Abstract

This study compares severe weather reports associated with the nine convective system morphologies used in a recent study by Gallus et al. to an additional morphology, supercell storms. As in that previous study, all convective systems occurring in a 10-state region covering parts of the Midwestern United States and central plains were classified according to their dominant morphology, and severe weather reports associated with each morphology were then analyzed. Unlike the previous study, which examined systems from 2002, the time period over which the climatology was performed was shifted to 2007 to allow access to radar algorithm information needed to classify a storm as a supercell.

Archived radar imagery was used to classify systems as nonlinear convective events, isolated cells, clusters of cells, broken lines of cells, squall lines with no stratiform precipitation, trailing stratiform precipitation, parallel stratiform precipitation, and leading stratiform precipitation, and bow echoes. In addition, the three cellular classifications were subdivided to allow an analysis of severe weather reports for events in which supercells were present and those in which they were not.

As in the earlier study, all morphologies were found to pose some risk of severe weather, and differences in the two datasets were generally small. The 2007 climatology confirmed the theory that supercellular systems produce severe weather more frequently than other morphologies, and also produce more intense severe weather. Supercell systems were especially prolific producers of tornadoes and hail relative to all other morphologies, but also produced severe wind and flooding much more often than nonsupercell cellular morphologies. These results suggest that it is important to differentiate between cellular morphologies containing rotation and those that do not when associating severe weather reports with convective morphology.

1. Introduction

Many studies have attempted to classify mesoscale convective systems by organizational mode. Bluestein and Jain (1985) classified squall lines in terms of their development as broken line, back building, broken areal, and embedded areal. Parker and Johnson (2000) considered squall lines with trailing stratiform precipitation, parallel stratiform precipitation, and leading stratiform precipitation. Jirak et al. (2003) used satellite and radar data to separate mesoscale convective systems into four categories: mesoscale convective complexes, persistent elongated convective systems, meso-β circular convective systems, and meso-β elongated convective systems. The same study also classified systems by development on radar in terms of the presence of stratiform precipitation, whether the initial convection was linear or areal in coverage (or a combination), and whether systems merged with others. Baldwin et al. (2005) used 1-h rainfall amounts to develop an automated classification procedure that separated rainfall events into stratiform nonconvective, convective linear, and convective cellular categories. Other studies used isolated cells as an organizational mode (Grams et al. 2006), and Baldwin et al. (2005) alluded to classifying systems as both isolated cells and clusters of multicells. Gallus et al. (2008; hereafter G08) used several of these morphologies and related severe weather reports in the midwestern United States to morphology type and added clusters of cells, squall lines with no stratiform precipitation, and nonlinear convective systems to the typology.

No matter which classification system is used, classification of convective system morphology can be difficult. Some subjectivity is inherent in the classification since some systems exhibit aspects of multiple morphologies (G08) with changes occurring both spatially and temporally. For example, Parker and Johnson (2000), Parker (2007), and Storm et al. (2007) noticed that squall lines with leading stratiform and parallel stratiform precipitation had a tendency to transform gradually to trailing stratiform precipitation squall lines. In addition, the assignment of severe weather reports to particular morphologies also poses challenges (see G08 for a detailed discussion). Many of the difficulties are related to the methods used to report storms and how they appear in the National Climatic Data Center’s (NCDC) Storm Events Database. Such issues include the overreporting or underreporting of severe wind and hail events (Trapp et al. 2006), the affects of population density on the reporting of severe wind events (Weiss et al. 2002), the methods by which tornadoes are reported (Doswell and Burgess 1988; Trapp et al. 2005b; Verbout et al. 2006), and the fact that most wind and hail reports are given as point measurements rather than as swaths, as tornado reports are.

