Service Interruption

The NSIDC Web site and data services are currently having intermittent problems and may be unavailable. We are working to restore these services as soon as possible and apologize for any inconvenience this may cause. Please contact NSIDC User Services for assistance.

NASA Distributed Active Archive Center (DAAC) at NSIDC


Moderate Resolution Imaging Spectroradiometer

Great Lakes, 09 March 2003

Three of North America's Great Lakes -- Lake Huron, Lake Superior and Lake Erie -- have frozen over for the first time in nearly a decade after icy weather lasted from February through March 2003, according to experts at Environment Canada. Lake Michigan, however, still had a large percentage of open water in early March, as shown in the first two pairs of images below.

The images on the left are MODIS true-color reflectance composites (bands 1, 4, and 3) at 1 km resolution of the Great Lakes on 09 March 2003. These images are derived from Level-1B calibrated radiances, available from the MODIS Characterization Support Team (MCST): Level 1B Product Information and Status Web site.

The images on the right, which correspond to roughly the same areas as the left images, show the extent of snow-covered lake ice at 500 m resolution. These are derived from the MODIS/Terra Snow Cover Daily L3 Global 500m Grid, Version 4 data set, available from NSIDC. Data were reprojected from a sinusoidal projection into a Universal Transverse Mercator (UTM) projection, using the MODIS Reprojection Tool.

A 1 km resolution land/water mask classifies pixels that represent inland water. A Normalized Difference Snow Index (NDSI) algorithm further identifies snow-covered ice in these pixels. The images often show lake ice on land because of limitations with the cloud mask and NDSI algorithms. Refer to Figure 4. These areas of lake ice are actually inland water bodies identified by the land/water mask. This mask is liberal in identifying areas of shallow inland water, particularly in regions like Minnesota that have an abundance of inland water. The NDSI algorithm assigns snow-covered lake ice to these pixels. This limitation is known, and Boston University researchers are currently working on a more restrictive land/water mask that will accurately map inland water.

The MODIS snow algorithm still has limitations, particularly in discriminating clouds from snow-covered lake ice, as the northern Lake Michigan ice image shows. Also, the MODIS cloud mask often classifies translucent, or thin clouds as cloud, precluding observations of the Earth's surface.

The following files are the source files used to create this image:

The following legend is used with these colored images:

Snow cover legend

Figure 1. Lake Michigan, South
Lake Michigan South visible image Lake Michigan South ice cover

Figure 2. Lake Michigan, North
Lake Michigan North visible image Lake Michigan North ice cover

Figure 3. Lake Erie
Lake Erie visible image Lake Erie ice cover

Figure 4. Lake Superior, West
Lake Superior visible image Lake Superior ice cover

Figure 5. Lake Huron
Lake Huron visible image Lake Huron ice cover