This is the third of a three part series about parallax issues with satellite data. Part 1 was an introduction to the problem and Part 2 focused on specific concerns with regard to geostationary data, but parallax is not only a geostationary satellite problem. In this section we’ll more closely examine parallax considerations with polar satellite data.
Polar weather satellites take around 90 minutes to complete one orbit, and as the earth rotates beneath them, each pass covers a section of earth that is further west. This provides nearly global high resolution coverage without the large parallax displacement errors over higher latitudes that is characteristic of geostationary satellites.
As the satellite orbits on its polar track, it scans the earth from side to side creating a swath of data. The “nadir” is the location where the track of the satellite is directly overhead and the viewing angle increases toward the edge of the scan.
Since polar satellites orbit at a much lower altitude than geostationary satellites, resolution is much improved and cloud displacement is negligible near the nadir, the central portion of a swath. However, for lower orbits the viewing angle and parallax displacement grows more quickly with increasing distance from the nadir.
In the figure above, the viewing angle to the cloud feature for the lower satellite A is greater than the angle for satellite B, so the displacement error of the cloud due to parallax is greater for satellite A. Note that the apparent footprint of the cloud feature is larger with satellite A. This illustrates that there are limits to the width of good polar data in a single pass because of the viewing angle. Nwer polar satellites are able to significantly reduce image degradation near the edge of the scan by “oversampling”, which will be discussed in a later blog.
The graph below is a plot of the Normalized Cloud Offset (NCO) as a function of distance. This plot is similar to the geostationary version in the previous post, but for polar satellites with units of kilometers on the x-axis rather than degrees of latitude. The blue line is the polar NCO with respect to distance from the nadir, and for comparison the dashed grey line shows the NCO for geostationary data at the same distance.
As mentioned in the previous post, the NCO is a ratio of the cloud displacement away from the nadir to the height of the cloud (in equivalent units). So, at a distance of 1300 kilometers from the nadir the NCO is approximately 2.0. For a cloud top at 20,000 ft (3.8 miles) at this distance you could expect a displacement error of around 7.6 miles away from the center of the satellite track (2.0 * 3.8). Note that for surface-based features such flooded rivers, burn scars, and open leads in the sea ice there would be negligible parallax error regardless of the distance from the nadir (NCO is near zero).
To see an example of cloud displacement at the swath edge, two overlapping passes from VIIRS and MODIS were selected that were only a few minutes apart (image below). Two boxes outline areas that are near the center of one pass and the edge of another. Box A is near the center of the MODIS pass and the edge of VIIRS, Box B is near the center of VIIRS and the edge of MODIS.
In the animated close up of Box A at right, you can see that the convective clusters in the VIIRS image are displaced about 6 miles to the west of the position in the MODIS image. The location of the convection in the MODIS image is fairly accurate since parallax error is negligible near the center of the swath. Since the distance from the nadir to the edge of this VIIRS image is around 1400 km the NCO is around 2.2. In order to get a displacement of 6 mi (31600 ft) the cloud tops should be around 14400 ft (31600 / 2.2). Note there is some loss of resolution near the edge of the VIIRS swath.
The animated close-up of Box B shown at left is the reverse of the scene above with MODIS near the edge of the pass and VIIRS close to the center. In this case the VIIRS image has minimal parallax displacement for the high cumulus bands and MODIS has displaced the cells 8 miles to the east southeast toward the edge of the swath away from the nadir. The distance from the nadir for the MODIS edge is less than it was for VIIRS (1100 km) so the NCO is actually less than the previous example. The reason this region has greater displacement error is likely due to higher cloud tops.
In northern latitudes, polar satellite data is a critical complement to geostationary data with better resolution and minimal parallax error, especially near the center of the swath. However, as the examples above illustrate, regardless of satellite type, viewing angle and parallax displacement should always be taken into account when the exact location of a cloud feature is important.
By Carl Dierking