SEARCH

SEARCH BY CITATION

Keywords:

  • VHF radar;
  • wind profiler;
  • stratosphere-troposphere

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Radar
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

[1] The Mount Gambier wind profiling radar was installed in September 1997 and has operated almost continuously since the beginning of 1998. The site is at the Australian Commonwealth Bureau of Meteorology Mount Gambier Meteorological Office, close to the Mount Gambier airport. The radar was developed as a prototype operational profiler by the Atmospheric Physics Group at the University of Adelaide, in collaboration with the Australian Commonwealth Bureau of Meteorology. It was designed to operate as a spaced antenna radar, as a Doppler beam swinging radar, and as a hybrid Doppler interferometer. Here we present a summary of the operation of the system and a comparison with radiosonde observations concentrating on operation in the spaced antenna mode. The utility of the system and recommendations for the further development of similar systems are addressed.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Radar
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

[2] The Mount Gambier VHF (44.75 MHz) profiler was developed to exploit the advantages of the lower VHF band and the spaced antenna (SA) technique when applied to wind profiling. At the time of its development, most operational profilers utilized the Doppler beam swinging (DBS) technique. Now, the SA technique is being increasing used, particularly in the boundary layer troposphere (BLT) mode at lower VHF [see, e.g., Vincent et al., 1998], but also at UHF [see, e.g., Cohn et al., 2001]. In many situations, the SA technique utilizing the full correlation analysis (FCA) offers advantages. These include the need for only relatively small antenna arrays, no requirement for beam steering, and an analysis very tolerant of antenna deterioration. Like all techniques, however, there are advantages and disadvantages, and it was one of the major aims of the present project to determine the applicability of the SA technique in an operational system. Although the Mount Gambier radar is capable of operation in both DBS and SA modes, the main mode of operation for the six years of operation reported in this study has been the FCA analysis of SA data, and it is this mode we investigate in detail here. However, the radar was designed as development system, and we do present some brief examples of operation in the hybrid Doppler interferometer (HDI) mode.

[3] The use of the lower VHF frequency band for profilers has advantages because of its reduced susceptibility to bird, insect and precipitation echoes, and although it does generally require physically larger antenna arrays for Doppler work than higher frequencies, lower VHF BLT systems utilizing the SA technique have a footprint of less than 25 × 25 m. Another aspect that must be considered with radars operating in the lower VHF band is the effect of the atmospheric aspect sensitivity on the antenna polar diagrams. For the SA technique, this results in a narrowing of the vertically directed beam pattern, and this generally has little effect on the technique. For the DBS technique, with off-vertical beams, the effect is to bias the polar diagrams back toward the zenith. With relatively broad beams (or relatively small antenna arrays; say beam widths >4°) the effect can result in severe underestimation of the radial velocities. However, the effect can be corrected for if the mean angle of arrival can be determined. This suggests a combined Doppler-interferometer mode of operation for smaller arrays, in which hardware beam steering is applied on transmission to form a relatively narrow transmit beam, and the returned signal is received on multiple subsections of the array, so that the receive beam can be formed in software. This arrangement allows the effective beam direction to be determined, and permits a smaller antenna array to be utilized.

[4] This, or very similar, arrangements are used at MF with the large Buckland Park MF radar [Reid et al., 1995], and at VHF with the ESRAD radar at ESRANGE in Kiruna, Sweden [Chilson et al., 1999], the ALWIN radar in Andenes, Norway [Latteck and Singer, 2001], the OSWIN radar in Kühlungsborn, Germany [Zecha et al., 2003], the CRL radar in Wakkanai, Japan (Y. Murayama, private communication, 2000), the Australian Antarctic Division radar at Davis Station in Antarctica [Morris et al., 2004] and the Mount Gambier prototype operational profiler described in this paper. The six VHF radars just mentioned are very similar in design in that they are all based on a six-module transmitter system and a six-receiver data acquisition system, and all have a dual SA and HDI/DBS capability. Apart from the Mount Gambier profiler, these radars are used for basic research rather than operational applications, although the ESRAD and ALWIN radars do provide input data into the European CWINDE profiler demonstration network [see, e.g., Dibbern et al., 2003].

[5] The Mount Gambier system is located adjacent to the Mount Gambier airfield, approximately equidistant from the Adelaide and Melbourne regional offices of the Bureau of Meteorology (Bureau).

