Meteorological radiosondes that use navigation systems to determine winds (and horizontal location) can be susceptible to data loss in thunderstorm environments. This paper reports on tests of a radiosonde that uses the Global Positioning System (GPS) for windfinding. Tests were made by flying the GPS radiosonde into three thunderstorms on free balloons that also carried an electric field meter and a long-range navigation (loran) radiosonde of a type previously tested. The GPS radiosonde performed without any significant loss of wind or thermodynamic data in in-storm maximum electric fields of up to −104 kV m−1. Also, no obvious deleterious effect on radiosonde data was found from the presence of nearby lightning. The radiosonde was further tested in a laboratory-produced electric field in an ambient atmospheric pressure of about 70 kPa, in which the radiosonde functioned normally in a vertical electric field up to 160 kV m−1 and in a horizontal electric field up to 100 kV m−1, the respective maximum applied. Radiosondes that were sprayed with water to simulate flight in rain performed correctly in an electric field of 135 kV m−1—the maximum that could be applied safely. The hypothesized reason for the excellent windfinding performance in high electric fields is partly the very short antenna length needed for GPS reception. Other factors, which could not be assessed in this study, may include the inherent low-noise susceptibility of the GPS signals and the processing circuitry. The tests showed that the GPS radiosonde obtains wind data in larger electric fields than does the loran radiosonde. It is concluded that GPS radiosondes will acquire windfinding data in most, if not all, thunderstorm and nonthunderstorm clouds that contain high electric fields. The thermodynamic data were also very good in the large electric fields.
Since the late 1980s, the National Severe Storms Laboratory has operated mobile laboratories with mobile ballooning capabilities (Rust 1989; Rust et al. 1990) that evolved from the Cross-Chain loran Atmospheric Sounding System (CLASS) developed at the National Center for Atmospheric Research (Lally and Morel 1985; Lauritsen et al. 1987; NCAR/ATD 1997). A major reason for our interest in mobile ballooning has been our desire to obtain soundings near and within severe thunderstorms and mesoscale convective systems. In addition, we have wanted to place these measurements in the storm context by comparing balloon-borne instrument location with Doppler radar data in terms of storm structure, in situ microphysical measurements, etc. The constraints of ballooning in severe thunderstorms and mesoscale convective systems, along with the need to receive radiosonde data as the mobile laboratories are in motion, led us to use radiosondes that have their own navigation and wind-finding capabilities and do not require radio-tracking antennas.
Our experience had been confined to the use of the CLASS technology, which includes the Vaisala RS80-15L radiosonde. Tests by Rust et al. (1990) in and near thunderstorms showed that the wind and position information were usually lost but that the thermodynamic data were usually acquired and of usable to excellent quality. [The results of these tests on the thermodynamic data are reconfirmed here, but the emphasis of this study is to test windfinding.] Rust et al. (1990) deduced that the cause of the loran (long-range navigation) data loss was corona discharge (the electrical breakdown of air) from the radiosonde’s 2.6-m-long loran antenna in the high electric field near and within storms. Rust et al.’s (1990) research indicates that an unmodified Vaisala RS80 loran radiosonde loses its windfinding data in an electric field magnitude |E| of approximately 10 kV m−1, which is well below the maximum electric field of about 75–150 kV m−1 typically found inside thunderstorms and stratiform regions of mesoscale convective systems (e.g., Marshall and Rust 1991; Shepherd et al. 1996). In subsequent tests, it was learned that halving the loran antenna length to 1.3 m and encasing it in a dielectric tube consistently extended the windfinding capability to electric field values of up to 40 kV m−1, without apparently adversely affecting the loran reception. More recently, when the straight-wire loran antenna was replaced with an antenna that had a center-loading coil, dielectric foam-encased upper and lower vertical elements, and a total length of approximately 0.6 m (Fig. 1), the radiosonde’s performance was improved, so that windfinding data were obtained in electric fields of up to about 100 kV m−1, although adequate performance does not always occur in such high fields. Since this type of data loss can also occur in winter storms and in electrified deep-layer clouds without lightning, the need to acquire winds is not limited to special research use in thunderstorms: it extends to operations in which the winds through deep layers of clouds are needed.
