Data Acquisition

The Interactive Multisensor Snow and Ice Mapping System (IMS) product is produced at the National Ice Center (NIC) from a variety of sources, including:

Visible and Infrared Spectral Data

• Advanced Very High Resolution Radiometer (AVHRR) Channels 1 & 3
• Geostationary Operational Environmental Satellite (GOES) East and West
• Meteosat 7
• Meteosat Second Generation (MSG)
• Moderate Resolution Imaging Spectroradiometer (MODIS) Channel 8
• Multi-functional Transport Satellite (MTSAT)
• Suomi National Polar-Orbiting Partnership

Ancillary Data

• Surface Observations (METAR)

Derived Data

• Air Force Weather Agency (AFWA) Snow Depth
• Advanced Microwave Scanning Radiometer for EOS (AMSR-E)
• Advanced Microwave Sounding Unit (derived snow, ice, and rain)
• Advanced Scatterometer (ASCAT)
• Envisat Advanced Synthetic Aperture Radar (ASAR) operating in Global Monitoring Mode
• SSMI/S (Derived snow, ice and rain)
• NASA Quick Scatterometer (QuikSCAT)
• NOAA Office of Satellite Data Processing and Distribution (OSDPD) Automated Multisensor Snow and Ice
• National Operational Hydrologic Remote Sensing Center (NOHRSC) Snow Data Assimilation System (SNODAS)
• United States Air Force Snow and Ice Analysis Product

Trained analysts at NIC produce IMS daily 24 km snow maps once per day, using an interactive workstation application to incorporate a wide variety of satellite imagery, derived mapped products, and surface observations. IMS daily ASCII files were acquired from NIC, converted to binary using the NCAR Command Language (NCL), and regridded to EASE-Grid 2.0 with Mapx.

Processing Steps

The data were compared with the BU-MODIS LOCI 25 km mask. In cases where an ocean pixel was converted to land, a 3x3 moving box filter was used to determine snow cover. With the converted cell in the center, surrounding land cells were examined and if more than 50 percent of the original surrounding land cells were snow covered, the cell was assigned a snow cover value. Otherwise the cell was determined to be snow-free land.

In cases where an island pixel was created by converting ocean to land, the same 3x3 box filter was utilized to determine the percentage of adjacent cells with IMS values for ice. If more than 50 percent of the surrounding water cells were ice, the new pixel was assigned a snow cover land value.

Once reconciled with BU-MODIS LOCI 25 km mask, the data were written to the ims_snow_cover_extent variable.

For more information about the Interactive Multisensor Snow and Ice Mapping System, visit the U.S. National Ice Center IMS Products Web page.

Data Acquisition

Snow cover extent was derived from passive microwave brightness temperatures using the DMSP SSM/I-SSMIS Pathfinder Daily EASE-Grid Brightness Temperatures data set. In a series of ongoing missions, the Defense Meteorological Satellite Program (DMSP) has been acquiring brightness temperature data since 09 July 1987 with the Special Sensor Microwave/Imager (SSM/I) and Special Sensor Microwave Imager/Sounder (SSMIS) instruments onboard DSMP satellites F8, F11, F13, F17.

Table 7 lists the satellites, instruments, and date ranges that were used for this data set.

Processing Steps

The following steps were used to convert brightness temperatures into snow cover:

1. SSMIS F17 brightness temperatures were converted to SSM/I F13 based on the methodology in Meier et al., 2011. Ascending and descending passes were averaged into daily average brightness temperatures.
2. The presence of snow was identified where the 19/37 GHz and 22/85 GHz (91 GHz for SSMIS) vertical polarization frequency gradients exceed minimum thresholds. Additionally, the brightness temperature for the 37 GHz channels (horizontal and vertical polarization) and 85GHz (vertical polarization) must not exceed a maximum threshold. The approach is conceptually similar and uses the same frequency gradients and channels as the AMSR-E Snow Water Equivalent Algorithm.  Refer to the text by Chang and Rango, 2000 for more information. Thresholds were set to achieve a 60 percent probability of detecting snow, determined by: hemispheric comparison with IMS snow; snow depth observations from available stations including the Global Historical Climatology Network, cooperative stations, and the Meteorological Service of Canada; and daily average brightness temperatures during the 2005-2006 snow season. The thresholds are listed in Table 8.

   Table 8. Criteria for Identifying Snow

   Channel	Threshold
   19V–37V	> 7 K
   22V–85V (22V–91V)	> 8 K
   37V	< 256 K
   37H	< 243 K
   85V	< 253 K

3. At locations where the elevation exceeds 1500 meters above mean sea level, the 19/37 GHz and 22/85 GHz frequency gradients were reduced by 0.001 K and 0.002 K per meter elevation, respectively. This correction was applied because passive microwave snow retrievals overestimate snow-covered area in high elevation locations, and in particular the Tibetan Plateau, because the decrease in atmospheric thickness increases the spectral gradient. Refere to the text by Savoie et al., 2009 for more information. The coefficients were determined by a comparison with high elevation (>1500 m) locations in the same manner discussed in Step 2.
4. In regions with a maximum snow-covered albedo ≤58 percent, the 19/37 GHz and 22/85 GHz frequency gradients were increased by 3 K and 4K, respectively. Because snow that falls through the canopy may not be apparent when on the ground (Foster et al. 1991), passive microwave snow retrievals underestimate snow covered area in densely forested locations such as the boreal forests. To account for this effect, a maximum snow covered albedo data set (Robinson et al., 1985) was regridded to EASE Grid 2.0 and used to identify regions in which snow cover is spectrally masked by vegetation. Coefficients were determined by comparing areas with maximum snow covered albedo ≤58% in the same manner discussed in step 2.
5. The criteria described in Step 2, as modified in steps 3 and 4, were applied to the daily average brightness temperatures to determine the presence or absence of snow. Internally, separate codes were used to track cells with snow cover identified by only the 19 GHz and 37 GHz channels or the 85 (91) GHz channels.
6. Areas of missing data were reduced by employing a six-day gap fill. If data were missing for a given date, the most recent previous date with valid data (up to five days prior) was retained as the observation for the current day.
7. The resulting gap-filled snow cover map was regridded to EASE-Grid 2.0 using Mapx.
8. Areas of permanent ice and ocean were masked using the BU-MODIS LOCI 25 km mask. A maximum snow cover extent mask was also applied based on all cells with IMS data that indicated snow on any date during a given month.
9. Intermediate, daily two-byte flat binary files in EASE-Grid 2.0 were created for the microwave snow layer. The data were then written to the passive_microwave_gap_filled_snow_cover_extent variable.