The Super Dual Auroral Radar Network (SuperDARN)
is an international collaborative network of HF radars that
monitors ionospheric plasma convection over the majority of the
northern and southern polar regions. SuperDARN currently is
comprised of 13 radars in the northern hemisphere and 7 radars in
the southern hemisphere.
Fig. 1: The SuperDARN fields of view in (a) the northern and (b) the southern hemispheres (geographic co-ordinates).
The location and boresite direction for all SuperDARN radars is listed in the following table:
|Radar Name||Radar Code||Geog. Lat.||Geog. Lon.||Boresite Heading|
The radar network's vast coverage extends longitudinally over more than 18 hours of local time in the northern hemisphere, and latitudinally from equatorward of the auroral electrojet to well into the polar cap, thus sampling the ionospheric footprint of many magnetospheric regions. Each SuperDARN radar has a very large field of view, covering approximately four million square kilometers. A complete radar scan is performed in one or two minutes in the common modes of operation, leading to data with time resolution good enough to improve the understanding of the dynamics of ionospheric and magnetospheric convection.
The radars normally sound 16 beams sequentially to form a full 52° azimuth scan. During the common mode of operation each of the SuperDARN radar pairs is synchronised to perform a full scan every two minutes, with a dwell time of 7 seconds per beam. During the fast common mode of operation each of the SuperDARN radar pairs is synchronised to make a complete scan every minute, with a dwell time of 3 seconds per beam. The beams are separated by 3.2°. 75 range gates, each of 45 km in length, are measured along each beam and the range to the first gate is 180 km. The beam is produced by an array of sixteen log-periodic antennas and an electronically controlled phasing matrix that steers the radar beam through its 16-position scan.
Many operating modes are possible with SuperDARN. In addition to the common and fast common modes, which together make up at least 50% of the total operating schedule, the schedule includes discretionary and special time to test and run alternative operating modes, as well as to repair and upgrade the radars. For example, in one of the discretionary modes a radar scans through all sixteen beams sequentially but returns to a selected beam position after every other beam (e.g., 0,8,1,8,2,8,3,8,...,8,15), resulting in better time resolution along that beam. This mode is useful for studies requiring better temporal resolution in a smaller area.
Each radar site also has, separated by 100 m from the main array and
parallel to it, a four-antenna interferometer array that is used
to make angle-of-arrival measurements. By measuring the phase
difference between the antenna arrays it is possible to deduce
the elevation angle of arrival of the incoming radio waves. This
is important for identifying the HF propagation mode.
2. Ionospheric Sounding by Pulsed Radars
The three principal parameters determined by the radars are the component of the plasma drift velocity along the beam, the backscattered power and the spectral width of the signal. The complex autocorrelation function (ACF) of the received signal is used to determine these three parameters in each of the SuperDARN range cells. The technique by which this is achieved utilises a multi-pulse transmission sequence. During the presently run common mode of operation, the radars transmit a seven-pulse sequence of the form illustrated below:
Fig. 2: The 7-pulse sequence transmitted by the SuperDARN radars. The pulses are 300 µs long and they are separated by integral multiples of the lag time, τ, which is normally set to 2400 µs.
Each transmitted pulse has a duration of 300 µs. During the common mode of operation the radars transmit their seven-pulse pattern (0,9,12,20,22,26,27) during a 100 ms transmission window. After each pulse is transmitted the radars switch to receiver mode and the return signal is sampled periodically and processed in order to reduce the ACF for each range, each a function of the lag time. For a 7 second dwell time the multi-pulse sequence is usually repeated 70 times at each beam position and the 70 resulting ACFs are integrated and averaged to increase the signal to noise ratio.
The pulse separation of a multi-pulse pattern is chosen in such a way as to maximise the number of unique lags between the pulses. For example, the seven-pulse pattern illustrated above has unique lags for all integral multiples of the lag times except for 16τ, 19τ, 21τ, 23τ, 24τ, and 25τ, which are missing. For this reason, SuperDARN uses only the first 18 lags to determine the ACFs for this pulse sequence. An example of a SuperDARN ACF is presented below:
Fig. 3: (a) The complex ACF measured along beam 3 at range gate 59 at 11:21:06 UT on 24 November 1998 by the SuperDARN radar at þikkvybær, Iceland. (b) The phase of the ACF (crosses), which aliases at ±π, and the fitted phase (solid line) determined by the SuperDARN ACF fitting routines. (c) The normalised power spectrum (black line) obtained from the Fourier transform of the ACF, and the power spectrum calculated from the velocity, power and spectral width parameters determined from the ACF fitting routines.
