Chi-Fan Shih (NCAR) noticed that the input data for CDAS was smaller than normal for May 10-11, 2003. Under normal circumstances, missing data is a fact of life. If we don't receive the data over the GTS, there is little we can do. However, the loss of the satellite data on May 10-11 was an exception. The data was received at NCEP and we were able to recover the data. As result, it was possible to reprocess CDAS with the previously missing satellite data.
The injested satellite data consists includes winds and ATOVS temperature retrievals. For temperatures, the ATOVS is more important than the satellite winds. The addition of the ATOVS satellite data eliminated the drop in the 100 mb temperatures. The red line is the global 100 mb temperature for CDAS with missing satellite data. The black line is the revised CDAS which used the previously missing satellite data.
The spatial structure of the satellite impact is similar to that found by the TOVS filtering problem. The difference in 100 mb temperatures on the affected times is shown by the following plot.
The vertical structure for the 20S-20N shows a strong negative impact around 100 mb, a small positive impact at 500 mb and a small negative impact at 850 mb.
The impact of losing the satellite data is consistent with what was found with the TOVS filtering problem. The 100 mb tropical temperature is sensitive to the TOVS data. Any change in the bias of the TOVS retrievals or the number of profiles injested will affect the 100 mb tropical temperatures.
Users can tell that they are using v1.1 cdas using
wgrib -V grib_file
CDAS v1.0 has a process id of 180 while v1.1 uses 181. This revision affects 2003050900 - 2003062618.