Supercooled Drizzle Development in Response to Semi-Coherent Vertical Velocity Fluctuations Within an Orographic Layer Cloud

Observations of supercooled liquid water are nearly ubiquitous within wintertime, orographic layer clouds over the intermountain west; however, observations of regions containing supercooled drizzle drops (SCDDs) are much rarer and the factors controlling SCDD development and location less well understood. As part of the Seeded and Natural Orographic Wintertime clouds—the Idaho Experiment (SNOWIE) goal of improving understanding of natural cloud structure, this study examines the role of fine-scale (sub-kilometer) vertical velocity fluctuations on the microphysical evolution and location of 10 SCDDs within the observed mixed-phase, wintertime orographic clouds from one research flight in SNOWIE. For the case examined, SCDDs developed in an elevated, postfrontal layer cloud with cold cloud tops (T < -30 °C) and low number concentrations of both ice (less than 0.5 L-1) and cloud droplets (less than 30 cm-3). Regions of supercooled drizzle at flight level extended more than a kilometer along the mean wind direction and were first located at and below layers of semi-coherent vertical velocity fluctuations (SCVVFs) embedded within the cloud and subsequently below cloud top. The 15 microphysical development of SCDDs in this environment is catalogued using size and mass distributions derived from in situ probe measurements. Regions corresponding to hydrometeor growth are determined from radar reflectivity profiles retrieved from an airborne W-band cloud radar. Analysis suggests that SCVVF layers are associated with local SCDD development in response to the kinematic perturbation pattern. This drizzle development and subsequent growth by collision-coalescence is inferred from vertical reflectivity enhancements (-20 dBZ km-1), with drizzle production confirmed 20 by in situ measurements within one of these SCVVF layers. The SCDD production and growth occurs embedded within cloud over shallow (km or less) layers before transitioning to drizzle production at cloud top further downwind, indicating that wind shear and resultant vertical velocity fluctuations may act to enhance or speed up SCDD development compared to classic cloud top broadening mechanisms in orographic (or similarly sheared) cloud environment(s).


Introduction 25
Over the last forty years, there have been numerous field campaigns either directly or indirectly examining mixed-phase, orographic layer clouds over the intermountain western United States (Hobbs, 1975;Cooper and Saunders, 1980; Heggli and on active ice nucleating particle (INP) and ice crystal number for SCDD formation, else ice will more rapidly scavenge the available vapor and cloud water, inhibiting growth of cloud droplets to drizzle sizes (Rasmussen et al., 2002;Geresdi and Rasmussen, 2005). A byproduct is that SCDD observations are infrequent in clouds with cloud tops colder than -15 °C, with few observations of SCDD formation found in the literature with cloud tops colder than -20 °C (Lawson et al. 2001;Korolev et al., 2002;Rosenfeld et al., 2013;Silber et al., 2019).
Collision-coalescence initiation and growth often depend on broadening mechanisms for the largest droplets to begin collection of smaller droplets in the population in all but the cleanest of clouds (Wood et al., 2018), and this is true regardless if clouds are supercooled. Steady condensational growth alone leads to a narrowing of the drop size distribution (DSD) around a large drop mode (D ~ 30-40 μm), such that DSD broadening mechanisms (e.g. turbulent or isobaric mixing, eddy 70 hopping, etc.) are necessary to provide the differential fall speed conducive to collision-coalescence onset and subsequent rapid collectional growth. Pobanz et al. (1994) found that when SCDDs formed in clouds with cloud droplet number concentrations of more than 100 cm -3 , layers of cloud top shear were correlated with vertical location of drizzle development, presumably due to turbulent broadening or mixing. Shear-induced turbulent mixing, especially at cloud top, is thought to be responsible for relatively rapid DSD broadening (Grabowski and Abade, 2017). Any isobaric mixing of 75 different temperature parcels near the cloud boundary (e.g. for clouds with a strong capping inversion) are expected to further accelerate this process. This is why the warm rain process is understood to start at or near cloud top, with drizzle mass principally increasing with cloud layer depth (Comstock et. al., 2007).
Supersaturation history provides an analytical framework for understanding several mechanisms (e.g. vertical velocity fluctuations, turbulent eddy hopping, mixing events, etc.) that may be responsible for the rapid spectral broadening and 80 subsequent collision-coalescence enhancement in warm stratiform clouds (Cooper et al., 1989;Korolev and Mazin, 1993;Politovitch and Cooper, 1994;Korolev, 1995). For instance, Korolev found that when modeled cloud parcels are subjected to repeated vertical velocity fluctuations, DSDs broaden and may even see a second, small-diameter droplet mode develop from interstitial CCN activation (hereafter, secondary droplet activation). Turbulence and wave motions were both suggested as possible sources for these vertical velocity fluctuations, but the lack of parcel-following in situ measurements made 85 validating these behaviors an observational challenge (Pobanz et al., 1994).
Between the orographic SLW case studies (Rauber and Grant, 1986;Rauber 1992), SCDD climatologies (Rauber et al., 2000;Bernstein et al. 2007), mechanistic understandings of SCDD production (Rosenfeld et al. 2013), and exceptional cases (Korolev and Isaac, 2000;Pobanz et al., 1994), a clear picture of SCDD formation develops: clouds formed in gradual updrafts within low CCN and INP populations are most likely to produce SCDD. The frequency, spatial extent, and 90 thermodynamic extremity of SCDD production is a function of CCN and INP abundance (Rosenfeld et al., 2013). Wind shear and dynamic instability appear to lead to SCDD development in clouds with exceptionally high CCN concentrations given low enough ice concentrations (Korolev and Isaac, 2000;Pobanz et al., 1994). Mixed-phase clouds throughout the western U.S. in which the phase partitioning is mostly liquid are common even well away from the coast (Hindman, 1986).
