5.1 Atmospheric conditions

Figure 5.1 shows data from a mobile CLASS sounding launched from the Fort Morgan Airport at 1356 MDT on 12 July 1996. Plotted on the Skew-T/Log-p diagram are vertical profiles of temperature, dewpoint, and winds. The sounding appears to be somewhat more moist than the 10 July 1996 sounding. The CAPE is only 441 J kg^-1, the shear over the lowest 6 km is 13.6 m s^-1, and the Bulk Richardson Number (BNR) is 4.7. The BNR is even lower than 10 July. It is suggestive of possible supercellular development (Weisman and Klemp, 1982), but again it is so low because the CAPE is also low. The ambient wind shear is weak as well. However, the Lifted Index is -3.45, which is greater than 10 July and implies that convection was at least possible. But overall, the atmospheric environment did not appear conducive to the development of intense convection. However, this sounding may not have been entirely representative of the storm’s environment, since it initially developed on the over 100 km northwest of the sounding launch location. Also, conditions could have changed between the time of the sounding and the time of storm development (e.g., a short wave may have propagated through the region after the sounding was taken). The steering winds appear to be mostly westerly, whereas the storm itself traveled to the southeast. This is perhaps further evidence that the sounding was not representative of the storm environment.

5.2 Overview of storm evolution and major transition

Figure 5.2 (a-h) shows horizontal cross-sections of radar reflectivity (dBZ) obtained from the CSU-CHILL radar at 0.5 km AGL for the 12 July 1996 storm. The cross-sections are approximately 30 minutes apart. Full radar volume coverage of the storm began at 1647 MDT. At this time the storm was over the Cheyenne Ridge in extreme southern Wyoming. It then moved to the southeast toward the CHILL radar, passing just east of the radar site, and continued moving to the southeast away from the radar. In its early stages it was relatively well-positioned in the western lobe of the interferometer (ITF). However, as it approached CHILL, its eastern extent crossed the baseline of the ITF into the eastern lobe, while the western half of the storm remained in the western lobe. After the storm passed CHILL it moved beyond the southern boundaries of either ITF lobe.

At 1658 MDT (Figure 5.2a), the storm mainly consisted of five distinct cells or groups of cells. (Cell groups will be referred to as “cells” for simplicity.) The southern central cell, Cell 1, was the most intense, in terms of the areal extent of its low-level reflectivity. The cell immediately north of Cell 1, Cell 2, also was strong. The other cells in the storm complex - Cells 0, 3, and 4 - were much less intense.

By 1730 MDT (Figure 5.2b), Cell 4 had intensified, while Cells 1 and 2 had decreased in strength. The other cells from the 1658 MDT observation period had predominantly dissipated, or were not easily distinguishable within the large area of contiguous echo. Between this time and 1804 MDT (Figure 5.2c), Cell 4 continued to strengthen, and an additional cell, Cell 5, had developed on its forward flank (to the southeast) and began to merge with Cell 4. To the east of these cells a large contiguous echo region persisted, with maximum dimensions of approximately 40-50 km in either direction. Many weak cells continued to be embedded within this region of echo.

From this time until 1832 MDT (Figure 5.2d), Cell 5 completed its merger with Cell 4, and another cell, Cell 6, formed on the forward flank of the main cell. Within the large contiguous echo region, a weak cell - Cell 7 - began to form and dominate the area. At 1855 MDT (Figure 5.2e), the storm complex was divided into two main parts, the western and the eastern. The western region was comprised of Cells 4 and 6, which were undergoing a merger. These cells were quite intense, with extended areas of 60 dBZ and greater reflectivity at 0.5 km AGL. The eastern region was comprised mostly of stratiform echo with some embedded weak cells (Cell 7) which had not yet reached 50 dBZ in the low-level reflectivity.

