Observations of two sprite-producing storms in ColoradoTimothy J. Lang1, Walter A. Lyons2, Steven A. Cummer3, Brody R. Fuchs4, Brenda Dolan4,Steven A. Rutledge4, Paul Krehbiel5, William Rison5, Mark Stanley5, and Thomas Ashcraft6

1National Aeronautics and Space Administration George C. Marshall Space Flight Center, Huntsville, Alabama, USA, 2FMAResearch, Inc., Fort Collins, Colorado, USA, 3Duke University, Durham, North Carolina, USA, 4Colorado State University, FortCollins, Colorado, USA, 5New Mexico Institute of Mining and Technology, Socorro, New Mexico, USA, 6HeliotownObservatory, Lamy, New Mexico, USA

Abstract Two sprite-producing thunderstorms were observed on 8 and 25 June 2012 in northeasternColorado by a combination of low-light cameras, a lightning mapping array, polarimetric and Dopplerradars, the National Lightning Detection Network, and charge moment change measurements. The 8 Juneevent evolved from a tornadic hailstorm to a larger multicellular system that produced 21 observed positivesprites in 2 h. The majority of sprites occurred during a lull in convective strength, as measured by totalflash rate, flash energy, and radar echo volume. Mean flash area spiked multiple times during this period;however, total flash rates still exceeded 60min�1, and portions of the storm featured a complex anomalouscharge structure, with midlevel positive charge near �20°C. The storm produced predominantly positivecloud-to-ground lightning. All sprite-parent flashes occurred on the northeastern flank of the storm, wherestrong westerly upper level flow was consistent with advection of charged precipitation away fromconvection, providing a pathway for stratiform lightning. The 25 June event was another multicellularhailstorm with an anomalous charge structure that produced 26 positive sprites in less than 1 h. The spritesagain occurred during a convective lull, with relatively weaker reflectivity and lower total flash rate butrelatively larger mean flash area. However, all sprite parents occurred in or near convection and tappedcharge layers in adjacent anvil cloud. The results demonstrate the sprite production by convective groundstrokes in anomalously charged storms and also indicate that sprite production and convective vigor areinversely related in mature storms.

1. Introduction

After more than 25 years of research on sprites, stemming from their fortuitous capture on video by Franzet al. [1990], a general understanding of how they are related to thunderstorm structure and evolution hasbegun to coalesce. Sprites are most likely to be associated with powerful positive cloud-to-ground (+CG)lightning strokes [Boccippio et al., 1995; Bell et al., 1998;Williams, 1998], especially those associated with largecharge moment changes (CMCs) [Huang et al., 1999; Hu et al., 2002; Cummer and Lyons, 2005; Hiraki andFukunishi, 2006; Qin et al., 2012]. Production of such strokes is favored to occur in stratiform precipitationregions, such as might be found in mesoscale convective systems (MCSs) [Lyons, 1996; Lyons et al., 2003;Lyons, 2006; Williams and Yair, 2006]. Research has shown that sprite production appears to be related tothe development and intensification of the stratiform region [Lyons, 1996, 2006; Williams and Yair, 2006;Soula et al., 2009; Lang et al., 2010]; however, it also appears that a robust convective line or region is requiredsince many sprite-parent CGs originate there, followed by propagation into the stratiform precipitation [Langet al., 2010; van der Velde et al., 2010, 2014; Soula et al., 2015]. The charge layers involving these flashes appearto be fed by charged ice particles that detrain from the convective line, which then descend and furtherdevelop in the weak mesoscale updrafts typically found inside stratiform precipitation regions [Carey et al.,2005; Ely et al., 2008].

However, there are notable exceptions to this scenario. Some sprites are caused by �CGs [Barrington-Leighet al., 1999, 2001; Taylor et al., 2008; Soula et al., 2009; Li et al., 2012; Lu et al., 2012], and these appear tonot be strictly tied to stratiform precipitation; they often occur in deep convection [Lang et al., 2013]. Somesprites may even be caused by purely intracloud (IC) lightning [Neubert et al., 2005], and in any event it hasbeen established that sprite evolution is strongly related to the development of in-cloud portions of theparent flash [van der Velde et al., 2006, 2010, 2014; Lu et al., 2013]. Furthermore, positive sprites can occur over+CGs in convective regions, and there is some limited evidence that many of these are related to the

LANG ET AL. TWO SPRITE-PRODUCING STORMS IN COLORADO 9675

PUBLICATIONSJournal of Geophysical Research: Atmospheres

RESEARCH ARTICLE10.1002/2016JD025299

Special Section:Deep Convective Clouds andChemistry 2012 Studies (DC3)

Key Points:• Positive sprites were observed overanomalously electrified convection inColorado

• Sprites did not occur until intense,mature convection began weakening

• Polarimetric and multi-Doppler radaranalyses indicated effects of stormmicrophysics and kinematics onsprite-parent lightning

Supporting Information:• Supporting Information S1• Movie S1• Movie S2• Movie S3• Movie S4• Movie S5• Movie S6• Movie S7• Movie S8• Movie S9• Movie S10• Movie S11• Movie S12• Movie S13• Movie S14• Movie S15• Movie S16• Movie S17• Movie S18• Movie S19• Movie S20• Movie S21• Movie S22• Movie S23• Movie S24• Movie S25• Movie S26• Movie S27• Movie S28• Movie S29• Movie S30• Movie S31• Movie S32• Movie S33• Movie S34• Movie S35• Movie S36• Movie S37• Movie S38• Movie S39• Movie S40• Movie S41• Movie S42• Movie S43• Movie S44• Movie S45• Movie S46

presence of anomalous charge structures [Lyons et al., 2008; Lang et al., 2010]; that is, the presence of positivecharge layers near midlevels (e.g., �20°C) rather than the more common negative charge [Lang et al., 2004;Rust et al., 2005; Wiens et al., 2005; Bruning et al., 2014; Fuchs et al., 2015]. Moreover, it is not strictly the casethat an MCS is required for sprites; they have been observed over smaller (i.e., sub-MCS) storms [Lyons et al.,2008; São Sabbas et al., 2009; van der Velde et al., 2014], including winter storms [Takahashi et al., 2003; Suzukiet al., 2011].

These exceptions demonstrate the limits of our current understanding concerning the meteorology ofsprites. Moreover, studies of most sprite-producing storms often lack a comprehensive data set includingdetailed information on storm kinematics and microphysics, such as can be provided by polarimetric andmulti-Doppler radar networks. In addition, while increasing emphasis has been placed on three-dimensional(3-D) mapping of individual sprite-parent CGs [Lyons et al., 2003; Lang et al., 2010, 2011; Lu et al., 2013; van derVelde et al., 2014], relatively little focus has been placed on the evolution of total lightning relative to spriteproduction, nor has storm tracking been implemented to characterize the full temporal evolution of theconvection in sprite-producing systems. This would be especially helpful with smaller, sub-MCS storms whereambiguity in convective state can be reduced due to the presence of fewer cells. Overall, this informationdeficit prevents the development of a fully realized theory of sprite production relative to thunderstormstructure and evolution.

In order to make further progress on these questions, this study presents observations of two sprite-producing storms that occurred in Colorado during the Deep Convective Clouds and Chemistry (DC3) fieldcampaign [Barth et al., 2014]. The first occurred on 8 June 2012 and produced 21 observed sprites overa 2 h period (0500–0700UTC). The second storm, on 25 June, produced 26 observed sprites over a 1 hperiod (0415–0515UTC). These storms were observed by a network of sprite-observing camera systems,multiple Doppler radars (including one polarimetric radar), a 3-D lightning mapping array (LMA), andnational CG- and CMC-observing networks. This enabled one of the most comprehensive examinationsof the relationships between sprite production and thunderstorm evolution to date. These storms featuredanomalous charge structures, and the observations clearly indicated the influence of convective strengthon sprite production in mature storms. The data also will show how the microphysical, kinematic,and electrical evolution of these storms strongly influenced the structure, location, and timing of sprite-parent CGs.

2. Data and Methodology2.1. Sprite Cameras

Optical observations of sprites were made on 8 and 25 June 2012 frommultiple locations via a variety of cam-eras. The main camera type that was used in this study was the Watec 902H, which is reviewed extensively inLu et al. [2013]. These are steerable, low-light video cameras. They are monochrome and record either con-tinuously or via an automated trigger on an optical transient. Time stamping is provided by a connectedGlobal Positioning System antenna. For the 8 and 25 June cases, the cameras were steered manually andrecorded observations continuously. Sprites were identified during later manual analysis of the video. Amodified digital single-lens reflex (DSLR) camera with the near-infrared filter removed was used to performautomated, long-exposure color photography of sprites and other TLEs.

On 8 June 2012, Watec cameras near Lamy, New Mexico, Bennett, Colorado, and on Cerro Pelado mountain,New Mexico, operated and captured sprites. On 25 June 2012, a DSLR and Watec at Lamy and a Watec inBennett were available. Watecs captured all but one sprite on that day, and the DSLR captured the othersprite. See Figure 1 in Lu et al. [2013] for a map of the sprite-observing network. Phantom v7.3 high-speedimagers were available in the network, but not used in this study.

Time-stamped sprite video and imagery were matched to parent CG lightning (data set described later) viacomparison of times and locations (i.e., azimuth from a given camera) of occurrence, following standard prac-tices [e.g., Lu et al., 2013]. Timing for the DSLR-only sprite was matched to a +CG in the camera’s field of viewthat occurred during the 3 s exposure window. Sprite durations were available via manual video analysis on 8June, but this analysis was not done for 25 June. This study mainly used the sprite observations to establishtime and approximate location of occurrence, relative to storm evolution.

Journal of Geophysical Research: Atmospheres 10.1002/2016JD025299

LANG ET AL. TWO SPRITE-PRODUCING STORMS IN COLORADO 9676

Correspondence to:T. J. Lang,timothy.j.lang@nasa.gov

Citation:Lang, T. J., W. A. Lyons, S. A. Cummer,B. R. Fuchs, B. Dolan, S. A. Rutledge,P. Krehbiel, W. Rison, M. Stanley, andT. Ashcraft (2016), Observations of twosprite-producing storms in Colorado,J. Geophys. Res. Atmos., 121, 9675–9695,doi:10.1002/2016JD025299.

