1. Introduction
Cold pools of air form beneath precipitating convection by evaporation, melting, and sublimation of hydrometeors within convective downdrafts. Upon reaching the surface, the cooler, denser air spreads out radially from the downdraft, analogous to a density current (Simpson 1969; Zipser 1977), although its heterogeneity and expansion rate may deviate substantially from such an idealization (Borque et al. 2020).
Convective cold pools can enhance, suppress, and/or organize subsequent shallow and deep convection. The leading edge of the cold pool (commonly referred to as the “gust front” or “outflow boundary” due to its gusty winds) from a precipitating cloud is a favorable location for initiation of additional (“secondary”) convection, by mechanical lifting of potentially buoyant environmental air to its level of free convection (e.g., Droegemeier and Wilhelmson 1985; Wilson and Schreiber 1986; Tompkins 2001; Böing et al. 2012; Schlemmer and Hohenegger 2014; Torri et al. 2015). A new generation of storms initiated by the cold pools from their predecessors can result in “upscale growth” of convection, from single-cell to multicellular convective systems (e.g., Coniglio et al. 2010). Behind the gust front, stabilization of near-surface lapse rates and enhanced subsidence within the cold pool can suppress the initiation of new convection (e.g., Tompkins 2001; Trapp and Woznicki 2017). In numerical simulations, when cold pools are suppressed by eliminating the evaporation of precipitation, the convection remains shallow, even in the presence of high CAPE and marginal CIN that would normally encourage deep convection (Khairoutdinov and Randall 2006; Böing et al. 2012), and substantially reduces smaller cumulus and cumulus congestus (Khairoutdinov et al. 2009). Thus, in numerical weather prediction and climate models where neither the convection nor their cold pools can be explicitly represented due to inadequate resolution, some parameterized representation of cold pools and their potential for forcing or suppression of additional convection is necessary.
The known dependencies of cold pools upon their parent convection have been used to implement their effects in parameterizations used within large-scale weather or climate models (Grandpeix and Lafore 2010; Park 2014; Del Genio et al. 2015; Rio et al. 2009; Rooney et al. 2022). Current schemes within convective parameterizations initiate cold pools when sufficient evaporation or cooling within unsaturated downdrafts occurs at or near the lowest model level; the parameterized cold pools are typically represented as homogeneous circular areas that evolve as pure density currents, with the speed, depth, and thermodynamic properties of the cold pool prognosed by the downward convective mass flux.
Despite the latent cooling from hydrometeor phase changes in convective downdrafts being the origin of cold pools, current cold pool parameterizations omit any representation of the contributing microphysical processes, including aerosol effects. It has long been recognized from theoretical, observational, and numerical modeling that increased CCN number concentrations slow the warm rain process by delaying the onset of collision and coalescence, and as a result delaying or even preventing rainfall from this process (e.g., Alpert 1955; Squires 1952a,b, 1958a,b). In deeper clouds and storms with tops that extend above the 0°C isotherm, raindrops produced by the warm rain process in the bottom of the storm can be carried to the upper, colder regions of the cloud. These supercooled raindrops then assist in the production of frozen hydrometeors (ice, snow, graupel, hail) by freezing, riming after freezing, and secondary ice production (e.g., Koenig 1963; Braham 1964; Chisnell and Latham 1976; Lasher-Trapp et al. 2018). Mallinson and Lasher-Trapp (2019) analyzed idealized numerical simulations of multiple realizations of a mesoscale convective system and its attendant cold pool. While artificially perturbing various aspects of the microphysical scheme to investigate the sensitivity of the cold pool characteristics to the latent cooling attributable to different hydrometeor types, they found that the cold pool was delayed by over 2 h by increasing initial CCN concentrations from 350 to 1700 cm−3 in the simulations. These results suggest that CCN effects on the timing of cold pool initiation are worthy of additional study.
The effects of CCN upon precipitation in mixed-phase clouds has been studied mainly with numerical simulations but has often produced conflicting results, due to the myriad of microphysical pathways that precipitation can be created in such clouds, and also due to differences in the dynamics of different storm morphologies that are intimately linked to environmental factors other than CCN (e.g., Khain 2009; Kalina et al. 2014). The sensitivity of the precipitation produced by deep convection to CCN has been found to vary with CAPE (e.g., Storer et al. 2010; Kalina et al. 2014), vertical wind shear (e.g., Fan et al. 2009; Kalina et al. 2014; Lebo and Morrison 2014; Lasher-Trapp et al. 2018), and environmental relative humidity (Tao et al. 2007; Kalina et al. 2014; Grant and van den Heever 2015). Kalina et al. (2014) performed suites of numerical simulations over an extremely wide range of CCN (100–10 000 cm−3) for four very different idealized storm environments. They found that the storm environment was much more deterministic upon the cold pool area and strength (coldness) than CCN. However, within a given environment, their simulated cold pools did show some sensitivity to CCN, mostly for values less than 1000 cm−3. The sensitivity to CCN has also been found to be significantly decreased in simulated supercell thunderstorms (e.g., Seifert and Beheng 2006; Storer et al. 2010; Loftus and Cotton 2014; Grant and van den Heever 2015) compared to other storm types. Given evidence that local aerosol concentrations and environmental variables can be intimately connected (e.g., Stevens and Feingold 2009; Varble 2018; Veals et al. 2022), CCN effects may be less important to precipitation production than the concomitant environmental effects on storm strength, but no study has investigated if the timing of cold pool initiation is influenced by CCN.
Measurements of CCN in the environments of deep convection have been rare, and no observational studies have investigated their possible effects upon convective cold pools. Recently, however, the Clouds, Aerosols, And Complex Terrain Interactions (CACTI; Varble et al. 2021) and Remote sensing of Electrification, Lightning, And Mesoscale/microscale Processes with Adaptive Ground Observations (RELAMPAGO; Nesbitt et al. 2021) field campaigns, based in Argentina in the austral summer of 2018/19, provided datasets that include airborne measurements of CCN, ground-based measurements of deep convection and its cold pools, and environmental soundings. These datasets provide an opportunity to investigate if CCN influences on cold pools can be witnessed in nature.
The overarching hypothesis to be tested in this study is that increased CCN in the storm environment delays the onset of surface rainfall, and thus delays cold pool initiation in land-based deep convection. To address the study hypothesis, observations of shallow and deep convection are first used to test the relationships between the observed CCN, surface precipitation, and timing of cold pool initiation from deep convection. Then, idealized numerical modeling of select cases is performed to further examine the CCN–cold pool relationship within the deep convection, as well as other cold pool characteristics like strength and depth, that could not be directly observed.
2. Observations and analysis methods
a. Observations from CACTI and RELAMPAGO
The Department of Energy CACTI field campaign (Varble et al. 2021) occurred from October 2018 through April 2019 near the Sierras de Córdoba mountain range in south-central Argentina. An intensive observing period (IOP) occurred from 1 November to 15 December, when an instrumented Gulfstream-1 research aircraft sampled state and microphysical variables within shallow clouds and their immediate environment, including aerosol sampling, augmenting the multiple ground-based radars and environmental soundings deployed over the entire field campaign. The CACTI IOP was coincident with the National Science Foundation RELAMPAGO field campaign (Nesbitt et al. 2021). Its focus was primarily on deep convection initiation, hailstorms, and upscale growth, and consisted of sonde launches, mobile radar data, and surface observations.