2. Radar

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Radar
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

[6] The radar is a six receive channel 36 kW solid-state system that operates at a maximum 5% duty cycle. The basic transmitter unit has a peak power of 1 kW, and these are grouped into six independent 6 kW power blocks. The antenna array consists of 144 Yagi antennas grouped in fours, so that 36 basic units are available in the array. However, the basic antenna unit used to connect to a transmitter power block or receiver consists of one sixth of the array, or 24 antennas. These can be configured either as a 4 × 6 rectangular group, or as a 2 × 12 antenna row or column. The antenna layout is shown in Figure 1. The antenna design was a compromise to allow operation in both SA and Doppler modes, and although it is not completely ideal for either, as an evaluation system, it has proved to be extremely useful. Overall system specifications are summarized in Table 1.

image

Figure 1. Antenna arrangement for the prototype profiler. Each diagonal dash represents a three-element Yagi antenna, and the Yagis are grouped into quads as the basic subunit of the array. Each receiver may be connected to 24 antennas (six groups of four) in one of two ways. For spaced antenna operation, three receivers are connected to three noncollinear groupings of 24 Yagi antennas, as indicated by the dashed and shaded boxes. Because there are six receivers, two triangles can be utilized. For Doppler beam steering, six rows or columns of 24 antennas are connected to the six receivers so that beam swinging in the orthogonal direction is possible. Interpulse beam steering is possible within this orthogonal plane. The north-south axis of the array is nominally aligned 27° east of north.

Download figure to PowerPoint

Table 1. Summary of Mount Gambier Radar Parameters
ParameterDescription
Peak power36 kW
Transmitter typesolid-state class E operation
Maximum duty cycle5%
Power aperture product2.5 × 106 Wm2
Pulse coding 4–32 bit complementary, Barker codes
Sample widths150, 300, 600, and 1200 m
Receiverssix independent coherent channels
Frequency44.75 MHz
Wavelength6.7 m
Doppler mode one beam at a time at the zenith, ±7°, ±14° and ±21° off zenith at ±27° (N-S) and ±117° (E-W) azimuth
Half-power beam width6.5°
SA modethree subgroups of 24 antennas arranged on an isosceles triangle of spacing about 35, 35, and 38 m (two independent triangles available)

[7] A relatively crude figure of merit used with this kind of system is the power aperture (PA) product. This is the product of the average power with the antenna array area, and is a proxy for the power per solid angle in the main lobe of the radar. For the Mount Gambier system, the PA product is 4.9 × 106 Wm2. This figure really only provides a rule of thumb indication of capability. In the presence of aspect sensitive regions of the atmosphere, which is most of the time, the actual effective beam is determined by the combination of the antenna polar diagram, and the polar diagram of the atmospheric irregularities. This is discussed in more detail in section 2.1. This PA product applies when the radar is operated using the full array for transmission and reception. In the case of the SA technique as applied here, only half the array is used for reception. May [1990] has shown that the effective antenna area Aeff is then

  • equation image

where AT and AR are the antenna transmit and receive areas, respectively. This equation applies in the case of isotropic volume scatter, which as we have noted above, is not usually the case at lower VHF. For the Mount Gambier system, Aeff = equation imageAT, so the PA product is 3.2 × 106 Wm2.

[8] In SA mode, the present arrangement allows for up to six independent rectangular groupings. In this mode the full correlation analysis is applied. The FCA of SA data is well known, and a detailed discussion will not be presented here (however, see, e.g., Briggs [1984]). For the FCA experiment, only three noncollinear groupings are required. One such grouping is shown as the large shaded rectangles in Figure 1. The centers of these groupings lie on the vertices of an isosceles triangle with sides 34.5, 34.5 and 38.3 m. The FCA analysis can also be applied to the three nonshaded groupings shown in Figure 1, yielding two independent estimates of the wind. As we noted above, this arrangement is not entirely optimum, because for FCA operation, the ideal configuration would be three symmetrical subarrays arranged on an equilateral triangle.

[9] Multichannel FCA or spatial correlation analysis (SCA) operation [see, e.g., Holdsworth, 1999a] is also possible in SA mode, but generally these applications would not offer operational advantages in terms of convincing benefits for the cost of the additional channels involved. Such operation would, however, still be cheaper than DBS operation. As we have noted above, an operational SA FCA radar only requires three noncollinear subgroups, and this is the design that has been pursued with smaller BLT type systems.

[10] In Doppler mode, each of six rows or columns of 24 antennas is connected to a 6 kW power block. By phasing the six outputs appropriately through the insertion of cable delays, the transmitter beam can be steered off zenith. Three off-zenith angles are available. These are 7°, 14°, and 21°. The grouping of antennas into quads produces grating lobes of increasing severity with increasing off-zenith angle, because the basic spacing of the quads is more than one wavelength. These lobes are tolerable at 7° off zenith, but make routine operation at 14° difficult, and very difficult at 21°, where the beam pattern is represented by symmetric lobes at ±21° off zenith. This design compromise was considered acceptable for a prototype system. The half-power beam width of the main lobe of the entire array is about 7°. Two Doppler analysis modes are available on the system. These are the DBS and HDI modes. Like the FCA of SA data, the DBS technique is well known, and so we will not describe it further here. However, see Woodman and Guillen [1974] for a classic description. The hybrid Doppler interferometer (HDI) mode is less well known, and we now briefly describe in more detail.

