world: Northeast China (NE), Northwest China (NW), the Qinghai–Tibet Plateau (QT), Northern America (NA), and European Union (EU). The two categories of snow products are utilized: The L3 Daily Tiled products, referenced by their Earth Science Data Type (ESDT) names of VJ110A1, VNP10A1, MOD10A1, MYD10A1, and L3 Daily Cloud-Gap-Filled (CGF) products, VJ110A1F, VNP10A1F, MOD10A1F, MYD10A1F. The important conclusions demonstrated as follows.
(1) During the snow season, the four types of 10A1 snow products demonstrated good consistency, with higher R values and smaller BIAS under clear sky. VIIRS exhibited a higher snow cover percentage than MODIS. By combining the four 10A1snow products, it is effective and feasible to produce cloud-free snow products.
(2) The consistency of the four 10A1F snow products was lower than that of the 10A1 products under clear skies. SNPP showed good consistency with JPSS-1, and the same to TERRA with AQUA.
(3) In the 10A1F products based on the previous day's clear-sky cloud-filling algorithm, VJ1 and VNP products exhibited larger fluctuations compared to MOD and MYD products. Among the 10A1F products, the smaller fluctuations and higher snow cover percentage of MODIS, along with a cloud persistence duration higher than VIIRS, led to an overestimation in MODIS's 10A1F snow products.
(4) The snow-cloud confusion is existing both in products with the same sensor and with different sensors for the 10A1products, and the latter is much larger than the former, the percentage of which is approximately 10% in the five regions.
(5) High-resolution snow product validation indicates that VIIRS has higher accuracy in both snow products than MODIS. 
(6) The newest JPSS-1 snow cover products display good agreement with that of SNPP. The pixels with the flag of ‘no decision’ in VNP10A1, MOD10A1, MYD10A1 are labelled as land, waterbody, and mostly clouds in VJ110A1 product, respectively.               
Above all, in spite of existing sensor differences affecting consistency of snow cover products, the paired comparisons indicated that under clear skies, the four snow products exhibit good consistency, with higher consistency observed in snow products from the same sensor. The evaluations by higher resolution snow products assured the high accuracy. It is effective and feasible to produce cloud-free snow products considering the overestimation of 10A1F products.