Such clouds must contain low concentrations of cloud droplets and ice to develop SCDDs (Saleeby, 2011). Where 95 encountered in orographic environments, these supercooled, relatively clean clouds are expected to encounter vertical and turbulent motions at both broad and fine scales (Houze and Medina, 2005).
This study examines an individual case from a field campaign located in the IMW in which SCDDs developed in a winter orographic cloud system despite cold cloud tops (T ~ -30 °C) which are typically associated with more active ice nucleation and more abundant natural ice (DeMott et al., 2010). Persistently low droplet number concentrations (less than 50 cm -3 ) and 100 frequent SCDD observations from about half of the cases throughout the field campaign (Tessendorf et al., 2018) inspired this analysis and are consistent with the climatological maxima of wintertime SCDD frequency that stretches from the coastal barrier mountains into Idaho (Bernstein et al., 2007). The analysis focuses on the spatial kinematic patterns and their effect on the liquid phase precipitation development in these mixed phase clouds.

Study Area and Data 105
The Seeded and Natural Orographic Wintertime clouds-the Idaho Experiment (SNOWIE) was designed to observe and analyze the evolving wintertime orographic cloud structure in a series of prescribed airborne cloud seeding experiments (Tessendorf et al., 2018). As part of this process, it was necessary to establish the evolution of the natural cloud structure and microphysics as a baseline for evaluating cloud seeding effects. Separately, the extensive dataset and state-of-the-art measurements were expected to yield new insights toward the natural cloud structure, microphysical evolution, and 110 precipitation patterns of mixed phase winter orographic clouds. Understanding how fine-scale (km or less) dynamical processes impact cloud microphysical development and spatial distribution, amount, and phase, of observed precipitation in such clouds is at the forefront of the remote sensing and cloud microphysics observational literature (e.g. Houze and Medina, 2005) and further provides valuable insight to cloud modeling and microphysical parameterizations.
To characterize and describe the development of precipitation hydrometeors (e.g. SCDDs) at flight level requires direct 115 measurements of cloud hydrometeor spectra, thermodynamic and dynamic conditions (which govern the development of the spectra), and characterization of the spatial variability of each. Remote profiling radar, in situ cloud probes, temperature and humidity sensors, and gust probes, onboard the University of Wyoming King Air (UWKA) research aircraft catalogued the evolving cloud structure and precipitation patterns for repeated fixed flight legs oriented along the mean wind direction through cloud (Fig. 1), at as low an altitude as practical. UWKA legs, anchored above the Packer John (PJ, see Fig. 1) 120 ground site, recurrently sampled coincident spatial cross sections through the evolving orographic cloud structure, often between the -10 to -15 °C level. Flight legs (blue line in Fig. 1) were generally no longer than 100 km, with the western end located over the valley and the eastern end extending over the Sawtooth Mountain Range. Soundings launched at Crouch, ID (KCRH, Fig. 1) before and during each flight were used to characterize bulk thermodynamic and dynamic conditions. Measurements from the W-Band Wyoming Cloud Radar (WCR) documented the orographic cloud structure above and 125 below flight level and provided context for the in situ cloud microphysics measurements (as in Vali et al., 1998;Wang and Geerts, 2003;Wang et al., 2012). Previous studies demonstrated that the WCR resolves fine-scale details of orographic clouds (~30 m spatial resolution), observing aspects of their dynamical and microphysical structure technologically impossible in previous decades (Aikins et al., 2016). The WCR is sensitive to cloud droplets and drizzle in the Rayleigh regime, with Mie effects starting at around 600 µm and reflectivity increasing monotonically with diameter up to millimetric 130 sizes (D > 0.95 mm). Radar reflectivity for volumes containing even large drizzle drops was therefore dominated by the contribution of the largest drops, and throughout SNOWIE no drizzle drops larger than 0.5 mm diameter were observed such that Mie effects were non-existent for purely liquid volumes. Doppler velocity measurements from the WCR captured the near-vertical, reflectivity-weighted motions of the distributed hydrometeor targets. In the data presented here, no attempt has been made to separate hydrometeor terminal fall velocity with vertical air motions. Since the antennas point nearly vertical, 135 the influence of horizontal wind in the Doppler measurements is negligible for straight and level flight.
In situ probes on the UWKA measured cloud hydrometeors with diameters from a few microns to several millimeters (Table   1). Two probe types were used to collect these data-a forward scattering cloud probe (i.e. the Cloud Droplet Probe, CDP), and two optical array probes (OAPs) for larger hydrometeors. The CDP (Lance et al., 2010) provided 5 Hz cloud droplet (1 to 50 μm) size spectra in bins 1 to 2 μm wide. The CDP RMS accuracy of mean droplet diameter of 0.7 μm was determined 140 after the campaign using the University of Wyoming droplet generator (Faber et al., 2018).
The OAPs image larger hydrometeors (D > ~50 μm) as particles pass through an illuminated sample volume and shadow individual members of a linear photodiode array. The 2D Stereo probe (2DS; Lawson et al., 2006) imaged particles at a 10 μm resolution across a 1.28 mm diode array, accurately resolving the hydrometeor spectra for particles 50 μm < D < 1 mm.