Based on previous experience, and given that the Global Positioning System (GPS) antenna on a radiosonde is shorter by a factor of 10 than our custom short loran antenna, we hypothesized that windfinding from GPS will be more reliable than windfinding from CLASS in the high electric field beneath and in thunderstorms and in other highly electrified but nonthunderstorm clouds. [Kaisti (1995) makes an equivalent statement regarding GPS performance in high electric fields but did not provide any details.] The high frequency band that carries GPS data has also been credited with increasing reliability (Saarnimo 1998). Furthermore, the digital GPS signals and processing may be less prone to problems even if corona noise enters the circuitry.
For nonstormy conditions, comparisons between loran and GPS windfinding and accuracy evaluations have been reported recently by Jaatinen and Pälä (1998), Nash et al. (1998), and Saarnimo (1998). For a baseline comparison of our systems, flights were also made in nonstormy weather conditions to determine if, as found by Jaatinen and Pälä (1998), Nash et al. (1998), and Saarnimo (1998), the GPS and loran radiosondes, along with their data processing capabilities, gave similar thermodynamic and wind profiles. Note that the tests reported here were designed to give rough comparisons, not to be precision tests. Since the two radiosondes were positioned vertically within about 2 m, they sampled the same volume of atmosphere essentially simultaneously as they ascended at approximately 5 m s−1. Some, if not all, of the apparent differences in temporal resolution and variability of the data shown here between the two systems resulted from the different data processing and recording intervals. Since this test was not aimed at detailed comparisons of radiosonde performance, no attempt was made to reconcile these. Given that this test was designed to search for major problems and for total loss of windfinding in large electric fields, no attempt was made to research loran versus GPS windfinding. For profiles in storms, the examples of thermodynamic data are shown to document whether the profiles appear plausible: they indicate that the radiosondes were functioning reasonably well in the thunderstorms. The primary goals of the tests reported here were to determine if the GPS radiosonde would provide windfinding in the presence of nearby lightning and in high electric fields and if the maximum electric field in which windfinding was obtained would be greater than for the loran system.
The GPS radiosonde tested was the Vaisala RS80-15G. This radiosonde has the same sensor suite as other Vaisala RS80 radiosondes. There can be a difference in the humicap sensor, which was a type A on the GPS radiosondes, and types A and H on the loran radiosondes. The differences are minor (Schmidlin 1998) and not important in this study. In lieu of the loran circuitry and antenna, the RS80-15G has a codeless, eight-channel, digital GPS receiver and antenna. The GPS portion of the radiosonde measures the satellite carrier Doppler shifts, and the radiosonde telemeters its GPS information, along with the thermodynamic data, to a ground station. The receiving system used during this test was the commercially available Vaisala DigiCORA II MW15. To improve windfinding accuracy, errors common to the receivers in the radiosonde and at the ground station are removed (Saarnimo 1998). A GPS radiosonde requires a minimum of four satellites to determine the horizontal wind (e.g., Jaatinen and Pälä 1998). The GPS radiosonde data shown here were taken in the system’s research mode, which recorded thermodynamic and windfinding data every 2 s. The GPS windfinding data were calculated every 0.5 s and were filtered to give 2-s raw data points (Kaisti 1995). The data from CLASS are a point every 10 s. Each thermodynamic data point is calculated from a linear least squares smoothing of raw data obtained every 1.5 s. Each 10-s windfinding data point is calculated from 60 s of loran information (Lauritsen et al. 1987; NCAR/ATD 1997).
Pertinent to the behavior of the RS80-15G in a large electric field is the packaging of the radiosonde. The circuit cards, water-activated battery, and GPS receiving antenna are housed in a Styrofoam case. This dielectric housing not only holds the components but also inhibits corona. Surrounding most of the Styrofoam is a water-repellent, thick-paper type of outer packaging that has four, 1.5-cm-diameter metal grommets at the top to hold instrument rigging lines. The 400-MHz transmitting antenna protrudes vertically from the bottom of the radiosonde. The antenna is made of copper wire. It is 16 cm long and 1.5 mm in diameter and is sheathed with a thin dielectric except at its tip. The transmitting antenna tip was straight, which is a worst-case configuration. (We had confirmed in our previous tests that if the antenna tip is tightly curled, the corona-onset field is larger.)
As with earlier testing of loran radiosonde performance during in-storm flights (Rust et al. 1990), an electric field meter was flown on the same balloon-borne instrument train (Fig. 1) to provide a quantitative measure of the electric field in which the radiosondes were embedded. The electric field meter and procedures for analysis of the electric field data have been described in several papers (e.g., Marshall et al. 1995). Both a GPS and a loran radiosonde were flown on each thunderstorm flight reported here. Because it is discussed later in this paper, note that the electric field meter provides substantial drag: the instrument train quickly stabilizes after launch, and the radiosondes do not swing much.