Nonetheless, it appears potentially useful to attempt this classification since certain morphologies have been shown to favor producing one or more types of severe weather. Parker (2007), among others, has shown that parallel stratiform and leading stratiform lines tend to produce more flooding than other systems. G08 also noted the tendency for trailing stratiform lines and nonlinear convective events to produce more flooding reports, and they also showed that cellular systems tended to produce more hail and tornado reports. Bow echoes and trailing stratiform events have been shown to produce a greater percentage of all severe wind reports and tend to have a large wind-to-hail report ratio (Klimowski et al. 2003; G08). One shortcoming of those studies, however, is the exclusion of supercells as a morphology or storm type. G08, for instance, found that although cellular systems were most common in the summer, most of the severe weather reports associated with them occurred in spring, and speculated that the spring events might be supercellular in environments of greater wind shear while summer events might be nonsupercellular. Additional data and more detailed analysis are needed to identify supercells.

Supercells have long been known as producers of some of the most intense severe weather (Doswell and Burgess 1993; Moller et al. 1994), and several methods have been devised to allow forecasters to identify supercells using radar or satellite (e.g., Forbes 1981; Johns and Doswell 1992; Burgess and Lemon 1990; Moller et al. 1994). More recently, teams from the National Severe Storms Laboratory have written two algorithms that aid in the identification of supercells and tornado vortex signatures from the Weather Surveillance Radar-1988 Doppler (WSR-88D) network: the Mesocyclone Detection Algorithm (MDA) (Stumpf et al. 1998) and the Tornado Detection Algorithm (TDA; Mitchell et al. 1998). Stumpf et al. (1998) showed the MDA to be a better identifier and predictor of supercells than the recent operational algorithm, the WSR-88D Build 9.0 Mesocyclone Algorithm, by comparing the probability of detection and critical success index on a dataset. Mitchell et al. (1998) showed the TDA to be a better identifier and predictor of tornadoes than the recent operational algorithm, the WSR-88D Tornadic Vortex Signature Algorithm, by comparing the probability of detection, false alarm rate, critical success index, and Heidke skill score on a different dataset. The MDA may be particularly useful for identifying supercells since a defining characteristic of a supercell is the presence of a deep, persistent mesocyclone (Doswell and Burgess 1993). The MDA enables meteorologists to detect rapid rotation in all kinds of storms including ones in which the rotation may be difficult to see due to cluttering of reflectivity, distance from radar, or other reasons.

The present study takes the morphologies used in G08 and expands them to include supercellular versions of the three cellular morphologies. To make use of improved supercell detection algorithms, however, 2007 data were used instead of the 2002 data used in G08. Two hypotheses are tested: 1) trends in severe weather reports associated with each morphology found for the 2002 dataset in G08 remain true for the 2007 dataset and, more importantly, 2) supercell morphologies will produce more severe weather more frequently and produce more intense severe weather than will nonsupercellular morphologies. Section 2 outlines the data sources and methodology for the study, while section 3 provides the results and analysis of the study. Conclusions and discussion follow in section 4.

2. Data and methodology

To preserve continuity between the present study and G08, as many aspects as possible of the data collection and methodology used in G08 were followed in the present study. Radar data came from the University Corporation for Atmospheric Research/Mesoscale and Microscale Meteorology Division’s (UCAR/MMM) image archive for warm season precipitation episodes (information online at http://locust.mmm.ucar.edu/case-selection/). The images are mosaics from various sources, but most are of composite reflectivity with spatial and temporal resolutions of 2 km × 2 km and 30 min, respectively. For the few periods in which data from this archive were unavailable (the longest such period being 24 h), the interactive radar feature on the Iowa Environmental Mesonet Web site (http://mesonet.agron.iastate.edu) was used instead, with settings matched as closely as possible to those of the UCAR image archive. Output from the MDA, which was accessed both through the NCDC mesocyclone product and level III storm attribute data, was used to identify supercells. Severe storm reports were obtained from NCDC.