2.1. Hybrid Doppler Interferometer Operation

[11] In hybrid Doppler interferometer mode, the beam is steered on a pulse-to-pulse basis along the ±27° azimuth (nominally north-south) and the ±117° azimuth (nominally east-west). The zenith angles available are ±7°, ±14°, and ±21°. It follows that a five-beam cycle in the vertical and (nominally) cardinal directions at one off-zenith angle can be completed in two subsequent acquisitions.

[12] The analysis employed in HDI mode is based on the time domain interferometry technique (TDI) of Vandepeer and Reid [1995], rather than the standard Doppler analysis [e.g., Woodman and Guillen, 1974]. The relatively wide transmission half-power half width of the system results in the effective beam direction being biased toward the zenith because of the aspect sensitivity of the atmosphere [e.g., Vandepeer and Reid, 1995]. The use of the transmit beam direction rather than the effective beam direction can therefore lead to substantial velocity biases. Application of time domain interferometry allows the effective beam direction to be determined, thus permitting compensation for the effects of aspect sensitivity to be applied.

[13] The time domain interferometry technique is implemented as follows. After correction for complex gain differences, the poststatistic steering (PSS) technique of Kudeki and Woodman [1990] is used to steer the receive beam. The returned power for each receive beam direction is calculated, and a Gaussian fit is applied about the peak of the power variation to determine the direction of maximum power. If the power variation of the transmit beam as a function of zenith angle θ is assumed Gaussian as

  • equation image

where θb is the e−1 beam width and θt is the transmit beam zenith, and the returned power due to aspect sensitivity is also assumed Gaussian

  • equation image

where θs is the aspect sensitivity parameter [e.g., Hocking et al., 1986; Vandepeer and Reid, 1995], it is easily shown that the effective beam direction corresponds to the direction of maximum power. Furthermore, the effective beam direction will also correspond to the mean angle of arrival [e.g., Röttger and Ierkic, 1985] determined using the phase differences between receiving antennas. The advantage of using poststatistic steering rather than mean angle of arrival determination to determine the effective beam direction is that the combination of the six receiver time series employed for poststatistic steering increases the SNR by 9 dB, thereby reducing the error statistics and increasing the range extent. The receive beam is then resteered into the effective beam direction, and the radial velocity, spectral width, power and SNR are estimated. The aspect sensitivity parameter θs is then calculated using equations (2) and (3) [e.g., Vandepeer and Reid, 1995] as

  • equation image

where θe is the effective beam direction. The time domain interferometry analysis can be configured to use either correlation or spectral analyses. The correlation analysis is recommended except in environments where ground clutter, sea scatter or interfering signals are a problem. In this case the spectral analysis is recommended, as the user can configure the experiment to minimize the effects of the nonatmospheric returns.

2.2. SA FCA Operation

[14] The FCA of SA data was the standard mode of operation for the radar. Two modes were utilized. These were “high mode” and “low mode.” In high mode, a pulse repetition frequency (PRF) of 4096 Hz was utilized. 512 coherent integrations were performed to yield a 360-point time series with a sampling interval of 0.125 s for each of the six 24 antenna subgroups. Thus a time series was 45 s in duration. The FCA analysis was performed on data from each 45 s time series from three antenna groups arranged on one of the triangles shown in Figure 1, providing a potential sampling interval of about 1 min for a determination of horizontal velocity. FCA analysis was performed on data from each of 63 range gates of 300 m starting at 1.2 km. The transmitter pulse width used was 600 m, so this represents an oversampling of the pulse volume. This yielded a sampling range of 1.2 to 20.1 km. A detailed discussion of the FCA technique, including acceptance criteria is given in the papers of Holdsworth and Reid [1997, 2004] and Holdsworth [1999b].

[15] Operation in low mode was similar, except that a PRF of 10 kHz was utilized, along with 1024 coherent integrations, yielding a time series of 440 points with a sampling interval of 0.1024 s. Thus FCA analysis was performed on a time series of 45.06 s duration in this mode. The transmitter pulse was 150 m, oversampled at 100 m between 0.7 and 6.7 km. The parameters for both modes are summarized in Table 2.

Table 2. Standard Spaced Antenna Operating Parameters
ParameterDescription
High Mode
Transmitter pulse width600 m
Codingnoncoded pulses
Sampling resolution300 m
Sampling range1.2–20.1 km (63 heights)
PRF4096 Hz
Number of coherent integrations512
Number of points in time series 360 points
Time series sampling interval0.125 s
 
Low Mode
Transmitter pulse width150 m
Codingnoncoded pulses
Sampling resolution100 m
Sampling range0.7–6.7 km (60 heights)
PRF10,000 Hz
Number of coherent integrations1024
Number of points in time series440 points
Time series sampling interval0.1024 s

[16] All analyses are implemented in real time on the system, and are available online. Data in the bureau display format are exported into the Bureau network as 15, 30 and 60 min averages of the mean horizontal and vertical winds, the SNR, and acceptance rates for each range bin for use in the Bureau Australian Integrated Forecasting System (AIFS).

