Service Interruption

Jobs | Contact Us

Global EASE-Grid 8-day Blended SSM/I and MODIS Snow Cover

Summary

This data set comprises global, 8-day Snow-Covered Area (SCA) and Snow Water Equivalent (SWE) data from 2000 through 2008. Global SWE data are derived from the Special Sensor Microwave Imager (SSM/I) and are enhanced with MODIS/Terra Snow Cover 8-Day Level 3 Global 0.05 degree Climate Modeling Grid (CMG) data. Global data are gridded to the Northern and Southern 25 km Equal-Area Scalable Earth Grids (EASE-Grids). These data are suitable for continental-scale time-series studies of snow cover and snow water equivalent. The data are in netCDF data files and PNG browse image files and are available via FTP.

Citing These Data

The following example shows how to cite the use of this data set in a publication. For more information, see our Use and Copyright Web page.

The following example shows how to cite the use of this data set in a publication. List the principal investigators, year of data set release, data set title and version number, dates of the data you used (for example, December 2003 to March 2004), publisher: NSIDC, and digital media.

Brodzik, M. J., Richard Armstrong, and Matt Savoie. 2007. Global EASE-Grid 8-day Blended SSM/I and MODIS Snow Cover. [list the dates of the data used]. Boulder, Colorado USA: National Snow and Ice Data Center. Digital media.

Overview Table

Table of Contents

1. Contacts
2. Detailed Data Description
3. Data Access and Tools
4. Data Acquisition and Processing
5. References and Related Publications
6. Document Information

1. Contacts

Investigator(s) Name and Title

Mary J. Brodzik
National Snow and Ice Data Center
CIRES, 449 UCB
University of Colorado
Boulder, CO 80309-0449 USA

Richard L. Armstrong
National Snow and Ice Data Center
CIRES, 449 UCB
University of Colorado
Boulder, CO 80309-0449 USA

Matt Savoie
National Snow and Ice Data Center
CIRES, 449 UCB
University of Colorado
Boulder, CO 80309-0449 USA

Technical Contact

2. Detailed Data Description

Format

The data files are in Network Common Data Form (netCDF) format and are identified with the .nc file extension. Each netCDF file, except for the current year file, contains a full year of 8-day data. The current year file only contains data currently available and will be updated as new data becomes available. The files contain data arrays with dimensions of 721 columns by 721 rows. For more information on the netCDF file format, please see the NetCDF Resources at NSIDC Web site or visit UCAR Unidata's NetCDF Web page. Browse images of snow data in Portable Network Graphics (PNG) format are also included.

File and Directory Structure

Data are on the FTP site in the nsidc0321_blended_ssmi_modis directory. Within this directory, there are two subdirectories as described in Table 1.

File Naming Convention

NetCDF Files

The netCDF files are named according to the following convention:

GG.YYYY.nsidc0321vXX.nc

Where:

Browse Image Files

The browse image files are named according to the following convention:

GG.YYYYMMDD-YYYYMMDD.nsidc0321vXX.png

Where:

File Size

The netCDF files range in size from 82 to 100 MB per yearly file, and the browse PNG files range in size from 12 to 44 KB per 8-day image file.

Spatial Coverage

These data are provided in two different spatial coverages: Northern and Southern Hemispheres. Please see the Grid Extent Table on the EASE-Grid: A Versatile Set of Equal-Area Projections and Grids Web page for specific latitude and longitude values.

Spatial Resolution

This data set is derived from multiple sources. While the files are gridded at 25-kilometer spatial resolution, the actual resolution of the component data, SWE or SCA, depends on the input remote sensing data. For SWE data, the satellite passive microwave sensors at the frequencies used for these algorithms have sampling resolutions of 25 km. For SCA data, the spatial resolution of the MODIS/TERRA Snow Cover 8-Day L3 Global CMG (MOD10C2) data are 0.05 degrees, itself derived from 1 km MODIS data.

Projection/Grid Description

These data are stored in the Northern and Southern Hemisphere EASE-Grids. For more information about EASE-Grids, please see All About EASE-Grid.

Temporal Coverage

This data set ranges from 05 March 2000 to 24 January 2008 and has an 8-day resolution. Please see the Data Set Release History section of this document for more information.

