The Dines pressure tube anemometer was the primary wind speed recording instrument used in Australia until it was replaced by Synchrotac cup anemometers in the 1990s. Simultaneous observations of the gust wind speeds recorded using both types of anemometers during tropical cyclones have, however, raised questions about the equivalency of the gust wind speeds recorded using the two instruments. An experimental study of the response of both versions of the Dines anemometer used in Australia shows that the response of the anemometer is dominated by the motion of the float manometer used to record the wind speed. The amplitude response function shows the presence of two resonant peaks, with the amplitude and frequency of the peaks depending on the instrument version and the mean wind speed. Comparison of the gust wind speeds recorded using Dines and Synchrotac anemometers using random process and linear system theory shows that, on average, the low-speed Dines anemometer records values 2%–5% higher than those recorded using a Synchrotac anemometer under the same conditions, while the high-speed Dines anemometer records values 3%–7% higher, depending on the mean wind speed and turbulence intensity. These differences are exacerbated with the adoption of the WMO-recommended 3-s moving average gust wind speed when reporting the Synchrotac anemometer gust wind speeds, rising to 6%–12% and 11%–19% for low- and high-speed Dines anemometers, respectively. These results are consistent with both field observations and an independent extreme value analysis of simultaneously observed gust wind speeds at seven sites in northern Australia.
Long-term historical records of mean and gust wind speeds are essential in a number of applications, including the determination of the design wind speeds used in wind loading codes and standards, and the assessment of possible effects of projected climate changes. One of the issues in dealing with such records is that wind speeds are impacted by a considerably larger range of factors than other meteorological variables, such as temperature. These factors include the observing site surroundings, both immediate and upwind of the site, the anemometer height and type used, as well as the observing practices used over the period of the record. Changes in one or more of these factors over time lead to wind speed records with changing characteristics over the period of the record, which then need to be taken into account during any analysis of the record. This is particularly true of gust wind speeds, which are extremely sensitive to factors such as the anemometer response characteristics.
While the need to correct observed mean and gust mean wind speeds for differing terrain conditions both at and upstream of the anemometer site has long been recognized, the issue of correcting for changes in both instrumentation type and observing practices that have taken place over the period of the record at a particular site has often been overlooked. Several different approaches have been taken in the past to account for the latter effects following changes in either instrumentation type or observing practices, or indeed the occurrence of both simultaneously at a particular observing site. The simplest of these approaches involves comparing the mean or gust wind speed measurements made before the changeover with those made afterward, before calculating a correction factor based on some measure of the difference between the two sets of wind speed records (see, e.g., Smith 1981; Department of the Environment, Heritage and Local Government 2009). A slightly more sophisticated approach is to base the correction factors on simultaneous measurements made by collocated instruments over some period of time before phasing out the earlier instrument (Logue 1986). Finally, several authors have used an approach based on that of Davenport (1964), in which random process and linear system theory is used in combination with knowledge of the frequency response characteristics of both the instrumentation types and observing practices being used to calculate a set of correction factors to be applied to the corresponding wind speed records. Using the latter approach, Masters et al. (2010) determined correction factors for Automated Surface Observing System (ASOS) stations in the United States following the changeover from Belfort cup anemometers to Vaisala ultrasonic anemometers in combination with the adoption of the World Meteorological Organization's (WMO) 3-s moving average gust wind speed definition (World Meteorological Organization 2008) in place of the 5-s block-averaged gust definition that was used previously.
In Australia, the Dines anemometer was the main wind speed recording instrument for over 50 years, before being replaced by Automatic Weather Stations (AWSs) with Synchrotac 706 cup anemometers in the 1990s. At selected stations in regions of Australia likely to be impacted by tropical cyclones, however, the Dines anemometers were retained as backups to the newer Synchrotac cup anemometers because of their ability to still continue recording using a clockwork-driven chart recorder should the AWS power supply fail during severe weather. In several cases where both Dines and Synchrotac anemometers continued to record during a landfalling tropical cyclone, it was noted that the Dines anemometers recorded peak gust wind speeds that were considerably higher than the gust wind speeds simultaneously recorded by the Synchrotac cup anemometers. Reardon et al. (1999) noted that at Learmonth Airport, Western Australia, during the landfall of Cyclone Vance in March 1999, a Dines anemometer recorded a peak gust wind speed of 74 m s−1 while a Synchrotac cup anemometer located less than 25 m away recorded a peak gust of 63 m s−1, a difference of over 17% between the two observations. Similar observations have subsequently been made during the landfalls of Cyclone Tessi in April 2000 and Severe Cyclone Yasi in February 2011. In the case of the latter, the Synchrotac cup anemometer at Townsville Airport, Queensland, Australia, registered a gust of 38 m s−1, but a Dines anemometer at the same site concurrently recorded a gust of 45 m s−1, the difference between the two values being over 18% (Boughton et al. 2011).
