1. Introduction
Past efforts to use zeroth, first, and second moments of Doppler spectra collected by wind profiling radars (WPRs) to infer profiles of meteorological quantities other than winds have been broadly unsuccessful on operational wind profilers. In an experiment by Gossard et al. (1990), a 915-MHz (32.8-cm wavelength) radar was located near Stapleton Field at Denver, Colorado. Radisonde sounding data for the month of October 1989 were compared with radar sounding data to judge the capability of ground-based radars to estimate profiles of atmospheric refractive index (RI), from which humidity profiles could be inferred. Forty-five radiosondes were compared with radar data, and it was concluded that layers of insects, birds, and debris (sometimes from wildfires), as well as reflection from spurious targets, contaminated radar measurements and prohibited good comparisons. Contaminating signals could not be removed because raw Doppler spectra from the radar were not available for analysis and it was necessary to edit moments of the spectra instead. Thus, the 1990 experiment was inconclusive regarding the usefulness of radars for sensing profiles of RI or humidity.
In 1995 an experiment (Stankov 1998) was designed to eliminate most of the uncertainties described above by placing the radar in a relatively isolated area where large gradients of humidity were present most of the time. For this experiment, a 449-MHz (66.8-cm wavelength) wind profiling radar was taken to Point Loma near San Diego and located in a gully on the west side of the Point, where it was shielded from contamination by North Island Naval Air Station and most manmade emissions from the city. Contamination by spurious targets (aircraft, birds, radio frequency interference, and surface traffic), however, was still severe. Frequent radiosonde releases provided a large comparison dataset. These data clearly indicated a need to develop software for editing the raw Doppler spectra and rejection of the unwanted spectral features.
Historically, the first moment of the Doppler spectra from wind profilers has been of primary interest, providing the radial wind components from which wind profiles can be calculated. For this purpose it has been economical to calculate total moments, then edit the moments by removing outliers in an orderly way (e.g., Weber and Wuertz 1991; Woodman 1985). This has generally worked well for fairly long averaging times (30 min to an hour). That method, however, is poorly suited to the analysis of the short data samples necessary for direct comparison with balloon soundings, and it is not generally appropriate for analyses using zeroth-moment or second-moment data where the moments are sensitive to the shape and bounds of individual spectral peaks.
To obtain accurate moments for the desired atmospheric spectral peak, the National Oceanic and Atmospheric Administration's (NOAA) Environmental Technology Laboratory (ETL) has developed an operational advanced Signal Processing System (SPS; Wolfe et al. 2001). SPS differs substantially from more traditional operational signal processing systems, in which signal processing is linear and sequential throughout. Traditionally, signal processing is conducted independently on all data channels (i.e., over time, over range, across antenna beams) before combining data from multiple beams to generate wind profiles. Processing algorithms typically identify a single dominant spectral peak as the clear-air return, which can result in errors when contamination is significant. SPS is based on the recognition that even with attempts to suppress possible contamination from ground clutter, radio frequency interference, spurious echoes, noise, etc., wind profiler Doppler spectra can contain multiple signals (i.e., multiple spectral peaks), none of which may be recognizable as radar return from the atmosphere. SPS simultaneously uses all moment data to reliably extract atmospheric signals. Profiles of the zeroth (power), first (velocity), and second (spectral width) moments associated with the identified atmospheric spectral peak, the confidence in signal identification, the uncertainties in measurement estimates, and the computed horizontal and vertical winds are provided by SPS.
In this study we carry the analysis of the SPS-selected atmospheric signals one step further and use the identified, quality-controlled atmospheric signals, together with the radiosonde-measured temperature profiles, to compute the magnitude of the atmospheric humidity gradient profiles. The radiosonde-measured temperature profiles served as a proxy for WPR measurements of temperature using a radio acoustic sounding system (RASS) capability that is part of some profiling systems, but was not available on the systems used in this study. High vertical resolution but relatively low altitude RASS-measured temperature profiles can be extended throughout the atmosphere by combining ground- and space-based remote sensing observations in a physically consistent manner (Schroeder et al. 1991; Stankov 1996, 1998). We use data collected by a 3-beam 915-MHz radar on board the R/V Ronald H. Brown during the Department of Energy's Atmospheric Radiation Measurement Program (ARM) NAURU-99 experiment in June/July 1999 in the tropical Pacific (Westwater et al. 2003, hereafter WES) and data collected by a 5-beam 449-MHz radar located near Boulder, Colorado, during a typical winter upslope icing snow storm. We show detailed analysis of two cases for each radar system and compare the radar-obtained humidity gradient profiles with the humidity gradient profiles computed from the radiosonde-measurements.
