NOTES AND CORRESPONDENCE. On the Use of 50-MHz RASS in Thunderstorms

Transcription

1 936 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 0 NOTES AND CORRESPONDENCE On the Use of 50-MHz RASS in Thunderstorms PETER T. MAY Bureau of Meteorology Research Centre, Melbourne, Victoria, Australia CHRIS LUCAS Department of Physics and Mathematical Physics, Adelaide University, Adelaide, South Australia, Australia RICHARD LATAITIS NOAA, Environmental Technology Laboratory, Boulder, Colorado IAIN M. REID Department of Physics and Mathematical Physics, Adelaide University, Adelaide, South Australia, Australia 19 July 00 and 13 December 00 ABSTRACT An experiment describing the use of a 50-MHz wind profiler Radio Acoustic Sounding System (RASS) to observe the vertical velocity, precipitation, and buoyancy structure of convection is discussed. In most cases the RASS signal was lost when convection was overhead. A theoretical argument that this is the effect of turbulence on the acoustic wave fronts is presented. 1. Introduction Wind profilers have been used to study precipitating systems for some time. Techniques for the retrieval of raindrop size distribution (DSD; Wakasugi et al. 1986; Gossard 1988), snow (Rajopadhyaya et al. 1994), and mixed phase precipitation (Drummond et al. 1996; May et al. 001) have been developed and applied to the study of precipitating systems (e.g., May and Rajopadhyaya 1996; Cifelli et al. 000). Such studies have also aimed to examine the vertical motion field and its relation to the precipitation characteristics in convective systems (May and Rajopadhyaya 1996; May et al. 001). In particular, the combination of 50-MHz and 90-MHz wind profiler data have been used to obtain fairly complete kinematic descriptions of convective systems. However, these studies have lacked corresponding thermodynamic information at comparable resolution to the profiler observations. In principle, Radio Acoustic Sounding System (RASS) measurements provide profiles of virtual temperature and hence buoy- Corresponding author address: Dr. Peter T. May, BMRC, GPO Box 189K, Melbourne 3001, VIC, Australia. p.may@bom.gov.au ancy. From the results of previous 50-MHz RASS observations, it may be expected that such measurements should yield nearly continuous temperature data up to or above 6 km (May et al. 1988). Therefore, experiments using 50-MHz RASS have been performed in an attempt to include detailed thermodynamic information in addition to the vertical motion and precipitation fields within convective clouds. The aim of the experiment was to measure the thermodynamic structure of storms in the Darwin area through the freezing level to heights where significant acceleration of the vertical air motion is observed in deep convection to obtain as complete a picture as possible of the kinematics, thermodynamics, and precipitation microphysics of tropical thunderstorms. This paper reports results that indicate that this objective was very optimistic and that such RASS measurements are much more difficult than expected. A possible explanation for these difficulties is explored.. The experiment There are two collocated profilers located near Darwin, northern Australia: one operating at 50 MHz and the other at 90 MHz. These systems sample at common 003 American Meteorological Society

