Radar detection of individual raindrops

1 Radar Detection of Individual Raindrops 1 2 Jerome M. Schmidt1, Piotr. J. Flatau2, Paul R. Harasti1, Robert. D. Yates3, David J. 3 Delene4, Nicholas J. Gapp4, William J. Kohri5, Jerome R. Vetter5, Jason E. Nachamkin1, 4 Mark G. Parent6, Joshua D. Hoover7, Mark J. Anderson7, Seth Green7, and 5 James E. Bennett8 6 7 1Marine Meteorology Division, Naval Research Laboratory, Monterey CA 8 2Scripps Institution of Oceanography, University of California, San Diego, CA 9 3AECOM, Dahlgren, VA 10 4The University of North Dakota, Grand Forks, ND 11 5The Johns Hopkins University Applied Physics Laboratory, Laurel, MD 12 6Radar Division, Naval Research Laboratory, Washington, DC 13 7Naval Surface Warfare Center Dahlgren Division, Dahlgren, VA 14 8The Naval Ordnance Test Unit, Cape Canaveral, FL 15 16 Corresponding Author: Jerome M. Schmidt, Marine Meteorology Division, Naval Research 17 Laboratory, 7 Grace Hopper Ave, Monterey, CA, 93943 18 Email: jerome.schmidt@nrlmry.navy.mil 19 20 Early Online Release: This preliminary version has been accepted for publication in Bulletin of the American Meteorological Society, may be fully cited, and has been assigned DOI The final typeset copyedited article will replace the EOR at the above DOI when it is published. © 2019 American Meteorological Society 10.1175/BAMS-D-18-0130.1.

2 Abstract 21 Descriptions of the experimental design and research highlights obtained from a series of four 22 multi-agency field projects held near Cape Canaveral, FL, are presented. The experiments 23 featured a 3MW, dual polarization, C-band Doppler radar which serves in a dual capacity as both 24 a precipitation and cloud radar. This duality stems from a combination of the radar’s high 25 sensitivity and extremely small resolution volumes produced by the narrow 0.22 degree 26 beamwidth and the 0.543 m along-range resolution. Experimental highlights focus on the radar’s 27 real-time aircraft tracking capability as well as the fine-scale reflectivity and eddy structure of a 28 thin non-precipitating stratus layer. Examples of precipitating storm systems focus on the 29 analysis of the distinctive and nearly linear radar reflectivity signatures (referred to as “streaks”) 30 that are caused as individual hydrometeors traverse the narrow radar beam. Each streak leaves a 31 unique radar reflectivity signature that is analyzed with regard to estimating the underlying 32 particle properties such as size, fall speed, and oscillation characteristics. The observed along-33 streak reflectivity oscillations are complex and discussed in terms of diameter-dependent drop 34 dynamics (oscillation frequency and viscous damping time scales) as well as radar-dependent 35 factors governing the near-field Fresnel radiation pattern and inferred drop-drop interference. 36 Capsule Summary: A high resolution, hybrid cloud and precipitation radar reveals the 37 fascinating structure and behavior of individual hydrometeors within observed cloud systems. 38 39 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

3 Introduction 40 41 At first glance, the discrete detection of individual raindrops using either surface or airborne 42 radar platforms would seem to be an impossible und ertaking. Calculations using standard 43 Marshall and Palmer (1948 ) type size distributions, after all, would suggest precipitation sized 44 particle concentrations on the order of 1000 m -3 even in light rainfall conditions of only 2 .0 mm 45 hr-1. Such concentrations would easily translate into over a billion or more precipitation sized 46 particles within modern day WSR-88D operational radar resolution volumes (on the order of 10 6 47 m-3) that stem from this radar’s 250 m range resolution and 0.925 degree beamwidth (see Table 48 1). Inferences on the internal cloud structure derived from remote s ensing using even higher -49 resolution research radars have thus long relied, with great success, on the statistical properties of 50 the returned microwave signal to extract critical particle characteristics such as the particle habit, 51 phase, and canting angle (Seliga and Bringi 1976; Illingworth et al. 1987; Zrnić et al. 1993; Orr 52 and Kropfli 1999; Bringi and Chandrasekar 2001; Matrosov et al. 2001; Houze 2004; Houser and 53 Bluestein 2011; Matrosov et al. 2012; Chandrasekar et al. 2013, Hubbert et al. 2018). 54 55 The more elusive undertaking of single particle detection to date has lar gely been relegated to 56 highly idealized environmental settings. Early r adar studies of indiv idual hydrometeors were 57 attempted by Atlas et al. (1960) and Willis et al. (1964) who studied the returns from large hail 58 stones tethered to or released from weather balloo ns. More cont rolled studies have been 59 conducted in laboratory wind tunnels where individual raindrops could be suspended in the air 60 stream for prolonged periods of time. Drop properties, such as the natural drop oscillation 61 frequencies, could then be recorded using photographic techniques and microwave 62 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

4 measurements (Brook and Latham 1968 ). Additional studies over the past several decades have 63 provided detailed insight on the factors impacting the drop shape, canting angle, oscillation 64 frequencies, viscous damping time scales, and fall s peed characteristics (Rayleigh 1879; Gunn 65 1949; Blanchard 1950; List and Hand 1971 ; Musgrove and Brook 1975; Pruppacher and Klett 66 1978 [pgs 315-322]; Beard et al. 1983; Beard 1984; Beard and Kubesh 1991; Testik et al. 2006; 67 Szakáll et al. 2010). 68 69 A series of specialized experiments utilizing an advanced ground-based radar operated by the 70 U.S. Navy have demonstrated that it is now possible to extend the studies of individual raindrops 71 beyond the wind tunnel to observed precipitating cloud systems (Schmidt et al. 2012) . The 3 72 MW, C-Band, dual-polarization Mid-Course Doppler Radar used in that study (hereafter referred 73 to as the MCR; Fig. 1) has a maximum radar resolution volume on the order of 14 m3 at a range 74 of 2 km from the radar (a comparison of the MCR attributes to other well -known radars is 75 provided in Table 1). The MCR makes the observations of individual raindrops more tenable 76 provided they exceed the radar’s minimal detectable diameter at a given range [Dmin > 0.4 mm @ 77 2 km when the drops are embedded in a background reflectivity field of -40 dBZ; see Schmidt et 78 al. (2012) for details]. 79 80 The CAPE Field Experiments 81 82 The MCR-derived raindrop signatures were first observed in 2009 during the second of four field 83 experiments held bet ween 2008 and 2015 (Table 2). The original purpose of these experiments 84 was to assess the feasibility of using the MCR for basic cloud research and model validation 85 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

5 applications. Historically, the MCR had been utilized up to 2011 to study the debris field 86 associated with the NASA Space Shuttle launches in the return -to-flight missions after the 1986 87 Challenger disaster. A subsequent effort to incorporate the MCR for cloud studies initially gre w 88 from a small group of scientists and rad ar engineers in 2008 to a much larger multi -agency 89 research effort by 2015. Each follow-on experiment in the series included additional instruments 90 to help guide the placement of the research aircraft and the MCR’s specialized higher -resolution 91 waveforms (Fig. 2 and Table 2). 92 93 The MCR is a dual polarization linear frequency modulated (LFM) pulse compression radar that 94 transmits at a pulse repetition frequency (PRF) of either 160.1 or 320.2 Hz. When used at the 95 lower PRF, the MCR typically transmits two wave forms (every other pulse) each of which have 96 a right-hand circular polarization and a one-way 3 dB far-field beamwidth of 0.22°. A quadratic 97 phase modulation is applied to the transmitted pulse which, after the 2010 radar upgrade, covers 98 a swept frequency bandwidth (B) of either 8.0 or 500 MHz (subsequently referred to as the 99 narrowband and wideband waveforms, res pectively). A matched filter data processing technique 100 is applied to the received signal and this leads to 6 dB range resolution values (R6 = Gc/2B) of 101 34.02 and 0.543 m for the narrowband and wideband waveforms, respectively [where c is the 102 speed of light (m s -1) and G is the normalized 6dB width that has a value of either 1.812 103 (wideband) or 1.816 (narrowband)]. The MCR’s high 3MW transmit power and large 15.24 m 104 diameter dish help overcome the loss in sensitivity that can otherwise arise in radars which 105 transmit at longer wavelengths (Kollias et al. 2016). When combined with the other radar 106 attributes noted above, the MCR maintains a relatively high sensitivity in comparison to other 107 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