Similar differences have been noted in other countries where the Dines anemometer has been the main wind recording instrument until being replaced in the 1990s by cup anemometers with the introduction of AWSs. In reviewing the available wind speed data for South Africa, Kruger (2011) noted that the Dines anemometers at all sites considered recorded consistently higher peak gust wind speeds than those recorded at the same sites after 1992, when the Dines anemometers were replaced by AWSs equipped with cup anemometers. Without an adequate explanation for the differences, it was decided to discard all wind speed data recorded prior to 1992 using Dines anemometers, and to base an analysis of the South African design wind climate on the shorter post-1992 wind speed record, rather than using a longer record with the attendant problem of trying to standardize the gust wind speed measurements to account for the observed differences between the Dines and cup anemometers. The same issue arose in Ireland when determining the design wind speeds for use with the Irish National Annex to the Eurocode (Department of the Environment, Heritage and Local Government 2009), and where Dines anemometers were once again in use until the 1990s, when they were replaced by either Vaisala or Vector cup anemometers. An extreme value analysis of the gust wind speeds recorded using each anemometer type led to the conclusion that the mode of the annual extremes of the Dines anemometer was some 10% larger than the mode of the Vaisala cup anemometer predictions, while the difference between modes for the Vector cup anemometers was even higher at 16%.
In this paper we consider the response of the Dines anemometer to gusts, and how the measured gusts relate to those recorded using the Synchrotac 706 cup anemometers currently in use at Bureau of Meteorology AWS sites. The next section provides a brief review of the principles of the Dines anemometer and the history of its use in Australia, together with the research that has been published previously on its response to gusting winds. This is followed by a description of the experimental system used to explore the response of the Dines anemometer to varying wind speeds, and the results of this exploration. Finally, using the experimentally determined response curves, a comparison is made of the performance of the Dines anemometer relative to the Synchrotac cup anemometers currently in use, before presenting the conclusions of the present study.
2. The Dines anemometer
The Dines pressure tube anemometer is a robust pitot-static device that measures the difference in pressure between moving air brought to rest and a reference static pressure to determine the wind speed. The basic components of the Dines anemometer are a head, consisting of an open tube kept facing into the wind by a vane at the rear of the head, and a specially designed float manometer that sits in a chamber containing water. Sectional views of both components are shown in Fig. 1, with more detailed cross sections being found in Gold (1936). Figure 2 shows a typical Dines installation in Australia, where the head is mounted at a height of 10 m above ground on a lattice mast above a hut that contains the float manometer and recording instrumentation, together with a close-up view of the float manometer and recording instrumentation located in the hut at the base of the mast. The open tube of the head measures the total stagnation pressure on the assumption that the head is aligned with the flow by the tail vane while a series of staggered holes on the vertical mounting just below the head measures the reference static pressure. Both pressure measurements are then transmitted through an appropriate length of 1-in. (2.54 cm) tubing to the float manometer located in the hut at the base of the mast supporting the head. The total stagnation pressure is applied to the interior of the float, while the reference static pressure is applied to the space above the float. The shape of the float is designed so that it moves in proportion to changes in , where ΔP is the difference in pressure between the stagnation and static pressures. The result is that the movement of the float then scales linearly with the velocity recorded at the mouth of the open tube of the head. Further information on the design of the float, and in particular the determination of its shape, may be found in Gold (1936). The movement of the float is then either recorded using a chart recorder mounted directly on top of the float chamber or electrically, so that the float output can be recorded remotely.
Two main types of float manometer have been used in Australia. The first of these is the standard, or low-speed, type described by Giblett (1932) and Gold (1936) and which was initially used at all observing sites in Australia. Subsequently, it was found that when wind speeds exceeded approximately 68 m s−1, air bubbles tended to escape from under the bottom of the float, leading to a nonlinear response of the float and erratic vertical movements. This led to the development of a modified float manometer with a taller chamber and heavier float by R. W. Munro, the makers of the standard Dines anemometer, to increase the maximum wind speed able to be recorded by the system, particularly under tropical cyclone conditions. This high-speed Dines anemometer was subsequently used from the mid-1970s onward at observing sites in regions of Australia affected by tropical cyclones, while the low-speed system was retained at observing sites located outside of this region. A third intermediate type of float was developed in the early 1970s (Young 1987), prior to the adoption of the high-speed float type, and used at several observing sites; however, very little is known about the design and use of this intermediate type of float.
In general, the assumption has been that the Dines anemometer records a peak gust wind speed with an averaging time of the order of 2–3 s. In an Australian context, this assumption would appear to date to Whittingham (1964) who, on the basis of information on the response of the float manometer alone presented in Giblett (1932), concluded that the Dines anemometer gave a good indication of the speed of strong gusts of 2–3-s duration. The data contained in Giblett (1932) were obtained in the course of an investigation into the characteristics of the Dines anemometer undertaken by the National Physical Laboratory (NPL) in the United Kingdom between 1925 and 1934. Two key findings of the initial investigation were that the use of 1-in.-diameter connecting tubing gave a significantly improved frequency response when compared to the ½-in. (1.27 cm)-diameter tubing then in use, and that some form of shielding was required for the total stagnation and reference static tubing connections at the base of the head to avoid significant variations in the static pressure with wind direction due to interference effects from the tubing. The summary contained in Giblett (1932) is the only formally published record of the major findings of this investigation, with some of this information subsequently being reproduced in Meteorological Office (1956).
