4.1 Atmospheric conditions

Figure 4.1 shows data from a mobile CLASS sounding launched from the Fort Morgan Airport at 1450 MDT on 10 July 1996. Plotted on the Skew-T/Log-p diagram are vertical profiles of temperature, dewpoint, and winds. The sounding appears to be relatively dry. The CAPE is 792 J kg^-1, the shear over the lowest 6 km is 14.4 m s^-1, and the Bulk Richardson Number (BNR) is 7.7. The BNR is very low but is suggestive of possible supercellular development (Weisman and Klemp, 1982). It is so low because the CAPE is also low. The ambient wind shear is weak as well. However, the Lifted Index is -2.8, which 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 the storm initially developed on the order of 80-90 km north of the sounding launch location. Also, conditions could have changed between the time of the sounding and the time of storm development (i.e., a short wave may have propagated through the region after the sounding was taken). Note the relatively weak northwesterly winds at mid-levels. These may have been responsible for the storm’s slow propagation to the south and east after it developed.

4.2 Overview of storm evolution and major transition

Figure 4.2 (a-i) shows a series of horizontal cross-sections of reflectivity (dBZ) obtained from the CSU-CHILL radar at 0.5 km AGL for the 10 July 1996 storm. The cross-sections are spaced approximately 30 minutes apart. When full-volume coverage of the storm began at 1629 MDT, the storm consisted of three cells, arranged in a line from northwest to southeast. At this point the storm was in southwestern Nebraska, near the town of Kimball. Because of its initial proximity to this town, the storm also shall be referred to as the Kimball storm. At 1629 (Figure 4.2a), the northwest cell (Cell 1) is separated from the other two cells by approximately 20 km. The other two cells (Cells 2 and 3) are enclosed by a common border of 30-40 dBZ reflectivity. However, the strongest cell is Cell 2, with low-level reflectivities peaking at greater than 50 dBZ. By 1701 MDT (Figure 4.2b), the Cell 1 was nearly gone, and Cell 2 had absorbed Cell 3. At this time, another cell had developed between Cells 1 and 2. This cell (Cell 4) had reached over 40 dBZ in low-level reflectivity at this time. Cell 2 was quite intense, with greater than 50 dBZ reflectivities encompassing a relatively large area, as well as a well-defined low-level notch in the core. Thirty minutes later, at 1731 MDT (Figure 4.2c), Cell 1 was completely absent, and Cell 4 had developed to greater than 50 dBZ reflectivity at 0.5 km AGL. Cell 2 was still strong, though the reflectivity notch was gone. Both cores (Cells 2 and 4) were enveloped by a common 40-50 dBZ border, giving the appearance of a short, but intense squall line segment.

By 1803 MDT (Figure 4.2d), significant growth of new cells had occurred on the northwestern flank of the aggregate storm, and the azimuthal boundary of the radar scan was increased to scan these developing cells. Thus, Cells 2 and 4 from 1731 now comprised the southeasternmost cells of the line. Here there were still two separate reflectivity cores, but of these two cells, Cell 4 had decreased in intensity, while Cell 2 was still intense, and had elongated to a length of approximately 20 km. The developing cells (Cells 5 and 6) comprised the northwestern flank of the storm, and of these cells Cell 6 was the most intense in terms of low-level reflectivity at this time. Note the weak echo well to the southeast of the main line. This is a manifestation of the huge (radar-detected) anvil of this storm, which was advected to the south and east of the main line. This anvil was present in many low-elevation scans, indicating the presence of a significant area of stratiform rain/virga.

By 1834 MDT (Figure 4.2e), major changes had occurred in the storm complex. Cell 2 in the line from 1803, which at that time seemed quite intense, had dissipated, and the strongest reflectivity core was now Cell 4, which at 1803 seemed rather weak The northernmost cells (Cells 5 and 6) had weakened as well, with Cell 6 absorbing 5. Another cell, Cell 7, formed behind Cell 6.