The period of study extended from 1 April 2007 to 31 August 2007. The domain of the study (Fig. 1) matched that used in G08 and consisted of a 10-state region from the southern Great Plains through the upper Midwest. All convective events that formed within this domain and time period were included in the study as long as they met the following radar characteristics (as in G08):

  1. areal coverage of reflectivity greater than 10 dBZ of at least 6 km × 6 km,

  2. at least one pixel of data of at least 30 dBZ within the area in characteristic 1, and

  3. temporal maintenance of characteristics 1 and 2 for at least 1 h (at least three frames).

Fig. 1.

Ten-state domain used in the study (from G08).

Fig. 1.

Ten-state domain used in the study (from G08).

All convective systems meeting these criteria were classified using visual inspection according to their dominant morphology. Nine morphologies were used (Fig. 2): three were cellular, consisting of isolated cells (ICs), clusters of cells (CCs), and broken lines (BLs); five were linear, consisting of no stratiform precipitation squall lines (NSs), trailing stratiform squall lines (TSs), parallel stratiform squall lines (PSs), leading stratiform squall lines (LSs), and bow echoes (BEs); and the final morphology was the nonlinear convective morphology (NL). To be classified as one of the linear morphologies, a system had to be at least 75 km in length, have an eccentricity (ratio of major axis to minor axis) of at least 3:1, and persist for at least 2 h. Cellular systems had to contain identifiable cellular elements. If the elements were connected by relatively weaker reflectivities (around 30 dBZ), the systems were classified as CC; if not, or if only very weak reflectivities (less than 10 dBZ) connected individual cellular elements, the systems were classified as IC. If the cellular elements were organized in a discontinuous line, the systems were classified as BL. Linear systems were classified according to their pattern of stratiform precipitation, and lines with no stratiform precipitation, or in which the stratiform precipitation was narrower than the convective part of the line, were classified as NS. Bow echoes were not required to possess stratiform precipitation but did consist of a line in which part of the line bowed out at a speed that was clearly faster than the other parts of the line. If a system met the three main radar criteria but did not fit into one of the linear or cellular morphologies, it was classified as NL.

Fig. 2.

Schematic drawings of systems from the nine morphologies. Abbreviations are as follows: IC, isolated cells; CC, cluster of cells; BL, broken line of cells; NS, squall line with no stratiform precipitation; TS, trailing stratiform precipitation; PS, parallel stratiform precipitation; LS, leading stratiform precipitation; BE, bow echo; and NL, nonlinear convective system (from G08). Storm motion can be assumed left to right in all cases.

Fig. 2.

Schematic drawings of systems from the nine morphologies. Abbreviations are as follows: IC, isolated cells; CC, cluster of cells; BL, broken line of cells; NS, squall line with no stratiform precipitation; TS, trailing stratiform precipitation; PS, parallel stratiform precipitation; LS, leading stratiform precipitation; BE, bow echo; and NL, nonlinear convective system (from G08). Storm motion can be assumed left to right in all cases.

In classifying systems, only the dominant morphology was considered. All severe reports that occurred with that system were assigned to that morphology. However, if a system displayed properties of a different morphology for more than 1 h during any time other than the initial and decaying stages of its life, then severe reports that occurred during that time were attributed to the other morphology. Some systems in this study did change morphologies, occasionally several times. Effort was taken to prevent duplicated reports, especially hail and tornado reports (several of which were found), from being overcounted. It is recognized that classifying convective systems by mere visual inspection of radar is subjective, but the quantitative guidelines used for classification should reduce the subjectivity. Nonetheless, systems exhibit a spectrum of morphologies, and a given system may exhibit characteristics of multiple morphologies both between successive scans and within one scan, causing difficulty in the assignment of a morphology. Approximately 5% of the systems proved difficult to classify, either because they evolved rapidly (i.e., did not resemble a particular morphology for at least an hour), or because they exhibited characteristics of disparate morphologies simultaneously. However, a morphology was still assigned to these systems. In fact, an additional morphology was suggested by Schumacher and Johnson (2005), called the training line/adjoining stratiform (TL/AS) morphology. A few of the systems in this study resembled TL/AS characteristics and would have been labeled as such had that morphology been included. However, since the TL/AS morphology was not included in G08, it was not employed in this study.