The 2D Precipitation probe (2DP) measured hydrometeors larger than a millimeter with an image resolution of 200 μm. The 145 data from the OAPs were processed using the University of Illinois OAP Processing Software (Jackson et al., 2014, Finlon et al., 2016, to perform standard image rejection and dimension corrections. Image-derived size and particle timing information and a sample volume estimate following Heymsfield and Parrish (1978) were used to produce particle size distributions. Shattering artifacts were avoided using anti-shattering tips on the 2DS and by filtering of particles with a short, static inter-arrival time threshold in the software processing. 150 From these 1 Hz particle size spectra, several integrated water content metrics were calculated to estimate the mass distribution within certain drop size categories. The total liquid water content-i.e. across the entire measured liquid hydrometeor size spectrum-was integrated from the combined CDP and 2DS size spectra and is hereafter referred to as LWCtot. The Cloud Water Content (CWC) and Drizzle Water Content (DWC) metrics contain the mass from the 2 to 50 μm and 50 μm to 1 mm parts of the cloud hydrometeor spectrum, respectively, and hence sum to LWCtot. The calculated LWCtot 155 was compared to the bulk estimate from the Rosemount icing probe, which is sensitive to all sizes of SLW drops. A comparison performed over two flight legs validated these estimation methods. The only remarkable disagreement between the metrics came for LWC values of the Rosemount greater than 0.4 g m -3 , where the integrated LWCtot was larger compared to the Rosemount icing probe measurement. This may be an overestimation of LWCtot related to mis-sizing of drizzle drops that are near the edge of the depth-of-field in the 2DS and appear as hollow images. Conversely, this may also be due to an 160 underestimation from the Rosemount probe due to splashing of SCDDs that are not completely captured by the probe's icing rod. Regardless, the error is almost certainly associated with the liquid mass of SCDDs and the Rosemount and the integrated LWCtot estimates provide a lower and upper bound, respectively.
The following results and analysis produced from the WCR profiles, in situ bulk probes, and cloud microphysics datasets from the first UWKA flight in SNOWIE, highlight the role of sub-kilometer vertical velocity fluctuations on the 165 spatiotemporal distribution of SCDDs and the inferred cloud microphysical response.

Results
The results presented are from the period of 0245 to 0405 UTC (legs 1, 2, and 5) during the first flight of the field campaign on January 8, 2017. Two distinct layer clouds developed in the wake of a precipitating frontal cloud system. Of these two clouds, the elevated cellular cloud layer contained both low background number concentrations of ice and cloud droplets and 170 embedded kilometer or longer regions of SCDDs that formed in a larger pattern of orographic lift.

Synoptic and Thermodynamic Context
The UWKA research flight followed the passage of a deep snow band associated with a weak jetstreak in the 500 mb wind field. The deep, saturated atmosphere present in the upstream sounding during the heavily-precipitating period roughly 4 hours prior to the start of leg 1 (Fig. 2a), experienced mid-tropospheric drying, and veering and strengthening of the winds 175 above 8 km MSL. This led to lowered cloud tops and a pronounced dry slot from 7 to 9 km in the pre-flight sounding just 3 hours later (Fig. 2b). This dry layer contained thin layers of expected dynamic instabilities-defined by bulk Richardson number from 0 to 0.5 ( Fig. 2b; blue shading). The layer below, between 4 and 7 km, saw several vertical humidity variations accompanied by evaporational cooling of the radiosonde upon exiting cloud layer tops, resembling conditional instabilities (orange shading). These layers were not expected to correspond to real convective motions in cloud. 180 By the start of the first flight leg at 2:45 UTC, a shallow orographic cloud layer persisted over the study region on the western end of the flight track, with cloud tops around 4 km MSL (Fig. 3a)-matching the top of the lower saturated layer in the pre-flight sounding (Fig. 2b). This orographic cloud layer was capped on the eastern end by a layer of broken, cellular cloud structures roughly 1 to 3 km wide-hereafter the elevated cellular layer-resembling, at times, either coherent K-H billows or incoherent generating cells. This elevated cellular layer was consistently strongest in terms of layer depth and 185 highest radar reflectivities over the highest terrain at the east end of the leg.
The final upstream sounding launched one hour after the start of leg 1 (Fig. 2c) indicated a deeper saturated layer through 6.5 km and further strengthening and veering of the wind above, with more vertically homogeneous, near-zonal winds between 3 and 6 km. This shear profile resulted in several layers of possible dynamic instability within 500 m both above and below the top of the saturated layer and matched well with the 6 to 6.5 km cloud tops observed with the WCR during flight legs 4 and 5 190 ( Fig. 3d and 3e).
Variations in humidity and wind, superimposed on the background zonal winds and low-level orographic clouds, appeared responsible for an elevated cloud layer that was at times unstable and variable in vertical location and depth (Fig. 2b).
Additionally, a surface inversion and attendant low-level static stability was present in all the upstream soundings around the time of the flight (Fig. 2a, b and c). As a result, calculated bulk Froude numbers were consistent with blocked flow below 2 195 km MSL, matching the overall low-level static stability pattern that was present through much of the field campaign (Tessendorf et al., 2018). The stability from this surface inversion may have helped to decouple the surface airmass from the free troposphere above the Sawtooth Range barrier.

General Cloud Structure and Vertical Motions
There were several differences between the orographic cloud layer (4.5 km MSL and below) and the cellular layer above. 200 The orographic cloud layer persisted over the nearest 1 to 2 km above the terrain, with cloud tops that rose slightly (no more than 500 m) from west to east with the average height of the topography beneath (e.g. Fig. 3a). The cellular layer, however, was transient-discrete layers of cells advected into the target area at varying altitudes. Some of these layers appeared coupled to the lower orographic cloud layer (as in legs 1, 2, 4, and 5), while others appeared totally separate (as in legs 3, 9, and 10). This behavior is consistent with the large vertical variations in wind shear and humidity between the three 205 soundings in this layer (Fig. 2), including several dynamically unstable layers. Consistent with this, several of the elevated layers appeared to contain overturning (or breaking) cells in the reflectivity profiles, for example, within the elevated cellular layer of leg 4 from 10 to 15 km downwind of PJ (Fig. 3d).