3. Results

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Radar
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

3.1. Radiosonde Data

[17] Routine balloon flights are made from the Mount Gambier Meteorological Office. Over the period between 1998 and 2000, which we discuss here, between two and four flights were made per day. The weather balloons were tracked with a WF-44 radar (S band, 10 cm wavelength). Winds were calculated and smoothed using the Vaisala RWIND software from azimuth, elevation and slant range data gathered at 2 s intervals. The zonal and meridional components of the balloon wind data were averaged over all data points within the same height range to produce a representative averaged wind for comparison with the profiler data. The profiler data used were the 60 min data averages for the period best corresponding to the radiosonde flight time. For operational use, the Bureau display software filters out hourly averaged wind estimates that are calculated from data with less than 33% of possible acceptances within the hour. However, for the purposes of completeness, we have included all hourly averages in this comparison, finding that the conclusions are little changed by their inclusion.

3.2. Hybrid Doppler Interferometer Data

[18] This mode of operation was not used routinely on the system, and we only present a brief intercomparison of the results from the HDI and FCA techniques here. More detailed comparisons will be presented elsewhere. When both SA and HDI modes are required, the acquisition sequence involves a 60 s vertical beam SA experiment, followed by a 30 s pulse-to-pulse steered EW 7° HDI experiment, followed by 30 s pulse-to-pulse steered NS 7° HDI experiment. The SA data are also used for vertical beam HDI analysis.

[19] This sequence allows SA and HDI estimates to be determined every 2 min. A typical example of the SA and HDI velocity profiles obtained over a half hour period is shown in Figure 2. The agreement between the two techniques is very good. The velocity magnitudes and directions generally agree to within 3 m s−1 and 10° respectively, at all heights.

image

Figure 2. Half-hourly averaged (left) velocity magnitudes and (right) directions obtained for the hybrid Doppler interferometer Doppler beam steering (HDI-DBS) technique (line with diamonds) and the spaced antenna full correlation analysis (SA-FCA) (line with squares) for the period between 0430 and 0505 UT on 5 May 1998.

Download figure to PowerPoint

[20] The effective beam zeniths and aspect sensitivity parameters estimates for the 7° beams are shown in Figure 3. The effective beam zenith is significantly smaller than the transmit beam zenith of 7° thereby confirming the need to use the TDI analysis. Inspection of these diagrams indicates that this region of the atmosphere can be very aspect sensitive. Such results are routinely observed and are similar to those reported by Hobbs et al. [2001] for VHF radar observations using another ST near Adelaide (35°S).

image

Figure 3. Half-hourly hybrid Doppler interferometer (left) derived effective beam direction θe and (right) aspect sensitivity parameter θs estimates for the period between 0430 and 0505 UT on 5 May 1998. The θs parameter corresponds to the half width of the angular polar diagram of the backscattering irregularities. A small value corresponds to more aspect sensitive backscatter. These results indicate the very aspect sensitive nature of the scattering irregularities in this region of the atmosphere and the variability of this parameter with height. Note that the effective beam direction is more than one half beam width closer to the zenith than the apparent beam direction at 5.5 km.

Download figure to PowerPoint

3.3. Spaced Antenna Acceptance Rates

[21] Figure 4 (left) shows the data acceptance rate as a height time intensity plot in terms of the percentage of 1 min wind determinations per hour for the period from the beginning of 1998 until the end of 2001 in the 1 to 15 km height interval for spaced antenna high mode. Data acceptance generally falls off with height, with acceptances reaching about 50% at 7.5 to 8 km. In the 9 to 13 km height region, values vary between from 10 and 30%, maximizing near the tropopause between 11 and 12 km. Figure 4 (right) shows height profiles of the acceptance percentage rates for both high and low modes of operation.

image

Figure 4. (left) Height-time intensity plot of the percentage of accepted 1 min FCA wind determinations in each hour of observation for the period from 1998 to 2000 for spaced antenna high mode. (right) Height profiles of the acceptance rates for hourly average FCA winds for both high and low modes of operation. Note the general falloff in acceptances with height in high mode which continues until about 9 km, after which the number of acceptances is steady at around 10% until 12 km.

Download figure to PowerPoint

3.4. Comparisons of Full Correlation Analysis and Radiosonde Winds

[22] For the observational period discussed in this paper, there are three balloon launches a day from the Mount Gambier site. Radiosonde results in the 2 to 14 km height range were binned into 1 km height averages and an hourly average calculated. Hourly averaged wind profiles of 1 km height averaged winds from the profiler were calculated for the hour nearest the sonde launch. A total regression fit, which assumes errors in both techniques, indicates a slope of 0.93 and an intercept of −0.1 for the total of 18,994 points in the entire 2–14 km height interval. Similar fits for the wind directions give the slope as 1.0 and the intercept as 0.8. Figures 5 and 6 show scatterplots of the radar and radiosonde wind speeds and directions for the 2 to 6 km and 6.1 to 14 km height ranges respectively. The results are consistent with those for the entire height range, and indicate that the radar wind speeds are smaller than the radiosonde wind speeds, and that there is a bias in the direction of about one degree. Like other radars in the Australian Bureau of Meteorology network, the Mount Gambier wind finding radar has been accurately calibrated against a target of known location. In contrast, the Mount Gambier profiler antenna array was aligned during installation using a compass. This suggests the Mount Gambier profiler radar array is misaligned with respect to the wind finding radar by about 1 degree, and that it is the Mount Gambier array that is aligned incorrectly. The bias in the wind speed is discussed below.