Received 29 APR 2016Accepted 25 JUL 2016Accepted article online 29 JUL 2016Published online 26 AUG 2016

©2016. American Geophysical Union.All Rights Reserved.

2.2. Radars

There were two research radars and two operational radars of most relevance to this study. The ColoradoState University-University of Chicago-Illinois State Water Survey (CSU-CHILL) radar is a dual-wavelength,S/X-band, polarimetric Doppler radar with an 8.5m, dual-offset, Gregorian antenna [Bringi et al., 2011;Junyent et al., 2015], located near Greeley, Colorado. CSU-Pawnee was an S-band Doppler radar situated47.7 km NW of CSU-CHILL. Together, they formed a dual-Doppler network that has been frequently usedto study combined kinematic and microphysical characteristics of storms [e.g., Lang and Rutledge, 2002;Dolan and Rutledge, 2007; Basarab et al., 2015]. Additional coverage was provided by U.S. National WeatherService Doppler radars near Denver, Colorado (KFTG), and Cheyenne, Wyoming (KCYS). These S-bandDoppler radars can often contribute to multi-Doppler syntheses because of their close proximity to CSU-CHILL and CSU-Pawnee [Dolan and Rutledge, 2007].

The CSU radars were operated from ~2100UTC on 7 June 2012 through ~0440UTC on 8 June. A major focusof the scanning was coordinated plan position indicator (PPI) sectors to support multi-Doppler syntheses. Inaddition, CSU-CHILL performed frequent range-height indicator (RHI) scan volumes during this time period.On 24–25 June 2012, CSU-CHILL radar was the only research radar operated, and it primarily scanned stormsin RHI mode from ~2030UTC to ~0450UTC. When performing RHIs, CSU-CHILL would focus on volumetricscanning with 1–6min updates, as opposed to high temporal resolution updates along a single azimuth.On both days KFTG and KCYS scanned full 360° PPI volumes, from 0.5° up to 19° elevation.

Multi-Doppler synthesis was performed for the 8 June 2012 case. Prior to gridding and analysis, velocitydata from all radars were first manually dealiased, and nonmeteorological artifacts were removed using amixture of thresholds on polarimetric variables (e.g., correlation coefficient) and manual editing. Then thedata were interpolated to a common Cartesian grid using the National Center for Atmospheric Research(NCAR) Sorted Position Radar Interpolator (SPRINT) software [Mohr and Vaughn, 1979; Miller et al., 1986].Resolution of the grid was 1 km in all dimensions, and the vertical coordinate stretched from 2.5 to18.5 km above mean sea level (msl; CSU-CHILL is at 1.43 km msl). Wind analyses were performed on themerged data set using the Custom Editing and Display of Reduced Information in Cartesian Space(CEDRIC) software [Mohr and Miller, 1983; Mohr et al., 1986]. In this study we present only the off-baselinehorizontal wind results in and near the stratiform precipitation, as the primary convective core was alongthe CHILL-Pawnee baseline during our focus period for wind analysis (04–05UTC). Radars were used inan analysis if their scans started within 3min of one another, and analyses were available roughly every6min during 0200–0440UTC.

Reflectivity data from all available National Weather Service (NWS) radars are also regularly included in 3-Dnational mosaics [Zhang et al., 2011]. During 2012, these mosaics (now called Multi-Radar/Multi-Sensor orMRMS) were available every 5min, at 0.01° latitude/longitude resolution, and with a vertical coordinate thatstretched from 0.25 km to 18.0 km [Lang et al., 2014]. Vertical resolution varied from 0.25 km near the surfaceto 2 km near the top. For the domain of this study, KFTG and KCYS were the primary radars that contributed tothese mosaics in the study domain. These mosaics provided volumetric information on all storms in theregion, since the research radars often were narrowly focused in their scanning.

2.3. Colorado Lightning Mapping Array

The Colorado Lightning Mapping Array (COLMA) is a very high frequency (VHF)-based network that mapsflashes in three dimensions plus time [Rison et al., 1999; Lang et al., 2014; Fuchs et al., 2015]. During June2012 the COLMA consisted of 15 separate stations, spread throughout northeastern Colorado. During mostof the analysis periods, the storms in this study were well within the 100 km distance of network centerthat is generally considered optimal for 3-D analysis [MacGorman et al., 2008]. As is standard practice, aminimum of seven station detections and a chi-square error of 1 or less were required for a locationsolution to be considered valid [e.g., Lang and Rutledge, 2011]. Individual VHF sources were clustered intoflashes via the algorithm developed by Fuchs et al. [2015], which required no more than 3 km distance and0.15 s time between individual sources, for them to be considered part of the same flash. Flash energywas computed following the methodology of Bruning and MacGorman [2013]. For more informationabout the performance of COLMA during June 2012, see Lang et al. [2014], Basarab et al. [2015], andFuchs et al. [2015].

Journal of Geophysical Research: Atmospheres 10.1002/2016JD025299

LANG ET AL. TWO SPRITE-PRODUCING STORMS IN COLORADO 9677

2.4. National Lightning Detection Network

The National Lightning Detection Network (NLDN) is a well-known lightning network that in 2012 primarilydetected cloud-to-ground (CG) lightning with a detection efficiency at the flash level of 95% or better anda location accuracy better than 0.5 km [Cummins and Murphy, 2009]. Flash-level data were used in this study.In order to filter out possible ICs, a 15 kA threshold was already applied to the positive CG flashes beforereceipt of the data [Biagi et al., 2007]. Prior to analysis we also applied the same magnitude threshold tothe negative CG data. This was because in anomalously charged storms, the negative CG population isdisproportionately filled with low-peak current flashes, and these are likely misclassified ICs [Montanyàet al., 2009; Lang et al., 2015]. This was also observed in the present study.

2.5. Charge Moment Change Network

In June 2012, the Charge Moment Change Network (CMCN) consisted of two stations, one near Fort Collins,Colorado, and one near Duke University Forest in North Carolina [Cummer et al., 2013]. The stations consist oforthogonal induction magnetic field sensors that are sensitive to signals from 2Hz to 25 kHz or from extre-mely low frequency (ELF) to very low frequency (VLF). Via specialized processing detailed in Cummer et al.[2013] and Beavis et al. [2014], the impulse charge moment change (iCMC) from individual CG strokes canbe determined, with a factor of 1.5 accuracy. The iCMC is the charge moment change within the first 2msof the return stroke. For this study we focused only on the strokes of either polarity with iCMC> 75 C km.Geolocation of these large-iCMC strokes was done via comparison with the NLDN, following the methodol-ogy described in Cummer et al. [2013].

2.6. Soundings

Regular National Weather Service (NWS) soundings from the Denver station (No. 72469) and the North Platte,Nebraska, station (No. 72562) were analyzed to determine environmental characteristics near the stormsstudied in this paper. Parameters such as convective available potential energy (CAPE), convective inhibition(CIN), lifting condensation level (LCL), level of free convection (LFC), equilibrium level (EL), shear, andstorm-relative helicity (SRH) were analyzed by considering the most unstable parcel (for the variables wherethat is relevant; e.g., CAPE). The analysis package used for these calculations was the Sounding/HodographAnalysis and Research Program in Python (SHARPpy) [Halbert et al., 2015]. Mobile soundings were availableduring the DC3 project, but none were launched on 8 or 25 June.

2.7. Colorado State University (CSU) Lightning, Environment, Aerosol, and Radar Analysis Framework

The CSU Lightning, Environment, Aerosol, and Radar (CLEAR) analysis framework provides a useful means ofintegrating lightning and radar data for this study. While the details of the framework are covered in Langand Rutledge [2011] and Fuchs et al. [2015], a brief synopsis of how CLEAR was used for this study is includedhere. We used the Fuchs et al. [2015] CLEAR-processed data set, which merged MRMS radar features withCOLMA and NLDN. Radar features were identified using two reflectivity thresholds applied to a 2-D compo-site reflectivity field. As in Fuchs et al. [2015], a minimum area threshold of 20 km2 was applied to the 30 dBZcontour, and a threshold of 13 km2 was applied to the 40 dBZ contour, when identifying radar features.This preferentially selected more intense, mature convective cells. By using these thresholds, CLEARdid not include distinct stratiform echo regions when producing time series analyses. Since this studywas focused on the convective behavior of multicellular sprite-producing storms, these thresholds areconsidered appropriate.

LMA flashes were linked to a radar feature if they initiated within it or if not, then to the closest radar featurewithin 10 km of the initiation point. If an NLDN flash was located within the horizontal area of a radar feature,it was matched to that feature. If it was located outside a feature, then it was assigned to the closest featurewithin 10 km distance. If it occurred beyond 10 km, it was not linked to any feature. CMCN-detected large-iCMC strokes were associated with radar features identically to how NLDN flashes were. Sprites were asso-ciated with the radar features that were determined by CLEAR to have produced their NLDN-detectedparent +CGs.

Once linked to individual radar features, which in turn were grouped together in time to form a track thatrepresented the evolution of the main storm producing sprite-parent flashes, all lightning and radar observa-tions were analyzed to produce time series for these storms. CLEAR is capable of automated feature tracking

Journal of Geophysical Research: Atmospheres 10.1002/2016JD025299

LANG ET AL. TWO SPRITE-PRODUCING STORMS IN COLORADO 9678

to enable this time series development [Fuchs et al., 2015]. However, in order to ensure maximum accuracy,and because we were only focused on two cases, the storms in this study were tracked manually by notingthe CLEAR identification numbers for each cell in a radar volume and creating a list of all the cells associatedwith the particular storm at a particular time. Examples of how this worked are shown in Figure 1 for the 8June 2012 storm and in Figure 2 for the 25 June 2012 storm. For example, at 0300UTC on 8 June 2012(Figure 1a) cell number 2732 comprised the main storm. By 0600UTC, the CLEAR-assigned cell number forthe main storm had changed and was now 2956. The manual tracking had to take account of these changeswhen they occurred, and assigning cells to the sprite storms was done via visual inspection and animationsusing plots like these.

At times, the storm of interest consisted of more than one CLEAR-identified radar feature in a particularvolume. When this occurred, metrics for each cell were included in time series analyses. Subjective deci-sions about which cells to include took into account whether cells eventually merged into the main stormor split from it. If cells merged into the main storm, they were kept in the analysis prior to the merger.However, when splits occurred, the smaller cell(s) that split away were not kept if they moved away fromthe main convective envelope associated with the storm. Generally, it was relatively straightforward toidentify the cell(s) responsible for the main storm of interest, and sensitivity studies found that includingor excluding nearby smaller and shorter-lived cells did not make a significant difference on the timeseries results.

Figure 1. Composite reflectivity from the MRMS mosaic valid at several times on 8 June 2012. Individual CLEAR-tracked cells are indicated by the numbers, with thedashed lines showing the convex hulls associated with each cell and the grey curves showing the fitted ellipses to each hull. Cells 2732 and 2956 are the primary cellsassociated with the sprite-producing storm. (a) 0300 UTC. (b) 0400 UTC. (c) 0500 UTC. (d) 0600 UTC.

Journal of Geophysical Research: Atmospheres 10.1002/2016JD025299

LANG ET AL. TWO SPRITE-PRODUCING STORMS IN COLORADO 9679

3. Results3.1. The 8 June 2012 Storm

The 0000UTC Denver sounding on 8 June is shown in Figure 3a (several derived sounding parameters arelisted in Table 1). The atmosphere was particularly unstable on this day, with the most unstable parcelmethod yielding 2662 J kg�1 of CAPE. In addition, significant shear and SRH were reported. This led to thedevelopment of a severe tornadic, multicellular hailstorm. Prior to the sprite analysis period (which wasroughly 0300–0700UTC; Figure 1), cells in Weld County, Colorado (the primary focus area of the DC3 cam-paign in Colorado), had produced multiple large hail reports (up to ~6 cm in diameter) as well as a tornadoreport. There were also wind gust reports up to ~40m s�1 [Storm Prediction Center, 2012a]. Even during thesprite analysis time, the storm was still strong, with multiple hail reports up to ~4.5 cm before 0500UTC aswell as severe-scale winds. After passing near the CSU-CHILL radar between 0430 and 0500UTC, one ofthe authors (B. Fuchs) noted significant hail accumulations leading to impassable roads in the storm’s wake.

Around 0300UTC (Figure 1a), the storm consisted of multiple discrete cells and was northwest of the Pawneeradar. These cells gradually coalesced until they formed a contiguous convective region by 0500UTC(Figure 1c). As this occurred, the storm became larger and more intense in terms of reflectivity until justbefore 0430UTC, when it began a long-term decline in size and intensity through 0600UTC (Figure 4a).This weakening was notably interrupted by a brief convective burst around 0450–0500UTC, which wasobserved in total flash rate and total flash energy (Figure 4b), and coincided with an upward shift in LMA

Figure 2. Composite reflectivity from the MRMSmosaic valid at several times on 25 June 2012. Individual CLEAR-tracked cells are indicated by the numbers, with thedashed lines showing the convex hulls associated with each cell and the grey curves showing the fitted ellipses to each hull. Cells 3670, 3680, 3696, and 3707 are theprimary cells associated with the sprite-producing storm. (a) 0300 UTC. (b) 0400 UTC. (c) 0500 UTC. (d) 0600 UTC.

Journal of Geophysical Research: Atmospheres 10.1002/2016JD025299

LANG ET AL. TWO SPRITE-PRODUCING STORMS IN COLORADO 9680

source density altitude (Figure 4d). Immediately after this burst subsided, the storm began producing most ofits observed sprites, roughly between 0500 and 0610UTC (Figure 4b). Spikes in mean flash area occurredwithin this time frame as well.

Starting during 0550–0600UTC, the storm intensified again, in terms of reflectivity structure (especially 40and 50 dBZ volumes; Figure 4a), total flash rate, and flash energy (Figure 4b). With one exception (near0700UTC), the sprite production ended when the peak of the reintensification was achieved. The reintensi-fication also featured enhanced source activity at lower altitudes (Figure 4d). Even at its lowest value,however, the total flash rate in this storm was ~60min�1. Visually, from a position west of the storm inFort Collins, Colorado, the first author (T. Lang) observed essentially continuous lightning as the stormmovednorth to south through Weld County.

Throughout this period the storm wasproducing very low CG flash rates, rarelyabove 10 per 5min period, marking it asa high IC:CG ratio storm common to theEastern Plains of Colorado [Lang et al.,2000; Boccippio et al., 2001; Lang andRutledge, 2002]. The majority of theseground flashes consisted of +CGs, andthe relative fraction of +CGs was highestduring the most intense periods of thestorm (roughly 0330–0430 and 0600–0700UTC; Figure 4c). If no 15 kA filterwas applied, then �CGs were 5–10times more frequent than +CGs, and

Figure 3. Skew-T/log-p sounding plots from Denver, CO. (a) 0000 UTC on 8 June 2012. (b) 0000 UTC on 25 June 2012.

Table 1. Derived Parameters From the Denver, CO, Sounding MostRelevant for the Two Colorado Sprite Casesa

Variable 6/8/2012 0000 UTC 6/25/2012 0000 UTC

LCL (m msl) 3,502 6,239LFC (m msl) 3,785 6,401Melting Level (m msl) 4,278 4,267EL (m msl) 14,021 11,417CAPE (J kg�1) 2,662 190CIN (J kg�1) �27 �0.3Lifted Index �6.7 0.70–6 km shear (m s�1) 15.4 13.20–3 km SRH (m2 s�2) 64.5 33.4

aStation elevation is 1625m msl.

Journal of Geophysical Research: Atmospheres 10.1002/2016JD025299

LANG ET AL. TWO SPRITE-PRODUCING STORMS IN COLORADO 9681

Figure 4. (a) Time series of echo volumes associated with various MRMS reflectivity thresholds, for the 8 June 2012 storm. (b) Time series of LMA-derived flash rate,mean flash area, and total flash energy for this storm. Also indicated are the times of the observed sprites. (c) Time series of NLDN-derived CG flash rates in this storm,separated by polarity, as well as the times of strokes that produced a positive iCMC above 75 C km. A 15 kA peak current threshold has been applied to the NLDNdata. (d) Time-height plot of LMA source density in the 8 June 2012 storm.

Journal of Geophysical Research: Atmospheres 10.1002/2016JD025299

LANG ET AL. TWO SPRITE-PRODUCING STORMS IN COLORADO 9682

largely tracked LMA total flash rates (not shown), similar to the El Reno storm studied by Lang et al. [2015].This behavior suggested that these low-peak current �CGs were likely misclassified ICs.

Large iCMCs (all positive) were not recorded until after 0600UTC. This is because power and Internet outagesduring the storm prevented the reporting of the CMCN sensor at Yucca Ridge Field Station (YRFS; near FortCollins, Colorado) to the national iCMC data set during 03–06UTC. Manual analysis of the CMCN data forselect sprite-parent strokes during 05–06UTC indicated that large-CMC strokes were occurring at this time,however. Based on that analysis, at least four large-iCMC strokes occurred during 0513–0525UTC [Lyonset al., 2012]. This is likely a very significant underestimate of the pre-0600UTC large-iCMC rate, however.Values for large-iCMC strokes after 0600UTC ranged between 75 C km and 273 C km.

Overall, the lightning behavior was indicative of an anomalously electrified storm [Bruning et al., 2014]. Thisinference was based on a number of observations. First was the high percentage of positive CGs (77.6% overthe analysis period), as well as the fact that only positive large-iCMC strokes were observed. Second, therewas a midlevel (6–8 km or approximately �10 to �20°C based on the Denver sounding) LMA source densitymaximum during much of the analysis period (Figure 4d), which has been linked to the presence of a mid-level positive charge [Lang and Rutledge, 2011; Fuchs et al., 2015]. Finally, manual analysis (following themethodology of Wiens et al., 2005) of select periods of COLMA data (not shown) indicated that small andinverted IC discharges (between midlevel positive and upper level negative charge) were frequently occur-ring near the strongest convection.

However, the charge structure in this storm featured substantial temporal and spatial heterogeneity. Forexample, after 0430UTC the LMA activity shifted upward and a midlevel maximum did not return until thestorm reintensified after 06 UTC (Figure 4d). In addition, between 0430 and 0530UTC the relative fractionof +CG lightning fell (Figure 4c). Moreover, lightning in the region where sprite-parent flashes occurred didnot typically feature inverted characteristics.

Sprite-parent flashes, without exception, occurred on the downshear (eastern) side of the convection, whichafter 0500UTC started to exhibit a quasi-linear north-south structure. These larger flashes tended to initiatemore than 10 km away from the most intense portion of the convection to the southwest, which featured fre-quent, small, and often inverted IC flashes. This pattern is reminiscent of the flash-size spectra results ofBruning and MacGorman [2013]. Figure 5 shows a very typical example of a sprite-parent flash in this storm.Near the convection, the flash had the classic “I” structure common in normal bilevel IC flashes (Figure 5b)[Rison et al., 1999], with inferred positive charge near 9 km msl (roughly �40°C in Figure 3a) and inferrednegative charge near 6 km msl (roughly �10°C). The flash then propagated eastward along the downwardsloping upper positive charge layer, behavior that is very typical of normal sprite parents [Lang et al.,2010]. During the sprite there was LMA-mapped activity within this positive charge layer. The parent +CGcame to ground just downshear of the convective line, while the majority of the flash activity occurred welleast, within the stratiform precipitation. As the storm was traveling from the north-northwest to the south-southeast, this was technically the forward flank of the storm. This is a departure from common sprite-parentbehavior in leading-line/trailing-stratiform MCSs, where stratiform lightning typically travels rearward andthus opposite of the direction of motion [Lang et al., 2010]. The closest analog to the 8 June case would bea cross between a parallel-stratiform and a leading-stratiform MCS [Parker et al., 2001].

