The Airborne Microwave Temperature Profiler (MTP)
Michael J. Mahoney, Bruce L. Gary and Richard F. Denning
Jet Propulsion Laboratory, California Institute of Technology
Pasadena, CA
Abstract
Airborne Microwave Temperature Profilers (MTPs) built at
the California Institute of Technology's Jet Propulsion Laboratory (JPL)
have been involved in atmospheric research for a quarter of a century. These
instruments use measurements of the natural microwave thermal emission
from oxygen molecules in the earth's atmosphere to retrieve a vertical
profile of the temperature field, T(z), along an aircraft's flight track.
It is the intent of this paper to describe three new instruments currently
flying on several airborne platforms, to explain issues related to their
performance, and to show some examples of results obtained using the new
instruments.
I. Introduction
The first airborne Microwave Temperature Profiler (MTP) flew aboard
a NASA CV-990 in January 1979 (Gary, 1981). This work, funded by NASA's
Aviation Safety program, was initially directed to solving the Clear Air
Turbulence (CAT) problem. It was a concept demonstration, which used
the engineering model of the Scanning Microwave Spectrometer (SCAMS) spacecraft
instrument, built for the series of Nimbus weather satellites. In
1981 the first MTP designed specifically for airborne use flew aboard
the NASA C-141, the former Kuiper Airborne Observatory (Gary, 1984). In
the mid-1980's the ER-2 high-altitude research aircraft, then at the NASA/Ames
Research Center, were taking on a greater role in atmospheric research.
Our group at the Jet Propulsion Laboratory (JPL) was asked to build a MTP
(Denning, 1989) for the ER-2 to support the Stratospheric-Troposheric Exchange
Project (STEP), to be based in Darwin, Australia in January 1987. This
MTP began flying in 1986. Then, in August, 1986, the recent discovery
of an "ozone hole" appearing over Antarctica every October prompted the
formation of a NASA program to study the ozone hole with NASA ER-2 and DC-8
research aircraft only 6 months after STEP. During this highly publicized
Airborne Antarctic Ozone Experiment (AAOE), based in Punta Arenas, Chile,
the MTP discovered that immense vertical displacements of air occurred
over Antarctic mountains, that these mountain waves extended into the stratosphere
and caused air parcels to undergo large temperature extremes during their
5-day circuit around Antarctica. Cold temperatures such as those produced
by the mountain waves would trigger the formation of polar stratospheric
clouds (PSCs), which is a key step in producing an "ozone hole" (Gary, 1989).
Since these early ER-2 flights, three new MTPs have been built and
currently fly on a number of research aircraft. They are referred
to by the names MTP/ER2, MTP/DC8, and MTP/WB57, which reflects the names
of the NASA research aircraft on which they most often fly. The original
DC-8 MTP was built to participate in the Airborne Arctic Science Experiment
II (AASE-II) campaign to study northern-hemisphere ozone depletion. This
instrument was upgraded in 2002 so that the entire instrument is now located
on a DC-8 window. A new ER-2 MTP was built in 1995 to reduce its weight and
free up space for more instruments on the Stratospheric Tracers of Atmospheric
Transport (STRAT) campaign. This instrument has also flown on the NASA WB-57F
since 1998. Finally, a new instrument was built in 2002 for the WB-57F for
the Cirrus Regional Study of Tropical Anvils and Cirrus Layers - Florida
Area Cirrus Experiment (CRYSTAL-FACE) campaign. All three intruments are
electronically identical, but the MTP/DC8 is packaged in a single box, while
from the other two are packaged in two boxes: the Sensor Unit and the Data
Unit.
To date MTPs have flown on 34 research missions, accumulating 678 flights
and 3826 flight hours with nearly flawless performance. The most recent
MTP platforms and campaigns have been:
- NASA DC-8 and ER-2 during the SAGE III Ozone Loss and Validation
Experiment (SOLVE, 1999-2000),
- NASA WB-57F during the Atmospheric Chemistry of Combustion
Emissions Near the Tropopause campaigns (ACCENT I and II, 1999 and 2000),
- NCAR Electra during the Texas Air Quality Study 2000 (TexAQS
2000, 2000),
- NASA DC-8 and ER-2 during the Fourth Convection and Moisture
Experiment (CAMEX-4, 2001),
- NASA ER-2 and WB-57F during the Cirrus Regional Study of
Tropical Anvils and Cirrus Layers - Florida Area Cirrus Experiment (CRYSTAL-FACE,
2002), and most recently
- NASA DC-8 during the second SOLVE campaign (SOLVE-2), and the
Myashischev Design Bureau's M-55 "Geophysica" during the European Union's
Polar Stratospheric Cloud and Lee Wave Experiment (EUPLEX) and the ENVISAT
Validation campaigns.
The last MTP instrument paper (Denning et al., 1989) describes an instrument
that is no longer in use (the original MTP/ER2). It is the purpose of this
paper to describe the three new instruments.
II. Principles of Operation
A MTP is a passive microwave radiometer that measures the brightness
temperature of the natural thermal emission from (primarily) oxygen molecules
in the earth's atmosphere along the instrument's line of sight. As shown
in the figure to the right, this is done by scanning a mirror canted 45-degrees
to a corrugated feed horn. The scans are made from near zenith to near nadir
(generally in the flight direction) so that radiation from specific elevation
angles with respect to the horizon is reflected into the feed horn.