The severe reports were divided into the same categories used in G08: hail ≥ 0.75 in. (1 in. = 2.54 cm) in diameter but <1 in. in diameter, hail ≥ 1 in. but <2 in. in diameter, hail ≥ 2 in. in diameter, severe wind gusts [severe being gusts ≥50 kt (1 kt = 0.514 m s−1)] <65 kt, severe wind gusts ≥65 kt, tornadoes, floods, and flash floods. In G08, the report of urban/small stream flooding was used. However, changes in the way NCDC’s Storm Events Database classified flooding reports caused the elimination of the term “urban/small stream flooding,” and consolidated it with other low-impact flooding events that no longer appear in the database. Other changes to the flooding reports listed in the database include continuing a flash flood report as a flood report if the definition of a flood event is met from an ongoing flash flood report. This occurred rarely in the study and was ignored. If a system met the radar requirements but was not associated with any reports of severe weather, the system was classified as a nonsevere case.

An additional classification of supercells was included in the present study to determine whether or not systems that contain supercells produce more frequent or more violent severe weather than other morphologies. To be classified as a supercell system, an event must have been classified as cellular and must have contained at least one supercell. Although it has been shown that noncellular systems likely contain embedded supercells (e.g., Miller and Johns 2000), those will not be considered in this study to keep the focus of the study on the system morphologies and not individual convective elements. If at least one supercell was found within a system, all reports for that system were attributed to the supercell morphology.

Because a supercell has been defined as a storm typically possessing a mesocyclone for at least 15 min (Glickman 2000), in the present study any identifiable cellular element from a cellular system that was flagged by the MDA consistently for a period of at least 15 min was considered a supercell. Because radar volume scans are generally produced at a rate of one volume scan every 4–6 min, the minimum number of scans in which a cellular element had to be flagged as a mesocyclone to be considered a supercell was chosen to be four. Several levels of rotation are marked by the MDA, including UNCO and 3DCO, which correspond to uncorrelated rotation at one isolated elevation angle and correlated rotation at two adjacent elevation angles of the radar, respectively, and MESO, which corresponds to correlated rotation at three or more adjacent elevation angles. Only the MESO level was used to mark a cell as possessing a mesocyclone. Because supercells may fluctuate in strength over time, a one-scan break in a sequence of four consecutive scans flagging a cell with MESO was allowed in the defining of supercells. An analysis of how many individual supercells a supercell system contained was not performed. Systems that were only partially inside the domain were only classified as a supercell system if any supercells that occurred within the system occurred within the domain. This process was used for both severe systems and those that did not produce severe weather.

3. Results

a. Comparison with G08

Because a different sample of cases than that analyzed in G08 had to be used to allow for the identification of supercells, a comparison was performed first to examine the generality of the G08 results. Table 1 shows that the sample sizes from the 10-state region in the 2 yr were roughly similar, although the 2007 dataset had more nonsevere cases (events that did not produce any severe reports) and fewer total severe reports. However, the 2007 events produced an average of 10.2 reports of severe weather per system, a value slightly less than the 11.4 found in G08 for the 2002 systems, with BE systems producing the largest average of about 22.5 reports per system in 2007 (not shown).

Table 1.

Number of events classified and producing severe weather or flooding in the current study using 2007 data and from G08, which used data from 2002.

Number of events classified and producing severe weather or flooding in the current study using 2007 data and from G08, which used data from 2002.
Number of events classified and producing severe weather or flooding in the current study using 2007 data and from G08, which used data from 2002.