Temporal Resolution

These data are provided as 8-day composites.

Parameters

The parameters of this data are SWE and SCA. SWE is a measurement of the amount of water contained within a snowpack. The SWE data for this data set are derived from DMSP SSM/I-SSMIS Pathfinder Daily EASE-Grid Brightness Temperatures. SCA, as the name implies, is the total area of land covered by snow. The SCA data used to enhance this data set are derived from the MODIS/Terra Snow Cover 8-Day L3 Global 0.05Deg CMG (MOD10C2) data, regridded to the NL and SL EASE-Grid.

Parameter Range

Table 4 describes the values of the SWE variable found in the netCDF files.

Table 5 describes the values of the SCA variable found in the netCDF files.

Sample Data Records

Figures 1 and 2 are sample browse images for the Northern and Southern Hemispheres, respectively. Click on the samples to view larger images.

Northern Hemisphere

Figure 1. Northern Hemisphere PNG Browse Image Average snow water equivalent (mm) from passive microwave with additional red area classified as snow by MODIS for more than 25% of the grid cell area for 25 November 2006 - 02 December 2006.

Southern Hemisphere

Figure 2. Southern Hemisphere PNG Browse Image Average snow water equivalent (mm) from passive microwave with additional red area classified as snow by MODIS for more than 25% of the grid cell area for 05 August 2007 to 12 August 2007.

Data Set Release History

Table 6 describes the release history for this data set. New releases are added to the top of the table.

Notes and Limitations

This data set is intended for studies of continental- to hemispheric-scale seasonal fluctuations of SCA and SWE. Due to the lack of in situ validation data for SWE at the spatial scale of the microwave sensors, SWE derived from satellite passive microwave sensors should be considered with caution. The effective field of view of these passive microwave sensors yields radiometric information from an area that is larger than 625 square kilometers. The gridded value represents a mean SWE for this large area; therefore, this value cannot capture localized maxima or minima. The large sensor footprint and other limitations of microwave sensors result in decreased confidence in the SWE reliability and possible undermeasure in the following circumstances:

• Mountainous areas with large topographic variability return low mean SWE values. Brightness temperatures in these areas may include mixed emission from deep snow on north-facing slopes, snow-free south-facing slopes, wind-scoured Alpine areas, etc.
• Forested areas return low mean SWE values because the mixed signal includes emission from trees and the snow canopy as well as the underlying surface.
• Areas near coastlines return low or no SWE values because the mixed signal includes frozen and unfrozen water and possibly snow-free land.
• Areas containing melting snow or wet snow packs typical of maritime snow conditions return low or no SWE values because the microwave emission from liquid water overwhelms scattering from the snow pack.
• Shallow or intermittent snow during fall and early winter typically does not result in sufficient microwave scattering to reliably detect SWE. See the Processing Steps section of this document for shallow snow improvement used in this data set.

Lower confidence in SWE reliability due to overmeasure may also occur due to the following circumstances:

• Areas with significant depth hoar formation. The conditions for depth hoar formation involve the combination of shallow snow exposed to strong temperature gradients driven by cold air temperatures over a period of weeks to months. This results in a snow cover with large grains that enhance the microwave scattering signal and cause overmeasure when a particular algorithm has been tuned to a smaller mean grain size. A typical region prone to this type of snow texture is Eastern Siberia. The seasonal snow cover consistently begins to form in this region as early as September, and then relatively shallow snow remains on the ground as air temperatures begin to approach the extremely cold conditions of winter. The SWE values in this data set indicate greater values for Eastern Siberia than for Western Siberia, although some climate models indicate the opposite. There could be some degree of overmeasure by the microwave algorithm in this region due to the persistent presence of depth hoar. Unfortunately, the investigators currently do not possess sufficient surface measurements of SWE to determine with any certainty which pattern is correct.
• Areas of extremely high elevation. Current passive microwave snow retrieval algorithms are empirically based and were typically developed using data from lower elevations. The investigators are currently developing an atmospheric correction for regions of extremely high elevation on the Tibetan Plateau (Armstrong 2004) (Savoie 2007), which they expect to implement in the next revision of this data set.