The real part of the ACF, the solid line, maximises at lag zero, as expected. The imaginary part of the ACF is offset from the real part by the phase, which is plotted in panel (b). The phase aliasing at ±π complicates the determination of the phase of the complex ACF. The distance, or lag, to the first range is determined from the delay time between the transmission of a pulse and the measurement of the first echoes. In order that the zero-lag correlation of a pulse with itself not be contaminated by the returned signal from the other pulses in the transmission sequence, the first pulse is usually separated from the rest of the pulse sequence by a comparatively large time lag with respect to the rest of the lags in the sequence. For the SuperDARN radars, where the transmitters and receivers are co-located, if the radars are transmitting a pulse, then the extremely sensitive receivers must be turned off. This results in missing lags in the ACF. For the far ranges this can result in the receiver being turned off when the first echoes return from the first pulse. For the current pulse pattern, the receiver cannot receive the first echoes of pulse 1 (0τ) at range gate 68 since the radar is transmitting pulse 2 (9τ). Therefore, the zero-lag power for range gate 68 cannot be determined, and the ACF fitting for this range is extremely poor. This is also the case for echoes from range gates 4, 12 and 20, which correspond to the 1τ, 2τ, and 3τ lags of the ACFs at those ranges. These first few lags are also very important to finding good fits to the ACFs and the fits determined for these ranges are therefore generally of slightly poorer quality.
The optimisation of the pulse sequence is a balance between minimising the number of missing lags and eliminating multiple occurrences of the same lag, which introduce range aliasing in the ACF. For a multi-pulse pattern with many sampled ranges, range aliasing, or "cross range noise," can become a problem when there are correlated signals at several ranges being received simultaneously. For example, echoes from pulse 2 (9τ) are received from range gate 47 at the same time as echoes from pulse 3 (9τ) are received from range gate 23. If the target irregularities are well correlated within each range gate, then the signal will not tend to zero with averaging of multiple ACFs. The method by which the SuperDARN radars deal with this cross-range noise is to compare the zero-lag powers for all ranges from which echoes are received. If the zero-lag power is much stronger in one range gate, then the data is attributed to that gate.
In general, if ambiguities exist in the received signal, the data
is disregarded if: (a) the receiver is turned off because
another pulse is being transmitted, (b) the power of the
returned signal from a range gate is similar to or much weaker
than the simultaneous returns from another range gate, or (c)
the ACF is unphysical, which is due, in general, to HF
3. Determination of Radar Spectral Characteristics
A common method of calculating the Doppler velocity of the backscattered radar wave employs the Wiener-Khinchin theorem, which states that the Fourier transform of the ACF is the power spectrum of the backscattered signal. The spectrum in Fig. 3(c), denoted by the black line, is the Fourier transform of the ACF in Fig.3(a). Because there are not a large number of data points in the SuperDARN ACFs and there can be missing lags, SuperDARN utilises a method of phase-fitting to determine the spectral characteristics (velocity, power, and spectral width) directly from the ACF.
The complex ACF can be written in terms of its amplitude and phase at lag k. The rate of change of the phase of the ACF as a function of the lag time τ can be used to determine the mean Doppler shift frequency of the backscattered signal. If the signal is dominated by a single Doppler frequency, as is the case in Fig. 3, there will be a linear variation of the phase over time. The SuperDARN software is designed to perform a linear fit to the phase, after accounting for the 2π phase jumps. The slope of this line is the measured Doppler frequency. The Doppler velocity of the plasma along the radar beam can then be determined as follows:
Vel = c * (Dopp_Freq)/(4*π*Radar_freq)
The operating frequency of the SuperDARN radars (Radar_freq) is between 8 and 20 MHz. The first few lags in the ACF are extremely important to the fitting procedure, and any missing lags in this portion of the ACF are very detrimental to the fit.
The other defining properties of the power spectra, namely the backscattered power and the width of the spectra, can be determined directly from the variation of the ACF amplitude. When there is more turbulence in the ionosphere the coherence time of the plasma structures from which the radio waves are scattering is lower. This is reflected in a faster decay of the ACF. The maximum backscattered power is the zero-lag power and the exponential decay envelope of the ACF is related to the width of the spectrum.
The main objective of SuperDARN is to measure ionospheric plasma convection with relatively high spatial and temporal resolution on a global scale. The overlapping fields of view of the radar pairs provide independent plasma drift measurements in two directions. Each of the radars in a SuperDARN pair has a 16-beam field of view, so the radar pair has, in principle, 256 beam intersection cells. Each beam measures the projection of the full velocity vector onto the beam, and from the two overlapping velocity measurements the full vector can be reconstructed, or "merged."
Another method by which global scale convection velocities are derived is the so-called "map potential" technique (Ruohoniemi and Baker, 1998). Equipotential contours are determined by fitting a series of spherical harmonic functions to the measured convection velocity components measured by the SuperDARN radars. In regions where there is little or no data, the fits are constrained by statistical models based on upstream IMF conditions. In the ionosphere the equipotential contours are also velocity streamlines. Therefore, these global-scale voltage maps are also equivalent to global-scale convection maps.