Chilbolton 94 GHz Galileo radar data
Robin Hogan
11 April 2002
Introduction
This dataset consists of effective radar reflectivity factor
(Z) measured by the 94 GHz Galileo cloud radar situated
at Chilbolton, England (51.1445°N, 1.4370°W). The instrument
is a bistatic system that operates continuously round the clock in a
vertically-pointing configuration. It was developed for the European
Space Agency by Officine Galileo, the Rutherford Appleton Laboratory
and the University of Reading.
This document is a revision of the original documentation of 26
February 2001, and describes the recalibrated data that was released
to BADC in spring 2002. If you have used the reflectivity values of
the original Galileo data for anything quantitative then you
should definitely download the newer data, as there was a steady and
significant loss of sensitivity during the 18 month duration of this
dataset, that we were previously unaware of.
The characteristics of the radar are as follows:
_________________________________________________
Frequency: 94.00 GHz
Antenna diameter: 0.6 m
Peak power: 1.6 kW
Pulse width: 0.5 µs
Pulse repetition frequency (PRF): 6250 Hz
System noise figure: 10 dB
Beamwidth: 0.5°
Range resolution: 60 m
_________________________________________________
History of operations
The radar was initially operated in summer 1996 for a few months,
but a problem with sensitivity was found. This was rectified when the
radar was Dopplerised, and from September 1999 until February 1999 the
instrument was operated from the `Receive Cabin' at Chilbolton. It
was then mounted on the side of the main 25 m antenna at Chilbolton to
permit scanning with the 3-GHz radar, although did not operate
round-the-clock until the end of April 1999. On 17 October 2000 it was
removed from the side of the main dish to solve the problem of gaps in
the data whenever the 25 m antenna was scanning through precipitation.
During this time the output power from the instrument dropped
steadily; the calibration section below discusses the implications of
this for sensitivity and calibration. Doppler velocities measured
during this time were unreliable and so have not been included in the
dataset. The data described here are from 1 May 1999 until 17 October
2000.
From January 2001 to March 2002 the radar was operated from the
ground at Chilbolton and a new data acquisition system installed which
permitted measurement of the mean Doppler velocity and Doppler
spectral width. Operations from the ground meant that the data stream
was not interrupted every time the 3-GHz radar was scanning, but
unfortunately it also meant that direct cross-calibration with the
3-GHz radar was not possible. Nonetheless, it is hoped that these
data can be released on BADC in the future.
Processing that has been applied to the data
The following stages of processing have been applied to the raw
data (in this order):
- True linear averaging of the original 1-s data to 30 s.
- Rejection of rays recorded when the radar was pointing more than
5° from zenith (i.e. when the 3-GHz CAMRa radar on to which
the 94 GHz radar was attached was scanning through precipitation).
- Rejection of `glitched' rays when the data acquisition system got
out of sync with the transmitter.
- Calculation of the noise level by looking at the assumed
cloud-free gates between 13 and 14.5 km. This is stored in the
nez variable.
- Subtraction of the noise component from the measured echo power,
and thresholding of the resulting signal.
- Clean-up of speckle noise by rejecting cloudy pixels or pairs of
cloudy pixels with cloud-free pixels above and below them.
- Range correction of the data to account for the inverse square law
and any range offset.
- Application of calibration figure (to both Zh and nez).
- Spring 2002: Recalibration based on numerous calibration
events with the 3-GHz radar.
The result of this processing is that the minimum-detectable signal at
a range of 1 km was around -50 dBZ in May 1999, increasing to around
-38 dBZ in October 2000. The sensitivity decreases with range
according to the inverse square law; therefore the minimum-detectable
signal at 10 km is 100 times (20 dB) higher than at 1 km. The
`chil2nc' program was used to process the data and convert it
to NetCDF; source code for the latest version may be downloaded from
http://www.met.rdg.ac.uk/radar/software.html.
Data format
The data is provided as daily
NetCDF
files containing the following variables:
- frequency
- The frequency of the radar in GHz (94.00).
- latitude
- The latitude of Chilbolton in degrees north
(51.1445).
- longitude
- The longitude of Chilbolton in degrees east
(-1.4370).
- altitude
- The altitude of the radar antenna above mean
sea level in metres (90 m when the radar was on the side of the main
CAMRa dish).
- time
- A vector containing the centred time of each ray
of data, in decimal hours UTC.
- range
- A vector containing the centred range from the
antenna, in km, of each range gate. Note that to get height above mean
sea level you should add the altitude to this value.
- nez
- A vector containing the `noise-equivalent radar
reflectivity factor' at 1 km in dBZ (i.e. the level of the
thermal and instrumental noise normalised to a range of 1 km), after
calibration. It is calculated for each ray using the range-gates
between 13 and 14.5 km which are assumed to be cloud free. It
increases through the dataset due to the steady loss of instrument
sensitivity, but can also vary on short timescales by as much as 2 dB
when low clouds pass overhead because of the change in sky brightness
temperature at 94 GHz.