Interestingly, late in the flash’s lifetime there is evidence for a negative charge region above the positiveregion, with recoil activity near 12 km msl in the stratiform region (Figures 5b and 5d). Inferred high-altitudenegative charge also was commonly observed closer to convection in some (nonsprite-parent) flashes (notshown), so this flash may have revealed that this layer also advected eastward. Figure 6a shows the time-range development of this flash, following the analysis style of van der Velde et al. [2014]. Here the bidirec-tional leader development during much of the flash is readily apparent. Multiple negative leaders (the steeplinear features in Figure 6a, with speeds ~105m s�1) occur in succession, both before and after the sprite-parent +CG. This negative development often coincides with slower-moving positive leaders (~104m s�1),which are revealed by the less sloped recoil activity that is located closer to the flash origin. The activityassociated with the upper level negative charge layer is readily apparent after 1 s and largely consistsof two high-altitude positive leaders separated in time by a lower altitude negative leader. Additionalsprite-parent flash animations can be found in the supporting information, and some of these flashes alsoindicated eastward advection of the lower negative charge region into the stratiform region.

Journal of Geophysical Research: Atmospheres 10.1002/2016JD025299

LANG ET AL. TWO SPRITE-PRODUCING STORMS IN COLORADO 9683

The CSU-CHILL was operational until about 0440UTC, prior to the first sprite observations. In addition, thePawnee radar shut down around the same time. However, prior to this CSU-CHILL was able to contributeto multi-Doppler syntheses, via combination with the Pawnee, KFTG, and KCYS radars. Figure 7 shows the lastsynthesis (valid 0423UTC) before CSU-CHILL terminated operations. At this timemuch of the stormwas alongthe radar baselines and was very nearly on top of CSU-CHILL. Thus, we do not attempt to consider verticalwind. However, the horizontal winds were still useful away from the baselines. They indicated strong westerlyto northwesterly flow (Figure 7a), with increasing vertical shear in the zonal component of the wind (Figure 7b), particularly on the eastern flank of the storm. This suggests that the sprite-parent flashes were respondingto the eastward motion of the charged hydrometeors detraining from near the top of the storm, enabled bythe strong vertical shear relative to storm motion (Figure 3a and Table 1).

3.2. The 25 June 2012 Storm

The 0000UTC Denver sounding from 25 June 2012 is shown in Figure 3b, with derived parameters listed inTable 1. The atmosphere on this day was significantly different than it was on 8 June. It was much drier, withthe lapse rate approximating a dry adiabat up to the LCL, which was nearly 2800m higher than it was on 8June. This led to the melting level being below the LCL and thus zero expected warm-cloud depth. CAPEwas greatly reduced, only 190 J kg�1, and the equilibrium level was also lower than 8 June. While shear

Figure 5. Typical example of a sprite-parent +CG during the 8 June 2012 storm. The flash occurred at the indicated time, with radar data coming from 0505 UTCMRMS mosaic. (a) Plan view with composite radar reflectivity (filled contours) and lightning data. The pink dots show the LMA sources from 50ms prior to thesprite-parent +CG through the end of the sprite, and the black dots are all other LMA sources during the flash. The red dashed crosshairs indicate the vertical crosssections in Figures 5b and 5c. (b) Vertical cross section through the indicated constant latitude. (c) Vertical cross section through the indicated constant longitude. (d)Time-height plot of the flash.

Journal of Geophysical Research: Atmospheres 10.1002/2016JD025299

LANG ET AL. TWO SPRITE-PRODUCING STORMS IN COLORADO 9684

was comparable to 8 June, the SRH was reduced by about 50%. Since the main storm on this day occurrednear the Nebraska border, the North Platte sounding at 0000UTC also was checked. Though this stationwas much farther from the storm than Denver (~370 km versus ~120 km), its sounding was moister andsignificantly more unstable, with the LCL reduced to 2890mmsl and CAPE increased to 3970 J kg�1. This gra-dient suggested that the storm likely had access to a more favorable convective environment than would beexpected based on the Denver sounding alone. In fact, an NWS analysis at 2336UTC on 24 June indicated asurface boundary and CAPE gradient in northeastern CO, near where this storm developed [Storm PredictionCenter, 2012b]. This storm occurred during the High Park fire, which flared up andwas reported on 9 June andwas not fully contained until July [Lang et al., 2014]. Thus, the region was affected by significant amounts ofsmoke aerosol [Barth et al., 2014]. Photographs by one of this study’s authors (W. Lyons) indicated that severalclouds were ingesting smoke aerosol on this day.

An intense storm developed during the local afternoon hours, northeast of CSU-CHILL near where Wyoming,Nebraska, and Coloradomeet. A westward moving outflow boundary from this convection merged with diur-nally forced storms moving off the Rocky Mountain foothills and formed a storm (Figure 2) that persisted intothe evening as it moved north-northeastward from Colorado into Wyoming. The period for which this stormwas analyzed was 0300–0700UTC. As it evolved, between 0400 and 0500UTC it transitioned from a small,quasi-linear collection of cells to a more unicellular system (Figure 2).

Interestingly, this consolidation process was coincident with a significant overall decrease in the convectivestrength, as indicated by reflectivity volumes (Figure 8a). This decrease in strength was also indicated by along-lived secular decline in both flash rate and flash energy (Figure 8b) from their relative maxima around0315UTC. However, after 0400UTC, mean flash area began to increase even as flash rate declined towarda short-lived relative minimum at 0415UTC. The first two sprites were observed at 04:18:24 UTC and then

Figure 6. Horizontal distance versus time for COLMA sources observed in two sprite-parent flashes. Coloring is by sourcealtitude. Reference location for horizontal distance is the median location of the first 10 sources in each flash. Eachsprite-parent +CG is indicated by a black cross. The dashed lines indicate 104 and 105m s�1 leader speeds. (a) Sprite parentoccurring around 05:05:07 UTC on 8 June 2012. (b) Sprite parent occurring around 04:18:23 UTC on 25 June 2012.

Journal of Geophysical Research: Atmospheres 10.1002/2016JD025299

LANG ET AL. TWO SPRITE-PRODUCING STORMS IN COLORADO 9685

again at 04:21:02 UTC. Total flash rate continued to decrease after 0425UTC to an absolute minimum of 46flashes during 0505–0510UTC (9.2min�1). Between 0437 and 0511UTC, during this total flash rate lull, therewas a burst of 24 observed sprites. The rapid production of sprites was coincident with a large spike in meanflash area, which peaked above 1500 km2 at 0455UTC. The occurrence of large flashes helped drive anincrease in flash energy around this time as well. Total flash rate recovered as sprite production terminated,although the storm never regained its pre-0400UTC intensity.

This storm also was anomalously charged. During the analysis period 92.1% of the filtered CG lightning waspositive (if no 15 kA filter were applied, the positive and negative rates were comparable to one another). Incontrast to 8 June, the greatest production of +CGs occurred during its lull period, between roughly 0400and 0600UTC (Figure 8c). Peak +CG rates were about twice the number on 8 June as well (compareFigures 4c and 8c). Large positive-iCMC strokes were most frequent during this lull as well, especially themainsprite period (0437–0511UTC). The values for large-iCMC strokes in this storm ranged between 75 and290 C km. Throughout the analysis, LMA source densities were maximized near 7 km msl (approximately�15°C; Figure 8d). Manual analysis of the LMA data [Wiens et al., 2005] indicated frequent in-cloud activitybetween midlevel positive charge and a negative layer above it. This is a classic anomalous storm signature[Lang et al., 2004;Wiens et al., 2005; Lang and Rutledge, 2011; Fuchs et al., 2015]. The burst of sprites occurredafter a sharp decline in source activity.

Figure 7. (a) Horizontal cross section of merged reflectivity (filled contours) at 7.5 kmmsl, from the multi-Doppler synthesis at 0423 UTC on 8 June 2012. Also shownare ground-relative horizontal wind barbs (black, m s�1). Wind barb density is thinned by a factor of 50, for clarity. The red dashed line denotes the cross section inFigure 7b, and the CSU-CHILL radar location is indicated by the white diamond. (b) Vertical cross section of the zonal wind (U) along 30 km north of CSU-CHILL.CSU-CHILL, along with the KCYS and KFTG NEXRAD radars, contributed to this synthesis.

Journal of Geophysical Research: Atmospheres 10.1002/2016JD025299

LANG ET AL. TWO SPRITE-PRODUCING STORMS IN COLORADO 9686

Figure 8. (a) Time series of echo volumes associated with various MRMS reflectivity thresholds, for the 25 June 2012 storm. (b) Time series of LMA-derived flash rate,mean flash area, and total flash energy for this storm. Also indicated are the times of the observed sprites. (c) Time series of NLDN-derived CG flash rates in this storm,separated by polarity, as well as the times of strokes that produced a positive iCMC above 75 C km. A 15 kA peak current threshold has been applied to the NLDNdata. (d) Time-height plot of LMA source density in the 8 June 2012 storm.