At 1900 MDT (Figure 4.2f), the storm complex had redeveloped on its northern flank, and the strongest cell at 1834, Cell 4, was now quite weak Cell 6 was the most intense in terms of low-level reflectivity. By 1932 MDT (Figure 4.2g), the radar scan continued to scan a larger sector to the west, in order to obtain data for a weak cell, Cell 8, that formed on the northern flank after 1900. The strongest cell complex from 1900, Cell 6, now had a large area of greater than 50 dBZ reflectivity at 0.5 km AGL. There was also a well-defined notch on its eastern flank, suggesting strong near-surface inflow. It was around this time that the storm exhibited characteristics of a supercell, including being single-cellular, having mid-level rotation (as indicated by Doppler velocity data), having a low-level reflectivity notch, and having a Weak Echo Region (WER). This similarity to a supercell will be discussed in more detail later.

Between 1932 and 2001 MDT, the weak northern cell (Cell 8) dissipated, and the main cell (Cell 6) split into two main reflectivity cores, a northern one (Cell 9) and a southern one (Cell 6). Because the storm was moving mostly southward during this period, the southernmost cell would be considered the “right-mover”, and the northernmost offshoot the “left-mover”. This kind of storm splitting is typical of supercells (Weisman and Klemp, 1982, 1984), and usually the right-mover survives because the storm dynamics forced by the ambient shear profile favors the development of the right-mover and the dissipation of the left-mover. Similar behavior was observed in this case. By 2001 MDT (Figure 4.2h), the right-mover (Cell 6) was still above 50 dBZ at 0.5 km AGL, but the left-mover (Cell 9) was dissipating. Cell 6 redevelops one last time, but by 2030 MDT (Figure 4.2i) it was well into its dissipating stage. In fact, this storm, which appeared supercellular only a few minutes earlier, dissipated so quickly that radar coverage was terminated after the 2039 MDT volume scan due to a lack of ITF-inferred lightning and because maximum near-surface reflectivities were under 40 dBZ.

Figure 4.3 (a-b) shows a vertical cross-section of reflectivity (Figure 4.3a) and single-Doppler velocity (Figure 4.3b) constructed from the radar volume begun at 1932 MDT. The cross-section is taken along the 45° azimuth. In Figure 4.3a, a Weak Echo Region (WER) is clearly visible at approximately 80 km from CHILL and 2 km AGL. The high reflectivities above this WER are suggestive of a strong updraft in this region suspending large precipitation particles aloft. From Figure 4.3b, strong convergence along this radial exists in the vicinity of the main precipitation shaft. (Negative velocities are toward the radar and positive velocities are away.) The low-level outbound velocities on the radar side of this shaft are suggestive of significant inflow to the WER/updraft region. Note the significant divergence aloft, above the core. The zero velocities in the areas of deleted reflectivity data are a consequence of the K_dp calculation program used on the radar data after it has been edited. The program unfortunately sets missing or deleted velocity data values to zero. Note, however, that in Figure 4.3a, the white pixels in the reflectivity core are not due to missing data but instead due to the reflectivities being higher than the given scale can represent. The existence of a WER and the significant mid-level convergence are suggestive of a supercell.

Figure 4.4 (a-b) show a horizontal cross-section of reflectivity (Figure 4.4a) and single- Doppler velocity (Figure 4.4b) at an elevation angle of 3.5° for the radar volume starting at 1932 MDT. As can be seen from Figure 4.4a, the main core of the storm resides along the 45° azimuth, justifying the choice of this azimuth in Figure 4.3. Note that Figure 4.4 uses the same magnification as Figure 4.3, and distances are in km. In the velocity data, there appears to be a pair of cyclonic couplets in the general vicinity of the core. The first resides at a distance of approximately 80 km, at an azimuth just less than 45°. The second exists at about the same distance, but further to the west along the 40° azimuth, and is more sharply defined than the first. These velocity couplets are suggestive of mid-level (approximately 5 km AGL at this distance and elevation angle) rotation. Such mid-level rotation, known as a mesocyclone, is present in all supercells by definition.

