List of Images
For a general description of the images, please refer to the Introduction to the Data Sets. For access to the images, please read the Administrative Details.
Here is the LIST OF IMAGES in ASCII format. The table contains 195 lines, one for each image. Here is an explanation of the fields in the table:
The image identifier (ID) is the name assigned by the Alaska Satellite Facility (ASF). For example, the first image ID in the table is: R110406261P3S010. All the image IDs start with "R1", which means "RADARSAT-1". The next 5 digits are the orbit number, such as 10406 in the example. The next 3 digits are the frame number, such as 261 in the example.
Each image from ASF comes with a leader file that includes all kinds of information about the image, most of which you will probably not need (I have tried to include all the relevant metadata information here on this web page). Nevertheless, here is a tar file that contains all 195 leader files:
Download tar file of leader files (6 megabytes)
To read the information in each leader file, you can use ASF's "metadata" program.
Alternatively, you can decode the bytes yourself (e.g. with the unix utility "od" - octal dump) and then look at the CEOS specifications to find out what the numbers mean.
Note that some of the fields in the leader files apply to the full, original images, and not to the sub-images available through this web site. In particular, the corner coordinates in the leader file apply to the full image. See Geolocation below for computing coordinates within the sub-images.
Animations
The animations are in MPEG format, a lossy compression format, and therefore are not suitable for quantitative analysis. Think of them as time series of browse images. A full set of 195 images at 800 x 800 pixels would be 125 megabytes, but each MPEG file is only about 3 megabytes - a compression factor of more than 40!
Download MPEG file of 40 x 40 km animation (3.1 megabytes) (Copyright CSA 1997-1998)
Download MPEG file of 200 x 200 km animation (2.7 megabytes)(Copyright CSA 1997-1998)
GIF animations: visit EOL's dataset page. These are much better quality than the MPEG animations but the files are too large to post.
Since the overall brightness of each image depends strongly on the incidence angle, which varies from about 19^{o} to 46^{o} (see the section on Incidence Angle below), an adjustment has been made in scaling each image so that the frames don't flash from dark to bright to dark, distracting the eye's ability to follow movement and features. Each image is nominally adjusted to 32.5^{o} incidence angle by adding (I-32.5)*0.135 (dB) to each pixel, where I is the center incidence angle. The factor 0.135 (dB/degree) is a typical slope for backscatter vs. incidence angle (see FIGURE 7 from Stern and Moritz [JGR, 2002]).
Many interesting features and events are evident in the animations. Some of the highlights are the large divergent and convergent events in January and early February, and the abrupt onset of melt on May 29. See the Introduction to the Data Sets for links to Stern and Moritz [JGR, 2002], which contains more details.
There are other ways to make better quality animations than the MPEG ones given here. For example, IDL has a routine called "xinteranimate" that puts image frames into an animated loop with speed controls. Here is an IDL PROGRAM that sets up the images and calls xinteranimate.
File Format and SAR Backscatter
There are two sets of sub-images, 40 x 40 km and 200 x 200 km. Each sub-image is centered on the SHEBA station. The 40 x 40 km images have a pixel size of 50 meters, while the 200 x 200 km images have a pixel size of 250 meters (more on this below). Thus all the sub-images are 800 x 800 pixels.
The sub-images are in GIF format. There are many programs available for reading and displaying GIF images, such as "xv" and IDL. For example, in IDL one could read and display a GIF image like this:
IDL> read_gif, 'file_name', image_name
IDL> window, /free, xsize=800, ysize=800
IDL> tv, image_name
Each pixel is one byte, hence values range from 0 to 255. This scaling is a result of the ASF "calibrate" program that was used to apply the calibration coefficients to the SAR images. SAR backscatter is the (dimensionless) ratio of backscattered to incident power, commonly expressed in decibels (dB) by taking the logarithm (base 10) of the ratio and multiplying by 10. The backscatter can be recovered from the byte pixel values via:
backscatter (dB) = ( (byte pixel value) - 255 ) / 10 | (1) |
For example, a pixel value of 155 is -10 dB. To remove the decibel transformation, compute:
backscatter (ratio) = 10 ^{ (backscatter (dB)) / 10 } | (2) |
The 40 x 40 km sub-images retain the original pixel size of 50 meters. The 200 x 200 km sub-images, with a pixel size of 250 meters, were created as follows: First, equations (1) and (2) were used to convert the 50-meter pixel values to backscatter (ratio) for the 200 x 200 km region. Then the backscatter values were averaged in non-overlapping 5 x 5 blocks, and the resulting values were scaled back to byte values by the inverse of equations (2) and (1).