In considering the response of the Dines anemometer to fluctuating wind speeds, it has also typically been assumed that the frequency response is a first-order function, not dissimilar to those obtained for cup and propeller anemometers, where the response monotonically decreases as the frequency increases. Goldie (1935) notes, however, that as a result of discussions about wind speeds obtained using a Dines anemometer at Bell Rock Lighthouse, experiments were undertaken in a wind tunnel at NPL in the early 1930s to compare the response of a Dines anemometer with simultaneous hot-wire anemometer measurements of the same fluctuating wind speeds. Rather unexpectedly these experiments showed the presence of a naturally occurring resonant frequency in the response to the Dines anemometer. For the two mean wind speeds considered in the tests, the peak response occurred at periods of the order of 5–7 s, with the amplitude of the associated peak decreasing with increasing mean wind speed.
Borges (1968) examined the frequency response of a Fuess 82a universal anemometer, similar in design and function to the Dines anemometer, and identified two separate resonant peaks in the frequency response function with periods of less than 5 s. For a mean wind speed of 20 m s−1 and a 6-m length of connecting pipe, a single peak was identified with a magnitude of the order of 2.1 between 3 and 5 s, while for a 2-m length of connecting pipe the magnitude of the initial peak response was reduced to around 1.7; however, a second resonant peak was apparent between 1.25 and 1.43 s. At a mean wind speed of 40 m s−1, the magnitude of the first resonant peak was reduced to around 1.4; conversely, the magnitude of the second resonant peak was increased, particularly for the shortest length of connecting pipe considered, and was in excess of 2.5. The resonant behavior observed was attributed to the interaction of the float and the water in the chamber in which it rests and the fact that the two elements form a 2-degree-of-freedom system, the first peak corresponding to a mode in which the float and water are moving in phase with each other and the second peak corresponding to a mode in which they are moving in opposition to each other.
A number of intercomparisons between Dines and cup anemometers have been made over the years, most of which have been primarily concerned with the differences in the mean wind speeds reported by the two instruments—see, for example, Smith (1981) and several earlier papers from the 1950s referenced in this paper. In terms of gust wind speed comparisons, perhaps the most relevant study is that of Logue (1986), who compared both mean and gust wind speeds measured using a Dines anemometer collocated with a standard Met Office Mark II cup anemometer at the Irish Meteorological Service's Galway observing site over the entire year of 1984. Although some monthly variation of the ratio between both the mean and gust wind speeds recorded by the two instruments was observed, it was found that overall the mean wind speeds from the two instruments compared well. On the other hand, the cup anemometer significantly underestimated the gust wind speeds when compared to those obtained using the Dines anemometer, with the mean ratio between the two being 0.947 for gust wind speeds exceeding 28.3 m s−1 (55 kt, where 1 kt = 0.51 m s−1).
3. Experimental setup
Examples of both low-speed and high-speed Dines anemometers were tested in this study. A complete low-speed Dines anemometer, consisting of both the head/vane unit and low-speed float manometer was supplied by the Bureau of Meteorology to the Cyclone Testing Station at James Cook University in Townsville, Queensland. The Bureau of Meteorology also granted access to a still operational high-speed Dines anemometer located at Townsville Airport. In both cases we make the assumption that the anemometers tested are representative of a well-maintained, properly calibrated low- or high-speed Dines anemometer as used by the Bureau of Meteorology. Since the Dines anemometer uses pressure to measure the wind speed, both instruments were tested by connecting a pressure loading actuator (PLA) capable of applying either a steady or time-varying pressure trace through a polyvinyl chloride (PVC) manifold to the open tube on the head of the anemometer, as shown in Fig. 3. The PLA itself consists of a blower used to generate the required pressure increase, together with a special valve controlled by a servomotor that regulates the pressure supplied by the PLA outlet hose, and was originally designed to apply fluctuating pressures to full-scale structures to examine their response to spatially and temporally varying wind loads, as described by Kopp et al. (2010). The practical limit to the frequencies that the PLA can accurately reproduce time-varying input pressure time histories is stated by Kopp et al. (2010) to be 10 Hz, which is more than adequate for the purposes of the tests described in this paper.
Two Sensortechnics CTEM7N350GL0 pressure transducers were fitted to the manifold surrounding the head of the Dines anemometer, the first of which was used to provide feedback to the PLA controller to ensure that the PLA was correctly reproducing the desired input pressure time history, while the second transducer was used to determine the equivalent wind speed being applied to the head of the Dines anemometer. This was calculated using the following equation:
where ΔP is the difference between the pressure being applied to the Dines head and the reference static pressure, ρ is the air density, V is the wind speed, and K is a constant experimentally determined to be 1.49 for the Dines anemometer (Giblett 1932; Meteorological Office 1956).
The movement of the float was recorded using a Gefran PZ12A150 displacement transducer that provides an electrical output linearly proportional to the movement of a shaft that passes through a series of circular solenoidal coils. In normal operation, a short vertical rod is connected to the top of the float and used to support a pen that records the movement of the float on the chart recorder that is fixed to the top of the float chamber. The rod also supports a shot cup that is used to adjust the zero position of the float by the addition or removal of lead shot from the cup. The displacement transducer was connected via a small universal joint to the shot cup on top of the float rod, with care being taken to ensure that the weight of the universal joint, modified shot cup, and displacement transducer shaft matched the weight of the original shot-cup, lead shot, and pen. The movement of the displacement transducer was then recorded electronically before being converted into an equivalent wind speed value, through the use of an appropriate calibration curve for the type of float being considered.