The CACTI aircraft data include the aerosol, microphysical, thermodynamic, and dynamical characteristics of shallow cumulus. These data are used to examine the CCN-warm rain relationship on days when shallow convection was sampled. For each flight, CCN measurements were made in the clear air above the cloud tops and below the cloud bases. In addition, cloud sampling at constant altitude flight legs at or near cloud base, at midcloud level, and near cloud top were flown.
Both CACTI and RELAMPAGO radar and surface observations of deep convection and cold pool characteristics are used here, as well as rawinsonde data characterizing storm environments. Four fixed-site radars are used to assess the initiation of deep convection and its cold pools. The CACTI radars included the C-Band scanning ARM Precipitation Radar 2 (CSAPR2), and the dual-mounted Ka-band and X-band Scanning ARM Cloud Radars (KaSACR and XSACR, respectively). The Colorado State University C-band Hydrological Instrument for Volumetric Observation radar (CSU-CHIVO; Arias et al. 2019) was also deployed as part of RELAMPAGO. The radar best suited for establishing the timing of deep convection initiation and/or cold pool initiation (due to its relative position and/or scans at critical times) is used for each case; times are cross-checked using data from any other radar as possible. Surface weather stations deployed during CACTI and RELAMPAGO, in addition to permanent stations within the Argentina Mesonet, verify the time and location of the cold pools. They measured state variables at 1 Hz, except for the Argentina Mesonet stations that provided data at 10-min intervals. Atmospheric soundings acquired from both field campaigns provide thermodynamic and dynamical context for the analyzed clouds and storms. Those closest in time and space to the deep convection are also used to initialize idealized numerical simulations.
b. CCN and the warm rain process in early shallow convection
CCN were measured during each CACTI flight using the airborne dual-column CCN Counter 200 (Roberts and Nenes 2005). It measures CCN activated at two different supersaturations; for this analysis, all CCN number concentrations refer to measurements made at the higher supersaturation, 1.12%. For two flights, measurements at 1.12% supersaturation were not taken (21 November and 4 December), so to enable comparison among all flights, those CCN number concentrations have been extrapolated to 1.12% supersaturation from the two measurements at lower supersaturations using a standard power-law relation (Twomey 1959) N = Csk, where N is the number of activated CCN (in cm−3), s is the supersaturation (in %), and C and k are constants dependent on the CCN properties in the air mass. The CCN measurements taken in clear air at a constant altitude near that of the shallow cumulus bases on each day were quantified over times ranging from 5 to 20 min.
Number concentrations of drizzle drops exceeding 50 μm diameter, those sizes easily capable of undergoing collision–coalescence, were measured using the 2-Dimensional Stereo Probe (2-DS; Lawson et al. 2006) at various altitudes within the clouds. The drizzle number concentrations within cloud updrafts exceeding 1 m s−1 are used to quantify the possible effectiveness of the warm rain process in the shallow convection sampled by the aircraft.
c. Analysis of deep convection initiation and cold pool onset
Of the nine CACTI research flights sampling early convection, only three cases developed into deep convection with well-observed cold pools suitable for meeting the study objective. Many other cold pools were observed during CACTI and RELAMPAGO, but these three storms and their first-generation cold pools were well-observed by radar and surface instrumentation. These cold pools were also not interacting with other cold pools.
Determining the cold pool initiation time (τCPI) begins as identifying a cold pool in surface meteorological data, and then tracing it backward in time in the radar data to when it was first discernible on radar. Cold pools are identified in surface station data by instances of sharp falling temperatures, shifts in wind direction, and increases in gustiness, signaling the passage of the gust front. The position of the gust front is then identified in the radial velocity patterns of RHI scans from the nearest radar to the storm, appearing as a shallow (0.5–1 km) volume of radial velocity perturbations (in either speed or direction), in contact with the ground, and extending from the precipitating convection. This signature is then tracked backward to its first appearance beneath the parent precipitating storm, defining τCPI. Any surface observation stations along the direction of the RHI scan are used to verify the position of the gust front.
Using Eq. (1) to calculate τCPO, its largest quantifiable uncertainty results from the time gap between subsequent radar scans: the timing of PPI scans used to estimate τCI and the timing of RHI scans used to estimate τCPI. Lower and upper bounds on τCPO are considered using the minimum and maximum time differences within which the criteria for τCI and τCPI are met.1
3. Results of observational analysis
a. CCN–drizzle relationship in shallow convection
For each of the nine CACTI research flights taking measurements within cumulus mediocris or cumulus congestus, median CCN number concentrations measured in clear air at altitudes near the cloud bases range from less than 400 to 1300 cm−3 (Table 1; Fig. 1). Spatial differences in the CCN measured on a given day mostly fluctuate about 100 cm−3. For the 5 December case, however, a low-level temperature inversion existed at the time of aircraft sampling. Above the inversion, the median CCN number concentration was 376 cm−3 (Fig. 1, lower box), but below the inversion, the median CCN number concentration was near 1200 cm−3 (upper box). The lower CCN number sampled above the inversion is used to relate to the warm rain process in the earlier shallow cumuli.
CACTI shallow cumulus cases analyzed in this study. The boldface font indicates the 3 days with well-observed cold pools.
Ten-second averages of CCN number concentration measured in clear air near shallow cumulus bases for each flight. Red bars denote median values measured at 1.12% supersaturation; magenta bars represent median values extrapolated to 1.12% supersaturation. Boxes encompass the 25th–75th percentiles; full data shown by extensions above and below. The upper set for 5 Dec denotes measurements beneath a temperature inversion as discussed in text. Aqua shading represents the three cold pool cases analyzed later.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0154.1
The warm rain process can be particularly sensitive to CCN number concentrations over this wide range (e.g., Hudson and Yum 2001), so the observed cumuli should show discernable variability in drizzle production. Vertical profiles of drizzle drop number concentrations within updrafts exceeding 1 m s−1 for all nine cases did exhibit large variability (Fig. 2), where maxima range from fewer than 10 L−1 to several 100 L−1. The cloud base median CCN number concentration exhibits a moderately strong correlation to the maximum drizzle observed at any altitude (Spearman correlation of −0.7). The correlation is not stronger because the cloud depth also plays a role.2 Those cumuli that are shallower, as indicated by the limited depth of aircraft sampling in Fig. 2, will have more difficulty producing raindrops due to insufficient liquid water content. Nonetheless, it is clear from this analysis that CCN number concentrations do have some effect on the warm rain process in these clouds.
Drizzle number concentrations in shallow cumulus updrafts vs altitude above their bases, arranged from lowest to highest median observed CCN number concentration for each flight, indicated in upper right of each panel. Vertical line at 100 L−1 in all panels is a reference to facilitate comparison of maximum drizzle concentrations.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0154.1
Because the aircraft cannot safely sample within the updrafts of deep convection, the amount of CCN ingested by storms on the 3 days with well-observed cold pools (boldface entries in Table 1) is unknown. Thus, the trends in the potential strength of the warm rain process from Fig. 2 are assumed valid in the updrafts of the deep convection on these days, except for the 5 December case. At the time of deep convection initiation on 5 December, soundings indicate that the temperature inversion had disappeared. Thus, the higher amount of CCN measured below the inversion in Fig. 1 is used as representative of CCN in the deep convection environment later that day. As indicated by the shaded box in Fig. 1, these three cases exhibit a wide range of ambient CCN concentrations that should allow for a test of their effects upon the timing of cold pool initiation.
b. Quantification of cold pool onset time in the deep convection
Radar and surface meteorological data from multiple sites are used to analyze the convection initiation and cold pool onset times for each of the three cold pool cases, and rawinsonde data quantify the storm environments (Fig. 3). For brevity, not all the data used are shown in the specific descriptions of the cases in the following subsections. The 4 December storm and cold pool were the best observed, having the most radar and surface observations for cold pool verification, and are thus presented first.
