1. Introduction
Clouds, including precipitating deep cumulus convection and thunderstorms, are among the most uncertain processes in models of the climate system (Morrison et al. 2020b; Stephens et al. 2010) and among the most challenging and relevant phenomena to predict on weather forecast time scales (Stensrud et al. 2009). Thunderstorms are influenced by the thermodynamics of the atmosphere, large-scale dynamical forcing from weather systems, mesoscale boundaries such as sea breezes, and atmospheric composition. Each of these factors can also be influenced by urban areas (Shepherd 2013). Recognizing that the Houston, Texas, metropolitan area was influenced by all of these processes, the Tracking Aerosol Convection Interactions Experiment (TRACER) and Experiment of Sea Breeze Convection, Aerosols, Precipitation and Environment (ESCAPE) field campaigns (Jensen et al. 2019, 2022; Kollias et al. 2024) were conducted in 2021–22 to observe thunderstorms as they developed within and moved across this unique natural laboratory.
2. Background
a. Updrafts and cloud microphysics
The life cycle of isolated precipitating deep convection in the atmosphere has long been understood as a thunderstorm cell that typically lasts 20–30 min and begins with an updraft, followed by a downdraft and the cell’s dissipation (Byers and Braham 1949). Any longer-lived complexes of storms result from a multicellular sequence of updrafts, often forced along rain-cooled outflow (a cold pool) that provides a low-level convergence to force new updrafts and sustain their life cycle. In environments without substantial wind shear, the up-scale mesoscale organization is rare (Weisman and Klemp 1982).
One possible model of atmospheric deep convection analogizes a thunderstorm updraft to a dry thermal plume, which is produced by a continuous point source of buoyancy and expands due to lateral entrainment (Squires and Turner 1962). These idealized plumes allow for the properties of turbulent deep convection to be treated in terms of their averages, making such plumes especially suitable for parameterization of deep convection.
In contrast to the steady-state plume model, observational studies have observed impulsive behavior in thunderstorm updrafts (Anderson 1960) that are more in line with the discrete, bubble-like idealization of convection as a spherical vortex (Levine 1959). Recent theoretical considerations and simulations have also questioned the validity of the plume model (Morrison et al. 2020a; Peters et al. 2020) and have proposed that a chain of thermal bubbles is a better conceptual model to represent moist convection, consistent with the evidence for impulsive updrafts in observations. These bubbles might be spaced closely or further apart, and that flexibility combined with their spherical-vortex-like dynamical structure makes them a better model for predicting the actual entrainment, detrainment, and dilution effects on buoyancy in updrafts, which are as large as tens of percent compared to the adiabatic maximum.
Perturbation flows and thermodynamics in the thermal bubble also influence the cloud microphysics (Hernandez-Deckers et al. 2022). Latent heat release (from condensation) along the central axis of the moist thermals imposes an additional updraft perturbation. As droplets grow, greater turbulence in the bubble also promotes raindrop growth through the enhancement of collision and coalescence. In thermals that rise above the melting level, there is additional potential for interaction with ice particles, and the droplets may support enhanced riming and diffusional growth conditions favorable for electrification.
Polarimetric weather radars routinely observe the lofting of large drops to altitudes higher than the environmental 0°C level due to a sufficiently strong updraft that overcomes the drop fall speed. These oblate raindrops are spatially concentrated into a column and initially produce corresponding radar signals of enhanced radar reflectivity Z and differential reflectivity ZDR (Kumjian et al. 2014). The columns are not steady in depth or width and so serve to indicate updraft unsteadiness. At times, the updraft is sufficiently strong to loft these large drops to the mixed-phase portion of the cloud (where temperature T is 0°C > T > −40°C) where they can subsequently glaciate (Jameson et al. 1996). Upon the formation of larger zero-ZDR hail and graupel particles, ZDR drops to background levels (closer to zero), while KDP increases. van Lier-Walqui et al. (2016) defined a method for quantifying these columns, which was later refined in Fridlind et al. (2019).
Each of these processes is also influenced by aerosol concentration and chemistry through aerosol effects on nucleation rates of cloud droplets and ice crystals (Kreidenweis et al. 2019). Changes to the latent heat profile in different environments owing to variability in aerosol activation rates with temperature have been hypothesized to have an observable influence on thunderstorm updrafts and their correlates, including precipitation rates (Fan et al. 2020; Rosenfeld et al. 2008; Lin et al. 2006). The relative importance of aerosol invigoration in the climate system remains a topic of considerable interest and debate (Romps et al. 2023; Varble et al. 2023; Igel and van den Heever 2021), with strong recent interest in entrainment influences on deep convection (e.g., Peters et al. 2020; Giangrande et al. 2023) as yet another factor that could modify latent heat release. A recent model intercomparison study (Marinescu et al. 2021; Stier et al. 2024) motivated by an international roadmap for improving climate-scale modeling of deep convective clouds (van den Heever et al. 2018) compared convective column updraft and total precipitation. They showed that model formulation uncertainties (including in cloud microphysics parameterizations) were as large as aerosol effects and concluded there was a need for systematic characterization of observations that can be used to improve and constrain models.
b. Electrification and lightning
A sufficiently vigorous updraft in the mixed-phase region can result in lightning because electrification is fundamentally tied to the production of rimed ice-phase precipitation, especially graupel (Bruning et al. 2014; Williams 1989, 1985). As illustrated in Figs. 1a and 1b, charge is separated when ice crystals collide with and rebound from graupel. Actively riming graupel is known to electrify more per collision (Reynolds et al. 1957; Takahashi 1978; Saunders 2008), and so the particle number concentrations, crystal–graupel collision rates, and riming rates within an ascending thermal (Fig. 1c) relative to the polarity reversal line in Fig. 1b all control the charging rate and polarity. Because ice crystals have a slower terminal fall speed, the graupel separates from them as it precipitates (Fig. 1d), and differential advection (Fig. 1e) can result in further charge separation, converting particle-scale charging into net electric charge and potential regions that can support lightning. The ice particle habit influences the terminal fall speed, so graupel separates faster from ice crystals than snow and collides with more ice crystals per unit time—a secondary way in which riming enhances net charge separation and flash rates.
The primary physical factors controlling (a)–(e) storm electrification in a rising thermal (dashed red line and red circle at one instant) and producing a (f) net storm electrical structure leading to (g)–(i) lightning (yellow lines) and neutralizing charge deposition. Red and blue contours and symbols indicate positive and negative electrical polarity, respectively. For details, see the text.
Citation: Monthly Weather Review 152, 12; 10.1175/MWR-D-24-0060.1
The charging, sedimentation, and advection of hydrometeors (Figs. 1a–e), including at the turbulence scale (Bruning and MacGorman 2013), result in variations in the spatial extent of charged regions. Accordingly, Bruning and Thomas (2015) further showed that there is a sequence of smaller and then larger flashes during the sequence of updraft, precipitation formation, and detrainment of precipitating cloud during the life cycle of a thunderstorm cell, consistent with a tendency toward more stratified charge as the cloud precipitates and detrains into a stratiform anvil region (Brothers et al. 2018; Souza and Bruning 2021). The two polarities of electrification to graupel (Fig. 1b) often result in three vertically stacked regions of net electrical potential and two primary regions of large field (yellow stars in Fig. 1f), where lightning can begin. Lightning develops from its origin toward and throughout the regions of potential (positive channels toward negative potential, and vice versa, as in Fig. 1h; Coleman et al. 2003), depositing charge (Fig. 1g) that neutralizes the charge regions as it attaches to hydrometeors. Lightning produced by this charge structure (Fig. 1i) tends to be intracloud discharges between the upper positive and center negative charge regions (+IC), or discharges that lower net negative charge from cloud to ground (−CG) due to an excess negative charge in the storm. (Other net electrical structures are possible, though they were not common in the Houston storms studied herein.) In this way, the spatial structure and rate of lightning is intimately coupled to the microphysics and the vertical and horizontal draft structures of thunderstorms.
c. Joining cloud dynamics, microphysics, and lightning
Jameson et al. (1996) and Bruning et al. (2007) illustrated how cloud microphysics observed with polarimetric radar were closely tied to the sequence of electrification and lightning in a multicellular storm. Fridlind et al. (2019) demonstrated this concept for a larger number of isolated thunderstorm cells in Houston, explicitly hypothesizing a sequence of ZDR and KDP columns followed by lightning. Connecting to the thermal bubble theory, Salinas et al. (2022) showed that flashes tended to initiate near a balanced region of perturbation strain and vorticity near thermal-bubble-like circulations.