By 1931 MDT (Figure 5.2f), Cells 4 and 6 had merged, and weakened. Cell complex 4, which comprised the western region of the storm, consisted of three distinct (> 50 dBZ) cores. The eastern region had continued to shrink in areal coverage, but Cell 7 had intensified and was now above 50 dBZ at 0.5 km AGL.

After this time, at 1959 MDT (Figure 5.2g) and 2031 MDT (Figure 5.2h), the storm complex (i.e., both Cells 4 and 7) weakened, though by 2031 Cell 4 had re-intensified to some degree on its western flank. The eastern region of the storm complex, the region once occupied by Cell 7, featured much weaker convection by 1959 and appeared to be comprised largely of stratiform echo with some very weak embedded convection (by 2031 MDT). After the volume scan starting at 2057 MDT radar coverage was terminated because peak low-level reflectivities were below 50 dBZ throughout the storm, and the storm was well south of the ITF lobes.

To summarize storm evolution, when full radar volume coverage was begun, the storm complex was very large and multicellular. Throughout its lifetime, the storm complex remained multicellular, usually with one cell dominating much of the reflectivity pattern, though there also were other intense cells at times. After 1900 MDT the storm complex continued to persist in some form for over 2 hours, even though it featured lower peak reflectivities, over a smaller area, than before this time. This transition in storm structure will be examined in greater detail in the next few sections.

5.3 Comparison of lightning and bulk precipitation rates

Figure 5.3 shows 5-minute CG lightning rates as functions of time for the entire storm complex, separated into positive and negative CG components. The lifetime of this storm can be separated into two relatively distinct phases of CG lightning. In the first phase, CG flash rates for either polarity were quite low, typically less than 5 per 5-minute period. During this phase a large percentage of the CGs were of positive polarity, on the order of 50% (although total CG flash rates were low). However, after 1900 MDT, positive CG rates remained similar for less than an hour and then declined to 0 until near the end of radar coverage. Negative CG flash rates rose sharply, peaking at 25 per 5-minute period at 1935 MDT, then declined to more modest values late in the storm complex’s lifetime.

Following the method of Chapter 2, hail and rain mass fluxes were calculated at 0.5 km AGL for all radar volume scans of the storm complex. However, because this storm came so close to the radar, the azimuthal boundaries of the radar volumes had to be limited, especially during 1830-1930 MDT. Thus, the entire storm was not always included in these volumes, even at low levels, so precipitation fluxes will be underestimates during this time period. However, because effort was made to include as much significant (i.e., > 30 dBZ) echo as possible, the flux trends should not be adversely affected.

In addition, following the method of Chapter 3, intra-cloud (IC) flash rates were extracted from the interferometer (ITF) data. However, this storm was not well-placed in relation to the ITF lobes. After 1800 MDT, the eastern extent crossed from the western lobe into the eastern lobe. However, until 1920 MDT, the main cell cores were well-placed in either ITF lobe, with the main western cores within the western lobe and the main eastern cores in the eastern lobe, with only weak echo actually in the vicinity of the baseline. As will be shown later, IC lightning was well-correlated in position with high reflectivity cores, so flash rates should not be affected significantly. However, the IC rates that were calculated are likely to be underestimates after 1800 MDT, though trends should not be affected appreciably because the high reflectivity cores - that is, the cores that were most likely to produce IC lightning - were still well-positioned. After 1920 MDT, the storm began moving out of the southern boundaries of both lobes, so flash rate calculations were not done for later times, nor were they computed for times prior to 1700 at which time the northern part of the storm complex extended beyond the northern ITF boundary.

Figure 5.4 shows precipitation fluxes and 5-minute IC flash rates as functions of time for the entire storm complex. Note the significant change in hail fluxes around 1900 MDT. Before this, hail fluxes were significant, though of a pulsing nature related to individual cell growth and decay. After this time, however, hail fluxes were largely insignificant. This change in storm complex hail fluxes is well-correlated with the change in storm complex negative CG rates. The drop-off in hail fluxes after 1900 MDT also seems correlated with a similar, yet not as severe, reduction in rain fluxes from a peak just before 1900.