Initial testing of the low-speed Dines anemometer focused on the response of the individual system components, that is, the head/vane unit, the float manometer, and the tubing connecting the two parts of the system. This also allowed any potential difficulties to be identified and then rectified in controlled laboratory conditions before conducting tests on the high-speed Dines in the field. A range of different pressure traces was applied, including steady-step increases in the pressure to check the calibration of both low- and high-speed Dines anemometers, sinusoidal pressure time histories with varying amplitudes and frequencies ranging from 0.2 to 6 Hz, and Gaussian white noise pressure time histories. In testing the high-speed Dines, the head of the low-speed unit was taken into the field and connected to the high-speed float manometer via an appropriate length of tubing. This removed the necessity of constructing a 10-m-high platform to bring the PLA up to the level of the high-speed anemometer head. Instead, the PLA could be connected to a similar head at ground level, which simplified field testing considerably.
For each test both the pressure applied to the head of the Dines anemometer and the movement of the float were recorded simultaneously at a sampling frequency of 100 Hz for typically 5 min. Prior to any processing, the recorded signals were first filtered using a first-order Butterworth filter with the half-power frequency set to 10 Hz to remove higher frequencies beyond the maximum frequency capable of being achieved by the PLA, along with the first 10 s of the record to allow the system to settle following the start of the test. The measured pressure being applied to the head of the anemometer was then converted into an equivalent wind speed using Eq. (1) in combination with a value of 1.23 kg m−3 for the air density ρ (Meteorological Office 1956), while the displacement transducer measurements of the float movement were converted into wind speed values using an appropriate calibration curve developed from applying steady-step pressures to the low- and high-speed Dines float manometers. The resulting curves are shown in Fig. 4 and yield linear relationships between the wind speed applied to the Dines anemometer head and the vertical movement of the float of 0.290 and 0.583 m s−1 mm−1 for low- and high-speed Dines anemometers, respectively. These values are virtually identical to those derived from the scaling on the paper charts used to record wind speeds, which for the low- and high-speed Dines anemometers works out to a vertical scale of 177 mm for maximum wind speeds of 100 and 200 kt, respectively, or 0.290 and 0.590 m s−1 mm−1, respectively.
The variation of the amplitude response of the Dines anemometer with frequency was then calculated using Welch's averaged modified periodogram method (Welch 1967) to obtain spectral estimates of the input and output wind speeds for frequencies ranging from 0.1 to 10 Hz, in steps of 0.1 Hz. This was achieved by splitting both input and output signals into a series of nonoverlapping 10-s-long segments, before applying a Hamming window to each segment and then calculating the modified periodogram for each windowed segment. Finally, the periodograms for all segments were averaged to obtain a spectral estimate for either the input or output wind speeds. The amplitude response curve was then calculated by dividing the output spectra by the input spectra and taking the square root of the result. The variation of the corresponding phase angle with frequency was obtained using the same method to calculate the cross-power spectral density function of the input and output signals, before dividing the result by the input power spectrum and forming a complex transfer function from which the phase angle could be calculated. For the Gaussian white noise pressure time histories, all frequencies between 0.1 and 10 Hz were considered when evaluating the response. For the sinusoidal pressure time histories, the response at the nominal frequency of the sinusoidal wave was extracted from the appropriate response curve, with all other frequencies being discarded. This approach was adopted in the case of the latter because the PLA is not capable of reproducing a pure sinusoidal wave at a specific frequency with exact fidelity, and there is some leakage into the adjacent frequencies that results in small nonzero responses being obtained for these frequencies. The magnitudes are, however, significantly reduced when compared to the response at the nominal frequency of the desired sinusoidal pressure time history.
In presenting the results, we first consider the response characteristics of the major components of the low-speed Dines anemometer, consisting of the anemometer head, the tubing connecting the anemometer head to the float manometer, and the float manometer, to establish which of the three components dominate the response of the entire system. We then consider the response of the low- and high-speed Dines anemometers and how this changes with both wind speed and anemometer type.
a. Anemometer components
Figure 5 shows the response of a 10-m length of 30-mm-diameter PVC tubing compared to that of a Dines anemometer head connected to an 8.1-m length of 30-mm-diameter PVC tubing, such that the total combined length of the head piping and the PVC tubing is 10 m. Both systems are connected to a dummy float chamber constructed from a 225-mm-long section of PVC pipe with an internal diameter of 100 mm, and designed to nominally represent the volume of the float chamber used in the low-speed Dines float manometer. The length of 10 m of tubing alone has a very clearly defined resonant peak at a frequency of 6.6 Hz, as evidenced by both the resonant peak in the amplitude response curve and the 180° change in the phase angle at the same frequency. The addition of the Dines anemometer head has the effect of shifting both the resonant peak and change in phase angle to a slightly lower frequency of 5.8 Hz; however, apart from a reduction in the magnitude of the resonant peak, there are no significant differences in the response characteristics of the two systems.