2 JUNE 003 NOTES AND CORRESPONDENCE 937 times and spacings (with height averaging of the 90- MHz data) and the region around the profilers is covered by dual-polarization radar data (Keenan et al. 1998). They have been used for a number of the studies cited above. In these, the 50-MHz profiler data have been used to directly measure the vertical motion within precipitation systems, and the 90-MHz profiler data have been used as a vertically pointing Doppler weather radar. The combination allows the fall speed spectrum of the hydrometeors to be measured and the precipitation characteristics to be analyzed in detail. As discussed in the introduction, an experiment focussed on using RASS was undertaken to obtain thermodynamic information within storms. RASS exploits the profiler ability to measure the radar backscatter from fluctuations in clear-air refractive index. For RASS, an acoustic source is placed next to the radar and the radar backscatter from the density perturbations induced by the sound waves allows the speed of sound, and hence virtual temperature, to be measured (e.g., May et al. 1990). At acoustic frequencies of about 10 Hz (corresponding to a wavelength ½ the radar wavelength), the main limitation for height coverage has been thought to be the advection of the acoustic wave fronts away from the radar beam (May et al. 1988). A single acoustic source was deployed for the RASS in January and February 000. In February 001, three acoustic sources were deployed to improve the system sensitivity. With multiple sources, each acoustic source produces a radar spot on the ground and the resulting larger illumination reduces the effect of wind (May et al. 1990). At higher frequencies acoustic attenuation becomes the limiting factor for height coverage. The site has housing nearby and therefore there were potential noise pollution issues with the RASS. Thus, the RASS was only turned on when storms were approaching the profiler. A linear frequency sweep acoustic waveform was used (May et al. 1990). The reflectivity field from operational weather radar was used to make the decision to take measurements. These data are displayed in real time. The 50-MHz profiler was recording the clear-air and RASS Doppler spectra, but at limited velocity resolution because the signal processing unit could only handle up to 104 point spectra. The corresponding velocity resolution was 0.8 m s 1. The spectra were averaged over a period of 1 and min and a least squares fitting approach was applied for the moment estimation. With the 50-MHz profiler, the clear-air peak is generally well separated and mostly larger than the precipitation peaks. Occasionally there is some ambiguity where there is a mixture of ice and water phase (e.g., in the bright band) and the precipitation peaks become very broad and merge into the clear-air peak and these data have been removed. The operational characteristics are summarized in Table 1. TABLE 1. Operating characteristics of the profilers. Sampling period Height resolution Height sampling Beamwidth 50-MHz profiler 60 s 450 m 300 m 3 90-MHz profiler 60 s 150 m 100 m 9 RASS acoustic power 1.5 kw 3. Some observations Two examples will be discussed in detail. The first of these will be used to illustrate the possibilities of the technique for obtaining detailed storm structure including the precipitation, thermal, and vertical velocity structure. However, as will be shown in the second case, on most occasions the RASS echoes were not detectable when convection was overhead. On 8 February 000, a weak monsoonal convective line with a trailing stratiform rain region was approaching the profiler from the north (Fig. 1). The polarimetric radar data show that the storm was decaying as it moved over the profiler. Figure shows time height cross sections of the 90-MHz radar reflectivity as well as the vertical motion and virtual temperature measured with the 50-MHz profiler during the storm passage. The main convective band was overhead between 0630 and There is evidence of a bright band beginning to develop, but there were still relatively strong vertical motions. Since the updrafts were about m s 1, it must be large crystals and graupel that are melting to begin the brightband formation. A brief transition zone was seen between 0655 and 0700 (all times in UTC), with weak vertical motion. After 0700, heavy stratiform rain began with an intense bright band evident in the reflectivity. After 0730, there was light stratiform rain with the bright band still present, but only weak vertical motion and the echo intensity was weakening with time. The maximum vertical velocities are relatively weak with maximum amplitudes less than 5ms 1. Furthermore, it is clear in this figure that the convective downdraft has essentially undercut the entire storm and the process of decay into a stratiform rain region has begun. This is frequently seen in profiler data taken in this area (May and Rajopadhyaya 1999). Another feature of this dataset is that the spectral widths of the clear-air signals,, are mainly small. The importance of this will become clear in the following sections. As in previous studies we can use the profiler data to obtain microphysical information. With the relatively small amplitude vertical motions there is no sign of the production of hail in either the profiler or polarimetric radar data, such as was noted in the study of the small intense thunderstorms by May et al. (001). We can paramaterize the retrieved DSD by fitting exponential distributions to the data [N(D) N 0 e ( D) ] using a similar method to Wakasugi et al. (1986). Differences in DSD parameters in the different rain regions are very

3 938 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 0 FIG. 1. Radar reflectivity field as the convection was passing over the profiler site (marked as X). This was part of a decaying monsoonal squall line moving from the northwest into a region where light widespread rain had already fallen (8 Feb 000). apparent (Fig. 3). Both the DSD slope and intercept N 0 tend to increase in the convective region. The slope ranges between 3 and 5, with large values of the intercept parameter (N 0 : to ) that is, many small drops and relatively few large drops giving a small median diameter. There is a distinct transition into the stratiform region with relatively few drops altogether (low N 0 ) and relatively more large drops (small slope). The preponderance of large drops is reflected in median diameter. This regime lasts until about 0745, where intercept and slope move to intermediate values during the last portion of stratiform stratiform rain. These numbers and behavior compare well with the previous literature, particularly the dichotomy between the convective and stratiform regions. The median temperatures in the region just outside of the convective cloud (within 15 min of the precipitation) have been compared to those within, and although it is not clear in Fig., the temperature within the cloud was about 0.5 C warmer than the surrounding areas, although this was mostly in the stratiform rain area. However, there is a distinct negative correlation between vertical velocity and temperature perturbations seen within the storm; that is, the updrafts were cooler than the downdrafts and a warming in the stratiform rain area. We had expected the reverse to be the case. The RASS analysis has been carefully checked to ensure that effect of vertical motion is properly corrected, and the correlation between temperature and vertical motion far exceeds that expected from correlated random errors produced from using the observed vertical motion to correct the RASS sound speed measurements. However, updrafts with little or negative buoyancy and positively buoyant downdrafts have often been observed in weak or decaying convection (e.g., Igau et al. 1999) so this single result is not unreasonable. The regions with the largest vertical motion are characterized by the largest lapse rates and are moist unstable. The areas of the cloud with small vertical motion are typically 1 C km 1 more stable than the convective region. Note that with 50-MHz RASS the minimum height that is sampled is about 1.5 km above the ground and hence above any convectively generated cold pool. There may also be issues related to the fact that the storm is decaying and the negative correlations may be a reflection of this. This example illustrates the potential of the technique for obtaining detailed storm structure. Although there are some features that are difficult to understand and a single case is insufficient to resolve if they are true physical measurements of the warm downdrafts or are an artefact of the observational technique, this example illustrates the potential of the approach. This case had by far the best RASS height coverage within the convection that we have obtained. Note that