Geolocation
The original images (and hence the sub-images) are geolocated to the SSM/I polar stereographic coordinate system. This is based on the Hughes ellipsoid with standard parallel at 70^{o}N. The origin is at the North Pole, the +X axis runs along 45^{o}E longitude, and the +Y axis runs along 135^{o}E longitude. The units are kilometers. For example, the coordinates (-1000, 0) would refer to a point along the -X axis (i.e. 135^{o}W longitude) and 1000 km from the North Pole. In general, the transformation between (X,Y) and (latitude,longitude) is rather complicated. Here are some IDL PROGRAMS that convert back and forth.
In each sub-image, the SHEBA station is located at column 400, row 399 (0 relative, with rows enumerated from the bottom up). The latitude and longitude, and the SSM/I (X,Y) of that pixel are given in the List of Images at the top of this page. The coordinates of other pixels in the image are easily computed. For example, let (X_{SS},Y_{SS}) be the coordinates of the SHEBA station in one of the images. The coordinates of a pixel at column C and row R would be:
X = X_{SS} + ( C - 400 ) * pixel_size | (3) |
Y = Y_{SS} + ( R - 399 ) * pixel_size | (4) |
where pixel_size = 0.250 kilometers for the 200 x 200 km images or pixel_size = 0.050 kilometers for the 40 x 40 km images.
Incidence Angle
The radar backscatter of most surface types decreases as the incidence angle goes from 0^{o} (nadir-looking) to 90^{o} (grazing). The incidence angle in the RADARSAT SAR images of SHEBA varies from about 19^{o} to 46^{o}, and there is a considerable drop in backscatter over this range - see FIGURE 7 from Stern and Moritz [JGR, 2002]. The drop is about 0.135 dB per degree.
The last column in the List of Images at the top of this page gives the incidence angle at the SHEBA station (center of sub-image, column 400, row 399). Within a single 40 x 40 km sub-image, the incidence angle only varies by plus-or-minus 1.2^{o} from the center value, so the contribution to backscatter variations is small (about plus-or-minus 0.17 dB). For the 200 x 200 km images, the contribution is about five times larger.
Computing the incidence angle at an arbitrary pixel location within an image is complicated, because it depends on the orbital parameters of the spacecraft and the shape of the earth. Fortunately, a simple and accurate linear approximation can be computed, if precise incidence angle values are needed. We now describe the procedure.
For each sub-image, we have computed the incidence angle at each pixel and fit it with a bilinear function of the column and row number. Thus four coefficients per sub-image are sufficient to compute the incidence angle at any location. If COL is the column number and ROW is the row number (which vary from 0 to 799), then
incidence angle = A + B * COL + C * ROW + D * COL * ROW | (5) |
where A, B, C, D are the four coefficients. Here is the TABLE OF COEFFICIENTS for all the sub-images. The table contains 390 lines, 195 for the 40 x 40 km sub-images and 195 for the 200 x 200 km sub-images. The fields in the table are:
The "center value" (field #2) is the value of the bilinear function, equation (5), at the center of the sub-image. The amount by which it differs from the actual incidence angle at the center (last column in the List of Images) is a measure of the goodness-of-fit of the bilinear function. Two other measures of the goodness-of-fit are given in fields #8 and #9. For the 40 x 40 km sub-images, the mean absolute deviation (MAD) of the bilinear fit from the true incidence angle is always less than 0.01 degrees. For the 200 x 200 km sub-images, the MAD is always less than 0.22 degrees. This shows that the bilinear fit is quite accurate.
As an example of how to use the numbers in the above table, consider the first line:
R110406261P3S010 41.37 c 40.04 9.874e-04 2.376e-03 -5.239e-08 0.0081 0.0031
which refers to the 40 x 40 km sub-image ("c" in field #3) whose ID is R110406261P3S010. The incidence angle at any COL and ROW is then approximately: