The Atmospheric Radiation Measurement Program Cloud Profiling Radars: Second-Generation Sampling Strategies, Processing, and Cloud Data Products

Pavlos Kollias Atmospheric Science Division, Brookhaven National Laboratory, Upton, New York

Search for other papers by Pavlos Kollias in
Current site
Google Scholar
PubMed
Close
,
Mark A. Miller Atmospheric Science Division, Brookhaven National Laboratory, Upton, New York

Search for other papers by Mark A. Miller in
Current site
Google Scholar
PubMed
Close
,
Edward P. Luke Atmospheric Science Division, Brookhaven National Laboratory, Upton, New York

Search for other papers by Edward P. Luke in
Current site
Google Scholar
PubMed
Close
,
Karen L. Johnson Atmospheric Science Division, Brookhaven National Laboratory, Upton, New York

Search for other papers by Karen L. Johnson in
Current site
Google Scholar
PubMed
Close
,
Eugene E. Clothiaux Department of Meteorology, The Pennsylvania State University, University Park, Pennsylvania

Search for other papers by Eugene E. Clothiaux in
Current site
Google Scholar
PubMed
Close
,
Kenneth P. Moran NOAA/ESRL/Physical Sciences Division, Boulder, Colorado

Search for other papers by Kenneth P. Moran in
Current site
Google Scholar
PubMed
Close
,
Kevin B. Widener Pacific Northwest National Laboratory, Richland, Washington

Search for other papers by Kevin B. Widener in
Current site
Google Scholar
PubMed
Close
, and
Bruce A. Albrecht Rosenstiel School of Marine and Atmospheric Science, University of Miami, Miami, Florida

Search for other papers by Bruce A. Albrecht in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

The U.S. Department of Energy Atmospheric Radiation Measurement (ARM) Program operates millimeter-wavelength cloud radars in several climatologically distinct regions. The digital signal processors for these radars were recently upgraded and allow for enhancements in the operational parameters running on them. Recent evaluations of millimeter-wavelength cloud radar signal processing performance relative to the range of cloud dynamical and microphysical conditions encountered at the ARM Program sites have indicated that improvements are necessary, including significant improvement in temporal resolution (i.e., less than 1 s for dwell and 2 s for dwell and processing), wider Nyquist velocities, operational dealiasing of the recorded spectra, removal of pulse compression while sampling the boundary layer, and continuous recording of Doppler spectra. A new set of millimeter-wavelength cloud radar operational modes that incorporate these enhancements is presented. A significant change in radar sampling is the introduction of an uneven mode sequence with 50% of the sampling time dedicated to the lower atmosphere, allowing for detailed characterization of boundary layer clouds. The changes in the operational modes have a substantial impact on the postprocessing algorithms that are used to extract cloud information from the radar data. New methods for postprocessing of recorded Doppler spectra are presented that result in more accurate identification of radar clutter (e.g., insects) and extraction of turbulence and microphysical information. Results of recent studies on the error characteristics of derived Doppler moments are included so that uncertainty estimates are now included with the moments. The microscale data product based on the increased temporal resolution of the millimeter-wavelength cloud radars is described. It contains the number of local maxima in each Doppler spectrum, the Doppler moments of the primary peak, uncertainty estimates for the Doppler moments of the primary peak, Doppler moment shape parameters (e.g., skewness and kurtosis), and clear-air clutter flags.

** Current affiliation: Department of Atmospheric and Oceanic Sciences, McGill University, Montreal, Quebec, Canada

Corresponding author address: Pavlos Kollias, Department of Atmospheric and Oceanic Sciences, McGill University, 805 Sherbrooke Street West, Montreal, QC H3A 2K6, Canada. Email: pavlos.kollias@mcgill.ca

Abstract

The U.S. Department of Energy Atmospheric Radiation Measurement (ARM) Program operates millimeter-wavelength cloud radars in several climatologically distinct regions. The digital signal processors for these radars were recently upgraded and allow for enhancements in the operational parameters running on them. Recent evaluations of millimeter-wavelength cloud radar signal processing performance relative to the range of cloud dynamical and microphysical conditions encountered at the ARM Program sites have indicated that improvements are necessary, including significant improvement in temporal resolution (i.e., less than 1 s for dwell and 2 s for dwell and processing), wider Nyquist velocities, operational dealiasing of the recorded spectra, removal of pulse compression while sampling the boundary layer, and continuous recording of Doppler spectra. A new set of millimeter-wavelength cloud radar operational modes that incorporate these enhancements is presented. A significant change in radar sampling is the introduction of an uneven mode sequence with 50% of the sampling time dedicated to the lower atmosphere, allowing for detailed characterization of boundary layer clouds. The changes in the operational modes have a substantial impact on the postprocessing algorithms that are used to extract cloud information from the radar data. New methods for postprocessing of recorded Doppler spectra are presented that result in more accurate identification of radar clutter (e.g., insects) and extraction of turbulence and microphysical information. Results of recent studies on the error characteristics of derived Doppler moments are included so that uncertainty estimates are now included with the moments. The microscale data product based on the increased temporal resolution of the millimeter-wavelength cloud radars is described. It contains the number of local maxima in each Doppler spectrum, the Doppler moments of the primary peak, uncertainty estimates for the Doppler moments of the primary peak, Doppler moment shape parameters (e.g., skewness and kurtosis), and clear-air clutter flags.

** Current affiliation: Department of Atmospheric and Oceanic Sciences, McGill University, Montreal, Quebec, Canada

Corresponding author address: Pavlos Kollias, Department of Atmospheric and Oceanic Sciences, McGill University, 805 Sherbrooke Street West, Montreal, QC H3A 2K6, Canada. Email: pavlos.kollias@mcgill.ca

1. Introduction

The U.S. Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) Program (e.g., Stokes and Schwartz 1994) millimeter-wavelength cloud radars (MMCRs; Moran et al. 1998) are one of the primary observing tools for quantifying the properties of nearly all radiatively important clouds over the ARM Climate Research Facility (ACRF) field research sites (e.g., Ackerman and Stokes 2003). This includes a wide range of cloud types, from shallow fair weather cumuli and stratus in the boundary layer to thin cirrus and convective anvils in the upper troposphere. These radars were designed to provide accurate radar reflectivity measurements over a dynamic range of approximately −50 to +20 dBZ. The need to detect clouds with reflectivities below −40 dBZ at different altitudes dictates the use of multiple operational modes for the cloud radar that can be cycled repetitively. Clothiaux et al. (1999) described the operational parameters currently installed on the ARM Program cloud radars. These operational modes were designed for maximum sensitivity, given the receiver processor on which they were run, while attempting to maintain accurate estimates of hydrometeor reflectivities and atmospheric motions.

The Environmental Technology Laboratory (ETL) of the National Oceanic and Atmospheric Administration (NOAA) recently completed an upgrade of the digital signal processors (DSPs) for the ARM Program cloud radars, vastly improving their signal processing efficiency (Widener et al. 2004). The new board with C-40 DSPs from Texas Instruments is actually a 5-DSP board processor. Advantages of the new board are use of higher clock frequencies and multiple DSPs working in parallel to accelerate radar processing power. However, Texas Instruments manufactured only three analog-to-digital C-40 boards before the C-40 processor board was slated for discontinuation. At the same time, the National Center for Atmospheric Research (NCAR) developed a PC-Integrated Radar Data Acquisition System (PIRAQ). The new PIRAQ-III was selected for upgrading the ARM Program Tropical Western Pacific (TWP) sites. The processor efficiency of the PIRAQ-III is close to 70%–75%, an improvement over the C-40 processors, which have efficiencies of approximately 50%. As a consequence of the upgrade, a four-mode sequence that took 36 s on the old processors for completion (Clothiaux et al. 2000) is now completed in 5–8 s (5 s for the PIRAQ-III and 8 s for the C-40), with simultaneous recording of Doppler spectra. Installation of the new processors (e.g., PIRAQ-III) in the ARM cloud radars was completed at all the ACRF sites in 2006.