6 well established cloud and precipitation radars ( Table 1). It is th is combination of factors which 108 thus help make the detection of individual hydrometeors possible within the free atmosphere. 109 110 The wideband and narrowband waveforms are each comprised of two range windows that are 111 300 m and 75 km in length, respectively. These range windows can be placed contiguously or in 112 separate range locations of the MCR’s boresight direction at any given time. When located closer 113 to the radar, the range windows can fall within the radar’s near-field or Fresnel region where 114 both amplitude and phase ripples may be present in the transmitted waveforms. These ripples can 115 be of significance in the analysis of individual hydrometeors as they can introduce oscillations in 116 the extracted particle time series that can impact the interpretation of the derived spectral 117 analysis. Based on the MCR’s 15.24 m dish diameter and transmitted 0.053 m wideband 118 wavelength, the Fresnel region is calculated to lie between the ranges of 0.160 and 8.76 km and 119 thus encompasses the location of the wideband observations shown later in the text. 120 121 The MCR’s radiation pattern in the Fresnel region was obtained numerically using the GRASP 122 Software (TICRA, Copenhagen, Denmark, www.ticra.com), which is the industry standard for 123 analyses of electromagnetic radiation emitted from large reflector antennas [see Appendix A for 124 details]. The numerical results are shown at a range of 1.9 km in Fig. 3 as this location 125 corresponds closely with the level of the primary wideband observations shown later in the text. 126 The simulated cross-beam amplitude and phase patterns are symmetric about the center of the 127 domain and appear to lack significant amplitude or phase ripples near the radar boresight (Fig. 128 3). There is a subtle weakness in the amplitude and phase gradient fields evident at ranges of 10 129 to 15 m from boresight and this is a result of a ripple on the transmitted pulse that appears nearly 130 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

7 20 dB down from the peak (Fig. 4). The cross-beam structure evident in Fig. 4 indicates that the 131 Fresnel region beam structure is slightly wider than otherwise assumed for the far-field 132 beamwidth change introduces a bias in our calculated returns which we estimate to be on the 133 order of 0.7 dBsm (see Appendix A for details). The possible impact of the main amplitude and 134 phase ripple will be addressed later in the text when the structure and significance of the 135 individual particle spectra are discussed. 136 137 The MCR incorporates a number of standard scanning strategies including vertical stares, Range 138 Height Indicator ( RHI), Plan Position Indicator ( PPI), and volume scans. To date, these scans 139 have been applied to document the structure of cirrus clouds, deep convective systems and their 140 associated stratiform cloud shields, mammatus clouds (Fig. 2), altocumulus cloud layers, low -141 level stratus, and multiple cases where individual raindrops were detected by the ra dar. The 142 MCR’s relatively small Nyquist interval (±λPRF/4 = ±2.1 m s-1) can hamper the analysis but this 143 has not been a significant issue to date in our studies of shallow statiform or cirrus cloud layers. 144 Velocity aliasing issues can arise , however, in the studies of deeper convective clouds or below 145 the melting layer of precipitating stratiform cloud systems. In such cases, we use the boot -146 strapping techniques of Bargen and Brown (1980) as well as independent Doppler information 147 provided by the METEK Mi cro Rain Radar to help unfold the Doppler velocities into the 148 appropriate Nyquist interval. 149 150 Finally, the MCR also has a special scan that can be used to track research satellites [such as the 151 A-train; Stephens et al. (2002)] or research aircraft in real-time (Fig. 5a,b). These scans will be 152 used to improve the MCR derived Z-LWC relationships and understanding of the internal cloud 153 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

8 properties using the co-located radar and in situ cloud observation (Fig. 5c,d). The presence of 154 liquid water and chain-like crystals composed of large frozen drops in sub -40° C conditions 155 within Florida anvil systems are of particular research interest (Fig. 5c,d). Such crystal types are 156 not frequently observed but are thought to be influenced by the cloud electric field (Gayet et al. 157 2012). Plans for our pending 2019 field study include additional instruments, such as the aircraft-158 based electric field measurements, to gain additional insight into the environmental factors 159 governing the development of these chain-like crystals. 160 161 Detection of Individual Hydrometeors 162 163 The MCR radar signatures attributed to the presence of individual hydrometeors were first 164 recorded during scans of a deep convective storm system observed on the final day of the 2009 165 field study (Fig. 6). Two regions where the wideband recorded unusual radar reflectivity patterns 166 within this storm system are given by the location of the two boxes labeled B1 and B2 in Fig. 6. 167 The reflectivity data shown in B1 were associated with a transient cell that passed over the radar 168 between 2149 and 2154 UTC (Fig. 7). Note the rapid transition in the reflectivity pattern that 169 occurred near 21:49:30 UTC as a series of short-lived and discrete reflectivity anomalies that 170 suddenly appear against the otherwise more uniform background reflectivity field. The 171 anomalies were randomly orientated in space and time and indicative of multiple inbound and 172 outbound “targets”. A second set of anomalies derived from B2 were associated with a vertically 173 oriented precipitation streamer that originated much higher in the parent cloud system (labeled 174 R1 in Fig. 6). In this case, the B2 anomalies all had linear structures indicative of inbound 175 targets exclusively (Fig. 8). These anomalies are evident within R1 and can be seen as well both 176 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

9 entering and exiting this feature along its lateral flanks suggesting the two features are possibly 177 linked. These transient reflectivity anomalies would later be recognized as the first MCR 178 measurements of individual cloud hydrometeors. Given their linear nature and relatively “bright” 179 reflectivity returns, these features were subsequently dubbed “radar streaks” by Schmidt et al. 180 (2012). 181 182 The unexpected appearance of the radar streaks in the final few moments of the 2009 field 183 campaign led to a concerted effort to replicate the observations in subsequent field programs. 184 Improvements were made to the MCR waveforms after 2009 in order to reduce the magnitude of 185 the range sidelobes, from 15 to 35 dB down from the peak. This was accomplished through the 186 application of more accurate phase compensation tables and post-processing waveform 187 equalizers. This was undertaken to reduce the anomalous reflectivity “shadows” found to arise 188 on both sides and parallel to particularly strong streaks evident in the 2009 data set (Fig. 7). The 189 wideband range windows were also subsequently extended from 150 m to 300 m after 2010. 190 Taken together, these changes now provide a much better opportunity to observe clean streak 191 patterns associated with the developing rain shafts and vertical velocity patterns over a much 192 deeper layer of the parent storm system (Fig. 9). 193 194 By further limiting the temporal and spatial scales of the wideband plots, one is able to discern 195 the radar signatures associated with each individual streak (Fig. 10). The bulk streak properties 196 related to their magnitude, slope, and duration can be used to estimate the particle size and the 197 translational and radial velocity of the particles as they enter and then exit the beam. These 198 properties can be impacted by a variety of factors that depend on the particle’s precise trajectory 199 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

10 through the beam, the storm’s vertical velocity pattern, the mean and turbulent structure of the 200 environmental flow and possibly drop-induced lateral drift which Testik et al. (2006) have shown 201 can be as high 1-2 m s-1. Based on the streak slopes and the radar cross section (RCS) values 202 evident in Fig. 10, one infers a net radial motion toward the radar of 4.9 to 5.7 m s-1 and particle 203 size estimates in the range of 2.9 to 3.5 mm (see Appendix A for details). As particles at this 204 altitude and size range should have terminal velocities on the order of 8.8 to 9.4 m s-1 (Beard 205 1976), the slower slope-based velocity estimates suggest the particles are embedded in updraft on 206 the order of 4 m s-1 (confirmed through a separate Doppler analysis not shown). 207 208 The data shown in Fig. 10 also reveal interesting variations in the along-streak temporal trends 209 for each particle. The pulse-to-pulse along-streak RCS time series extracted for S2, S3, and S5 210 highlight some of the apparent variability between particles (Fig. 11). The mean parabolic 211 temporal trends evident in each time series results from the movement of the particles across the 212 vertically orientated beam. The magnitude of the perturbations about these mean trends was 213 greatest in the highly periodic S2 time series (Fig. 11a). The corresponding spectral analysis of 214 the S2 perturbations indicates these oscillations were centered about a broad peak in frequency 215 near 26 Hz and a secondary maximum near 53.7 Hz (Fig. 12a). Similar higher amplitude and 216 periodic fluctuations were also evident in S3 but these oscillations rapidly dampened in time as 217 the particle began to recede from its closest approach to the beam boresight (Fig 11b). The e-218 folding reduction in the perturbation amplitude for S3 was well-observed and occurred over a 219 time-scale of approximately 0.3 seconds. The spectral plot for S3 indicates a strong contribution 220 near 38 Hz arising during the initial portion of the transect and weaker contributions below 25 221 Hz and near 60 Hz (Fig. 12b). The S5 perturbations were of intermediate magnitude throughout 222 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