image

Figure 5. Comparison of FCA wind magnitudes and directions in the 2–6 km height range derived from simultaneous radar measurements and 3000 radiosonde ascents. The dashed lines are for reference and indicate a slope of 1. The slope of the regression line for the radar versus radiosonde wind direction is 1.0 ± 0.1, with an intercept of 0.8 ± 0.1. The slope of the regression line for the radar versus radiosonde wind magnitude is 0.93 ± 0.07, with an intercept of −0.10 ± 0.02.

Download figure to PowerPoint

image

Figure 6. Comparison of FCA wind magnitudes and directions in the 6.1–14 km height range derived from simultaneous radar measurements and 3000 radiosonde ascents. The dashed lines are for reference and indicate a slope of 1. The slope of the regression line for the radar versus radiosonde wind direction is 1.0 ± 0.1, with an intercept of −3.1 ± 0.1. The slope of the regression line for the radar versus radiosonde wind magnitude is 0.98 ± 0.10, with an intercept of −0.9 ± 0.03.

Download figure to PowerPoint

[23] Hocking et al. [2001] discuss limitations of least squares fitting to variables with different uncertainties, and suggest an analysis procedure which was adopted by Holdsworth and Reid [2004] when comparing MF FCA and imaging Doppler interferometric (IDI) upper atmosphere winds. We refer to this paper here because the authors found the FCA wind magnitudes to be around 10% less than the IDI winds. We will discuss this further in section 4 below. The relative uncertainties in the measurements being compared here are unknown, and because of this, a histogram approach was implemented as an additional method of intercomparison as illustrated in Figure 7. The histograms of the wind speed differences and direction difference presented as a number of data points between 2 and 14 km. Inspection of Figure 7 confirms the conclusions drawn from Figures 5 and 6. The histograms are close to normally distributed, with about a 1 degree bias in the direction, and an underestimation of the radar wind speeds with respect to the radiosonde wind speeds of about 1.1 m s−1.

image

Figure 7. Radiosonde/radar FCA (top) wind direction differences and (bottom) wind speed differences presented as a number of data points in the height interval 2–14 km. These results indicate a bias in both direction and magnitude. The direction difference takes a mean value of 0.85° which we attribute to the misalignment of the VHF radar array. The radar wind speeds underestimate the radiosonde wind speeds, with the mean difference being 1.1 m s−1.

Download figure to PowerPoint

[24] Figure 8 shows the histograms of radiosonde/radar speed ratios and direction differences in the 2 to 6 km and 6.1 to 14 km height intervals. Once again, the radiosonde/radar speed ratio obtained is consistent with the results from the regression method, taking a mean value of 0.9 for both height ranges. The mean value of the direction difference is about 0.7° in the lower range and about 1.0° in the upper range.

image

Figure 8. Radar/radiosonde FCA wind speed ratios and direction differences presented as a percentage of available coincidences in the height interval (left) 2–6 km and (right) 6.1–14 km. In the lower height range, the wind speed ratio takes a mean value of 0.91 with a median value of 0.93. In the upper height range, the same ratio takes a mean value of 0.90 with a median value of 0.93. The direction differences take a mean value of 0.65° and a median value of 0.61° in the lower height range and a mean value of 0.96° and a median value of 1.04° in the upper height range.

Download figure to PowerPoint

[25] The comparisons between radar and radiosonde wind speeds/directions are extended in Figure 9, which illustrates the vertical profiles of the mean values of the radar/radiosonde wind speed ratios and direction differences. The error bars in Figure 9 (left) indicate the standard deviations of the population wind speed ratios. Figure 9 (right) shows the vertical profiles of the radiosonde-radar wind direction differences. Results are generally consistent through height, with a tendency for the direction differences to increase with height. Figure 9 also indicates that the mean direction difference is about 1 degree.

image

Figure 9. Vertical profiles of the median of (left) the wind speed ratios and (right) the radiosonde-radar direction differences. The speed ratios take mean values from 0.85 at 2.5 km to 0.95 at 14.5 km. Direction differences show behavior consistent with the results from Figure 8, with values generally increasing with height, with values of about 0.5° at 3.5 km to values of about 1.4° at 14.5 km.