Across the entire research flight, the radar reflectivity within the upper cloud layer was less than -5 dBZ except for discrete, individual fall streaks. This behavior suggests mostly liquid cloud species in the elevated layer, confirmed by the 99 th 210 percentile of precipitation-sized ice number (integrated from the 2DP probe) for each of the first four legs remaining below 0.1 L -1 and for leg 5 only marginally higher, with a 99 th percentile value of 0.3 L -1 . Some of the higher reflectivity fall streaks, especially towards the end of the flight, may have corresponded to seeding lines Tessendorf et al., 2018;Hatt, 2019) after the seeding period started at the end of leg 2, but are otherwise beyond the scope of this study.
The radar reflectivity within the lower orographic cloud layer, by comparison, was greater than in the cellular layer above. 215 Large regions within 1 km of the surface contained reflectivity greater than 5 dBZ, suggesting the presence of ice below the orographic cloud top. This conjecture is consistent with a significant reduction of about 4 m s -1 in downward Doppler velocity in the lowest ~1 km above the surface (Fig. 4). This reduction often occurred at a level corresponding to an increase in the radar reflectivity. The inferred relative abundance of ice in this shallow orographic layer may be due to more abundant aerosol (and INP) presumed to reside below the strong surface inversion (Fig. 2b) or from secondary ice multiplication in the 220 warm (-5 < T < -15 °C) temperatures in the lower layer, but no direct in situ measurements were available in cloud below flight level.
Mean reflectivity-weighted, near-vertical Doppler velocities (hereafter, hydrometeor vertical velocities or Doppler velocities) were available from the WCR to quantify cloud vertical motions (i.e. the convolution of vertical air motions and reflectivity-and phase inhomogeneity convoluted the observed Doppler velocities, making assumptions about a constant hydrometeor fall speed specious. In fact, the spread of fall speeds associated with observed hydrometeor size distributions were greater than the spread of air motions observed in the dynamic structures of focus (< 1.5 m s -1 amplitude where sampled at flight level).
Despite this complexity, there were several obvious and consistent trends in the observed Doppler velocities: nearly all legs 230 showed a distinct terrain-induced vertical velocity couplet centred roughly 24 km downwind of Packer John and directly above a pronounced N-S ridge, oriented perpendicular to the mean wind and flight direction (Fig. 4). This couplet consisted of up to 2 m s -1 upward Doppler velocities over the upwind slope immediately followed by as much as 4 m s -1 downward Doppler velocities on the downwind side, and frequently extended up to cloud top (as in leg 5). Despite the wave-like signatures present in the reflectivity profiles, Doppler velocity couplets away from flight level, and phase relationships at 235 flight level, between perturbation kinematic and thermodynamic quantities (not shown) were inconsistent with K-H waves.
For this reason, care was taken separately in (1) quantifying the effects of spatial variations in hydrometeor fall speed and (2) adopting the label of semi-coherent vertical velocity fluctuations (SCVVFs) to distinguish layers of these regularly-spaced, oriented vertical velocity perturbations from the more isotropic turbulent motions found elsewhere. Probable meteorological sources for SCVVFs in this environment include K-H waves, shear-driven mechanical overturning (Houze and Medina, 240 2005), and shallow convective overturning with some regular triggering mechanism; however, the actual sources did not seem to uniquely affect the microphysics and therefore remain undistinguished. What follows are descriptions of how SCVVFs affected the evolution and spatial distribution of precipitation in the elevated cellular cloud layer, significant for where drizzle development deviated from the expectation of starting at cloud top and collecting through the depth of SLW cloud. 245

Comparisons Between Drizzling Legs (1, 2, and 5)
The three flight legs of interest, 1, 2, and 5 (Table 2), were flown at altitudes ranging from 3.9 to 4.5 km MSL. During each of the legs the UWKA encountered kilometer-or-longer stretches of SCDD measured at flight level within the elevated cellular cloud layer. Significantly larger drops were observed on the first two legs compared to leg 5 despite similar cloud water contents across all three. The regions containing SCDDs were all located at or downwind of Packer John mountain 250 (PJ; the start of prominent terrain features along this transect), where reflectivities and cloud layer thicknesses were consistently near the leg maxima. Above the windward slope of the Sawtooth Range, from 10 to 25 km downwind of PJ, was a broad region of ascent observed on most legs (0 to 1 m s -1 hydrometeor upward velocities) which contributed to the relatively high reflectivities and cloud layer thicknesses compared to cloud further upwind (Fig. 4). From 10 to 60 km downwind of PJ, where SCDDs were encountered on all three legs, flight level vertical velocities varied from -0.5 to 2 m s -1 , 255 with perturbation magnitudes on legs 1 and 2 of up to 0.6 m s -1 and less than 0.2 m s -1 for leg 5 ( Table 2). The flight level temperatures on these legs ranged from -16 °C on legs 2 and 5 to -11 °C for the lowest altitude leg 5.