The addition of the float manometer to the system changes the response characteristics of the complete system significantly, as can be seen in Fig. 6. The amplitude response of the system is now dominated by the response of the float manometer, which shows two resonant peaks. The first of these is a broad peak at a frequency of 0.5 Hz, with a second, reduced, peak occurring at a frequency of 1.2 Hz. The effects of the float manometer response are also visible in the phase angle curve, where the phase angle steadily decreases to a value of −148° at a frequency of 0.8 Hz, before recovering to a value of −52° at a frequency of 1.1 Hz, before once again decreasing steadily beyond this frequency. The response of the tubing connecting the anemometer head to the float manometer is suppressed, although its presence is clearly visible in the phase angle curve, which again shows a 180° change at frequencies beyond 4.0 Hz.
The effect of removing the 10-m length of tubing between the anemometer head and float manometer and connecting them directly is shown in Fig. 7. In this case the initial broad low-frequency resonant peak associated with the float manometer is shifted to a slightly higher frequency of 0.6 Hz, while the second resonant peak associated with the float manometer response is relatively unchanged. The resonant peak at 6.5 Hz associated with the tubing response disappears, as does the 180° change in phase angle at frequencies higher than 4.0 Hz, which is also associated with the tubing response.
In considering the effect of the major components of the Dines anemometer system on the overall response characteristics of the complete system, we therefore conclude from the results presented above that over the range of frequencies likely to be of interest, the system response is dominated by the response of the float manometer.
b. Low-speed Dines anemometer
To explore the effect of changing mean wind speeds on the response of the low-speed Dines anemometer, we first consider the response to sinusoidal inputs at specific frequencies for two different mean wind speeds, as shown in Fig. 8. Both the amplitude response and phase angle curves show a difference, particularly at low frequencies, where the peak of the low-frequency resonant response shifts from 0.3 Hz at 23.0 m s−1 to 0.5 Hz at 30.5 m s−1, although the magnitude of the response remains relatively constant. The amplitude response curve suggests that the second resonant peak remains unchanged in terms of both the frequency at which it occurs and its magnitude, although the phase angle curve suggests that there is a difference as the minimum between the two peaks shifts from 0.7 to 1.0 Hz, and the second resonant peak shifts from 1.0 to 1.1 Hz.
The amplitude response of the low-speed Dines anemometer to white noise inputs at mean wind speeds of 21.2, 26.4, 28.5, and 29.9 m s−1 as seen in Fig. 9 shows similar behavior; that is, as the mean wind speed increases, the first resonant peak shifts toward higher frequencies, while the second resonant peak remains relatively constant in terms of both magnitude and frequency at which it occurs. A comparison of the curves shown in Figs. 8 and 9 does, however, reveal subtle differences between the two sets of curves, particularly in terms of the relative magnitudes of the first and second resonant peaks. For the sinusoidal inputs, the magnitude of the first resonant peak is slightly reduced relative to the white noise resonant peak, while the magnitude of the second resonant peak is somewhat higher.
Figure 10 shows a comparison of the results obtained from three independent runs with the same white noise inputs at a mean wind speed of 28.3 m s−1 for the low-speed Dines anemometer. Although there is clearly some variability from run to run, there is consistency across all three runs in terms of both the magnitudes of the resonant peaks and the frequencies at which they occur. Some differences in the measured resonant peak magnitudes were noted when a second set of three independent runs with the same white noise inputs were carried out for the same setup after the high-speed Dines tests were conducted at Townsville airport; however, there are no obvious reasons for these differences. It was noted during preliminary testing that the results did show some sensitivity to the alignment of the displacement transducer when connecting it to the float chamber, so it is possible that minor differences in the setup between the two sets of runs led to the observed differences.
The results of the low-speed Dines anemometer tests are also quantitatively similar to those of Borges (1968), who tested a Fuess 82a universal anemometer of similar design and function to the Dines anemometer. Both sets of results show the presence of two resonant peaks, although the frequencies at which these occur are different. Furthermore, the relative magnitudes of the two resonant peaks show considerable differences in behavior, with increasing wind speeds when those for the Fuess anemometer are compared to those of the Dines anemometer. While the results of the current study suggest that the relative magnitudes of the two peaks remain roughly constant for all wind speeds, those for the Fuess anemometer suggest that the magnitude of the second resonant peak becomes comparable to that of the first resonant peak with increasing wind speeds. This may be due to the much larger range of wind speeds considered by Borges, but it could also indicate that there are differences between the float designs used in the two anemometers. A numerical model developed as part of the larger project that this study forms a part of could be tuned to either the results of Borges or the current study by adjusting the physical constants used in the model, implying that there are some significant differences between the float designs used in the two instruments [see Ginger et al. (2011) for further details]. In this regard, the behavior of the Fuess anemograph is more consistent with that of the high-speed Dines anemometer to be discussed in the following section.
c. High-speed Dines anemometer
Figure 11 shows the results of applying a series of sinusoidal inputs at specific frequencies for a nominal mean wind speed of 46.0 m s−1 to the high-speed Dines anemometer, while Fig. 12 shows the variation of the amplitude response of the high-speed Dines anemometer to white noise inputs at mean wind speeds of 27.2, 32.2, and 37.8 m s−1. As for the low-speed anemometer, two distinct resonant peaks are visible in both the amplitude response and phase angle curves. The results suggest that by comparison to the low-speed Dines anemometer, the resonant frequencies are shifted to the left, and that the magnitude of the second resonant peak is broadly comparable to that of the first peak. Furthermore, Fig. 12 shows that while the resonant frequencies are relatively unchanged with mean wind speed, the magnitude of the second resonant peak shows a much greater dependency on the mean wind speed than it does with the low-speed Dines, becoming significantly larger with increasing mean wind speeds. The first resonant frequency remains consistent at 0.3 Hz, while the second resonant frequency is a consistent 1.1 Hz.