Recent evaluations of ARM Program cloud radar signal processing performance (Kollias et al. 2005) and high-temporal-resolution data from the University of Miami (UM) 94-GHz Doppler cloud radar (e.g., Kollias et al. 2001) have identified problems and limitations in the current processing paradigm of Clothiaux et al. (1999, 2000). Improvements that can, and should, be implemented on the new processor data from the ARM Program cloud radars include higher temporal resolution, wider Nyquist velocities through reduction of the number of coherently integrated pulses, operational dealiasing of the Doppler spectra, removal of pulse coding in the boundary layer, and continuous recording of Doppler spectra. To realize these improvements, while ameliorating current problems and limitations, a new set of operational modes has been implemented.

With continuous recording of Doppler spectra new methods of postprocessing become available. The postprocessing of Doppler spectra for insect-generated clutter and cloud signatures improves the quality of the Doppler moments and allows generation of parameters that describe the shapes of the Doppler spectra from which the moments are derived. These parameters characterize the number of local maxima in the spectra and the shape of spectra near local maxima, which ultimately are related to cloud microphysics and turbulence. The cloud data produced by these enhanced radar systems are much better suited for cloud and radiation studies. The microscale data product based on the new temporal resolution (i.e., ≤2 s or less) of the millimeter-wavelength cloud radars is described that contains the number of local maxima in each Doppler spectrum, the Doppler moments of the primary peak, uncertainty estimates for the Doppler moments of the primary peak (e.g., Doviak and Zrnic 1993), Doppler moment shape parameters (e.g., skewness and kurtosis), and clear-air clutter (insect) flags.

In the following section, the limitations of the past operational modes of the ARM cloud radars are described. Sections 3 and 4 detail the new operational settings and new sampling strategy of the ARM cloud radars that are currently operational. Section 5 describes the postprocessing of the recorded Doppler spectra, and section 6 outlines the new microscale data products derived from the postprocessing of the Doppler spectra.

2. Limitations of past operational modes of the cloud radars

Before describing the new operational modes and sampling strategy for the ARM Program millimeter-wavelength cloud radars, we first consider briefly the previous operational modes (Clothiaux et al. 1999) and their impact on the Active Remote Sensing of Clouds (ARSCL) value added products (VAPs; Clothiaux et al. 2000). The four previous operational modes of these radars are a boundary layer (BL) mode, a cirrus (CI) mode, a general (GE) mode, and a precipitation (PR) mode. The parameters of the previous operational modes that influence radar signal dwell and processing (SDP) and mode sensitivity are listed in Table 1. These parameters are the interpulse period (τipp), the number of coded bits (Nbits), the number of coherent averages (Ncoh), the number of fast Fourier transform (FFT) points (Nfft), the minimum range (Rmin) of useful data, the unambiguous range (Ru), the number of samples (Ns), the sampling rate (SR), the signal dwell time (SDT), the unambiguous velocity (Vu), the velocity resolution (ΔVu), and the temporal resolution (Ts) of the Doppler spectra and moments that result from this set of parameters.

Kollias et al. (2005) identified several problems in the parameters listed in Table 1. First, the sampling interval (Ts) of 9 s is much too coarse to resolve boundary layer cloud motions. Second, the unambiguous velocity range, or Nyquist velocity (Vu), is too small (3.0–3.5 m s−1) for the BL, CI, and GE modes under a variety of cloud conditions, leading to deleterious filtering of received power as a result of coherent averaging. Third, velocity resolutions (ΔVu) for the BL, CI, and GE modes are too coarse, often placing the Doppler spectrum power from boundary layer clouds into a single Doppler spectrum velocity bin. Fourth, treatment of noise in the processing of the Doppler spectra to Doppler moments produces spurious moments for low signal-to-noise ratio (SNR) returns. Fifth, nonlinear compression (i.e., receiver saturation) of voltages at the top of the PR mode dynamic range leads to reflectivities of questionable accuracy. The PR mode was designed to have sufficiently wide Nyquist velocity boundaries to observe large downward Doppler velocities of drizzle droplets and raindrops but saturates even in the presence of light precipitation with reflectivities of +15 dBZ. And sixth, inspection of BL mode returns from low-level clouds has indicated subtle problems with pulse coding of this mode, leading to biases in cloud-top heights of 100–200 m (Fig. 1a) and a minimum range of 465 m.

Analysis of the ARSCL data product indicates relatively weak representation of boundary layer clouds that is attributable to their shallow and broken nature, especially at the TWP Nauru site, and the presence of insects and other clutter at the Southern Great Plains (SGP) site. The 9-s dwell and processing time for the BL mode smears in time the cloud reflectivity and Doppler velocity fields and contaminates the Doppler moments with undesired contributions from gradients of reflectivity and wind shear that pass through the radar sample volume during the 9-s dwell. As a result of their proximity to the radar, boundary layer clouds are adequately detected by the noncoded, less-sensitive GE mode (Fig. 1b). The problems inherent in the 8-bit coding of the BL mode (e.g., Schmidt et al. 1979; Wakasugi and Fukao 1985) outweigh the advantages of the 9-dB gain in system sensitivity, and thus there is minimal use of BL mode data in cloud studies and data products based on all four previous modes (Figs. 1c,d). The 32-bit coding of the cirrus mode, with a 15-dB enhancement of system sensitivity, has proven to be valuable in detecting high-level clouds (Figs. 1c,d). These findings suggested elimination of pulse coding from the BL mode and preservation of 32-bit pulse coding of the CI mode.

Through a set of theoretical simulations and observations from the ARM Program cloud radars and the UM 94-GHz cloud radar (Albrecht and Kollias 1999), Kollias et al. (2005) demonstrated each of the problems outlined above and proposed a refinement in the operational parameters that removed it. They also provided criteria against which to judge radar system sensitivity and the minimum detectable signal for each mode of operation.

Overall, the four previous operational modes have 9-s temporal spacing with dwell times uniformly spread over the 9 s and varying from 0.4 to 3.0 s. Thus, previous digital signal processor efficiency, defined as the ratio of the dwell time to sample temporal spacing (i.e., the sum of the dwell and processing times), varied from 5% to 30%. The new processors substantially improve processing efficiencies of these cloud radars, leading to efficiencies of 50%–70% for the new modes. For optimal processing of radar returns, a processor that misses no pulses and processes the radar data in parallel with radar sampling is required. While such technology exists, the design of the new processor for the ARM Program cloud radars precludes the simultaneous acquisition and processing of radar samples.

Our approach in the new operational modes is to limit the SDP time to approximately 2 s or less for each of the modes. A fundamental limitation in the reduction of the signal dwell, which is equivalent to maximization of the radar sample production rate, is the interpulse period τipp, which cannot be reduced below a value for which there is an appropriate maximum unambiguous range with minimal impact of second trip echoes. Consequently, the new operational modes presented here are once again a compromise among competing effects and represent only small, albeit significant, changes from the previous operational modes.