11 (Fig. 11c) with several spectral peaks arising at different points along the transect (Fig. 12c). The 223 weaker fluctuations of S5 were similar to those found for the S1 and S6 streaks (not shown) 224 while those of S4 compared more favorably with that shown for S3. 225 226 Streak Analysis and Discussion 227 228 The ability to observe individual hydrometeors with radar naturally leads one to question 229 whether the along-streak data can be further analyzed in order to bring out additional particle 230 characteristics. The calculation of the natural vibrational and viscous damping characteristics as 231 a function of drop size might be two of many such parameters as each process occur over time-232 scales far less than that needed for a given streak to cross the beam (Brook and Latham 1968; 233 Beard et al. 1983). Isolating cause and effect can be difficult to undertake in practice, however, 234 due to nuances in the radar itself such as Fresnel region effects, possible sidelobe contamination, 235 the uncertainty of the drop’s position within the beam, or multi-particle interactions such as 236 described by Doviak et al. (1979). Additional physical processes are also likely to arise in 237 observed cloud systems such as drop-drop collisions, wake vortex shedding, multi-modal drop 238 vibrations, or turbulence induced fluctuations among others (Beard and Kubesh, 1991; Szakáll et 239 al. 2009). 240 241 An important initial concept in the analysis of individual hydrometers is that of determining the 242 radar’s minimum detectable particle size. For the MCR, this relationship is governed by an 243 expression of the form shown in Eq. A2 which is governed by the range to the particle, the 244 particle size and phase, and the magnitude of the particle and background signal to noise ratios. 245 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

12 The relationship is illustrated by the curves shown in Fig. 13 which indicate that the minimum 246 detectable particle size increases as the background reflectivity and the range to the target 247 increase. Particles easily identified in Fig. 8 along R1’s lateral flanks, where the background 248 reflectivity is weak (~ -20 dBZ), thus become much more difficult to discern within R1 where 249 the background reflectivity values are much higher (~10 dBZ). This is akin to the faintest stars 250 being the first objects to fade from view as daybreak ensues along the morning horizon. For the 251 conditions shown in this paper, particle sizes of 1 to 3.5mm would have been required for the 252 streaks to stand out against the observed background reflectivity values (Fig. 13). Rising 253 background reflectivity values also mean that particles must pass closer to boresight before they 254 exceed the minimum detection threshold. For conditions shown in this study, particles would 255 need to pass within 10 m of boresight before their SIR values would exceed the detection 256 threshold. This suggests that the ripple in the near field radiation pattern evident in Fig. 3 and 257 Fig. 4, would exert a stronger control on the observed particle oscillation characteristics in cases 258 where a low background reflectivity (< -20 dBZ) leads to earlier particle detection. 259 260 Other possible radar controls on the oscillation characteristics could arise from range sidelobe 261 contamination and constructive and destructive interference patterns caused by the superposition 262 of signals received from other nearby particles. The motivation for the inclusion of particle-to-263 particle interference stems from the nature of the streaks shown in Figs. 10-12. We first note 264 from the spectra shown in Fig. 12 that the main spectral peaks do not align particularly well with 265 those anticipated for first (or second) fundamental mode of naturally occurring drop oscillations 266 at the inferred diameter. Rather, we find better consistency between the inferred 3.3 mm particle 267 sizes with the secondary maxima evident in all the spectra in the 50 to 60 Hz range (Fig. 12). 268 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

13 Errors in the analysis of the drop size caused by their unknown trajectory through the beam may 269 be one culprit. In the case of S2, the spectral peak near 26 Hz would correspond to a drop 270 diameter of 5.5 mm (Fig. 12a). This would have required errors in the corresponding RCS of 271 approximately 14 dBsm at the observed range (see Eq. A2). 272 273 An alternative explanation for the main spectral peaks evident in Fig. 12 may lie in the nature of 274 the signal associated with the constructive or destructive interference patterns associated with 275 two or more closely located particles (ΔR ~ R6 or less). An expansion of Eq. 2.8 of Doviak et al. 276 (1979) for the simplest possible case of two interacting particles leads to a time fluctuating 277 component of the instantaneous echo power P(τs) given by a term of the form 278 1 2 𝐶(𝜏 )𝑐𝑜𝑠 [ 4𝜋(𝑣2−𝑣1) 𝜆 𝑚 𝑃𝑅𝐹]. In this expression C represents the composite amplitude, m/PRF 279 represents time, and fb=2(v2-v1)/λ is the frequency of the fluctuation as a function of the 280 differential radial velocities (v2-v1) between the two particles. This is a statement indicating that 281 the differential range phase between the two particles cycles more quickly as the velocity 282 between the two particles increases. 283 284 A look back at Fig. 10 indicates that at least one of the observed streaks (S2) is really comprised 285 of two nearby particles (labeled S2a and S2b). We can further determine from the plot that these 286 two particles are falling with a differential radial velocity of approximately ~0.7 m s-1. The use of 287 the above expression for fb leads to a predicted resonance frequency of 26.4 Hz, a value that is 288 found to be in good agreement with the location of the main spectral peak shown in Fig. 12a. 289 Numerical simulations of the S2 type drop-drop interference patterns provide further support that 290 this type of interaction can have a strong influence on the temporal evolution of closely located 291 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

14 streaks (Fig. 14; see Appendix A for the simulation details). Note the similarity of the simulated 292 periodicity of the along-streak structure with that of the observed S2 couplet shown in Fig. 10. 293 Offline calculations indicate this oscillation frequency is exactly 26.4 Hz in agreement with the 294 predicted result. This agreement arises even though the drops themselves are held at a fixed 295 diameter and are not allowed to oscillate or otherwise interact dynamically in any manner 296 throughout the course of the simulation. 297 298 The above analysis is meant to show some of the fascinating combinations of cloud particle 299 physics and radar attributes that are possibly contributing to the variability in the observed along-300 streak structure presented in this study. Resonance may be invoked in some instances to help 301 explain why some particles exhibit periodic fluctuations in their time series while other nearby, 302 but somewhat more isolated particles (such as S1 and S6) do not. Other particles, such as S5, 303 will likely have a more complex interpretation as they also appear more isolated yet contain 304 well-defined spectral peaks at specific frequencies not readily attributed to resonance effects 305 alone. In such cases, the inclusion of additional processes such as multi-modal oscillations, drop 306 collision rates, sidelobe impacts, or other factors, may have a greater bearing on the final 307 outcome. The structure of streaks, such as S3, represents an interesting hybrid between S2 and 308 S5 in that the oscillations rapidly dampen in time. This damping is perhaps in response to either 309 natural viscous dissipation processes which are known to have similar time-scales for particles of 310 this size (see Eq. 2 of Beard et al. 1983) or as a result of impacts from other nearby particles. In 311 short, the observed complexity can be challenging to fully understand but it appears promising 312 that enough spectral separation may exist in some instances to further help isolate cause and 313 effect with additional study. 314 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

15 315 Additional Case Studies 316 317 The MCR is capable of deriving the structure of layered cloud systems at a level of detail 318 comparable to high resolution Large Eddy Simulations. One previously published example using 319 the MCR includes a study of the gravity wave impacts on the melting level structure of a 320 convectively generated trailing stratiform cloud shield (Schmidt et al. 2017). They noted that 321 gravity wave updrafts observed near the melting level provided a favorable environment 322 conducive to the production of small water drops and pristine needles in a temperature regime 323 (-3 to -8 °C) known to be favorable for secondary ice particle production (Hallet and Mossop 324 1974). This is similar to the results described in Yuter and Houze (2003) and Houser and 325 Bluestein (2011). The gravity wave downdrafts were found to be associated with larger particle 326 production and aggregates. These contributed to the periodic negative velocity anomalies 327 observed to extend downward from the melting level in a series of wave-induced rain shafts. 328 329 The fine resolution and high sensitivity of the MCR also allows this system to operate as a 330 capable cloud radar. Cloud radars have traditionally been associated with shorter wavelength 331 radars, such as the Ka-band ARM Zenith Radar (KAZR), which have a proven track record for 332 documenting the internal cloud velocity and microphysical signatures [Chandra et al., 2015; 333 Kollias et al. 2016). Recent work has also shown promise for using such radars to estimate the 334 precipitation rates through knowledge of the radar attenuation characteristics (Chandra et al., 335 2015). A comparison of the MCR characteristics with that of KAZR is shown in Table 1. The 336 comparison reveals that the MCR wideband waveform has a higher sensitivity and a comparable 337 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

