DOCUMENTATION ON UCI PROCESSING OF NOAA/AOC P3 Aircraft

We have completed our processing of the slow-rate 1 Hz NOAA WP3D
aircraft N42RF and N43RF data for TOGA COARE.  If you are a user of
these data, you may be interested in the results of the UCI processing
which were obtained independently of both the NOAA/AOC standard
processed data and the NOAA/NSSL GPS correction and re-calculated AOC
horizontal winds.  Some of the differences in processing are described
below. 


The UCI processed data are stored at UCI in binary files (MATLAB
format) of about 33 Mbytes per flight and are accessible via both 
 Mosaic and  anonymous ftp. Some TOGA COARE investigators have 
accessed a previous version of the data and converted it from MATLAB
binary to their own desired format.  If you do not have access to
MATLAB please contact us: sean@cafws2.eng.uci.edu, dkhelif@uci.edu, or
cfriehe@uci.edu.

1. GPS Correction and Horizontal Winds.

Trimble 2100 GPS systems were installed on the WP3Ds for TOGA COARE to
allow for post-flight correction of the Inertial Navigation Equipment
(INE) drift and Schuler oscillation.  This was deemed necessary for
the boundary-layer turbulence and mean wind data in TOGA COARE; it may
also help the convection/radar flights as well.  The GPS data were
recorded at 1 Hz on the slow-rate data tape.  Overall, they appeared
to work well, although there are numerous spikes in turns probably due
to the antenna angle changing in roll, periodic jumps with a 30 sec
period, and a few unexplained periods of no changes.

Our initial processing technique was to correct only for the slow
drift and Schuler oscillation (approximately 84 min period) of the
Delco INE's with a zero-phase-shift low-pass filter using a cut-off
frequency of 0.0005 Hz (33 min period) applied to the differences
between INE and GPS positions and velocities.  These filtered
differences were then added to the original INE data to create 
" corrected " INE data.  This procedure corrected for a large
percentage of the INE errors.  Comments from the radar community (Dave
Raymond) last fall stated that better corrections in turns were
required.  We examined using an increased cut-off frequency, and Sean
Burns of UCI found by cross-spectral analysis that the GPS horizontal
velocity data had a phase shift corresponding to a 1.2 sec time lag as
well. (See memo from Sean Burns on Dec 20 1994 about the details 
of the UCI correction scheme.  There was no phase shift between the
GPS and INE position data.  Accounting for the velocity time lag,
which can only be done to the nearest second for the slow-rate data,
showed increased improvement in the corrections.  Additionally,
increasing the low-pass filter cut-off frequency to 0.0025 Hz (6.7 min
period) also was judged to further improve the correction without
allowing significant errors due to spikes in the turns.  In summary,
the UCI GPS correction technique is to shift the GPS velocity data by
1 sec, apply the 0.0025 Hz zero-phase-shift low-pass filter to the
position and velocity data.  We believe that this provides the best
over-all correction.  We have also manually corrected some of the
egregious GPS errors which were found in a few flights.  We examined
the possibility of differentiating the corrected position INE data to
obtain corrected ground speeds, but we found that excessive noise was
introduced.  We have compared our GPS correction to that of NOAA/NSSL.
The NSSL correction appears to retain the 30 sec periodic jumps and
spikes from the GPS.  A PostScript plot comparing the two GPS corrections 
to the east component of the groundspeed vector is available. A 
B & W version  of these plots is also available. All these plots are 
described here.

The main effect of the GPS velocity correction is of course on the
horizontal winds.  The AOC-processed data have no GPS correction. NOAA
NSSL recently released a data set which corrects the AOC winds with a
variational GPS scheme.  We have compared these data to ours.  In many
TOGA COARE boundary-layer flights, we performed precise
up-wind/down-wind maneuvers to check the winds.  We used these to
obtain our own values for calibrations of side-slip angle sensors and
dynamic pressure offsets for our own independent wind calculation.
The comparisons indicate that the original AOC winds and those with
the NSSL GPS correction do not pass the wind checks.  Part of this may
be attributed to dynamic and static pressure corrections used by AOC
which we believe are erroneous; we have used the static pressure
corrections obtained from the UCI-NCAR-AOC trailing cone test flights
performed for the TOGA COARE project.  In the convection flights, we
have examined some up-wind/down-wind segments that were flown as a
part of the research plan and the circle " pearls ", and find
that the UCI winds generally are acceptable at high altitudes and in
other maneuvers as well.  (There are exceptions when in heavy
precipitation the radome and/or the fuselage sensors disagree perhaps
due to water ingestion or icing.)

It should be noted that there are still unknowns in the WP3D winds.
These arise primarily from the calibration of the side-slip angle and
the heading accuracy from the INE's.  The Delco Carousel IV INEs which
have been used on the WP3Ds since they were built have a factory
specification accuracy of only 0.4 deg in true heading.  At 100 m/s
true airspeed, this gives a wind error of 0.7 m/s.  There is no
independent heading reference on the WP3Ds.  (Recall that a
single-antenna GPS system such as the Trimble 2100 measures track
angle, not heading.)  Examination of the difference in heading between
INE1 and INE2 reveals that they can drift during an 8-hour flight up
to 0.35 deg apart.  Also there is no record at AOC of the accuracy of
the alignment of the heading of the INEs with respect to the aircraft
axis, i.e., to the side-slip sensors.  (Initially we thought there
were errors in heading alignments, but have chosen instead to
incorporate the equivalent biases in slip angles.)