The strength of this radiation can be described by an atmospheric microwave
opacity model (P.W. Rosenkranz, 1993), which includes the effects of oxygen,
nitrogen, water vapor and water liquid. Except in heavy rain, oxygen is the
dominant contributor to the microwave emission; the oxygen absorption coefficient
is a function of pressure, temperature (and hence oxygen density) and,
to a small degree, water vapor density. (It is important to realize that
the MTP retrieval process described below only works because oxygen is a
well-mixed species in the atmosphere.) There are 40 spectral lines in the
opacity model ranging from 51GHz to 834 GHz, with the most important lines
being between 51 GHz and 67 GHz (wavelength of ~0.5 cm). Because any oxygen
emission is absorbed in proportion to its distance from the radiometer,
the integrated emission can be represented by a weighting function which
characterizes the weighted mean distance of the emission. In the special
case where the temperature lapses at a constant rate with distance, it
is easy to show that the brightness temperature measured by the radiometer
is exactly equal to the physical temperature at the e-folding distance for
the absorption. More complicated situations require more detailed treatment
(Gary, 1989).
All MTPs to date have used double-sideband receivers, meaning that
a local oscillator (LO) causes an intermediate frequency (IF) receiver
to be responsive to radio frequency radiation within two bands -- one above
the LO and one below the LO -- that fall within the IF pass band. At each
scan-mirror viewing position a local oscillator (LO) is sequenced through
two or more frequencies. Since a double-sideband receiver is used, the
LO is generally located near the "valley" between two spectral lines, so
that the upper and lower sidebands are located near the spectral line peaks
to ensure the maximum absorption. This is especially important at high
altitudes where "transparency" corrections become important if the lines
are too "optically thin." When choosing the pair of lines used, it
is important to select pairs that have similar absorption when weighted
across the IF pass band. If this is not done, the two sidebands will
have unequal absorption, and the retrievals will be degraded because the
effective look distance in each sideband will be different.
Because each LO frequency has a different effective viewing distance,
the MTP is able to "see" to different distances by changing frequency.
This is the way all satellite sounders and other airborne radiometers operate,
which limits them to looking down, and in the case of aircraft instruments,
not being able to make measurements near flight level. Because the atmospheric
opacity is temperature and pressure dependent, and therefore changes
with altitude, different effective viewing distances are also achieved
by scanning in elevation angle. If the scanning is done so that the applicable
altitudes (that is, the effective viewing distance times the sine of the
elevation angle) at different frequencies and elevation angles are the
same, then inter-frequency calibration can also be done, which improves
the quality of the retrieved profiles. For a three-frequency radiometer
with 10 elevation angles, each ~10-second observing cycle produces a set
of 30 brightness temperatures (or observables), which can converted by a
linear (or other) retrieval algorithm to a profile of air temperature versus
altitude, T(z).
Radiometric calibration is generally performed using the outside air
temperature (OAT) and an ambient reference target (see below the scan mirror
in the above figure) to determine the instrument gain. The gain can also
be determined from noise diode measurements, or by a gain equation based
on the temperature of sensitive receiver components. Complete calibration
of the system must also include "window corrections," which correct for
changing antenna spillover with scan angle, as well as other effects. This
requires tedious analysis and comparison with radiosondes near the aircraft
flight path. This is probably the most important single factor contributing
to reliable calibration. For stable MTPs, like that on the DC8, such calibrations
appear to be reliable for many years.
On an airborne platform, pitch and roll are needed in real time so
that the scanning-mirror angular position can be corrected for aircraft
attitude, allowing measurements at specific elevation angles with
respect to the horizon. This must be done because doing otherwise would
make the retrieval of temperature profiles computationally intractable.
Pressure altitude is also needed because gaseous absorption depends on pressure.
Finally, it is helpful, but not necessary, to also have outside air temperature
since this simplifies the retrieval process. In practice, these and many
more avionics parameters (e.g., universal time, latitude, longitude) are
available to facilitate data analysis and interpretation.
However, additional improvements are possible that overcome some of
the Backus-Gilbert limitations. "Statistical retrievals" add a degree
of sophistication that allow for even further improvements in the ability
to convert MTP observables to T(z) profiles. The use of statistical
retrievals is an immense field, and requires a great deal of user skill
to avoid pitfalls while achieving significant improvements in retrievals.
In this appendix only a rudimentary treatment will be given. Statistical
retrieval can be thought of in terms of linear regression theory.
Imagine a procedure in which a set of hypothetical T(z) are used to calculate
what a perfect MTP would measure for each T(z). With a set of N T(z)s
and M sets of MTP measurements, there exists a NxM matrix of "computer measurements."
If a mathematician were faced with the task of trying to estimate temperature
at a specific altitude (with respect to the MTP), and he is given the
NxM matrix, he would be wise to perform a least squares analysis solving
for the desired temperature (as the dependent variable) using the NxM measurements
(as independent variables). The LS solution will yield a set of
M+1 coefficients for use in determining the temperature at this one specific
altitude. The same procedure can be repeated for the other N-1 altitudes,
after which there will be a set of Nx(M+1) coefficients. These coefficients
are also called "retrieval coefficients." The set of retrieval coefficients
have the added value that they can be used with actual MTP measurements
for producing a solution for T(z). The resultant T(z) is likely to
provide more structure (have better altitude resolution) than provided
by any of the previously described procedures.