- Zh
- An array containing effective radar reflectivity
factor, in dBZ. A value of -999 dBZ indicates that no
signal was detected. The comments attribute contains the
following string: Calibration convention: in the absence of
attenuation, a cloud at 273 K containing one million 100-micron
droplets per cubic metre will have a reflectivity of 0 dBZ at
all frequencies. In addition to the usual sources of attenuation at 94
GHz (liquid water clouds, water vapour, oxygen, melting snow and
rain), it has been found that when the radomes covering the antennae
of the radar become wet, they strongly attenuate the beam, typically
by more than 10 dB. Reflectivity values measured at all altitudes
during rain (and shortly afterwards) should therefore not be
trusted.
The following global attributes are present:
- system
- A string identifying the instrument that took
the data, in this case `Galileo'.
- scantype
- A string identifying the type of scan that
the radar was performing, in this case `Fixed'.
- day
- The day of the month as a short (two-byte) integer.
- month
- The month as a short integer.
- year
- The year as a short integer.
- file
- The `tape' number of the original
Chilbolton-format radar data file from which the data in this NetCDF
file was obtained, as a short integer. Where a day of data spans more
than one such file, the tape number of the first is given.
- raster
- The number of the first raster of the first
Chilbolton-format data file that was used in producing this data, as a
short integer. Chilbolton-format radar files are divided into rasters
which, in the case of cloud radar data, are usually one hour long.
- options
- A string containing the command-line options
that were used with the chil2nc program to produce the
data. This is useful to see what processing has been done, such as
the calibration figure that has been applied in dB (the -Zcal
switch).
- software_version
- The version of chil2nc that
was used to produce the data. This field was not implemented in
versions of chil2nc earlier than around 0.7.6.
The following global attributes were added when the data were
recalibrated:
- history
- The record of programs that have operated on
the data file since its original generation.
- command_line
- The command line that was used to invoke
the program used to perform the recalibration.
- comments
- This attribute contains the following
message: It has been found that the transmitter of the Galileo
radar has been losing power steadily since around September
1999. Recalibration of the entire dataset from May 1999 to October
2000 (the parameters Zh and nez) has been performed by comparing with
the 3 GHz radar at Chilbolton in 13 scanning events. The 3 GHz radar
is itself calibrated to better than 0.5 dB using the method of Goddard
et al. (1994, Electronics Letters 30, 166-167). The resulting Galileo
calibration is believed to be accurate to around 1.5 dB. It should be
noted that this loss of sensitivity means that the fraction of clouds
not detected by the radar will increase through the dataset.
Simple programs to read NetCDF files of Chilbolton data into
Matlab, IDL and PV-WAVE can be found at
http://www.met.rdg.ac.uk/radar/software.html.
Interpretation of radar reflectivity factor and calibration issues
It is taken for granted that if quantitative use is to be made of
the Z values then the user has a fairly good understanding of
the concept of radar reflectivity and how it is related to the
particle size distribution. In this section therefore we concentrate
on calibration issues, although the problem of attenuation at 94 GHz
is also discussed since it is much stronger than at lower
frequencies.
Our approach to calibration of the 94 GHz radar is by reference to
the 3 GHz `CAMRa' radar at Chilbolton. CAMRa can be calibrated
absolutely by exploiting the non-independence of the radar parameters
Z, ZDR and KDP in heavy
rain, as described by Goddard et al. (1994). However, to compare the
two radars directly requires a Rayleigh-scattering target which is
both near enough for attenuation to be small at 94 GHz, yet far enough
that near-field and ground-clutter effects are small at 3 GHz. We
correct for gaseous attenuation at 94 GHz using the temperature and
humidity from either the ECMWF or Met Office models, and make the
small correction for the near-field effect at 3 GHz using the
expression of Sekelsky (2001). The results are expressed in the figure
below in terms of the noise-equivalent reflectivity at 1 km (see the
definition of nez above):
The error bars indicate the 13 calibration events and their estimated
uncertainty, and the numbers indicate the Chilbolton tape number. A
steady deterioration in sensitivity is evident, which is due to a loss
of transmit power from the tube. Given the uncertainties in the
intercomparison of the two radars, the error in the resulting 94-GHz
reflectivity factor after calibration is estimated to be around
±1.5 dBZ.
It is important to understand the convention used in the
intercalibration of radars of different frequencies because of the
temperature dependence of the |K|2 parameter of
liquid water at millimetre wavelengths. We have calibrated our radars
such that Rayleigh-scattering liquid water droplets at 0°C produce
the same reflectivity factor at all frequencies. For example, a
population of 100µm droplets with a concentration of
106 m-3 at 0°C would have a Z of 0
dBZ at all frequencies. Hence a radar at frequency f
(after calibration and correction for attenuation) will report an
effective reflectivity given by
Zf = Integral from
D=0 to D=infinity {
(|Kf|2/|Kf,0|2)
n(D) D6
Mf(D) dD },
where Mf is the Mie/Rayleigh backscatter
ratio. |Kf,0|2 is the dielectric
parameter of liquid water at 0°C, and is 0.93 at 3 GHz, 0.877 at
35 GHz and 0.668 at 94 GHz. Because the
|Kf|2 of liquid water varies with
temperature at 94 GHz, if the example above were repeated at 20°C,
a 94 GHz radar using this calibration convention would report a
Z of +0.82 dBZ while a 3 GHz radar would still report 0
dBZ. Formulae for the dielectric constants of ice and liquid
water can be found in Liebe et al. (1989).