5. Comparison with cup anemometers
Of particular interest in this study is whether the results of the experimental determination of the response of both low- and high-speed Dines anemometers described in the previous section can explain the observed gust wind speed differences between collocated Dines pressure tube and Synchrotac cup anemometers noted in the introduction to this paper. To do this we use an approach based on that of Davenport (1964) and Beljaars (1987), in which the mean gust factor occurring over some time period T is calculated as
where G is the mean gust factor, g is a peak factor, σu is the standard deviation of the along-wind speed fluctuations about the mean wind over T, and is the mean wind speed over the same time period. If the along-wind speed fluctuations follow a Gaussian distribution about the mean, then according to Davenport (1964) the peak factor can be calculated as
where υ is the cycling rate. If the spectral density function Su(f) of the along-wind speed fluctuations is known, then υ can be calculated as
while the standard deviation of the along-wind speed fluctuations can be calculated as
The effect of the anemometer response on both the cycling rate and standard deviation of the along-wind fluctuations can be accounted for by multiplying the spectral density function by an amplitude response squared function |H(f)|2, such that
The use of a finite T to determine the mean wind speed will result in filtering of the spectral density function at low frequencies, which, following Greenway (1980), can be accounted for by initially multiplying the spectral density function by an amplitude response squared function given by
For a typical cup anemometer, the amplitude response squared function takes the form
where D is the distance constant of the anemometer. One potential issue with the use of Eq. (9) to describe the response of a typical cup anemometer is that it does not account for overspeeding, which is a known issue with cup anemometers. Overspeeding of cup anemometers is a complex problem and, as shown by Kristensen (1998), is dependent not only on the response characteristics of the anemometer itself, but also to which the magnitude of the lateral velocity fluctuations that the instrument is exposed. The latter means that the degree of overspeeding associated with a particular type of cup anemometer is likely to be heavily influenced by local site conditions, and that the same anemometer may experience different degrees of overspeeding at two sites with widely differing terrain exposures. Accordingly, we do not address the issue of overspeeding in this paper when considering the response of a typical cup anemometer.
In Australia the gust wind speeds recorded by Synchrotac cup anemometers are further modified through the adoption of the WMO 3-s moving average gust wind speed definition (World Meteorological Organization 2008), in which a 3-s moving average is applied to the anemometer output signal before selecting the largest gust to have occurred in an appropriate period of time. This can be accounted for by multiplying the spectral density function by a third amplitude response squared function, such that
To define the appropriate amplitude response squared functions for the low- and high-speed Dines anemometers, we use the experimental results from the preceding section. For the low-speed Dines, we use the average of the three white noise runs shown in Fig. 10 for a mean wind speed of 28.3 m s−1, while for the high-speed Dines we use the white noise run shown in Fig. 12 for a mean wind speed of 32.2 m s−1. The resulting amplitude response squared functions are shown in Fig. 13, and are used to model the response of the low- and high-speed Dines anemometers in the same way that Eq.(9) is used to model the response of a typical cup anemometer. These curves account for the overall response of the Dines anemometer to fluctuating wind speeds, and include the effects of the float response, tubing lengths, orifice and frictional losses throughout the system, as well as other phenomena that may be occurring, such as Helmholtz resonance effects. Although the amplitude response squared functions shown in Fig. 13 were obtained for the defined mean wind speeds, they have been assumed to be applicable for the entire range of wind speeds likely to be of interest. Alternative calculations in which frequency shifting of the amplitude response squared function in proportion to the square root of the mean wind speed was assumed showed very little impact on the calculated gust factors for the Dines anemometer. Although the experimental results presented in the preceding section suggest that the amplitude response squared functions vary slightly with the mean wind speed, the theoretical model described in Ginger et al. (2011) indicates little change in the frequency at which the first resonant peak occurs with increasing mean wind speed. Furthermore, while the range of wind speeds considered experimentally is too small to reveal any changes in the frequency at which the second resonant peak occurs with mean wind speed, the results of the numerical model do show significant increases in the frequency at which the second resonant peak occurs with increasing mean wind speeds. It is found, however, that the response of both the low- and high-speed Dines anemometer over the range of frequencies likely to be of interest is dominated by the first resonant peak, and the second resonant peak has a negligible effect on the calculated gust factors.