Pitfalls abound when using statistical retrievals. The principal
consideration when preparing retrieval coefficients for use in a future
field campaign is to anticipate the types of T(z) profiles that are likely
to be encountered. Fortunately, there is a way to know if unusual,
or unanticipated T(z) situations are being encountered. The "safety
check" procedure makes use of the set of observables created during the
process of deriving retrieval coefficients from the set of hypothetical T(z)
profiles, which we'll call OBarc (the "observable" vector matrix, having
M elements, derived from the T(z) archive). Actually, OBarc can be
associated with a distribution of M values for each "observable" for each
of the N observables. Imagine an "observable space" (N-dimensional)
where it is possible to assign a probability of encountering a T(z) that
places one at that observable space locus. Some observable space locations
may never be "visited" during the retrieval coefficient procedure, and would
therefore be associated with a zero probability. Now, when using actual
MTP measurements if one of these zero probability locations in observable
space is encountered, this event should "ring a bell" that warns the user
that something is awry. Either the data is bad (interference, for
example), or unexpected T(z) conditions are being encountered. When
the latter occurs, it is time to reconsider the appropriateness of the T(z)
set that comprised the simulation archive during the retrieval coefficient
preparation process.
This example illustrates that although statistical retrievals can
be powerful, they can also be misused. It also illustrates that
there are methods for detecting when a misuse is occuring, and there are
responses for recovering. Those practiced in the art of Bayesian
probability theory will recognize that stastistical retrievals permit making
use of "external information" that appropriately enhance the information
content of the retrieval product. This represents an important endorsement
of statistical retrievals as a valid approach for converting observations
to retrievals of properties of the real-world. In the right hands,
statistical retrievals can be an important tool for converting MTP measurements
to T(z) profiles.
III. Hardware Description
There are currently three airborne MTP instruments in service, of two different
architectures. Two of these, known as MTP-X., were designed
originally for the ER-2 aircraft and have subsequently been adapted for
the NASA WB-57 and the Russian M-55 aircraft. The WB-57 has a ‘back-seat’
instrument manager, and so an analysis computer was added to provide real-time
profiles to guide the aircraft along certain temperature structures such
as the tropopause or a particular theta surface. These MTP’s are autonomous..
The second and newest instrument is the ‘Phoenix’ version, designed
for the DC-8, with real-time profiles distributed to other experimenters
during flight. This instrument is fully automated, and also may be controlled
over the aircraft Ethernet if an MTP experimenter is present. The Phoenix
MTP was used first on the SOLVE-II mission in Sweden in January of 2003.
The MTP’s are heterodyne double-sideband radiometers with an intermediate
frequency (IF) bandpass tailored to maximize the received energy from
the oxygen emission lines used, and to minimize the contribution from other
sources. Both the MTP-X and Phoenix radiometers incorporate the same receiver
module that combines most of the low-level functions into one block, and
tunable frequency synthesizers to provide the local oscillator signal to
the receiver mixer.
Figure 2. A block diagram of the MTP-X sensor unit and data unit.
A. The MTP-X
As is shown in Figure 2, the MTP-X instruments are comprised of two
major assemblies, the Sensor Unit (SU), and the Data Unit (DU). Radiometer
functions are implemented in the SU, while the DU includes the control and
data acquisition functions. Aircraft installation is arranged so that the
Sensor Unit scanning mirror has a clear view from near Zenith to near Nadir
during its scan cycle. The Data Unit may be mounted some distance
from the SU. For example, when used with the WB57 Spear Pod the Data Unit
is installed near the center of the pod about 3 meters aft of the Sensor,
while on the ER-2 the boxes are within a few inches of each other.
The SU consists of a scanning parabolic mirror feeding a corrugated feed
horn, which passes the incoming radiation to the receiver inside the SU
box. The scanning system is protected from the air stream by a radome and
fairing that is tailored to the particular mounting arrangement on each
aircraft. Immediately aft of the scanning mirror is a microwave black body
target positioned so that the viewing beam of the instrument falls on this
reference target on each rotation. Inside the SU, the signal enters the
receiver module through a directional coupler, allowing a calibrated temperature
increment from a microwave noise diode to be added to the thermal radiation
from the antenna as part of a gain calibration algorithm.
Figure 3. A block diagram of the MTP-X receiver.
The receiver module shown in Figure 3 is a complete double-sideband heterodyne
receiver, including biased mixer, IF amplifiers, IF band pass filter, detector
and DC amplifier, made by Spacek Labs of Santa Barbara, California. These
units typically have a (double side-band) Noise Figure of less than 5 dB,
and produce a fairly high-level DC output voltage, (approximately 2 to 3
Volts with a 300K target), which is proportional to the total thermal emission
seen by the antenna plus the receiver’s system noise temperature. The sensitivity
or gain at this point is about 3mV/K.
A Voltage-to-Frequency Converter (VFC) then transforms the varying DC voltage
from the receiver into a train of digital pulses whose TTL output frequency
is proportional to its input voltage. This is the basic analog to digital
process in the radiometer.
