Bruce L. Gary


The following analysis is meant to show how well XS-file temperatures agree with radiosondes, RAOBs.  XS temperatures are expected lose quality far from RAOB sites, since the XS-data is heavily dependent upon RAOBs as input.  The first step in understanding the performance of XS temperatures is to evaluate their performance close to RAOB sites.

No Layer Averaging

This analysis is based on data in the SOLVE archive for 3 flights:  ER000111, ER000114 and ER000318.  No aircraft data is used.  Rather, RAOBs are interpolated in time for comparison with the time in the XS-file when the ER-2 flew near the RAOB site.

Figure 1.  Each of the 13 RAOB minus XS profiles.

Figure 2. Average (red dashed), standard deviation of the population of temperature differences (green dashed) and RMS temperature difference (thick blue).

Assuming the RAOBs are "true" the XS-fle temperatures exhibit an (positive) offset of about 1 K.  The RMS difference between XS and RAOB temperatres is about 2 K.  The bulge at 12 km must be associated with the tropopause, which is to be expected. There are large offsets above 23 km, and these dominate the RMS difference trace at these altitudes.  Assuming, for now, that the large offset above 23 km is not typical, it is approximately accurate to say that the XS-file temperatures exhibit an RMS uncertainty of about 2.0 K at all altitudes that are not close to the tropopause, where an RMS uncertainty of 2 or 3 K is found.

Layer Averaging

Because the XS data is a synoptic scale model data product, it seems only "fair" to evaluate it with RAOB profiles that have been subjected to altitude averaging.  I have chosen to use 1-km averages.  The following two figures are counterparts to the first two, differing only with the use of 1-km layer averaged RAOB data.

Figure 3.  Each of the 13 RAOB minus XS profiles for 1-km thick layer averages of RAOB data.

Figure 4.  Average (red dashed), standard deviation of the population of temperature differences (green dashed) and RMS temperature difference (thick blue), for layer average ROAB data.

The differences in XS RMS accuracy using layer-averaged RAOB data versus unaveraged RAOB data are miniscule!  This is shown in the next figure.

Figure 5.  RMS accuracy of XS-file temperatures as measured against RAOB raw data and RAOB layer-averaged data.

The altitude-averaged RMS accuracies in this figure are 1.828 K (raw RAOB as reference) and 1.751 K (layer-averaged RAOB as reference).  Their ratio is 0.96.  Thus, by subjecting the RAOB data to layer averaging before evaluating the XS-file temperature field improves the apparent XS-file temperature field accuracy by only 4%!   


This site opened:  January 5, 2001.  Last Update: January 6, 2001