To summarize storm evolution, this storm complex began as a multicellular line, but late in its life it became single-cellular, and even displayed aspects common to supercells. However, this “supercellular” stage was very short-lived, and thus could not be described as quasi-steady. Since longevity and steadiness of strength are requirements for storms to be identified as supercells, this storm probably was not a true supercell during this stage. Thus, it shall be called a “quasi-supercell” at this time. This type of ambiguity in storm classification has been noted by other researchers (e.g., Burgess and Lemon, 1990). In fact, Foote and Frank (1983) have suggested the creation of an intermediate category between multicell and supercell: the Westplains storm. Regardless of what it should be classified as, it is clear that this storm underwent a major transition during its long lifetime, which makes it a prime candidate for study in this thesis.

4.3 Comparison of lightning and bulk precipitation rates

Five-minute IC flash rates from the ITF were calculated following the method outlined in Chapter 3. Five-minute NLDN CG rates also were calculated for this storm complex. The result is Figure 4.5, which shows these rates as functions of time. There are several notable features in this plot. The first is that the entire storm complex, throughout its lifetime, produced relatively few CGs (and the overwhelming majority of these were of negative polarity, so CG rates were not separated into positive and negative CG components). The peak flash rate was 9 in a 5-minute period, which is not very high (Uman, 1987), especially given the magnitude of this storm as viewed by radar. This peak flash rate occurred near 1805 MDT. Before and after this time, CG rates were quite low, generally varying between 0 and 2 per 5-minute period. After 1835 MDT, CGs were almost nonexistent. Throughout the storm’s lifetime, there were several time periods, lasting anywhere from 10 to 30 minutes, where CGs were not produced at all. The most striking examples of these are the periods of 1640-1710 MDT and 1920-1945 MDT. Turning to the ITF data, this second time period coincided with a maximum in the IC flash rate, with IC flash rates during this time often 200 or more per 5-minute period. This is also the time when the storm exhibited the most supercellular characteristics, with mid-level rotation and a Weak Echo Region (WER). Before this time, when the storm complex was more multicellular, IC flash rates were approximately half these peak values. Indeed, between 1810 and 1910 MDT, IC flash rates were remarkably steady, averaging near 20 per minute throughout the hour, with only minor fluctuations. Before 1810, IC rates were much less steady, undergoing two major peaks with a major lull in between. After the IC flash rates reached their maximum values around 1930, they fell off rapidly and then underwent another major surge, centered on 2005 MDT. After this time the IC rates decreased quickly as the final cell in the storm complex (based on radar observations) collapses.

Figure 4.6a shows NLDN CG ground strike locations overlaid on low-level reflectivity for the 1803 MDT volume scan. CGs shown are those that occurred during the volume scan period (approximately 6 minutes in duration). Note that from Fig. 4.5, this time period nearly coincided with the CG rate peak for the storm. It is apparent that most of the CGs were not associated with the most intense cells along the line. Instead, the northernmost cells, which were approaching maturation as their cores descended toward the ground, were producing the most CGs during this time. The southernmost cells are associated with a single positive CG. Figure 4.6b is the same as Fig. 4.6a, except for the 1809 MDT volume scan, also a time representative of the storm’s peak CG output. The cores of the northernmost cells appeared to have reached the ground, as low-level reflectivities had risen in these cells. The southernmost cells were still intense. Again, most CGs were associated with the weaker cells (in terms of low-level reflectivity), while a single positive CG was associated with the southernmost cells. So it appears that even during the peak CG production period of this storm, CGs were associated with cells just reaching maturation, not with the most intense cells along the line.

Figures 4.7a and 4.7b are similar to Figures 4.6a and 4.6b, respectively, except that mean horizontal positions of ITF ICs are plotted instead of NLDN ground strike locations. It is clear that the more intense southern cells are producing more ICs than the maturing northern cells.