2. Measuring systems
a. 915-MHz 3-beam boundary layer wind profiling radar
Wind profilers detect signals backscattered from turbulence-induced refractive index fluctuations with a scale of one-half the radar wavelength. The 915-MHz wind profiler (Fairall et al. 1997) is a special version of the mobile boundary layer wind profiler developed at NOAA's Aeronomy Laboratory in Boulder, Colorado (Ecklund et al. 1988). During the NAURU-99 experiment this profiler was deployed (Fig. 1a) on the R/V Ronald H. Brown (RHB) and operated in the three modes. The first mode sampled the boundary layer to 1500 m MSL in the vertical using only a 60-m resolution. The second used one vertical and two oblique beams to retrieve wind profiles in the lower atmosphere to 3500 m MSL with 100-m vertical resolution. The third mode was used for higher-altitude sampling (to 4500 km) with reduced vertical resolution (400 m). In this study we use data from the second mode only. Table 1 describes 915-MHz wind profiling radar parameters and the radar characteristics.
b. 449-MHz 5-beam lower tropospheric wind profiling radar
The 449-MHz radar wind profiler used in this study was designed to provide 15-min average winds from 0.2 to 4.0 km above the ground with 0.1-km vertical resolution. It is installed at Fort Huachuca, Arizona, to support operations of the aerostat-borne Tethered Atmospheric Radar System (TARS) deployed by the Air Force for the U.S. Customs Office. A network of 12 TARS sites spanning the southern border of the United States is used to detect low flying aircraft carrying drugs into the United States. However, when accompanied by WPRs, this network will provide valuable measurements of the moisture influx from the Gulf of Mexico. The system consists primarily of commercial-off-the-shelf (COTS) parts, and was integrated at a field site near Boulder, Colorado (Fig. 1b), during the 2000/01 winter. Data used in a part of this study were collected during the testing phase while the 449-MHz radar was located near Boulder, Colorado. The radar was operated in one mode with 100-m height resolution. A description of radar parameters is given in Table 1.
c. Radiosonde
On board the RHB we used the Vaisala RS80-15GH (GPS wind finding, H-Humicap, 403 MHz) radiosondes in two ways during this study; radiosonde temperature profiles were used as a proxy for RASS measurements which are needed to retrieve humidity gradients, and radiosonde-measured humidity profiles were used for verification of the retrieved humidity gradient profiles. Because Vaisala radiosonde humidity measurements deteriorate as the radiosondes age, we applied the age-dependent correction provided by the proprietary Vaisala algorithm. In addition, WES showed that the Vaisala radiosonde humidity measurements during the NAURU-99 experiment need to be constrained by the microwave radiometer measurements of the total precipitable water vapor (PWV). In this study we used the constrained, age-corrected radiosonde measurements. For the 449-MHz radar study we used the 6-s National Weather Service (NWS) radiosonde measurements at Stapleton, Colorado, with the average vertical resolution to 6-km altitude of 125 m.
3. Equations used to extract humidity gradient profiles from radar measurements
We used Eqs. (5) and (6) to retrieve moisture gradient profiles for WPR signals. The zeroth, first, and second moments of the backscatter Doppler spectra measured by the WPR yield
a. Estimate of the structure function parameter for the potential refractivity, C2ϕ
b. Estimate of the structure function parameter for the vertical velocity, C2w
The velocity structure parameter,
4. The ETL Signal Processing System
The NOAA/ETL operational advanced SPS includes four signal processing modules: signal detection, multiple moments estimation, signal identification, and meteorological products estimation (Wolfe et al. 2001). Each of these modules runs as a separate process, and a number of signal processing algorithms are implemented in each.