4 JUNE 003 NOTES AND CORRESPONDENCE 939 FIG.. Time height cross sections of 90-MHz range corrected signal power (db), vertical motion, 50-MHz profiler spectral width, and RASS virtual temperature measured during the passage of a decaying monsoonal squall line. The contour overlays are of power (top), zero velocity (over the velocity), spectral widths greater than 1.5 m s 1 (over the spectral width), and zero velocity over the RASS virtual temperature measurements. The zero velocity over the virtual temperature are to highlight the correlation between the updrafts and cooler temperatures. the coverage in heavy rain was maintained. We will now discuss more typical cases (e.g., Fig. 4). These cases were associated with deep convective cells and heavy rain. The quality of the data in this case is representative of every other case sampled during these experiments. The example, observed on 6 February 001, that is shown in detail was much more convectively active than the 8 December 000 case. Data from the most active time are shown in Fig. 4. As for Fig., time height cross sections of 90-MHz reflectivity, vertical velocity clear-air spectral width, and the RASS measurements are shown. Features to note are the updrafts in excess of6ms 1 showing that this example is much more intense. The s during the convection are large and variable, typically having values of several meters per second. At this time the RASS signals completely disappear. This behavior was seen in almost every case observed during the 000 and 001 experiments. In this case, it is seen that the signal dropout occurred prior to the onset of the precipitation. However, note the increased reflectivity and the large spectral widths at low levels ahead of the storm. This is possibly associated with a turbulent storm outflow. The remaining discussion will focus on the causes of the dramatic dropouts in the temperature data. One possibility is that the presence of the precipitation itself could affect the sound waves. Precipitation may cause some scattering of the sound, but the RASS height coverage did not change significantly during the 8 December 001 case despite quite high rain rates being present. This suggests that the loss of RASS signal was not associated with the effect of rain on the acoustic wave fronts. We hypothesize that the loss of signal is associated with turbulent distortion of the acoustic waves decreasing the RASS intensity. This also explains the early dropout in Fig. 4 and will be examined next. 4. Some theoretical considerations The outstanding question is the cause of the dramatic loss in RASS echo intensity as convective systems move across the profiler/rass. While large water concentrations are seen in the most active convection, if the water concentration was the major factor in the coverage, as just noted, it is expected that some effects would have been seen in the 8 December 000 case and in the 6 December 001 case we see dropouts ahead of the rain. A more likely explanation is attenuation by turbu-