There is a persistent pattern of relatively high SWE values that develops during the winter season in a large portion of the Canadian Arctic, stretching roughly from the Western edge of Hudson's Bay to the North Slope of Alaska. Unfortunately, this is an area with few ground observing stations. Although a large-scale field experiment begun in the 2003-2004 winter season by Derksen and others (Derksen and MacKay 2006) indicates that the SWE gradient across this area appears to be real and measurable, it is not as large a gradient as the microwave algorithm indicates. This feature is still under investigation and should be treated with caution.

The investigators plan to release future versions of this data set with improved SWE algorithms that will address some of these problems.

3. Data Access and Tools

Data Access

Data are available via FTP.

Volume

The volume of the netCDF files averages 94 MB per year, and the volume of the PNG browse images averages 40 KB per year.

Software and Tools

For a list of tools for reading/viewing netCDF files, please see the NetCDF Resources at NSIDC: Software and Tools Web page.

The PNG browse images are viewable in many web browsers and graphics applications.

Related Data Collections

• Northern Hemisphere EASE-Grid Weekly Snow Cover and Sea Ice Extent Version 3
• DMSP SSM/I-SSMIS Pathfinder Daily EASE-Grid Brightness Temperatures
• EASE-Grid Version of Boston University's MODIS Land Cover Characterization
• MODIS/Terra snow cover 8-day L3 global 0.05deg CMG V005 (MOD10C2)

4.Data Acquisition and Processing

Processing Steps

The netCDF data files for each hemisphere and year contain 2 EASE-Grid data layers for each 8-day period, derived as follows:

1. The SCA layer is derived from the 8-day MOD10C2 data, regridded to the NL or SL EASE-Grid. An output 25 km grid cell is the drop-in-the-bucket (equally weighted) average of percent SCA from the component 0.05 degree cells with snow detected by MODIS. Input cells classified as cloud, night, or missing are ignored. At the time of processing, the MOD10C2 Version 4 (V004) data are being reprocessed to Version 5 (V005). V005 data are used if they are available, otherwise V004 data are used. The string variable bpInfo in the netCDF file contains the identifier MOD10C2.005 or MOD10C2.004 for each layer depending on the input used to derive that layer. The SCA data layer includes percent SCA for all grid cell locations regardless of passive microwave returns at this location. Microwave and visible data are blended in the SWE data layer as described in Step 3.
    
2. Input passive microwave data are daily, cold pass, DMSP SSM/I-SSMIS Pathfinder Daily EASE-Grid Brightness Temperatures. These data are generally available three to six months after acquisition. If, at the time of processing the blended data, brightness temperatures from this data set are not yet available, then a near-real-time substitute is used. The near-real-time data differ from the DMSP SSM/I-SSMIS Pathfinder Daily EASE-Grid Brightness Temperatures in two ways:
    
   1. Input swath data are obtained from the Global Hydrology Resource Center (GHRC), rather than from Remote Sensing Systems (RSS).
       
   2. Regridding from swath to grid space is done with an inverse-distance square interpolation, rather than the Backus-Gilbert interpolation.
       
   The data sets are otherwise identical. The string variable bpInfo in the netCDF file contains the identifier BG for Backus-Gilbert or ID2 for inverse-distance squared for each layer, depending on the input used to derive that layer.
    
3. Two microwave algorithms to derive SWE are used.
    
   1. Deep SWE is derived from 19 and 37 GHz brightness temperatures:
       
      1. Daily SSM/I brightness temperatures are adjusted to SMMR brightness temperatures via regression data at selected stable targets (Brodzik 2005):
          
         • SMMR_ADJ(18H) = 0.925 * SSMI(19H) + 10.110
         • SMMR_ADJ(37H) = 0.936 * SSMI(37H) + 10.74
            
      2. Daily SWE is derived from:
          
         • SWE (mm) = 4.77 (SMMR_ADJ(18H) - SMMR_ADJ(37H))
            
         The constant 4.77 is a combination of the multiplicative constant of 1.59 (Chang, Foster, and Hall 1987) with the assumption of a constant snow density of 300 kg m^-3.
          