The contribution from each morphology toward the total number of systems is shown in Fig. 3. The three most common systems were the same in both datasets (NL, IC, and CC), although the largest single contributors were CC systems for the 2007 dataset, contributing nearly 28% to the total number of systems, and NL systems from G08, consisting of 29% of all systems. Note that LS systems were relatively rare, a result similar to G08. Parker and Johnson (2000), in defining leading stratiform systems as a morphology, indicated that leading stratiform lines could also possess trailing or parallel stratiform precipitation. It is likely, then, that some systems were classified as TS or PS instead of LS even if some leading stratiform precipitation was present.

Fig. 3.

Percentage contribution of each morphology to the total number of events for the (top) 2007 and (bottom) 2002 data from G08. Shading implies general morphological type: nonlinear (dark gray), cellular (medium gray), and linear (light gray). Numbers in parentheses indicate the number of events that occurred for each morphology.

Fig. 3.

Percentage contribution of each morphology to the total number of events for the (top) 2007 and (bottom) 2002 data from G08. Shading implies general morphological type: nonlinear (dark gray), cellular (medium gray), and linear (light gray). Numbers in parentheses indicate the number of events that occurred for each morphology.

Regarding broad morphological types, cellular systems dominated in both studies, consisting of 61% of all systems in 2007 and 49% in 2002. Although linear systems composed about the same percentage of all systems (24% and 22% for 2002 and 2007, respectively), the order of frequency between nonlinear and linear systems differed between the two studies. Nonlinear systems consisted of a greater portion of all systems in the 2002 dataset (29%) than they did in 2007 (15%).

Shown in Fig. 4 is the percentage of the total number of severe storm reports produced by each morphology for both datasets. In both years, cellular systems contributed the most severe reports, followed by linear systems. The larger sample size of the cellular morphological types, as well as the fact that a number of cellular events contained supercells (section 3b), probably plays some role in that outcome. Of note, the contribution of storm reports from linear systems in both years (Fig. 4) is noticeably larger than the fraction of systems identified as linear (Fig. 3). This result likely reflects both the relatively large sizes and the increased organization of the linear systems. Also common between the two datasets is that CC systems contributed the most reports, producing 33% of all severe reports in 2007 and 22% in 2002, and that LS systems contributed the fewest number of total severe reports, again results likely reflecting the relative numbers of those types of systems. Differences between the two datasets include substantial disparities in the percentages of total reports contributed by NL, BE, BL, and CC systems and different rankings of the top three most productive systems, as CC, BL, and BE systems were the top three (in that order) in 2007 compared to CC, NL, and IC in 2002.

Fig. 4.

As in Fig. 3, but for percentage contribution of each morphology to the total number of severe storm reports.

Fig. 4.

As in Fig. 3, but for percentage contribution of each morphology to the total number of severe storm reports.

Shown in Fig. 5 is the percentage of systems that produced at least one report of severe weather by morphology for both datasets. Linear systems were more likely to be associated with at least one severe weather report than cellular and nonlinear systems, again likely due to both their larger sizes and increased organization. In fact, about 75% of all linear systems in the 2007 dataset (85% in G08) produced severe weather, compared to around 55% for cellular and nonlinear systems (66% in G08). When only nonflooding severe reports were considered, all percentage values decreased somewhat due to the presence of systems that produced only flooding reports. The NL and linear morphologies contained more flooding-only systems than did others. The associated stratiform precipitation and large areal coverage of the NL and linear systems likely explains at least some of this difference. Another explanation is that NL systems are generally poorly organized convective systems and often develop in regions of weak vertical wind shear. Thus, given that surface wind speeds during the convective season are generally weak, winds at all levels of the atmosphere are weak, meaning that any convection would be slow moving and hence pose a heightened risk of flooding.

Fig. 5.

Percentage of systems from each morphology that produced (top) at least one report of severe weather and (bottom) at least one nonflooding report of severe weather for both years (2007, dark; 2002, light).

Fig. 5.

Percentage of systems from each morphology that produced (top) at least one report of severe weather and (bottom) at least one nonflooding report of severe weather for both years (2007, dark; 2002, light).