There are several possible forms for the spectral density function that are required to be integrated to find both υ and σu. In this paper we use the von Kármán form specified in the Australian wind loading code, AS/NZS1170.2 (Standards Australia 2011), although any suitable expression for the spectral density function can be used (see, e.g., Beljaars 1987). This takes the form
where ℓu is an integral length scale.
a. Comparison with the measurements of Logue (1986)
As noted earlier, Logue (1986) compared mean and gust wind speeds measured using a standard (low speed) Dines anemometer with those measured using a collocated cup anemometer at the Irish Meteorological Service's Galway observing site over an entire year. The cup anemometer used was stated to be a Mark II cup anemometer, which has been taken to refer to a Munro Mark II cup anemometer of the standard pattern used by the Met Office and others for many years. The cup anemometer output was recorded using a Mark 4 chart recorder. Testing by Sparks (1997) showed that the Munro Mark II cup anemometer has a distance constant of 12.2 m, while the response of the Mark 4 chart recorder is such that the overall response of the anemometer and chart recorder combined is dominated by the response of the cup anemometer. For the purposes of this paper, therefore, we model the response of the cup anemometer using the product of Eqs. (8) and (9), with D in Eq. (9) set equal to 12.2 m. The mean gust factor recorded at the site using the Dines anemometer was around 1.7, which suggests values for the turbulence intensity of around 0.22.
Using a value of 85 m for ℓu at a height of 10 m, as specified in Standards Australia (2011); a mean wind speed of 20 m s−1; a turbulence intensity of 0.22, from which σu can be calculated as ; and a long-term averaging time T of 600 s, the response parameters for both the Mark II cup anemometer and the Dines anemometer can be calculated as shown in Table 1. The values given in this table show that the low-speed Dines anemometer gives a higher cycling rate, peak factor, and gust factor when compared to the response of the cup anemometer because of the resonance in the float system. The ratio of the average gust factors can be used to calculate the average ratio between gusts recorded using the two anemometers, which for the conditions specified in Table 1 is 1.69/1.61 or 1.05. This value compares favorably with the average gust ratio of 1.06 (1/0.947) found by Logue (1986) for gusts greater than 28 m s−1.
b. Comparison with Synchrotach cup anemometers
In Australia Dines anemometers were replaced in general use in the early 1990s by Synchrotac 706 cup anemometers, with the WMO 3-s moving average gust wind speed definition subsequently being adopted by the Bureau of Meteorology at a later date. The Synchrotac 706 cup anemometer is a heavy-duty 3-cup anemometer with a distance constant of 13.0 m (J. Gorman 2010, personal communication), not dissimilar to the Mark II cup anemometer used by Logue (1986). A series of calculations of the average expected ratio between gust wind speeds measured using both low- and high-speed Dines and Synchrotac anemometers at a standard height of 10 m is made for a range of mean wind speeds and turbulence intensities using the theory outlined above. Furthermore, in calculating the Synchrotac cup anemometer response, we have considered both the anemometer alone [i.e., the product of Eqs. (8) and (9)] and the effect of applying a 3-s moving average to the sampled signal [i.e., the product of Eqs. (8), (9), and (10)]. The results are summarized in Tables 2 and 3 for low- and high-speed Dines anemometers, respectively.
Several observations may be made with respect to the results shown in Tables 2 and 3. First, the high-speed Dines anemometer tends to measure higher gust wind speeds relative to those measured using the Synchrotac cup anemometer than the low-speed Dines does for the same mean wind speed and turbulence intensity. This is a consequence of the fact that the first resonant peak for the high-speed Dines anemometer occurs at a lower frequency than it does for the low-speed Dines anemometer. There is some dependence of the gust ratio on mean wind speed, with the ratio slowly decreasing for the Synchrotac cup anemometer alone, and slowly increasing when a 3-s moving average is applied to the cup anemometer signal. For the same mean wind speed, increased turbulence intensities leads to significantly increased overshoot ratios for both low- and high-speed Dines anemometers when compared to Synchrotac cup anemometers. Finally, the effect of adopting the WMO 3-s moving average gust is to increase the overshoot ratio for both low- and high-speed Dines anemometers considerably when compared to the corresponding values for the Synchrotac cup anemometer alone. The reason for this is the effect of applying a 3-s moving average filter to the underlying spectral density function of the wind speed, which effectively truncates the high-frequency end of the spectrum, leading ultimately to much-reduced gust factors when compared to those for the cup anemometer alone.
Of particular interest are the overshoot ratios shown in Table 3 for the high-speed Dines anemometer when compared to those for the Synchrotac cup anemometer with a 3-s moving average applied to the sampled signal, since this is the case that is directly comparable to the field observations made during Cyclone Vance and Severe Cyclone Yasi. For a turbulence intensity of 0.20, the results are commensurate with the individual values reported during these events, particularly when it is kept in mind that the values in Table 3 represent an average value over a large number of observations, whereas the individual values are drawn from the expected distribution about the average value. An independent study of gust wind speeds measured using collocated high-speed Dines and Synchrotac cup anemometers at seven northern Australia observing sites was also undertaken as part of the larger project of which the study reported in this paper forms a part (Ginger et al. 2011). That study found that for gust wind speeds exceeding 15 m s−1, the high-speed Dines anemometer tended to report gusts that were 5%–10% higher than those obtained from the Synchrotac cup anemometer for gust wind speeds of about 45 m s−1, increasing to between 12% and 17% at gust wind speeds approaching 60 m s−1. Once again these values are consistent with both the values reported in Tables 2 and 3, as well as the field observations made during landfalling cyclones.
