Getting to this point has been a tedious process, but it has taught me some interesting things about the quality of the outside air temperature (OAT) measured on the DC8. There are two such measurements: one from the air data computer (ADC), and one which is calculated by DADS using their own total air temperature (TAT) probe and Mach number from the ADC. More on this later. The PEM Tropics B flights were particularly amenable to this analysis because the frequent, large altitude excursions of the DC8 allowed OAT-derived profiles of the temperature field to be determined; these in turn could be compared to RAOBs, and if the agreement was good, the RAOBs could be used to determine the temperature field above the aircraft. In addition, the repeatability of the OAT-derived temperature field gave an useful indication of the variability of the temperature field throughout the flight. Given RAOBs which adequately represent the temperature field over the course of each flight, it is then a simple matter to calculate a number density profile using the ideal gas law and taking into account effect of water vaopr of the mean molecular weight of the atmosphere. The number density profile is important to interpreting concentration of ozone as remotely measured by the LaRC/DIAL lidar.
I will now present two charts for each of these five flights which lacked reliable MTP data: the first chart shows one or two average RAOB profiles compared to the OAT-derived temperature field measured at the aircraft, and the other shows the variation of OAT and geopotential height with time (and in one case, longitude). Either two or three RAOBs were used to calculate the average profile, one before and after the flight, and sometimes one during the flight. Generally, two RAOB sites were used as mandated by the behaviour of the OAT-derived temperature field. On only one flight (19990309) was a single RAOB site sufficient to represent the entire flight. On the temperature profile plots (the left hand chart), the aircraft OAT is the dark blue trace, and markers on this trace (generally not legible on this scale) are at 1 ks UT intervals. Click each image to see plot 16X larger.
In trying to assess when to transition between RAOB sites, I use a rule
of thumb ("MJ's ROT") that states: if the temperature at 8 km
is 250 K or warmer, then the air is tropical. I have examined hundreds
of RAOBs from the PEM Tropics B deployment period, and this was always
the case. These results are discussed below.
Number |
Date |
|
|
|
|
|
|
|
|
||
|
|
||
|
|
||
|
|
|
19990317 - Click to Enlarge |
With a regression coefficient of 0.98, there can be no question that
one or both of these measurements has a problem! To understand this behaviour
further, I decided to use the 1999015 flight, which had take-off and landing
in Hilo, to compare the two DC8 measurements to the average of the three
Hilo RAOBs taken before, during and after the flight.
Although there is significant temperature variation over this 24 hour
period at the tropopause, the three RAOBs are virtually indistinguishable
between 4 and 11 km. Therefore, there was hope that this altitude region
might provide a useful region to compare against DC8 measurements. Although
a visual comparison would be easy (simply overlay the average RAOB between
4 and 11 km on the DC8 measurements during take-off and landing), I want
numerical results. To this end, I extracted from the 1-second data in the
PO-file, DC8 temperature measurements at pressure altitude steps of 250
meters from the ground to 12 km. I then interpolated the RAOB profiles
at these same altitudes so that a numerical comparison could be made. The
results are show in the following two figures; the right hand figure is
a zoom in to show more detail.
19990315 - DC8/RAOB Comparison from 0-12 km (Click to Enlarge) |
19990315 - DC8/RAOB Comparison from 4-10 km (Click to Enlarge) |
The error bars shown on these figures are the standard deviation at
each altitude for the temperature as measured by the three Hilo RAOBs.
Not very good statistics, but nevertheless it shows that near 6 km the
temperature variation is very small. Although not obvious on these charts,
the two DC8 temperatures agree best between 6 and 7 km, and yet this is
where the maximum difference (about 1.6 K) occurs between RAOBs and the
DC8 measurements. To make these differences a little clearer, the chart
below the shows difference between the DC8 take-off (up) and
landing (dn) measurements, and the average RAOB measrement. (I was careful
only use DC8 measurements during take-off and landing to minimize the impact
of non-co-location of the DC8 and RAOBs.) For reference I also included
on this chart, the previuosly determined variation of each DC8 OAT measurement
with pressure altitude. A 2 K peak difference is something to get to the
bottom of! For the record, when DC8 OAT has been used as a reference
to calibrate MTP data, we have subracted 0.8 K from the ADC OAT value to
bring it into agreement with careful RAOB comparisons. For the current
data set, I find that the average offset (OATadc-RAOB) is 0.7 K, in agreement
with our earlier determinations. What is different however is that the
RMS on this value is also 0.7 K, because of the strange dependence on altitude.
It will be interesting to examine more flights to see if the conclusions
hold up.