Overall, our up-wind/down-wind checks indicate that the UCI horizontal
winds are good to about +/- 0.5 m/s in straight and level flight.
Comparison of the UCI-WP3D winds with those from the WHOI IMET buoy,
adjusted for the height difference, is also good.  In the turns,
accuracy is degraded due partly to some spikes from the GPS passing
through the filter and unresolved errors as discussed above. PostScript 
 plots comparing UCI, AOC, and NSSL winds for a set of wind maneuvers 
are available. B & W version of these plots are also available 
(wind components and track, elevation and heading). These plots are 
described here. 

It appears that we have reached the accuracy limit for the horizontal
winds on a " production " basis for the 45 TOGA COARE WP3D
flights.  Some researchers may want to make detailed corrections for
specific portions of certain flights on a case-by-case basis.

2. Air Temperature

The standard WP3D total air temperature sensors are the Rosemount
de-iced hermetically sealed units.  These have a slow response time,
of the order of 10 sec.  To obtain ambient temperature, a correction
has to be made for dynamic heating which is about 6-12 K.  This
incorporates the dynamic pressure from a Pitot tube.  The frequency
response of the dynamic pressure transducers is greater than that of
the temperature sensor (about 1 Hz for the co-pilot's fuselage Pitot),
so that the calculated ambient temperature has high frequency "signal"
that is erroneous; i.e. above about 0.1 Hz the total temperature
signal is greatly attenuated and the dynamic heating correction is not
applicable.  Therefore, we have, for the purpose of the ambient
temperature calculation only, low-pass filtered (with zero phase
shift) the dynamic pressure to match the response of the de-iced total
temperature sensors. (There are two on each WP3D.)  For the
UCI-processed slow-rate 1 Hz data, we have used these filtered ambient
temperatures for the calculations of all thermodynamic quantities.

For those wanting a faster-response ambient temperature signal, the
UCI thermistors with response to at least 1 Hz are available in our
processed data. (For a description of the thermistor sensor, see
Fuehrer, Friehe and Edwards, J. Tech, v 11, pp 476-488, 1994.)  These
were tailored to the boundary-layer conditions, so sometimes go
off-scale at colder, high altitudes.  Also, they were subject to some
radio interference.  Please contact us if you are interested in using
these data.

3. Humidity

Several humidity sensors were flown on each WP3D in TOGA COARE; UCI
added a second-cooled mirror dewpoint (EG&G), a Lyman-alpha hygrometer
(AIR); AOC added their own modified cooled mirror and installed the
NCAR UV-hygrometer.  For the slow-rate data, we have used the AOC
General Eastern "cooled mirror" signal in all calculations (humidity
correction to true airspeed, thermodynamic variables, calibration of
the AIR Lyman-alpha fast-response humidiometers.)  The GE's exhibit
the usual characteristics: slow response, oscillations, slow recovery
from sharp transients, etc.  For those interested in faster response
humidity data, the calibrated Lyman-alpha signal (in terms of dew
point and absolute humidity) is available in the 1 Hz data set.  Some
cautions apply; at high altitudes one of them (on N42RF) oscillates.
Also the calibrations were obtained by a least-squares fit over a
large portion of each flight where the dew point was > 0 C and there
were no oscillations or other errors; generally these are quite good,
but some improvement may be possible for individual shorter segments
and conditions where the dew point is less than 0 C.  On some flights,
especially later in the IOP, the NCAR UV-hygrometers appeared to work.

4. Radiation

In the UCI-processed data set, we have added in the short and long
wave radiometer data using the calibrations supplied by NOAA AOC.

The short and long wave radiometer signals were not monitored during
the TOGA COARE field experiment.  In the preparation for the
subsequent CEPEX project in Hawaii immediately following TOGA COARE,
CEPEX radiation scientists noted errors in the radiation signals and
corrected these for WP3D N42RF which was used in CEPEX.  We hope that
the CEPEX radiation scientists can provide a correction scheme for the
TOGA COARE data set.

Also, the UK C130 participated in TOGA COARE and performed some
boundary-layer comparison flights on two days (17 and 18 Jan 1993).
Dr. Phil Hignett of the UK Met Office has worked-up some of the
radiation data and finds that the downward long-wave signal on WP3D
N42RF had reverse polarity and a large offset. The short-wave
radiation data agreed, although there is some question about the type
of Epply colored filters on N42RF. It may be possible to use the
excellent UK C130 radiometer data to obtain empirical corrections for
WP3D N42RF, and then use the many close formation flights with N43RF
to examine N43RF's radiation data.  (The NCAR Electra comparison
flights are another source.)