3. Balloon-borne flight tests
All the tests were conducted in New Mexico at the Irving Langmuir Laboratory for Atmospheric Research, which is located on a mountain ridge at 3.2 km MSL. As part of the fair-weather flights made with a GPS and a loran radiosonde on the same balloon, we investigated reports that the GPS radiosonde sometimes lost windfinding data immediately after a launch. For the 1996 version of the RS80-15G radiosonde and its line-releasing apparatus, we quickly concluded that the problem was the swinging of the instrument, even in light wind. In this simple test, the problem was apparently solved for most of the data collection through the use of a small drag parachute. (The manufacturer has subsequently modified the line-release system for a smoother release and doubled the line length that attaches the GPS radiosonde to the balloon to 60 m. The longer train reduces the pendulum effect experienced by the radiosonde.)
The tests also included windfinding by tracking the balloon with dual, optical theodolites. The theodolite data were intended to provide an independent, albeit coarse, check on the radiosondes’ windfinding in clear air. They were not intended to be precise enough to serve as highly accurate ground truth. Winds from the dual-theodolite data track the radiosonde data in both wind speed and direction (Fig. 2). In the example shown in Fig. 2, the theodolite data have at 625 s (at a P ≈ 468 mb and z ≈ 6.4 km) one obvious outlier in wind direction and a few possible outliers in wind speed. The spikes in the temperature data from the loran radiosonde are from a few isolated points with high noise. Other comparison flights in fair-weather conditions yielded similar levels of agreement.
Three flight tests were made in thunderstorms. In all three in-storm flights, the balloon burst at unusually low altitudes, most likely because of hail. Even with the premature end to the soundings, the flight test data are useful because they provide a measure of GPS radiosonde windfinding performance inside thunderstorms and in proximity to lightning. The lightning was verified to be nearby via observers’ reports of thunder, an interferometric lightning mapping system that was being operated by New Mexico Institute of Mining and Technology, and the ground field mills around Langmuir Laboratory. These storm soundings are shown versus altitude, as is typically done to place such data in the storm context.
A thunderstorm flight was launched at 1941 UTC 3 August 1996 (Fig. 3). Two storm cells were over the area, with the most active regions of lightning apparently within about 5 km. The two cells produced about 30 flashes during the balloon ascent to 7 km. The electric field meter data were noisy and unusable from about 4.4 to 4.9 km and above 6.2 km because of a problem in the telemetry-receiving hardware at the ground, which was later corrected. The first 19 2-s data points from the GPS radiosonde were discarded. At first glance, these initial data appear valid. However, between 32 and 38 s into the flight, the pressure stays constant at 679 mb (≈3.4 km). Then the pressure data repeat values from within the range of values recorded during the first 38 s. The temperature recorded in the first 38 s appears to be too warm in comparison with the subsequent data and the data from the other radiosonde, which indicate a normal ascent. Thus, the GPS radiosonde data points in these 38 s may be an artifact of the processing or recording. After the first 38 s, the data are well behaved, the temperature plot follows that from the other radiosonde, and all possible 2-s thermodynamic data from the GPS radiosonde were obtained. We did not see this problem in any other flights: its cause remains unknown but appears unrelated to the weather.
The outliers in thermodynamic data from the loran radiosonde are of interest in this flight. Of the real-time processed 10-s data points, about 12% are scattered in the data as outliers. In contrast, an examination of the raw 1.5-s data shows that less than 3% of the individual data points appear to be outliers. Of all the test flights and laboratory tests, this flight contains the largest percent of thermodynamic data outliers. Even so, the quality of the overall thermodynamic data is quite high.
The GPS radiosonde had five satellites at launch, but no other satellite count was logged. (The number of satellites was not recorded by the system software. In addition, because of other duties during storm flights, the operator made only intermittent comments in the log regarding the number of satellites.) There were three short gaps in the windfinding data during a 50-s period between approximately 5.38 and 5.64 km. The maximum electric field recorded during this flight was about 40 kV m−1, which occurred at about 6.1 km. One close and large lightning-caused field change is seen at 5.45 km. The first dropout in windfinding occurred about 20 s before the lightning. Very little windfinding data were lost. Figure 3 also shows profiles of wind speed, wind direction, and temperature from the GPS and loran radiosondes for this flight. The profiles are similar. Again, this rough comparison is not meant to compare absolute accuracies but to show that the radiosondes performed in a reasonable way inside the storm.