Journal of Geophysical Research: Atmospheres 10.1002/2016JD025299

LANG ET AL. TWO SPRITE-PRODUCING STORMS IN COLORADO 9687

Analysis of individual sprite-parent flashes indicated that they were inverted in structure, and occurred in ornear convection, with leaders often tapping regions within the surrounding storm anvil. Figure 9 shows anXLMA-style plot [Rison et al., 1999] of a typical sprite parent in this storm. The flash in question was the firstsprite parent, which initiated at 04:18:23 UTC and lasted about 2 s. The flash struck ground very near its initia-tion point, during a tightly clustered burst of LMA source activity. The subsequent activity indicated thatnegative leaders spread in all directions after the flash struck ground, though they traveled farthest to thesoutheast, toward the CSU-CHILL radar. Figure 6b shows the time-range behavior of this flash. It indicates thatthe sprite-parent +CG marked a transition toward a single negative leader that neutralized positive charge

Figure 9. LMA observations (dots colored by time) of the first sprite-parent +CG in the 25 June 2012 storm (04:18:24 UTC).In all subplots the black cross indicates the NLDN-detected +CG. (a) Time-height evolution. (b) Altitude versus longitudediagram, overlaid on a vertical cross section of MRMS radar reflectivity from 04:20 UTC. (c) LMA source altitude frequencyhistogram. (d) Plan view of the flash overlaid on MRMS composite radar reflectivity. The diamond indicates the CSU-CHILLradar, and the dashed black line is the 328° azimuth of the RHI shown in Figure 10. The dashed red lines indicate theradar cross sections in Figures 9b and 9e. (e) Latitude versus altitude diagram, overlaid on a vertical cross section of MRMSradar reflectivity.

Journal of Geophysical Research: Atmospheres 10.1002/2016JD025299

LANG ET AL. TWO SPRITE-PRODUCING STORMS IN COLORADO 9688

within the cloud during the TLE. This simpler behavior contrasts with the more complex bidirectional devel-opment that was common on 8 June (Figure 6a).

Up until 0450 UTC, the CSU-CHILL radar performed a series of RHI volumes, providing detailed verticalstructure information as the storm was producing sprites. Figure 10 shows a relevant sweep near the timeof the flash, through the center of the storm. The storm was comparable to 8 June in terms of verticalstructure, with echo tops near 14 km above ground level (~15.5 kmmsl). The core contained a classic meltinghail signature [Ryzhkov et al., 2013], with hydrometeor identification indicating a transition from hail to heavyrain with large drops (Figure 10f) as the precipitation fell from the storm core. This was associated with anincrease in differential reflectivity below the melting level (Figure 10b), as well as strong propagation andspecific differential phase shifts near the ground (Figures 10d and 10e). There were also alternating positive

Figure 10. RHI sweep at 328° azimuth, from the CSU-CHILL radar volume starting at 0413 UTC. In all subplots, the black dots show the LMA observations of the sprite-parent flash shown in Figure 9, the white diamond is the median position of the first 10 LMA sources in the flash, and the black cross is the sprite-parent +CG. (a)Reflectivity. (b) Differential reflectivity. (c) Dealiased radial velocity. (d) Specific differential phase. (e) Differential phase shift. (f) Hydrometeor identification. Note thatthe vertical axis is above radar altitude (1432m msl), not msl.

Journal of Geophysical Research: Atmospheres 10.1002/2016JD025299

LANG ET AL. TWO SPRITE-PRODUCING STORMS IN COLORADO 9689

and negative differential phase shifts above the melting level, likely associated with the presence of dendriticcrystals [Thompson et al., 2014] as well as crystals vertically aligned by strong ambient electric fields [Krehbielet al., 1996; Carey and Rutledge, 1998].

The flash initiated to the southwest of this melting hail core (Figure 9), in a region where the RHIs indicatedlimited hail but considerable volume of low-density graupel (not shown). Despite the less intenseprecipitation structure, the upper level diffluence and midlevel confluence pattern in Doppler velocity(Figure 10c) were more pronounced toward the southwest near the center of the flash (particularly near320° azimuth from CSU-CHILL), and an overshooting top was observed between the flash initiation pointand the hail core shown in Figure 10. This indicates that the sprite-parent nonetheless initiated near astrong updraft.

Animations of all the sprite-parent flashes from this storm (available in the supporting information) show avery similar pattern to the flash in Figures 9 and 10. The flashes initiated near, but not in, the strongestprecipitation cores. They showed no strongly preferred region of initiation, however, with slightly morethan half initiating on the northern side of the storm and the rest the southern side. In addition, theirhorizontally propagating negative leaders (likely helping drive continuing current in the sprite-parent+CG) showed no preferred direction of travel. This is consistent with the weaker upper level flowcompared to 8 June (Figure 3) and the more omnidirectional anvil structure on 25 June (e.g., Figures 2cand 10a).

4. Discussion

Lyons et al. [2008] demonstrated that sprites could occur over anomalously charged warm-season storms, butthe present study is the first to robustly document this phenomenon. Both the 8 and 25 June 2012 stormswere anomalous, with LMA-inferred positive charge near �10 to �20°C, in contrast to the negative chargethat is commonly observed [Rust et al., 2005]. It is, of course, apt that such sprite-producing storms wouldbe observed in Colorado, as the presence of anomalous charge structures is very common in this region[Lang and Rutledge, 2011; Fuchs et al., 2015]. The reasons for this are not totally clear, but the preferencefor higher cloud bases in and near Colorado (and hence reduced warm-cloud depths) likely plays a major role[Williams et al., 2005; Fuchs et al., 2015], especially when combined with strong instability and shear [Lang andRutledge, 2011].

The hypothesis is that such a combination would produce large and vigorous updrafts, leading to increasedliquid water contents (LWCs) [Williams et al., 2005]. This would tend to favor net positive charge transfer toriming particles (e.g., graupel) undergoing collisions with ice crystals [Saunders and Peck, 1998], as opposedto negative charge that would be expected under more typical LWCs [Takahashi, 1978]. Since the rimingparticles are heavier, they would fall relative to the ice crystals, leading to negative charge layer overlayingpositive charge [Bruning et al., 2014] and thus an inverted electrical structure.

While the electrical structures of this study’s storms were complicated and featured substantial temporaland spatial heterogeneity (e.g., Figure 4d), the process described above is supported by the weight of theavailable evidence. Detailed updraft information was not available during the analyzed time periods,but both storms featured strong horizontal (Doppler-derived) flow structures (e.g., Figures 7 and 10), alongwith intense, vertically developed reflectivity structures. In addition, flash rates often were in excess of60min�1. All this is indicative of peak updrafts likely exceeding 20m s�1 [e.g., Basarab et al., 2015]. In combina-tionwith the high cloudbases (especially on 25 June), these collective observations are entirely consistentwithanomalous electrical structures.

However, on 8 June the sprite-parent flashes displayed typical behavior, as documented by multiple paststudies [Lang et al., 2010, 2011; Lu et al., 2013; van der Velde et al., 2014; Soula et al., 2015]. The sprite parentsinitiated as normal intracloud flashes (positive over negative), and then negative leaders traveled alongdownward sloping upper positive charge layers into the stratiform region, while the positive leaders struckground. The charge structure in the 8 June storm was heterogeneous, with midlevel positive charge morecommonly observed in the southwestern portion of the convection, while the northeastern portion featureda higher-altitude positive charge layer and was responsible for many of the sprite parents (e.g., Figure 5). Thiscomplexity is suggestive of the continuous variability in thunderstorm charge distributions described by

Journal of Geophysical Research: Atmospheres 10.1002/2016JD025299

LANG ET AL. TWO SPRITE-PRODUCING STORMS IN COLORADO 9690

Bruning et al. [2014] and may have indi-cated important differences in effectivesupercooled LWCs between the south-western and northeastern regions ofconvection in this storm.

The heterogeneity is also demonstratedby the LMA source density time series(Figure 4d). During 0430–0500UTC, thestorm was consolidating recent mergerswith smaller cells (Figure 1c). During thistime, LMA source density modal altitudeclearly lifted. Based on manual analysisof the LMA data, this reflected the pre-sence of a high-altitude (>10 km msl)source maximum within fresh convec-tion on the northeastern side of thestorm, suggestive of a convective surge[Zhang et al., 2004; Calhoun et al.,2013]. Once this core weakened around

0500UTC, both flash rate and LMA modal altitude declined, while mean flash area increased (Figures 4b and4d). We surmise that the weakening updraft led to a preference for less frequent flashes with larger areas onthe storm’s northeastern flank, away from the main core of the storm, and a number of these flashes wereable to access the ample charge that was detraining toward the eastern stratiform region, thereby enablingoccasional large-CMC +CGs that produced sprites.

By contrast, there was very little charge layer heterogeneity observed on 25 June; the storm was anomalousthroughout the analysis period. While we lack enough data to determine the precise reason 25 June wasfully inverted while 8 June was more heterogeneous, it is interesting that the cloud base on 25 June wasmuch higher than 8 June. This may have assisted with the anomalous charging, via the Williams et al.[2005] mechanism. Alternatively, but perhaps complementarily, the presence of wildfire smoke aerosolmay have aided the anomalous electrification of this storm by helping suppress large precipitation growthand thereby invigorating updrafts and increasing their LWC [Lyons et al., 1998; Rosenfeld et al., 2008]. Inaddition, smoke has been implicated in increasing sprite production in storms [São Sabbas et al., 2010].Detailed aerosol observations were not made for this case, however, so these hypotheses cannotbe confirmed.

Sprite-parent flash characteristics for both storms are summarized and compared in Table 2. There arenotable differences. In particular, 8 June sprite-parent flashes had larger areas and were longer livedcompared to 25 June, although the flashes for both storms were much smaller than those reported forlarger sprite-producing storms [Lang et al., 2010]. Peak currents for the sprite-parent +CGs were higheron 25 June, but iCMC magnitudes were similar for both storms. However, these iCMC magnitudes weresmaller on average compared to past studies of sprite-producing flashes [Hu et al., 2002; Cummer andLyons, 2005; Lang et al., 2011; Qin et al., 2012]. Indeed, iCMC values ~100 C km are at the lower limits ofthose that have been associated with sprites [Lyons et al., 2009; Beavis et al., 2014]. Lyons et al. [2012]demonstrated that the full CMC estimates (including continuing current) from several sprite parents on 8June were well in excess of the 200 C km threshold found by Qin et al. [2012], however. Thus, in additionto the possibility that the lower ionosphere was in a state more favorable for the production of spritesfrom weaker flashes, it is also possible that the continuing current made up for any low impulse CMC.However, full CMC retrieval has not been done for all sprites on these two days. Moreover, iCMC data weremissing for part of 8 June. Thus, more research into the detailed coupling between individual flash charac-teristics and sprite production may be warranted for these cases, but this was beyond the scope of thepresent study.