To gain a better sense of the vertical extent of each cell around this time, Figure 4.8 (a-d) was created. Each plot is a vertical cross-section of reflectivity for the 1809 MDT volume taken along the given distance north of the CHILL radar. Figure 4.8a is for the northernmost cell in the line from Figure 4.7, Figure 4.8b is for the next cell to the south, and so on through Figures 4.8c and 4.8d, which are for the southernmost cells along the line. From Figure 4.8 it is apparent that, although the vertical extents of the 30 dBZ contours for all cells are similar (approximately 10.5 km AGL or slightly greater), the vertical extents of the 50 dBZ contours for the two southernmost cells, the ones which produced very few CGs around this time, are much greater than the northern two cells, which accounted for most of the CGs. The low-CG cells show 50 dBZ up to 6 km AGL or more, whereas the CG-producing cells either lack 50 dBZ altogether (as in Figure 4.8b), or it extends to a significantly lower altitude (4 km in Figure 4.8a). This is strong evidence that the low-CG cells were much more intense than CG-producing ones, and that the internal structures of the two types of cells were different, even though both cell types extended to similar heights. Thus, throughout the lifetime of this storm, very few CGs can be attributed to the most intense cell(s) at any specific time.

Using the methods detailed in Chapter 2, low-level (0.5 km AGL) rain and hail mass fluxes were calculated for each CHILL radar volume obtained during this storm. Total radar-inferred precipitation mass fluxes, along with 5-minute ITF IC flash rates, are plotted versus time in Figure 4.9. Note that these values are for the entire storm complex, not any individual cell. ITF IC rates are plotted at the start of each 5-minute period, and total mass fluxes are plotted at the start of each volume scan. Note that this mass flux includes both rain and hail. Following the method of Chapter 2, these rates were determined separately for each precipitation type, then were added together to create Figure 4.9. Recall that the ITF IC flash rates exclude all flashes lasting less than 1 ms.

Before any trends are inferred from this figure, it is important to note two times when radar volume scan coverage was opened up to the west to include more storm-associated cells that were in their developing stages. These times were 1751 MDT and 1907 MDT. For several minutes before these times, precipitation rates are artificially low because the radar was not scanning the entire storm complex, but not excessively low since the excluded cells were still in their developing stages and probably would not have made major contributions to the total precipitation flux. Thus, the trends around these times are not necessarily corrupted, but the magnitudes of the trends may be misleading. The dip in precipitation flux around 1745 probably is somewhat more severe than it was in reality, because the developing cells to the northwest were not being scanned at this time. Also, the peak that occurs at 1907 is more sharply defined than it should be, because the immediately previous volumes excluded another developing cell to the northwest. Despite these issues, the overall trends should be correct, and there probably is little risk in interpreting them more in terms of storm-related phenomena, rather than the effects of radar scanning strategies.

From Figure 4.9, IC flash rates and near-surface precipitation mass fluxes appear to be highly correlated at times, and at other times the low-level precipitation flux appears to lag the IC flash rate by up to several minutes. The most notable example of the precipitation flux lagging IC flash rate occurs early in the analysis period, when IC flash rate peaks near 1720 MDT and precipitation flux doesn’t peak until just before 1730 MDT. Subsequent peaks in IC flash rate and precipitation flux – around 1810, 1825, and 1840 MDT – appear to line up roughly in time. Precipitation flux again lags IC flash rate from roughly 1915 MDT through the end of the observational period. Note, however, that the final peak in the storm’s IC flash rate, around 2005 MDT, is associated with a very small peak in the precipitation mass flux. At this time the storm consisted of a single cell, so it is perhaps not surprising that the peak should be smaller than earlier times when there were more cells. Recall that flash rates and precipitation fluxes are for the entire storm complex, not just individual cells. The minima in both IC flash rates and precipitation mass flux that occur between 1930 and 2000 MST coincide with the storm splitting, which was discussed earlier. The final peak in the precipitation flux and flash rate coincides with the right-mover redeveloping while the left-mover dissipated.