Maps of observational domain with positions of radars used for observing convection and cold pools (triangles), rawinsonde release sites (circles), and surface meteorological stations (squares) for (a) 29 Nov, (b) 4 Dec, and (c) 5 Dec. G-1 flight patterns flown in smaller cumuli before deep convection (dashed blue line) overlaid. Radar echoes at time of convection initiation τCI (given in text) also shown, with red box around storm of interest. For (a), radar echoes occurring at longitudes west of −64.85° have been masked for clarity.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0154.1
1) 4 December
At 1600 UTC, three discrete convective cells initiated on the ridge of the Sierras de Córdoba, approximately 80 km north of the CACTI radar site. Two of these cells (Fig. 3b) continued to strengthen, developing narrow reflectivity cores that exceeded 55 dBZ. Near 1700 UTC, the southern cell had grown into a convective cluster that expanded north and west, but also began to propagate southward along with the edge of a strong cold pool (Fig. 4a). The cold pool was later clearly visible as a sharp wind shift and a temperature drop greater than 5°C at the CSWR Pod L surface station (Fig. 4c), and as an area of in-bound velocities (red box in Fig. 4b) in the CSAPR2 RHI. Working backward in time through the CSU-CHIVO radar scans, the first detectable sign of the cold pool (τCPI) on 4 December is estimated as 1658 UTC. The convection initiation time (τCI) is estimated using the CSU-CHIVO radar as 1610 UTC, when the first contiguous five-pixel region of 30 dBZ echo is met within the cell that later intensified and spawned the cold pool (Fig. 3b).
Examples of observed 4 Dec storm and cold pool at times noted in each panel: (a) low-level radar reflectivity at 1720 UTC with locations of the CSAPR2 radar and CSWR Pod L surface station used in other panels noted; (b) CSAPR2 RHI scan of radar radial velocity along blue dotted line shown in (a), with radar located at lower-left corner and negative velocities indicating winds directed toward the radar, velocity perturbations indicative of the cold pool within the red box, and 20-dBZ radar reflectivity contour in black; and (c) time series from surface meteorological station CSWR Pod L [blue square shown in (a) and (b)] with time of cold pool passage denoted by the shaded region. Terrain heights plotted in (b) from the Shuttle Radar Topography Mission (SRTM) 30 m digital elevation model (Farr et al. 2007).
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0154.1
2) 5 December
Deep convective cells developed on this day, some being supercells as indicated by the presence of steady velocity couplets on PPI scans from the CSU CHIVO radar (not shown). The convection initiation time (τCI) for the supercell storm of interest is estimated using the CSU CHIVO radar as 1620 UTC (Fig. 3c). The storm intensified rapidly by 1648 UTC, with reflectivity increasing above 60 dBZ throughout most of the core and extending to the ground. A zone of inbound radial velocities is very clear in RHIs at 1756 UTC (Fig. 5b). Tracing the radar signature backward in time, the upper limit of τCPI is estimated to be 1658 UTC. The convection strengthened rapidly over the course of the next hour as it advanced eastward down the terrain, and various meteorological surface stations recorded progressively stronger temperature and wind deviations, with a nearly 10°C temperature drop and 200° wind shift occurring at Pod O (Fig. 5c) as the cold pool passed.
As in Fig. 4, but for the 5 Dec cold pool case. The storm producing the cold pool is the southern cell located near −32° latitude in (a).
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0154.1
3) 29 November
By 1630 UTC, multiple small convective cells had initiated 20–30 km west of the CACTI radar site. The strongest of these cells initiated a shallow and short-lived cold pool that advanced a short distance south (Fig. 6a), overtaking the AMF1 surface instrumentation at the CACTI radar site (Fig. 6c), where falling temperatures, as well as a wind shift from northeast to north, helped to confirm the radar radial velocity feature (Fig. 6b) as the leading edge of a cold pool. When traced backward in time, the velocity feature is first observed on the 1648 UTC XSACR RHI scan, setting τCPI. The value of τCI for the storm occurs at 1630 UTC based on the CSAPR2 radar data (Fig. 3a).
As in Figs. 4 and 5, but for the 29 Nov cold pool case. (b) A hemispheric RHI scan that was performed while the storm was directly overhead of the radar.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0154.1
c. Calculation and comparison of cold pool onset with CCN
The time of cold pool onset (τCPO) is then calculated for each case using Eq. (1). Although establishing a trend with only three cases is difficult, the estimated τCPO does not increase ubiquitously with increasing CCN as hypothesized (Fig. 7). The 29 November case, having the fewest CCN, did initiate its cold pool 30 min earlier than that of 4 December that had more CCN in its environment. But the 5 December cold pool onset time is nearly the same as the 29 November case, despite having approximately double the amount of environmental CCN. It is tempting to consider that if the 5 December CCN sampled earlier in the day above the inversion is instead used for the 5 December case (dashed red cross in Fig. 7), then the expected trend in CCN versus τCPO would hold. Unfortunately, the observations cannot establish the amount of CCN ingested by the storm.
Plot of τCPO vs ambient CCN median number concentration for each case. Range of CCN for each day along x axis represents the interquartile range. Range of τCPO for each day along y axis represents possible range of values due to times between successive radar scans for determining τCI and τCPI. Dashed line for 4 Dec denotes that the CCN measurements had to be extrapolated to 1.2% supersaturation on this day. Dashed red cross represents results for the 5 Dec case if the lower CCN value was assumed to have been ingested by the storm.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0154.1
d. Other environmental factors
The storm environment can exert a significant influence upon storm morphology and development, which can also affect the rate at which the cold pool develops. Three environmental characteristics relevant to storm morphology are shown in Table 2. None of these parameters alone is strongly related to τCPO. However, their combined influence undoubtedly governed the convective morphology: the 29 November and 4 December cases were “ordinary” buoyancy-driven thunderstorms, while the stronger shear of the 5 December environment allowed a supercell thunderstorm. Supercells, driven primarily by dynamic pressure perturbations, have been shown to be more resistant to CCN effects on their precipitation than ordinary thunderstorms in numerical simulations (e.g., Storer et al. 2010; Loftus and Cotton 2014; Grant and van den Heever 2015).
Convective parameters of the observed storm environments.
4. Experimental design of numerical simulations
The observational analysis motivates further study requiring numerical modeling, where the CCN ingested by each storm can be modified to eliminate the convolution of CCN and environmental effects present among the observed cold pool cases. Numerical modeling also allows an investigation of the important microphysical processes governing τCPI, impossible to measure directly and unambiguously.3 The sensitivities of other cold pool properties to CCN, including strength, expansion rate, and depth, which are not able to be quantified with confidence in the observations, can also be easily evaluated with the simulation results.