Interestingly, while sprite parents in both storms initiated at a similar altitude (~6.6 km), the 25 June spriteparents averaged ~0.5 km higher both around the sprite time and during the median lifetime of the flash.

Table 2. Median Characteristics of Sprite-Parent Flashes That Occurredon 8 and 25 June 2012a

Variable 6/8/2012 6/25/2012

NLDN peak current (kA) 54 86Impulse CMC (C km) 105 114LMA flash area (km2) 2040 978LMA flash altitude (m msl) 6840 7313Initiation altitude (m msl) 6606 6657Sprite delay (s) 0.014 0.059Altitude at sprite time (m msl) 6694 7286Sprite duration (s) 0.068 NATime to +CG (s) 0.385 0.111Flash duration (s) 1.686 0.485

aSprite durations were not determined for 25 June. The two storms’distributions for the listed parameters are significantly different at>99% confidence based on a two-tailed rank-sum test, except for flashinitiation altitude and impulse CMC. Sprite delay, or the time betweenthe +CG and the first video image with a sprite, has an uncertainty of±0.017 s due to video frame rate. Sprite duration also is affected by thisuncertainty. LMA source altitude during the sprite time is set by LMAsource altitudes from 50ms before the +CG to 100ms after.

Journal of Geophysical Research: Atmospheres 10.1002/2016JD025299

LANG ET AL. TWO SPRITE-PRODUCING STORMS IN COLORADO 9691

We interpret this as an indication of the downward sloping of the stratiform components of flash channels,which was observed during 8 June but not in 25 June. Overall, the differences between the sprite parentsin the two storms are consistent with the more convective nature of the sprite-parent lightning on 25 June(compare, e.g., Figure 5 to Figure 9).

On both case days, sprites did not begin occurring until the convection in the mature storms began weaken-ing, especially as indicated by total flash rate and 40 dBZ echo volume. The sprites also often were associatedwith short-term spikes in mean flash area. In addition, the reduced convective strength was indicated by ageneral decrease in LMA-estimated flash energy during the sprite period, although the energy sometimesdid undergo short-lived increases along with flash area, especially on 25 June. A transition to less frequentflashes with larger areas is a commonly observed behavior in weakening thunderstorms or away from updraftregions [Bruning and MacGorman, 2013]. Essentially, with fewer flashes to neutralize them, charge layers havemore time to advect and spread out, so that when the flashes do occur, they cover a larger area. Many studies[Williams and Yair, 2006; Soula et al., 2009; Lang et al., 2010] have found that sprites were associated withweakening convective lines or intensifying stratiform regions, but these studies mostly looked at largeMCSs, which often have many cells in various lifecycle stages. The June 2012 storms were smaller than typicalMCSs [Houze et al., 1990]. In fact, during their sprite-producing stages both the 8 and 25 June storms wereperhaps best described as unicells (e.g., Figures 1c and 2c). Thus, this study’s results provide a much moredirect link between the production of sprites and convective lifecycle. In these cases, sprite production clearlyindicated weakening convection and a transition to less frequent flashes with larger areas. This is consistentwith the reduction in flash rate during a transition toward sprite production that was observed in a sprite-producing MCS studied by Yang et al. [2013b].

However, the weakening should be understood in relative terms. During its sprite stage the 8 June storm wasamature, long-lived storm that remained dangerous and severe and was still producingmore than a flash persecond. And while its flash rate wasmuch lower than that during its sprite stage, 25 June was nonetheless stillproducing significant amounts of hail, even if most of it appeared to melt before reaching the ground (e.g.,Figure 10). The 25 June storm also was mature and long lived prior to sprite production. While total convec-tive plus stratiform echo size was not calculated (CLEAR analysis focused on mainly the convective regions),based on Figures 1 and 2, 10 dBZ composite echo area for these storms was roughly ~10,000–20,000 km2

during their sprite periods. This put them near or perhaps somewhat below the rough threshold of20,000 km2 for sprite production that was reported by Lyons [2006].

5. Conclusions

This study adds to the growing body of evidence [Lyons et al., 2003; São Sabbas and Sentman, 2003; Lyonset al., 2008; Lang et al., 2010; Yang et al., 2013a; Soula et al., 2009, 2015] that sprite production is an indicatorof weakening convection in mature long-lived storms and is also an indicator of a transition to less frequentflashes with larger areas. Moreover, due to the small and quasi-unicellular structure of the analyzed cases, thisstudy provides the strongest evidence to date that sprite production can be directly related to convectivetrends in long-lived storms and does not necessarily require a robust stratiform precipitation region suchas those typically observed in mature MCSs. Access to large reservoirs of stratified charge provides the bestchance for a flash to produce a large-CMC ground stroke, which in turn maximizes electrical stress in theupper atmosphere. However, the stratified charge layers can be the result of detraining hydrometeors fromvigorous and long-lived, but sub-MCS convection.

This study is also the first to fully document the production of sprites from anomalously charged storms, withLMA-inferred positive charge near midlevels. The results support the pioneering work of Lyons et al. [2008] onthis issue. Such storms are capable of producing sprites from +CGs within their convective regions, and thisimplies that large and long-lived (but still submesoscale) anomalous storms are good candidates for the pro-duction of sprites from convective +CGs, due to their ability to produce and distribute widespread positivecharge layers within their convective and adjoining anvil/stratiform regions. Such storms can produce spriteseven when they are producing lightning at rates greater than 60min�1, but sprite production appears to bemost favored during relative lulls in convective strength. This study thus provides additional support to thethesis of Lang et al. [2015] that the production of large-CMC and sprite-class lightning provides usefulinformation about the meteorological state of thunderstorms.

Journal of Geophysical Research: Atmospheres 10.1002/2016JD025299

LANG ET AL. TWO SPRITE-PRODUCING STORMS IN COLORADO 9692

ReferencesBarrington-Leigh, C. P., U. S. Inan, M. Stanley, and S. A. Cummer (1999), Sprites triggered by negative lightning discharges, Geophys. Res. Lett.,

26, 3605–3608, doi:10.1029/1999GL010692.Barrington-Leigh, C. P., U. S. Inan, and M. Stanley (2001), Identification of sprites and elves with intensified video and broadband array

photometer, J. Geophys. Res., 106, 1741–1750, doi:10.1029/2000JA000073.Barth, M. . C., et al. (2014), The Deep Convective Clouds and Chemistry (DC3) field campaign, Bull. Am. Meteorol. Soc., doi:10.1175/BAMS-D-13-

00290.1.Basarab, B. M., S. A. Rutledge, and B. R. Fuchs (2015), An improved lightning flash rate parameterization developed from Colorado DC3

thunderstorm data for use in cloud-resolving chemical transport models, J. Geophys. Res. Atmos., 120, 9481–9499, doi:10.1002/2015JD023470.

Beavis, N., T. J. Lang, S. A. Rutledge, W. A. Lyons, and S. A. Cummer (2014), Diurnal, seasonal, and regional behavior of cloud-to-groundlightning with large impulse charge moment changes, Mon. Weather Rev., doi:10.1175/MWR-D-14-00034.1.

Bell, T. F., S. C. Reising, and U. S. Inan (1998), Intense continuing currents following positive cloud-to-ground lightning associated with redsprites, Geophys. Res. Lett., 25, 1285–1288, doi:10.1029/98GL00734.

Biagi, C. J., K. L. Cummins, K. E. Kehoe, and E. P. Krider (2007), National Lightning Detection Network (NLDN) performance in southern Arizona,Texas, and Oklahoma in 2003–2004, J. Geophys. Res., 112, D05208, doi:10.1029/2006JD007341.

Boccippio, D. J., E. R. Williams, W. A. Lyons, I. Baker, and R. Boldi (1995), Sprites, ELF transients and positive ground strokes, Science, 269,1088–1091.

Boccippio, D. J., K. L. Cummings, H. J. Christian, and S. J. Goodman (2001), Combined satellite- and surface-based estimation of theintra-cloud-cloud-to-ground lightning ratio over the continental United States, Mon. Weather Rev., 129(1), 108–122, doi:10.1175/1520-0493(2001)129<0108:CSASBE>2.0.CO;2.

Bringi, V. N., R. Hoferer, D. A. Brunkow, R. Schwerdtfeger, V. Chandrasekar, S. A. Rutledge, J. George, and P. C. Kennedy (2011), Design andperformance characteristics of the new 8.5-m dual-offset Gregorian antenna for the CSU–CHILL radar, J. Atmos. Oceanic Technol., 28,907–920.

Bruning, E. C., and D. R. MacGorman (2013), Theory and observations of controls on lightning flash size spectra, J. Atmos. Sci., 70,4012–4029.

Bruning, E. C., S. A. Weiss, and K. M. Calhoun (2014), Continuous variability in thunderstorm primary electrification and an evaluation ofinverted-polarity terminology, Atmos. Res., 135, 274–284.

Calhoun, K. M., D. R. MacGorman, C. L. Ziegler, and M. I. Biggerstaff (2013), Evolution of lightning activity and storm charge relative todual-Doppler analysis of a high-precipitation supercell storm, Mon. Weather Rev., 141, 2199–2223, doi:10.1175/MWR-D-12-00258.1.

Carey, L. D., and S. A. Rutledge (1998), Electrical and multiparameter radar observations of a severe hailstorm, J. Geophys. Res., 103,13,979–14,000, doi:10.1029/97JD02626.

Carey, L. D., M. J. Murphy, T. L. McCormick, and N. W. S. Demetriades (2005), Lightning location relative to storm structure in a leading-line,trailing-stratiform mesoscale convective system, J. Geophys. Res., 110, D03105, doi:10.1029/2003JD004371.

Cummer, S. A., and W. A. Lyons (2005), Implications of lightning charge moment changes for sprite initiation, J. Geophys. Res., 110, A04304,doi:10.1029/2004JA010812.