Signal detection is accomplished in a two-step process. First, the system noise threshold is determined using a statistical model (Hildebrand and Sekhon 1974). Second, spectral values above the noise threshold are separated into different signal domains for later identification. The recognition of multiple signals in the presence of noise is accomplished by detecting maxima above the noise threshold based on a defined minimum detection level (1.25σ above the mean noise floor). Each maximum is searched to the right and left until either the noise floor or a local minimum is encountered. Uncertainty in the spectral estimates, calculated as part of the statistical model, is used in assigning significance to the maxima and local minima. Overlapping signals are identified as part of this process.
Multiple spectral moments (Doppler velocity, power, and spectral width) are estimated using the centroid method for all peaks identified in a signal detection module. The combined set of spectral moments is used to assist in the identification of each signal with its physical source. An uncertainty is also computed as part of the estimates of the spectral moments and used later in calculating a confidence factor. No attempt is made at this point to separate out the true atmospheric signal, but only to pass on all potential signals with additional statistical information.
Signals from the atmosphere must be classified by their physical characteristics. In general these signals fall into two categories: precipitation and clear air. Precipitation signatures are related to the types of precipitation (snow, rain, hail, etc.), which are known to have different fall velocities (Ralph et al. 1996; Wuertz et al. 1988). The broadening effect precipitation has on the spectra and the fact that many different types of precipitation exist simultaneously can hide or overlap weaker clear-air signals.
The signal identification module selects which of the multiple moment estimates best describes an atmospheric signal. The identification is based on four characteristics: 1) magnitude of signals, 2) persistence of signal over range, 3) persistence of signal over time, and 4) persistence of signal across antenna beams. Each of these characteristics is calculated for each of the first three spectral moments: power, velocity, and spectral width. Therefore, there are 12 (4 characteristics times 3 spectral moments) characteristic measurements (Charms) calculated for each signal identified. These 12 Charms are used to select the atmospheric signals that will be further processed by a time–height continuity process (Weber and Wuertz 1991; Weber et al. 1993).
For each signal identified in a spectrum a confidence factor is calculated. This confidence factor is based on the uncertainty in estimates of the first three spectral moments, signal-to-noise ratio, signal-to-clutter ratio (clutter is the total power in all signals), range continuity, time continuity, and cross-beam continuity. The moments, calculated from the one signal in each beam that has the most consistency in terms of its Charms, are retained as the signal for processing into wind data. These same moments can also be used for calculation of turbulence quantities and humidity gradient profiling.
For each beam, the meteorological product module calculates an estimate of the radial wind velocity and its uncertainty from identified atmospheric moments using a least squares fit over a userdefined time–space grid. These radial velocities are combined to produce the standard WPR product and wind speed and direction profile over a user-defined time–space grid.
5. Humidity gradient profiles
a. Retrieval module
After processing radar data with SPS, the user possesses quality-controlled zeroth- (power), first- (velocity), and second- (spectral width) moment data associated with an atmospheric signal. In addition SPS provides a display of the time series of moments, winds, noise, and confidence measures for each beam and mode separately, or for all the beams and modes together.
To assess whether WPRs are capable of observing humidity gradient profiles, we developed an algorithm that builds onto the SPS analysis package. This algorithm was developed using the Math Works, Inc. technical computing language MATLAB, which allows easy interaction with the SPS. With minor future adjustments we expect that the developed humidity gradient algorithm will eventually be applied operationally at each wind profiler site of the Wind Profiler Network to provide continuous high vertical resolution humidity profiles (Stankov et al. 1996; Stankov 1998) together with continuous observations of the wind and temperature profiles from RASS. The method used is based on the theory described in section 3. It requires that all the Doppler spectral moments and the wind shear are available simultaneously. As the moments and horizontal winds are being read, the software allows the user to interrupt and display Doppler spectra, moments, and winds for each beam for additional interactive investigation of the raw and processed input data. Using the vertical beam measurements only,
b. 15–16 June 1999 (915 MHz)
During June–July 1999, the RHB sailed from Darwin, Australia, to the Republic of Nauru, a small island 8 km in diameter at 0.52°S latitude and 168°E longitude. RHB was equipped with a suite of ground-based remote observing systems that included a three-beam 915-MHz wind profiling radar. The goal of the mission was to evaluate how representative the island measurements are of the atmospheric conditions over the surrounding ocean. In addition to the remote sensing measurements, frequent radiosonde releases were conducted onboard the RHB.