In addition to the issues identified in the previous set of operational parameters (Kollias et al. 2005), the merging strategy for data from the four modes developed by Clothiaux et al. (2000) was revised. The goal of the processing by Clothiaux et al. (2000) was to produce best estimates of the Doppler moments with a nominal temporal resolution of 10 s. Although the atmospheric column was sampled in 36 s (i.e., the time period required for a complete four-mode cycle), the merged data contained estimates of Doppler moments every 10 s obtained from mode data comparison and interpolation. The new processor upgrade eliminates the need for mode data interpolation since the four-mode cycle is now completed in less than 10 s. While this alone constitutes a significant improvement over previous data acquisition and processing, we nonetheless decided to alter the mode sequence and sample more frequently the lower portion of the troposphere, where coupling with surface processes and high levels of turbulence produce fast changing and complicated cloud structures. In the following section we discuss the new operational modes, the new postprocessing methods, and the data products that result from them.

3. New operational parameters of the cloud radars

The number of coded pulses (Nbits) in the previous set of operational modes was 8 for the BL mode and 32 for the CI mode. In the new operational modes (Table 2) the pulse coding for the BL mode is removed, while pulse coding for the CI mode is preserved. We reduced the number of coherent integrations for each of the three modes that use this approach in order to increase the Nyquist velocity of the Doppler spectra. Wider Nyquist velocities better accommodate large-bandwidth cloud signals at all of the sites and reduce the deleterious effects of coherent integration of these signals (Kollias et al. 2005). By reducing the number of coherent integrations, a few decibels of sensitivity are lost. Kollias et al. (2005) have used simulations of Doppler radar cloud signals to demonstrate that the apparent loss of sensitivity as a result of fewer coherent integrations might actually be compensated by a real gain in sensitivity by keeping the receiver bandwidth large compared to the signal bandwidth of turbulent clouds.

The operational parameters of the PR mode remain unchanged in the new scheme. The PR mode does lose 20–25 dB of sensitivity with its increase in dynamic range at high signal levels obtained by switching of one of the four receiver transmit/receive (T/R) circulators into the closed position during signal reception. That is, the received atmospheric signal is attenuated at the front of the radar receiver, before the mixer and preamplifier, in order to preserve the receiver noise level and to maintain receiver calibration. This change in the PR mode prevents the radar from saturating in light to moderate drizzle, allowing for accurate measurements of radar reflectivity in these conditions.

A polarization (PO) mode was introduced at the SGP site that provides co- and cross-channel Doppler spectra and moments. During the PO mode, returns from (right-hand) circularly polarized transmitted signals are received by both left- (co-channel) and right- (cross-channel) hand circular receivers on a pulse-to-pulse basis. Utilizing the strong cross-channel returns [i.e., large circular depolarization ratios (CDRs)] of nonspherical scatterers (e.g., insects), the PO mode is useful for identifying them in the boundary layer. While the new PO mode is useful for insect studies, it is not useful for cloud studies due to the poor antenna cancellation ratio of 15 dB for the radar antenna at the SGP site. For a vertically pointing radar the circular depolarization ratio of a particle depends on its shape but not on its orientation in the polarization plane. The reflectivity differential ratio (ZDR) of any particle type is almost always 0 dB for a vertically pointing radar, thereby providing little useful information. The linear depolarization ratio (LDR) depends on both particle shape and orientation and is usually low for vertically pointing radars (S. Matrosov 2004, personal communication). Overall, circular depolarization ratio improves our ability, relative to reflectivity differential and linear depolarization ratios, to distinguish between particle habits (e.g., hexagonal columns or irregular crystals) and also provides for larger signals from them (e.g., Jameson 1987; Matrosov et al. 1996). Future ARM Program plans include installation of a PO mode on the North Slope of Alaska (NSA) cloud radar and removal of any antenna cancellation problems for this radar.

If the previous mode sequence (i.e., cycling between the BL, CI, GE, and PR modes) were adopted for the new processors, there would be no compelling reason to alter further the mode parameters, except for reducing the number of coherent integrations in all but the PR mode with resultant higher Nyquist velocities, increasing the 64-point FFT of the BL mode to 256 points, the CI and GE mode 64-point spectra to 128 points, and the 128-point spectra of the PR mode to 256 points. With these changes the velocity resolutions of the mode Doppler spectra would increase to approximately 5–6 cm s−1 with a four-mode repetition time of 10 s, which is down by a factor of 3.6 from the current repeat time of 36 s. In this configuration, the ARSCL data product would have a true 10-s resolution, as opposed to an interpolated resolution of 10 s with a true temporal resolution somewhere between 9 and 36 s.

In the new mode cycling that is discussed in the following section, the main objectives are mode repeat times that are sufficiently fast to capture cloud motions, with at least two boundary layer sampling periods per 3–4 s, and Nyquist velocities sufficiently wide to accommodate turbulent clouds. All of the signal dwell times in Table 2 are less than approximately 1 s. While the CI mode does not suffer a significant loss of sensitivity from the previous (Table 1) to the new (Table 2) modes, there is a significant apparent reduction in the sensitivity of the new BL mode. However, as we have already discussed, BL mode sensitivity in the previous mode sequence was often not realized because of pulse compression artifacts and power losses associated with large bandwidth cloud signals. The previous GE mode data were similar to the BL mode data but with no pulse compression artifacts; hence, it tended to be of more use in boundary layer cloud studies. General mode sensitivity is also somewhat reduced (i.e., 1–2 dB) in the new scheme due to the collection of a smaller number of radar samples, but the increase in bandwidth of this and the BL mode, with smaller resulting losses from coherent integration of turbulent cloud signals, will offset some of these losses. Overall, the parameters of the new mode sequence (Table 2) reduce radar sensitivity primarily for the BL mode. If we account for the modes cycling 4–5 times faster, we anticipate that our capability to observe all cloud types is actually enhanced.

4. New mode sequence for the cloud radars

In the new mode sequence, use of different modes is still dictated by the objectives of observing clouds at all altitudes, including clouds with low reflectivities in the range of −45 to −40 dBZ, and producing accurate Doppler moment estimates for them. As a result, we keep all four modes since they serve different purposes toward these objectives (Clothiaux et al. 1999). We could adopt the same exact sequence of modes as Clothiaux et al. (1999) with implementation of the new processors, decreasing the temporal spacing between modes to 1.5–2.0 s with an overall cycle time of 5–8 s (5 s for the PIRAQ, 8 s for the C-40). In this paradigm all four modes would be used to determine the best Doppler moments profile at the 10-s resolution of the ARSCL data product. This would substantially improve the quality of this product by eliminating the need to interpolate to 10-s resolution and by limiting the averaging of radar signals across 2.5-s dwell and processing intervals. Such an approach, however, does not eliminate poor sampling of boundary layer clouds.

To improve observation of boundary layer clouds we implemented the following sampling strategy: boundary layer, to general, to boundary layer, to cirrus, to boundary layer, to general, to boundary layer, to precipitation or polarization mode:
i1520-0426-24-7-1199-eq1
Each mode has a ∼1.5-s or less signal dwell and processing period. In this eight-mode sequence there are four BL mode dwells, two GE mode dwells, and one dwell each for the CI and PR or PO modes. At the SGP site the PR and PO modes alternate from one cycle to the next, whereas other sites use the PR mode for each cycle. This proposed scheme observes turbulent and fast-changing clouds in the boundary layer with better fidelity than the previous scheme, with modes sampling the upper troposphere once every 3–4 s. Every mode is sampled on a fixed temporal spacing, allowing for spectral analysis of their data. Ample evidence from the ARSCL data product from the various sites supports frequent sampling of the BL mode. The repeat cycle for this eight-mode sequence is approximately 12 s. The upgrade of the processors with the new operational modes results in 12 BL, 8 GE, 2 CI, and 2 PR, or PO, mode dwells in 36 s, as compared to a total of 4 dwells with the old processors. This is a substantial improvement in the temporal sampling of the cloud radars. As a result, we consider new cloud data products that utilize the higher-temporal-resolution data with their accompanying Doppler spectra.