Mansell and Ziegler (2013), Fuchs et al. (2015), Hu et al. (2019), and Sun et al. (2023) also investigated the influence of aerosol on mixed-phase processes, including the processes leading to lightning. These studies found that lightning variability was associated with both aerosol and thermodynamic controls, with lightning increasing as CCN increased to about 1000 cm−3 and then saturating or decreasing thereafter. Stolz et al. (2015) showed that lightning in clouds with large warm (T > 0°C) cloud depths had relatively large aerosol sensitivity, while clouds with smaller warm-cloud depths had greater sensitivity to normalized convective available potential energy.
A stronger updraft will deliver more water mass flux to the mixed-phase region, which is then partitioned among a continuously interacting population of hydrometeors of all phases. In isolated storms like those in Houston, vertical wind shear is not significant enough to provide mesoscale organization and enhancement of updrafts, so the updraft morphology and intensity are very closely linked to the spatial scale and rate of input of buoyant thermals from the atmospheric boundary layer, the water vapor mixing ratio of those thermals, and the water vapor condensation and deposition rates following the thermals. Entrainment of dry air from the free troposphere further modifies the water budget. In addition to condensation rates that (at least initially) depend on aerosols, the succession of thermals results in the interaction of fresh condensate with precipitating hydrometeors that have exited previous thermals.
Fully disambiguating each of these processes is beyond the scope of this study, but it is clear that the basic physics of electrification and therefore the lightning flash rate depend on the production of both precipitating and nonprecipitating ice and liquid hydrometeors in the mixed-phase portion of the cloud, and that without sufficient graupel the formation of net charge regions would be impossible in sufficient time to produce lightning on the observed time scales. A necessary first step is to characterize the basic covariability of observed precipitation microphysics in the mixed phase with lightning so that the untangling of aerosol and meteorological processes can be situated within the distribution of possible storms during TRACER.
Thunderstorms in Houston are advantageous for the study of thermal-driven deep convection because they tend to be isolated cells. During the summer months (Wang et al. 2022), a ridge of 500-hPa high pressure is typically present over Texas, leading to weak midlevel wind speeds and therefore little vertical wind shear with height. Surface high pressure centered to the east near Bermuda brings a flow of warm, moist air from the Gulf of Mexico over Houston. The summer of 2022 exhibited an anomalously strong Bermuda high at 850 hPa, with heights above climatology extending westward toward the study domain, producing anomalously strong onshore flow (Fig. 2) on average.
The 850-hPa height and wind anomalies for the TRACER IOP months (June–September 2022) relative to the 2000–20 mean for June–August. Data from the MERRA-2 reanalysis (GMAO 2015; Gelaro et al. 2017).
Citation: Monthly Weather Review 152, 12; 10.1175/MWR-D-24-0060.1
Sea-breeze dynamics in this pattern tend to initiate thunderstorms in the afternoon while providing marine-continental thermodynamic and aerosol contrast, with the urban corridor contributing additional perturbations to this pattern. The isolated cells are relatively easy to track. The humid subtropical climate of Houston also leads to a relatively large fraction of storms that produce precipitation, primarily through warm-cloud (non-ice) processes. Some cells develop deeper into cumulonimbus and have a mixed-phase region, providing a contrast in precipitating storm depths and microphysical processes.
The purpose of this study is to characterize the distribution of radar polarimetry and lightning in isolated deep convection in Houston and the relative prevalence and depth of the unsteady updrafts that we associate with thermal bubbles. The unsteady ZDR and KDP columns tied to these bubbles are relatively short-lived and so indicate an impulse in vertical velocity that is observed through their microphysical signatures of large drop lofting and glaciation. While ZDR columns are expected to be prevalent (Fridlind et al. 2019), the relative frequency with which heavily rimed, mixed-phase precipitation and lightning are produced in storms in the Houston region of the Gulf Coast is uncertain. While it is straightforward to observe a cell with lightning and find the expected ZDR and KDP column sequences, we are not aware of any explicit theoretical prediction or empirical characterization for the fraction of clouds in any given environment that will have impulsive updrafts, glaciation, and lightning. Since operational radars can readily track clouds with vanishingly small surface precipitation rates, the population of tracked clouds across a few days is expected to be large. Such a statistical study is necessary to provide an observational constraint on simulations of clouds and their ice processes in the climate system. Hence, characterization of the overall population of cells gives later studies a useful reference when searching for effects of the land/sea and urban/rural contrasts in the thermodynamic and aerosol environments, as well as in case studies investigating dynamics of individual storms that track thermal bubbles using research radars.
3. Data and preprocessing
The data used in this study cover the TRACER field campaign’s intensive observation period (IOP) during 1 June–30 September 2022. Additional field observations with mobile radars and aircraft were conducted during June in the ESCAPE field campaign (Kollias et al. 2024), and with an additional research C-band radar in August and September but will not be utilized here. We manually identified 25 “golden” days on which there were thunderstorms and lightning within 90 km of the TRACER research radar assets and on which at least some cells were isolated. During the TRACER IOP, numerical weather prediction forecasts were conducted with the NASA Unified-Weather Research and Forecast (NU-WRF) Model (Matsui et al. 2022). Simulations were at a 1-km horizontal grid spacing, with electrification and 3D bulk lightning discharge parameterizations, inline polarimetric radar instrumental emulation, and updated cloud condensation nuclei schemes (EPIC). The NU-WRF EPIC real-time forecasts also performed well on the golden days, which are 2, 4, 17, and 22 June; 2, 6, 12, 13, 14, 28, and 29 July; 1, 2, 3, 6, 7, 8, 13, 21, 25, 27, and 31 August; and 1, 15, and 17 September. As will be detailed later, radar-based tracking of storms identified 7488 storm tracks across each of the 24-h golden-day periods, giving a large statistical sample in varying environments across these days.
Radar polarimetry for this study was provided by the KHGX S-band (wavelength of 10 cm) operational WSR-88D (Crum and Alberty 1993; Kumjian 2013) radar located just south-southeast of Houston in League City, Texas. Level-II data were obtained from the NOAA Big Data project, which replicates the official archive on an Amazon Web Services S3 storage bucket. KHGX sampled the full volume of storms every 5–6 min in a surveillance scan mode. The equivalent radar reflectivity factor at a horizontal polarization Zh, differential reflectivity ZDR, and differential phase shift ϕDP products in the level-II data were processed with the algorithms in the CSU-radartools package (Lang et al. 2019). The Lang et al. (2007) method was used to calculate the specific differential phase KDP from ϕDP. The ZDR was assumed to be well-calibrated, though calibration biases and standard deviations of about 0.2 dB are expected even after operational correction methods are applied (Holleman et al. 2022). Valid radar range gates were required to have a correlation coefficient ρhv > 0.9 and Zh > 10.0 dBZ.
Lightning data for this study were provided by the Houston Lightning Mapping Array (LMA; Logan 2021). During the TRACER campaign, two additional sensors from Texas Tech University were deployed to the southwest and east of Houston to supplement the nine permanent stations operated by Texas A&M University. Figure 3 gives the effective source and flash detection efficiency calculated using the observed station receiver thresholds following the methodology of Chmielewski and Bruning (2016); a source height of 7000 m, and a minimum of six stations were used in the calculations. LMA locations were considered valid if they had a maximum reduced χ2 value of 1.0 and were detected by a minimum of six stations. Flashes were identified using the density-based spatial clustering of applications with noise (DBSCAN)-based algorithm of Fuchs et al. (2016), and each flash was required to have five LMA source locations. For this work, the lmatools (Bruning 2015) flash algorithm architecture and output data formats were modernized into a new Python package (xlma-python) that unifies event, flash, and gridded variables (such as flash extent density) into a unified NetCDF file format. These files are available on the ESCAPE field campaign data archive (Logan et al. 2023).