Attenuation by both atmospheric gases and liquid water is much
stronger at 94 GHz than at lower frequencies. At 10°C, 1013 mb and
100% humidity, the one-way attenuation due to gaseous attenuation at
94 GHz is 0.636 dB km-1. In summer the typical two-way
gaseous attenuation to top-of-atmosphere is 2 dB. If the temperature
and humidity profile is known with some degree of accuracy (such as
from a model or a radiosonde ascent) then gaseous attenuation can be
corrected for. However, when low clouds are present then the liquid
water attenuation can easily exceed the gaseous attenuation, and of
course the profile of cloud liquid water content is generally far more
uncertain. This makes quantitative use of the reflectivity data in ice
difficult if there is any low cloud present. At 10°C and 1013 mb,
the one-way attenuation of 1 g m-3 of liquid water is 4.34
dB km-1. Attenuation by rainfall is even greater, and in
moderate and heavy rain can extinguish the signal completely. Wetting
of the radomes of the radar also introduces a large attenuation; see
point 4 in the next section.
Known problems with the data
- Gaps in the data are present due to:
- The CAMRa radar (to which Galileo was attached from
February 1999 to October 2000) scanning through precipitation. Cloud
data were not recorded whenever the pointing angle was more than
5° from zenith.
- Glitches in the data caused by the data acquisition system losing
sync with the transmiter. These events were easy to detect and reject.
- A 50 kHz interference was present in the raw data which, if
unchecked, resulted in anomalous horizontal `lines' of cloud every 3 km
in processed time-height plots of reflectivity. This problem has been
tackled in the following ways:
- Characterising the shape of the oscillation in the cloud-free
gates at the top of each ray, and then using this knowledge when
subtracting the noise from the remainder of the ray.
- Removing isolated cloudy pixels or pairs of pixels from the
processed ray.
- Manual removal of any remaining erroneous echos.
Removing this interference has unavoidably compromised the sensitivity
to some extent. Also, some anomalous echos may still be present in
some of the data, although they are fairly easy to locate
subjectively. The removal of pairs of pixels (indicated by the
options attribute containing the string
`-doubleclean') obviously will remove any genuine cloud that
is only two range gates thick.
- The reflectivity values in the lowest gates are affected by:
- Ground clutter and leakage of the transmit pulse into the
receiver: this appears as a fairly constant return (in time) in the
lowest 6 to 8 gates that falls off rapidly with range.
- The near field effect: this appears as decrease in the Z of
rain in the lowest 2 or 3 gates. The far-field approximation is generally
applicable beyond 2×antenna
diameter2/wavelength, which for the
Galileo radar is 225 m. A slight error in the range calibration
could also have contributed to the apparent reduction of Z in
the closest few gates.
- Insects: although the problem is much less at 94 GHz than 35 GHz
(due to the fact that insects are usually Mie scatterers), on some hot
summer days insects are apparent between dawn and dusk as fairly low
Z values up to 3 km. They can be distinguished from cloud using
data from the Chilbolton lidar ceilometer.
- Wetting of the radomes covering the two antennas of the radar has
been found to cause a two-way attenuation of 9-14 dB. The data should
therefore not be used quantitatively during rain events, which can be
identified using the Chilbolton rain-gauge data also provided by
BADC. Other radars at 94 GHz presumably also suffer from this problem.
Conditions of use
If data from the Galileo radar is used in any publication or
report then acknowledgement must be given to RCRU at the Rutherford
Appleton Laboratory for providing the data. The acknowledgement should
be of the form:
We thank the Radiocommunications Research Unit at the
Rutherford Appleton Laboratory for providing the 94 GHz Galileo radar
data. The Galileo radar was developed for the European Space Agency
by Officine Galileo, the Rutherford Appleton Laboratory and the
University of Reading, under ESTEC Contract
No. 10568/NL/NB.
Who to contact
If you have any problems obtaining the data, please contact the
British Atmospheric Data Centre. If you have problems, queries or
comments regarding the data themselves that are not covered
adequately by this document, or would like to know if any data was
recorded on specific dates outside the period available on BADC,
please contact Charles Kilburn (C.Kilburn@rl.ac.uk) and Robin
Hogan (r.j.hogan@reading.ac.uk). Even
if you use the data and have no difficulties at all, we are very
interested in knowing the uses to which our cloud radar data is being
put, so please contact us!
See also
References
- Goddard, J. W. F., J. Tan and M. Thurai, 1994: Technique for
calibration of meteorological radars using differential
phase. Electronics Letters, 30, 166-167.
- Liebe, H. J., T. Manabe and G. A. Hufford, 1989: Millimeter-wave
attenuation and delay rates due to fog/cloud conditions. IEEE
AP, 37, 1617-1623.
- Sekelsky, S. M., 2001: Near-field corrections for meteorological
radars. Proc. 30th AMS Conference on Radar Meteorology,
Munich, Germany, 32-34.