5. Postprocessing of cloud radar Doppler spectra

In the ARM Program, FFTs are applied to the cloud radars’ received signals to produce a Doppler spectrum for each range gate at each 1.5-s time step. Standard algorithms (e.g., Sirmans and Bumgarner 1975; Doviak and Zrnic 1993) are subsequently applied to the Doppler spectra for estimation of the first three moments: radar reflectivity (zeroth moment), mean Doppler velocity (first moment), and Doppler spectrum width (second moment). The Hildebrand and Sekhon (1974) scheme is used to estimate the noise floor of the Doppler spectrum. In the past, the first three moments of the Doppler spectra were archived, and the Doppler spectra themselves were discarded (Kollias et al. 2005). For a Gaussian-shaped Doppler spectrum the first three Doppler moments do contain most of the information in it. Doppler spectra from clouds, however, are often not Gaussian-shaped.

Since April 2004, the cloud radar receivers at the SGP and NSA sites have been upgraded with new C-40 processors. The upgrades of the TWP site cloud radar receivers were completed in early 2006 using the new PIRAQ-III processors (Widener et al. 2004). The new processors at all of the sites offer the 1.5-s dwell and sampling period, 256-point FFTs, and archival of the Doppler spectra for all times at each range gate. Due to the new higher-temporal-resolution sampling strategy and continuous recording of Doppler spectra, 15 GB of radar data are generated per day per site. The volume of Doppler spectra data is reduced on site by removing spectra with no apparent atmospheric signals as determined by an SNR-based cloud mask. The first step of the mask is calculation of the mean and standard deviation of the SNR for a small subset of echo-free (i.e., clutter- and cloud-free) returns during each radar profile for each mode. The mean plus two standard deviations is then used as a dynamic threshold in subsequent processing. The number of power returns within each 5 × 5 (time × height) box of returns that exceed the threshold are tallied. If this number is equal to or greater than 9, the power return at the center of the 5 × 5 box is labeled as a significant return (e.g., Clothiaux et al. 1995; Uttal et al. 1993). Those Doppler spectra with significant atmospheric signals are archived and postprocessed (Fig. 2).

The first step in Doppler spectra postprocessing removes any residual dc spectral component. The dc spectral component is replaced using linear interpolation of spectral densities from FFT points neighboring to the zero frequency spectra bin. Singly wrapped aliased Doppler spectra are subsequently dealiased, which is a straightforward process (the FFT point with the highest spectral density is set as the half-way point of the dealiased Doppler spectra), and the Hildebrand and Sekhon (1974) noise estimation technique is applied to them (Figs. 3a,b). In the next step, all contiguous Doppler spectral points separated by FFT points with powers below the noise are labeled as a Doppler spectral peak (Figs. 3c,d). To reduce the influence of noise and eliminate the noise spikes and narrow, isolated clear-air returns from insects, only Doppler spectra peaks with five or more contiguous Doppler spectra points (e.g., 20 cm−1 or more in Doppler velocity) above the noise are labeled as hydrometeor-induced Doppler spectra peaks. This two-step threshold technique is based on Clothiaux et al. (1994) and facilitates processing of multimodal Doppler spectra. The peak with the maximum spectral bin power is identified as the primary Doppler spectral peak (Figs. 3c,d). If the maximum Doppler spectral bin power is more than 30 dB above the noise threshold, it may have an image at the bin location with the same Doppler velocity magnitude but opposite sign. If an image does exist for the peak with maximum power, the image peak is rejected from further postprocessing (Figs. 3c,d). The presence of a spectral image is attributed either to gain and phase imbalance (Barber 1947) of the in-phase (I) and quadrature-phase (Q) signal components at the output of the I/Q demodulator or to nonlinear matched-filter effects within the receiver. Difficulties in implementing a 90° shift to create the Q signal generate gain imbalances between I and Q signals, dc offsets, amplitude differences, and phase rotations.

The first steps of Doppler spectra postprocessing eliminate radar hardware- and software-related artifacts in the Doppler spectra and identify the primary Doppler spectral peaks. Next, an insect mask (Fig. 2) is applied to BL mode Doppler spectra from all of the ARM Program sites, which identifies signatures of insects in the Doppler spectral peaks and classifies clutter-contaminated peaks as such (Fig. 3e) (N.B., the use of “clutter” in this manuscript refers to radar returns from airborne insects that are frequently observed in the boundary layer). The clutter mask is based on a neural network algorithm (e.g., Luke et al. 2006; Kollias et al. 2006) that utilizes information of the morphology of the primary Doppler spectral peaks, including slopes, curvatures, and integrated peak powers, to identify clutter (Fig. 3e). Insects are generally associated with extremely narrow, steeply sloped peaks and are a major problem at the SGP site.

The clutter mask was extensively tested on a large summertime dataset from the SGP site and was highly consistent with ceilometer cloud-base detections and polarimetric measurements from the cloud radar (Fig. 4). Small deficiencies in the technique occurred near cloud boundaries, with turbulence near the boundaries leading to noisy results, and for quiescent periods of time when cloud returns were weak and exhibited almost no turbulent broadening, producing narrow Doppler spectral peaks indistinguishable from those of insects. For this small subset of observations, with reflectivities less than −30 dBZ and Doppler spectral widths less than 10 cm s−1, circular depolarization ratios from the PO mode are used as an additional input to the neural network scheme. If the CDR is more than −7 dB for a radar return, the return is classified as a clutter return (Fig. 4). For clutter-contaminated range gates we estimate the three Doppler moments for the primary Doppler spectral peak, and no further processing is performed. Overall, the Doppler spectra–based clutter mask has a 95% success rate in identifying correctly the presence of insects in the radar resolution volume (Luke et al. 2006).

If a Doppler spectrum is classified as a hydrometeor return and the Doppler spectrum width of the primary peak exceeds a threshold value (0.1 m s−1), additional processing on the peak is performed that attempts to assess the presence of multiple peaks within the apparent single-peak spectrum (Figs. 3c,d). That is, an attempt is made to identify multimodal Doppler spectral peaks. Careful analyses of Doppler spectra collected from millimeter-wavelength cloud radars during various field experiments (e.g., Shupe et al. 2004) have demonstrated that many Doppler spectra have multimodal features resulting from mixed-phase cloud processes, coexisting cloud and drizzle drop size distributions, and collections of different ice crystals. While multimodal Doppler spectra occur in only a fraction of the observations, their study will improve our understanding of cloud and precipitation processes.

The multimodal-peak detection algorithm is based on a spectral densities sorting algorithm that identifies local maxima and minima using a 3-dB decrease or increase in spectral density between local “peaks” and “valleys” for their identification (e.g., Sato et al. 1990; Kollias et al. 2003). The total number of significant Doppler spectral peaks, either clearly separated by spectral bins with noise or within the primary Doppler spectral peak (e.g., Figs. 3c,d), is an output parameter of this processing. Time–height mapping of the number of significant Doppler spectral peaks emphasizes the presence of areas that contain particles of different size distributions or phase and are of interest in the study of small-scale cloud and precipitation processes.