Plan-view map of source and flash detection efficiency performance for the Houston LMA as configured during the TRACER campaign. Station locations are indicated by letters marking locations colored (blue–green–yellow) according to that station’s receiver threshold. Color shading (purple–red–yellow) indicates VHF source detection efficiency, and thick black contours indicate the corresponding flash detection efficiency. Thin black circles mark range rings every 50 km from KHGX. Thin gray lines indicate counties in Texas and a small portion of Louisiana, with a shoreline demarcating the Gulf of Mexico. Galveston Bay is located between stations M and H.
Citation: Monthly Weather Review 152, 12; 10.1175/MWR-D-24-0060.1
The environmental 0° level was assessed using operational model guidance and observed soundings at 2100 UTC for each day. These soundings are available in the archived TRACER forecast slides prepared on each day of the campaign and were found to be consistent with the ARM sounding synthesis product discussed later. Sensitivity tests on the final analyses showed little impact of the observed O(100) m variability from day to day.
4. Methodology
After quality control and polarimetric preprocessing, the KHGX data were interpolated to a three-dimensional Cartesian grid with a 500-m grid spacing using the PyART (Helmus and Collis 2016) package. Thunderstorms were identified using the Tracking and Object-Based Analysis of Clouds (TOBAC) tracking software (Heikenfeld et al. 2019; Sokolowsky et al. 2024). The horizontal extent of each storm at each time step (a TOBAC “segmented feature”) was identified using composite Zh (i.e., the maximum Zh at each column). Critical feature detection and segmentation parameters were as follows. A reflectivity threshold of Zh = 15 dBZ was used for watershed segmentation after a Gaussian blur filter was applied to Zh with smoothing parameter σ = 1. Feature centroids were identified using a weighted difference of Zh from the threshold.
Segmented features were linked across time into cell track segments using motion prediction. If more than 100 features were identified for a cell trajectory within a search range implied by the maximum allowed cell motion of (17 m s−1, or 1 km min−1 in dataset units), the search range was reduced by 5% until ≤100 features were found or until the search range was 20% of its starting value. Finally, any cell track segments where features approached within 15 km were joined into tracks using the cell merge–split functionality recently added to TOBAC. The result of tracking was a 2D feature mask giving the ID of each feature at each time step, and a data table linking the feature to its parent cell, and the cell to its parent track.
The microphysical and electrical properties of each track were quantified within the vertical column above the 2D feature mask at each time step. To count the lightning flashes along each track, the flash location was defined as the average of all constituent VHF events in the flash, i.e., the flash centroid. Radar measurements within the mixed-phase region were used to quantify the ZDR, KDP, and ρhv columns. To eliminate noisy polarimetric measurements in low signal-to-noise regions along cloud edges, ZDR, KDP, and ρhv values were only taken from grid boxes with Zh > 10 dBZ. The mixed-phase region was defined using typical environmental sounding values: a 0°C melting level altitude of 4.4 km, and a −40°C homogeneous freezing altitude of 10.6 km. Polarimetric columns were defined for each variable as follows: ZDR > 1.0 dB, KDP > 0.75° km−1, and ρhv < 0.98, similar to that used in Fridlind et al. (2019). The count of grid boxes meeting these criteria was recorded to quantify the column volume. The sum of values in these grid boxes was used to quantify column strength, except for ρhv, where the deficit below 1.0 was summed. To distinguish deeper columns, column strength S was also weighted by its depth above the melting level zm (Fridlind et al. 2019) to give Sw = ∑S(z)(z − zm), where z is the altitude of each grid box.
The final analysis considered track-level properties. At each time step, quantities were normalized by feature area, and the values of each feature were summed along the track and then normalized by the track duration. To ensure at least 3–4 unique radar elevation angle scans over the depth of the mixed-phase region, tracks had to be within 90 km of KHGX. A lower bound on range was not applied, so the radar cone of silence could have impacted storms within about 30-km range from the radar, though that impacts less than 10% of the observation area. No corrections were performed for the LMA detection efficiency and nonuniform source location errors because the storm range limit kept them within or very close to the outer perimeter of the LMA stations (Fig. 3).
The number of neighboring features within a 20-km distance of each track were also counted, summed over the track, and normalized by feature area and track duration. The neighbor count is a simple way to detect whether polarimetry and flash properties varied as trackable clouds became less isolated from one another.
5. Results
a. Lightning mapping
Lightning activity during June–September 2022 was greater over land (Figs. 4a,d,g,j) than ocean. June had less than half the lightning (50 782 flashes) of the next largest month, while August had the largest amount (289 920 flashes), and lightning on nearly every day (Figs. 4c,f,i,1). VHF source counts by time of day for each month (Figs. 4b,e,h,k) were largest during the local afternoon and evening (roughly 1700–0100 UTC), as solar heating of the land resulted in greater convective instability. VHF sources were distributed in altitude from the surface to about 18 km above mean sea level, with a maximum near 10 km and a secondary maximum near 5 km MSL. These maxima roughly correspond to the upper- and lower-level positive storm charge regions inferred from an analysis of individual lightning flashes on a few days, with the relative minimum in VHF sources in between corresponding to the negative storm charge. This charge structure is a normal tripole (Fig. 1f), as expected in isolated subtropical thunderstorms where moisture is plentiful (Bruning et al. 2014). A detailed, case-by-case examination of charge structure for every case is beyond the scope of this study but would be a valuable activity for future work to assess any departures from the normal tripolar structure.
Houston Lightning Mapping Array for (a)–(c) June, (d)–(f) July, (g)–(i) August, and (j)–(1) September 2022. (a),(d),(g),(j) Plan view of the monthly total flash extent density, shaded by the number of flashes at that location on a logarithmic scale. The Texas Gulf Coast (black line) and range rings (red circles) are shown every 50 km. (b),(e),(h),(k) Monthly sum of time–height VHF source density (color shading, logarithmic scale) and total number of VHF sources in each 5-min interval (black line) across the full diurnal cycle. (c),(f),(i),(l) Hourly flash rate for each day with lightning.
Citation: Monthly Weather Review 152, 12; 10.1175/MWR-D-24-0060.1
b. Track types
In total, 7488 tracks were identified across the 25 golden days during the 4-month observing campaign from June to September 2022. The number of tracks observed on each day are indicated in Fig. 5. The 25th, 50th, 75th, and 95th percentile track durations were 35, 45, 80, and 155 min, respectively. Track data for 7 August are shown in Fig. 6. A shift in track direction between 0000 UTC (late evening on 6 August) and the next diurnal cycle after 1500 UTC is evident. Many tracks on this day lasted longer than 1 h, illustrating the ability of the tracking algorithm to capture the full life cycle of storms beginning from the relatively weak 15-dBZ reflectivity threshold for feature detection.
Track count and fraction of track types on the golden days during TRACER. Each color corresponds to a different combination of the presence of a polarimetric column and/or lightning as indicated by the legend.
Citation: Monthly Weather Review 152, 12; 10.1175/MWR-D-24-0060.1
A map of tracks identified on 7 Aug 2022, filtered to only those that exceeded 35 dBZ to aid visualization. The time (UTC) of individual feature centroids (filled circles) along each track is indicated by the color bar. Apparent jumps or gaps in features along a track are due to algorithmic uncertainty. Each track is labeled with its ID as assigned by TOBAC.