Download figure to PowerPoint

[26] Figure 10 shows the mean magnitude difference between radar and sonde measured winds as function of height (left) and as function of sonde speed (right). The increase of magnitude differences with height agrees with the measurements of slope obtained from the scatterplots. The underestimation is found to depend on the actual wind velocity, increasing with increasing wind speed as indicated by the sonde. Since larger wind speeds tend to occur at larger heights, this result is consistent with a larger separation between the profiler and the sonde in this case.

image

Figure 10. Profiles of magnitude differences between radiosonde and FCA radar winds as a function of (left) height and (right) sonde speed. The velocity differences tend to be consistent with height but do increase with increasing wind speed as measured by the sonde. The largest sonde values correspond to values that only occur at upper heights (see Figure 6) and so correspond to greater distances between the profiler measurements at the launch site and the sonde location.

Download figure to PowerPoint

[27] Figure 11 shows the root mean square (RMS) difference between radar and sonde wind speeds and wind directions as a function of height. The RMS differences in the speed between the two estimates range from 2 to 3 m s−1. The RMS difference between the sonde and radar wind directions is just over 6 degrees. The RMS error for sonde measurements is generally around 2 m s−1 [see, e.g., Dibbern et al., 2003]. We conclude that the sonde and radar wind estimates have similar RMS errors, but that there is a bias in the profiler estimates of the wind magnitude.

image

Figure 11. Profiles of root mean square (RMS) differences between radiosonde and FCA (left) radar wind speeds and (right) directions as a function of height. RMS values of the magnitudes all lie in the range 2–3 m s−1, with a tendency to increase with height. The majority of the RMS direction differences lie in the range 5–7°, with larger values at the lowest and greatest heights.

Download figure to PowerPoint

4. Discussion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Radar
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

[28] The FCA SA results presented here are highly correlated with the radiosonde wind speeds and directions, underestimate the radiosonde wind speeds, but are in excellent agreement with the radiosonde wind directions. This result appears to be typical, but the extent of the underestimation may depend on the antenna array configuration. Using a rather more limited data set and three groups of four antennas, Vincent et al. [1998] found a tendency for Buckland Park (BP) boundary layer radar FCA winds to underestimate radiosonde winds, with a correlation coefficient of 0.91 between sonde and boundary layer radar wind directions and a correlation of 0.85 in the magnitudes. Also using a limited data set, Hobbs [1998] found a similar result, with a correlation of 0.88 between DBS and FCA meridional wind velocities using 3 subgroups of the BP VHF ST radar coaxial-collinear antenna [see Hobbs et al., 2001], each 50 × 50 m in size.

[29] Klein Baltink and Reid [2000] found that with three groups of nine antennas, BLT derived winds underestimated radiosonde winds with a slope of 0.85, and also that they underestimated UHF BL Doppler winds with a slope of 0.87. They also noted that the UHF profiler underestimated the radiosonde winds, with a slope of 0.94. Klein Baltink [1998] found that the same UHF profiler underestimated tower winds by about 5%. Using the same antenna arrangement, MacKinnon [2001] compared BP VHF BLT SA radar wind magnitudes with GPS sondes launched from the BP site, and found a slope of 0.93. He found a slope of 0.89 when they were compared with radiosondes launched from Adelaide airport, about 40 km distant from BP, and a slope of 0.96 when compared with wind measurements from a GROB G109B aircraft operating near the radar. Also using the same antenna arrangement, Vincent et al. [2004] found a correlation of 0.91 when comparing BLT FCA winds with radiosonde winds during the DARWEX campaign at Darwin (12°S). A preliminary analysis of Sydney VHF BLT (same antenna configuration) wind speeds indicates a correlation of 0.88 when are compared to those from sondes launched from the same location. The reasons that the FCA tends to underestimate the wind speed are discussed in detail by Holdsworth [1999b], but briefly, result from the tendency of factors such as noise and antenna coupling to reduce cross correlation values, and as a consequence, wind speeds.

[30] The RMS errors between the techniques used in the present study are around 2 m s−1. This turns out to be a typical value for the studies referred to above, and also typical of the value obtained when sondes are compared with sondes. The bias in the wind speed appears to be consistent across the entire period of the intercomparisons featured here, and although somewhat unappealing, we suggest that for operational use, that a scaling factor be applied to calibrate the FCA of SA data for to the radiosonde winds.

[31] The experience with the Mount Gambier profiler has suggested a number of design considerations for an operational profiler based on the SA technique. The system has proven to be very reliable, and tolerant of antenna degradation and the loss of power blocks. As we have noted, the Mount Gambier system is a compromise, and a dedicated SA system would only need three noncolinear receiving arrays. This would halve the number of antennas required. Of course, the antenna arrangement would also need to be such that there were three groups on the vertices of an equilateral triangle.

[32] Whether the present power and antenna configuration are appropriate for the application needs to be further discussed. The height coverage in SA mode is generally not adequate for the requirements of the Australian Bureau of Meteorology, in that observations are desired to heights of around 18 km. Trials using the DBS technique indicated that it does have the potential to extend the height coverage for the same power aperture product, at least for hourly averages and in some conditions. This is because it is easier to average spectra, to apply time and spatial consistency checks to the spectral peaks, and to apply effective outlier rejection algorithms to the Doppler spectra [see, e.g., Merritt, 1995], which are directly related to the radial velocities, than to the analysis of SA data, which depends on the cross correlation or cross spectral analysis of the raw time series.

