Steps In MTP Post-Campaign Data Analysis

MJ Mahoney

7. Determine OATnavCOR

OATnavCOR is defined as the temperature correction which must be ADDED to the measured outside air temperature (Tnav) to make it agree with radiosondes (Traob) that the aircraft flew close to; that is, OATnavCOR = Traob - Tnav. OATnavCOR is positive if Tnav is colder than radiosondes, and negative if Tnav is warmer than radiosondes.

Generally, it is unlikely that a radiosonde launch site will be flown by when the sonde is launched, and in any event, it could take up to 2 hours for a sonde to reach maximum altitude, so temporal coincidence is not ever the case. Therefore, when we use sondes to calibrate the flight level OAT, we must interpolate temporally between two soundings. In addition, we examine the two earlier and later soundings to get a sense of the temporal variability of the air. If it is substantial, that particular site cannot be used for temperature calibration. In addition to the temporal variability, we must also be concerned about the presence of spatial temperature gradients when the aircraft does not fly close to a particular sonde launch site.
cTbiasWrange.GIF
Figure 1. Temperature bias with range during the SOLVE-2 campaign.

This is illustrated in Figure 1 for the SOLVE-2 flights on January 12, 16, 19, and 21, 2003. To generate this figure, we compared the temperature on level flight legs at 35, 39 and 41 kft over distances ranging from 0 to 220 km. Note that the error bars increase with distance both because of spatial gradients, but also because there are fewer samples and hence poorer statistics. It is evident that at FL390 an absolute bias of ~0.5 K can be expected on average over a distance of 225 km with ~0.5 K standard deviation (white trace and error bars in Figure 1).

If the aircraft does not fly close to a radiosonde launch site, we should perform a spatial interpolation between two sound launch sites (after the temporal interpolation at each of the sites).  RAOBman allows this to be done manually, but this is rather tedious. The temporal interpolation can also be done manually, but because it was also tedious, it has been completely automated. Now dozens of temporal comparisons between MTP retrievals and radiosondes can be performed in minutes.

Although determination of OATnavCOR only requires comparison of measured flight level temperatures with nearby radiosonde temperatures at flight level, we in fact compare the entire retrieved altitude temperature profile (ATP) to the radiosonde temperature profile. The reason for doing this is to be able to evaluate the accuracy of the retrievals above or below flight level. It is very important in this process to be completely objective in selecting which radiosondes should be used. The question of objectivity will be discussed further below.

To perform the RAOB and ATP comparison, a number of files are used:

images/RAOBman_Blend.png
Figure 2. The RAOBman Blend tab.

Before proceeding with the comparison we need to specify a number of options on the RAOBman Blend tab, which is shown in Figure 2. These options are all in the lower right side of the tab and default values are normally "safe" to use.
Once these options have been set, all that remains is to depress the Import RAOBs button. You will then see information appear on the remainder of the Blend tab - information that previously had to be manually transterred or entered from other places.

R34_20050120VBG.PNG
Figure 3. A RAOB/ATP comparison from the January 20, 2005, PAVE fligth near Vandenburg AFB, CA.

Figure 3 shows a file named R34_20050120VBG.PNG from the PAVE mission folder PNG subdirectory. The R34 at the beginning of the filename indicates that this was the 34th comparison made for this campaign, and is noteworthy in that there is substantial temperature variability near the tropopause between these soundings (yellow) which were 24 hours apart. (Note that if there were no missing or burst soundings, they should always be <12 hours apart during comparisons.) Nevertheless, the retrieved ATP (cyan) is in excellent agreement with the temporally interpolated sounding (white). The horizontal white bar is the average aircraft flight level at the time of the flyby. The retrieved ATP does have a ~5 K cold bias in the troposphere, but this was uncalibrated preliminary data. As mentioned earlier, comparisons near the tropopause should be avoided if at all possible. It is possible to filter these out later in the OATnavCOR determination.

RAOBcomparison1.png
Figure 4. The RAOBcomparison.txt file imported into an Excel spreadsheet.

As indicated above, the output from the comparison is written into a text file named RAOBcomparison.txt in the mission folder. This file is tab-delimited and readily imported into an Excel spreadsheet as shown in Figure 4. Before proceeding further, find the RAOBcomparisonMISSION.xls file from the last field campaign to use as a template. Using the Excel File menu, Open it and then Save As in the current mission folder, in our example, C:\MTP\Data\DC8\PAVE\.

RAOBcomparison2.png
Figure 5. The RAOBcomparisonPAVE.xls window.

Copy the data in the RAOBcomparison.txt window and paste it at the location of the yellow square in the newly created RAOBcomparisonPAVE.xls window H  tab as shown in Figure 5. Only a tiny fraction of the data is visibile in this screenshot. We will show the rest of it below.

The information on line 12 and below is as follows:
RAOBcomparison1b.png
Figure 6. The RAOBcomparisonPAVE.xls window extending beyond Figure 5. and showing the temperature comparisons between -10 and 0.5 km with respect to flight level. It actually cover -10 to +10 km with respect to flight level if all could be shown.

The information show in Figure 6 is the continuation to the right of what was shown in Figure 5. When we first began doing these comparisons, they were done in 500 m steps over a range of altitudes appropriate for the aircraft. To do this, both the RAOBs and ATPs were interpolated to a fixed grid of altitudes. The difficulty with this approach is that it does not take account of the fact that the comparisons are made with the aircraft possibly flying at different altitudes. This would be fine if the performance was not pressure altitude dependent, but it is. For the sake of argument, assume that you have a fixed altitude grid on which you superimpose measurements made when the plane is flying at FL330 and FL410. Clearly the retrieval at 41, 000 feet is not going to be as good when flying at 33,000 feet, as would be if the plane was actually at 41,000 feet. As a result the performance is degraded the broader the range of altitudes at which comparisons are made.

It is better to do the comparisons relative to flight level. The works reasonably well as long as there are not huge optical depth differences. For example, if a given flight had a boundary layer run when most of the flight was at much higher cruise altitudes, then the boundary layer run's performance should be handled separately. This is what is now shown in Figure 6. We cover the range -10 km to +10 km in 0.5 km steps starting at -10 km in column W, reaching flight level in column AQ, and +10 km in column BK (not visible). Rows 13 downward simply report the difference between the RAOB and MTP temperature profiles relative to flight level.

When the comparisons in Figure 5 and 6 were made, there were originally 86 soundings available before additional editting criteria were applied. In this case we simple deleted all soundings that had temperature differences >3 K. This removed 9 of the soundings, leaving 77. This number is entered in the cell F10, which is shown in red to remind you to do it. If you don't do this, the accuracy statistics will be incorrect.

The really interesting results appear in the top 10 rows of the spreadsheet. Referring to Figure 5, they show the following parameters starting with row 4:
The bottom line in doing all this of course is the value of OATnavCOR, which can be found in Figure 5 cell I2 to be -0.77 K. When the data is reprocessed using OATnavCOR = -0.77, and all the above steps repeated, the result in cell I2 should be 0.0 K. This should be done to verify that the sign of the correction is correct.

rTrange.GIF
Figure 7.

rTthreshold.GIF
Figure 8.


Previous  |  Next  |  Index  |