Figure 4.10 shows the storm’s separate rain and hail mass fluxes as functions of time. This plot was created by separating the contributions of the two classes of hydrometeors to the total precipitation mass flux, as detailed in Chapter 2. By comparing Figures 4.9 and 4.10, it is obvious (and not surprising) that the rain flux contributes the most to the total precipitation mass flux, with the hail contribution approximately one-tenth the rain contribution. Thus, no further discussion of the rain flux trends will be presented since these trends are nearly identical to the total precipitation flux trends, which have already been discussed. Turning to the hail flux trends, it is apparent that they are more variable than the rain (and hence total precipitation flux) trends. However, the hail fluxes are still relatively well-correlated with the rain fluxes, with some exceptions. The major exception occurs just after 2000 MDT, when the rain flux achieves its last major peak, but the hail flux is at a minimum. Also, the relative amplitudes of the hail and rain maxima and minima do not match well. For example, a minor, perhaps insignificant (in terms of radar estimation error) peak in rain flux at 1827 MDT is coincident with a relatively sharp peak in hail flux.

By comparing Figures 4.9 and 4.10, it can be seen that IC flash rates and low-level hail fluxes are well correlated, though there is a tendency for the low-level hail flux to lag the IC flash rate at times, especially near the beginning and end of radar coverage.

Recall, however, that rainfall estimates are subject to at least a 10-20% error, and hail rate estimates are subject to errors on the order of 50%, possibly more. Thus, some of the smaller peaks and troughs in the flux plots are likely within the error bars of this analysis, though it is interesting that the variations still trend the IC flash rate well.

This precipitation flux calculation algorithm was compared with the one described by Carey and Rutledge (1997), to gain a better sense of how it compares to that technique. The method of Carey and Rutledge (1997) was applied to flux calculations for all volumes from the Kimball storm and compared to the present study'’s method in Figure 4.11, which shows the time history of total precipitation flux at 0.5 km AGL according to each method. Note that both algorithms use the same methods to calculate hail fluxes, but their rain flux calculation techniques differ. Carey and Rutledge (1997) utilize Z_dr as well as K_dp and reflectivity, whereas our method uses only the latter two. It is apparent that Carey’s and Rutledge’s method almost always produced higher rain (and hence higher precipitation) fluxes than this method. Going back to Figure 4.10, the periods of greatest disagreement tend to coincide with the periods of greatest hail flux. In addition, during the approximately 20-minute period after 1900 MDT when agreement between the two methods is best, hail fluxes are low. This is consistent with hail contamination if the reflectivity field, upon which rain rate calculations based on Z_h and Z_dr critically depend, is affected by the presence of hail. Carey and Rutledge (1997) used the results of Golestani et al. (1989) to correct these calculations for hail contamination, but based on this simple comparison, it appears that this method still may be subject to significant hail contamination in certain cases.

4.4 Identification of bulk precipitation types

Following the method outlined in Chapter 2, bulk hydrometeor identification was performed on each radar volume from this storm. To review, the following categories of bulk hydrometeors below the freezing level are identified by the algorithm: small (< 2 cm) wet (i.e., melting) hail, small wet hail mixed with rain, large (> 2 cm) wet hail, large wet hail mixed with rain, and rain alone. See Table 2.3 for a synopsis of the hydrometeor identification matrix.

Based on the afternoon (1450 MDT) CLASS sounding launched from the Fort Morgan airport on this day, the freezing level was determined to be 3.29 km AGL (where AGL is defined from the elevation of the CHILL radar - 1432 m MSL).

As mentioned in Chapter 2, the algorithm makes use of a reflectivity gradient threshold to avoid the possibility of Z_dr bias in the gridded data. This threshold eliminates from consideration those grid points that have excessive reflectivity gradients. This threshold was set at some typical reflectivity gradients values - 10, 15, and 20 dBZ km^-1. Additionally, sensitivity tests were performed in order to better estimate the proper threshold setting. For this storm, it was found that the most grid points (below the freezing level) fell under the category of small hail. Figure 4.12 shows the time history of the volume of small hail below the freezing level, as determined using the three different thresholds, for the 10 July storm. Clearly, agreement is superb between the different runs, except for some minor discrepancies at a few times. However, due to the possible errors in this estimation method, these differences are probably insignificant. Regardless, the most restrictive reflectivity gradient, 10 dBZ km^-1, was used in further analysis. This more restrictive threshold is also appropriate because of the long range to the storm throughout the observational period, as estimation errors caused by reflectivity gradients are more likely to occur at farther ranges.

