This directory contains an archive of global daily snow and ice maps produced with the Global Multisensor Automated Snow and Ice (GMASI) algorithm. The dataset includes binary data files and corresponding reduced resolution (browse) images. CHANGES/MODIFICATIONS February 19, 2019: Product Version 3 has been uploaded. An updated normalization of the AVHRR visible and near IR bands response has been used. An additional filter has been applied to reduce the amount of spurious snow identifications in the Southern Hemisphere. FILE NAMING CONVENTION Filenames incorporate the date in the yyyyddd format where yyyy is a four-digit year and ddd is a three digit day of the year. Files produced through reprocessing of historical satellite observations (year 1988 to the beginning of 2018) are identified by "reproc_v003" in the file name. OUTPUT DATA Map data files are flat binary, 1-byte per pixel, 9000 elements per record, 4500 records per file, compressed with linux compress command. Global snow and ice maps are provided on a latitude-longitude (Plate Carree) grid oriented north to south and west to east with the grid cell size of 1/25 of a degree or about 4 km at the equator. The location of the upper left corner of the upper left pixel of the map is 90N, 180W. Snow and ice map byte values are as follows 0: water 1: snow-free land 2: snow-covered land 3: ice over 200: undetermined/unclassified In the browse .png images snow is shown in white and ice is in yellow. Background colors over the land surface represent surface elevation. Information on the surface elevation is not included in the binary snow and ice maps. ALGORITHM Information on the snow and ice cover is derived from combined satellite observations in the visible/infrared and in the microwave spectral bands. Data from AVHRR sensor onboard NOAA and METOP satellites are used as the source of observations in the visible and infrared. Microwave measurements are provided by SSMI and SSMIS sensors onboard DMSP satellites. For snow and ice retrievals the current version of the algorithm utilizes observations from one AVHRR sensor and from all available SSMI and SSMIS sensors. The GMASI retrieval system includes separate processing streams for satellite observations in the visible/infrared and in the microwave spectral bands. At the followng stage information on the snow and ice cover derived from the two groups of sensors is combined. The final step incorporates a recurrent gap-filling procedure which ensures spatial continuity of the derived daily snow and ice maps. The full description of the algorithm is provided in the system ATBD (available at http://www.star.nesdis.noaa.gov/smcd/emb/snow/documents/Global_Auto_Snow-Ice_4km_ATBD.pdf). Information on the accuracy of snow and ice maps and on the general performance of the algorithm since 2011 is provided in Romanov P. (2017) Global Multisensor Automated satellite-based Snow and Ice Mapping System (GMASI) for cryosphere monitoring, Remote Sensing of Environment, 196, 42-55. INPUT SATELLITE DATA To generate global snow and ice maps within the GMASI reprocessing system the following datasets have been used: 1. Observations in the microwave (DMSP SSMI and SSMIS) Fundamental Climate Data Record (FCDR) of microwave brightness temperature observed by SSMI and SSMIS sensors onboard DMSP satellites. The dataset have been produced by Colorado State University. It currently covers the time period from 1987 to the beginning of 2018 and includes Level 1B satellite in swath format. Data files have been obtained from National Centers for Environmental Information (NCEI, former NCDC)at https://www.ncdc.noaa.gov/cdr/fundamental/ssmis-brightness-temperature-csu 2. Observations in the visible/infrared (NOAA and METOP AVHRR) NOAA and METOP AVHRR level 1B data in the swath format have been acquired from the NESDIS STAR Central Data Repository (SCDR). Calibration of the AVHRR optical sensors and corrections for their degradation with time have been performed using the algorithm and data provided in the AVHRR Cloud Properties - PATMOS-x processing package at https://www.ncdc.noaa.gov/cdr/atmospheric/avhrr-cloud-properties-patmos-x. Additional calibration correction for AVHRR reflective bands have been established using the statistics of cold cloud observations over oceans in the tropical zone. INPUT AUXILIARY DATA Several static datasets are used in the GMASI system to filter out possible false snow and ice identifications and hence to improve the quality of snow and ice retrievals. These dataset are as follows 1. Snow cover weekly climatology (source: NOAA IMS) 2. Land surface temperature climatology (source: ISCCP) 3. Sea surface temperature climatology (source: Reynolds SST, NCDC) 4. Surface elevation (GTOPO30, source: USGS) KNOWN ISSUE/WEAKNESSES [not finished] Although all efforts have been made to provide the best quality product, a number of issues still remains. Below we provide a short list of issues/weaknesses we have noticed and identify possible ways to make proper correction in the next versions of the dataset. 1. At this time the dataset covers the time period from 1988 to 2017. The quality of the maps may be lower in the end of 1980s and in the beginning of 1990s primarily due to the fact that at this time SSMI observations from only one DMSP satellite (F-8) were avaialble. Within he current system the best quality microwave-based snow and ice retrievals are made with data avaialble from at least three platforms. 2. Snow misses over forested areas during the snow melt period. Snow masking by the tree canopy hampers snow identification with visible/infrared data. Microwave observations are not sensitive to the melting snow. 3. Reduced quality retrievals prior to the year 2001. Mostly associated with the lack of observations in the shortwave infrared band (centered at 1.6 micron) with older AVHRR instruments. Derived reflectance in the middle inrared band (centered at 3.7 micron) is used instead. The derived values of mid-infrared reflectance are associeated with larger noise. They are are less sensitive to the land surface cover properties and provide less efficient discrimination between clouds and cloud-clear scenes as compared to the reflectance in the shortwave infrared band. The most noticeable effect of using observations in the middle infrared instead of shortwave infrared in the AVHRR-based snow retrievals is a frequent overestimate of the snow cover over mountainous areas, particulalry over Tibet. 4. Gradually decreasing quality of snow retrievals in 1993-1994 and 1999-2000. These periods of time are associated with a considerable shift of orbits of NOAA-11 and NOAA-14 satellites towards late overpass time. Due to this shift the AVHRR observation geometry gradually shifted from the cross-prinicipal plane to observations in the pricipal plane. In the principal plane the surface reflectance anisotropy effects are the largest both for the snow-covered and the snow-free land surface. Large azimuthal variation of the land surface reflectance complicated consistent snow and ice retrievals acrross the scan and resulted in an increasing number of both snow misses and false identifications. These effects get more pronounced towards the edge of the scan. 5. For most of the years prior to 2000 coverage of the Southern Hemisphere regions affected by seasonal snow is poor. This is due to late overpass times of NOAA afternoon satellites over these areas.