Cloud Water Content (CWC) measured at flight level were similar for these drizzling sections of cloud across all three flight legs, with maximum values approaching 0.6 g m -3 in legs 1 and 5. Slightly lower maximum CWCs were measured in the drizzling sections of leg 2, only as high as 0.4 g m -3 , possibly reduced due to scavenging and removal of cloud water by 260 drizzle in the time between legs 1 and 2 (Table 2). Cloud droplet number concentration measured at flight level during all three legs never exceeded 35 cm -3 and decreased to values less than 5 cm -3 within portions of cloud in which there appeared significant SCDD sedimentation from above. Within these plumes of SCDDs, which appeared only in flight legs 1 and 2, DWC measured at flight level were at times as high as 1 g m -3 . Also within SCDD plumes, the mean-volume diameter of the DSD approached 80 μm (Table 2). Unlike the first two legs, the SCDDs sampled in leg 5 were much smaller, and the DSD 265 mean-volume diameter did not exceed 45 μm.
The primary microphysical differences for these three legs were the smaller SCDDs in leg 5 relative to legs 1 and 2. The following section provides an analysis of where SCVVFs may have acted to enhanced hydrometeor growth and the subsequent evolution of cloud downwind.

Semi-Coherent Vertical Velocity Fluctuations 270
The primary structural difference within the elevated cellular cloud layer across these three legs, which appeared responsible for cloud microphysical characteristics and SCDD development, were the presence and vertical location of layers of SCVVFs. A train of these velocity fluctuations were sampled at flight level during leg 1 from 24 to 35 km downwind of PJ (Fig. 5). The SCVVFs appeared as a series of ± 0.5 m s -1 vertical velocity perturbations with a wavelength of roughly 1 to 2 km (Fig. 5b). The vertical velocity fluctuations drove both a thermodynamic (Fig. 5e) and microphysical response (Fig. 5c  275 and d), which saw positive perturbation vertical velocities paired with lower temperatures, higher cloud droplet number, and lower CWC relative to the mean trend. Appreciable drizzle mass was only present in the perturbation downdrafts (Fig. 5c, pink curve).
From size distributions averaged across individual perturbation up-and downdrafts (Fig. 6), it is apparent that secondary droplet activation was primarily responsible for the increased droplet number concentration within perturbation updrafts. 280 DSDs corresponding to perturbation updrafts show that much of the increased droplet number concentration can be explained by an increased number of 6 to 8 μm droplets, which are an order of magnitude more abundant than within the perturbation downdrafts and nearly as abundant as the number of droplets in the primary mode from 25 to 35 μm. Given that these legs were flown at a constant altitude, the secondary droplet activation in perturbation updrafts, paired with a reduction in the CWC, may indicate kinetically limited parcel behavior and is examined in the discussion. The perturbation downdrafts 285 contained increased DWC, larger droplets, and lower total number concentration relative to perturbation updrafts. The decreased number and increased DWC are likely explained by scavenging by the larger drops, which were as large as 150 μm (Fig. 6), and indicate an active collision-coalescence process. Furthermore, collision-coalescence likely began very near or just above flight level, as the reflectivity values were between -25 and -15 dBZ within the nearest 400 m above flight level are indicative of populations of cloud droplets with very few, if any, drizzle drops (Fig. 5a). 290 Spatiotemporal cross-sections of Doppler velocity (Fig. 7) highlight the difficulty in identifying layers of SCVVFs away from the aircraft using the WCR. During leg 1, from 25 to 30 km downwind of PJ, a region where in situ measurements indicate a regular perturbation velocity pattern with 1 to 2 km spacing (Fig. 5b), there appears no similar Doppler velocity pattern from the WCR within the nearest few hundred meters of flight level (Fig. 7a). However, within 200 m of cloud top, from 30 to 35 km downwind of PJ, a clear train of vertical velocity fluctuations can be seen (Fig. 7a). These Doppler 295 velocity fluctuations match the crests of the wavelike reflectivity structures near cloud top in the corresponding reflectivity profile (Fig. 5a, top circled), but do not extend as far downward into cloud as the reflectivity structures. This perturbation velocity pattern is clearest in the highest 200 m of cloud, presumably due to the smaller sizes and resulting lower terminal velocities of scatterers there. In regions lower in cloud, the radar volumes contain more and larger drizzle drops and the resulting Doppler velocities become gradually more negative, eventually dominating the overall Doppler velocity pattern. 300 Very near flight level, it is possible to estimate the hydrometeor terminal fall speed by subtracting the in situ measured air velocity from the WCR measured Doppler velocity in the nearest range gates. Near flight level, 29 km downwind of PJ, we note an increase in hydrometeor terminal velocity (Fig. 7b, red and blue lines). This matches well with increases in DWC and DSD mean volume diameter beginning at nearly the same location illustrated in Fig. 5 c and e.

The link between SCVVFs and hydrometeor growth is also apparent in Contoured Frequency by Altitude Diagrams 305
(CFADs) generated from WCR radar reflectivity measurements. For the region in leg 1 corresponding to the sampled SCVVF train at flight level, 25 to 30 km downwind of PJ (Fig. 8a), the median reflectivity rapidly increased from a roughly constant -25 dBZ above 5 km MSL (~500 m above flight level) to greater than -15 dBZ just below flight level, suggesting rapid growth from cloud droplets to drizzle drop sizes for the low number concentrations observed in these clouds. This increase was characterized by a roughly -20 dBZ km -1 slope in the reflectivity CFAD which appeared consistently within 310 layers of SCVVFs elsewhere in cloud this day. For example, in a layer of SCVVFs near cloud top at 6 km MSL, located at 30 to 35 km downwind of PJ, a similar reflectivity slope with altitude is measured (Fig. 8b). The reflectivity enhancement tied to both of these layers of SCVVFs is discrete, in comparison to the more gradual growth (roughly -7 dBZ km -1 ) that occurred further downwind on this leg, starting at cloud top and extending through the entire cloud layer (Fig. 8c).