Radiometers measure temperature, and the measurement and management of
the physical temperatures of the radiometer components is one key to the
successful operation of these devices. Of primary importance is knowledge
of the temperature of the reference target, which provides the ‘anchor’
point against which all other observations will be compared. The reference
target temperature is a ‘first-order’ variable in the radiometer equation.
A 500 ohm platinum RTD is embedded in the target, and its resistance is
converted to a voltage sent to the ADC, resulting in about 0.04 C resolution.
Since the properties of many of the radiometer components have strong temperature
dependencies, a proportionally controlled heater keeps the receiver and
associated components at about 42 C., and thermistors monitor the temperature
of individual devices in the Sensor Unit.
A PC-104 stack in the Data Unit includes a CPU module, digital I/O, ADC
and a bootable PCMCIA Flash drive that holds the DOS operating system, the
control program, and provides configuration settings and MTP output data storage.
In addition there are three other circuit boards in the Data Unit.
The Digital Board implements a programmable integration timer that gates
the VFC output of the Sensor Unit into a 16-Bit counter that performs the
actual integration. The timer can be set from 1 to 255 intervals of 1.67
mS each, for a total of 425 mS. Integration time is typically 200 mS for
all of the MTP's, but this system allows the possibility of optimizing the
relationship between scan cycle time and sample-to-sample noise.
A hardware watchdog timer is implemented on the Analog Board to provide
an indication to the pilot that a serious failure has occurred during flight.
An ‘MTP Fail’ signal is sent to the cockpit by a normally closed relay.
The fail timer holds the relay in the ‘not fail’ state for up to 90 seconds,
and is restarted by the control program at the successful completion of each
scan cycle. Any failure that prevents the scan or data acquisition from completing
will stop this reset and allow the relay to release, which alerts the pilot.
Once the relay is in the fail state, the only way of resetting it is for
the pilot to cycle MTP power off and on according to a pre-arranged procedure.
Also on the Analog Board is the signal conditioning and scaling for a vertical
accelerometer that is used to record statistics on the relationship between
the atmospheric temperature profile and turbulence. Other ADC channels on
the board are devoted to recording various power supply voltages and the
temperatures inside the Data Unit for housekeeping or diagnostic purposes.
Stepper motor control is handled by a stand-alone smart stepper control
board with RS-232 communication to the CPU. This card provides a ‘high-level’
command structure for the motor, and eases the real-time burden on the control
computer. For example, a command to move a specific number of steps may be
given and the CPU can do other tasks while the move is taking place. The computer
can check back later to see if the move was completed before starting an
integration or other scan-dependent task. All stepper parameters, such as
starting speed, acceleration and maximum speed can be set in software, allowing
great adaptability to changes in scan hardware or observing strategy.
MTP’s require knowledge of certain flight parameters, like aircraft pitch
and roll to adjust the scan pointing to the proper angles with respect to
the horizon. Several other values are recorded for use in data analysis,
including time, altitude, latitude and longitude. These come from the aircraft
Flight Data Recorder into the Data Unit via an RS-232 navigation data stream,
and are parsed and recorded in the MTP recorded data frame.
The last major function in the MTP Data Unit is the aircraft power and
control interface. Power is provided to MTP in two forms, the aircraft 28
VDC, which is used in MTP primarily for the stepper power, and as 120 VAC
400 Hz., from which all other DC voltages in the instrument are derived.
There are two different states of power for MTP, ‘Standby’ and ‘Operate’.
In Standby, the only systems operating are the temperature controllers, to
keep the instrument warm when it is not being used. In Operate, MTP boots
up and will continue to take and record data until turned off or back to
standby.
These switching tasks are done by a pair of relays, one AC and the other
DC, which are controlled by a command from the cockpit. The pilot’s switch
console typically would have one button that sends power to the MTP interface
and often some other instruments as well. This puts MTP into Standby mode.
Another button and lamp assembly puts MTP into Operate, switching the relays
in the Data Unit, and displaying the state of the Operate switch and the
MTP Fail signal.
Table I. MTP-X Weight and Power
Component
|
Weight (kg)
|
Power
|
Peak Current (A)
|
Sensor Unit
|
5.2
|
28 VDC
|
2.20
|
Data Unit
|
4.2
|
120 VAC (400 Hz)
|
0.84
|
Radome/Fairing
|
0.7 - 3.0
|
|
|
Cables
|
0.7 - 2.0
|
|
|
Total
|
10.8 - 14.4
|
|
|
Table I summarizes the MTP-X weight and power. Because of different fairing
sizes and cable runs between the SU and DU, the MTP weight can range from
10.8 to 14.4 kg. The average power is about 100 W, but is dependent on the
amount of heat required to maintain temperature control.
B. The Phoenix
MTP’s on the DC-8 have always been intended to provide real-time temperature
profiles for guidance and advice for other experiments on board, and so have
required sophisticated hardware and software for data analysis and display.
Previous models had separate control and analysis computers, with MTP data
coming from an MTP-X data system to the data reduction computer via RS-232.
There was little communication from the display computer to the data system,
and consequently it was very difficult to change the observing strategy in
flight.