There is good agreement between the 2 yr in the temporal patterns of the number of convective events (Fig. 6). A general increase can be seen in the number of all systems from April through August, while severe-storm-report-producing systems only increase in number through June and then tend to remain constant. These trends suggest that although fewer convective systems occur in the spring, a greater proportion of them produce severe weather compared to those that occur in the summer.

Fig. 6.

Breakdown by month of the number of systems in each of the 2 yr, and the number of severe-storm-report-producing systems.

Fig. 6.

Breakdown by month of the number of systems in each of the 2 yr, and the number of severe-storm-report-producing systems.

To best compare the frequency with which the systems of each morphology produced severe weather, reports were normalized as in G08 to determine the average number of reports produced per system for each morphology (Figs. 7 –10). Tests of statistical significance were applied via the Wilcoxon rank sums test, and a threshold p value of 0.05 was used to determine statistically significant differences in means. Figure 7 shows that BL and PS systems were the leading producers of tornadoes, although the differences were not statistically significant, as p values were greater than 0.5000 in all tests. G08 mentioned that BL systems often occur near baroclinic boundaries, which would likely be regions of enhanced vertical wind shear. One noticeable difference between the datasets is that PS systems did not produce tornadoes in 2007 as frequently as in 2002, but ranked third in average number in 2007 behind BL and CC systems. In general, the systems in the 2002 dataset associated with the most tornadoes produced about twice as many tornadoes as those in the 2007 dataset, even though the average number of tornadoes produced by all of the systems combined was nearly identical (0.39 for 2002 compared to 0.38 for 2007). Otherwise, the numbers agree fairly well between the two datasets. Except for PS systems, it is clear that cellular systems produced more tornadoes per event than other systems, although the difference in the average numbers of tornadoes per system between combined cellular and combined noncellular morphologies was not statistically significant in either year (p = 0.1065 and p = 0.0575 in 2002 and 2007, respectively).

Fig. 7.

Average number of reports of tornadoes per system (left-side axis) and average (E)F-scale rating (right-side axis) for the 2007 and 2002 datasets for each morphology.

Fig. 7.

Average number of reports of tornadoes per system (left-side axis) and average (E)F-scale rating (right-side axis) for the 2007 and 2002 datasets for each morphology.

Fig. 10.

As in Fig. 8, but for (a) flooding, (b) flash flooding, and (c) all flooding reports.

Fig. 10.

As in Fig. 8, but for (a) flooding, (b) flash flooding, and (c) all flooding reports.

Further analysis was performed to determine which morphologies produced the most intense tornadoes. An average of the enhanced Fujita (EF) scale rating (F-scale rating for the 2002 G08 dataset) was computed for the tornadoes produced by the systems in each morphology. Due to the large number of EF0 tornadoes produced by many systems, the average ratings are all below 1.0 (Fig. 7). In general for both years, cellular systems had the highest average tornado intensity rating. The difference was significant in 2002 (p = 0.0007), but not in 2007 (p = 0.1368), when cellular morphologies were combined and tested against the noncellular morphologies combined. Exceptions include PS and BE systems in 2007, which earned the highest average ratings in that year, and LS systems in 2002. However, these PS, BE, and LS systems produced only 15, 3, and 14 tornadoes, respectively, compared to 50 or more for each of the cellular morphologies, and the small sample size may be affecting the results. In fact, as will be discussed in section 3b, the differences in tornado ratings between PS and BE systems and supercellular systems in 2007 were not statistically significant.

The 2007 and 2002 datasets agree well for the average number of reports of hail of all sizes per system for the cellular morphologies (Fig. 8). BL systems were the most productive of all morphologies, although not significantly (p values in all tests comparing BL systems to BE, CC, or TS systems for the average numbers of reports of hail per system for all sizes were greater than 0.6000). There is less agreement between the two datasets for the linear morphologies, particularly for LS, PS, and BE systems, as the average numbers of hail reports per system differed by up to 90% for these morphologies. Again, caution should be used in interpreting these results due to small sample sizes.

Fig. 8.