5. Empirical adjustments

Prior to the 1994 TOGA COARE International Data Workshop held in
Toulouse, we made empirical adjustments to some of the aircraft
measurements.  The adjustment consisted of adding or subtracting a
constant to/from the measurements. These offsets were determined from
the many aircraft-to-aircraft intercomparisons in TOGA COARE with the
goal of increasing the consistency between the various aircraft
platforms.  They are subject to further investigation and revision.
The empirical offsets that have been used in the UCI data processing
are summarized in the table below:



---------------------------------------------------------------------
UCI Empirical Adjustments for TOGA COARE                    May 20, 1994

        Formula: X_adj= X_orig + Empirical Adjustment

Variable        N42RF           N43RF           308D            C-130
--------        -----           -----           ----            -----

 Total          ttf2            ttf1            N.A.            N.A.
Temperature  (+0.35 C)       (reference)

 Ambient        N.A.            N.A.            atb             None
Temperature                                  (-0.2 C)          (~ref)

 Dewpoint       dpgef           dpgef         dptc (ge)          dp
Temperature  (-0.50 C)       (reference)        (~ref)         (~ref)

Sea Surface     prt5            prt5         rstb   sst_dh      sst
Temperature  (+0.80 C)       (+0.70 C)    (-0.70 C) (~ref)      N.A.
---------------------------------------------------------------------


6. UCI Data Access

The UCI WP3D data set for TOGA COARE can be accessed by either Moasaic
or FTP.  Also available (by either Mosaic or FTP) are the PostScript
plots, text files, and memos which are previously mentioned. In
addition, this memo is available as is further documentation about the
UCI processing.

 A. Via NCSA Mosaic

Access to the UCI Mosaic page is done using the URL:

                 http://cafws4.eng.uci.edu

NCSA Mosaic can be used to view and obtain PostScript graphs of the WP3D
data. For each of the two planes, there are nine pages of graphs per
flight, consisting of a 3-D flight track and eight pages of time
series "strip-chart" graphs of important variables. Other information
about the UCI data processing (as listed in part B below) is also
provided in the Mosaic pages.

 B. Via Anonymous FTP
  
ftp cafws4.eng.uci.edu  (OR ftp 128.200.56.33)
anonymous               (Name)
your E-mail address     (Password)
cd pub/TOGA_COARE/

There are three directories off of 'TOGA_COARE/', they are:
DATA_MATLAB/ (contains all processed data files in MATLAB binary format)
Documents/   (contains documentation about the data)
Plots_var/   (contains PostScript time series plots of selected variables
              described in paragraph below) 

As mentioned above, the UCI data are in MATLAB format.  For the
slow-rate 1 Hz data, there are some 160 variables.  A listing of these
with definitions and units is available in the PostScript file
"pub/TOGA_COARE/Documents/0togac_slow_var.ps" and also in the ASCII
Latex text file "pub/TOGA_COARE/Documents/0togac_slow_var.tex". More
information on format and file structure is given in
"pub/TOGA_COARE/Documents/0READ_ME".

A brief description of the files and plots which reside in
pub/TOGA_COARE/Documents is given here for the convenience of the user.

 0READ_ME                * A general description of the UCI processed data.

 0togac_slow_var.ps      * PostScript file of UCI variables' nomencalture.

 0togac_slow_var.tex     * ASCII Latex text file of UCI variables' nomencalture.

 GPS_corr_uci.tx         * Sean Burns memo of Dec 20 1994 regarding the UCI
                         * GPS correction of the INE groundspeeds.

 plot_descrip.txt        * A Description of three PostScript plots (plot1*.ps,
                         * plot2*.ps, and plot3*.ps) which evaluate the
                         * UCI processed data and compare it to the
                         * NOAA/AOC and NOAA/NSSL processed data.

 plot1_gpscomp_clr.ps    * A color PostScript plot comparing the GPS correction
                         * scheme of UCI with NOAA/NSSL.
                         * (see plot_descrip.txt for details)

 plot1_gpscomp_bw.ps     * A black and white PostScript plot comparing the
                         * GPS correction scheme of UCI with NOAA/NSSL.
                         * (see plot_descrip.txt for details)

 plot2a_windcomp_clr.ps  * Two color plots which compare the calculated winds
 plot2b_windcomp_clr.ps  * by UCI, NOAA/AOC, and NOAA/NSSL during a set of wind
                         * maneuvers.  (see plot_descrip.txt for details)

 plot3a_windcomp_bw.ps   * Two black and white plots which compare the
 plot3b_windcomp_bw.ps   * calculated winds by UCI, NOAA/AOC, and NOAA/NSSL
                         * during a set of wind maneuvers.
                         * (see plot_descrip.txt for details)

Note: The blanks appearing in the file names listed above are in fact
      under-scores ("_") and not blanks. Because they are underlined by
      Mosaic, they look like blancs. 


7. Registration of UCI Data Users

   If you use the UCI-processed WP3D TOGA COARE data, we would
appreciate it if you fill out the form below and send it
electronically to us (dkhelif@uci.edu, sean@cafws3.eng.uci.edu). This
will enable us to keep you informed about revisions and future
updates.  Also, it would be extremely helpful to us if you would
inform us of any problems you encounter with these data.

---------------------------------------------------------------------------
Name:
Institution:
E-mail address:
----------------------------------------------------------------------------