With the Phoenix MTP, this limitation has been eliminated, and several other
features are now available. A ‘Panel PC’ computer running with Windows 2000
controls the operation of the radiometer, records the data, analyses and
displays the temperature profiles as they are taken. Between the Phoenix
and this computer is an RS-422 serial connection allowing the computer to
be located practically anywhere in the cabin. Since all control functions
are in one computer with access to and from the Ethernet, the entire scan
and tuning sequence is available to the MTP experimenter for modification
during flight. Parameters such as mounting offsets, integration time, frequencies
and scan angles are contained in a configuration file that may be changed
at any time. Other experimenters can obtain numerical data during flight
over the DC-8 Ethernet.
A VGA-to-NTSC video converter generates a composite video signal that is
distributed to color monitors throughout the cabin. If an MTP experimenter
is on board, he can connect a laptop computer to any available Ethernet port
to control the instrument or download data files, or even edit programs on
the Panel PC.
In the current configuration, the Phoenix MTP is mounted on an aluminum DC-8
window plate behind the Mission Manager’s console, and its computer is in
a rack on the left (forward) end of the console. The radiometer section of
the Phoenix MTP is nearly identical to that in the MTP-X; however, the data
acquisition functions have been implemented in the same enclosure rather
than a separate data unit. This was made possible by the use of several microcontrollers
and “smart” peripheral chips to replace many standard logic and analog integrated
circuits.
As an example of how the system is simplified in this way, consider the radiometer
temperature controller. This function on the previous DC-8 MTP required a
14-pin quad op-amp, 12 discrete parts (resistors and capacitors), and a solid-state
AC relay. In the Phoenix MTP, an 8-pin microcontroller (Microchip 12F675),
a resistor, capacitor and solid-state relay provides the same proportional
control function but adds an error integration term to reduce the error.
Even more striking is the integration timer/counter that reads the output
of the radiometer voltage-to-frequency converter. The standard-logic design
we had been using required nine 74HC DIP packages of 14 to 16 pins. The new
integrator is a single 18 pin DIP (Microchip 16F628), plus a three-lead ceramic
resonator. A similar improvement came from replacing the PC-104 ADC board
with two 20-pin chips (MAX186). Communication between the various functions
is over a Serial Peripheral Interface (SPI) bus, which requires only four
lines per device, three of which are shared among all the SPI circuits. Another
PC-104 board (Parallel I/O) was eliminated by this simplification.
In the Panel PC, the MTP control software starts upon boot-up of the Windows
2000 operating system. This stand-alone application sequences the radiometer
and records the data. At the same time, the analysis/display program starts
and will produce profiles as soon as data is available. The data transfer
between these two processes is handled by file sharing. If an MTP experimenter
is present, the panel PC may be mounted as a disk drive on the experimenter’s
laptop, making all data and programs controlling the radiometer available.
IV. Hardware Calibration and Alignment
MTP calibration involves several steps. Before and after each field
campaign LO frequencies are checked to ensure that they are within 10 MHz
of the desired frequency. The shape of the IF bandpass is also recorded
so that it can be modelled accurately when doing the forward radiative transfer
calculation to obtain retrieval coeffients. In addition, the calibration
of thermistors that provide engineering data are carefully calibrated. This
is most important for the reference target used in the gain calibration,
and the IF mixer temperature, which can be used as a gain proxy.
Because it is computationally intractable to calculate retrieval coefficients
for every possible elevation angle that the MTP might view, retrieval coefficients
are calculated for specific elevation angles with respect to the horizon.
Since the aircraft attitude is always changing, pitch and roll are required
in real time to correct the MTP pointing so that the proper elevation angles
are viewed. In addition, account must be taken of how the MTP is mounted
with respect to the aircraft. We have developed techniques that allow us
to easily measure the instrument pitch, roll, and yaw with respect to an aircraft.
We also have built a double, 3-axis gimbal apparatus that allows us to check
the MTP pointing algorithm in the lab.
Beamwidth effects
Pointing error simulations
Using pressure vs geometric height in abs coeff calc.
V. Gain Calibration
A. Flagging Bad Data
There are many steps involved in the MTP gain calibration. The first
is to read in the raw counts data recorded by the Data Unit and check for
possible interference. This is done by using an iterative algorithm that
compares a data sample for a particular frequency and elevation angle to
a box-car running average of its neighbors to see if it exceeds a specified
threshold. If it does, the data sample is flagged and the process repeated
without this sample. This is continued until no more data samples are flagged.
B. Temperature Calibration
MTP temperatures are generally calibrated by comparing the measured
horizon view brightness temperature to the temperature of radiosondes
that the aircraft flies near interpolated to flight level. Care must be
taken to ensure that the aircraft altitude or attitude is not changing, and
that there aren't significant spatial or temporal gradients in the atmosphere.
The latter is assessed by examing not only the radiosondes before and after
the flyby, but also the radiosondes the day before and the day after. In addition,
several MTP measurements are averaged to beat down the radiometric noise
(~0.4 K). At least 10-20 such comparisons are needed to achieve a reliable
result. Typically the population standard deviation for these comparisons
is ~1 K, and the standard error on the bias determination is ~0.2-0.3 K. It
should also be noted that these comparisons should be done at different altitudes
to avoid the possibility that the pressure altitude might be in error, in
which case the bias correction would be pressure altitude dependent. They
should also be done, if possible, at altitudes where the lapse rate is smallest
to minimize the impact of aircraft altitude excursions.