When full radar coverage began at 1647 MDT, the low-level precipitation fluxes appeared to be declining rapidly from an earlier peak. Both the rain and hail flux trends seem in phase at first, but the rain flux recovers from its reduction earlier than the hail flux. By 1720 MDT, though, both fluxes have recovered and peaked again, albeit at much smaller values than before. ITF IC flash rates seem to peak earlier than either of the precipitation fluxes, at 1705, and actually achieved a minimum by 1720. After 1720 MDT, IC flash rates began a sharp rise, but the low-level rain and hail fluxes declined to minima at 1730 MDT. They recovered after this time, and then an interesting series of events occurred, with IC flash rate peaking at 1740 MDT, hail flux peaking after this at 1759, and then the rain flux reaching a maximum at 1812. This implies, in a total storm complex sense, that the IC rate peaks before the fallout of significant hail, which itself precedes the fallout of significant rain. The tendency for IC lightning to lead the low-level hail fluxes is consistent with what was seen in the 10 July storm, though the lag between the hail and rain fluxes was not observed in that storm as it is in the present case.

After this time, all three variables began another decline and subsequent recovery. Once again, hail flux led the rain flux, but the IC flash rate took a much longer time to reach its next peak at 1905 MDT, which occurred after the hail peak at 1849 and the rain peak at 1855. Thus, whereas before IC flash rate led the precipitation fluxes, it now lagged both. The decline in IC rates after 1905 again continued lagging the precipitation flux reductions, though IC rate calculations were terminated after 1920 MDT. After 1900 MDT, hail fluxes remained largely negligible, while rain fluxes were lower than their peak at 1855, but continued to be fairly significant. The rain flux made a small recovery at 1918, and remained very steady until 1937 when it underwent a sharp decline. After reaching a minimum around 2000 MDT, the rain flux recovered to one final peak, around 2020. This last rain flux peak was roughly coincident with a slight increase in hail flux, though the hail flux was still very low compared to earlier times. After this time, the fluxes again declined. Radar coverage was terminated after 2057 MDT.

Note, in general, that the fluxes and IC rates are much higher than the storm of 10 July 1996. This is likely due to two causes. One is that the 12 July storm complex itself was larger, both in volume and areal extent. The second is that it appeared more intense, based on the gridded reflectivity data seen in Figure 5.2, with higher peak reflectivities and larger areal extent of high reflectivity regions.

There are two main questions to be posed regarding Figures 5.3 and 5.4. The first is: Why do the negative CG rates and the storm complex hail fluxes seem so anti-correlated, in a bulk sense? The second is: Why did the IC flash rate lead the low-level precipitation fluxes early in the storm’s lifetime, yet lag them later? These questions will be examined in order in the following discussion.