This study has assessed the response of both low- and high-speed Dines anemometers to gusts using experimental methods, as well as comparing the gust wind speeds measured using both versions of the Dines anemometer with those measured using Synchrotac 3-cup anemometers, which replaced the Dines anemometer as the official wind measuring instrument at AWSs in Australia in the 1990s, using analytical methods. Although the Dines anemometer produces mean wind speed measurements that are close to those reported by the Synchrotac cup anemometer, an experimental investigation of its response to fluctuating wind speeds shows a characteristic amplitude response curve dominated by the motion of the float manometer used to record the wind speed that is quantitatively similar to the one determined by Borges (1968) for a Fuess pressure tube anemometer, similar in design to the Dines anemometer. The amplitude response curves for both low- and high-speed Dines anemometers contain two resonant peaks associated with the interaction of the float and the water in which it rests. For the low-speed Dines, the first resonant peak lies at about 0.5 Hz, while the second peak lies between 1.0 and 1.1 Hz for the range of wind speeds tested. For the high-speed Dines anemometer, the location of the first resonant peak is shifted to the left of the corresponding peak for the low-speed Dines with a value of about 0.3 Hz, while the magnitude of the second resonant peak shows some reliance on the mean wind speed. For practical purposes, however, this dependency can be ignored, since it is the first resonant peak that dominates the response of both the low- and high-speed Dines anemometer over the range of frequencies of interest when determining peak gust wind speeds.
The use of random process and linear system theory allows the response of the Dines anemometer to gusts to be compared to that of a cup anemometer, and in particular to determine the expected average value of the ratio of the gust wind speeds recorded by the two instruments. The approach used is validated by reproducing the results of Logue (1986), who compared mean and gust wind speeds observed using collocated Dines and cup anemometers over a period of one year at an observing site in Ireland, before using the same theory to compare gust wind speeds measured using low- and high-speed versions of the Dines anemometer and the Synchrotac cup anemometer that replaced them. The results show that, on average, the low-speed Dines anemometer records gust values that are 2%–5% higher than those recorded using a Synchrotac cup anemometer alone under the same conditions, while the high-speed Dines anemometer records values that are 3%–7% higher, the exact value being dependent on both the mean wind speed and the turbulence intensity at a particular site. These differences are exacerbated with the adoption of the WMO-recommended 3-s moving average gust in which a 3-s moving average is applied to the digitally sampled Synchrotac cup anemometer output before reporting the peak gust wind speed, rising to 6%–12% and 11%–19% for the low- and high-speed Dines anemometers, respectively. This is because the effect of applying a 3-s moving average filter to the Synchrotac anemometer output is to further truncate the high-frequency end of the wind spectrum, which then leads to significantly lower peak gust values relative to those measured using the anemometer output directly. In the case of the high-speed Dines anemometer, these differences are consistent with both field observations made during tropical cyclones (Reardon et al. 1999; Boughton et al. 2011) and an independent extreme value analysis of the simultaneously observed gust wind speeds at seven sites in northern Australia with collocated high-speed Dines and Synchrotac cup anemometers (Ginger et al. 2011).
The results of this study also suggest that provided the turbulence intensity can be estimated, it is possible to use the calculated ratios of the gust wind speeds recorded by both low- and high-speed Dines anemometers relative to those recorded using the Synchrotac 706 cup anemometers currently in use in Australia to adjust both the older gust wind speeds recorded using Dines anemometers and more recent gust wind speeds recorded using Synchrotac cup anemometers to a common equivalent gust duration. This would result in significantly longer, standardized gust wind speed data records being available at many anemometer sites in Australia for long-term wind climate analyses of the type required to determine the design wind speeds used in wind loading codes and standards, or the assessment of possible effects due to climate change. The same methodology could also be used in countries other than Australia, such as South Africa or Ireland, where Dines anemometers were the main wind recording instrument for many years, until being replaced by AWSs typically equipped with cup anemometers in the 1990s.
There are, however, several potential issues that have not been addressed in this study. First, we have neglected any potential overspeeding effects on the cup anemometer measurements, since these are not only dependent on the response characteristics of the anemometer itself, but also on the magnitude of the lateral velocity fluctuations at a particular observing site. Furthermore, in undertaking comparisons of the gust wind speeds recorded by low- and high-speed Dines anemometers relative to those recorded by cup anemometers in both Ireland and Australia, we have assumed that all of the anemometers examined are representative of a well-maintained, properly calibrated example of the type of anemometer being considered. In addition to this, we have neglected any potential effects of long-term instrument wear over many years of service on the resulting wind speed measurements. All of these issues would warrant further examination to establish their impact on the accuracy of long-term wind speed measurements.
The authors would like to acknowledge the contributions of both the Bureau of Meteorology and Geoscience Australia to this project. We would also like to thank Bruce Harper, Jeff Kepert, and two anonymous reviewers for their comments on an earlier draft of this paper. The work described in this paper was undertaken as part of the project Extreme Windspeed Baseline Climate Investigation, funded by the Department of Climate Change and Energy Efficiency.