Cummer, S. A., W. A. Lyons, and M. A. Stanley (2013), Three years of lightning impulse charge moment change measurements in the UnitedStates, J. Geophys. Res. Atmos., 118, 5176–5189, doi:10.1002/jgrd.50442.

Cummins, K. L., and M. J. Murphy (2009), An overview of lightning location systems: History, techniques, and data uses, with an in-depth lookat the U.S. NLDN, IEEE Trans. Electromagn. Compat., 51(3), 499–518.

Dolan, B. A., and S. A. Rutledge (2007), An integrated display and analysis methodology for multivariable radar data, J. Appl. Meteorol.Climatol., 46, 1196–1213.

Ely, B. L., R. E. Orville, L. D. Carey, and C. L. Hodapp (2008), Evolution of the total lightning structure in a leading-line, trailing-stratiformmesoscale convective system over Houston, Texas, J. Geophys. Res., 113, D08114, doi:10.1029/2007JD008445.

Franz, R. C., R. J. Nemzek, and J. R. Winckler (1990), Television image of a large upward electrical discharge above a thunderstorm system,Science, 249, 48–51.

Fuchs, B. R., S. A. Rutledge, E. C. Bruning, J. R. Pierce, J. K. Kodros, T. J. Lang, D. R. MacGorman, P. R. Krehbiel, and W. Rison (2015),Environmental controls on storm intensity and charge structure in multiple regions of the continental United States, J. Geophys. Res.Atmos., 120, 6575–6596, doi:10.1002/2015JD023271.

Halbert, K. T., W. G. Blumberg, and P. T. Marsh (2015), SHARPpy: Fueling the Python Cult, paper presented at 5th Symposium on Advances inModeling and Analysis Using Python, Am. Meteorol. Soc., Phoenix Ariz.

Hiraki, Y., and H. Fukunishi (2006), Theoretical criterion of charge moment change by lightning for initiation of sprites, J. Geophys. Res., 111,A11305, doi:10.1029/2006JA011729.

Houze, R. A., Jr., B. F. Smull, and P. Dodge (1990), Mesoscale organization of springtime rainstorms in Oklahoma, Mon. Weather Rev., 118,613–654.

Hu, W, S. Cummer, W. A. Lyons, and T. Nelson (2002), Lightning charge moment changes for the initiation of sprites, Geophys. Res. Lett., 29(8),1279, doi:10.1029/2001GL014593.

Huang, E., E. Williams, R. Boldi, S. Heckman, W. Lyons, M. Taylor, T. Nelson, and C. Wong (1999), Criteria for sprites and elves based onSchumann resonance observations, J. Geophys. Res., 104, 16,943–16,964, doi:10.1029/1999JD900139.

Junyent, F., V. Chandrasekar, V. N. Bringi, S. A. Rutledge, P. C. Kennedy, D. Brunkow, J. George, and R. Bowie (2015), Transformation of theCSU–CHILL Radar Facility to a dual-frequency, dual-polarization doppler system, Bull. Am. Meteorol. Soc., 96, 975–996, doi:10.1175/BAMS-D-13-00150.1.

Krehbiel, P., T. Chen, S. McCrary, W. Rison, G. Gray, and M. Brook (1996), The use of dual channel circular-polarization radar observations forremotely sensing storm electrification, Meteorol. Atmos. Phys., 59, 65–82.

Lang, T. J., and S. A. Rutledge (2002), Relationships between convective storm kinematics, precipitation and lightning,Mon. Weather Rev., 130,2492–2506.

Lang, T. J., and S. A. Rutledge (2011), A framework for the statistical analysis of large radar and lightning dataset: Results from STEPS 2000,Mon. Weather Rev., 139, 2536–2551.

Lang, T. J., S. A. Rutledge, J. E. Dye, M. Venticinque, P. Laroche, and E. Defer (2000), Anomalously low negative cloud-to-ground lightning flashrates in intense convective storms observed during STERAO-A, Mon. Weather Rev., 128, 160–173.

Lang, T. J., et al. (2004), The severe thunderstorm electrification and precipitation study, Bull. Am. Meteorol. Soc., 85, 1107–1125.

AcknowledgmentsPat Kennedy, Dave Brunkow, JimGeorge, and Bob Bowie all contributedto the CSU-CHILL and CSU-Pawneeradar scanning on the storm daysanalyzed in this study, and thus,important data sets would not havebeen available without their help. CSUradar data and COLMA data from DC3are available from the National Centerfor Atmospheric Research (NCAR; http://data.eol.ucar.edu/master_list/?project=DC3). NEXRAD volumetric radardata are available from Amazon WebServices (https://aws.amazon.com/noaa-big-data/nexrad/), while MRMSdata are available from the NationalSevere Storms Laboratory (http://www.nssl.noaa.gov/projects/mrms/). NLDNdata are available from the NASA GlobalHydrology Resource Center (https://ghrc.nsstc.nasa.gov/home/). Soundingdata were obtained from the Universityof Wyoming (http://weather.uwyo.edu/upperair/sounding.html). Key opensource software packages used in thisstudy include Py-ART (http://arm-doe.github.io/pyart/), ARTview (https://github.com/nguy/artview), lmatools(https://github.com/deeplycloudy/lma-tools), CSU_RadarTools (https://github.com/CSU-Radarmet/CSU_RadarTools),DualPol (https://github.com/nasa/DualPol), MMM-Py (https://github.com/nasa/MMM-Py), SkewT (https://pypi.python.org/pypi/SkewT), and SHARPpy(http://sharppy.github.io/SHARPpy/).CEDRIC and SPRINT can be obtainedfrom NCAR (https://wiki.ucar.edu/dis-play/raygridding/Home) along withother useful radar software, such asRadx (https://www.ral.ucar.edu/pro-jects/titan/docs/radial_formats/radx.html). Contact the first author (timothy.j.lang@nasa.gov) for access to other datasets, such as sprite imagery and CMCNmeasurements. Lang also can provideaccess to customized analysis software,such as CLEAR and XLMA. Funding forthis work was provided by the NASALightning Imaging Sensor (LIS) project,the Defense Advanced Research ProjectAgency (DARPA) Nimbus program, andthe National Science Foundation (NSF)Physical Meteorology and LowerAtmosphere Observing Facilitiesprograms. DC3 was made possible bythe financial and logistical support ofNSF, NASA, and the National Oceanicand Atmospheric Administration(NOAA). The views, opinions, andfindings in this report are those of theauthors and should not be construed asan official NASA or U.S. Governmentposition, policy, or decision.

Journal of Geophysical Research: Atmospheres 10.1002/2016JD025299

LANG ET AL. TWO SPRITE-PRODUCING STORMS IN COLORADO 9693

Lang, T. J., W. A. Lyons, S. A. Rutledge, J. D. Meyer, D. R. MacGorman, and S. A. Cummer (2010), Transient luminous events above twomesoscale convective systems: Storm structure and evolution, J. Geophys. Res., 115, A00E22, doi:10.1029/2009JA014500.

Lang, T. J., J. Li, W. A. Lyons, S. A. Cummer, S. A. Rutledge, and D. R. MacGorman (2011), Transient luminous events above two mesoscaleconvective systems: Charge moment change analysis, J. Geophys. Res., 116, A10306, doi:10.1029/2011JA016758.

Lang, T. J., S. A. Cummer, S. A. Rutledge, and W. A. Lyons (2013), The meteorology of negative cloud-to-ground lightning strokes with largecharge moment changes: Implications for negative sprites, J. Geophys. Res. Atmos., 118, 7886–7896, doi:10.1002/jgrd.50595.

Lang, T. J., S. A. Rutledge, B. Dolan, P. Krehbiel, W. Rison, and D. T. Lindsey (2014), Lightning in wildfire smoke plumes observed in Coloradoduring summer 2012, Mon. Weather Rev., 142(2), 489–507.

Lang,T. J., S.A.Cummer,D.Petersen, L. Flores-Rivera,W.A. Lyons,D.MacGorman,andW.Beasley (2015), Largechargemomentchange lightningon 31 May to 1 June 2013, including the El Reno tornadic storm, J. Geophys. Res. Atmos., 120, 3354–3369, doi:10.1002/2014JD022600.

Li, J., S. Cummer, G. Lu, and L. Zigoneanu (2012), Charge moment change and lightning-driven electric fields associated with negative spritesand halos, J. Geophys. Res., 117, A09310, doi:10.1029/2012JA017731.

Lu, G., S. A. Cummer, R. J. Blakeslee, S. Weiss, and W. H. Beasley (2012), Lightning morphology and impulse charge moment change of highpeak current negative strokes, J. Geophys. Res., 117, D04212, doi:10.1029/2011JD016890.

Lu, G., et al. (2013), Coordinated observations of sprites and in-cloud lightning flash structure, J. Geophys. Res. Atmos., 118, 6607–6632,doi:10.1002/jgrd.50459.

Lyons, W. A. (1996), Sprite observations above the U.S. High Plains in relation to their parent thunderstorm systems, J. Geophys. Res., 101,29,641–29,652, doi:10.1029/96JD01866.

Lyons, W. A. (2006), The meteorology of transient luminous events—An introduction and overview, in NATO Advanced Study Institute onSprites, Elves and Intense Lightning Discharges, edited by M. Fullekrug et al., pp. 19–56, Springer, New York.

Lyons, W. A., T. E. Nelson, E. R. Williams, J. A. Cramer, and T. R. Turner (1998), Enhanced positive cloud-to-ground lightning in thunderstormsingesting smoke from fires, Science, 282, 77–80.

Lyons, W. A., T. Nelson, E. R. Williams, S. A. Cummer, and M. A. Stanley (2003), Characteristics of sprite-producing positive cloud-to-groundlightning during the 19 July STEPS mesoscale convective systems, Mon. Weather Rev., 131, 2417–2427.

Lyons, W. A., S. A. Cummer, M. A. Stanley, G. R. Huffines, K. C. Wiens, and T. E. Nelson (2008), Supercells and sprites, Bull. Am. Meteorol. Soc., 89,1165–1174, doi:10.1175/2008BAMS2439.1.