The second flight on 3 August 1996 was launched at 2034 UTC (Fig. 4). The sounding has several large field changes from lightning, which coincided with observations by the lightning mapping system and other observations. All possible GPS windfinding data were obtained during the flight. The GPS radiosonde had five to seven satellites during the ascent. The maximum electric field was about 38 kV m−1. Figure 4 also shows profiles of wind speed, wind direction, and temperature from the GPS and loran radiosondes. There are no obvious outliers in the thermodynamic data from either radiosonde.
The third test flight into a thunderstorm was at 2249 UTC 7 August 1996 (Fig. 5). Although the balloon achieved little altitude before it burst, the maximum electric field was −104 kV m−1. During this sounding, at least one flash occurred (at about the time the balloon entered the cloud), but no lightning-caused field changes are obvious in the electric field profile. The first 6 s of GPS windfinding data are missing, after which four or five GPS satellites were received, and all possible winds from the GPS radiosonde were calculated for the ascent. Figure 5 also shows profiles of wind speed, wind direction, and temperature from the GPS and loran radiosondes. The substantial disagreement in the winds occurs when the loran windfinding data had a sudden increase in the error in the 10-s averaged data. Thus, the loran windfinding data appear to be incorrect, but this cannot be determined conclusively. There were no obvious outliers in the thermodynamic data from either radiosonde.
4. Tests in laboratory-produced high electric fields
A controllable electric field that simulated the interior of a highly electrified cloud was created in a parallel-plate capacitor connected to a direct-current high-voltage supply (Fig. 6). An advantage of such laboratory tests is the ability to place the radiosonde in a large electric field that can be sustained, repeated, etc. The parallel-plate system was embedded in ambient, mountaintop air in the balloon hangar at Langmuir Laboratory where the atmospheric pressure was typically about 70 kPa. A GPS repeater transmitter was used to obtain the GPS satellite signals inside the metal hangar and between the parallel plates. Known voltages were applied to the upper plate, which was about 1 m above the lower plate. (The measured separation was 0.93 m in the center of the plates and 0.96 m at the edges.) A negative voltage was applied to the upper plate and created a positive electric field whose magnitude was simply the applied voltage divided by the plate separation. While the electric field for corona onset will vary slightly for positive and negative polarities, the results from a single-polarity test will provide the approximate values for corona onset and, more importantly, the approximate electric field for any failure in the radiosonde system. As the electric field was increased, the data reception from the radiosonde through its telemetry link to its ground receiving station was monitored for loss of either thermodynamic data or GPS windfinding data. In addition, the number of GPS satellites received was frequently logged by an observer.
The first test was made in daylight, and corona, which is dimly luminous, could not be observed visually. The radiosonde was initially dry and was suspended in ambient air whose relative humidity, measured by the radiosonde, was about 50%. The electric field was increased slowly, and audible corona was heard at an E ≈ 60 kV m−1. The radiosonde was receiving three satellites and acquired the needed fourth for winds as the electric field reached 112 kV m−1 (Fig. 7). The electric field was increased further to approximately 160 kV m−1, and the system continued to receive and process four GPS satellites for windfinding. Then the voltage was turned down to zero very rapidly to produce a large field change that simulated nearby lightning. No detrimental effect on wind or thermodynamic data was detected from the large field change. The test continued after the radiosonde had been sprayed with water. Audible corona began at 40 kV m−1. As the electric field was increased to 100 kV m−1, there were short periods when the number of satellites dropped from four to three. The dropouts of wind (at three satellites) can be seen in Fig. 7. When four satellites were received, the radiosonde continued to perform correctly in an electric field up to 100 kV m−1. There were some lags of a few tens of seconds between changes in the number of satellites and the loss or gain of wind data. Whether these delays were real or from possible delays in entering the observer’s comments in the data log is unknown, but the differences in time are small. The test was repeated (at 1900–2100 s in Fig. 7), with the radiosonde suspended horizontally in the plate capacitor to analyze the effect of a horizontal electric field. Five satellites were received, and all data were received with no dropouts in an electric field up to 100 kV m−1. Also, the wind speed data were always 0.0 m s−1 during the entire test, indicating good data accuracy in high electric fields. There were no obvious outliers in the thermodynamic data during this test.