There were 156 radiosonde soundings collected during a month-long cruise, and the 915-MHz radar operated almost continuously with only minor interruptions. We chose two radiosonde observations, 2254 UTC 15 June 1999 and 1045 UTC 16 June 1999, which were taken at the very beginning of the cruise while the RHB was near Darwin, Australia, crossing the Arafura Sea approaching the Torres Strait. The data represent typical maritime atmospheric conditions and are in no special way distinguished from the weather generally encountered in that region of the world.
Figure 2 shows 16-h-long time–height cross section of radar-observed signal-to-noise ratio, spectral width, and wind barbs. Arrows point to the times of radiosonde releases. Signal-to-noise ratio shows layers of enhanced Σ at 1500–2000 m MSL. Time–height cross section of the spectral width shows the same layered structure as Σ. The boundary layer was 700–800 m deep during this 16-h period. Between 0500 and 0800 UTC on 16 June 1999, the RHB encountered a period of rain evident from the radial velocity having high positive values. This period has been edited out in the time-height cross section of the horizontal wind barbs which shows generally easterly winds. Figure 3 shows the skewT-logp diagram of temperature and dewpoint, and 3D maps of radiosonde ascents with the observed horizontal wind for both radiosonde observations. During the first 5 h of this 16-h period, a lidar ceilometer was working and it showed intermittent but persistent low-level clouds with cloud fraction of 0.3–0.4 but rising to 0.8 for a brief period. At the time of the first radiosonde release, however, the ceilometer showed clear sky. During the release of the second radiosonde the ceilometer was not working. Winds were consistently easterly with zones of directional shear in the vertical. The 3D maps of radiosonde ascents show that they drifted away only about 10 km from the 915-MHz WPR on board the RHB and the skewT-logp diagrams show alternating layers of moist and dry air.
For the humidity gradient computation we used 24 min of radar data centered on the time of radiosonde release. Before computing means of radar measurements we applied a two-point two-dimensional median filter to the raw radar data and interpolated both the radar mean profiles and the radiosonde profiles to a 100-m vertical resolution grid. The ratio of turbulence outer length scales in Eq. (6a) was determined empirically to be 9 for this dataset. The results of comparison and the details of computation are shown in Fig. 4 for 2254 UTC 15 June 1999 and in Fig. 5 for 1045 UTC 16 June 1999. Solid lines represent radar data and dashed lines represent radiosonde data. The top row in each figure shows mean signal-to-noise ratio and spectral width from the 915-MHz radar obtained from the SPS, mean potential temperature and specific humidity from radiosonde, and mean horizontal components from both radar and radiosonde. The bottom row in each figure shows computed
In the first case, the Σ profile shows a very strong return from the middle of the boundary layer and two smaller peaks at 1400 and 2050 m MSL. The peak in the signal-to-noise ratio at 2050 m MSL, Σ = 0.335, coincides with the strong gradients of potential temperature and specific humidity observed by the radiosonde. The spectral width profile drops to a small value just at the top of the boundary layer inversion and the horizontal wind components obtained from the radar agree with the radiosonde-observed wind measurements. Here,
One of the main sources of error in the measurements of
c. 27 January 2001 (449 MHz)
In January 2001 the new 449-MHz WPR designated for support of the Ft. Huachuca TARS system was installed at its testing site near Boulder, Colorado. On 27 January 2001 a winter storm reached the Colorado Front Range. The 700-mb weather maps during this storm showed moist air influx from the south caused by a 700-mb level low-pressure center moving into the area from the West Coast. This moist air encountered cold air to the north of it, producing layers of enhanced humidity that were visible to the radar. Eventually the moist air influx produced a moderate upslope winter icing storm in the Front Range area of the Rockies.