The impact that SCVVF layers had on the broader microphysical character of cloud during leg 1 was a trend of increasing 315 hydrometeor size with distance downwind. At the location of broad 0.5 to 1 m s -1 updraft 20 to 25 km downwind of PJ (Fig.   5b), the DSD contained mostly cloud droplets with diameters less than 40 μm (Fig. 9a, red). In the region of SCVVFs at flight level 25 to 35 km downwind of PJ, the diameter of the cloud droplet mode shifts to larger sizes and the steep exponential tail flattens out into a drizzle shoulder (Fig. 9a, green and blue). Even further downwind (Fig. 9a, orange and purple), a mature drizzle shoulder (100 µm < D < 300 µm) becomes apparent. Here, downwind of the SCVVFs at flight 320 level, the sampled drizzle originates from the layer near cloud top.
Observations from flight leg 2 indicate that the SCVVF layers present in leg 1 had broken down into incoherent turbulence.
A prominent drizzle precipitation plume was present from 45 to 53 km downwind of PJ, capped by a turbulent and variable cloud top height (circled, Fig. 10a). Still present were juxtaposed perturbation updrafts and downdrafts, especially near cloud Within the drizzle plume clearly evident in the reflectivity field (Fig. 10a), in situ measurements revealed DWCs in excess of 0.4 g m -3 (Fig. 10d). While several short wavelength perturbations appeared in the flight level vertical velocity profile (Fig.   10c), there did not appear a consistent correlation for either the thermodynamic (Fig. 10e) or the bulk microphysical measurements (Fig. 10d), unlike leg 1.
Leg 5, by comparison, contained a long and shallow layer of SCVVFs located 12 to 33 km downwind of PJ between 4.5 and 330 4.8 km MSL, about 500 to 1000 m below cloud top (Fig. 11a, circled) and just above the flight level. The horizontal scale of these fluctuations was smaller than in leg 1, with the width of a complete up/down perturbation couplet less than 1 km (Fig.   11b). Perhaps because of both the thinness of the SCVVF layer and its nearness to flight level, drops were much smaller compared to those observed in leg 1. The DSD mean volume diameter remained below 45 μm ( Table 2) and size distributions at flight level just below these SCVVFs reveal significantly lower concentration of drizzle drops with D > 100 335 μm compared to those observed in both legs 1 and 2 (Fig. 9c). Unlike observations in legs 1 and 2 however, there did appear a relatively even partitioning of mass distribution between CWC and DWC (Fig. 11d). Also, the presence of ice was corroborated by 2DS probe images (not shown) indicating that any vertical reflectivity enhancements from layers of SCVVFs for this leg are complicated by the increased linear growth rates (and hence reflectivity response) of ice in a mixed phase environment. 340 Reflectivity and Doppler velocity CFADs for three 5 km-wide drizzling columns from legs 1, 2, and 5 were generated for comparison (Fig. 12). The incoherent turbulence at cloud top for leg 2, seen in the large spread of Doppler velocities in the highest 1 km of cloud (Fig. 12e), produced a similar vertical reflectivity enhancement pattern as in the eastern end of leg 1 (Fig. 8c), where reflectivity gradually increases with distance downward through the elevated cellular layer. This pattern also appears in drizzling marine stratocumulus clouds where drizzle production typically occurs at cloud top and drizzle drops 345 grow throughout the entire cloud layer (e.g. Comstock et al., 2007). The broadening processes associated with incoherent turbulence and entrainment at cloud top are sufficient for drizzle production and subsequent accretional growth through the whole cloud layer. By comparison, the thin embedded layer of SCVVFs present in leg 5 led to a shallow growth layer with larger reflectivity-altitude gradients (i.e. more horizontal slope in the thinner shaded growth region; Fig. 12g) than in either legs 1 or 2. The larger ice particles present in the tail of the corresponding size distribution for the column from leg 5 (Fig.  350 9c) explain the similar median radar reflectivity up to 0 dBZ at flight level observed in legs 2 and 5 ( Fig. 12 d and g) despite the comparatively smaller, more numerous drizzle drops in leg 5 compared to legs 1 and 2. All three drizzling columns contained reverse S correlation patterns between reflectivity and Doppler velocity in the vertical, associated with hydrometeor growth and fallout over the layer (Fig. 12 c, f, and i).
Much of the previous work describing SCDD development in orographic, mixed phase cloud systems focused on the necessary conditions for development-namely the low cloud droplet and ice number concentrations coupled with condensate supply rates sufficient to support condensational growth to the droplet sizes required for active collisioncoalescence (Rauber, 1992;Ikeda et al., 2007). Several other studies suggest conditions which may be responsible for accelerated drizzle development or for relaxing these necessary conditions, introducing broadening mechanisms important 360 for SCDD production in cloud (Pobanz et al., 1994;Korolev and Isaac, 2000). Of these, the relationship between fine wind shear levels, spatial supersaturation fluctuations, and SCDD development has yet to be connected mechanistically by in situ measurements, despite being identified both as associated with SCDD development (Pobanz et al., 1994) and, separately, as important for the spectral broadening seen in certain layer clouds (Cooper, 1989;Korolev, 1995;Korolev and Mazin, 1993).
The observations here seem an important continuation of the work by Pobanz et al. (1994), which called for further airborne 365 research investigating the link between layers of strong wind shear and SCDD development. While their explanation called for observations of K-H billows to understand the production mechanisms, the microphysical behavior in layers of SCVVFs here seems to provide similar insight towards understanding these mechanisms.