C. Brightness Temperature Fit
Once the MTP horizon temperature (Tmtp) has been calibrated against
radiosonde temperatures (Traob), the gain can be determined. If an
outside air temperature (OAT) is available from the aircraft navigation
system or some other instrument (Toat), it is easiest to use this OAT
to transfer the MTP temperature calibration against radiosondes by adding
a bias correction of (Traob - Toat) to Toat to obtain the corrected temperature
(Tcorr). The gain is then adjusted to minimize the quantity (Tmtp - Tcorr)
for level flight segments for each frequency, with each frequency being
weighted inversely to its optical depth. This weighting is done because
each frequency measures the temperature at a different distance in front
of the airplane. There are a number of other important considerations
that should be noted when the brightness temperature fit is performed. Since
the MTP measures the temperature ~1-2 km in front of the airplane, a "time-tag
offset" is applied to the MTP data in order to synchronize it with the OAT
measurement. In addition, important engineering temperatures such as the
reference target (Tref), mixer (Tmxr), and noise diode (Tnd) are either
led or lagged in time to account for thermal time constants related to how
a particular thermistor is mounted relative to the component temperature
that it is measuring.
The MTP has two basic methods of monitoring the gain: a gain equation
and a noise diode. The gain of the MTP depends largely on the temperature
of the mixer and IF amplifier. If the mixer is LO-starved, there can also
be a dependence on the temperature of the LO amplier and/or the synthesizer
temperature. We have also seen a dependence on pressure altitude, which must
be a surrogate for something else, such as the effectiveness of thermal conduction
versus radiation.Historically, we always used the gain equation to calculate
the gain because the noise diodes that we used displayed some hysteresis
in their gain. The new MTPs, however, have much better noise diodes and for
the past year we have used the noise diode to determine the gain. They are
a little better at determining the gain early in a flight when their are
large thermal gradients on the RF plate.
Once the gain (Gi) for each frequency in counts/K is calibrated
on the horizon view, it is possible to calculate the brightness temperature
(TBij) at any frequency (i) and elevation angle (j) by using
the relationship:
(1)
TBij = Tref + (Cij - Crefi)
/ Gi
where Cij are the counts for frequency i and elevation angle
j, and Crefi is the reference counts for frequency i. Obviously
when Cij = Crefi, TBij = Tref, as it should.
Note that Tref can be either the horizon view temperature or the reference
target temperature. In practice we prefer to use the horizon view temperature
since this minimizes the impact of gain variations. (The actual TBij
calculation that we perfom is a little more complicated than Eqn 1 as it
also accounts for the emissivity and reflectivity of the HDPE window that
measurements are made through.)
D. Window Corrections
Before brightness temperatures can be used in a retrieval one final
correction must be applied, the window correction. When measured brightness
temperatures near a radiosonde launch site are compared to the expected
brightness temperatures based on the radiosonde temperature and humidity
profile, small differences are seen that have a clear and repeatable dependence
on elevation angle and frequency. These differences are generally <1
K, but can become much larger for high altitude aircraft at the highest elevation
angle. The window correction appears to be very stable during a campaign,
but most often changes when the instrument is remounted. It is believed
that most of the window correction is due to beam sidelobes viewing different
thermal structure inside the fairing as the scan mirror moves. There may
also be a component due to LO leakage and reflection off the HDPE window.
post-mission RCs
Use WCT errors in RC calc
Explain cause of WCT due to sidelobes when there is a large temperature contrast,
note altitude dependence
E. Retrieval Coefficients
To convert the raw brightness temperatures measured by a MTP into temperature
profiles, it is necessary to calculate retrieval coefficients (RCs). A
decade ago, we would simply take hundreds of archived radiosondes (RAOBs)
and calculate a single set of RCs for a campaign. This would work well in
the tropics where there is not a large amount of temperature variability;
however, in the Arctic winter the formal retrieval errors are substantial
at moderate distances from flight level because of the huge dispersion in
the RAOB profiles over even a one month period. During the first SOLVE campaign
in the winter of 1999-2000, an attempt was made to minimize the formal retrieval
errors by separating the RAOBs into temperature bins at a nominal flight
altitude. This worked well for the ER-2, but was less successful for the
DC-8 because there is much more temperature structure near the much lower
DC-8 flight altitudes. For SOLVE-2 RC calculations RAOBs launched near the
DC-8 flight track were used as templates for selecting additional RAOBs
to calculate RCs. This worked very well. Although the MTP doesn’t measure
many photons beyond ~6 km, the range of accurate retrievals was much greater
than in the past when within several hundred kilometers of a RAOB site.
Nearly 50 sets of RCs were calculated to represent all the temperature profile
shapes encountered during SOLVE-2. With so many RC sets to choose from,
it was generally the case that the archive average observables of several
sets might match the measurements quite well. However, by using an information
content metric, it was always the case that, when near a radiosonde site,
the RCs based on the template sonde were selected.
OB a priori error refinement
F. The MTP Retrieval Algorithm
As discussed earlier, MTP retrievals have evolved from assigning a brightness
temperature to an applicable altitude, to a more versatile Backus-Gilbert
retrieval, and finally to a statistical retrieval. Over the years we have
made many refinements to help improve the accuracy and range of the retrievals.