Average numbers of reports per system for hail for the 2007 (dark) and 2002 (light) datasets: (a) ≥0.75 in. but <1 in., (b) ≥1 in. but <2 in., (c) ≥2 in. in diameter, and (d) in all size ranges for each morphology.

Fig. 8.

Average numbers of reports per system for hail for the 2007 (dark) and 2002 (light) datasets: (a) ≥0.75 in. but <1 in., (b) ≥1 in. but <2 in., (c) ≥2 in. in diameter, and (d) in all size ranges for each morphology.

As in G08, BE systems produced the greatest average number of reports of severe wind (Fig. 9), with TS systems ranking second. The difference in the average numbers of total reports of wind between BE and TS systems was statistically significant (p = 0.0108 and p = 0.0027 for 2007 and 2002, respectively), as well as between TS systems and BL systems (which ranked third in average numbers of total reports of wind) in 2007 (p = 0.0129), but not in 2002 (p = 0.2725). The structures of the BE and TS systems likely explain the high amounts of severe wind reports, with rear-inflow jets transporting momentum downward in both types of systems (e.g., Smull and Houze 1987; Duke and Rogash 1992). Strong downdrafts and resulting large-scale cold outflows, in association with the acceleration produced from mesohighs under the downdrafts produced by these systems, could also explain the high amount of severe wind reports. There is rather good agreement between the 2007 and 2002 events for the other morphologies as well.

Fig. 9.

As in Fig. 8, but for wind gusts of magnitude (a) ≥50 kt but <65 kt, (b) ≥65 kt, and (c) from all wind ranges.

Fig. 9.

As in Fig. 8, but for wind gusts of magnitude (a) ≥50 kt but <65 kt, (b) ≥65 kt, and (c) from all wind ranges.

The average number of flooding reports per system in both years is shown in Fig. 10. The average number of flash flooding (flash flooding typically occurs suddenly on shorter time and smaller distance scales than flooding) reports per system (Fig. 10b) was nearly the same in both years for TS and PS systems, but differed much more for flooding reports (Fig. 10a). Also, a large disparity between 2002 and 2007 existed for the average number of reports of flash flooding for BL systems, a disparity absent for flooding reports. A similar pattern of behavior can be seen for LS systems. Systems that usually have significant stratiform precipitation (NL, TS, LS, PS, and BE) were all among the top four or five morphologies for average numbers of flooding reports per system in both datasets. Overall though, there is a general disagreement between the leading producers of flooding reports. One area of agreement, however, and a finding supported by Parker (2007), is that PS systems produced the greatest average number of reports of flash flooding per system in both datasets, although not significantly (p values were greater than 0.5000 in both tests comparing PS to TS systems). However, LS systems, which produced the most average reports of flooding per system in G08, were associated with far fewer reports in 2007, while BE and NL systems produced far more reports of all types of flooding in 2007 than in G08.

b. Supercellular versus nonsupercellular systems

A key difference between the present study and that of G08 is the inclusion of supercell morphologies in the 2007 dataset. The remainder of this section discusses the differences between 2007 systems that were classified as supercellular and all other systems, emphasizing differences between the supercellular and nonsupercellular cellular systems. Using the methodology outlined earlier, 207 supercell systems (23% of all systems) were classified. Of those 207 supercell systems, the majority were CC, numbering 118 (57%), while IC events totaled 47 (23%), and BL events produced the remaining 42 (20%). These numbers are all greater than the numbers of LS and PS systems present in the 2007 dataset and are comparable to several other morphologies.