From Figures 5.3 and 5.4, it can be seen that around 1804 MDT, CG flash rates were quite low, and storm complex hail and rain fluxes were near their maxima. Figure 5.5 shows a vertical cross-section of reflectivity at 11 km west of CHILL at this time. Referring to Figure 5.2c, this cross-section cuts through the center of Cell 4. Based on this cross-section, the cell appears intense, with the 30 dBZ contour extending to 11 km AGL, and the 50 dBZ contour extending to 7.5 km AGL. Figure 5.6 shows a horizontal cross-section of reflectivity at 0.5 km AGL for the radar volume starting at 1937 MDT. Also shown are the ground-strike positions and polarities of NLDN-detected CG lightning that occurred during this volume (approximately 6 minutes in duration). At this time the negative CG rates were at their peak, and the storm was divided into two main regions, Cell 4 and Cell 7. (Refer to the discussion of Figure 5.2.) Likewise, CG lightning around this time was divided nearly equally between the two major cell complexes. Figures 5.7a and 5.7b show vertical cross-sections of reflectivity at this time at 24 km east of CHILL and 56 km east of CHILL, respectively. These two cross-sections cut through the approximate centers of the major cells. Contrasting these vertical cross-sections with Figure 5.5, it is apparent that cell structure is radically different between the two times. The 30 dBZ contours at 1937 MDT extend to only 9.5 km AGL for either cell complex. The 50 dBZ contours extend to 4 km AGL or less. There is no region of 60 dBZ and greater reflectivity like there was at 1804 MDT, when the 60 dBZ contour extended to 4 km AGL. Overall, the storm complex at 1937 MDT appears much less intense, based on both horizontal and vertical structure, compared to the storm at 1804. Thus, marked differences in storm intensity between the two times correspond well to the marked differences in CG rates. This is further confirmation of what was seen in the hail flux data, where significant hail fluxes - even though they underwent major fluctuations - corresponded to low CG production, but very low hail fluxes corresponded to significant production of negative CGs. It appears that, for this storm at least, storm intensity - as determined by the existence of hail and the maximum heights of the 30 and 50 dBZ reflectivity contours - is inversely correlated with negative CG lightning, at least in a bulk sense. This will be discussed further in Section 5.5.

The second question deals with why IC lightning led low-level precipitation fluxes at first, but lagged them later. The answer to this question relates to the multicellular nature of the storm complex. Figure 5.8 (a-b) shows horizontal cross-sections of reflectivity at 0.5 km AGL at 1741 and 1855 MDT, respectively. The mean horizontal positions of IC flashes occurring during each radar volume (approximately 6 minutes in duration) are overlaid on these plots of reflectivity. At both times IC lightning rates were near their maximum values, and it appears that the western half of the storm was producing the most ICs. Note, however, that especially at 1855, the storm complex is straddling the ITF baseline. However, the two main cell regions were still well-positioned in the ITF lobes. At 1741 MDT, there was mainly one cell producing IC lightning. The same was basically true at 1855 MDT, but note the large area of 50 dBZ and greater echo north of the main IC lightning centroid. This was an older cell, which due to its areal coverage and intensity likely was contributing significantly to the low-level precipitation fluxes, even though it was producing little IC lightning. At 1741 MDT there was no other major cell to contribute to the low-level precipitation fluxes. Thus at this time IC rates and precipitation fluxes were largely dominated by one cell, but at 1855 MDT the IC lightning was still dominated by one cell, but the fluxes had an additional major cell contributing to their total values.

Thus, the apparent change in the IC flash rate trend relative to the low-level precipitation flux trend (Refer to Figure 5.4.) probably is due to the multicellular nature of the storm. The precipitation fluxes and IC rates are for the entire storm complex. Near both major IC flash rate peaks, only one cell was contributing significantly to the IC flash rate. But during the second peak, two major cells were contributing to the low-level precipitation fluxes, whereas during the earlier peak there was only one major cell. The superposition of the flux contributions from these two cells around the time of the second IC peak are probably what caused the apparent phase change between the fluxes and IC lightning. This superposition of cells also probably was causing the slow increase in IC rate toward its final peak. On an individual cell basis, based on the available data, IC lightning preceded the development of significant precipitation. This is seen with the first major IC peak, when only one major cell was in existence. The same pattern likely repeated itself in subsequent cells, but because of the superposition of two major cells this pattern is not evident.

5.4 Identification of bulk precipitation types

Following the method of Chapter 2, bulk hydrometeor identification was performed on the radar volumes from this storm. Unfortunately, unlike the storm of 10 July, this storm came extremely close to the radar. Due to its proximity to the radar for much of its lifetime, the storm was not “topped” and medium- to high-altitude data were not obtained for the entire storm complex. Thus, trends in bulk precipitation volumes below the freezing level will not be examined in detail, as such trends will be corrupted by the lack of data from a significant percentage of the radar volumes. Instead, areal coverage of bulk precipitation types at 0.5 km AGL will be examined, as such trends will not be affected by the loss of higher altitude data.