16 beamwidth to KAZR along with a significant increase in peak power and range resolution. This 338 sensitivity was demonstrated in a study of a mixed-phase altocumulus cloud layer examined by 339 Schmidt et al. (2014). The MCR was able to resolve fine-scale circulations in the altocumulus 340 associated with cloud top entrainment, deeper penetrating updrafts and downdrafts, and subcloud 341 eddy structures associated with descending virga shafts. This sensitivity is also evident in Fig. 13 342 as well as through the detection of clear air signals such as apparent in Fig. 2 near an elevation of 343 4.5 km. These signals near such a high altitude of 4 km are likely not due to insects but rather 344 Bragg scatter from refractive index fluctuations caused by turbulent mixing in a layer at that 345 level of the atmosphere. 346 347 A more recent use of the MCR as a cloud radar is shown in our final example of the higher 348 resolution wideband waveform data obtained for the shallow stratus cloud depicted in Fig. 15a. 349 Broad weak echo regions are evident within the primary updrafts (Fig. 15b,c). There is evidence 350 of very thin layers of enhanced reflectivity (< 10 m thick) that ring the tops and peripheral 351 regions of these rising plumes (labeled ER). These structures are suggestive of preferential 352 increases in either the drop concentration or particle sizes in these regions. Very narrow and 353 vertically coherent channels of weaker reflectivity also extend downward from cloud top and 354 appear to be associated with the cloud-top entrainment process (labeled EC). Very narrow 355 filaments of higher reflectivity tend to form along the periphery and base of these features and 356 near the base of the virga shafts extending below cloud base. A similar feature is found between 357 the U2 and U3 updrafts. Similar features were noted to arise in the study of thin altocumulus 358 cloud layers presented by Heymsfield et al. (1991) and Schmidt et al. (2014). 359 360 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

17 Summary and Future Work 361 362 Highlights from a series of experiments using the MCR have demonstrated this radar serves as a 363 highly capable cloud and precipitation radar. The radar has a high peak power, narrow 364 beamwidth, and extremely small radar resolution volumes that help provide a detailed look at the 365 interior cloud structure at a resolution that may be of interest in many basic research and model 366 validation applications. It has also been shown that it is now possible to observe individual cloud 367 hydrometeors in real cloud systems using a 500 MHz LFM bandwidth waveform. The individual 368 particles generated unique, and at times complex, radar signatures referred to as streaks which 369 provide clues into the underlying cloud hydrometeor structure and behavior. These structures are 370 being analyzed in terms of radar and physically based processes such as multi-drop radar 371 resonance, sidelobe contamination, the onset and viscous decay of diameter-dependent single or 372 multi-modal drop oscillations, drop-drop collisions, and other factors. Future work is required to 373 determine the prevalence of drop couplets such as S2 and to ascertain whether their origins can 374 be tied to specific processes such as the break-up of larger drops. Future field studies will be 375 conducted to obtain additional high-resolution observations of clouds and hydrometeor-induced 376 streaks. It is hoped such observations may help motivate new radar designs and/or research 377 applications that can be used to study other phenomena such as the formation and evolution of 378 damaging hail stones. Possible future collaboration efforts involving dual-polarization cloud 379 radars could be run in a coordinated fashion to further document the radar inferred cloud 380 microphysical structure measurements from multiple points of view. The use of co-located Ka-381 band radars, may be of particular interest to attenuation-based precipitation estimates or other 382 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

18 studies if the MCR can help determine the properties of the individual particles in nearly the 383 same resolution volumes. 384 385 We anticipate that the interaction of two or more drops oscillating at their natural diameter-386 dependent frequencies will provide an even wider variety of responses than those modeled or 387 observed here. Idealized radar-based wind tunnel studies could also be conducted with two or 388 more drops to see if the drop-drop interference signatures reported here could be replicated in a 389 more controlled environment. More realistic numerical solutions could also certainly be designed 390 that include a wider range of differential velocity groupings and additional physical processes. 391 The MCR’s dual-polarization capability which can be used to study drop canting angles, drop 392 aspect ratios, and tumbling ice particles has also yet to be fully exploited. 393 394 Acknowledgements 395 396 This research was supported under grants N0017818WX00186 from the Naval Surface Warfare 397 Center Dahlgren Division. The Authors benefitted greatly from the suggestions and insight of 398 two anonymous Reviewers and the review of Dr. Paul Smith, South Dakota School of Mines. 399 The Authors also benefitted with discussion with Dr. Viswanathan Bringi, Colorado State 400 University; Dr. Jeffery Keeler, NCAR; and Dr. Kevin Knupp, University of Alabama, 401 Huntsville. We also would like to acknowledge the support received from the Cape Canaveral 402 Air Force Weather Station Facility site manager (Dave Chapman), the NASA Space Shuttle 403 Program Manager (Dr. William Parsons), the NASA Debris Radar Project Manager (Anthony 404 Griffith), and personnel at the Naval Ordnance Test Unit (NOTU) for their logistical support, and 405 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

19 air traffic support provided by Mr. Hank Tracy and local air traffic controllers in the Titusville, 406 FL region working within the Federal Aviation Administration. Jamie VanderHeiden of 407 NSWCDD is acknowledged for her help in the experimental planning. Dr. Stephen Schindler of 408 NASA helped secure radiation permission for the ground-based instrumentation. Satellite 409 support for the field programs were provided by Mr. Kim Richardson of the Marine Meteorology 410 Division of the Naval Research Laboratory. Ed Garcia, Nen Nguyen, Daniel Margulieux, Jim 411 Niger, Peter Jones, Joe Ellis, and Fernard Purnell of L3 Interstate Electronics Corporation, 412 Anaheim, CA, provided critical support with the balloon releases, radar data processing, and 413 aircraft tracking algorithms. The Authors are also indebted to the dedicated MCR radar operators 414 Ricky Littleton, David Botterbausch, and James Erbaugh. A special thanks to Dr. David Lando 415 (retired NSWCDD) for his vision, persistence, and patience. And finally, this work would not be 416 possible without the pioneering studies and MCR technical development and direction provided 417 early on by Dr. Julius Goldhirsh (The Johns Hopkins University Applied Physics Laboratory) 418 and Dr. Ed Fletcher (Radar Technology Specialist, Anaheim, CA) who are now both deceased. 419 420 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

20 Appendix A 421 422 Streak Analysis: 423 Factors governing the raindrop size estimates for the MCR were described by Schmidt et al. 424 (2012) and are summarized in Fig. 6 of that study. Individual particle size estimates are derived 425 from the mean peak RCS values (taken near the apex of parabolic plots shown in Fig. 11), the 426 particle slope, or through the use of well-known diameter-dependent factors such as the terminal 427 velocity and particle oscillation frequency (Rayleigh 1879; Gunn 1949; Beard 1976; Pruppacher 428 and Klett, 1978; pgs 315-322; Szakáll et al. 2010). Reflectivity based estimates of the particle 429 diameter (D) take the form D = (R2V1Z)1/6 where R [km] is the range to the target, V1 [m3 km-2] 430 is the range normalized radar pulse volume (V1 = 3.4328), and Z [mm6 m-3] is the along-streak 431 radar reflectivity factor (Z = 10dBZ/10). It is assumed (where applicable) that the background 432 reflectivity does not significantly alter the derived diameter estimates and that the application of 433 the Rayleigh approximation provides reasonable size estimates. The pre-multiplication of Z by 434 the resolution volume (R2V1) in the equation for the particle diameter [D = (R2V1Z)1/6 ] is 435 equivalent to converting reflectivity (which is related to a R2 dependence in the distributed target 436 form of the radar equation) to the sixth power of the diameter of a single-particle 437 (whose received power dependence goes as R4 in the single-particle version of the radar 438 equation). This follows from the Rayleigh approximation used in the weather radar equation (see 439 Doviak et al (1979), Eqs. 2.20 and 2.21). The streak observations shown in this study happen to 440 occur in the near field of the radar (as was first suggested to us by Dr. Paul Smith, South Dakota 441 School of Mines; personal communication). This can lead to a negative reflectivity bias of ~0.71 442 dB for the MCR at a range of 2 km and an under-estimate in the calculated diameter of 443 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