      3. Daily SWE is adjusted for surface forest cover (Chang, Foster, and Hall 1996) using the 25 km EASE-Grid version of BU-MODIS Land Cover (Knowles 2004):
          
         • Let forest_percent = {
           0 : no forest,
           0.01-0.49 : 1-49% total forest,
           0.50 : >= 50% total forest }
           Then:
           Forest-Adjusted SWE (mm) = SWE / (1.00 - forest_percent)
            
      4. Forest-Adjusted SWE values less than 7.5 mm are considered unreliable and are set to zero (Chang, Foster, and Hall 1987)
          
      5. In the Northern Hemisphere, false SWE signals from lower latitude features such as deserts are filtered using frequency climatologies derived from the Northern Hemisphere EASE-Grid Weekly Snow Cover and Sea Ice Extent Version 3 data from 1966-2005 (Armstrong and Brodzik 2005). Pixels where Northern Hemisphere EASE-Grid Weekly Snow Cover and Sea Ice Extent Version 3 data never recorded snow in the given month are set to zero SWE. In the Southern Hemisphere, false SWE signals from tropical atmospheric phenomena are filtered using a monthly SWE frequency climatology derived from SSM/I. The Southern Hemisphere SWE frequency climatology limits legitimate SWE data to the Andes Mountains region and New Zealand.
          
      6. Daily SWE files for the eight days that correspond to the MOD10C2 data are combined using the maximum at each grid cell for the component eight days.
          
   2. Shallow SWE is derived from vertically-polarized 19, 37, and 85 GHz SSM/I cold pass brightness temperatures according to Nagler and Rott (1992). The day with the most cloud-free brightness temperatures for the component 8-day period is determined at each grid cell as the maximum positive temperature difference (37V - 85V). The brightness temperatures for the most cloud-free day are used to derive snow depth (cm) as:
       
      1. A grid cell is considered snow-covered if and only if:
          
         • 19V <= 266 K
         • (19V - 37V) >= 4K or (37V - 85V) >= 3K
            
      2. In a grid cell that is classified as snow-covered, snow depth is:
          
         • Depth (cm) = (-2.41 + 1.2 * (19V - 37V) - 0.16 * (37V - 85V))
            
      3. Snow depth is converted to SWE by multiplying depth (cm) by a factor of 3, which assumes a constant snow density of 300 kg m^-3
          
4. The SWE layer in the netCDF files is derived from a series of steps to combine the deep and shallow microwave SWE with the visible data in a reasonable way, based on our knowledge of the relative strengths and reliability of each algorithm. Refer to Table 4 for SWE values and Table 5 for SCA values.
    
5. Grid cells classified as permanent ice such as ice sheets, ice shelves, and large glaciers are determined using a 50 percent threshold for permanent ice from the 25 km EASE-Grid version of the BU-MODIS Land Cover data (Knowles 2004).

Forest Percent Mask

The forest percent map is derived from NSIDC's EASE-Grid version of the BU-MODIS Land Cover data set. Data values represent the sum of the percent area classified as any of the International Geosphere-Biosphere Programme (IGBP) forest categories. These include:

• Evergreen Needleleaf Forest
• Evergreen Broadleaf Forest
• Deciduous Needleleaf Forest
• Deciduous Broadleaf Forest
• Mixed Forest

Any pixels with forest percent higher than fifty percent are set to the fifty percent threshold, thereby bounding the forest correction of the SWE value to a maximum factor of two.

Figures 3 and 4 are samples of the Northern and Southern Hemisphere forest percent masks derived from BU-MODIS land cover data. Click on the samples to view larger images.

Figure 3. Northern Hemisphere Forest Percent Mask Derived from BU-MODIS Land Cover Data		Figure 4. Southern Hemisphere Forest Percent Mask Derived from BU-MODIS Land Cover Data

Permanent Ice Masks

Areas with permanent ice such as ice sheets, ice shelves, and large glaciers are masked using a fifty percent threshold for permanent ice from the 25 km EASE-Grid version of the BU-MODIS Land Cover data.

Figures 5 and 6 are samples of the Northern and Southern Hemisphere permanent ice masks derived from BU-MODIS land cover data. Click on the samples to view larger images.

Figure 5. Northern Hemisphere Permanent Ice Mask Derived from BU-MODIS Land Cover Data		Figure 6. Southern Hemisphere Permanent Ice Mask Derived from BU-MODIS Land Cover Data

Snow Frequency Climatologies

Snow frequency climatologies for the Northern Hemisphere are shown in Figure 7.