Lyons, W. A., M. Stanley, J. D. Meyer, T. E. Nelson, S. A. Rutledge, T. J. Lang, and S. A. Cummer (2009), The meteorological and electricalstructure of TLE-producing convective storms, in Lightning: Principles, Instruments and Applications, edited by H. D. Betz et al., pp. 389–417,Springer, New York, doi:10.1007/978-1-4020-9079-0_17.

Lyons, W. A., S. A. Cummer, W. Rison, P. R. Krehbiel, T. J. Lang, S. A. Rutledge, G. Lu, M. Stanley, T. Ashcraft, and T. E. Nelson (2012),Sprite-producing convective storms within the Colorado Lightning Mapping Array, Abstract 0256 presented at 2012 Fall Meeting, AGU,San Francisco, Calif.

MacGorman, D. R., et al. (2008), TELEX: The Thunderstorm Electrification and Lightning Experiment, Bull. Am. Meteorol. Soc., 89, 997–1013.Miller, J., C. G. Mohr, and A. J. Weinheimer (1986), The simple rectification to Cartesian space of folded radial velocities from Doppler radar

sampling, J. Atmos. Oceanic Technol., 3, 162–174.Mohr, C. G., and L. J. Miller (1983), CEDRIC—A software package for Cartesian space editing, synthesis and display of radar fields under

interactive control, in 21st Conference on Radar Meteorology, Edmonton, AB, Canada, pp. 569–574, Am. Meteorol. Soc, Boston, Mass.Mohr, C. G., and R. L. Vaughn (1979), An economical approach for Cartesian interpolation and display of reflectivity factor data in three-

dimensional space, J. Appl. Meteorol., 18, 661–670.Mohr, C. G., L. J. Miller, R. L. Vaughn, and H. W. Frank (1986), The merger of mesoscale datasets into a common Cartesian format for efficient

and systematic analyses, J. Atmos. Oceanic Technol., 3, 143–161.Montanyà, J., S. Soula, N. Pineda, O. van der Velde, P. Clapers, G. Solà, J. Bech, and D. Romero (2009), Study of the total lightning activity in a

hailstorm, Atmos. Res., 91, 430–437.Neubert, T., et al. (2005), Co-ordinated observations of transient luminous events during the EuroSprite2003 campaign, J. Atmos. Sol. Terr.

Phys., 67, 807–820.Parker, M. D., S. A. Rutledge, and R. H. Johnson (2001), Cloud-to-ground lightning in linear mesoscale convective systems,Mon. Weather Rev.,

129, 1232–1242.Qin, J., S. Celestin, and V. P. Pasko (2012), Minimum charge moment change in positive and negative cloud to ground lightning discharges

producing sprites, Geophys. Res. Lett., 39, L22801, doi:10.1029/2012GL053951.Rison, W., R. Thomas, P. Krehbiel, T. Hamlin, and J. Harlin (1999), A GPS-based three-dimensional lightning mapping system: Initial

observations in central New Mexico, Geophys. Res. Lett., 26, 3573–3576, doi:10.1029/1999GL010856.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, doi:10.1126/science.1160606.Rust, W. D., D. R. MacGorman, E. C. Bruning, S. A. Weiss, P. R. Krehbiel, R. J. Thomas, W. Rison, T. Hamlin, and J. Harlin (2005), Inverted-polarity

electrical structures in thunderstorms in the Severe Thunderstorm Electrification and Precipitation Study (STEPS), Atmos. Res., 76(1–4),247–271.

Ryzhkov, A. V., M. R. Kumjian, S. M. Ganson, and P. Zhang (2013), Polarimetric radar characteristics of melting hail. Part II: Practicalimplications, J. Appl. Meteorol. Climatol., 52, 2871–2886, doi:10.1175/JAMC-D-13-074.1.

São Sabbas, F. T., and D. D. Sentman (2003), Dynamical relationship of infrared cloudtop temperatures with occurrence rates ofcloud-to-ground lightning and sprites, Geophys. Res. Lett., 30(5), 1236, doi:10.1029/2002GL015382.

São Sabbas, F. T., V. T. Rampinelli, J. Santiago, P. Stamus, S. L. Vadas, D. C. Fritts, M. J. Taylor, P. D. Pautet, G. Dolif Neto, and O. Pinto (2009),Characteristics of sprite and gravity wave convective sources present in satellite IR images during the SpreadFEx 2005 in Brazil, Ann.Geophys., 27, 1279–1293, doi:10.5194/angeo-27-1279-2009.

São Sabbas, F. T., et al. (2010), Observations of prolific transient luminous event production above a mesoscale convective system inArgentina during the Sprite2006 Campaign in Brazil, J. Geophys. Res., 115, A00E58, doi:10.1029/2009JA014857.

Saunders, C. P. R., and S. L. Peck (1998), Laboratory studies of the influence of the rime accretion rate on charge transfer duringgraupel/crystal collisions, J. Geophys. Res., 103, 13,949–13,956, doi:10.1029/97JD02644.

Soula, S., O. van der Velde, J. Montanya, T. Neubert, O. Chanrion, and M. Ganot (2009), Analysis of thunderstorm and lightning activityassociated with sprites observed during the EuroSprite campaigns: Two case studies, Atmos. Res., 91, 514–528, doi:10.1016/j.atmosres.2008.06.017.

Soula, S., et al. (2015), Time and space correlation between sprites and their parent lightning flashes for a thunderstorm observed during theHyMeX campaign, J. Geophys. Res. Atmos., 120, 11,552–11,574, doi:10.1002/2015JD023894.

Journal of Geophysical Research: Atmospheres 10.1002/2016JD025299

LANG ET AL. TWO SPRITE-PRODUCING STORMS IN COLORADO 9694

Storm Prediction Center (2012a), Severe weather event summaries. [Available at http://www.spc.noaa.gov/climo/online/.]Storm Prediction Center (2012b), Mesoscale Discussion 1263. [Available at http://www.spc.noaa.gov/products/md/2012/md1263.html.]Suzuki, T., Y. Matsudo, T. Asano, M. Hayakawa, and K. Michimoto (2011), Meteorological and electrical aspects of several winter thunder-

storms with sprites in the Hokuriku area of Japan, J. Geophys. Res., 116, D06205, doi:10.1029/2009JD013358.Takahashi, T. (1978), Riming electrification as a charge generation mechanism in thunderstorms, J. Atmos. Sci., 35, 1536–1548.Takahashi, Y., R. Miyasato, T. Adachi, K. Adachi, M. Sera, A. Uchida, and H. Fukunishi (2003), Activities of sprites and elves in the winter season,

Japan, J. Atmos. Sol. Terr. Phys., 65, 551–560.Taylor, M. J., et al. (2008), Rare measurements of a sprite with halo event driven by a negative lightning discharge over Argentina, Geophys.

Res. Lett., 35, L14812, doi:10.1029/2008GL033984.Thompson, E. J., S. A. Rutledge, B. Dolan, V. Chandrasekar, and B. L. Cheong (2014), A dual-polarization radar hydrometeor classification

algorithm for winter precipitation, J. Atmos. Oceanic Technol., 31, 1457–1481, doi:10.1175/JTECH-D-13-00119.1.van der Velde, O. A., Á. Mika, S. Soula, C. Haldoupis, T. Neubert, and U. S. Inan (2006), Observations of the relationship between sprite

morphology and in-cloud lightning processes, J. Geophys. Res., 111, D15203, doi:10.1029/2005JD006879.van der Velde, O. A., J. Montanyà, S. Soula, N. Pineda, and J. Bech (2010), Spatial and temporal evolution of horizontally extensive lightning

discharges associated with sprite-producing positive cloud-to-ground flashes in northeastern Spain, J. Geophys. Res., 115, A00E56,doi:10.1029/2009JA014773.

van der Velde, O. A., J. Montanyà, S. Soula, N. Pineda, and J. Mlynarczyk (2014), Bidirectional leader development in sprite-producing positivecloud-to-ground flashes: Origins and characteristics of positive and negative leaders, J. Geophys. Res. Atmos., 119, 12,755–12,779,doi:10.1002/2013JD021291.

Wiens, K. C., S. . A. Rutledge, and S. A. Tessendorf (2005), The 29 June 2000 supercell observed during STEPS. Part II: Lightning and chargestructure, J. Atmos. Sci., 62, 4151–4177.

Williams, E. R. (1998), The positive charge reservoir for sprite-producing lightning, J. Atmos. Sol. Terr. Phys., 60, 689–692.Williams, E. R., and Y. Yair (2006), The microphysical and electrical properties of sprite-producing thunderstorms, in NATO Advanced Study

Institute on Sprites, Elves and Intense Lightning Discharges, edited by M. Fullekrug et al., pp. 57–83, Springer, New York.Williams, E. R., V. Mushtak, D. Rosenfeld, S. Goodman, and D. Boccippio (2005), Thermodynamic conditions that lead to superlative updrafts

and mixed-phase microphysics, Atmos. Res., 76, 288–306.Yang, J., X. Qie, and G. Feng (2013a), Characteristics of one sprite-producing summer thunderstorm, Atmos. Res., 127, 90–115.Yang, J., M. R. Yang, C. Liu, and G. L. Feng (2013b), Case studies of sprite-producing and non-sprite-producing summer thunderstorms, Adv.

Atmos. Sci., 30(6), 1786–1808, doi:10.1007/s00376-013-2120-5.Zhang, J., et al. (2011), National Mosaic andMulti-Sensor QPE (NMQ) system: Description, results, and future plans, Bull. Am. Meteorol. Soc., 92,

1321–1338.Zhang, Y., Q. Meng, P. R. Krehbiel, X. Liu, and X. Zhou (2004), Spatial and temporal characteristics of VHF radiation source produced by

lightning in supercell thunderstorms, Chin. Sci. Bull., 49, 624–631.

Journal of Geophysical Research: Atmospheres 10.1002/2016JD025299

LANG ET AL. TWO SPRITE-PRODUCING STORMS IN COLORADO 9695