The factors that determine how long it takes to calculate a set of retrieval
coefficients are the number of flight levels, frequencies, IF bandpass segments,
elevation angles, and radiosondes. For aircraft that fly at fixed flight
levels such as the NASA DC-8, interpolation errors can be minimized by ensuring
that an RC set is calculated for each flight level.
Further improvement can be achieved by using radiosondes (RAOBs) launched
near the aircrafts flight track as templates for selecting the several hundred
additional soundings needed to calculate a set of RCs. This ensures that
the actual atmospheric conditions are reflected in the RC sets, which might
total 50 or more for an Arctic winter campaign. This also produces much
better retrievals at larger distances above or below the aircraft where the
MTP does not receive many photons. The MTP retrievals have also been improved
by modifying the quality metric that compares MTP measurements to each RC
set of archive average observables to put more emphasis on the variation of
the shape of the measurements with elevation angle, and less emphasis on
the bias. In addition, we correct the bias contribution, which has a second
order effect on the oxygen absorption coefficient, by using a sensitivity
matrix to get into the temperature regime of the RC set which best matched
the shape of the MTP measurements with elevation angle. After performing the
retrieval, we then bias the profile back to measurement space
Figure 2 shows how the new method of selecting RCs to perform retrievals
can improve the accuracy of the retrieval at large distances from flight
level (white horizontal line at ~10 km) when near a radiosonde launch site.
The pink profiles are the MTP retrievals, and the yellow traces are the radiosondes
launched before and after the DC-8 flew near the island of Jan Mayen (ENJA)
on 2003.02.02. The left two panels show retrievals performed 144 km from
the island. For the upper panel retrieval, a RC set was used that was not
based on the ENJA soundings for this day. It does a poor job of matching
the soundings above 20 km. The lower panel, on the other hand, did use an
RC set based on the ENJA sounding. We believe that the warm bias above 18
km is real. The two right panels show retrievals performed 15 km from the
island. Even though the upper panel does not use a RC set based on the ENJA
soundings, it does an excellent job. The lower panel is almost identical
to the upper panel, but careful examination shows that it does a little bit
better at ~26 km. Since both retrievals near the island did so well, it supports
the suggestion that the ~ 2 K warmer temperature at ~20 km in the left, lower
panel is real. Such a change is consistent with expected temperature gradients
over 144 km.
MRI
V. Performance
Add Simulations
Include a priori rms error, WCT error, pointing error
Poorer M55 performance because of tropopause and bursting RAOBs in RC calculations
(which go to 50 km) only 1 RAOB in 25000 reached 40 km.
Figure 2. DC-8 MTP performance
|
Figure 3. Geophysica MTP performance
|
Figure 2 summarizes the result of comparing MTP temperature profiles
to the temperature profiles of 17 radiosonde launch sites that the DC-8
flew close to during the SOLVE-2 campaign. The average flight altitude for
these comparisons was 11.6 km, and the average distance to the radiosonde
launch sites was 84 km. The heavy black trace is the average bias of the
MTP temperatures compared to radiosondes, and the error bars represent
the standard error on the average bias. The dotted trace is the population
standard deviation for the 17 temperature comparisons, and the thin black
trace is an estimate of the retrieval error, which is calculated by removing
1 K in quadrature from the dotted trace to correct for the fact the the
MTP and radiosondes are not co-located. For level flight, the expected standard
deviation in flight level temperatures separated by 84 km is ~1 K; this is
due to real temperature gradients in the atmosphere. Based on these comparisons
and assuming an average flight level of 11.6 km, the retrieved MTP temperature
profiles have an accuracy of <1 K from 8.5 to 16.5 km, <2 K from 5
to 21 km, and <3 K from 4 to 26 km.
Figure 3 summarizes the result of comparing MTP temperature profiles
to the temperature profiles of 21 radiosondes (RAOBs) that the Geophysica
flew close to during the EUPLEX and ENVISAT Validation campaigns. The average
flight altitude for these comparisons was 17 km, with a population standard
deviation of 3 km. The average distance to the radiosonde launch sites was
117 km. The heavy black trace is the average bias of the MTP temperatures
compared to RAOBs, and the error bars are the standard error of the average
biases. The error bars are larger at the higher altitudes because the RAOBs
burst and fewer comparisons were possible. Note that near 12 km the population
standard deviation increases. This is because when flying at an average
altitude of 17 km, the MTP is not able to resolve sharp tropopause temperature
structure, or alternatively, there is significant variability in the tropopause
temperature. The dotted trace is the population standard deviation for the
21 comparisons, and the thin black trace is an estimate of the retrieval
error, which is arrived at by removing 1 K in quadrature from the dotted
trace to correct for the fact the the MTP and RAOBs are not co-located. For
level flight, the expected standard deviation in flight level temperatures
separated by 117 km is ~1 K; this is due to real temperature gradients in
the atmosphere. Based on these comparisons, and assuming an average flight
level of 17 km, the retrieved MTP temperature profiles have an accuracy of
<1 K from 13.5 to 20 km, <2 K from 13 to 21.5 K, and <3 K from 10
to 22 km.
Causes of poorer performance
- RAOBs do not go to 50 km, only 1 in 25000 made it to 40 km
- WCT error at high elevation angles.