[33] What is striking is the ability of small VHF systems to obtain results in the BLT region. For example, the height coverage for the Sydney BLT system [Lawrence and Jardine, 2000] obtains 50% coverage to a height of between 4 and 5 km. Although the conditions at Sydney are different to those at Mount Gambier, the results do illustrate the capabilities of a small system, with a power aperture product of 3.5 × 104 Wm2. An identical system operated during the wet season component of the DARWEX campaign, obtained 50% coverage at 6 km [Vincent et al., 2004]. This is because these radars make good use of the gradients in the humidity which dominate the radar refractive index in the lowest part of the atmosphere, and only moderate powers are required to obtain such coverage. The Adelaide BLT radar, with an even lower power aperture product of 4.7 × 103 Wm2 (corresponding to a peak power of 1 kW) operating in conditions similar to those at Mount Gambier, reaches an acceptance rate of about 50% at 2.5 km [MacKinnon, 2001]. If we exclude the recovery in power that occurs near the tropopause, as a very rough rule of thumb, each doubling of power gains about an additional kilometer of height coverage, all else being equal. Application of pulse coding would improve this.

[34] We suggest that the most appropriate technique for a VHF profiler with an array size greater than around 8 to 10 wavelengths diameter is the DBS technique, applied with a single receiver channel and off-zenith beam angles of around 15°. This was the configuration used with the NOAA VHF profilers installed across the Pacific [see, e.g., Gage et al., 1994], and was the design adopted for both the Woomera VHF ST radar system [see, e.g., MacKinnon, 2003], and the Vaisala LAP-12000 wind profiler installed for the UKMO in northern Scotland [see, e.g., Winston, 2004]. Of course, successful application in DBS mode requires the appropriate feed arrangements so that the antennas are less than one wavelength apart, so the grating lobes that are an issue on the Mount Gambier system are not present. When observations are only required to altitudes of 6 to 8 km, the SA technique applied on small (27 antenna) arrays with peak powers of around 10 kW may be the most appropriate technology.

5. Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Radar
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

[35] In the present study, we have used three years of wind data from the Mount Gambier profiler to compare with winds from over 3000 radiosonde ascents from the same site. The results suggest excellent agreement in direction as determined by both techniques, and a bias in the SA winds, which leads them to underestimate the radiosonde winds by about 10%.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Radar
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