21 approximately 0.1 mm for the observed peak particle RCS values evident in Fig.11. Field 444 measurements will be made in 2019 in an effort to validate the numerical simulations shown in 445 Fig. 3 and Fig. 4 of this study to further quantify the near-field structure of the MCR. We also 446 caution that the various estimates used in this study may not always agree as errors can be 447 expected to arise from a variety of other factors such as the presence of updrafts or downdrafts, 448 the off boresight placement of the streak, the contribution of two or more large particles to the 449 observed reflectivity. Use was made of the regime 2 and regime 3 relationships of Beard (1976) 450 as well as the streak slopes and Doppler analysis to estimate the particle sizes. The lateral motion 451 can be deduced knowing the temporal length of the streak, the range to the target, and knowledge 452 of where the particle crosses of the beam provided a monopulse analysis is conducted. 453 454 Idealized Drop-Drop Simulations: 455 The drop simulations were written in MATLAB for the complex return (V) from two drops at a 456 time (t) and matched filtered range (R) as: 457 458 𝑉(𝑡, 𝑅) = [𝐴(1, 𝑡, 𝑅) + 𝐴(2, 𝑡, 𝑅]𝑒−𝑖𝜃(𝑘0,𝑡) Eq. 1A 459 460 where 461 462 𝐴(𝑘, 𝑡, 𝑅) = 𝐴(𝑘, 𝑡)𝑃[𝑅 − 𝑅(𝑘, 𝑡)]𝑒−𝑖𝜃(𝑘,𝑡) is the complex return for drop k, 𝜃(𝑘, 𝑡) = −4𝜋𝑅(𝑘,𝑡) 𝜆 463 is the range phase for drop k, 𝜃(𝑘0, 𝑡) = −4𝜋𝑅(𝑘0,𝑡) 𝜆 is the motion compensation phase for 464 dealiasing the frequencies, 𝑅(𝑘, 𝑡) = 𝑅0(𝑘) − 𝑣𝑘(𝑘)(𝑡 − 𝑡0) is the range from the reference 465 value (R0) and initial time (t0) due to the terminal velocity of the particle (vk), λ is the radar 466 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

22 wavelength (0.053 m), and P[R] is the matched filter output pulse shape using an assumed 467 normal pulse shape fitted to the 6 dB range width of the actual MCR pulse shape (𝑝𝑘 = 𝑒−0.5𝑦2 468 where 𝑦2 = 𝑅𝐺−𝑅(𝑘) 𝜎𝑅 with RG representing the range axis with a gate spacing of 9.1 cm and σR = 469 0.2293 m representing the Gaussian pulse shape approximation to the response antenna 470 weighting as the particle traverses the beam). This approximation only qualitatively accounts for 471 the beam shape in the near field which is shown in Fig. 4 to have imposed ripples in the Fresnel 472 region which we do not attempt to simulate. The plotted time-height data in Fig. 14 are given by 473 10𝑙𝑜𝑔10(|𝑉(𝑡, 𝑅)|2). The motion compensation phase θ(k0,t) is selected from one of the two 474 specified terminal velocities for each particle. We do not attempt to model the hydrodynamic 475 interactions between the drops, the internal drop oscillations, or the viscous damping. 476 477 Near- and Far-Field Region Modeling: 478 479 The MCR’s radiation pattern in the Fresnel region was obtained numerically using the method of 480 spherical expansion provided by GRASP Software (TICRA, Copenhagen, Denmark, 481 www.ticra.com). GRASP is the industry standard for analyses of electromagnetic radiation 482 emitted from large reflector antennas (Jorgensen et al 2018; Cappellini and Salghetti-Drioli 2016). 483 A representative down-range of 1940 m discerned from Fig. 10 was selected for the Fresnel 484 region simulations using a 0.01 m grid spacing over a 15 m by 15 m mesh. An arbitrary down-485 range of 10,000 m was selected for the Far-field region simulations using a 0.5 m grid spacing 486 over a 1250 m by 1250 m mesh. These numerical simulations account for the size of the MCR’s 487 dish, transmitted power, wavelength, and right-hand-circular emitted polarization, as well as the 488 size, location and alignment of the feed horn, sub-reflector and struts relative to the dish, 489 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

23 obtained from MCR design documents. The Fresnel (far-field) simulation results were projected 490 onto a cross-range plane located at a down-range distance of 1940 m (10,000 m), perpendicular 491 to the boresight direction. 492 493 The inverses of Eqs. (10) and (16) from Sekelsky (2002) were used to derive estimates of the 494 boresight gain and reflectivity factor corrections in the Fresnel region, using the power density 495 function derived from the near field simulation at 1940 m down-range and the far-field 496 simulations at 10,000 m down-range. The derived corrections were 0.71 dB and 0.66 dBZ for 497 the gain and reflectivity factor corrections, respectively, which are smaller than those predicted 498 by Eqs. (17) and (20) of Sekelsky (2002), likely due to the unique structural and radiation 499 characteristics of the MCR (e.g., -12 dB sub-reflector illumination, and dish-edge taper level of -500 20 dB). Regarding the calculations of signal-to-interference ratio (SIR) of Fig. 13, Eq. (2.2) of 501 Doviak et al (1979), and Eqn. 1A of Schmidt et al (2012) were utilized along with the 2-way, 502 normalized radiation function ( 𝑓4(𝜃, 𝜙)) derived from the near field simulation at 1940 m 503 down-range to derive the following expression: 504 505 𝑆𝐼𝑅 = 𝜋5|𝐾𝑝2| 2 𝐷6 𝐿𝐺 𝑓4(𝜃,𝜙) 𝜆4𝑅4(𝑍𝑁𝑅+1) Eq. 2A 506 507 where D is the particle diameter, λ is the radar wavelength, LG is the MCR loop gain expressed 508 as a power ratio of the signal power to noise power and has a value of 1027 m2 (or 270 dB), Kp is 509 the dielectric constant of the particle, R is the down-range to the particle, and ZNR is the 510 background (ZNR) signal-to-noise power ratio given by ZNR = dBZbk -20log10(R) + 111.21 511 where dBZbk is the background reflectivity value (dBZ). Owing to the small angular spread in 512 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

24 the cross-range compared to R, the products of R with both the transverse and elevation angles, 513 𝜙 and 𝜃, are equated to the Cartesian x and y coordinates, respectively, which locate the points 514 in the cross-range plane of the simulations. Since the antenna pattern shown in Fig. 3a is 515 circularly symmetric, and only varies by ~0.1 dB with no significant change in shape (not 516 shown) over the ~10 m range of altitude spanned by the falling raindrop streaks shown in Fig. 517 10, for simplicity, it is assumed that the 3mm raindrop traverses the MCR beam while travelling 518 along the x-axis of Fig. 3a between -15 m < x < 15 m. 519 520 References 521 522 Atlas, D., W. G. Harper, F. H. Ludlam, and W. C. Macklin, 1960: Radar scatter by large hail. Q. 523 J. Roy. Meteor. Soc., 86, 468–482. 524 525 Bargen, D. W., and R. C. Brown, 1980: Interactive radar velocity unfolding. Preprints, 19th 526 Conf. on Radar Meteorology, Miami, FL, Amer. Meteor. Soc., 278–283. 527 528 Beard, K. V., 1976: Terminal velocity and shape of cloud and precipitation drops aloft. J. Atmos. 529 Sci., 33, 851–864. 530 531 Beard, K. V., D. B. Johnson, and A. R. Jameson, 1983: Collisional forcing of raindrop 532 oscillations. J. Atmos. Sci., 40, 455–462. 533 534 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