The freezing level based on the afternoon (1356 MDT) CLASS sounding launched from the Fort Morgan airport was determined to be 3.42 km AGL, CHILL relative. The reflectivity gradient thresholds again were set to the values of 10, 15, and 20 dBZ km^-1. Sensitivity tests were performed to better estimate the proper threshold setting. For this storm complex, most low-level (0.5 km AGL) grid points were split between rain only and small (< 2 cm) hail mixed with rain. Small hail only, and any form of large (> 2 cm) hail, were not diagnosed to be present at the lowest grid level except perhaps once or twice, and at those times their areal coverage was quite small (1-2 km² at the most). When higher altitude small hail calculations were available, typically small hail mixed with rain was most predominant closer to the surface, and small hail only was predominant closer to the freezing level. This suggests that significant melting of hail occurred as it passed below the freezing level and approached the surface. Figure 5.9 shows the time history of the low-level areal coverage of small hail mixed with rain for all three reflectivity gradient thresholds. The general pattern is quite similar to the hail flux time history in Figure 5.4, and is independent of the reflectivity gradient criteria. In both plots, there are indications of a few major pulses of hail before 1900 MDT, and afterward there was little to no hail.

Figure 5.10 is the same as Figure 5.9, except for the low-level areal coverage of rain only. Note that for rain, all three reflectivity gradient thresholds produced the same values. The general trends are similar to the rain flux trends in Figure 5.3 - the relative maxima around 1815, 1900, 1930, and 2020 MDT match up between the plots, though relative magnitudes do not necessarily match. Note that before 1800 MDT the areal coverage of rain at 0.5 km AGL was small, especially when its relative magnitude is compared to the relative magnitude of the rain flux at this time. However, recall that Figure 5.9 shows small hail and rain mixed together. These areas also will contribute to the rain flux to some extent.

Based on this limited precipitation identification analysis, it appears that the storm - when it produced hail - produced mostly small hail, and also had few hail-only or hail-dominated precipitation shafts. However, at 1749 MDT and again at 1759 MDT, the National Weather Service reported the occurrence of hail with a diameter of 1.75 inches (4.5 cm) near the town of Carr, Colorado, 50 km NNW of the CHILL radar. This suggests that even though the bulk of the frozen precipitation was in the form of small (< 2 cm) hail (according to the radar) this does not mean that larger hailstones did not develop and reach the surface. Based on the precipitation flux data as well as the areal coverage data, it appears that the storm produced vast quantities of rain. After 1900 MDT, nearly all the precipitation was in the form of rain.

Following the method of Chapter 4, searches for regions of probable wet growth of hail and graupel were done for the volumes preceding the major hail flux peaks at 1759 and 1832 MDT. Indications of small regions of enhanced LDR above the freezing level were found in the radar volumes starting at 1741 and 1753 MDT (which preceded the first peak), as well as 1818 MDT (which preceded the second peak). However, these regions (not shown here) were small compared to the one plotted in Figure 4.16. So, while wet growth seems to have occurred in this storm, it probably did not occur to any significant extent.

5.5 Discussion and summary

Based on the available data, some general observations can be made about this case. The time of 1900 MDT serves as a turning point in the storm’s evolution. Before this time, hail fluxes and areal coverage were significant, high-reflectivity (50 dBZ and greater) regions extended to high altitudes, and CG production (especially that of negative CGs) was at a minimum. After this time, hail was nearly non-existent, high-reflectivity regions extended to lower altitudes, and negative CG production was at a maximum.