5 940 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 0 FIG. 3. Time series of derived precipitation characteristics from the 50-MHz spectra. The top panel is the rain rate deduced from the DSD measurements and the lower panel shows the median volume diameter of the DSD (solid) and the intercept parameter (N 0 ) of the fitted exponential distribution. The median volume diameter is the diameter such that 1/ of the rain volume lies in drop with diameters exceeding this value and is related to by D /. lence. Theoretical studies have examined the impact of turbulence on echo intensity (e.g., Lataitis 199) and it is expected that severe turbulence will cause increased attenuation of RASS echoes and a very strong range dependence with signal decreasing at approximately R 6/5 instead of R, where R is the range (in the absence of horizontal winds which displace the acoustic wave fronts and cause further losses). It is seen in Fig. 4 that the data dropouts are associated with large values of, an indication of strong turbulence, rather than heavy precipitation per se. The dropouts occur outside of the precipitation if there are highly turbulent zones in the lower levels and the regions with the largest vertical motions and very wide spectral widths where the temperature perturbations would be expected to be largest. However, Lataitis (199) expressed the intensity in terms of the value of the acoustic turbulent structure parameter C n, which is not directly measured with profiler. 1 However, for inertial subrange turbulence, both 1 The subscript n of the turbulent structure parameter refers to the refractive index; other subscripts refer to the variable of interest: T for temperature, q for specific humidity, w for vertical motion, for velocity, etc. the acoustic and radar C n, have been related to the eddy dissipation rate, and the radar spectral width has in turn been related to the eddy dissipation rate. Thus, we can use the from radar Doppler spectra and the expressions in Lataitis (199) to infer the degree of turbulence-induced attenuation in the RASS signal under the assumption that mechanical turbulence dominates acoustic C n. The horizontal wind speeds in Darwin are quite low and for the examples with very large widths it can be safely assumed that he spectral broadening due to wind shear or finite beamwidth is relatively small. If we assume we are observing radar scatter from turbulent fluctuations in the inertial subrange of threedimensional turbulence, we can relate the radar to the eddy dissipation rate. In practice, the assumption of the inertial subrange turbulence is probably not valid as a significant component of the vertical velocity variations are associated with convective motions rather than pure turbulence, but these results should at least give us a guide on the effects expected in very turbulent convective motions. For example, we can use the Gossard et al. (1998) results to relate to eddy dissipation rate although other algorithms provide similar results (e.g., Hocking 1983; White et al. 1999). In order to use this theory, we are assuming a transverse wind speed

6 JUNE 003 NOTES AND CORRESPONDENCE 941 FIG. 4. As in Fig., except for 6 Feb 001 and contours of spectral width 1.5 m s 1 are drawn over the virtual temperature. of 10 m s 1 and a sample time of 60 s although the results are not sensitive to this. For the Darwin radar characteristics and the above assumption m s 3. With estimates of, we can use results Ostashev (1997, 04 05) to relate to the acoustic C by C 5.13 /3 / c 0 where c 0 is the speed of sound and assuming that the velocity term is dominant over the humidity and temperature terms. For many conditions this is the case and at worst they are comparable. Note also that the temperature and humidity terms are additive to the total, so that the actual attenuation for the inertial subrange C n turbulence will be greater than our calculations. However, this approximation will at least provide us with a qualitative indication on whether turbulent effects explain our signal loss. Then from Lataitis (199) we have the intensity of the RASS signal I: where 1 I, 16/5 R 1 R 0 5/6 G 3/4 R0, 3/8 40 (C ) n n n and G is the antenna gain and is the antenna wavelength. The value of C n should be a height average (see Lataitis 199 for details), but for the case of a uniform profile of C n it reduces to the mean. Our case has a small height variation but is a second-order effect. Figure 5 shows curves of the turbulent loss in signal intensity as a function of height for a variety of. The loss rates are clearly very high in very turbulent conditions. In fact, the turbulent attenuation rates clearly explain the loss of the signal in section 4 given that our best observations were still being made with low signalto-noise ratios. The loss of signal is even seen at the lowest heights even though the loss at 1.5 km is relatively modest in these calculations. This is a reflection of the low SNR of the RASS signals measured here. The attenuation rates are comparable to acoustic attenuation for 1-GHz RASS so it is reasonable to ask why this effect has not been seen in 90-MHz observations in turbulent conditions, for example the gust front results discussed by May (1999). However, we would expect turbulent effects on 90-MHz RASS at low altitudes to be less noticeable for several reasons. The ranges are smaller, so turbulent distortion of the wave fronts is less and to a certain degree the additional loss is probably masked by the already high attenuation due to absorption of the sound by the air at these frequencies (May et al. 1988). Another factor is that the