25 Beard, K. V., 1984: Oscillation models for predicting raindrop axis and backscatter ratios. Radio 535 Sci., 19, 67–74. 536 537 Beard, K.V., and R. J. Kubesh, 1991: Laboratory measurements of small raindrop distortion. Part 538 2: Oscillation frequencies and modes. J. Atmos. Sci., 48, 2245–2264. 539 540 Beard, K.V., and R. J. Kubesh, 1993: Laboratory measurements of spontaneous oscillations for 541 moderate-size raindrops. J. Atmos. Sci., 50, 1089–1098. 542 543 Blanchard, D. C., 1950: The behavior of water drops at terminal velocity in air. Eos Trans. 544 AGU, 31, 836–842. 545 546 Bringi, V. N., and V. Chandrasekar, 2001: Polarimetric Doppler Weather Radar. Cambridge 547 University Press, 636pp. 548 549 Brook, M., and D. J. Latham, 1968: Fluctuating radar echo: Modulation by vibrating drops. J. 550 Geophys Res., 73, 7137–7144. 551 552 Cappellin1, C, and L. Salghetti-Drioli, 2016: Near field 3D reconstruction of the search and 553 rescue antennas on the GALILEO satellite, Proc. EuCAP’2016, 10th European Conference on 554 Antennas and Propagation, 10-15 April 2016, Davos, Switzerland. 555 556 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

26 Chandra, A., C. Zhang, P. Kollias, S. Matrosov, and W. Szyrmer, 2015: Automated rain rate 557 estimates using the Ka-band ARM Zenith Radar (KAZR). Atmos. Meas. Tech., 8, 3685–3699. 558 559 Chandrasekar, V., R. Keränen, S. Lim, and D. Moisseev, 2013: Recent advances in classification 560 of observations from dual polarization weather radars. Atmos. Res., 119, 97–111. 561 562 Delene, D. J., 2011: Airborne data processing and analysis software package. Earth Sci. Inf., 4, 563 29–44, 564 565 Doviak, R. J., D. S. Zrnić, and D. S. Sirmans, 1979: Doppler Weather Radar, Proceedings of the 566 IEEE, 67, No. 11, 1522–1556. 567 568 Gayet, J. F., G. Mioche, L. Bugliaro, A. Protat, A. Minikin, M. Wirth, A. Dörnbrack, V. 569 Shcherbakov, B. Mayer, A. Garnier, and C. Gourbeyre, 2012: On the observation of unusual 570 high concentration of small chain-like aggregate ice crystals and large ice water contents near the 571 top of a deep convective cloud during the CIRCLE-2 experiment. Atmos. Chem. Phys., 12, 727–572 744. 573 Gunn, R., 1949: Mechanical resonance in freely falling raindrops. J. Geophys. Res., 54, 383–385. 574 575 Hallett, J., and S. C. Mossop, 1974: Production of secondary ice particles during the riming 576 process. Nature, 249, 26–28. 577 578 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

27 Heymsfield, A. J., L. M. Miloshevich, A. Slingo, K. Sassen, and D. O’C. Starr, 1991: An 579 observational and theoretical study of highly supercooled altocumulus. J. Atmos. Sci., 48, 923–580 945. 581 582 Houser, J. L., and H. B. Bluestein, 2011: Polarimetric Doppler radar observations of Kelvin-583 Helmholtz waves in a winter storm. J. Atmos. Sci., 68, 1676–1702. 584 585 Houze Jr, R. A., 2004: Mesoscale convective systems. Rev. Geophys., 42, 586 10.1029/2004/RG00150(2004). 587 588 Hubbert, J. C., J. W. Wilson, T.M. Weckwerth, S. M. Ellis, M. Dixon, and E. Loew, 2018: S-589 Pol’s polarimetric data reveals detailed storm features (and insect behavior). Bull. Amer. Meteor. 590 Soc., 99, 2045–2060. 591 592 Illingworth, A. J., J.W.F. Goddard, and S. M. Cherry, 1987: Polarization radar studies of 593 precipitation development in convective storms. Quart. J. Roy. Meteor. Soc., 113, 4694–89. 594 595 Jorgensen, E., P. Meincke, N. Vesterdal and O. Borries, 2018: Advances in commercial software 596 for high-gain antennas and electrically large platforms, Proc. 2018 IEEE AP-S Symposium on 597 Antennas and Propagation, 8-13 July 2018, Boston, MA. 598 599 Kollias, P., E. E. Clothiaux, T. P. Ackerman, B. A. Albrecht, K. B. Widener, K. P. Moran, E. P. 600 Luke, K. L. Johnson, N. Bharadwaj, J. B. Mead, M. A. Miller, J. Verlinde, R. T. Marchand and 601 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

28 G. G. Mace, 2016: Development and Applications of ARM millimeter-wavelength cloud radars. 602 Meteor. Monogr., 57, 17.1–17.19. 603 604 List, R., and M. J. Hand, 1971: Wakes of freely falling water drops. Phys. Fluids, 14, 1648–605 1655. 606 607 Marshall, J. S., and W. Mc K. Palmer, 1948: The distribution of raindrops with size, J. Meteor., 608 5, 165–166. 609 610 Matrosov, S. Y., R. F. Reinking, R. A. Kropfli, B. E. Martner, and B. W. Bartram, 2001: On the 611 use of radar depolarization ratios for estimating shapes of ice hydrometeors in winter clouds. J. 612 Appl. Oceanic. Meteor., 40, 479–490. 613 614 Matrosov, S. Y., G. G. Mace, R. T. Marchand, M. D. Shupe, A. G. Hallar, and I. B. McCubbin, 615 2012: Observations of ice crystal habits with a scanning polarimetric W-band radar at slant linear 616 depolarization ratio mode. J. Atmos. Oceanic. Technol., 29, 989–1008. 617 618 Musgrove, C., and M. Brook, 1975: Microwave echo fluctuations produced by vibrating water 619 drops. J. Atmos. Sci., 32, 2001–2007. 620 621 Orr, B. W., and R. A. Kropfli, 1999: A method for estimating particle fall velocities from 622 vertically pointing Doppler radar. J. Atmos. Oceanic Tech., 16, 29–37. 623 624 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

29 Pruppacher, H. R, and J. D. Klett, 1978: Microphysics of Clouds and Precipitation. Reidel , 625 Boston, Massachuesetts, pp 315–322. 626 627 Rayleigh, J.W.S, 1879: On the capillary phenomena of jets. Proc. R. Soc. London, 19, 71–97. 628 629 Rinehart, R. E., 2004: Radar for meteorologists. Rinehart Publications. 381–392. 630 631 Schmidt, J. M., P. J. Flatau, P. R. Harasti, R. D. Yates, R. Littleton, M. S. Pritchard, J. M. 632 Fischer, E. J. Fischer, W. J. Kohri, J. R. Vetter, S. Richman, D. B. Baranowski, M. J. Anderson, 633 E. Fletcher, and D. W. Lando (2012). Radar observations of individual rain drops in the free 634 atmosphere. Proc. Natl. Acad. Sci., 109, 9293–9298. 635 636 Schmidt, J.M., P.J. Flatau, and R. D. Yates, 2014: Convective cells in altocumulus observed with 637 a high-resolution radar, J. Atmos. Sci., 71, 2130–2154. 638 639 Schmidt, J.M., P. J. Flatau, and P. R. Harasti, 2017: Evidence for nimbostratus uncinus in a 640 convectively generated mesoscale cloud shield, J. Atmos. Sci., 74, 4093–4116. 641 642 Sekelsky, S., 2002: Near-field reflectivity and antenna boresight gain corrections for millimeter-643 wave atmospheric radars, J. Atmos. Oceanic Technol., 19, 468-477. 644 645 Seliga, T. A., and V.N. Bringi, 1976: Potential use of radar differential reflectivity measurements 646 at orthogonal polarizations for measuring precipitation. J. Appl. Meteor., 15, 69–76. 647 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