Figure 4.13 shows the volume of small wet (i.e., melting) hail alone (i.e., not mixed with rain) below the freezing level as a function of time for this storm. Note that, although the relative magnitudes of the minima and maxima are different, the volume of small hail below the freezing level very closely trends the low-level radar-inferred hail flux in Figure 4.10. Figure 4.14 shows time histories of the volumes of small hail mixed with rain and rain only, below the freezing level. These volumes are much lower than the volumes of small hail only, suggesting that most of the significant precipitation below the freezing level is small hail. Large hail, and large hail mixed with rain, were not detected by the algorithm. Again, although the relative magnitudes are different, in general the peaks and troughs in small hail mixed with rain match up well with the peaks and troughs in small hail only. The rain-only volume trend does not line up well with either of these trends, but the rain-only volume is quite low, so whether any significance should be given to its trend is uncertain.

Figure 4.15 shows radar-inferred areal coverage at 0.5 km AGL for small hail, small hail mixed with rain, and rain. Interestingly, here the overall trends in the areas of small hail only and rain only match up better than the trends in small hail only and small hail mixed with rain.

In order to better understand the sensitivity of this precipitation identification algorithm to changes in category requirements, some tests were performed on a single volume, 1726 MDT. At this time the small hail category is at its peak, both in volume below the freezing level (167 km³) and areal coverage at 0.5 km AGL (19.5 km²). The K_dp threshold to determine between regions of hail only, and regions of hail and rain, is 0.5° km^-1. This threshold was lowered to 0.25° km^-1. The result was that a significant portion of the small hail category was shifted over to the small hail and rain category. The volume of small hail below the freezing level dropped to 123 km³, while the volume of small hail mixed with rain grew from 20.5 to 52.5 km³. A similar trend was seen in the areal coverage of the two categories at 0.5 km AGL. This suggests that a significant portion of the small hail category may be more properly placed in the small hail and rain category. This would tend to increase agreement between this precipitation identification method and the precipitation flux method of Section 4.3, since the precipitation identification method found most classifiable precipitation areas to be small hail, even though the flux method found significant rain fluxes. However, the precipitation identification algorithm was not rerun for the entire storm using this lower specific differential phase threshold, mainly because it would have not changed the result that there was little or no radar-detectable large hail in this storm, so that small hail accounted for most of the hail fall. This is perhaps the most important result of this precipitation identification study. Due to the subjective nature of the precipitation identification matrix, the strongest trends are probably real, but absolute values should be taken with less faith, as the preceding sensitivity study and comparison with the more quantitative flux method show.

According to the results of Saunders et al. (1991) and Saunders and Brooks (1992), extensive charge separation should not occur during collisions in a wet growth regime, since the rebounding efficiency goes to zero. If one accepts these results, then charge separation may not be very active in the cloud region above the freezing level but below the level of charge reversal (found to range from approximately -10 °C to -20 °C depending on the laboratory study). This could suppress the development of a lower positive charge region that normally could provoke CG lightning.

The radar data were examined in order to determine whether or not wet growth may have occurred over an extensive region in this storm. Based on several studies (Bringi et al., 1986; Holler et al., 1994; Waterman, 1969), it is believed that LDR values in excess of approximately -26 to -25 dB (at S-band frequencies) above the freezing level may be indications of hail; in particular, hail in a wet growth mode. Note that these enhanced LDR values should be co-located with an extended region of high radar reflectivity (~ 40-50 dBZ or greater) in order to avoid artifacts caused by low signal-to-noise ratios and large reflectivity gradients. These results have been used by researchers (e.g., Kennedy et al., 1997) to identify probable regions of wet growth hail above the freezing level.

In the present study, mid-level elevation angle scans from radar volumes preceding each major peak in the hail flux (1726, 1809, 1827, 1925, and 2019 MDT) were examined for extended regions of enhanced LDR values that could be signatures of wet growth hail. In particular, since the radar volumes were spaced typically on the order of 6 minutes apart, the second to last radar volume before each major flux peak was examined since the high-elevation scans from this volume were spaced on the order of convective time scales (6-10 minutes) from the low-elevation scans (which were the most influential in the hail flux calculations) from the volume occuring during the hail flux peak.