[36] The development and operation of the Mount Gambier prototype operational wind profiler was supported by Australian Research Council Industry collaboration grant C195301238, by the Australian Bureau of Meteorology, by Atmospheric Radar Systems Pty Ltd. (ATRAD), and by the University of Adelaide.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Radar
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information
  • Briggs, B. H. (1984), The analysis of spaced sensor records by correlation techniques, in Handbook for MAP, vol. 13, edited by C. H. Liu, and B. Edwards, pp. 166186, Sci. Comm. on Sol. Terr. Phys. Secr., Univ. of Ill., Urbana.
  • Chilson, P. B., S. Kirkwood, and A. Nilsson (1999), The Esrange MST radar: A brief introduction and procedure for range validation using balloons, Radio Sci., 34, 427436.
  • Cohn, S. A., W. O. J. Brown, C. L. Martin, M. E. Susedik, G. D. Maclean, and D. B. Parsons (2001), Clear air boundary layer spaced antenna wind measurement with the multiple antenna profiler (MAPR), Ann. Geophys., 19(8), 845854.
  • Dibbern, J., W. Monna, J. Nash, and G. Peters (Eds.) (2003), EUR 20614-COST action 76—Development of VHF/UHF wind profilers and vertical sounders for use in European observing systems, final report, 433 pp., Off. for Off. Publ. of the Eur. Communities, Luxembourg.
  • Gage, K. S., J. R. McAfee, W. L. Ecklund, D. A. Carter, C. R. Williams, P. E. Johnston, and A. C. Riddle (1994), The Christmas Island wind profiler: A prototype VHF wind-profiling Doppler radar for the tropics, J. Atmos. Oceanic Technol., 11, 2231.
  • Hobbs, B. G. (1998), Measurements of tropospheric scatter with a new multi-receiver VHF Doppler radar, Ph.D. thesis, 228 pp., Univ. of Adelaide, Adelaide, S. Aust., Australia. (Available at http://adt.caul.edu.au).
  • Hobbs, B. G., I. M. Reid, and D. A. Holdsworth (2001), A comparison of tropospheric VHF Doppler beam steering and spaced antenna measurements of aspect sensitivity, Radio Sci., 36, 955964.
  • Hocking, W. K., R. Ruester, and P. Czechowsky (1986), Absolute reflectivities and aspect sensitivity of VHF radio wave scatterers measured with the SOUSY radar, J. Atmos. Terr. Phys., 48, 131144.
  • Hocking, W. K., T. Thayaparan, and S. J. Franke (2001), Method for statistical comparison of geophysical data by multiple instruments which have differing accuracies, Adv. Space Res., 27, 10891098.
  • Holdsworth, D. A. (1999a), Spatial correlation analysis revisited: Theory and application to “radar backscatter model” data, Radio Sci., 34, 629641.
  • Holdsworth, D. A. (1999b), Influence of instrumental effects upon the full correlation analysis, Radio Sci., 34, 643655.
  • Holdsworth, D. A., and I. M. Reid (1997), An investigation of biases in the full correlation analysis technique, Adv. Space Res., 20, 12691272.
  • Holdsworth, D. A., and I. M. Reid (2004), Comparisons of full correlation analysis (FCA) and imaging Doppler interferometry (IDI) winds using the Buckland Park MF radar, Ann. Geophys., 22, 38293842.
  • Klein Baltink, H. (1998), A long term intercomparison of windprofiler/RASS and tower measurements, Meteorol. Z., 7, 271279.
  • Klein Baltink, H., and I. M. Reid (2000), CaPrix: An intercomparison of a VHF and UHF lower tropospheric windprofiler, paper presented at 5th International Symposium on Tropospheric Profiling, Natl. Cent. for Atmos. Res., Adelaide, S. Aust., Australia, Dec.
  • Kudeki, E., and R. F. Woodman (1990), A post-statistics steering technique for MST radar applications, Radio Sci., 25, 591594.
  • Latteck, R., and W. Singer (2001), Multi-beam radar observations of polar mesosphere summer echoes during the MIDAS/DROPPS/MINIDUSTY campaign at Andenes, Norway in July 1999, Adv. Space Res., 28, 10651070.
  • Lawrence, S., and R. Jardine (2000), Use of boundary layer profiler data for aviation forecasting, paper presented at 5th International Symposium on Tropospheric Profiling, Natl. Cent. for Atmos. Res., Adelaide, S. Aust., Australia, Dec.
  • MacKinnon, A. D. (2001), VHF boundary layer radar and RASS, Ph.D. thesis, 266 pp., Univ. of Adelaide, Adelaide, S. Aust., Australia. (Available at http://adt.caul.edu.au).
  • MacKinnon, A. D. (2003), Radar results compared with radiosonde measurements: Woomera VHF DBS, Sci. Rep. 10, 12 pp., Atmos. Radar Syst., Thebarton, S. Aust., Australia.
  • May, P. T. (1990), Spaced antenna versus Doppler radars: A comparison of techniques revisited, Radio Sci., 25, 11111119.
  • Merritt, D. A. (1995), A statistical averaging method for wind profiler Doppler spectra, J. Atmos. Oceanic Technol., 12, 985995.
  • Morris, R. J., D. J. Murphy, I. M. Reid, and R. A. Vincent (2004), First polar mesosphere summer echoes observed at Davis, Antarctica (68.6°), Geophys. Res. Lett., 31, L16111, doi:10.1029/2004GL020352.
  • Reid, I. M., B. G. W. Vandepeer, S. C. Dillon, and B. M. Fuller (1995), The new Adelaide medium frequency Doppler radar, Radio Sci., 30, 11771190.
  • Röttger, J., and H. M. Ierkic (1985), Postset beam steering and interferometer applications of VHF radars to study winds, waves and turbulence in the lower and middle atmosphere, Radio Sci., 20, 14611480.
  • Vandepeer, B. G. W., and I. M. Reid (1995), Some preliminary results obtained with the new Adelaide MF Doppler radar, Radio Sci., 30, 11911203.
  • Vincent, R. A., S. Dullaway, A. MacKinnon, I. M. Reid, F. Zink, P. T. May, and B. H. Johnson (1998), A VHF boundary layer radar: First results, Radio Sci., 33, 845860.
  • Vincent, R. A., A. D. MacKinnon, I. M. Reid, and J. M. Alexander (2004), VHF profiler observations of winds and waves in the troposphere during DARWEX, J. Geophys. Res., 109, D20S02, doi:10.1029/2004JD004714.
  • Winston, H. (2004), Vaisala wind profilers support: Modernization of upper-air data collection network in the UK, Vaisala News, 164, 2425.
  • Woodman, R. F., and A. Guillen (1974), Radar observations of winds and turbulence in stratosphere and mesosphere, J. Atmos. Sci., 31, 493505.
  • Zecha, M., J. Bremer, R. Latteck, W. Singer, and P. Hoffmann (2003), Properties of midlatitude mesosphere summer echoes after three seasons of VHF radar observations at 54°N, J. Geophys. Res., 108(D8), 9439, doi:10.1029/2002JD002442.

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Radar
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information
FilenameFormatSizeDescription
rds5093-sup-0001-t01.txtplain text document1KTab-delimited Table 1.
rds5093-sup-0002-t02.txtplain text document1KTab-delimited Table 2.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.