Testing was also performed in very dark conditions at night to allow observers to see—and a low-light-level video camera to record—the luminosity of corona from the radiosonde. Figure 8 is a time-exposure photograph of corona from the radiosonde during one such test. The ellipses of light confirm that corona occurred from the temperature and humidity element that extends upward and outward from near the top of the radiosonde. They also confirm that corona occurred from the 400-MHz radiosonde-transmitting antenna that extends downward from the radiosonde as it rotated slowly and swayed in the electric field. During this test, the mountaintop and test facility were in cloudy air much of the time, with a radiosonde-measured humidity of approximately 90% and a temperature of about 9°C. The radiosonde was hung vertically. Audible corona began at an E ≈ 40 kV m−1, with corona from the radiosonde visible at about 60 kV m−1. The system initially received wind data to an E ≈ 135 kV m−1 (Fig. 9) except for a 6-s dropout at 700 s. Within a few more kilovolts per meter, the satellite reception dropped to three, and wind data were lost. This event is the only occurrence that suggests a possible link between a high electric field and the loss of wind data. However, the event could not be reproduced in subsequent parts of this test, indicating that something else caused the wind data loss. For example, the number of satellites sometimes changed during tests, even in fair weather. This was also observed with an independent, portable GPS receiver outside the hangar on the mountain ridge.
The high voltage was turned off, the radiosonde was sprayed with water, and the test was repeated. Corona was observed and recorded from the top and bottom of the radiosonde and from the 400-MHz transmitting antenna. The test was terminated at an E ≈ 135 kV m−1 because of concern that the observed sparking might lead to a full arc across the parallel-plate gap and damage to the high-voltage power supply. The radiosonde received five GPS satellites and all possible data. Corona and small sparks did not appear to affect the radiosonde’s thermodynamic and windfinding data. There were no obvious outliers in the thermodynamic data during this test.
The thermodynamic data were found to be of usable to excellent quality in all flights and tests in high electric fields and nearby lightning. From three balloon flights in thunderstorms, we find little evidence that GPS windfinding data are adversely affected by either a high electric field or the presence of nearby lightning. These flight results are corroborated and enhanced by the findings in larger laboratory-produced electric fields. The maximum laboratory-produced electric field in this test of 160 kV m−1 is greater than many thunderstorm electric fields reported in the literature. The number of GPS satellites obtained by the radiosonde during flights changed, but in a manner not obviously connected with the electric field or the occurrence of lightning. Furthermore, the maximum number of satellites in any given laboratory test was often received during the maximum electric field, suggesting no adverse effect from a large electric field. We think that the variations in the number of satellites received were due primarily to changing satellite reception.
The comparison of these tests with previous ones on loran radiosondes indicates that the GPS radiosondes are more reliable in obtaining winds (and horizontal location) in highly electrified clouds and in the vicinity of lightning than are radiosondes using loran. These results from thunderstorms and laboratory-produced electric fields are also applicable to nonthunderstorm, electrified clouds such as nimbostratus clouds, winter storms without lightning, etc. We had only a single manufacturer’s radiosonde system and GPS radiosondes available for testing, but we expect similar windfinding performance of all GPS radiosondes. The reason for their improved reliability is at least in part simply that the much shorter GPS antenna compared to a loran antenna increases the electric field needed to produce corona from the navigation receiving antenna. However, the test data showed good performance even in the presence of corona from the radiosonde. Other factors in the excellent performance, which were not evaluated, may include the frequency band of GPS, inherent low noise susceptibility of the GPS signals, and the processing circuitry. We believe the important conclusion is that GPS radiosondes will acquire windfinding data in most, if not all, thunderstorms and clouds that contain large electric fields.
We thank Dennis Nealson and Paul Griffin for help with sounding system maintenance, and Les Showell for assistance in data collection. We thank Patrick Stoller and Ivy Winger for their assistance in data collection and processing: 80% of their participation was funded through a National Science Foundation Research Experience for Undergraduates grant, NSF ATM-9508621, and 20% through grant NSF ATM-9403664. We thank Bill Winn for providing logistical support and making the Irving Langmuir Laboratory for Atmospheric Research available for this test, Mark Stanley and Paul Krehbiel for the video of corona and the observations of lightning with the New Mexico Institute of Mining and Technology interferometric lightning mapping system, and C. Andrew Niesen for the photograph of corona from the radiosonde. We thank Rich Thomas for making this test possible by providing the radiosondes and the National Weather Service sounding system.
Disclaimer: Reference in this paper to any specific commercial product or manufacturer does not constitute or imply its endorsement, recommendation, or favoring by the United States Government.
Corresponding author address: Dr. W. David Rust, NSSL, 1313 Halley Circle, Norman, OK 73019.