The next step in postprocessing of Doppler spectra is to assign uncertainty and bias estimates to the three Doppler moments of the primary Doppler spectral peak (Fig. 5). These uncertainty and bias estimates for the cloud radar Doppler moments were obtained from extensive numerical simulations of radar signals for a variety of SNRs and signal bandwidths (Doppler spectrum widths σD; Kollias et al. 2005). Lookup tables for a range of SNRs and signal bandwidths consistent with the parameters of each operational mode were generated. As Fig. 5 illustrates, the uncertainties are insignificant for large SNRs and narrow signal bandwidths. Large uncertainties are present at low SNRs and large signal bandwidths. Fortunately, few cloud radar returns have these values.

The large negative bias in radar reflectivity during wideband (large signal bandwidth σD) conditions results from use of coherent integration and the implicit assumption that the full gain of coherent integration [i.e., 10log10(Ncoh)] is applicable for all signal bandwidth conditions (Kollias et al. 2005). Large signal bandwidths lead to radar returns that decorrelate quickly in time. Under such conditions coherent integration is not as efficient as it is for more narrow signals. Using these lookup tables, each Doppler moment (i.e., reflectivity Ze, mean Doppler velocity VD, and Doppler spectrum width σD) is assigned an uncertainty and bias estimate, and these values are reported in the processing output.

In addition to the standard three Doppler moments, the third (i.e., skewness) and fourth (i.e., kurtosis) moments of the primary Doppler spectral peak are estimated. These higher Doppler moments are highly sensitive to small deviations of the peak from a Gaussian shape. The left- and right-hand slopes of the primary Doppler spectral peak are also estimated using a linear fit from the edge of the peak to the bin location of maximum spectral power. The motivation for estimating these four additional parameters can be found in Gossard et al. (1997) and Kollias et al. (2001). According to these earlier studies, the shape of Doppler spectral peaks, as described by these additional parameters, is modulated by two primary physical mechanisms: microphysics, or the shape of the hydrometeor size distribution, and small-scale turbulence unresolved within the radar resolution volume. Under certain conditions, these additional parameters can be used for extraction of microphysical and turbulence properties in clouds and precipitation.

The same Doppler spectra postprocessing will be applied to the ARM Program cloud radars operating at 94 GHz, such as the ones now operated at the SGP site (Mead and Widener 2005) and within the program’s mobile facility. Due to scattering by particles large compared to the wavelength, 94-GHz radar Doppler spectra often exhibit oscillations with increasing downward Doppler velocities, that is, with increasing sizes of large raindrops and snowflakes (Kollias et al. 2002). These oscillations are an additional source of multimodality within the Doppler spectra. In the presence of precipitation this phenomenon modulates the Doppler spectra, creating bimodal, and even trimodal, Doppler spectra from single-mode precipitation drop size distributions (e.g., Lhermitte 1988; Firda et al. 1999; Kollias et al. 2003). Once Doppler spectral peaks that result from this modulation are identified, the first minimum in the spectrum can be used to estimate the mean vertical air motion. The Doppler spectrum can then be shifted by this velocity to produce a still-air Doppler spectrum, and a nonlinear least squares fitting procedure can be applied to it for the retrieval of the raindrop size distribution (Kollias et al. 2002).

6. New “active remote sensing of clouds” data products

Doppler spectra postprocessing that is more sophisticated than current (and conventional) methods is used for estimation of the three Doppler moments (i.e., reflectivity Ze, mean Doppler velocity VD, and Doppler spectrum width σD). The new processing provides information on the number of peaks in each spectrum together with the Doppler moments for the primary peak. The new set of data products, the microscale data products (“micro” for microphysics), are produced after postprocessing of the recorded Doppler spectra from each mode (Fig. 6). There is one product for each mode with no merging of data from different modes. The microscale data products have the temporal resolutions of the modes from which they are derived, that is, 3–4 s for the BL mode, 6–7 s for the GE mode, 12–14 s for the CI mode, and 12–14 s for the PR/PO mode.

The microscale data products include the three Doppler moments based on all of the spectral bin power estimates above the noise (Fig. 7a), the total number of peaks in each spectrum resulting from mixed-phase clouds or cloud and drizzle particle-size distributions (Fig. 7b), the three Doppler moments of the primary peak, uncertainty estimates for both the Doppler moments based on all bin powers above the noise and those for the primary peak (Fig. 7c), Doppler spectrum asymmetry information (e.g., Doppler spectrum skewness) (Fig. 7d), clutter mask flags (Fig. 4), and flags that characterize the overall quality of the Doppler spectra upon which the Doppler moments are based, including information on Doppler spectra aliasing, in-phase and quadrature-phase power imbalances, and other Doppler spectra artifacts. These new microscale data products facilitate new studies using ARM Program cloud radar data. Time–height maps of uncertainty estimates for radar reflectivity (Fig. 7c), as well as of the other two Doppler moments, can be used as input to retrieval techniques to quantify retrieval accuracy. In Fig. 7b, the pockets of bi- and trimodal Doppler spectra (see Fig. 3b for an example of these Doppler spectra) are located above the melting layer and reveal the presence of large and small ice-particle distributions. Such time–height information on the presence of multimodal size distributions facilitates identification of Doppler spectra rich in microphysical and turbulence information (e.g., Shupe et al. 2004) and offers new insights on cloud and precipitation processes. The skewness of Doppler spectra peaks (Fig. 7d) in conjunction with the slope estimates from linear fits to the left- and right-hand sides of the primary peaks of the Doppler spectra, can be used to infer information about the particle-size distributions and turbulence intensity (e.g., Gossard et al. 1997; Kollias et al. 2001). The information derived from the new Doppler spectra postprocessing, together with the substantial increase in temporal resolution of the ARM Program cloud radar data, will make the new microscale data products suitable for the study of cloud microphysics and turbulence (e.g., Kato et al. 2001; O’Connor et al. 2005).

7. Summary

Recent upgrades of the Atmospheric Radiation Measurement Program cloud radars have led to new radar operational parameters, mode sequences, and microscale data products. The temporal resolution of these cloud radars is now 1.5–2.0 s per mode, a significant reduction from the previous sampling time of 9 s per mode. A new polarization mode was implemented on the Southern Great Plains site radar to facilitate detection of clutter in the radar returns. Future plans include the installation of the polarization mode at the North Slope of Alaska site to help in the identification of nonspherical particles. The precipitation mode for all of the cloud radars has an extended dynamic range to prevent nonlinearities in strong radar returns from precipitation. The cloud radar mode sequence is now uneven with the boundary layer mode, repeating every other time to improve observations of boundary layer clouds. These changes in the operational sampling of the cloud radars led to a change in the strategy for combining the mode data to produce a product that best describes the overall properties of clouds. The new microscale data products are at the highest temporal and spatial resolution possible to facilitate cloud retrievals and process studies.

The recording of Doppler spectra for all of the ARM Program cloud radars and new Doppler spectra postprocessing methods applied to them lead to new information related to cloud microphysics and turbulence and substantially improved overall cloud radar data quality. The availability of Doppler spectra also allows for implementation of a neural network–based technique to identify the presence of clutter (e.g., insects) in the boundary layer mode returns of all cloud radars.

The new microscale data products will contain the highest-resolution data and information related to Doppler spectra morphology. These products will be suitable as input to radar-based retrievals of cloud microphysics and turbulence. As such, they will be of value in assessing small domain models, such as large eddy simulation (LES) models.