Citation: Monthly Weather Review 152, 12; 10.1175/MWR-D-24-0060.1
On each day with deep convection, the number of tracks having lightning, ZDR, and/or KDP columns were counted. Some tracks had none of the above—so-called “nothing” tracks. Figure 5 shows the fraction of tracks having various combinations of these properties. The most frequently observed cell types were nothing cells (ranging from 15% to 90% of tracks on a day, total across all days of 62%), followed by cells with ZDR, KDP, and lightning (5%–50%, total 21%). The nothing cells did not have strong enough updrafts to raise raindrops to the mixed-phase region, consistent with a precipitation process that was primarily through warm-phase collision–coalescence. Lightning was not observed in the absence of a ZDR column, and KDP (and therefore lightning) was not observed without ZDR. The ZDR-only tracks were observed 2%–20% (total 5%) of the time, and ZDR–KDP-only tracks were also about 2%–25% (total 4%) of the tracks on any given day. The ZDR–lightning tracks were also observed (2%–25%, total 7%). All other types were observed <1% of the time.
c. Polarimetry and lightning distributions by track type
The fraction of track types on each day does not distinguish between weak and strong polarimetric columns, nor high and low flash rates. To assess whether greater ZDR and KDP column strengths were associated with lightning, the distributions of altitude-weighted ZDR and KDP column strengths for tracked cells were compared for tracks with and without lightning.
The distribution of flash counts in tracks with ZDR only compared to cells with both ZDR and KDP is shown in Figs. 7a,b). For a nominal cell width of a few tens of kilometers, values span from a weakly electrified storm with a stray lightning flash (10−7–10−6 flashes per square kilometer per second) up to tens of flashes per minute (10−3–10−2 flashes per square kilometer per second).
Counts of tracks having various combinations of polarimetric and lightning properties. Each measure is normalized by track duration and area, and the 5th, 50th, and 95th percentiles of the distribution are indicated. The title of each plot corresponds to the categories in Fig. 5 and gives the total number of tracks in that category. (a) Flash count for tracks with ZDR and lightning but no KDP column and (b) with all three. (c) Altitude-weighted ZDR column strength for tracks with a ZDR and KDP column and no lightning and (d) with all three. (e) Altitude-weighted KDP column strength for tracks with a ZDR and KDP column and no lightning and (f) with all three.
Citation: Monthly Weather Review 152, 12; 10.1175/MWR-D-24-0060.1
The median normalized flash count in tracks with ZDR only (Fig. 7a) was 1.12 × 10−5, while adding KDP (Fig. 7b) increased the normalized flash count by 48% to 1.66 × 10−5. The right tail of the distribution differed even more, increasing the 95th percentile value from 1.45 × 10−4 to 5.28 × 10−4 or a factor of 3.64 with the addition of KDP. Only a few dozen ZDR-only tracks had persistent, moderate electrification, while there were hundreds with the addition of KDP. The highest flash counts (>5 × 10−3, normalized) were only observed when KDP was present.
A comparison of tracks with and without lightning might also be expected to show differences in ZDR and KDP column strengths. Figures 7c–f compare tracks that always have both ZDR and KDP columns and separate those without lightning (327 tracks) and with lightning (1600 tracks).
The altitude-weighted ZDR column strength in tracks without lightning (Fig. 7c) was lower (at all percentiles). Tracks with lightning (Fig. 7d) had a median ZDR column strength that was 31% higher and a 95th percentile strength that was 42% higher.
The altitude-weighted KDP column strength in tracks without lightning (Fig. 7c) was lower at the median and 95th percentile. Tracks with lightning (Fig. 7d) had a median KDP column strength that was 68% higher and a 95th percentile strength that was 124% higher. The shape of the KDP distribution without lightning was also flatter at lower values, and conversely, the peak in the distribution with lightning was substantially more peaked near the median, especially when accounting for the log scale of the plot.
Summarizing, the population of tracks with lightning had a right tail that was shifted to higher values of column strength, and the increase in the column strength magnitude was about twice as strong for the KDP strength compared to ZDR. The greater number of tracks with lightning also meant that the right tails of the lightning tracks had hundreds of members, and while the nonlightning tracks with ZDR and KDP columns only had dozens of members.
Together, these results are consistent in showing larger values of ZDR and KDP column strengths in tracks with lightning. The signal in KDP is especially large. As indicated by the relative track counts, the presence of both ZDR and KDP means a track is much more likely to produce lightning than not.
d. Joint distributions
To further assess tracked storm properties, joint histograms of pairs of variables were calculated to identify any correlations between variable pairs. For brevity, the results here focus on the weighted column strength measures, though results were similar for column volume and unweighted column strength. Because the analysis aggregates variables along each storm track, the analysis here characterizes the net intensity of the polarimetric columns and lightning on time scale of the whole storm.
All variable pairs exhibited a substantial spread of at least one, and usually several, orders of magnitude, so relationships are visualized in a log–log space (Fig. 8). The Pearson correlation coefficient r was calculated for a power-law fit to each pair. All fits exhibited two-sided p values less than 0.001. The best correlations were between the polarimetric column strengths, especially ZDR–ρhv (Fig. 8f), followed by ZDR–KDP (Fig. 8e) and KDP–ρhv (Fig. 8h).
Joint histograms of track-total lightning flash counts, measures of altitude-weighted polarimetric column strength, and neighbor count, as defined in the text. The left column and bottom row in each histogram correspond to values of zero. Units have been normalized by the feature area and track duration. For a 40 km × 40 km feature of a 1-h duration, the normalization factor would be 5.76 × 106, such that the lowest bin of the flash and neighbor counts correspond to a count of one at some time during that track. The polarimetric column values are about three orders of magnitude larger, corresponding to an altitude weighting factor typically on the order of 1000 m. Power-law fits (excluding the left and bottom rows) of the form y = 10bxm are shown, with coefficients given alongside Pearson correlation coefficient r and the two-sided p value.
Citation: Monthly Weather Review 152, 12; 10.1175/MWR-D-24-0060.1
Correlations between the polarimetric column strength and lightning flash counts (Figs. 8a–c) were lower than among the polarimetric measures, though the largest flash rates were associated with the largest column strengths, as expected. Nonzero flash rates required the presence of ZDR and ρhv columns. There were significant numbers of tracks, including with larger flash counts, that did not have a KDP column.
The polarimetric and lightning variables were generally poorly correlated to the track neighbor count (Figs. 8d,g,i,j), though the neighbor counts for tracks without a column or without lightning were generally higher, perhaps implying that some aggregation of individual convective cloud elements favors electrification.
6. Discussion
a. Track types
The daily track-type data characterized the relative prevalence of warm- and cold-cloud precipitating deep convection in Houston. In aggregate, nothing tracks (which we take to be dominated by warm-phase precipitation processes) were the most common type, followed by deep cold clouds that produced ZDR and KDP columns and lightning, together accounting for a total of 83% of all tracks. There was also a small (but on some days considerable) fraction of cells that had a ZDR column but that did not have lightning.
The presence of tracks with a ZDR column and lightning but no KDP column was a somewhat unexpected result, though Sharma et al. (2021) did find better correlation of lightning with ZDR than to KDP in a supercellular thunderstorm. The KDP–lightning joint histogram (Fig. 8b) also shows that there was a significant overlap in the distribution of flash rates in cells with and without KDP columns. It is possible that the well-known difficulties in reliably retrieving KDP (Reimel and Kumjian 2021) are hiding some of the smaller but real KDP signals or that other sources of charge are involved beyond that generated in unsteady updraft pulses (e.g., advection and residual from prior storms).
On some days (e.g., 29 July, 31 August), clouds either remained in the warm phase with no impulsive, deep updrafts, or the cloud produced lightning after an updraft pulse, consistent with the general expectation that the vast majority of clouds were either shallow or deep, with few ZDR-only tracks. However, on other days (e.g., 2 June), there was an order of magnitude more tracks that had updraft pulses, but mixed-phase precipitation formation remained more marginal. The variability in the track-type fraction from day to day points to a likely role of regional (or even local) thermodynamic, kinematic, and aerosol variabilities in controlling updraft and microphysical behavior. On these days, the addition of lightning to the dataset helped to distinguish the degree of heavily rimed cold-cloud precipitation formation.