30 648 Stephens, G. L., D. G. Vane, R. J. Boain, G. G. Mace, K. Sassen, Z. Wang, A. J. Illingworth, E. 649 J. O’Connor, W. B. Rossow, S. L. Durden, S. D. Miller, R. T. Austin, A. Benedetti, C. Mitrescu, 650 and the CloudSat Science Team, 2002: The CloudSat mission and the A-Train: A new dimension 651 of space-based observations of clouds and precipitation. Bull. Amer. Meteor. Soc, 83, 1771–652 1790. 653 654 Szakáll, M., K. Diehl, and S. K. Mitra, 2009: A wind tunnel study on the shape, oscillation, and 655 internal circulation of large raindrops with sizes between 2.5 and 7.5mm. J. Atmos. Sci., 66, 755–656 765. 657 658 Szakáll, M., S. K. Mitra, K. Diehl, and S. Borrmann, 2010: Shapes and oscillations of falling 659 raindrops–A review. Atmos. Res., 97, 416–425. 660 661 Testik, F. Y., A. P. Barros, and L.F. Bliven, 2006: Field observations of multimode raindrop 662 oscillations by high-speed imaging. J. Atmos. Sci., 63, 2663–2668. 663 664 Willis, J. R., K. A. Browning, and D. Atlas, 1964: Radar observations of ice spheres in free fall. 665 J. Atmos. Sci., 21, 103–108. 666 667 Yuter, S. E., and R. A. Houze, 2003: Microphysical modes of precipitation growth determined 668 by S-band vertically pointing radar in orographic precipitation during MAP. Q. J. Roy. Meteorol. 669 Soc., 129,455–476. 670 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

31 671 Willson, J.W., T. M. Weckwerth, J. Vivekanandan, R. M. Wakimoto, and R.W. Russell, 1994: 672 Boundary layer clear-air radar echoes: origin of echoes and accuracy of derived winds. J. Atmos. 673 Oceanic Technol, 11, 1184–1206. 674 675 Zrnić, D. S., V. N. Bringi, N. Balakrishnan, K. Aydin, V. Chandrasekar, and J. C. Hubbert, 1993: 676 Polarimetric measurements in a severe hailstorm. Mon. Wea. Rev., 121, 2223–2238. 677 678 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

32 679 Parameter MCR UND North Pol CSU CHILL NCAR S- Pol WSR-88D KAZR Frequency (GHz) 5.4 – 5.9 5.6 2.725 2.7 – 2.9 2.7-3.0 34.86 Transmit Polarization Right Circular Linear H, V Linear H, V, Slant 45/135, Right and Left Circular Linear H, V Linear H, V Linear H Receive Polarization Right and Left Circular Linear H, V Linear H, V, Slant 45/135, Right and Left Circular Linear H, V Linear H, V Co-/Cross- Polar Antenna Dish Diameter (m) 15.24 3.66 8.5 8.5 8.5 2 to 3 Beamwidth (degrees) 0.22 0.99 1.1 0.91 0.925 at 2850 MHz 0.31 to 0.19 Peak Power (MW) 3 0.25 0.8 to 1.0 1.0 0.7 0.002 Maximum Pulse Rate (Hz) 320 1000 1250 1300 1304 10000 Pulse Width (s) 12.5 (LFM) 0.6 to 2.0 0.2 to 1.6 (Gaussian) 0.3 to 1.4 (Tapered) 1.57 to 4.71 0.03 to 12 (NLFM) Maximum/Minimum Range Resolution (m) 0.543/34.0 90/300 30/150 38/1000 250/1000 30 Sensitivity at 5 Km Range [Z (dBZ)] -57 (NB) -38(WB) -28 -24 -29 -28 to -37 -31 Sensitivity at 150 km Range [Z (dBz)] -26 (NB) -8 (WB) 2 6 1 2 to -7 -1 (150 km range not attainable) -21 (at 15km) Typical Scan Range Interval (km) 75 (NB) 0.375 (WB) 150 150 150 300 (Doppler) 460 (dBZ, dual-Pol) 15 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

33 680 Table 1. The Table provides characteristics of the MCR and a selection of standard weather 681 research radars including the UND North Pol, the CSU-CHILL, the NCAR S-Pol, WSR-88D, 682 KAZR Ka-band (Kollias et al. 2016; Chandra et al. 2015) Doppler radars. The UND North Pol is 683 a Weather Surveillance Radar designed in 1974 (WSR-74C) that has since been upgraded to a 684 dual polarization system. See text for more details (Adapted from Rinehart (2004) with updates 685 from the CHILL and S-Pol Radar websites, and personal communication with UND). 686 687 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

34 Instrument Manufacturer Parameter Resolution Frequency/ Wavelength CL51 Ceilometer Vaisala Backscatter 10 m Gate Spacing @ 1 min 910 nm Micro Pulse Lidar Sigma Space Backscatter 5/15/30/75 m Gate Spacing @ 1 min 532 nm Micro Rain Radar METEK dBZ, Ur, PSD, Nt, LWC 30 Range Gates (30-6000 m) @ 10-3600 s 24 GHz MP3000A Radiometrics T, Qd, LWC 50-500 m Gate Spacing @ 2 min V & K-band Parsivel2 Disdrometer OTT Nt, VT, dBZ, Visibility 20 Bins: 200 µm-8 mm (liquid) 200 µm-25 mm (solid) 650 nm All-Sky Camera Prede Cloud Images 5 Megapixel Image @ 1 min Visible Snow White Meteolabor T, Td, P, U, V 0.5 to 2 s Response Times Doppler Lidar Halo Photonics Ur, Back Scatter 18 m @ 10 Hz 1.5 µm Sonic Anemometer Gill U, V, W, T 20 Hz Li-7500DS LiCor H2O, CO2 20 Hz FM120 Droplet Measurement Technologies PSD, Nt 30 Channels (2-50 um) AQT420 Vaisala Gas Concentration (NO2, SO2, CO, O3) Particle Matter (PM2.5, PM10) 60 s Averaging WXT520 Vaisala T, P, RH, Precipitation, SPD, DIR 4 Hz Ka-Band Doppler Radar ProSensing dBZ, Ur 30, 75, 100 or 150 m 35.4 GHz *Liquid Water Content Probe King LWC 1 Hz *Two-Dimensional Optical Array Imaging Probe Particle Measuring Systems Particle Images, PSD, Nt 32 Diodes of 25 um Each, Asynchronous Data *Water Content Probe Nevzorov Total Water Content, LWC 25 Hz *TT102 Rosemount T 1 Hz (2 s Response Times *FSSP Particle Measuring Systems PSD, Nt 30 Channels (2 – 47 µm), 1 Hz Data *239 Pressure Probe Setra ±0.05% & ±0.14% Full Scale 20 ms Response Time *Td Probe Model 137 EdgeTech Td 25 Hz Data (1.5 °C s-1 Response Time) *TDL Maycomm Research Co. Qv 1.6 Hz *POSAV 310 and 5 Port Gust Probe Applanix U,V,W 25 Hz *2D-S Stratton Park Engineering Company, Inc. Particle Images, PSD, Nt 128 Diodes of 10 µm Each, Asynchronous Data *HVPS3 Stratton Park Engineering Company, Inc. Particle Images, PSD, Nt 128 Diodes of 150 µm each, Asynchronous Data *CDP Droplet Measurement Technologies PSD, Nt 30 Channels (2-50 µm), 10 Hz, Particle-by-Particle Data 688 Table 2: A list of the primary surface-based and airborne (denoted by asterisk) instrumentation 689 used during the 2008-2015 field campaigns and/or planned for 2019. The research aircraft have 690 included the North Dakota Cessna Citation II jet aircraft and the dual prop Cessna Cheyenne 691 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

35 aircraft operated by Weather Modification International. Aircraft measurements were acquired 692 using a Science Engineering Associates model M300 data acquisition system and processed 693 using the Airborne Data Processing and Analysis Software Package (Delene 2011). The symbols 694 have the following meanings: pressure (P), temperature (T), dewpoint temperature (Td), liquid 695 water content (LWC), relative humidity (RH), water vapor density (Qd), water vapor mixing ratio 696 (Qv), east-west, north-south, and vertical wind components (U,V,W), wind speed (SPD), wind 697 direction (DIR), particle concentration (Nt), particle size distribution (PSD), High Volume 698 Particle Spectrometer Version 3 (HVPS3), Two-Dimensional Stereographic Optical Array Probe 699 (2D-S), Position and Orientation System for Airborne Vehicles Model 310 (POSAV 310), 700 Tunable Diode Laser Hygrometer model AC19-400 (TDL), Forward Scattering Spectrometer 701 Probe (FSSP), microwave radiometer (MP3000A), Cloud Droplet Probe (CDP), Fog Monitor 702 Model 120 (FM120), reflectivity (dBZ), Doppler radial wind component (Ur), terminal velocity 703 (VT), and Total Temperature Probe Model 102 De-iced (TT102). The four field experiments 704 were conducted during the following periods: 7-18 July 2008, 13-24 July 2009, 16-27 August 705 2010, 27 July to 9 August 2015. 706 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