Acknowledgments

Support for this research was funded in part by the Office of Biological and Environmental Research, Environmental Sciences Division of the U.S. DOE (under Grants DE-FG02-90ER61071 and DE-FG02-97ER62337) as part of the Atmospheric Radiation Measurement Program.

REFERENCES

  • Ackerman, T. P., and Stokes G. , 2003: The Atmospheric Radiation Measurement Program. Phys. Today, 56 , 3845.

  • Albrecht, B. A., and Kollias P. , 1999: Observations of tropical cloud systems with a mm-wavelength cloud radar—An overview. Preprints, 29th Int. Conf. on Radar Meteorology, Montreal, QC, Canada, Amer. Meteor. Soc., 454–456.

  • Barber, N. F., 1947: Narrow band-pass filter using modulation. Wireless Eng., 24 , 132134.

  • Clothiaux, E. E., Penc R. S. , Thomson D. W. , Ackerman T. P. , and Williams S. R. , 1994: A first-guess feature-based algorithm for estimating wind speed in clear-air Doppler radar spectra. J. Atmos. Oceanic Technol., 11 , 888908.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Clothiaux, E. E., Miller M. A. , Albrecht B. A. , Ackerman T. P. , Verlinde J. , Babb D. M. , Peters R. M. , and Syrett W. J. , 1995: An evaluation of a 94-GHz radar for remote sensing of cloud properties. J. Atmos. Oceanic Technol., 12 , 201229.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Clothiaux, E. E., and Coauthors, 1999: The Atmospheric Radiation Measurement Program cloud radars: Operational modes. J. Atmos. Oceanic Technol., 16 , 819827.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Clothiaux, E. E., Ackerman T. P. , Mace G. G. , Moran K. P. , Marchand R. T. , Miller M. A. , and Martner B. E. , 2000: Objective determination of cloud heights and radar reflectivities using a combination of active remote sensors at the ARM CART sites. J. Appl. Meteor., 39 , 645665.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Doviak, R. J., and Zrnic D. S. , 1993: Doppler Radar and Weather Observations. 2d ed. Academic Press, 592 pp.

  • Firda, J. M., Sekelsky S. M. , and McIntosh R. E. , 1999: Application of dual-frequency millimeter-wave Doppler spectra for the retrieval of drop size distributions and vertical air motion in rain. J. Atmos. Oceanic Technol., 16 , 216236.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gossard, E. E., Snider J. B. , Clothiaux E. E. , Martner B. , Gibson J. S. , Kropfli R. A. , and Frisch A. S. , 1997: The potential of 8-mm radars for remotely sensing cloud drop size distributions. J. Atmos. Oceanic Technol., 14 , 7687.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hildebrand, P. H., and Sekhon R. S. , 1974: Objective determination of the noise level in Doppler spectra. J. Appl. Meteor., 13 , 808811.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Jameson, A. R., 1987: Relations among linear and circular polarization parameters measured in canted hydrometeors. J. Atmos. Oceanic Technol., 4 , 634645.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kato, S., Mace G. G. , Clothiaux E. E. , Liljegren J. C. , and Austin R. T. , 2001: Doppler cloud radar derived drop size distribution in liquid water stratus clouds. J. Atmos. Sci., 58 , 28952911.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kollias, P., Albrecht B. A. , Lhermitte R. , and Savtchenko A. , 2001: Radar observations of updrafts, downdrafts, and turbulence in fair-weather cumuli. J. Atmos. Sci., 58 , 17501766.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kollias, P., Albrecht B. A. , and Marks F. Jr., 2002: Why Mie? Accurate observations of vertical air velocities and raindrops using a cloud radar. Bull. Amer. Meteor. Soc., 83 , 14711483.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kollias, P., Albrecht B. A. , and Marks F. Jr, 2003: Cloud radar observations of vertical drafts and microphysics in convective rain. J. Geophys. Res., 108 .4053, doi:10.1029/2001JD002033.

    • Search Google Scholar
    • Export Citation
  • Kollias, P., Clothiaux E. E. , Albrecht B. A. , Miller M. A. , Moran K. P. , and Johnson K. L. , 2005: The Atmospheric Radiation Measurement Program cloud profiling radars. An evaluation of signal processing and sampling strategies. J. Atmos. Oceanic Technol., 22 , 930948.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kollias, P., Luke E. , Miller M. , and Albrecht B. A. , 2006: Radar Doppler spectra recording at the ARM sites: An insight to cloud microphysics and turbulence. Proc. Fourth European Conf. on Radar Meteorology and Hydrology, Barcelona, Spain, Copernicus Online Service and Information System (COSIS), 233–235.

  • Lhermitte, R., 1988: Observations of rain at vertical incidence with a 94 GHz Doppler radar: An insight of Mie scattering. Geophys. Res. Lett., 15 , 11251128.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Luke, E., Kollias P. , and Johnson K. L. , 2006: A method for the automatic detection of insect clutter in Doppler-radar returns. Proc. Seventh Int. Symp. on Tropospheric Profiling, Boulder, CO, NCAR, 431–432.

  • Matrosov, S. Y., Reinking R. F. , Kropfli R. A. , and Bartram B. W. , 1996: Estimation of ice hydrometeor types and shapes from radar polarization measurements. J. Atmos. Oceanic Technol., 13 , 8596.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Mead, J., and Widener K. , 2005: W-band ARM cloud radar system. Proc. 2005 ARM Science Meeting, Daytona Beach, FL, U.S. Dept. of Energy.

  • Moran, K. P., Martner B. E. , Post M. J. , Kropfli R. A. , Welshand D. C. , and Widener K. B. , 1998: An unattended cloud-profiling radar for use in climate research. Bull. Amer. Meteor. Soc., 79 , 443455.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • O’Connor, E. J., Hogan R. J. , and Illingworth A. J. , 2005: Retrieving stratocumulus drizzle parameters using Doppler radar and lidar. J. Appl. Meteor., 44 , 1427.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sato, T., Doji H. , Iwai H. , Kimura I. , Fukao S. , Yamamoto M. , Tsuda T. , and Kato S. , 1990: Computer processing for deriving drop-size distributions and vertical air velocities from VHF Doppler radar spectra. Radio Sci., 25 , 961973.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Schmidt, G., Ruster R. , and Czechowsky P. , 1979: Complementary code and digital filtering for detection of weak VHF radar signals from the mesosphere. IEEE Trans Geosci. Electron., 17 , 154161.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Shupe, M. D., Kollias P. , Matrosov S. Y. , and Schneider T. L. , 2004: Deriving mixed-phase cloud properties from Doppler radar spectra. J. Atmos. Oceanic Technol., 21 , 660670.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sirmans, D., and Bumgarner B. , 1975: Numerical comparison of five mean frequency estimators. J. Appl. Meteor., 14 , 9911003.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Stokes, G. M., and Schwartz S. E. , 1994: The Atmospheric Radiation Measurement (ARM) Program: Programmatic background and design of the Cloud and Radiation Test Bed. Bull. Amer. Meteor. Soc., 75 , 12011221.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Uttal, T., Church L. I. , Martner B. E. , and Gibson J. S. , 1993: CLDSTATS: A cloud boundary detection algorithm for vertically pointing radar data. NOAA Tech. Memo. ERL WPL-233, 28 pp. [Available from the National Technical Information Service, 5285 Port Royal Rd., Springfield, VA 22161.].