A suite of idealized 3D numerical simulations is conducted for the two best-observed cold pool cases, 4 and 5 December. This set includes one case that appeared to follow, and another to contradict, the hypothesized CCN effect on cold pool initiation time, as well as different storm morphologies. The 4 December storm also had a colder (∼1°C) and higher (3600 m MSL) base; the 5 December storm base was comparatively warmer (∼8°C), and lower (2350 m MSL).
a. Numerical model overview
Cloud Model 1 (CM1; Bryan and Fritsch 2002) version 19.6, is used for all simulations; it is a nonhydrostatic, 3D, idealized, cloud-resolving numerical model designed and often used for the analysis of deep precipitating convection. CM1 solves the fully compressible form of the Navier–Stokes equations on a staggered Arakawa C grid for the following variables: the three velocity components in the horizontal and vertical directions, nondimensional pressure, potential temperature, subgrid-scale turbulent kinetic energy, and mixing ratios of water vapor, liquid water, and ice.
All simulations use the NSSL two-moment microphysics scheme (Mansell et al. 2010; Mansell and Ziegler 2013) within CM1. This scheme predicts both mass and number concentration of cloud water, cloud ice, rain, snow, graupel and hail. The rime densities of graupel and hail are also predicted, allowing for more accurate representations of graupel and hail fall speeds. The NSSL scheme incorporates both the warm rain parameterization of Ziegler (1985) and, for this study, the ice nucleation parameterization of Phillips et al. (2001), which includes contact freezing (Meyers et al. 1992), rime splintering (Cotton et al. 1986), and immersion freezing of raindrops (Bigg 1953). During model integration, microphysical budgets are calculated and output for each contribution to latent cooling by evaporating rain, melting or sublimating graupel, and melting or sublimating hail, as in Mallinson and Lasher-Trapp (2019).
Coriolis forcing is included in the calculations, and Rayleigh damping is applied within 3 km of the domain top to limit wave reflection. Open radiative boundary conditions at horizontal boundaries, and a free-slip lower boundary, are also used. Radiative processes and terrain effects are neglected.4
b. Simulation setup
The storm environments for the simulations are initialized as horizontally homogeneous fields, using soundings taken during the RELAMPAGO field program, with some modifications (Fig. 8). Both soundings were launched into the inflow of the observed storms, within 30 min of τCI. Shallow superadiabatic layers in the lower levels of the soundings have been removed and replaced by dry adiabatic layers. Shallow cloudy layers in the soundings have also been removed by decreasing the dewpoint temperatures by 0.5°C for each data point where relative humidity exceeded 95%. Neither sounding recorded data up to the altitude of the top of the model domain, so the uppermost 100 mb of the soundings taken one hour earlier are spliced onto these soundings. The effects of these changes from the original soundings are minimal upon the MUCAPE and vertical wind shear in both environments, being within 50 J kg−1 and 1 m s−1, respectively, of the original values.
Model base-state conditions, modified from original soundings, as discussed in the text, from (a) 1600 UTC 4 Dec CSU mobile sounding and (b) 1704 UTC 5 Dec UIUC2 mobile sounding. Dashed teal line highlights 0°C isotherm; wind barbs plotted in m s−1.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0154.1
For all simulations, the model domain is 120 km in both horizontal directions, and 18 km deep. The horizontal and vertical model grid spacing is a constant 200 m. For the 5 December simulations, the base of the model grid corresponds to the lowest level of the 5 December sounding, 1030 m MSL. The 4 December sounding was released from higher terrain (1140 m MSL), and because the deep convection and its cold pool initiated on the mountain ridge 400 m above that site, the lowest 400 m of the sounding are removed, essentially “lifting” the base of the model grid up to 1540 m MSL, to be representative of the environment into which the cold pool initiated. The model integration for all simulations uses a 1-s time step, and results are output at 1-min intervals.
To force the storm in each environment, a surface heat flux is imposed near the center of the domain, following the formulation of Carpenter et al. (1998), where a Gaussian sensible heating maximum is applied, decreasing horizontally with a standard deviation σ and decaying exponentially in the vertical with an e-folding height α (Table 3). The heating function, initially 0 W m−2, increases linearly for 5 min, holds constant at a maximum value Hg for 24 min, and linearly decreases for 1 min to 0 W m−2, for a total heating period of 30 min. The specific combination of α, σ, and Hg for each simulation is chosen to best reproduce the storm base temperature (estimated from the height of the LCL in the sounding) and radar-observed storm widths and heights.
Details of Gaussian heating function used to initiate convection in simulations.
Each case is simulated multiple times changing only the number of CCN. The NSSL microphysics scheme allows users to prescribe the number of CCN activated at 1% supersaturation as an input variable at the start of the simulation, given by C in the same Twomey (1959) power-law relation used in section 2b. New cloud droplets are created according to the value of the supersaturation that is implicitly a function of the updraft speed. The value of C for the control simulations for 4 and 5 December is initialized with the aircraft-observed median CCN number concentration, 850 and 1200 cm−3, respectively; additional simulations are then conducted with fewer and more CCN as indicated in Table 4. Two additional simulations are performed to supply more overlap in the two cases: 4Dec_600 and 5Dec_850.
CCN parameters and naming conventions for the 4 Dec and 5 Dec deep convection simulations.
c. Defining cold pools
Cold pools are defined in all simulations as a contiguous area having a potential temperature perturbation (θ′) threshold of −1 K at the lowest model level (100 m) from the horizontally homogeneous environment. Experimentation showed that the −1 K threshold is best spatially correlated with horizontal velocity perturbations at the cold pool leading edge, analogous to the temperature, wind shift, and velocity perturbations used to define the cold pool leading edge in the observations. The cold pool depth is defined as the maximum height of the contiguous −1 K θ′surface exceeding 1 km2 in area and coincident with a downdraft weaker than 1 m s−1. The cold pool strength is evaluated as the minimum surface potential temperature perturbation. The cold pool expansion rate is approximated as the time rate of increase in the area at the lowest model level.
5. Numerical modeling results
a. Comparison of control simulations with observations
Control simulations for the 4 and 5 December cases generally agree well with CACTI-RELAMPAGO observations up to and including the initiation of the cold pool. The maximum simulated storm widths are within 2 km, and maximum storm heights are within 600 m, of those observed by radar. The respective storm types (ordinary thunderstorm on 4 December; supercell thunderstorm on 5 December) are also simulated in the model. The simulated storm base heights (as determined by the lowest altitude of the 10−3 g m−3 contour of cloud water) are 1700 and 900 m for the 4 and 5 December cases, respectively.
The 4 December CTRL simulation (Fig. 9, top row) initiates convection as a single ordinary cell, with the first 30 dBZ echo occurring at 39 min (τCI). The cold pool initiates at 71 min into the simulation (τCPI; Fig. 9d). The value of τCPO is therefore 32 min, 8 min faster than the shortest estimate from the observations.