Figure 8 shows the time–height cross section of signal-to-noise ratio, spectral widths, and the horizontal winds during the first 20 h of this storm. Signal-to-noise ratio shows a persistent layer of enhanced scattering starting at 3.5 km MSL at 0000 UTC 27 January and descending down to 2.5 km MSL—that is, to about 0.7 km AGL—by 1800 UTC the same day. Wind barbs show weak horizontal winds southeasterly near the ground and switching to southwesterly at the level coinciding with the enhanced signal-to-noise ratio layer. There was almost an hour long period of radio frequency (RF) interference between 1000 and 2000 UTC, which was edited out of the wind barbs plot. The NWS released two radiosondes at Stapleton, Colorado, about 70 km east of the radar site during this period. Because the storm persisted for three days and the system was fairly stationary it was decided to compare the humidity gradient profiles computed from the Stapleton radiosonde profiles with the radar-retrieved humidity gradient profiles in spite of the spatial separation. We compensated for the terrain height difference of about 180 m between the two sites in the same way that Gossard et al. (1995) did. Figure 9 shows the GPS PWV measurements for Platteville and Boulder, Colorado and the PWV obtained from Stapleton radiosonde measurement. It is seen that PWV measurements during the period of interest are about the same at all three sites indicating that the storm was spatially quite uniform. SkewT-logp diagram and the 3D map of the radiosonde ascents are shown in Fig. 10. The conditions were cold, with the radiosonde at 0000 UTC showing almost saturated air at the 680-mb level overlaid by the drier air above. The 1200 UTC radiosonde also shows a moist layer with a little drying above. However, by this time the radiosonde-measured humidity of the entire layer up to 500-mb level is quite uniform in the vertical, but the temperature profile is still showing a slight inversion between the cold surface air and the air above at about 750-mb level. The temperature at the surface dropped by 7°C during this 12-h period. Both radiosondes remained just above the release site throughout the ascent, with no significant balloon drift associated with the ascents. The 0000 UTC radiosonde showed a stronger directional shear at the top of the surface layer than the 1200 UTC radiosonde. The following radiosonde ascent, 12 h later, showed the atmosphere saturated up to 250-mb level.
Figures 11 and 12 show the corresponding humidity gradient computation comparisons. The top panels show the profiles of radar-measured (solid line) signal-to-noise ratio, spectral width, and wind components and the radiosonde-measured (dashed line) profiles of potential temperature, specific humidity, and the winds. The bottom panel shows the profiles of computed terms in Eqs. (5), (6a), (6b), and (12), which compose the steps toward obtaining dQ/dz, and the absolute humidity gradient profiles from radar and radiosonde measurements. The median of the length scale ratio in this data was estimated to be 2. In the first case the Σ profile exhibits no contribution from any other layer other than the layer where radiosonde-observed the dθ/dz is in the maximum. Spectral width does go to zero slightly above the inversion layer and radar-observed wind components agree with the radiosonde-observed. At the level of the maximum in
Figures 13a and 13b show the dQ/dz comparisons in more detail. At 0000 UTC, the mean of profile differences (radar − sonde) is m = 0.82 × 10−4 g kg−1 m−1, and the standard deviation is std dev = 0.83 × 10−3 g kg−1 m−1 and the statistics for the dQ/dz based on Brunt–Väisälä sign are m = 0.58 × 10−4 g kg−1 m−1, and the standard deviation is std dev = 0.95 × 10−3 g kg−1 m−1. At 1200 UTC, the mean of the profile differences (radar − sonde) is m = 0.40 × 10−3 g kg−1 m−1, and std = 0.59 × 10−3 g kg−1 m−1. Combining the data from both profiles together shows the statistics of differences to be: m = 0.24 × 10−3 g kg−1 m−1, std = 0.73 × 10−3 g kg−1 m−1 and a correlation coefficient r = 0.77. Although the RASS temperature profile observations were not available we interpolated radiosonde temperature soundings from 0000 and 1200 UTC 27 January and 0000 UTC 28 January to the time grid of the radar observations using a quadratic fit. This interpolated temperature field was then used to compute humidity gradient profiles for the entire period. Figure 14 shows the resulting humidity gradient time–height cross section (top panel) and the interpolated potential temperature field (bottom panel). The RFI period has been edited out (white areas). The main humidity gradient features are clearly visible. However, result points to the importance of having the simultaneous high temporal resolution RASS measurements at each WPR site.