  • Wakasugi, K., and Fukao S. , 1985: Sidelobe properties of a complementary code used in MST radar observations. IEEE Trans. Geosci. Remote Sens., GE-23 , 5759.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Widener, K. B., Moran K. P. , Clark K. A. , Chanders C. , Miller M. A. , Johnson M. A. K. L. , and Koontz A. S. , 2004: MMCR upgrades: Present status and future plans. Proc. 2004 ARM Science Meeting, Albuquerque, NM, U.S. Dept. of Energy.

Fig. 1.
Fig. 1.

Example of 24 h of cloud radar observations at the ARM SGP site. (a) BL mode radar reflectivity, (b) GE mode radar reflectivity, (c) the ARSCL merged radar reflectivity, and (d) mode usage in the ARSCL merged radar reflectivity.

Citation: Journal of Atmospheric and Oceanic Technology 24, 7; 10.1175/JTECH2033.1

Fig. 2.
Fig. 2.

Flowchart of Doppler spectra postprocessing.

Citation: Journal of Atmospheric and Oceanic Technology 24, 7; 10.1175/JTECH2033.1

Fig. 3.
Fig. 3.

(a) Example of cloud radar Doppler spectrogram, i.e., stack of Doppler spectra across many different heights. (left) The original Doppler spectra as recorded by the cloud radar with the color bar showing dB (i.e., analog-to-digital counts), which is proportional to dBm. (right) The Doppler spectra after dealiasing. (b) One hour (approximately 900 BL mode profiles) of Doppler spectra velocity observations at the same height with the color bar showing dB (i.e., analog-to-digital counts) and the white line the location of the Doppler spectra image. In addition to the Doppler spectra image, one or two additional Doppler spectra peaks that correspond to ice particles with different sizes are observed. (c) Example of a trimodal Doppler spectrum. (d) Example of a bimodal Doppler spectrum with a spectral image identified by the blue dot. (e) Example of a typical insect-contaminated Doppler spectrum collected in the boundary layer at the SGP site.

Citation: Journal of Atmospheric and Oceanic Technology 24, 7; 10.1175/JTECH2033.1

Fig. 4.
Fig. 4.

(a) Example of cloud radar BL mode reflectivities at the SGP site that shows the extent of insect contamination. The reflectivities from 0000 to 0900 UTC and from 1400 to 2400 UTC that are below the ceilometer cloud-base heights (black line) are from insects. (b) The insect mask produced by the new Doppler spectra–based neural network algorithm.

Citation: Journal of Atmospheric and Oceanic Technology 24, 7; 10.1175/JTECH2033.1

Fig. 5.
Fig. 5.

Uncertainty estimates for the first three Doppler moments of the BL mode: (a) reflectivity, (b) mean Doppler velocity, and (c) Doppler spectrum width. The contours represent the uncertainty estimates produced from numerical simulations of Doppler signals with the same properties as the BL mode. The units are dBZ for the reflectivity and m s−1 for the mean Doppler velocity and Doppler spectrum width. The black dots correspond to actual data points from boundary layer cumuli clouds observed at the SGP site.

Citation: Journal of Atmospheric and Oceanic Technology 24, 7; 10.1175/JTECH2033.1

Fig. 6.
Fig. 6.

Flowchart of Doppler spectra postprocessing with the production of the new micro-ARSCL data products. The data from each cloud radar mode are processed separately with the new Doppler spectra postprocessing methods.

Citation: Journal of Atmospheric and Oceanic Technology 24, 7; 10.1175/JTECH2033.1

Fig. 7.
Fig. 7.

(a) Example of GE mode reflectivities with corresponding mapping of (b) the number of significant peaks in the Doppler spectra, (c) uncertainty estimates of the radar reflectivity estimates, and (d) skewness of the primary Doppler spectra peak.

Citation: Journal of Atmospheric and Oceanic Technology 24, 7; 10.1175/JTECH2033.1

Table 1.

Previous operational parameters for the four modes of the ARM Program cloud radars. These parameters are the interpulse period (τipp), the number of coded bits (Nbits), the number of coherent averages (Ncoh), the number of FFT points (Nfft), the minimum range (Rmin) of useful data, the unambiguous range (Ru), the number of samples (Ns), the sampling rate (SR), the signal dwell time (SDT), the unambiguous velocity (Vu), the velocity resolution (ΔVu), and the temporal resolution (Ts) of the Doppler spectra and moments that result from this set of parameters.

Table 1.
Table 2.

New operational parameters for the five modes of the ARM Program cloud radars. These parameters are the interpulse period (τipp), the number of coded bits (Nbits), the number of coherent averages (Ncoh), the number of FFT points (Nfft), the minimum range (Rmin) of useful data, the unambiguous range (Ru), the number of samples (Ns), the sampling rate (SR), the signal dwell time (SDT), the unambiguous velocity (Vu), the velocity resolution (ΔVu), and the temporal resolution (Ts) of the Doppler spectra and moments that result from this set of parameters.

Table 2.
Save
  • Ackerman, T. P., and Stokes G. , 2003: The Atmospheric Radiation Measurement Program. Phys. Today, 56 , 3845.

  • Albrecht, B. A., and Kollias P. , 1999: Observations of tropical cloud systems with a mm-wavelength cloud radar—An overview. Preprints, 29th Int. Conf. on Radar Meteorology, Montreal, QC, Canada, Amer. Meteor. Soc., 454–456.

  • Barber, N. F., 1947: Narrow band-pass filter using modulation. Wireless Eng., 24 , 132134.

  • Clothiaux, E. E., Penc R. S. , Thomson D. W. , Ackerman T. P. , and Williams S. R. , 1994: A first-guess feature-based algorithm for estimating wind speed in clear-air Doppler radar spectra. J. Atmos. Oceanic Technol., 11 , 888908.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Clothiaux, E. E., Miller M. A. , Albrecht B. A. , Ackerman T. P. , Verlinde J. , Babb D. M. , Peters R. M. , and Syrett W. J. , 1995: An evaluation of a 94-GHz radar for remote sensing of cloud properties. J. Atmos. Oceanic Technol., 12 , 201229.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Clothiaux, E. E., and Coauthors, 1999: The Atmospheric Radiation Measurement Program cloud radars: Operational modes. J. Atmos. Oceanic Technol., 16 , 819827.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Clothiaux, E. E., Ackerman T. P. , Mace G. G. , Moran K. P. , Marchand R. T. , Miller M. A. , and Martner B. E. , 2000: Objective determination of cloud heights and radar reflectivities using a combination of active remote sensors at the ARM CART sites. J. Appl. Meteor., 39 , 645665.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Doviak, R. J., and Zrnic D. S. , 1993: Doppler Radar and Weather Observations. 2d ed. Academic Press, 592 pp.