Microphysical Response to SCVVF Layers
The insight provided from sampling one of these SCVVF trains with the in situ cloud hydrometeor probes (Fig. 5) allows for 370 some characterization of the microphysical processes in clouds of this type. Based on the flight level measurements, a conceptual model is presented to consistently describe the microphysical response to SCVVFs (Fig. 13). The kinematic structure and LWC response for leg 1 saw positive (negative) perturbation updrafts (downdrafts) paired with negative (positive) CWC perturbations from the trend and positive (negative) cloud droplet number concentration perturbations associated with droplet activation (evaporation). For these regular vertical velocity fluctuations in clouds with sufficiently 375 low concentrations of cloud droplets, the supersaturation response to vertical velocity fluctuations as described by Korolev (1995), is responsible for (re-)activating interstitial CCN as small (6 to 8 µm) droplets in the sub-adiabatic perturbation updrafts and separately broadening the primary droplet mode from repeated supersaturation fluctuations. Sub-adiabatic implies LWC values below what is expected from the adiabatic LWC formulation, where ГLWC represents the adiabatic lapse rate of liquid water content determined by cloud base temperature and pressure 380 (Albrecht et al., 1990) and z -zCB is the height above cloud base. The mean CWC for the SCVVFs sampled at flight level was 0.25 g m -3 with regularly spaced oscillations ± 0.05 to 0.08 g m -3 about that mean (Fig. 5c).
In a well-mixed (i.e. nearly constant equivalent potential temperature; Fig. 2), non-precipitating orographic layer cloud, the adiabatically-constrained CWC is expected to remain nearly constant at a given altitude, with only small perturbations that are the result of variations in cloud base thermodynamic conditions. Back of the envelope calculations estimate the specific 385 adiabatic CWC lapse rate of this elevated cellular layer cloud is about 0.001 g m -4 , taking the thermodynamic conditions from the sounding at the interface between the orographic and elevated cellular layers as a pseudo cloud base for this upper layer. Given mean CWC of 0.25 g m -3 observed at flight level, this indicates roughly 250 m of ascent for the cloud parcels sampled at this altitude. Variations of ±5 °C at cloud base would then correspond to ±0.05 g m -3 perturbations in CWC, and variations of ±50 mb would correspond to ±0.01 g m -3 perturbations, respectively. While the orographic environment does 390 predispose clouds to experience more variation in cloud base conditions than similar layer clouds associated with fronts or boundary layers, cloud base thermodynamic variations of this magnitude are not expected over spatial scales of 0.5 to 2 km and are therefore insufficient to explain the regular CWC perturbation observed. Instead, the perturbations of up to 40% of the mean CWC at a constant altitude were likely the result of dynamic and/or precipitation processes that were tied to the

SCVVFs. 395
The primary effect on CWC if only condensational effects are considered and where drizzle is not falling through parcels from above may be explained due to the kinetic effect as described by Korolev (1995). The negative CWC perturbations in leg 1 were accompanied by local supersaturation sufficient for secondary droplet activation (i.e. saturation ratio large enough to activate interstitial CCN), inferred from the presence of small droplets (6 to 8 µm) within perturbation updrafts (Fig. 6a, red and blue curves). Such sub-adiabatic behavior is linked to the kinetic limitation on condensational growth. As noted 400 earlier, cloud droplet number concentrations were less than 30 cm -3 and the "condensational inertia" of droplet populations to condense excess water vapor supply governed the supersaturation response, associated CWC response, and secondary droplet activation behavior. For the droplet populations less than 30 cm -3 and mean count diameter of between 20 and 30 µm, the corresponding phase relaxation time is around 10 s (using estimation methodology by Fukuta and Walter, 1970;Polotivitch and Cooper, 1988;and Korolev, 1995). This phase relaxation time corresponds to expected perturbations from 405 the adiabatic mean of as much as 0.02 g m -3 at flight level, indicating that, while the kinetic effect cannot fully explain the perturbation magnitude in the CWC field, it acts in the proper observed direction and explains the primary adiabatic (i.e. closed parcel) effect in these clouds. This zero-lag anticorrelation between vertical velocity and CWC perturbations results in the spatial pattern illustrated in Fig. 13.
The remaining magnitude of CWC variation is likely related to the precipitation dynamics. Removal of cloud water by 410 scavenging from drizzle in perturbation updrafts would lead to lower CWCs and reduced cloud droplet number. While the lower CWCs are indeed observed, and this may account for the greater magnitude reduction expected from the kineticadiabatic model alone, cloud droplet number concentrations increase. However, the increase in activation due to the kinetic limitation as noted previously is likely greater than the reduction in number concentration due to scavenging. Within the interspersed perturbation downdrafts, greater DWC and larger drizzle drops are observed, indicating active collision-415 coalescence. CWC and cloud droplet number concentrations are therefore expected to be further depressed relative to the mean than expected from the kinetic condensational effect alone. These regions are the likely origin of drizzle fall streaks observed in the WCR profiles and are represented by slightly larger drizzle drops in the downdraft region in Fig. 13.