Figure 7. Northern Hemisphere Long-term Snow Cover Frequencies (1966-2003) These frequency maps are used for Northern Hemisphere processing and are derived from Northern Hemisphere EASE-Grid Weekly Snow Cover and Sea Ice Extent Version 3 data.

Snow frequency climatologies for the Southern Hemisphere are shown in Figure 8.

Figure 8. Southern Hemisphere Long-term Snow Cover Frequencies (1987-2003) These frequency maps are used for Southern Hemisphere processing and are derived from SSM/I-derived SWE frequency of occurrence. Areas with likely snow are limited to the Andes Mountains region and New Zealand. Frequency thresholds are a function of month, for example, twenty percent for October through May and seven percent for June through September.

5. References and Related Publications

Armstrong, R. L., and M. J. Brodzik. 2005. Northern Hemisphere EASE-Grid Weekly Snow Cover and Sea Ice Extent Version 3. Boulder, Colorado USA: National Snow and Ice Data Center. Digital media.

Armstrong, R. L., M. J. Brodzik, J. R. Wang, M. H. Savoie, O. Frauenfeld, T. Zhang. 2004. Solutions to the Snow Cover Mapping Anomaly Over the Tibetan Plateau (poster). EOS, Transactions, American Geophysical Union, 85(47), Fall Meeting Supplement Abstract C31A-0282.

Armstrong, R. L. and M. J. Brodzik. 2002. Hemispheric-scale Comparison and Evaluation of Passive Microwave Snow Algorithms. Annals of Glaciology, 34:38-44.

Armstrong, R.L. and M.J. Brodzik. 2001. Recent Northern Hemisphere Snow Extent: A Comparison of Data Derived from Visible and Microwave Sensors. Geophysical Research Letters, 23(19):3673-3676.

Brodzik, M. J., R. L. Armstrong, K. Knowles, M. H. Savoie. 2005. The Effect of Sensor Differences in Deriving Long-Term Trends from Satellite Passive Microwave Snow Extent and Snow Water Equivalent. EOS, Transactions, American Geophysical Union, 86(52), Fall Meeting Supplement Abstract U21A-0804.

Brodzik, M. J. and K. W. Knowles. 2002. "EASE-Grid: A Versatile Set of Equal-area Projections and Grids." Discrete Global Grids National Center for Geographic Information & Analysis. http://www.ncgia.ucsb.edu/globalgrids-book/ease_grid/.

Chang, A. T. C., J. L. Foster, D. K. Hall. 1996. Effects of Forest on the Snow Parameters Derived from Microwave Measurements During the BOREAS Winter Field Campaign. Hydrological Processes, 10:1565-1574.

Chang, A. T. C., J. L. Foster, D. K. Hall. 1987. Nimbus-7 SMMR Derived Global Snow Cover Parameters. Annals of Glaciology, 9:39-44.

Derksen, C. and M. MacKay. 2006. The Canadian Boreal Snow Water Equivalent Band. Atmosphere-Ocean, 44(3):305-320.

Knowles, K. 2004. EASE-Grid Land Cover Data Resampled from Boston University Version of Global 1 km Land Cover from MODIS 2001, Version 4. Boulder CO, USA: National Snow and Ice Data Center. Digital Media <ftp://sidads.colorado.edu/pub/EASE/Nl.BU-MODIS.tar.gz>.

Nagler, Thomas, and H. Rott 1992. Development and Intercomparisons of Snow Mapping Algorithms Based on SSM/I Data. IEEE Proceedings International Geoscience and Remote Sensing Symposium 1992, Houston, N.p., 812-814.

Savoie, M. H., J. Wang, M. J. Brodzik, R. L. Armstrong. 2007. Improved Snow Cover Retrievals from Satellite Passive Microwave Data Over the Tibet Plateau: The Need for Atmospheric Corrections Over High Elevations (poster). EOS, Transactions, American Geophysical Union, 88(52), Fall Meeting Supplement Abstract C23A-0942.

Table 7 lists related documents that are available on NSIDC's Web site.

6. Document Information

Acronyms

Table 8 lists the acronyms used in this document.

Document Creation Date

November 2007

Document URL

http://nsidc.org/data/docs/daac/nsidc0321_8day_ssmi_modis_blend/index.html