d. Discussion
Stankov (1998), considered 17 dQ/dz cases of comparison for the data collected with the 449-MHz WPR at the Point Loma, California, site, where there is always a maritime boundary layer present, and it is overlaid by very dry and warm air assuring strong temperature and humidity gradients during much of the year. The average boundary layer height for those 17 cases was about 600 m AGL and the data were normalized by that height. The radar data used provided only moment measurements and not the spectra editing of the raw data. The results showed that 449 MHz radar in that case obtained humidity gradient measurements only to 1200 m AGL, the mean of the differences was m = 1.0 × 10−3 g kg−1 m−1, the standard deviation was std = 10.41 × 10−3 g kg−1 m−1, and the correlation coefficient was r = 0.7. Comparison of the present study with Stankov (1998) results indicates that although here we considered typical (not specifically selected) cases in widely varied atmospheric conditions, tropical versus winter continental, editing the data in the spectral domain with SPS provided improved results. The standard deviation of the differences with the radiosonde observations for the 449 MHz cases dropped by a factor of 14, the mean of the differences dropped by a factor of 4 and the correlation coefficient increased to 0.77. For the 915-MHz cases standard deviation of the radar/radiosonde differences dropped by a factor of 2, mean dropped by a factor of 2.5 and the correlation coefficient increased from 0.7 to 0.72. Our data clearly suggest that the dominant term in Eq. (5) is dϕ/dz and this is why properties of the humidity gradient so closely approximate (except for a constant) profiles of refractivity derived from radar measurements.
Encouraged by our results and aware of the community need to know humidity profiles instead of the humidity gradient profiles only, we integrated dQ/dz profiles using the GPS-measured PWV measurements as a constraint and using the surface value of the specific humidity from the radiosonde as a boundary condition. We used radiosonde humidity profile to estimate percentage of PWV spent up to the radar measurement reach. Figure 15 shows the results for the June 1999 cases (Figs. 15a,b) and for the January 2001 cases (Figs. 15c,d). Except for the first kilometer above the ground during the 1200 UTC 27 January 2001 case, the agreement of the radar derived humidity profile with the radiosonde observed profile is rather good for both 915-MHz and 449-MHz radars. Integrating humidity gradient profiles directly requires knowledge of the percentage of the GPS-observed total precipitable water used up to the height of the radar measurements. In general this is not possible without simultaneous radiosonde measurements but they are not available at the same temporal resolution as the radar measurements.
Stankov (1998) proposed a different approach for obtaining humidity profiles. It consists of using a statistical retrieval technique based on the RASS-measured temperature profiles, WPR-measured humidity gradient profiles, GPS-measured total precipitable water vapor, and satellite measured brightness temperatures, to obtain both the temperature and humidity profiles throughout the atmosphere. In this way, since the profiles cover the entire column of the atmosphere the GPS-obtained PWV applies directly. Gossard et al. (1999) compared humidity profiles obtained in the two different ways and found a good agreement between the methods.
6. Conclusions
An algorithm for computing the magnitude of the humidity gradient profiles from the zeroth-, first-, and second-moment measurements of the radar Doppler spectrum was developed and tested in four cases of widely different atmospheric conditions and using two different radar systems. The gradient algorithm builds upon the NOAA/ETL Advanced Signal Processing System (SPS) algorithm (Wolfe et al. 2001) at the ETL. This algorithm is based on identifying all the spectral peaks and using several continuity criteria to select the one peak that is associated with an atmospheric contribution.