Reflectivity-Inferred Hydrometeor Growth in SCVVF Layers
Comparisons between vertical reflectivity, Doppler velocity, and their cross correlation suggest two main microphysical 420 behaviors within layers of SCVVFs. The first is rapid, and often discrete, drop growth in the vertical tied to layers of vertical velocity fluctuations and not confined to cloud top. This vertical growth rate appears as large for these SCVVF layers in leg 1 as for the drizzle production at cloud top in leg 2. The second behavior is a reverse S cross correlation pattern (cf. Vali et al., 1998) in layers of SCVVFs, irrespective of hydrometeor phase differences, which further corroborates the local hydrometeor growth and fallout tied to these layers. 425 Layers of SCVVFs in legs 1 and 5 were responsible for vertical reflectivity enhancements similar in magnitude, roughly -20 dBZ km -1 , as produced by the drizzling cloud in leg 2 where layers of SCVVFs were not present. However, these SCVVF layers, especially in the relatively upwind cloud elements closer to PJ, were responsible for discrete growth layers that did not begin at cloud top. This indicates that the vertical velocity fluctuations were likely responsible for the initiation of collision-coalescence and drizzle production which occurred earlier and at a different location in cloud compared to the 430 classic idea of production at cloud top. Further downwind, corresponding to later in time from the upwind edge, drizzle production and growth did occur at cloud top and subsequent growth of the SCDDs occurred through the depth of the SLW layer, even without the presence of SCVVFs. This was most apparent in the transition between legs 1 and 2 from discrete growth at the level of these SCVVFs to growth over the entire layer, starting at cloud top, in leg 2. While qualitative, this observation suggests the importance of SCVVFs in other layered liquid clouds where embedded shear or shallow layers of 435 static instabilities may be responsible for enhancing the collision-coalescence process. Layers of SCVVFs may also be important in clouds where condensational growth and cloud top spectral broadening occur too slowly for active warm rain production, although with the caveat that any condensational kinetic effects are bound to be smaller than reported here. This, however, agrees with the observations of both Pobanz et al. (1994) and Korolev and Isaac (2005).
A distinct feature of the layers of SCVVFs is the bimodal DSD with populations of large (D > 30 µm) and small (D < 10 440 µm) droplets of similar number, not present elsewhere in cloud. This small droplet mode contains much less mass compared to the large droplet mode, and collisions between the large and small droplets are likely inefficient (E ~ 1 to 3% for drops of these sizes in laminar flow; Rogers and Yau, 1996), but the effect of such numerous possible collision events, especially given the large fall speed separations, in a turbulent environment may be enough to break the colloidal stability of the narrow large drop mode for a few lucky drops such that subsequent self-collection within this mode becomes favored. Furthermore, 445 parcel model results (Korolev 1995) have shown repeated supersaturation variations driven by vertical velocity fluctuations produce a local broadening about the larger droplet mode. This broadening may provide enough fall speed separation for self-collection without the need for larger droplets to physically interact with the newly activated smaller droplets. The increases in drop size and drizzle mass with distance downwind within SCVVF layers where parcels have undergone repeated supersaturation fluctuations are in qualitative agreement with this hypothesis. 450 A reverse S cross correlation pattern between reflectivity and Doppler velocity with altitude across these SCVVF layers further corroborates the drop growth in these layers. Vali et. al (1998) demonstrated this pattern in drizzling coastal stratus as the result of upward transport of drizzle and dilution of downward moving parcels near cloud top (region of positive correlation) which transitioned to the dominance of precipitation terminal fall speed increases below (region of negative correlation). Here the same trend is present in leg 5 (Fig. 12g), where the very low background reflectivities (-25 dBZ) above 455 the growth layer transition to rapid reflectivity increases below 5 km MSL correlated with positive Doppler velocities (Fig.   12i). As the Doppler velocities become more negative below this layer (Fig. 12h), the pattern reverses to the falling drizzle (and ice) dominating the reflectivity signature-with strong anticorrelation between reflectivity and Doppler velocity. This strong anticorrelation is dominated by the terminal fall speed-size relationship (e.g. terminal fall speed is proportional to the square of the diameter for drizzle drops). At the top of the growth layer, where weaker positive correlation exists between 460 reflectivity and Doppler velocity, it is important to consider both the contribution of hydrometeor terminal velocity and air motion to the observed Doppler velocities. For the populations just above the growth layer, terminal velocities for the largest cloud droplets are much less than the magnitude of the vertical velocity perturbations (±0.5 to 1.0 m s -1 ) and therefore the Doppler velocity signal is dominated by air motions. This suggests that the regions of upward relative air motion are correlated with higher reflectivities near the top of SCVVF layers, though without in situ measurements nearer the top of 465 these layers it is impossible to determine whether this is due primarily size or concentration. A more expansive conceptual model (cf. Fig. 13) would incorporate the vertical gradient of these growth and fallout effects across the SCVVF layer but is too conjectural without more penetrations through SCVVF trains at different altitudes.

Conclusions
Low cloud droplet number concentrations, less than 30 cm -3 , and precipitation-sized ice number concentrations, less than 470 0.5 L -1 , despite cold cloud top temperatures (T < -30 °C), provided favorable conditions for the development of SCDDs in a postfrontal orographic layer cloud forming over the Sawtooth Mountains in the American Intermountain West. This cloud, while transient and variable in vertical location and depth, consistently was strongest over the prominent terrain features downwind of Packer John mountain, and frequently contained layers of SCVVFs. Where present, SCVVFs were associated with local enhancement of the development and growth of SCDDs in response to the kinematic perturbation pattern. This 475 was demonstrated by strong vertical enhancements in CFADs of reflectivity, on the order of -20 dBZ km -1 , and attributed to hydrometeor growth through collision-coalescence. This drizzle production and growth occurred embedded within cloud and over relatively shallow layers before transitioning to drizzle production at cloud top and growth over the entire elevated cellular layer cloud. Compared to quiescent clouds, those containing SCVVFs will have more active DSD broadening processes and larger CWC gradients coincident with regions of probable turbulent mixing. This appears to explain the 480 observation that initial SCDD production can be enhanced by SCVVF layers and can lead to SCDD production in vertical regions other than just cloud top.

Author Contributions
AM performed the analysis and prepared the manuscript. JRF contributed to interpretation of results and provided critical edits in preparing the manuscript. 485