In general, few extended regions of enhanced (i.e., > -25 dB) LDR above the freezing level (3.29 km AGL in this case, though it may have been slightly higher inside the storm itself) were found. Sometimes there were a handful of contiguous gates of enhanced LDR that were coincident with a high reflectivity region, but overall no significant signatures were found, save for the radar volume beginning at 2357 MDT (which preceded the second major hail flux peak at 1809 MDT).

Figure 4.16 (a-b) shows a vertical cross-section of radar reflectivity in dBZ (Figure 4.16a) and LDR in dB (Figure 4.16b) taken along the 50° azimuth in the 2357 MDT volume. This cross-section passes through the northern end of Cell 2 in Figure 4.2. Approximately 95 km from the radar, at a height of approximately 5 km AGL, an elevated high-reflectivity core is centered (indicated by the arrow in Figure 4.16). Co-located with this core is a an extended region of enhanced (-25 dB and greater) LDR. Note that the LDR scale is set such that values less than -26 dB are not displayed. The other regions of enhanced LDR are likely due to low signal-to-noise ratio or reflectivity gradients, but the region of interest probably is real since it is situated in an area of high reflectivity which extended significantly in both the vertical (as seen here) and horizontal (not shown) directions. Peak LDR values approach -21 dB here, though most values range between -25 and -22 dB. Since this core is centered at 5 km, according to the sounding in Figure 4.1 much of it exists above the freezing level, and below the lower limit for the charge revesal temperature (-20 °C). Thus, it appears that, at least at this time, significant wet growth could have been occuring, which may have led to reduced charge separation in the vicinity of the usual location of the lower positive charge region (Williams, 1989).

4.5 Discussion and summary

Given the uncertain and possibly large errors in hail estimation using multiparameter radars like the CHILL, care needs to be taken when interpreting any trends seen in the precipitation data. This is especially true because the storm was far from the radar (> 50 km throughout its lifetime) and multiparameter observations are less reliable at this distance. However, some general observations can be made about this case.

When this storm was in its most supercell-like stage, around 1930 MDT, its IC flash rate was maximum and CGs were not detected by the NLDN. In addition, even when CG rates were modest, after 1800 MDT, the CGs were almost exclusively associated with maturing cells along the multicellular line, not with the most intense cells.

Based on data obtained from the National Weather Service, there was one report of 1 inch (2.5 cm) hail around 2010 MDT. This is near the time the storm complex reached its peak in low-level reflectivity. However, based on the radar data, large hail was likely not present in great quantity with this storm, and most of the precipitation could be described as being small hail, small hail mixed with rain, or rain. Small hail and small hail mixed with rain were present throughout the storm’s lifetime, though amounts fluctuated with the growth and decay of individual cells. The trends in the output of small hail, whether alone or mixed with rain, tended to follow the low-level hail flux relatively well - which along with the lack of detected large hail suggests that, when the storm produced hail, it was almost exclusively small (< 2 cm) hail.

IC flash rates and low-level precipitation fluxes (either rain or hail, or both) are well-correlated, though sometimes there is a tendency for the low-level precipitation flux to lag IC flash rate by up to several minutes. Lags between IC flash rates and low-level precipitation fluxes is what would be expected from precipitation-based charging theories. According to these theories, as precipitation develops aloft, charge separation occurs which would lead to IC discharges. As the cores descend, IC discharges should decrease (due to increased distance between the upper positive and lower negative charge centers - Williams, 1989) and low-level precipitation flux should increase. This is seen in the Kimball storm at certain times, but normally the IC flash rates and low-level precipitation fluxes trend well, with little or no lag. Due to errors in the precipitation estimation method, as well as the IC flash rate calculations, it is unclear how much significance should be placed on the changes in the existence or non-existence of lag between these two trends. Also, these changes in lag times to some extent may be artifacts of the superposition of many cells, all in differing states of growth and decay, since precipitation fluxes and flash rates were for the entire storm complex. What is clear, however, is that correlation between the two trends exists, which is expected based on precipitation-based charging theories.