  • Firda, J. M., Sekelsky S. M. , and McIntosh R. E. , 1999: Application of dual-frequency millimeter-wave Doppler spectra for the retrieval of drop size distributions and vertical air motion in rain. J. Atmos. Oceanic Technol., 16 , 216236.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gossard, E. E., Snider J. B. , Clothiaux E. E. , Martner B. , Gibson J. S. , Kropfli R. A. , and Frisch A. S. , 1997: The potential of 8-mm radars for remotely sensing cloud drop size distributions. J. Atmos. Oceanic Technol., 14 , 7687.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hildebrand, P. H., and Sekhon R. S. , 1974: Objective determination of the noise level in Doppler spectra. J. Appl. Meteor., 13 , 808811.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Jameson, A. R., 1987: Relations among linear and circular polarization parameters measured in canted hydrometeors. J. Atmos. Oceanic Technol., 4 , 634645.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kato, S., Mace G. G. , Clothiaux E. E. , Liljegren J. C. , and Austin R. T. , 2001: Doppler cloud radar derived drop size distribution in liquid water stratus clouds. J. Atmos. Sci., 58 , 28952911.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kollias, P., Albrecht B. A. , Lhermitte R. , and Savtchenko A. , 2001: Radar observations of updrafts, downdrafts, and turbulence in fair-weather cumuli. J. Atmos. Sci., 58 , 17501766.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kollias, P., Albrecht B. A. , and Marks F. Jr., 2002: Why Mie? Accurate observations of vertical air velocities and raindrops using a cloud radar. Bull. Amer. Meteor. Soc., 83 , 14711483.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kollias, P., Albrecht B. A. , and Marks F. Jr, 2003: Cloud radar observations of vertical drafts and microphysics in convective rain. J. Geophys. Res., 108 .4053, doi:10.1029/2001JD002033.

    • Search Google Scholar
    • Export Citation
  • Kollias, P., Clothiaux E. E. , Albrecht B. A. , Miller M. A. , Moran K. P. , and Johnson K. L. , 2005: The Atmospheric Radiation Measurement Program cloud profiling radars. An evaluation of signal processing and sampling strategies. J. Atmos. Oceanic Technol., 22 , 930948.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kollias, P., Luke E. , Miller M. , and Albrecht B. A. , 2006: Radar Doppler spectra recording at the ARM sites: An insight to cloud microphysics and turbulence. Proc. Fourth European Conf. on Radar Meteorology and Hydrology, Barcelona, Spain, Copernicus Online Service and Information System (COSIS), 233–235.

  • Lhermitte, R., 1988: Observations of rain at vertical incidence with a 94 GHz Doppler radar: An insight of Mie scattering. Geophys. Res. Lett., 15 , 11251128.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Luke, E., Kollias P. , and Johnson K. L. , 2006: A method for the automatic detection of insect clutter in Doppler-radar returns. Proc. Seventh Int. Symp. on Tropospheric Profiling, Boulder, CO, NCAR, 431–432.

  • Matrosov, S. Y., Reinking R. F. , Kropfli R. A. , and Bartram B. W. , 1996: Estimation of ice hydrometeor types and shapes from radar polarization measurements. J. Atmos. Oceanic Technol., 13 , 8596.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Mead, J., and Widener K. , 2005: W-band ARM cloud radar system. Proc. 2005 ARM Science Meeting, Daytona Beach, FL, U.S. Dept. of Energy.

  • Moran, K. P., Martner B. E. , Post M. J. , Kropfli R. A. , Welshand D. C. , and Widener K. B. , 1998: An unattended cloud-profiling radar for use in climate research. Bull. Amer. Meteor. Soc., 79 , 443455.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • O’Connor, E. J., Hogan R. J. , and Illingworth A. J. , 2005: Retrieving stratocumulus drizzle parameters using Doppler radar and lidar. J. Appl. Meteor., 44 , 1427.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sato, T., Doji H. , Iwai H. , Kimura I. , Fukao S. , Yamamoto M. , Tsuda T. , and Kato S. , 1990: Computer processing for deriving drop-size distributions and vertical air velocities from VHF Doppler radar spectra. Radio Sci., 25 , 961973.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Schmidt, G., Ruster R. , and Czechowsky P. , 1979: Complementary code and digital filtering for detection of weak VHF radar signals from the mesosphere. IEEE Trans Geosci. Electron., 17 , 154161.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Shupe, M. D., Kollias P. , Matrosov S. Y. , and Schneider T. L. , 2004: Deriving mixed-phase cloud properties from Doppler radar spectra. J. Atmos. Oceanic Technol., 21 , 660670.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sirmans, D., and Bumgarner B. , 1975: Numerical comparison of five mean frequency estimators. J. Appl. Meteor., 14 , 9911003.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Stokes, G. M., and Schwartz S. E. , 1994: The Atmospheric Radiation Measurement (ARM) Program: Programmatic background and design of the Cloud and Radiation Test Bed. Bull. Amer. Meteor. Soc., 75 , 12011221.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Uttal, T., Church L. I. , Martner B. E. , and Gibson J. S. , 1993: CLDSTATS: A cloud boundary detection algorithm for vertically pointing radar data. NOAA Tech. Memo. ERL WPL-233, 28 pp. [Available from the National Technical Information Service, 5285 Port Royal Rd., Springfield, VA 22161.].

  • Wakasugi, K., and Fukao S. , 1985: Sidelobe properties of a complementary code used in MST radar observations. IEEE Trans. Geosci. Remote Sens., GE-23 , 5759.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Widener, K. B., Moran K. P. , Clark K. A. , Chanders C. , Miller M. A. , Johnson M. A. K. L. , and Koontz A. S. , 2004: MMCR upgrades: Present status and future plans. Proc. 2004 ARM Science Meeting, Albuquerque, NM, U.S. Dept. of Energy.

  • Fig. 1.

    Example of 24 h of cloud radar observations at the ARM SGP site. (a) BL mode radar reflectivity, (b) GE mode radar reflectivity, (c) the ARSCL merged radar reflectivity, and (d) mode usage in the ARSCL merged radar reflectivity.

  • Fig. 2.

    Flowchart of Doppler spectra postprocessing.

  • Fig. 3.

    (a) Example of cloud radar Doppler spectrogram, i.e., stack of Doppler spectra across many different heights. (left) The original Doppler spectra as recorded by the cloud radar with the color bar showing dB (i.e., analog-to-digital counts), which is proportional to dBm. (right) The Doppler spectra after dealiasing. (b) One hour (approximately 900 BL mode profiles) of Doppler spectra velocity observations at the same height with the color bar showing dB (i.e., analog-to-digital counts) and the white line the location of the Doppler spectra image. In addition to the Doppler spectra image, one or two additional Doppler spectra peaks that correspond to ice particles with different sizes are observed. (c) Example of a trimodal Doppler spectrum. (d) Example of a bimodal Doppler spectrum with a spectral image identified by the blue dot. (e) Example of a typical insect-contaminated Doppler spectrum collected in the boundary layer at the SGP site.

  • Fig. 4.

    (a) Example of cloud radar BL mode reflectivities at the SGP site that shows the extent of insect contamination. The reflectivities from 0000 to 0900 UTC and from 1400 to 2400 UTC that are below the ceilometer cloud-base heights (black line) are from insects. (b) The insect mask produced by the new Doppler spectra–based neural network algorithm.

  • Fig. 5.

    Uncertainty estimates for the first three Doppler moments of the BL mode: (a) reflectivity, (b) mean Doppler velocity, and (c) Doppler spectrum width. The contours represent the uncertainty estimates produced from numerical simulations of Doppler signals with the same properties as the BL mode. The units are dBZ for the reflectivity and m s−1 for the mean Doppler velocity and Doppler spectrum width. The black dots correspond to actual data points from boundary layer cumuli clouds observed at the SGP site.

  • Fig. 6.

    Flowchart of Doppler spectra postprocessing with the production of the new micro-ARSCL data products. The data from each cloud radar mode are processed separately with the new Doppler spectra postprocessing methods.

  • Fig. 7.

    (a) Example of GE mode reflectivities with corresponding mapping of (b) the number of significant peaks in the Doppler spectra, (c) uncertainty estimates of the radar reflectivity estimates, and (d) skewness of the primary Doppler spectra peak.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 644 212 12
PDF Downloads 1442 117 9