The following describes my derivation of "window correction tables" for the MTP/WB57 ACCENT Mission.
ACCENT consisted of two deployemnts, which I shall refer to as ACCENT/Spring (990326-990423) and ACCENT/Fall (990903-990924). I have determined that the WCT differs between the two deployments. Hence, I shall be descriibng two WCT determinations.
Flights WB990415 and WB990416 have been used to determine the Spring WCT. They provide a total of 9 RAOB/MTP comparisons (6 plus 3). The combined data are presented inthe following two figures.
Figure 1. Spring deployment's Channel 1 WCT.
Figure 2. Spring deployment's Channel 2 WCT.
The formal statistical SE for the average trace is approximately 0.14 K, for both channels, for all scan angles at and above the horizon. For the four "below horizon" scan angles the average trace has SE = 0.33 K, for both channels.
Flights WB990903 and WB990908 have been used to determine the Spring WCT. They provide a total of 12 RAOB/MTP comparisons (7 plus 5). The combined data are presented inthe following two figures.
Figure 3. Fall deployment's Channel 1 WCT.
Figure 4. Fall deployment's Channel 1 WCT.
The formal statistical SE for the average trace is approximately 0.16 K and 0.15, for channels 1 and 2 respectively, for all scan angles.
Figure 4 exhibits a good example of the need for relying somewhat upon the "median" of the individual RAOB/MTP comparison traces. At the upper elevations there seems to be one comparison trace that is unusually high, and can almost be thought of as an "outlier." Since outliers can be produced by a violation of our assumption that the temperature field varies linearly over time between the 12-hour RAOB times, thus invalidating the interpolated RAOB profile for use with the aircraft measurement, we should be prepared to reduce the influence of outliers somehow. A median analysis is well-suited for this situation. But since I am prepared to reject "blatant" outliers it would be unfair to simply use the median results only. My compromise is to average the median trace and the average trace. The following four figures show the results of doing this.
Figure 5. Spring deployment's Channel 1 WCT using the median/average analysis. The "thick dashed red" trace is the average of the individual traces, and the "thick solid black" trace is the average of the average trace and the median trace (not shown).
Figure 6. Spring deployment's Channel 2 WCT using the median/average analysis.
Figure 7. Fall deployment's Channel 1 WCT using the median/average analysis.
Figure 8. Fall deployment's Channel 2 WCT using the median/average analysis.
The differences are small, but to my eye the "thick solid black" traces do a better job of representing the individual traces than either the "thick dashed red" traces or the "median" traces (not shown).
Visual inspection of the Spring and Fall figures (for each channel separately) shows a difference. This is shown quite clearly in the following two figures.
Figure 9. Channel #1's Spring (red) and Fall (blue) WCT, with SE limits (dotted).
Figure 10. Channel #2's Spring (red) and Fall (blue) WCT, with SE limits (dotted).
It is clear that there are statistically-significant differences between the Spring and Fall WCTs, for both channels. It will therefore be necessary to use different WCTs for these two deployments.
Final Recommended WCTs
I recommend using the WCT values in the following table for the ACCENT
|Sequence #||Elevation Angle [deg]||SPRING Deployment Ch #1||SPRING Deployment Ch #2||FALL Deployment Ch #1||FALL Deployment #2|
I think it is significant that the WCT traces have such a high degree of autocorrelation! Not only does this suggest that the solutions are correct, but it has implications for understanding the physical nature of whatever process is causing WCT to be non-zero. If the principal cause for non-zero WCT is a leaky LO that reflects off the window surfaces, then the auto-correlation might be accounted for by a condition in which the two-way path from the horn throat to the window and back varies by amounts that are small compared with a wavelength. The significant change from Spring to Fall could then be accounted for by assuming that the physical mounting location during the Fall deployment was a significant fraction of a wavelength different from its Spring location. If window absorption is the principal cause for non-zero WCT, then the auto-correlated behavior could be accounted for, but it would be difficult to explain the Spring Fall difference. Someday we'll understand what causes WCT to be non-zero, and I hope I live to see that day!
This site opened: December 8, 2000. Last Update: December 8, 2000