This correlation also is consistent with the convective theory of cloud electrification (Vonnegut, 1963, 1982; Moore, 1977). According to this theory, strong updrafts should enhance the entrainment and mixing of ambient space charge, while also enhancing precipitation development. Thus, charge separation and subsequent IC flashing still should be correlated to precipitation development in some way, even though according to this theory charge separation need not occur during collisions between precipitation particles.

Thus, to summarize, it is clear that this storm had at least one intense, active cell throughout almost its entire lifetime. It also featured strong enough updrafts to produce at least small hail throughout its lifetime, though such hail production was clearly highly variable. Very low production of CGs occurred throughout this storm’s lifetime, but when CGs did occur in (modest) numbers, they were associated with cells of less intensity (based on comparisons of vertical cross-sections of reflectivity), while the most intense cells favored IC discharges. In addition, at this storm’s most intense stage, the quasi-supercellular stage, IC flash rates were maximum and CGs were non-existent. All of these facts are consistent with the elevated dipole hypothesis of MacGorman and Nielsen (1991). The apparently strong updrafts of this storm were perhaps enough to loft the negatively charged core to a higher altitude, so that ICs were preferred over CGs.

However, this does not explain why, when cell updrafts collapsed and the cores approached ground, CGs were not favored over ICs. This is perhaps where the observed high IC flash rates could become important.

The IC flash rate for this storm was relatively high, averaging near 100 flashes or more per 5-minute period for much of the storm’s lifetime. It could be possible that these high IC flash rates neutralized most of the charge before it began to descend, so that CGs were not favored even as the cores approached ground.

It appears that, at least around 2357 MDT, significant wet growth of hail and graupel may have occurred. According to the results of Saunders et al. (1991) and Saunders and Brooks (1992), this could have reduced charge separation and suppressed the development (or maintenance) of a lower positive charge region, and thus may have suppressed CG lightning. However, extended regions of possible wet growth were not found at other times. Also, based on the available data, it appears that dry growth (corresponding to reduced LDR values) probably occurred more often than wet growth, due to the paucity of enhanced LDR signatures. Thus, it seems doubtful that wet growth could have significantly affected the charge distribution within this storm.

Radar-inferred precipitation fluxes for this storm are not especially high, considering that all cells in the grid are included in the calculations, and that other severe storms have exhibited higher fluxes (Carey and Rutledge, 1997). However, for precipitation current to substitute for lightning current, instantaneous precipitation rates may be more important, rather than the area- and time-averaged rates used to calculate the fluxes. However, throughout the storm’s lifetime, peak rain rates for each radar volume are almost always below 60 mm h^-1, and do not even exceed 40 mm h^-1 in many volumes. Hail rates are similarly low (especially when compared to other severe storms - Carey and Rutledge, 1997). Though these estimates are made from gridded and averaged radar data, it seems likely that, because of the predominance of relatively low precipitation rates throughout the storm’s area and lifetime, instantaneous precipitation rates probably were not much higher than the peak values observed in the gridded data. Furthermore, similar precipitation rates can be observed in different storms which produce more CGs. So the precipitation current hypothesis, too, seems unlikely based on the available data.

Thus, based on the available data, a reasonable explanation of the low CG production by this storm would be the following. For many cells in the Kimball storm, strong updrafts loft the negative-charge region to higher altitudes than normal. This tends to favor ICs over CGs. In addition, the strong updrafts also enhance charging and subsequent IC flash rates. The high IC flash rates act to neutralize the separated charge, so that when the updrafts weaken and the cores descend, little charge is available to force CG flashing. This hypothesis appears to be consistent with the available data. However, because the other hypotheses have not been adequately tested, they cannot be ruled out completely.

The implications of these results and inferences will be examined in light of the results of Chapter 5 (the case study of the 12 July storm), and explored further in Chapter 6.