We applied the method to 3-beam 915-MHz and 5-beam 449-MHz radar system measurements obtained in the tropical atmosphere over the ocean and a winter continental upslope icing storm in the Front Range of the Rockies, respectively. We found generally good agreement between the radar measured humidity gradient profiles and the humidity gradient profiles computed from the radiosonde measurements for four different soundings. In addition we found good agreement between the humidity profiles obtained by integrating the radar-derived humidity gradients and the radiosonde-observed humidity profiles. The more powerful 449-MHz radar system performed better than the 915-MHz system even though the 915-MHz system operated in the tropics where the humidity is much higher than the humidity in the Front Range of the Rocky Mountains during the winter. We attributed this lesser performance of the 915-MHz system to the presence of clouds and the fact that 915 MHz has a wider beam. However, our results are in agreement with the earlier results obtained with the 404-MHz system during the FIRE. This technique is applicable to radars with frequencies other than the ones used here. However, we expect that other high frequency radars such as 1250-MHz radar would have problems similar to 915-MHz radar and that the longer wavelength radars with narrow beam would perform better.
The humidity algorithm with only minor modifications can become part of the SPS which, if used at each Wind Profiler Network site, can serve to obtain much needed error characteristics necessary for the assimilation of the remotely sensed humidity gradient information directly into numerical forecast models and thus improve mesoscale model forecasts.
Acknowledgments
We thank Drs. Christopher Fairall and John Bates for providing support during this study, Dr. M. J. Post for leading an excellent scientific research cruise on board R/V Ronald H. Brown, Dr. Shelby Frisch for suggesting use of the Sloss and Atlas (1968) study for estimating the antenna beam broadening due to cross wind, Dan Law for providing information about losses in the radar transmission lines, and Dave Wuertz of NCDC who provided 6-s radiosonde data from Stapleton, Colorado.
REFERENCES
Battan, L. J., 1973: Radar Observations of the Atmosphere. University of Chicago Press, 323 pp.
Doviak, R. J., and Zrnic D. S. , 1984: Doppler Radar and Weather Observations. Academic Press, 458 pp.
Ecklund, W. L., Carter D. A. , and Balsley B. B. , 1988: A UHF profiler for the boundary layer: Brief description and initial results. J. Atmos. Oceanic Technol., 5 , 432–441.
Fairall, C. W., White A. B. , Edson J. B. , and Hare J. E. , 1997: Integrated shipboard measurements of the marine boundary layer. J. Atmos. Oceanic Technol., 14 , 338–359.
Gossard, E. E., 1990: Radar research on the atmospheric boundary layer. Radar in Meteorology, D. Atlas, Ed., Amer. Meteor. Soc., 477–527.
Gossard, E. E., Chadwick R. R. , Neff W. D. , and Moran K. P. , 1982: The use of ground based Doppler radars to measure gradients, fluxes and structure parameters in elevated layers. J. Appl. Meteor., 21 , 211–226.
Gossard, E. E., Welsh D. C. , and Strauch R. G. , 1990: Radar-measured height profiles of C2 N and turbulence dissipation rate compared with radiosonde data during October 1989 at Denver. Tech. Rep. ERL 442-WPL 63, 115 pp. [Available from NOAA/ERL/ETL, 325 Broadway, Boulder, CO 80305.].
Gossard, E. E., Strauch R. G. , Stankov B. B. , and Wolfe D. E. , 1995: Measurements of property gradients and turbulence aloft with ground-based Doppler radars. NOAA Tech. Memo. ERL 453- ETL 67, Environmental Technology Laboratory, Boulder, CO, 31 pp. [Available from the National Technical Information Service, 5285 Port Royal Rd., Springfield, VA 22161.].
Gossard, E. E., Wolfe D. E. , Moran K. E. , Paulus R. A. , Anderson K. D. , and Rogers L. T. , 1998: Measurement of clear-air gradients and turbulence properties with radar wind profilers. J. Atmos. Oceanic Technol., 15 , 321–342.
Gossard, E. E., Gutman S. , Stankov B. B. , and Wolfe D. E. , 1999: Profiles of radiorefractive index and humidity derived from radar wind profilers and the Global Positioning System. Radio Sci., 34 , 371–383.