7 94 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 0 FIG. 5. Plot showing theoretical calculations of the signal loss due to turbulence as a function of height assuming uniform turbulence for six different values of spectral width (m s 1 ). RASS signal-to-noise ratios at 90 MHz and low latitudes are often very high, greater the 0 db, so some turbulent attenuation can be tolerated. However, even when 50-MHz RASS has very good height coverage, the signal-to-noise ratios are quite small for profilers similar to the Darwin system, but decrease only slowly in range (May et al. 1988). Thus, even moderate attenuation by turbulence may cause the complete loss of detectable signals. 5. Conclusions This paper has discussed some results from a systematic attempt to measure the vertical motion, thermal structure, and precipitation characteristics within convection. The results clearly indicate that there is potential for these measurements, but that there is a need for further experimentation. Reasonable data were obtained for a case where the convection was decaying and the vertical motions and turbulence were both small. Some of the results were interesting in particular, the negative correlation between temperature and vertical motion is puzzling. This has been observed in some aircraft data but is not usual. With a single case it is not possible to determine if this was an instrument or measurement technique artifact or a genuine meteorological signal. However, there were severe data dropouts that were induced by the high levels of turbulence seen with the storms and occasionally in the near-storm environment. These dropouts limit the utility of the RASS observations using a system with the sensitivity of the Darwin profiler/rass. Further experiments are required with improved RASS instrumentation, particularly more and better RASS sources. At best, the Darwin 50-MHz RASS was only achieving RASS signals with modest to low signalto-noise ratios. The RASS power loss by turbulence was enough to destroy the signals, but experiments with more powerful sources and radars may have more success, both from the perspective of obtaining a complete picture of storm structure and exploring the impact of turbulence on RASS in more detail. Acknowledgments. This experiment was supported by an Australian Research Council small grant. Helpful discussions and suggestions from Dr. V. E. Ostershev are gratefully acknowledged. REFERENCES Cifelli, R., C. R. Williams, D. K. Rajopadhyaya, S. K. Avery, K. S. Gage, and P. T. May, 000: Drop size distribution characteristics in tropical mesoscale convective systems. J. Appl. Meteor., 39, Drummond, F. J., R. R. Rogers, S. A. Cohn, W. L. Ecklund, D. A. Carter, and J. S. Wilson, 1996: A new look at the melting layer. J. Atmos. Sci., 53, Gossard, E. E., 1988: Measuring drop-size distributions in clouds with a clear-air-sensing radar. J. Atmos. Oceanic Technol., 5,

8 JUNE 003 NOTES AND CORRESPONDENCE 943, D. E. Wolfe, K. P. Moran, R. A. Paulus, K. D. Anderson, and L. T. Rogers, 1998: Measurement of clear-air gradients and turbulence properties with radar wind profilers. J. Atmos. Oceanic Technol., 15, Hocking, W. K., 1983: On the extraction of atmospheric turbulence parameters from radar backscatter Doppler spectra 1. Theory. J. Atmos. Terr. Phys., 45, Igau, R. C., M. A. LeMone, and D. Wei, 1999: Updraft and downdraft cores in TOGA COARE: Why so many buoyant downdraft cores? J. Atmos. Sci., 56, Keenan, T. D., K. Glasson, F. Cummings, T. S. Bird, J. Keeler, and J. Lutz, 1998: The BMRC/NCAR C-band polarimetric (C-POL) radar system. J. Atmos. Oceanic Technol., 15, Lataitis, R. J., 199: Signal power for radio acoustic sounding of temperature: The effects of horizontal winds, turbulence and vertical temperature gradients. Radio Sci., 7, May, P. T., 1999: Thermodynamic and vertical velocity structure of two gust fronts observed with a wind profiler/rass during MCTEX. Mon. Wea. Rev., 17, , and D. K. Rajopadhyaya, 1996: Wind profiler observations of vertical motion and precipitation microphysics of a tropical squall line. Mon. Wea. Rev., 14, , and, 1999: Vertical velocity characteristics of deep convection over Darwin, Australia. Mon. Wea. Rev., 17, , R. G. Strauch, and K. P. Moran, 1988: The altitude coverage of temperature measurements using RASS with wind profiler radars. Geophys. Res. Lett., 15, ,,, and W. L. Ecklund, 1990: Temperature sounding by RASS with wind profiler radars: A preliminary study. IEEE Trans. Geosci. Remote Sens., 8, 19 8., A. R. Jameson, T. D. Keenan, and P. E. Johnston, 001: A comparison between polarimetric radar and wind profiler observations of precipitation in tropical showers. J. Appl. Meteor., 40, Ostashev, V. E., 1997: Acoustics in Moving Inhomogeneous Media. E & FN Spon, 59 pp. Rajopadhyaya, D. K., P. T. May, and R. A. Vincent, 1994: The retrieval of ice particle size information from VHF wind profiler Doppler spectra. J. Atmos. Oceanic Technol., 11, Wakasugi, K., A. Mizutani, M. Matsuo, S. Fukao, and S. Kato, 1986: A direct method for deriving drop-size distribution and vertical air velocities from VHF Doppler radar spectra. J. Atmos. Oceanic Technol., 3, White, A. B., R. J. Lataitis, and R. S. Lawrence, 1999: Space and time filtering of remotely sensed velocity turbulence. J. Atmos. Oceanic Technol., 16,