While the track-type fractions fluctuated from day to day, later days in the campaign had an increasing fraction of nothing tracks and a smaller fraction of tracks with lightning, implying a shift toward warm rain-dominant microphysical processes. Among the motivating factors for the TRACER campaign was to assess the relative importance of thermodynamic and aerosol factors in controlling cloud types. It was also our experience during real-time forecast activity in support of the field campaign that lower-tropospheric dry layers had a deleterious effect on chances for deep convection, with the month of June having an unusually sparse convective coverage.
To assess these factors, radiosonde-measured (ARM User Facility 2022b) and (prerelease, with preliminary quality control) CCN (ARM User Facility 2022a) data from the ARM LaPorte site near the northwest corner of Galveston Bay were analyzed. The reported CCN values are the counts activated at a value of 0.45% supersaturation. While some aerosol composition measurements were taken during the campaign, their analysis is outside the scope of this study. The datasets were averaged into 3-h windows, and RH vertical profiles and CCN are shown in Fig. 9.
(a) Radiosonde measurements of the relative humidity as a function of height and time and (b) average cloud condensation nucleus concentration, both measured at the LaPorte, Texas, ARM field site. Data are 3-h averages for June–September 2022. There are two periods of missing CCN data, including 13 Aug and after 2 Sep.
Citation: Monthly Weather Review 152, 12; 10.1175/MWR-D-24-0060.1
To capture preconvective conditions, the 1800–2100 UTC average values of entraining convective available potential energy (ECAPE; Peters et al. 2023) and NCCN were calculated and compared to the track fraction data (Fig. 10). Correlation of ECAPE to CCN was poor (Pearson correlation coefficient r = 0.11, two-sided p = 0.617). For all track fraction types versus ECAPE, |r| ≤ 0.21, and two-sided p ≥ 0.340 was large, except for ZDR-only tracks that had r = −0.41 and the smallest p = 0.061. Correlation to CCN was also poor for all track fraction types except for nothing tracks (r = 0.43, p = 0.048) and ZDR–lightning-only tracks (r = −0.65, p = 0.001). The nothing track fractions had perhaps two regimes: a wide spread of fraction values at CCN < 800 cm−3 and then fraction > 0.5 for larger CCN concentrations. Similarly, the ZDR–lightning fraction was between 0.05 and 0.20 for CCN < 800 cm−3 and <0.1 for larger CCN concentrations.
(a) Entraining CAPE and (b) cloud condensation nucleus concentration at the La Porte ARM site vs the fraction of tracks in each category for the days shown in Fig. 5. The legend of each plot indicates the Pearson correlation coefficient r and two-sided p value for each track category. The NCCN value on each day is also indicated on each ECAPE plot in the black text, with the vertical position scaled in proportion to the CCN value, with r and p indicated for NCCN vs ECAPE. (c),(d) As in (a) and (b), but for the subset of days with NCCN; 1000 cm−3.
Citation: Monthly Weather Review 152, 12; 10.1175/MWR-D-24-0060.1
Bivariate ordinary least squares regression of ECAPE and CCN to predict the nothing and ZDR–KDP–lightning track fractions (65% and 21% of the total number of tracks, respectively) did not improve correlation, while the ZDR–lightning track fraction (7% of the total number of tracks) was correlated about as well as the univariate correlation with CCN. It is curious that the ZDR-only tracks (5% of the total number tracks) and any tracks with KDP were much more poorly correlated to CCN and that ECAPE on any given day offered little skill in predicting tracked storm type. As we discuss below, heterogeneity in the aerosol and thermodynamic environment is likely a significant confounding factor in driving variability from track to track.
Prior work (see section 2c) has shown that lightning activity might be expected to be maximized at 1000 CCN cm−3. As already noted, the nothing track fractions did seem to exhibit a regime shift at a bit less than that CCN value. However, storms with fully realized precipitating mixed-phase microphysics and lightning (21% of tracks) did not exhibit any clear trend relative to the 1000 CCN cm−3 threshold. The implied richer set of microphysical processes in these more mature and more numerous storms apparently masked any CCN effects on this population of the tracked storms.
Figures 10c and 10d show the subset of days with CCN < 800 cm−3. The ZDR-only and ZDR–KDP-only tracks on these days (collectively 10% of the total number of tracks and always less than 0.3 total track fraction) were positively correlated with CCN (r ≥ 0.61, p ≤ 0.02) up to a track fraction of about 0.25. Beyond CCN > 1000 cm−3, the track fractions decreased. The correlations of lightning to CCN in the subset were the same as or worse than in the full dataset, while the nothing track fraction was negatively correlated at r = −0.52, p = 0.055, opposite the full dataset. Correlations of ECAPE to nothing, ZDR-only and ZDR–KDP lightning track fractions were about the same as in the full dataset. The ZDR–KDP-only (r = −0.54, p = 0.044) and ZDR-lightning tracks (r = 0.52, p = 0.055) were much better correlated. The correlation of NCCN to ECAPE (r = −0.47, p = 0.087) was also much larger in the subset, and the negative correlation was of similar magnitude and significant to the most significant correlations between track categories and ECAPE in the subset. The correlation of the environmental parameters makes it difficult to attribute any improved correlation in the track fractions in the subset to either NCCN or ECAPE. Even though the correlation of CCN to nonlightning track fractions was improved in the low-CCN subset, the more-numerous tracks with fully developed mixed-phase microphysics and lightning were poorly correlated to CCN, as in the full dataset.
None of these signals are clear enough to make robust claims about the precipitation process dynamics. If anything, this simplistic check for a correlation between thermodynamics, aerosols, and deep mixed-phase activation illustrates the need for a much more careful and complete assessment of the convective and microphysical parameters and their role in cumulus convective dynamics on a storm-by-storm basis, especially including measurements of spatiotemporal heterogeneity. In future work, we will add mobile aerosol and sounding measurements by Rapp et al. (2024) and the ESCAPE aircraft measurements (Kollias et al. 2024) to our track dataset to more carefully account for heterogeneity. The TRACER and associated datasets are also uniquely well suited to observation-constrained simulations with prognostic lightning, enabling another route to improve insights into the physical processes that lead to the onset of lightning flashes.
b. Covariances and microphysical–dynamical information content
The ZDR and KDP column strength data were analyzed because of their utility as proxies for episodic updraft and microphysical processes. These measures covaried, but with significant scatter, consistent with the expectation that they emphasize different aspects of the coupled updraft and precipitation generation processes. When compared to the radar-derived variables, lightning also covaried but with even greater scatter.
While the radar variables selected here are straightforward to understand as qualitative proxies of updraft and microphysical processes, it must be remembered that their relationship to updraft strength (Kumjian et al. 2014) and hydrometeor particle size distributions (Kumjian et al. 2019) is still uncertain, with the retrieval of the more fundamental physical state reliant on one or more relatively complex retrieval methodologies and emphasizing the precipitation-sized particles in any mixture, with less skill in retrieving the cloud-sized particle distributions.
Likewise, lightning flash rate correlates positively with updrafts and the presence of precipitating ice (Lang and Rutledge 2002; Deierling and Petersen 2008; Schultz et al. 2015), though with significant scatter. Lightning must also, in some way, covary with cloud liquid water content and cloud ice number concentrations, since electrification most fundamentally depends on the rebounding collision rates of those hydrometeor species. But as with radar, we do not observe each of those quantities directly. Rather, we observe the aggregate result of the physical process of electrification and one that requires additional interaction of particles—a step beyond the radar’s direct detection of microwave scattering from the particles themselves.
The physical considerations described in the preceding paragraphs help to make sense of the varying degrees of observed scatter: the scatter grows with process complexity while preserving the expected qualitative correlation.
Consider an alternative result in which lightning was observed to correlate directly with ZDR or KDP column strength. This would imply that lightning is interchangeable with updrafts and mixed-phase microphysics inferred from ZDR or KDP columns. Quantitative retrievals of cloud state using radar polarimetry are arguably more mature than those relying on lightning data, so such a finding could have some advantages in, for example, climate system characterization, as lightning is easier to observe globally than ZDR. However, we do not find such a result here.