Simulated column-maximum reflectivity (shaded) and cold pool (contoured) for (top) 4 Dec control simulation and (bottom) 5 Dec control simulation at times indicated above each column.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0154.1
The 5 December CTRL simulation (Fig. 9, bottom row) produces a supercell that splits into two storms with counterrotating updrafts between 50 and 60 min. All quantities in the analyses are calculated over both storms because they did not separate far spatially. The first 30 dBZ echo occurs at 36 min (τCI), and the cold pool initiates at 66 min (τCPI). The simulated value of τCPO is thus 30 min, only 3 min longer than the longest observed estimate. The model also reproduces a period of early rainfall observed before and during the period of storm splitting. In the simulated storm, this early rain is not sufficient to initiate the −1 K cold pool; however, it was coincident with the initiation of the cold pool in the observed storm.
b. Sensitivities of simulated cold pool characteristics to initial CCN concentration
Across the range of CCN number concentrations used for the 4 and 5 December simulations, updraft strengths (evaluated by the 50th and 90th percentiles and maximum values) do not deviate greatly during the analysis period (not shown). Storm strengths and general behavior are relatively similar as CCN are varied, allowing a detailed study of microphysical effects upon the cold pools. Therefore, only the initiation time of the cold pool (τCPI) is discussed from the simulation results, as τCI is essentially unchanged among the simulations.
The value of τCPI does show some sensitivity to the CCN number concentrations used in the simulations (Fig. 10). Increasing the CCN number concentrations by a factor of 8 between the 0.25CCN and 2CCN runs for both the 4 and 5 December cases results in a delay in τCPI by 26 and 27 min for 4 and 5 December, respectively. However, for the 5 December supercell, most of that sensitivity occurs across the four lowest CCN simulations, with barely any change when the CCN exceeded the CTRL simulation. For both 4 and 5 December simulations, considering the variability in measured CCN number concentrations observed over this dataset (shaded region in Fig. 10), the delay in the cold pool in either environment is not substantial, only about 15–20 min.
Scatterplot of CCN number concentrations vs τCPI for all 4 Dec (blue circles) and 5 Dec (red squares) model simulations. Black markers of similar shape denote values for CTRL simulations (using observed CCN), and shaded rectangle represents observed CCN variability across the CACTI flights presented in Table 1.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0154.1
Interestingly, other cold pool characteristics, such as cold pool strength, expansion rate, and depth, are more sensitive to CCN, at least for the 4 December “ordinary thunderstorm” case. Even over the factor-of-3 increase in CCN measured during CACTI IOPs (shaded rectangles in Fig. 11), the cold pool strength decreases by a factor of nearly 2, the expansion rate decreases by a factor of 3, and the cold pool depth decreases by over a factor of 2. For the 5 December supercell thunderstorm simulations, however, these trends are nearly absent (Figs. 11b,d,f), except for the 0.25CCN simulation that lies outside the range of CCN observed. The insensitivity of the cold pool to CCN is consistent with the supercell precipitation response found in past modeling studies. The 5 December cold pools overall are also less cold, expand more slowly, and are shallower than the 4 December cold pools, resulting in part from a lower cloud base and a moister boundary layer, decreasing the depth and amount of latent cooling, but also attributed later in section d to differences in storm dynamics.
As in Fig. 10, but for cold pool characteristics averaged over the first 60 min of its lifetime vs CCN for (a),(c),(e) 4 Dec and (b),(d),(f) 5 Dec simulations.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0154.1
c. CCN influence on microphysics in updrafts
An investigation of the hydrometeors in the early updrafts preceding surface precipitation provides context for the differences in latent cooling discussed in the next section. In strong updrafts (here, exceeding 20 m s−1), increases in hydrometeor mass with increasing altitude can approximate growth in time, for hydrometeors with much smaller fall speeds than the updraft speed. Increasing CCN in the 4 December simulations results in more supercooled cloud water extending to progressively higher altitudes, and less rain by the warm rain process, as expected. Simulations initialized with fewer CCN produce much more graupel and much less ice than those initialized with more CCN, as a result of the larger mass of supercooled rain available for immersion freezing to create graupel. In contrast, simulations with more CCN transport more supercooled cloud water above 4 km, where it converts to small ice (and in time, snow) by heterogeneous and homogeneous freezing, and thus can be carried higher in the updrafts, delaying surface precipitation. For the 5 December case, the vertical distribution of all hydrometeor classes within the early updrafts (before storm splitting) reveals similar trends with increasing CCN as the 4 December case. Due to the warmer storm bases in the 5 December case, the magnitudes of the hydrometeor masses are all greater in the 5 December early updrafts than in the 4 December early updrafts.
d. CCN influence on latent cooling and cold pools
Time–height plots of maxima in latent cooling due to evaporating rain, sublimating graupel, melting graupel, and melting hail (Fig. 12) show the microphysical chain of events leading to the timing of cold pool initiation in the simulations. They also provide some insight into why cold pool strengths and depths differ.
Time–height diagrams of (a) 4 Dec maximum latent cooling from evaporating rain and melting; (b) 4 Dec maximum latent cooling from graupel sublimation and melting hail; (c) as in (a), but for 5 Dec; and (d) as in (b), but for 5 Dec. Bold white contour marks boundary of −1 K cold pool. Fine white horizontal line denotes height of storm base; fine white dashed horizontal line denotes height of environmental 0°C isotherm. Contour interval for all panels is 1 × 109 J.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0154.1
Comparing across the CCN extremes of the 4 December simulations (Figs. 12a,b), the impact of reduced graupel production by increased CCN upon cold pools becomes clear. As CCN are increased, sublimating graupel at higher altitudes is reduced substantially (Fig. 12b), as is melting graupel at midlevels (Fig. 12a). The reduction in graupel in turn reduces evaporating rain at the lowest levels (Fig. 12a), culminating in delayed cold pools (shown by the first appearance at the ground of the −1 K perturbation). Decreases in evaporating rain also result in shallower (and weaker; cf. Fig. 11) cold pools as CCN are increased.
For the 5 December simulations, the relationship between the timing of the first surface precipitation (affected by CCN) and storm splitting controls cold pool initiation. As CCN are increased, graupel production again decreases (Figs. 12c,d), but not for CCN amounts beyond the CTRL simulation. The 0.25CCN simulated storm, like the 4 December simulations, produces abundant graupel that sublimates (Fig. 12d) and melts (Fig. 12c), producing a very deep, early cold pool in that simulation before the supercell splits (between 50 and 60 min).5 Due to the low number of CCN and warm storm base encouraging the warm rain process, evaporating rain is a major source of latent cooling throughout the storm up to 5 km altitude (Fig. 12c), unlike any of the 4 December simulations. This major source of evaporating rain above 2.5 km disappears as CCN are increased to the CTRL value and beyond, as expected from a less productive warm rain process. In the CTRL and 2CCN simulations, rain evaporation continues to occur near 2 km altitude (Fig. 12c), yet still above the cloud base, during the splitting of the supercell (between 50 and 60 min). This evaporating rain is insufficient to produce a −1 K thermal perturbation at the ground until later (right before 70 min) when larger quantities of melting graupel (Fig. 12c) augment it. The unique dynamics within a splitting supercell thunderstorm (with counterrotating updrafts) redistribute hydrometeors into different regions across the storm horizontally (not shown), altering where the earliest precipitation falls and thus its latent cooling. After 60 min when the storms have split, magnitudes of latent cooling from evaporating rain, melting graupel and even melting hail decrease to nearly similar values in the CTRL and 2CCN simulations, with maxima in evaporating rain becoming increasingly relegated to beneath the storm base. While CCN do somewhat control the speed of precipitation formation in the 5 December simulations, the precipitation delay for greater amounts of CCN allows the supercell dynamics to overpower its role on the cold pool timing and features.