This is shown in the figure above, where the percentage difference between each RAOB-derived number density profile and the model atmosphere number density profile used by LaRC/DIAL is plotted. Clearly there are large (nearly 10%) differences from the default model atmosphere, so using real measurements is important to more accurate determinations of ozone altitude and concentration. This flight is very representative of all the flights, with the exception that the temperature variations at the tropopause (12 km) are much larger than for all the other flights. So this flight is really a worst case. Since number density determinations near the tropopause are generally suspect in any case (because of the likely temperature variations), it can be concluded that the variation of individual RAOB-derived number density profiles with respect to the average is less than 1 %, and that this provides a significant (up to 10%) improvement over the default model atmosphere number density profile.
OAT/RAOB Comparison |
OAT [UT] for 19990320 |
NSTU RAOB/MTP Comparison |
The Bottom Line on the usefulness of RAOB in determining number density
for PEM Tropics B flights is shown in the left most figure above. Except
for a cloud inversion near Fiji at take off and landing, the RAOB from
Fiji (NFFN) does an excellent job of reproducing the OAT, as does the RAOB
from Pago Pago (NSTU). You will note that the RAOB is consistently BELOW
the aircraft OAT measurement, in agreement with the fact that we know the
aircraft OAT is on average 0.8 too warm. The middle figure shows that the
OAT is consistently 250 K at 8 km, which by MJ's
ROT says that we are in tropical air throughout the flight. Finally
the right most figure compares the MTP temperature profile (cyan)
to Pago Pago soundings before, during and after this flight (yellow). The
agreement with the Fiji soundings is better. Given that the aircraft is
near 10 km, or about 6 km from the tropopause, the MTP does a remarkably
good job of retrieving the tropopause altitude. Tropical tropopauses are
generally very "sharp" (like this one), and the poor old MTP just can't
do much better than this given the degraded vertical resolution of the
instrument this far from the tropopause. The MTP in this situation might
under estimate the tropopause temperature by as much at 10 K or more. For
this particular case, the difference between the RAOB number density determination
and the LaRC/DIAL model atmosphere was -5% at the tropopause and largest
(almost 10%) at 12 km, suggesting that the model is too warm (that is,
the RAOB on this day was colder, producing a higher number density). It
is hard to reconcile the 20% difference near the tropopause which the ECMWF
model sees (compared to the DIAL model atmosphere and MTP) with this data.
Also, note the variability of the three Pago Pago soundings near the tropopause
over 24 hours. This variability is typical and will always make number
density determinations near the tropopause suspect. For the period of PEM
Tropics B, a typical tropical RAOB site showed an RMS variability of the
tropopause temperature of about 3 K. Finally, this is basically raw MTP
data. The instrument has not yet been recalibrated since it was returned
to the aircraft in Fiji. In addition the final retrievals will use retrieval
coefficients from the period that we were in the field, and this should
improve the quality of the MTP retrievals.
Figure 1 |
Figure 2 |
Figure 3 |
Radiosondes for the period March 1, 1999 to April 18, 1999, which were launched from sites in the Western and Central Pacific and North America were used to study the variability of soundings with latitude, and with altitude, over this period of time. The initial motivation for this work was to use soundings from them PEM Tropics B mission period to calculate new retrieval coefficients so that the MTP retrievals could be refined. Figure 2 shows the temporal average sounding for 15 RAOB sites from Tahiti to British Columbia, while Figure 2 is a zoom in on this data to show the behaviour near the tropopause more clearly. The behaviour with latitude is basically as would be expected. Note that the two Hawaiian site, Hilo (ITO) and Lihue (LIH), have been binned into soundings with tropopauses above 16 km and those with tropopauses below 16 km. During this period about half the soundings were in each bin, with the <16 km trops being closer to 12 km, reflecting the influence of mid-latitude air on the sounding.
Figure 3 illustrates the RMS variability of the temperature with altitude at each RAOB station over this one and one half month period. Note that West Coast sites such as San Diego, CA (NKX), Medford, Oregon (MFR) and Vancouver, British Columbia (YZT) show by far the most variability below the tropopause (about 12 km), reaching a peak value of 7 K. This is obviously associated with storm systems coming and going. Some tropical sites on the other hand, such as Nadi, Fiji (NFFN), and Ponape Island (PTPN), show less than 1 K variability up to 12 km, reaching their maximum variability (about 3 K) at and above the tropopause (> 16 km).
Finally, referring to Figure 1 again, we note the origin of MJ's
ROT (Rule of Thumb): if the temperature is >250 K at 8 km, then
the air is tropical. Satisfying this criteria are four sites: Agana
NAS (PGUM, latitude= 15N), Ponape Island (PTPN, latitude= 7N ), Marshall
Islands (PKMJ, latitude= 7N), and Kwajulein (PKWA, latitude= 7N); Fiji
(NFFN, latitude= 18S), New Caldonia (NWWN, latitude= 22S), and Tahiti (NTAA,
latitude= 18S) also satisfy this criteria, but show cooler tropopause temperatures.