Instead, the conclusion is that lightning is a complementary measure of coupled dynamical and microphysical processes in the mixed-phase region that adds information to that observed by radars. As with radar-only microphysical retrievals, there remains considerable uncertainty in practical inference of the physical state of the cloud from the observations. One of the unique pieces of information added by lightning to the radar picture may be its dependence on the small liquid and ice particles that are difficult to detect with radar, but which are more directly affected by the concentration of cloud condensation and ice nuclei. Studies to date have emphasized lightning’s correlation with precipitating ice but have not attempted to assess the dependence of flash rates on the smaller particles. Further use of this study’s dataset to control for similar ZDR and KDP properties across a range of electrification rates would help to isolate any variance in rebounding collision rates of smaller ice particles with graupel where riming is active, which could then be checked against local aerosol and meteorological conditions. Such a work is a possible direction for future theoretical and modeling studies.
7. Concluding remarks
This study summarized the radar polarimetric and electrical properties of days with deep convection and thunderstorms during the TRACER campaign. Daily track statistics showed that the day-to-day variability of deep convection in Houston was significant enough to impact the bulk polarimetric and electrical characteristics of tracked thunderstorms. Such a diversity bears out the rationale for conducting the TRACER and ESCAPE field campaigns in Houston, where it was thought that environmental factors would produce a diversity in storm characteristics while cells remained relatively isolated and unorganized in an environment of weak shear. This study provides a benchmark for how any future case studies of individual days or storms relate to the distribution of thunderstorm properties in the Houston region.
Lightning added a novel perspective to radar-only microphysical characterization. On days with marginal updrafts, lightning helped distinguish between degrees of activation of the mixed-phase region of cumuliform clouds, as seen in the lightning track fractions on each day. Storms with lightning had greater ZDR and KDP column strengths, with the KDP signal being a factor of about two larger. Correlations among polarimetric variables were stronger among each other than between each polarimetric variable and lightning. The lack of strong correlation implies that lightning is adding unique information about cloud microphysical interactions not offered by radar variables and therefore offers an additional physical constraint with different sensitivities, including cloud droplet and ice crystal concentrations.
The dataset analyzed here provides a foundation for future studies using the TRACER dataset. Such studies will need to carefully assess the physical drivers of the observed variables, accounting for spatiotemporal heterogeneity of observed properties. The range of properties observed also suggests that the TRACER dataset is well suited to assessing the dynamics of individual storms and development of observation operators that can be used to infer more fundamental physical processes of storms.
Acknowledgments.
This work was supported by DOE DE-SC0021247 and NSF AGS-2019939. We thank Michael Jensen and the DOE/ARM program for their capable coordination of the field campaign and the field forecasting teams led by Scott Collis (TRACER) and Andrew Dzambo (ESCAPE) that led to the collection of a quality dataset. Bobby Jackson and Jessica Souza contributed to helpful early discussions on cell tracking and to data processing. Ann Fridlind lent crucial guidance to our early formulation of this study and provided comments on a draft of the manuscript. We are also grateful for the participation of students, too numerous to mention, for their participation in forecasting and in maintenance of field equipment. This research has made use of the NASA Goddard Science Managed Cloud Environment (SMCE), which is a service of the Computational and Information Sciences and Technology Office at the NASA Goddard Space Flight Center.
Data availability statement.
Analysis code to reproduce the analyses in this study can be found in the TRACER-PAWS-NEXRAD-LMA repository on GitHub (https://github.com/deeplycloudy/TRACER-PAWS-NEXRAD-LMA). Key upstream libraries utilized include TOBAC (https://github.com/tobac-project/tobac), xlma-python (https://github.com/deeplycloudy/xlma-python), and pyart (https://arm-doe.github.io/pyart/). The Houston Lightning Mapping Array data are available at NCAR/EOL (Logan et al. 2023). NEXRAD data were downloaded from the NOAA AWS S3 bucket (https://registry.opendata.aws/noaa-nexrad). ARM sounding and CCN data were downloaded from the ARM Data Center (ARM User Facility 2022b,a).
REFERENCES
Anderson, C. E., 1960: A study of the pulsating growth of cumulus clouds. Ph.D. thesis, Massachusetts Institute of Technology, 165 pp., http://hdl.handle.net/1721.1/58153.
ARM User Facility, 2022a: Cloud condensation nuclei particle counter (AOSCCN2COLAAVG), 2022-06-01 to 2022-09-30, ARM mobile facility (HOU) Houston, TX; AMF1 (main site for TRACER) (M1). ARM Data Center, https://doi.org/10.5439/1323894.
ARM User Facility, 2022b: Interpolated sonde (INTERPOLATEDSONDE), 2022-06-01 to 2022-09-30, ARM mobile facility (HOU) Houston, TX; AMF1 (main site for TRACER) (M1). ARM Data Center, https://doi.org/10.5439/1095316.
Brothers, M. D., E. C. Bruning, and E. R. Mansell, 2018: Investigating the relative contributions of charge deposition and turbulence in organizing charge within a thunderstorm. J. Atmos. Sci., 75, 3265–3284, https://doi.org/10.1175/JAS-D-18-0007.1.
Bruning, E. C., 2015: lmatools: lmatools-v0.5z-stable. Zenodo, https://doi.org/10.5281/zenodo.32510.
Bruning, E. C., and D. R. MacGorman, 2013: Theory and observations of controls on lightning flash size spectra. J. Atmos. Sci., 70, 4012–4029, https://doi.org/10.1175/JAS-D-12-0289.1.
Bruning, E. C., and R. J. Thomas, 2015: Lightning channel length and flash energy determined from moments of the flash area distribution. J. Geophys. Res. Atmos., 120, 8925–8940, https://doi.org/10.1002/2015JD023766.
Bruning, E. C., W. D. Rust, T. J. Schuur, D. R. MacGorman, P. R. Krehbiel, and W. Rison, 2007: Electrical and polarimetric radar observations of a multicell storm in TELEX. Mon. Wea. Rev., 135, 2525–2544, https://doi.org/10.1175/MWR3421.1.
Bruning, E. C., S. A. Weiss, and K. M. Calhoun, 2014: Continuous variability in thunderstorm primary electrification and an evaluation of inverted-polarity terminology. Atmos. Res., 135–136, 274–284, https://doi.org/10.1016/j.atmosres.2012.10.009.
Byers, H. R., and R. R. Braham, 1949: The Thunderstorm: Report of the Thunderstorm Project. U.S. Government Printing Office, 287 pp.
Chmielewski, V. C., and E. C. Bruning, 2016: Lightning Mapping Array flash detection performance with variable receiver thresholds. J. Geophys. Res. Atmos., 121, 8600–8614, https://doi.org/10.1002/2016JD025159.
Coleman, L. M., T. C. Marshall, M. Stolzenburg, T. Hamlin, P. R. Krehbiel, W. Rison, and R. J. Thomas, 2003: Effects of charge and electrostatic potential on lightning propagation. J. Geophys. Res., 108, 4298, https://doi.org/10.1029/2002JD002718.
Crum, T. D., and R. L. Alberty, 1993: The WSR-88D and the WSR-88D operational support facility. Bull. Amer. Meteor. Soc., 74, 1669–1688, https://doi.org/10.1175/1520-0477(1993)074<1669:TWATWO>2.0.CO;2.
Deierling, W., and W. A. Petersen, 2008: Total lightning activity as an indicator of updraft characteristics. J. Geophys. Res., 113, D16210, https://doi.org/10.1029/2007JD009598.
Fan, J., Y. Zhang, Z. Li, J. Hu, and D. Rosenfeld, 2020: Urbanization-induced land and aerosol impacts on sea-breeze circulation and convective precipitation. Atmos. Chem. Phys., 20, 14 163–14 182, https://doi.org/10.5194/acp-20-14163-2020.
Fridlind, A. M., and Coauthors, 2019: Use of polarimetric radar measurements to constrain simulated convective cell evolution: A pilot study with Lagrangian tracking. Atmos. Meas. Tech., 12, 2979–3000, https://doi.org/10.5194/amt-12-2979-2019.