VI. Data Products
Figure 4. DC-8 MTP tropopause data (black dots) from the TOTE/VOTE transit
flight on December 11, 1995 from Fairbanks, AK, to Barbers Point, HI. The
cross section is near 155 W longitude. The PV (solid contours) and U wind
(dashed contours) fields are from UKMO data. The MTP thermal tropopause
shows a large jump at 37 N near the subtropical jet, and a smaller jump at
54 N near the polar jet.
Figure 5. A color-coded temperature curtain from a WB-57 flight on May
6, 1998 during the WAM campaign. This was the first time that the real time
backseat MTP display was used to guide the aircraft with repect to the tropopause.
After climbing out, the WB-57 (black trace) descended to just above the
tropopause (white dots) at ~56 ks. It tracked the tropopause until ~59 ks
before climbing to much higher altitudes.
Figure 6. There were many occasions during the DC-8 SOLVE-2 flights
when the vortex edge was approached and MTP-derived stratospheric isentropes
cut across the tropopause and dipped several kilometers into the troposphere.
This occurred twice during the flight of 2003.01.14; this figure shows the
second of these edge encounters. The closest approach to the vortex edge
occurred at ~44 ks UT, when the DC-8 heading changed from W to NE. The DC-8
pressure altitude is shown in blue. As would be expected, the wind speed
increased to a maximum value of ~110 knots as the vortex edge was approached
(green trace with units on right hand ordinate scale). The MTP-derived isentropes
dropped several kilometers, consistent with the LaRC DIAL nadir ozone isopleths
superimposed on the figure. The black trace at the bottom of the figure
is the lidar ground return. The MTP tropopause is indicated by small red
triangles, and the black theta surfaces, or isentropes, range from 295 K
to 370 K in 5 K steps. This stratospheric intrusion teaches us some lessons.
The five temperature profiles across the top of the figure show how the MTP
tropopause weakened at 42.5 ks and 45.5 ks compared to deeper inside the
vortex; they also show a tropopause at 44.0 ks which is ~1 km higher than
inside the vortex, and several kilometers higher than the chemical tropopause
represented by the LaRC DIAL measurements. In these small figures, the abscissa
axis is temperature ranging from 180 to 260 K, the left ordinate axis is
pressure altitude ranging from 0 to 30 km, and the right ordinates axis
is pressure altitude ranging from 0 to 98 kft.
Figure 7. This isentrope altitude cross-section for the SOLVE DC-8 flight
on January 25, 2000, shows the largest mountain lee wave ever recorded
by a MTP. The isentrope surfaces range from 320 K to 360 K in 5 K steps.
These lee waves were seen as the DC-8 was flying from the west coast of
Finland (left) towards the west coast of Norway (right), crossing the Norwegian
Mountains shown at the bottom of the image. The time period for 58 to 62
ks UT covered a distance of ~700 km; the wind was blowing across the mountains
from west to east, or right to left in the image.The flight track was chosen
because large lee waves had been predicted by several models. At the time
of the encounter, the DC-8 was flying at 12.5 km, about 0.8 kilometers above
the tropopause. The wavelength of the lee waves was about 30 km, and they
were also observed at two other periods early in the flight.
MFA
Figure 8. The MTP temperature curtain observed during three transects
of Hurricane Humberto on September 24, 2001 by the NASA DC-8 research aircraft.
This was the first airborne measurement of a temperature anomaly associated
with the eye of a hurricane; the anomalies are at ~77.4, 82.8 and 88.1 ks.
The DC-8 flight altitude is shown as the solid black trace, and the tropopause
altitude is shown as the upper white dots. The temperature color scale
is to the right.
Mesoscale T Variability
VII. Future Directions
SSB
1 PS
1 Box
PCB
Baysian
Two loads
VIII. Acknowledgements
We would like to thank Dr. Mike Kurylo of NASA's Upper Atmosphere
Research Program for the support needed to build all the existing Microwave
Temperature Profilers, as well as Dr. Don Anderson of NASA's Radiation
Program for his support to build the WB-57 MTP Sensor Unit. Work performed
at the Jet Propulsion Laboratory, California Institute of Technology
was performed under contract with the National Atmospheric and Space Administration.
IX. References
Denning, R. F., S. L. Guidero, G. S. Parks and B. L. Gary, Instrument
Description of the Airborne Microwave Temperature Profiler, J. Geophys.
Res., 94, 16,757-16,765, 1989.
Gary, B. L., An Airborne Remote Sensor for the Avoidance of Clear
Air Turbulence, AIAA Pap., AIAA-81-0297, 1981.
Gary, B. L., Clear Air Turbulence Avoidance Using an Airborne Microwave
Radiometer, AIAA Pap., AIAA-84-0273, 1984.
Gary, B. L, Observational Results Using the Microwave Temperature
Profiler During the Airborne Antarctic Ozone Experiment, J. Geophys.
Res., 94, 11,223-11,231, 1989.
Liebe, H. J. et al, JQSRT V.48, pp.629-643 (1992).
P. W. Rosenkranz, Chapter 2 and Appendix in Atmospheric Remote
Sensing by Microwave Radiometry, M.A. Janssen, ed., Wiley, New York,
(1993).
Schwartz, M.. J., Ph.D. thesis, M.I.T. (1997).
[Back to Home Page]