36 707 708 709 Fig 1: The U.S. Navy Mid-Course Doppler Radar located near Titusville, FL as pictured during a 710 study of an altocumulus layer on 8 August 2015. 711 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

37 712 Fig. 2: Experimental design for the CAPE Weather Experiments held in the Titusville, FL 713 region. The shaded backdrop is a time-height image of the MCR-derived cirrus and mammatus 714 cloud reflectivity obtained between 22:16 and 22:54 UTC 8 August 2015. The bold red and 715 green lines represent the temperature and dewpoint temperature data derived from a balloon 716 launch at 1749 UTC 8 August 2015 (scaled in arbitrary units). The primary surface instruments 717 include the Sigma Space micropulse lidar (black rectangle) and the MCR dual polarization C-718 band Doppler radar. The pictured aircraft is the instrumented University of North Dakota’s 719 Cessna Citation II aircraft. The MCR is depicted as tracking the research aircraft in real-time 720 while the lidar is used to guide the placement of the aircraft and the short MCR wideband range 721 window. 722 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

38 723 724 Fig. 3: The simulated MCR near-field calculated at a range of 1.94 km from the radar showing 725 the (a) power density (dB) and (b) phase (degrees) for a transmitted pulse having a right hand 726 circular polarization. The contour interval in (a) is 5 dB and 30 degrees in (b). The dashed white 727 line in each plot shows the location of the cross-section plots shown in Fig. 4. 728 729 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

39 730 731 Fig. 4: The MCR simulated cross-beam antenna patterns showing: (a) the power density (dB), 732 and (b) the phase (degrees). The black curve in (a) shows the power density at 1.94 km range 733 whereas the red curve in (a) shows the power density at 10 km range, normalized in the across 734 range direction by the range-angular factor of 1.94/10. The black dashed lines in (a) and (b) are 735 the 3 dB beamwidth, and the equivalent phase angle level corresponding to the 3dB beamwidth 736 level measured across boresight in (a), respectively. 737 738 739 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

40 740 741 742 743 Fig 5: Multi-component analysis of the aircraft pass through the 1 August 2015 anvil system 744 pictured in (a) as the aircraft was tracked by the MCR: (b) range-time and along-track MCR 745 narrowband reflectivity [dBZ], (c) in situ measurements of temperature (°C), pressure (hPa), 746 total water (black) and liquid water content (blue) [g m-3], and (d) 2D-S images from the 2D-S 747 showing ice crystal chain aggregates (circled in red). The vertical width of each panel in (d) 748 represents 1280 µm and each panel represents less than one second of elapsed time. The particle 749 images were acquired during the anvil penetration when the aircraft was at a temperature of 750 ~-43.5 °C. The bold letter ‘o’ in (b) denotes the location of the crystal images shown in (d). 751 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

41 752 753 Fig. 6: The time-height cross-section of the MCR narrowband reflectivity recorded between 754 2145 and 2215 UTC 24 July 2009 (shaded) at a time when streaks were being observed with the 755 MCR wideband waveform. The white boxes labeled B1 and B2 denote the location of the 756 wideband reflectivity shown in Fig. 7 and Fig. 8, respectively. The precipitation streamer labeled 757 R1 denotes the primary rain shaft depicted with the same labeling in Fig. 8. The length-scale is 758 based on an estimate of the reflectivity cell movement of 5 m s-1 as observed with the Melbourne 759 WSR-88D Doppler radar during the time of interest. 760 761 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

42 762 763 Fig. 7: Time-height plot of the MCR wideband reflectivity structure (shaded) for the portion of 764 the storm associated with the fast rising turret discussed in the text (encapsulated in the box 765 labeled B1 of Fig. 6). The discovery occurred 10 minutes prior to shutting down the radar for the 766 2009 field campaign as the last scan of the day was being executed. 767 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

43 768 769 Fig. 8: As in Fig. 7 except the data are taken at the time corresponding to the box labeled B2 in 770 Fig. 6. The linear reflectivity features in this time-height plot represent individual hydrometeors 771 and are denoted as radar reflectivity “streaks”. 772 773 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

44 774 775 776 Fig. 9: Rain shaft and drop structure derived from the MCR wideband waveform as the storm 777 system pictured in (a) approached the radar on 30 July 2015. The wideband derived reflectivity 778 of the rain shaft (b) and of the individual raindrops (c) are taken at different times during the 779 storm passage. The vertical scale and elapsed time shown in (b) and (c) are ~320 m and 180 s, 780 respectively. The images are inverted so that time increases from right to left. 781 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

45 782 Fig. 10: High-resolution time-height plots of the MCR wideband radar cross section (RCS) for a 783 3.4 second time window taken during the 30 July 2015 storm system. The data reveal the 784 detailed structure of the individual cloud hydrometeors passing through the vertically orientated 785 radar beam. The data represent the pulse-to-pulse structure and is presented without any temporal 786 or spatial averaging. The “streaks” labeled S1, S2, S2a, S2b, S3, S4, S5, and S6 are discussed in 787 the text. The bold white arrows labeled S2a and S2b highlight the position of the two streaks 788 comprising the S2 streak. 789 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

46 790 791 Fig 11: The pulse-to-pulse RCS values (dBsm) obtained along the primary axis of the streaks: (a) 792 S2a, (b) S3, and (c) S5 shown previously in Fig. 10. The bold dashed line in each panel 793 represents a fourth-order polynomial fit to the data. The maximum drop diameter taken at the 794 apex of the fitted dashed curve is denoted in each panel. 795 796 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

47 797 798 799 Fig. 12: A power spectra analysis of the amplitude fluctuations for the streaks labeled: (a) S2b, 800 (b) S3, and (c) S5 shown previously in Fig. 10 and Fig. 11. The shading in each plot is the 801 spectral power [dBsm Hz-1] derived from a sequence of 128 pulse subsets (ni = ns:ns+128) taken 802 from the original streak time series. The starting pulse number (ns) for each successive subset 803 was incremented by a value of one from that used in the previous subset. The red solid curve 804 represents the mean value of the power at each frequency derived over all subsamples. The plot 805 was normalized and then scaled by the total number of subsamples in each panel so each mean 806 spectral just covers the entire plot. The bold dashed line white line represents the theoretical drop 807 oscillation frequency (fd) as a function of diameter (right ordinate) derived from Eq. 2 of Szakáll 808 et al. (2010). The frequency of the main spectral peaks is high-lighted by the white text. 809 810 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

48 811 Fig. 13: The minimum detectable particle diameter (mm) for a water drop at the specified ranges 812 of 2.0 km (solid), 4.5 km (dashed), and 10 km (dot-dashed) derived from Eq. A2. The solid black 813 dots labeled F7, F8, F9, F10, and R1 represent mean conditions for the streaks shown Fig.7, Fig. 814 8, Fig.9, Fig.10, and within the precipitation streamer labeled R1 in Fig. 8, respectively. The 815 value of the Signal to Interference Ratio (SIR) used in Eq. A2 was set to 3 dB. 816 817 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

49 818 819 820 Fig. 14: Idealized simulations of the RCS [dBsm] for the case of two drops with specified fixed 821 diameters representing the S2a (D2) and S2b (D1) drop interaction. The differential velocity is 822 set to 0.7 m s-1 as observed. The drops are initially separated by a range of 1.4 m with specified 823 fall speeds of either 5.5 and 4.8 m s-1 for the D1 and D2 particles, respectively. 824 825 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.

50 826 827 828 Fig. 15: Data from the low-level stratus case of 08 August 2015 case day showing: (a) the all-sky 829 camera imagery of the cloud layer at 1347 UTC, (b) the MCR-derived narrowband reflectivity 830 [dBZ], and (c) the MCR-derived total vertical velocity [ms-1]. The labels U1, U2, and U3 in (b) 831 and (c) denote regions of updrafts discussed in the text. The labeled arrows denote very thin 832 enhancements in the reflectivity at the top and periphery of the U2 (and other) updrafts (labeled 833 ER) and a narrow reflectivity minimum indicative of cloud top entrainment (labeled as EC). The 834 horizontal scale shown is based on an advective speed of ~10 m s-1. The reflectivity and velocity 835 values were derived using 64 temporally consecutive pulse samples. 836 Accepted for publication in Bulletin of the American Meteorological ociety. DOI S 10.1175/BAMS-D-18-0130.1.