6. Summary and discussion
The study hypothesis was that increased amounts of CCN in the storm environment can delay precipitation and thus the timing of the first cold pool produced by land-based deep convection. Analyses of observations from cases observed during the CACTI and RELAMPAGO field campaigns in Argentina in 2018, as well as numerical simulations of some of the cases, produced the following findings:
-
CCN and cloud observations from nine research flights showed that shallow cumuli ingesting more CCN contained fewer drizzle drops within their updrafts than those ingesting fewer CCN, although cloud depth also likely had an effect.
-
Three of those cases later produced deep precipitating convection with isolated, well-observed cold pools. Estimates of the timing of cold pool onset using radar and surface meteorological observations did not follow the hypothesized trend with CCN, although the one outlier has some uncertainty regarding the amount of CCN it ingested.
-
Environmental variables other than CCN (CAPE, wind shear, moisture) governed the morphology of the deep convection on these three days, including the supercell thunderstorm case that did not adhere to the study hypothesis.
-
Idealized numerical simulations of two deep convection cases did show the expected sensitivity of the cold pool initiation time to CCN number concentration, especially for the ordinary thunderstorm case, but much less so for the supercell thunderstorm case. However, even in the ordinary thunderstorm case, the delay in the cold pool was minimal as the environmental CCN were increased across the range of observed values.
-
Simulated cold pool strength, expansion rate, and depth all showed a marked decrease with increasing CCN in the simulated ordinary thunderstorm. On the other hand, the cold pool from the simulated supercell thunderstorm showed little sensitivity to CCN except at very low CCN values. This deviant behavior resulted from the overpowering effects of the unique supercell dynamics, including storm splitting and horizontal redistribution of hydrometeors.
-
Across all simulations, the primary microphysical pathway to surface rainfall was warm rain in the updrafts lofted above the 0°C isotherm, its conversion to graupel, and melting of graupel falling beneath the 0°C isotherm. Evaporative cooling of this falling rain coincided with cold pool initiation. Increasing CCN number concentrations decreased graupel production, delaying and decreasing surface rainfall, and in turn delaying and weakening the cold pool.
This study was motivated by the need to enhance general knowledge controlling cold pool properties to improve future cold pool parameterizations for large-scale weather and climate models. Environmental parameters have been shown to be strong determinants of cold pool properties in past studies, and the current study reinforces their importance, particularly in dictating storm morphology. Storm morphology was capable of obscuring CCN effects, specifically in the case of supercell thunderstorms, both in the simulations and as suggested by an observed case.
There are several important observational limitations on this study. The amount of CCN actually ingested by the observed storms is unknown; it had to be approximated from CCN observations acquired near the bases of much shallower cumulus occurring earlier in the day. Observations of the progression of the warm rain process within the updrafts of the observed storms also had to be inferred from aircraft measurements within the earlier shallow cumuli for each case. Until the atmospheric science community acquires a new storm-penetrating aircraft, which it has lacked for nearly two decades, understanding the microphysical details within the updrafts (or downdrafts) of deep convection will be extremely hampered, including a very pertinent question here regarding how many microphysical details can be predicted by ground-based aerosol measurements.
The limitations of the idealized numerical modeling used here are also important. Using idealized modeling was logical, to produce high-resolution simulations with easily quantified and easily output variables such as latent cooling rates from individual hydrometeor phase conversions and other microphysical-related quantities. However, the model did not include terrain, and the observed storms were likely generated as a result of its presence. Because the focus of this study was not upon convection initiation, nor on terrain effects on cold pools, the lack of terrain was deemed acceptable, but absolute characteristics such as simulated cold pool depth and expansion rate found here would undoubtedly be affected by terrain. A limited choice of microphysical schemes in the model made the use of a single two-moment scheme the only practical selection, as the other schemes lack separate categories for graupel and hail, and/or only simulate them with constant densities, and/or do not address the difficulty of two-moment schemes in artificially enhancing hydrometeor size-sorting, and/or do not allow user-input values of any specific CCN number concentration desired. Thus, comparisons with these other microphysical schemes would not have been meaningful. The microphysical scheme used here shares similar uncertainties in its representation of hydrometeor evaporation, sublimation, and melting rates as other state-of-the-art schemes used in other models, especially with respect to ventilation effects and variability in hydrometeor fall speeds. As microphysical schemes continue to improve, revisiting the results found here will be essential.
Based on only three observed cases, and idealized simulations of two of those cases, it is difficult yet to conclude how important CCN are for future cold pool parameterizations; more cases are needed, particularly observational-based studies like that conducted here. The CACTI and RELAMPAGO data provided an opportunity to attempt this kind of study, but the difficulty in having radars and surface meteorological sites in the right places for robust characterization of the cold pool properties remains. These initial results do encourage caution in universally using any cold pool parameterization based solely on data from either maritime or continental storms alone, as the microphysical processes that produce surface rainfall, and affect other cold pool characteristics, may differ considerably due to the differences in the CCN in those environments.
Acknowledgments.
This research was supported by the Department of Energy (Grant DE-SC0021042). The contributions of all CACTI and RELAMPAGO field campaign participants for their data collection efforts, and all DOE-ARM and NSF-funded personnel performing data processing, are gratefully acknowledged. The authorship and maintenance of the CM1 model by Dr. George Bryan of the National Center for Atmospheric Research, and of the NSSL microphysical scheme by Dr. Ted Mansell of the National Severe Storms Laboratory, are also greatly appreciated.
Data availability statement.
CACTI data are available at https://adc.arm.gov/discovery/#/results/site_code::cor/start_date::2018-10-01/end_date::2019-04-30. RELAMPAGO data are available at https://data.eol.ucar.edu/master_lists/generated/relampago/. The CM1 model is available at https://www2.mmm.ucar.edu/people/bryan/cm1/.
Footnotes
Minimum and maximum time gaps between successive PPI or RHI scans used for the uncertainty of τCPO are: 10 and 33 min (29 Nov), 38 and 58 min (4 Dec), and 8 and 27 min (5 Dec).
Actually, both cloud-base temperature and cloud depth play a role, but only the 4 Dec case was an outlier in cloud-base temperature, being colder by 6°C or more compared to other days.
Although dual-polarization radar data could potentially provide some information, the radar scanning strategies did not explicitly target downdrafts nor their attendant cold pools, precluding a useful time history of microphysical information.
CM1 does not include the capability to consider topography at the base of the model grid. The neglect of terrain in the simulations would mostly affect the location and timing of convection initiation and storm propagation in the model. Neither is particularly important for the present study. Neglecting terrain effects may slow or quicken the speed of movement of the simulated cold pool compared to that observed, but should not affect the timing of its initiation, which is the primary focus of this study.
It is interesting that the earlier cold pool formation in the 0.25CCN run is more consistent with the 5 Dec observed storm, where the surface precipitation and cold pool appeared before the storm split. This perhaps argues that the smaller observed CCN number is indeed representative of that ingested by the storm, which would make all observed cases adhere to the hypothesis (Fig. 7). It is impossible to determine from the observations how many CCN the storm ingested, unfortunately.
REFERENCES
Alpert, L., 1955: Notes on warm-cloud rainfall. Bull. Amer. Meteor. Soc., 36, 64–68, https://doi.org/10.1175/1520-0477-36.2.64.