Unfortunately, IC flash data do not extend far into this high CG production phase, but the available data show that high IC flash rates occurred during the most intense portions of the storm'’s lifetime, when significant precipitation was developing aloft. Following the typical pattern, the first time the IC flash rate peaked it did so before the fallout of significant precipitation. This is consistent with the observations of other researchers (Williams et al., 1989a,b; Changnon, 1992; Carey and Rutledge, 1996). Note also that, during this first major pulse, the peak in the fallout of significant hail preceded the peak in the fallout of significant rain. This is reasonable since the hail will have a larger terminal fall speed, so the hail should reach the ground in bulk before the rain. The second major IC peak occurred after the fallout of significant precipitation. However, this second pulse involves the superposition of the outputs of two distinct and strong cells, so this second peak is not necessarily inconsistent with the typical pattern. However, even though there was a superposition of two major cells, the fallout of significant hail still led the fallout of significant rain, as was observed for the previous storm pulse.

After 1900 MDT, the storm evidently lost its ability to produce significant hail. It also featured a weaker vertical structure, in terms of radar reflectivity. This implies that, at least in an average sense, the cell updrafts after 1900 MDT were weaker than they were before this time. Thus, the increase in negative CG production after this time appears to be consistent with the elevated dipole hypothesis of MacGorman and Nielsen (1991). Because IC data only extend to 1920 MDT, corroborative evidence cannot be found in them. However, before 1920, IC flash rates - though still high - are decreasing. Thus, there is rough consistency, based on the elevated dipole hypothesis, between the decreasing IC flash rates with the rising CG flash rates after 1900 MDT - keeping in mind that these are all storm complex, and not individual cell, quantities. However, whether this is truly the collapse of the postulated elevated dipole cannot be verified.

Like the storm of 10 July, the high IC flash rate of this storm may explain why CG flash rates do not change appreciably after the first IC flash rate peak. The large number of ICs may have helped to neutralize the charged core before it began descending after this time. However, it appears that comparable flash rates did not cause a similar reduction in CGs after 1900 MDT. After this time, though, the available evidence implies that the average updraft strength in the storm complex had decreased. Perhaps this decrease in updraft strength allowed the negatively charged core to descend before it was largely neutralized by IC flashes.

Wet growth of hail and graupel probably did occur during this storm, though the available LDR data seems to indicate that, not unlike 10 July, wet growth was overshadowed by dry growth, and thus may not have played a significant role in this thunderstorm’s charge distribution.

Rain fluxes were high for this storm, especially compared to 10 July. In addition, peak rain rates were high. Peak rates in the gridded data typically exceeded 150 mm h^-1 from 1730 to 1930 MDT. Before and after this period, peak rates generally varied between 60 and 90 mm h^-1. These averaged rates imply that instantaneous rates could have been significantly higher. High precipitation rates conceivably could allow the precipitation current to substitute for the CG lightning current, provided the rain was negatively charged. However, it does not necessarily explain why there is a dearth of CGs before 1730 and a relatively large number of CGs after 1930, even though peak rain rates and fluxes were comparable during the two different time periods.

Alternatively, the anti-correlation between the hail fluxes and the negative CG rates could imply that - if the hail were charged negatively - it may have been able to substitute for the CG lightning current. This hail current hypothesis seems to explain the CG lightning pattern better than the rain current hypothesis. Non-inductive charging theory (e.g., Takahashi, 1978) predicts that, for conditions near storm mid-levels, net negative charge is transferred to large ice particles like graupel and possibly small hail, while net positive charge is transferred to small ice crystals (if there is no charge-reversal process active). However, as the large ice (i.e., graupel and hail) falls to the ground, it should begin to melt and form rain, which would still carry net negative charge. Thus, based on non-inductive theory, both hail and rain should be carrying net negative charge. Thus, this apparent inability of the rain current hypothesis to explain the observed CG pattern seems to detract from the likelihood of the precipitation current hypothesis to be true.

Because precipitation currents were not measured during STERAO-A, the precipitation current hypothesis and the charge reversal hypothesis cannot be dismissed completely. However, the elevated dipole mechanism, acting in concert with the high IC flash rates, is an explanation most consistent with the available data.

The implications of these results and inferences will be examined in light of the results of Chapter 4 (the case study of the 10 July storm), and explored further in Chapter 6.