DO NOT use F15 data from 2006-Aug-14 forward for climate research.
Since 2006-Aug-14 UTC, 22GHz(V) on SSM/I F15 has been degraded by a RADCAL beacon.
On 2009-April-19, we applied RADCAL beacon correction. This temperature dependent correction is much improved for times when the F15 satellite is eclipsed by Earth's shadow, which cools the instrument and amplifies the beacon's interference. Residual errors are larger during periods of eclipse.
F15 - F13 plots estimate residual errors after RADCAL beacon correction:
Time series of the difference between F15 and F13 retrievals using the Version-2 temperature dependent RADCAL correction (above). The hot load temperature is shown in shades of grey, with darker being colder. The start of beacon operation is marked with a black vertical line in August 2006. Notice that 2008 has two cool events, one in the spring and one in the fall. The very cold event at the beginning of 2009 has temperatures about 50 K less than the nominal values.
Former (obsolete) correction:
Obsolete Version-1 Hilburn and Wentz (2008) RADCAL correction is shown below. The RADCAL signature of low wind (green) and rain (black) retrievals and of high vapor (red) and cloud (blue) retrievals can easily be seen during the cold events.
In August 2006, a radar calibration (RADCAL) beacon was activated on the F15 DMSP satellite. The beacon was found to severely interfere with the 22 GHz SSM/I channel. The effect of the RADCAL beacon on geophysical retrievals was to bias wind and rain retrievals low, and bias vapor and cloud retrievals high. Retrieval biases were on the order of 30-40%. We developed a correction for the RADCAL beacon interference (Hilburn and Wentz, 2008). The correction consisted of a static table of brightness temperature offsets, with one value for each of the 64 cells in a scan line. The mean offset value is 10 K with higher offset values on the left side of the swath and lower values on the right side of the swath. The simple correction was able to reduce the wind and vapor biases to less than 1% and reduce the cloud and rain biases to a few percent.
The quality of the corrected data began to degrade in 2008, and by the beginning of 2009, F15 was once again exhibiting biases larger than 40% causing us to remove the data from our web site and ftp access. During certain times of the year (February/March and September/October), F15 enters the Earth’s shadow and the satellite’s electronics cool off. This cooling worsens the RADCAL interference. As the orbit of F15 drifts to earlier times-of-day, the cooling has resulted in more severe biases.
We have subsequently found that these temperature-dependent effects can be successfully characterized in terms of the hot load temperature. The relationship between temperature and interference is quadratic, with the interference becoming much stronger at cooler temperatures. The temperature dependence modulates the Hilburn and Wentz (2008) tabular correction in a multiplicative way. The Version-1 figure above shows the familiar RADCAL signature of low wind and rain retrievals and high vapor and cloud retrievals when just the Version-1 Hilburn and Wentz (2008) correction is applied. The temperature-dependent correction removes most of these problems, except during the most severe cold periods, where it seems that biases as large as 10% are unavoidable.