Arias, I., V. Chandrasekar, and S. S. Joshil, 2019: Cross-validation of CSU-CHIVO radar and GPM during RELAMPGO. 2019 IEEE Int. Geoscience and Remote Sensing Symp. (IGARSS 2019), Yokohama, Japan, IEEE Geoscience and Remote Sensing Society/IGARSS 2019 Organizing Committee, 7586–7589.
Bigg, E. K., 1953: The formation of atmospheric ice crystals by the freezing of droplets. Quart. J. Roy. Meteor. Soc., 79, 510–519, https://doi.org/10.1002/qj.49707934207.
Böing, S. J., H. J. J. Jonker, A. P. Siebesma, and W. W. Grabowski, 2012: Influence of the subcloud layer on the development of a deep convective ensemble. J. Atmos. Sci., 69, 2682–2698, https://doi.org/10.1175/JAS-D-11-0317.1.
Borque, P., S. W. Nesbitt, R. J. Trapp, S. Lasher-Trapp, and M. Oue, 2020: Observational study of the thermodynamics and morphological characteristics of a midlatitude continental cold pool event. Mon. Wea. Rev., 148, 719–737, https://doi.org/10.1175/MWR-D-19-0068.1.
Braham, R. R., Jr., 1964: What is the role of ice in summer rain-showers? J. Atmos. Sci., 21, 640–645, https://doi.org/10.1175/1520-0469(1964)021<0640:WITROI>2.0.CO;2.
Bryan, G. H., and J. M. Fritsch, 2002: A benchmark simulation for moist nonhydrostatic numerical models. Mon. Wea. Rev., 130, 2917–2928, https://doi.org/10.1175/1520-0493(2002)130<2917:ABSFMN>2.0.CO;2.
Carpenter, R. L., Jr., K. K. Droegemeier, and A. M. Blyth, 1998: Entrainment and detrainment in numerically simulated cumulus congestus clouds. Part I: General results. J. Atmos. Sci., 55, 3417–3432, https://doi.org/10.1175/1520-0469(1998)055<3417:EADINS>2.0.CO;2.
Chisnell, R. F., and J. Latham, 1976: Ice particle multiplication in cumulus clouds. Quart. J. Roy. Meteor. Soc., 102, 133–156, https://doi.org/10.1002/qj.49710243111.
Coniglio, M. C., J. Y. Hwang, and D. J. Stensrud, 2010: Environmental factors in the upscale growth and longevity of MCSs derived from rapid update cycle analyses. Mon. Wea. Rev., 138, 3514–3539, https://doi.org/10.1175/2010MWR3233.1.
Cotton, W. R., G. J. Tripoli, R. M. Rauber, and E. A. Mulvihill, 1986: Numerical simulation of the effects of varying ice crystal nucleation rates and aggregation processes on orographic snowfall. J. Appl. Meteor. Climatol., 25, 1658–1680, https://doi.org/10.1175/1520-0450(1986)025<1658:NSOTEO>2.0.CO;2.
Del Genio, A. D., J. Wu, A. B. Wolf, Y. Chen, M.-S. Yao, and D. Kim, 2015: Constraints on cumulus parameterization from simulations of observed MJO events. J. Climate, 28, 6419–6442, https://doi.org/10.1175/JCLI-D-14-00832.1.
Droegemeier, K. K., and R. B. Wilhelmson, 1985: Three-dimensional numerical modeling of convection produced by interacting thunderstorm outflows. Part II: Variations in vertical wind shear. J. Atmos. Sci., 42, 2404–2414, https://doi.org/10.1175/1520-0469(1985)042<2404:TDNMOC>2.0.CO;2.
Fan, J., and Coauthors, 2009: Dominant role by vertical wind shear in regulating aerosol effects on deep convective clouds. J. Geophys. Res., 114, D22206, https://doi.org/10.1029/2009JD012352.
Farr, T. G., and Coauthors, 2007: The shuttle radar topography mission. Rev. Geophys., 45, RG2004, https://doi.org/10.1029/2005RG000183.
Grandpeix, J.-Y., and J.-P. Lafore, 2010: A density current parameterization coupled with Emanuel’s convection scheme. Part I: The models. J. Atmos. Sci., 67, 881–897, https://doi.org/10.1175/2009JAS3044.1.
Grant, L. D., and S. C. van den Heever, 2015: Cold pool and precipitation responses to aerosol loading: Modulation by dry layers. J. Atmos. Sci., 72, 1398–1408, https://doi.org/10.1175/JAS-D-14-0260.1.
Hudson, J. G., and S. S. Yum, 2001: Maritime-Continental drizzle contrasts in small cumuli. J. Atmos. Sci., 58, 915–926, https://doi.org/10.1175/1520-0469(2001)058<0915:MCDCIS>2.0.CO;2.
Kalina, E. A., K. Friedrich, H. Morrison, and G. H. Bryan, 2014: Aerosol effects on idealized supercell thunderstorms in different environments. J. Atmos. Sci., 71, 4558–4580, https://doi.org/10.1175/JAS-D-14-0037.1.
Khain, A. P., 2009: Notes on state-of-the-art investigations of aerosol effects on precipitation: A critical review. Environ. Res. Lett., 4, 015004, https://doi.org/10.1088/1748-9326/4/1/015004.
Khairoutdinov, M., and D. Randall, 2006: High-resolution simulation of shallow-to-deep convection transition over land. J. Atmos. Sci., 63, 3421–3436, https://doi.org/10.1175/JAS3810.1.
Khairoutdinov, M., S. K. Krueger, C.-H. Moeng, P. A. Bogenschutz, and D. A. Randall, 2009: Large-eddy simulation of maritime deep tropical convection. J. Adv. Model. Earth Syst., 1 (4), https://doi.org/10.3894/JAMES.2009.1.15.
Koenig, L. R., 1963: The glaciating behavior of small cumulonimbus clouds. J. Atmos. Sci., 20, 29–47, https://doi.org/10.1175/1520-0469(1963)020<0029:TGBOSC>2.0.CO;2.
Lasher-Trapp, S., S. Kumar, D. H. Moser, A. M. Blyth, J. R. French, R. C. Jackson, D. C. Leon, and D. M. Plummer, 2018: On different microphysical pathways to convective rainfall. J. Appl. Meteor. Climatol., 57, 2399–2417, https://doi.org/10.1175/JAMC-D-18-0041.1.
Lawson, P. R., D. O’Connor, P. Zmarzly, K. Weaver, B. Baker, Q. Mo, and H. Jonsson, 2006: The 2D-S (stereo) probe: Design and preliminary tests of a new airborne, high-speed, high-resolution particle imaging probe. J. Atmos. Oceanic Technol., 23, 1462–1477, https://doi.org/10.1175/JTECH1927.1.
Lebo, Z. J., and H. Morrison, 2014: Dynamical effects of aerosol perturbations on simulated idealized squall lines. Mon. Wea. Rev., 142, 991–1009, https://doi.org/10.1175/MWR-D-13-00156.1.
Loftus, A. M., and W. R. Cotton, 2014: Examination of CCN impacts on hail in a simulated supercell storm with triple-moment hail bulk microphysics. Atmos. Res., 147–148, 183–204, https://doi.org/10.1016/j.atmosres.2014.04.017.