Hildebrand, P. H., and Sekhon R. S. , 1974: Objective determination of the noise level in Doppler spectra. J. Appl. Meteor., 13 , 808–811.
Ottersten, H., 1969: Mean vertical gradient of potential refractive index in turbulent mixing and radar detection of CAT. Radio Sci., 4 , 1247–1249.
Ralph, F. M., Neiman P. J. , and Ruffieux D. , 1996: Precipitation identification from radar wind-profiler spectral moment data: Vertical velocity histograms, velocity variance, and signal power-vertical velocity correlations. J. Atmos. Oceanic Technol., 13 , 545–559.
Schroeder, J. A., Westwater E. R. , May P. T. , and McMillin L. M. , 1991: Prospects for temperature sounding with satellite and ground-based RASS measurements. J. Atmos. Oceanic Technol., 8 , 506–513.
Sloss, P. W., and Atlas D. , 1968: Wind shear and reflectivity gradient effects on Doppler radar spectra. J. Atmos. Sci., 25 , 1080–1089.
Stankov, B. B., 1996: Ground- and space-based temperature and humidity retrievals: Statistical evaluation. J. Appl. Meteor., 35 , 444–463.
Stankov, B. B., 1998: Multisensor retrieval of atmospheric properties. Bull. Amer. Meteor. Soc., 79 , 1835–1854.
Stankov, B. B., Westwater E. R. , and Gossard E. E. , 1996: Use of wind profiler estimates of significant moisture gradients to improve humidity profile retrieval. J. Atmos. Oceanic Technol., 13 , 1285–1290.
Tsuda, T., Miyamoto M. , and Furumoto J. , 2001: Estimation of a humidity profile using turbulence echo characteristics. J. Atmos. Oceanic Technol., 18 , 1214–1222.
Weber, B. I., and Wuertz D. B. , 1991: Quality control algorithm for profiler measurements of winds and RASS temperatures. NOAA Tech. Memo. ERL WPL-212, 32 pp.
Weber, B. L., Wuertz D. B. , and McPeek R. , 1993: Quality control for profiler measurements of winds and RASS temperatures. J. Atmos. Oceanic Technol., 10 , 452–464.
Westwater, E. R., Stankov B. B. , Cimini D. , Han Y. , Shaw H. A. , Lesht B. M. , and Long C. N. , 2003: Radiosonde humidity soundings and microwave radiometers during Nauru99. J. Atmos. Oceanic Technol., in press.
White, A. B., 1997: Radar remote sensing of scalar and velocity microturbulence in the convective boundary layer. NOAA Tech. Memo. ERL ETL-276, Environmental Technology Laboratory, Boulder, CO, 127 pp. [Available from NOAA/ERL/ETL, 325 Broadway, Boulder, CO 80305.].
White, A. B., and Fairall C. W. , 1991: Radar observations of humidity variability in and above the marine atmospheric boundary layer. J. Atmos. Oceanic Technol., 8 , 639–658.
White, A. B., Lataitis R. J. , and Lawrence R. S. , 1999: Space and time filtering of remotely sensed velocity turbulence. J. Atmos. Oceanic Technol., 16 , 1967–1971.
Wolfe, D. E., Weber B. L. , Wilfong T. L. , Welsh D. C. , Wuertz D. B. , and Merritt D. A. , 2001: An advanced signal processing system for radar wind profilers. 11th Symp. on Meteorological Observations and Instrumentation, Albuquerque, NM, Amer. Meteor. Soc., 339–344.
Woodman, R. F., 1985: Spectral moment estimation in MST radars. Radio Sci., 20 , 1185–1195.
Wuertz, D. B., Weber B. L. , Strauch R. G. , Frisch A. S. , Little C. G. , Merritt D. A. , Moran K. P. , and Welsh D. C. , 1988: Effects of precipitation on UHF wind profiler measurements. J. Atmos. Oceanic Technol., 5 , 450–465.
Wind profiler characteristics and operating parameters