Fuchs, B. R., and Coauthors, 2015: Environmental controls on storm intensity and charge structure in multiple regions of the continental United States. J. Geophys. Res. Atmos., 120, 6575–6596, https://doi.org/10.1002/2015JD023271.
Fuchs, B. R., E. C. Bruning, S. A. Rutledge, L. D. Carey, P. R. Krehbiel, and W. Rison, 2016: Climatological analyses of LMA data with an open-source lightning flash-clustering algorithm. J. Geophys. Res. Atmos., 121, 8625–8648, https://doi.org/10.1002/2015JD024663.
Gelaro, R., and Coauthors, 2017: The Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2). J. Climate, 30, 5419–5454, https://doi.org/10.1175/JCLI-D-16-0758.1.
Giangrande, S. E., T. S. Biscaro, and J. M. Peters, 2023: Seasonal controls on isolated convective storm drafts, precipitation intensity, and life cycle as observed during GoAmazon2014/5. Atmos. Chem. Phys., 23, 5297–5316, https://doi.org/10.5194/acp-23-5297-2023.
GMAO, 2015: MERRA-2 instM_3d_ana_Np: 3d, monthly mean, instantaneous, pressure-level, analysis, analyzed meteorological fields V5.12.4. GES DISC, accessed 11 March 2024, https://doi.org/10.5067/V92O8XZ30XBI.
Heikenfeld, M., P. J. Marinescu, M. Christensen, D. Watson-Parris, F. Senf, S. C. van den Heever, and P. Stier, 2019: tobac 1.2: Towards a flexible framework for tracking and analysis of clouds in diverse datasets. Geosci. Model Dev., 12, 4551–4570, https://doi.org/10.5194/gmd-12-4551-2019.
Helmus, J. J., and S. M. Collis, 2016: The Python ARM Radar Toolkit (Py-ART), a library for working with weather radar data in the python programming language. J. Open Res. Software, 4, e25, https://doi.org/10.5334/jors.119.
Hernandez-Deckers, D., T. Matsui, and A. M. Fridlind, 2022: Updraft dynamics and microphysics: On the added value of the cumulus thermal reference frame in simulations of aerosol–deep convection interactions. Atmos. Chem. Phys., 22, 711–724, https://doi.org/10.5194/acp-22-711-2022.
Holleman, I., A. Huuskonen, and B. Taylor, 2022: Solar monitoring of the NEXRAD WSR-88D network using operational scan data. J. Atmos. Oceanic Technol., 39, 125–139, https://doi.org/10.1175/JTECH-D-20-0204.1.
Hu, J., and Coauthors, 2019: Polarimetric radar convective cell tracking reveals large sensitivity of cloud precipitation and electrification properties to CCN. J. Geophys. Res. Atmos., 124, 12 194–12 205, https://doi.org/10.1029/2019JD030857.
Igel, A. L., and S. C. van den Heever, 2021: Invigoration or enervation of convective clouds by aerosols? Geophys. Res. Lett., 48, e2021GL093804, https://doi.org/10.1029/2021GL093804.
Jameson, A. R., M. J. Murphy, and E. P. Krider, 1996: Multiple-parameter radar observations of isolated Florida thunderstorms during the onset of electrification. J. Appl. Meteor., 35, 343–354, https://doi.org/10.1175/1520-0450(1996)035<0343:MPROOI>2.0.CO;2.
Jensen, M., R. Jundt, and H. Powers, 2019: Tracking Aerosol Convection Interactions ExpeRiment (TRACER) science plan. Tech. Rep. DOE/SC-ARM-19-017, 39 pp., https://www.arm.gov/publications/programdocs/doe-sc-arm-19-017.pdf.
Jensen, M., and Coauthors, 2022: A succession of cloud, precipitation, aerosol, and air quality field experiments in the coastal urban environment. Bull. Amer. Meteor. Soc., 103, 103–105, https://doi.org/10.1175/BAMS-D-21-0104.1.
Kollias, P., and Coauthors, 2024: Experiment of Sea Breeze Convection, Aerosols, Precipitation and Environment (ESCAPE). Bull. Amer. Meteor. Soc., in press, https://doi.org/10.1175/BAMS-D-23-0014.1.
Kreidenweis, S. M., M. Petters, and U. Lohmann, 2019: 100 years of progress in cloud physics, aerosols, and aerosol chemistry research. A Century of Progress in Atmospheric and Related Sciences: Celebrating the American Meteorological Society Centennial, Meteor. Monogr., No. 59, Amer. Meteor. Soc., https://doi.org/10.1175/AMSMONOGRAPHS-D-18-0024.1.
Kumjian, M. R., 2013: Principles and applications of dual-polarization weather radar. Part I: Description of the polarimetric radar variables. J. Oper. Meteor., 1, 226–242, https://doi.org/10.15191/nwajom.2013.0119.
Kumjian, M. R., A. P. Khain, N. Benmoshe, E. Ilotoviz, A. V. Ryzhkov, and V. T. J. Phillips, 2014: The anatomy and physics of ZDR columns: Investigating a polarimetric radar signature with a spectral bin microphysical model. J. Appl. Meteor. Climatol., 53, 1820–1843, https://doi.org/10.1175/JAMC-D-13-0354.1.
Kumjian, M. R., C. P. Martinkus, O. P. Prat, S. Collis, M. van Lier-Walqui, and H. C. Morrison, 2019: A moment-based polarimetric radar forward operator for rain microphysics. J. Appl. Meteor. Climatol., 58, 113–130, https://doi.org/10.1175/JAMC-D-18-0121.1.
Lang, T. J., and S. A. Rutledge, 2002: Relationships between convective storm kinematics, precipitation, and lightning. Mon. Wea. Rev., 130, 2492–2506, https://doi.org/10.1175/1520-0493(2002)130,2492:RBCSKP.2.0.CO;2.
Lang, T. J., B. Dolan, N. Guy, C. Gerlach, and J. Hardin, 2019: Csuradarmet/csu_radartools: Csu_radartools v1.3. Zenodo. https://doi.org/10.5281/zenodo.2562063.
Lang, T. J., D. A. Ahijevych, S. W. Nesbitt, R. E. Carbone, S. A. Rutledge, and R. Cifelli, 2007: Radar-observed characteristics of precipitating systems during NAME 2004. J. Climate, 20, 1713–1733, https://doi.org/10.1175/JCLI4082.1.
Levine, J., 1959: Spherical vortex theory of bubble-like motion in cumulus clouds. J. Meteor., 16, 653–662, https://doi.org/10.1175/1520-0469(1959)016%3C0653:SVTOBL%3E2.0.CO;2.
Lin, J. C., T. Matsui, R. A. Pielke Sr., and C. Kummerow, 2006: Effects of biomass-burning-derived aerosols on precipitation and clouds in the Amazon Basin: A satellite-based empirical study. J. Geophys. Res., 111, D19204, https://doi.org/10.1029/2005JD006884.
Logan, T., 2021: An analysis of the performance of the Houston Lightning Mapping Array during an intense period of convection during Tropical Storm Harvey. J. Geophys. Res. Atmos., 126, e2020JD033270, https://doi.org/10.1029/2020JD033270.
Logan, T., E. Bruning, K. Brunner, and J. Souza, 2023: Houston Lightning Mapping Array (HLMA) Flash-level data, version 1.0. UCAR/NCAR - Earth Observing Laboratory, accessed 11 March 2024, https://doi.org/10.26023/GBKS-E7VT-HS11.
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.
Marinescu, P. J., and Coauthors, 2021: Impacts of varying concentrations of cloud condensation nuclei on deep convective cloud updrafts—A multimodel assessment. J. Atmos. Sci., 78, 1147–1172, https://doi.org/10.1175/JAS-D-20-0200.1.
Matsui, T., T. Iguchi, M. van Lier-Walqui, E. Mansell, E. C. Bruning, K. Brunner, and A. M. Fridlind, 2022: NUWRF-EPIC TRACER IOP forecasting and validation results using polarimetric radar simulator and cell tracking analyses. Abstracts, 2022 Joint ARM User Facility and ASR Principal Investigator Meeting, Rockville, MD, U.S. Department of Energy, https://asr.science.energy.gov/meetings/stm/posters/abstract/3065.