Mallinson, H. M., and S. G. Lasher-Trapp, 2019: An investigation of hydrometeor latent cooling upon convective cold pool formation, sustainment, and properties. Mon. Wea. Rev., 147, 3205–3222, https://doi.org/10.1175/MWR-D-18-0382.1.
Mansell, E. R., and C. L. Ziegler, 2013: Aerosol effects on simulated storm electrification and precipitation in a two-moment bulk microphysics model. J. Atmos. Sci., 70, 2032–2050, https://doi.org/10.1175/JAS-D-12-0264.1.
Mansell, E. R., C. L. Ziegler, and E. C. Bruning, 2010: Simulated electrification of a small thunderstorm with two-moment bulk microphysics. J. Atmos. Sci., 67, 171–194, https://doi.org/10.1175/2009JAS2965.1.
Meyers, M. P., P. J. DeMott, and W. R. Cotton, 1992: New primary ice-nucleation parameterizations in an explicit cloud model. J. Appl. Meteor., 31, 708–721, https://doi.org/10.1175/1520-0450(1992)031<0708:NPINPI>2.0.CO;2.
Nesbitt, S. W., and Coauthors, 2021: A storm safari in subtropical South America: Proyecto RELAMPAGO. Bull. Amer. Meteor. Soc., 102, E1621–E1644, https://doi.org/10.1175/BAMS-D-20-0029.1.
Park, S., 2014: A Unified Convection Scheme (UNICON). Part I: Formulation. J. Atmos. Sci., 71, 3902–3930, https://doi.org/10.1175/JAS-D-13-0233.1.
Phillips, V. T. J., A. M. Blyth, P. R. A. Brown, T. W. Choularton, and J. Latham, 2001: The glaciation of a cumulus cloud over New Mexico. Quart. J. Roy. Meteor. Soc., 127, 1513–1534, https://doi.org/10.1002/qj.49712757503.
Rio, C., F. Hourdin, J.-Y. Grandpeix, and J.-P. Lafore, 2009: Shifting the diurnal cycle of parameterized deep convection over land. Geophys. Res. Lett., 36, L07809, https://doi.org/10.1029/2008GL036779.
Roberts, G. C., and A. Nenes, 2005: A continuous-flow streamwise thermal-gradient CCN chamber for atmospheric measurements. Aerosol Sci. Technol., 39, 206–221, https://doi.org/10.1080/027868290913988.
Rooney, G. G., A. J. Stirling, R. A. Stratton, and M. Whitall, 2022: C-POOL: A scheme for modelling convective cold pools in the Met Office unified model. Quart. J. Roy. Meteor. Soc., 148, 962–980, https://doi.org/10.1002/qj.4241.
Schlemmer, L., and C. Hohenegger, 2014: The formation of wider and deeper clouds as a result of cold-pool dynamics. J. Atmos. Sci., 71, 2842–2858, https://doi.org/10.1175/JAS-D-13-0170.1.
Seifert, A., and K. D. Beheng, 2006: A two-moment cloud microphysics parameterization for mixed-phase clouds. Part 2: Maritime vs. continental deep convective storms. Meteor. Atmos. Phys., 92, 67–82, https://doi.org/10.1007/s00703-005-0113-3.
Simpson, J. E., 1969: A comparison between laboratory and atmospheric density currents. Quart. J. Roy. Meteor. Soc., 95, 758–765, https://doi.org/10.1002/qj.49709540609.
Squires, P., 1952a: The growth of cloud drops by condensation. I. General characteristics. Aust. J. Sci. Res., 5, 59–86, https://doi.org/10.1071/CH9520059.
Squires, P., 1952b: The growth of cloud drops by condensation. II. The formation of large cloud drops. Aust. J. Sci. Res, 5, 473–499, https://doi.org/10.1071/CH9520473.
Squires, P., 1958a: The microstructure and colloidal stability of warm clouds. Part I. The relation between structure and stability. Tellus, 10A, 256–261.
Squires, P., 1958b: The microstructure and colloidal stability of warm clouds. Part II. The causes of the variations in microstructure. Tellus, 10A, 262–271.
Stevens, B., and G. Feingold, 2009: Untangling aerosol effects on clouds and precipitation in a buffered system. Nature, 461, 607–613, https://doi.org/10.1038/nature08281.
Storer, R. L., S. C. van den Heever, and G. L. Stephens, 2010: Modeling aerosol impacts on convective storms in different environments. J. Atmos. Sci., 67, 3904–3915, https://doi.org/10.1175/2010JAS3363.1.
Tao, W.-K., X. Li, A. Khain, T. Matsui, S. Lang, and J. Simpson, 2007: Role of atmospheric aerosol concentration on deep convective precipitation: Cloud-resolving model simulations. J. Geophys. Res., 112, D24S18, https://doi.org/10.1029/2007JD008728.
Tompkins, A. M., 2001: Organization of tropical convection in low vertical wind shears: The role of cold pools. J. Atmos. Sci., 58, 1650–1672, https://doi.org/10.1175/1520-0469(2001)058<1650:OOTCIL>2.0.CO;2.
Torri, G., Z. Kuang, and Y. Tian, 2015: Mechanisms for convection triggering by cold pools. Geophys. Res. Lett., 42, 1943–1950, https://doi.org/10.1002/2015GL063227.
Trapp, R. J., and J. M. Woznicki, 2017: Convectively induced stabilizations and subsequent recovery with supercell thunderstorms during the Mesoscale Predictability Experiment (MPEX). Mon. Wea. Rev., 145, 1739–1754, https://doi.org/10.1175/MWR-D-16-0266.1.
Twomey, S., 1959: The nuclei of natural cloud formation Part II: The supersaturation in natural clouds and the variation of cloud droplet concentration. Geophys. Pure Appl., 43, 243–249, https://doi.org/10.1007/BF01993560.
Varble, A., 2018: Erroneous attribution of deep convective invigoration to aerosol concentration. J. Atmos. Sci., 75, 1351–1368, https://doi.org/10.1175/JAS-D-17-0217.1.
Varble, A., and Coauthors, 2021: Utilizing a storm-generating hotspot to study convective cloud transitions: The CACTI experiment. Bull. Amer. Meteor. Soc., 102, E1597–E1620, https://doi.org/10.1175/BAMS-D-20-0030.1.
Veals, P. G., A. C. Varble, J. O. H. Russell, J. C. Hardin, and E. J. Zipser, 2022: Indications of a decrease in the depth of deep convective cores with increasing aerosol concentration during the CACTI campaign. J. Atmos. Sci., 79, 705–722, https://doi.org/10.1175/JAS-D-21-0119.1.
Wilson, J. W., and W. E. Schreiber, 1986: Initiation of convective storms at radar-observed boundary-layer convergence lines. Mon. Wea. Rev., 114, 2516–2536, https://doi.org/10.1175/1520-0493(1986)114<2516:IOCSAR>2.0.CO;2.
Ziegler, C. L., 1985: Retrieval of thermal and microphysical variables in observed convective storms. Part 1: Model development and preliminary testing. J. Atmos. Sci., 42, 1487–1509, https://doi.org/10.1175/1520-0469(1985)042<1487:ROTAMV>2.0.CO;2.
Zipser, E. J., 1977: Mesoscale and convective-scale downdrafts as distinct components of squall-line structure. Mon. Wea. Rev., 105, 1568–1589, https://doi.org/10.1175/1520-0493(1977)105<1568:MACDAD>2.0.CO;2.