Morrison, H., J. M. Peters, A. C. Varble, W. M. Hannah, and S. E. Giangrande, 2020a: Thermal chains and entrainment in cumulus updrafts. Part I: Theoretical description. J. Atmos. Sci., 77, 3637–3660, https://doi.org/10.1175/JAS-D-19-0243.1.
Morrison, H., and Coauthors, 2020b: Confronting the challenge of modeling cloud and precipitation microphysics. J. Adv. Model. Earth Syst., 12, e2019MS001689, https://doi.org/10.1029/2019MS001689.
Peters, J. M., H. Morrison, A. C. Varble, W. M. Hannah, and S. E. Giangrande, 2020: Thermal chains and entrainment in cumulus updrafts. Part II: Analysis of idealized simulations. J. Atmos. Sci., 77, 3661–3681, https://doi.org/10.1175/JAS-D-19-0244.1.
Peters, J. M., D. R. Chavas, C.-Y. Su, H. Morrison, and B. E. Coffer, 2023: An analytic formula for entraining CAPE in midlatitude storm environments. J. Atmos. Sci., 80, 2165–2186, https://doi.org/10.1175/JAS-D-23-0003.1.
Rapp, A. D., and Coauthors, 2024: TAMU TRACER: Targeted mobile measurements to isolate the impacts of aerosols and meteorology on deep convection. Bull. Amer. Meteor. Soc., 105, E1685–E1702, https://doi.org/10.1175/BAMS-D-23-0218.1.
Reimel, K. J., and M. Kumjian, 2021: Evaluation of KDP estimation algorithm performance in rain using a known-truth framework. J. Atmos. Oceanic Technol., 38, 587–605, https://doi.org/10.1175/JTECH-D-20-0060.1.
Reynolds, S. E., M. Brook, and M. F. Gourley, 1957: Thunderstorm charge separation. J. Meteor., 14, 426–436, https://doi.org/10.1175/1520-0469(1957)014<0426:TCS>2.0.CO;2.
Romps, D. M., K. Latimer, Q. Zhu, T. Jurkat-Witschas, C. Mahnke, T. Prabhakaran, R. Weigel, and M. Wendisch, 2023: Air pollution unable to intensify storms via warm-phase invigoration. Geophys. Res. Lett., 50, e2022GL100409, https://doi.org/10.1029/2022GL100409.
Rosenfeld, D., U. Lohmann, G. B. Raga, C. D. O’Dowd, M. Kulmala, S. Fuzzi, A. Reissell, and M. O. Andreae, 2008: Flood or drought: How do aerosols affect precipitation? Science, 321, 1309–1313, https://doi.org/10.1126/science.1160606.
Salinas, V., E. C. Bruning, and E. R. Mansell, 2022: Examining the kinematic structures within which lightning flashes are initiated using a cloud-resolving model. J. Atmos. Sci., 79, 513–530, https://doi.org/10.1175/JAS-D-21-0132.1.
Saunders, C., 2008: Charge separation mechanisms in clouds. Space Sci. Rev., 137, 335–353, https://doi.org/10.1007/s11214-008-9345-0.
Schultz, C. J., L. D. Carey, E. V. Schultz, and R. J. Blakeslee, 2015: Insight into the kinematic and microphysical processes that control lightning jumps. Wea. Forecasting, 30, 1591–1621, https://doi.org/10.1175/WAF-D-14-00147.1.
Sharma, M., R. L. Tanamachi, E. C. Bruning, and K. M. Calhoun, 2021: Polarimetric and electrical structure of the 19 May 2013 Edmond–Carney, Oklahoma tornadic supercell. Mon. Wea. Rev., 149, 2049–2078, https://doi.org/10.1175/MWR-D-20-0280.1.
Shepherd, J. M., 2013: Impacts of urbanization on precipitation and storms: Physical insights and vulnerabilities. Climate Vulnerability, 5, 109–125, https://doi.org/10.1016/b978-0-12-384703-4.00503-7.
Sokolowsky, G. A., and Coauthors, 2024: tobac v1.5: Introducing fast 3D tracking, splits and mergers, and other enhancements for identifying and analysing meteorological phenomena. Geosci. Model Dev., 17, 5309–5330, https://doi.org/10.5194/gmd-17-5309-2024.
Souza, J. C. S., and E. C. Bruning, 2021: Assessment of turbulence intensity in different spots of lightning flash propagation. Geophys. Res. Lett., 48, e2021GL095923, https://doi.org/10.1029/2021GL095923.
Squires, P., and J. S. Turner, 1962: An entraining jet model for cumulo-nimbus updraughts. Tellus, 14 (4), 422–434, https://doi.org/10.1111/j.2153-3490.1962.tb01355.x.
Stensrud, D. J., and Coauthors, 2009: Convective-scale warn-on-forecast system. Bull. Amer. Meteor. Soc., 90, 1487–1500, https://doi.org/10.1175/2009BAMS2795.1.
Stephens, G. L., and Coauthors, 2010: Dreary state of precipitation in global models. J. Geophys. Res., 115, D24211, https://doi.org/10.1029/2010JD014532.
Stier, P., and Coauthors, 2024: Multifaceted aerosol effects on precipitation. Nat. Geosci., 17, 719–732, https://doi.org/10.1038/s41561-024-01482-6.
Stolz, D. C., S. A. Rutledge, and J. R. Pierce, 2015: Simultaneous influences of thermodynamics and aerosols on deep convection and lightning in the tropics. J. Geophys. Res. Atmos., 120, 6207–6231, https://doi.org/10.1002/2014JD023033.
Sun, M., X. Qie, E. R. Mansell, D. Liu, Y. Yair, A. O. Fierro, S. Yuan, and J. Lu, 2023: Aerosol impacts on storm electrification and lightning discharges under different thermodynamic environments. J. Geophys. Res. Atmos., 128, e2022JD037450, https://doi.org/10.1029/2022JD037450.
Takahashi, T., 1978: Riming electrification as a charge generation mechanism in thunderstorms. J. Atmos. Sci., 35, 1536–1548, https://doi.org/10.1175/1520-0469(1978)035<1536:REAACG>2.0.CO;2.
van den Heever, S. C., and Coauthors, 2018: Aerosol-Cloud-Precipitation-Climate (ACPC) initiative: Deep convective cloud group roadmap. ACPC Tech. Rep., 13 pp., http://acpcinitiative.org/Docs/ACPC_DCC_Roadmap_171019.pdf.
van Lier-Walqui, M., and Coauthors, 2016: On polarimetric radar signatures of deep convection for model evaluation: Columns of specific differential phase observed during MC3E. Mon. Wea. Rev., 144, 737–758, https://doi.org/10.1175/MWR-D-15-0100.1.
Varble, A. C., A. L. Igel, H. Morrison, W. W. Grabowski, and Z. J. Lebo, 2023: Opinion: A critical evaluation of the evidence for aerosol invigoration of deep convection. Atmos. Chem. Phys., 23, 13 791–13 808, https://doi.org/10.5194/acp-23-13791-2023.
Wang, D., and Coauthors, 2022: Linking synoptic patterns to cloud properties and local circulations over southeastern Texas. J. Geophys. Res. Atmos., 127, e2021JD035920, https://doi.org/10.1029/2021JD035920.
Weisman, M. L., and J. B. Klemp, 1982: The dependence of numerically simulated convective storms on vertical wind shear and buoyancy. Mon. Wea. Rev., 110, 504–520, https://doi.org/10.1175/1520-0493(1982)110<0504:TDONSC>2.0.CO;2.
Williams, E. R., 1985: Large-scale charge separation in thunderclouds. J. Geophys. Res., 90, 6013–6025, https://doi.org/10.1029/JD090iD04p06013.
Williams, E. R., 1989: The tripole structure of thunderstorms. J. Geophys. Res., 94, 13 151–13 167, https://doi.org/10.1029/JD094iD11p13151.