A breakdown of the contributions from each morphology to the total number of systems and to the number of severe storm reports can be seen in Fig. 11. Of all of the cellular systems, 37% contained a supercell. Nonsupercellular versions of the IC and CC morphologies were more common than the supercellular versions, whereas BL events were fairly evenly divided between those with and those without supercells. Of all of the supercellular morphologies, CC systems were most common, representing 13% of all systems. When only storm reports are considered, however, a significant change occurs. All of the supercellular morphologies contribute a larger percentage to the number of severe storm reports than they did to the number of all systems. In fact, supercellular CC systems produced the most severe reports (29%) of all morphologies, and the three supercell morphologies together accounted for 51% of all storm reports. Nonsupercellular cellular morphologies only accounted for 6% of severe reports. When combined with the fact that about 91% of all supercellular systems produced severe weather (Fig. 12, Table 2), it is clear that supercellular systems pose an especially large severe weather risk. These results support the speculation in G08 that higher frequencies of severe reports in spring from cellular events despite fewer numbers of cellular events than in summer likely implied the presence of supercells in spring.

Fig. 11.

Percentage contribution of each morphology in the 2007 dataset to the total number of (top) systems and (bottom) severe storm reports. Shading as in Fig. 3 except that additional light gray color distinguishes supercellular systems from nonsupercellular systems. Numbers in parentheses indicate the number of events that occurred for each morphology.

Fig. 11.

Percentage contribution of each morphology in the 2007 dataset to the total number of (top) systems and (bottom) severe storm reports. Shading as in Fig. 3 except that additional light gray color distinguishes supercellular systems from nonsupercellular systems. Numbers in parentheses indicate the number of events that occurred for each morphology.

Fig. 12.

As in Fig. 5, but for 2007 only, and separating cellular events into supercells and nonsupercells.

Fig. 12.

As in Fig. 5, but for 2007 only, and separating cellular events into supercells and nonsupercells.

Table 2.

Percentage of the total amount of storm reports or systems assigned to each morphology.

Percentage of the total amount of storm reports or systems assigned to each morphology.
Percentage of the total amount of storm reports or systems assigned to each morphology.

A breakdown by month and morphology of the number of systems to occur (Fig. 13) shows that surprisingly, with the exception of NS, PS, and TS systems, all systems had a peak of occurrence late in the period, either in July or August. Most systems were rarest in April, likely because cooler conditions common in the early spring do not typically favor convection nearly as much as do the warmer conditions common during the summer months. Of note in Fig. 13, far more supercellular CC systems occurred in August than in any other month, with a similar enhanced maximum in supercellular BL events. The supercellular CC systems also produced much severe weather in August with an average of about 22 reports per system, the second greatest of any morphology that month (not shown). These results differ greatly from G08, which found relatively small numbers of CC events and severe reports during August. This difference in results suggests August 2007 may have been unusually favorable for supercell thunderstorms and severe weather. Indeed, the jet stream was frequently located over the northern part of the domain during this period with stronger than normal flow aloft likely enhancing the shear needed for supercell development. An archive of severe storm events dating back to 2000 maintained by the Storm Prediction Center (information online at http://www.spc.noaa.gov) supports this assertion by showing more August days in 2007 with severe weather within the 10-state region, (21 days) than in any other year. August 2002 in contrast had slightly fewer days than normal with severe weather in this region.

Fig. 13.

Breakdown by morphology and month of the number of systems occurring.

Fig. 13.

Breakdown by morphology and month of the number of systems occurring.

The average number of tornadoes produced by supercellular systems was much greater than that from any other morphology (Fig. 14a, Table 3). The difference was statistically significant (Table 4). Unless otherwise noted, all of the following results were statistically significant, often with p values less than 0.0001. NS and PS systems were the next largest producers of tornadoes per event, but still produced about 0.4 tornadoes less per system (roughly 40%) than even the least productive supercellular system (IC). Supercellular BL systems produced the greatest average number of tornadoes per system at 1.55. The largest average number of tornadoes produced per system for the nonsupercellular cellular morphologies was 0.16 by nonsupercellular CC systems, a 90% reduction from the supercellular BL morphology. The average tornado intensity rating for supercellular CC systems was 0.63, the highest among the three supercellular morphologies. However, PS systems produced the largest average rating for all morphologies, 0.80, with BE systems producing the second highest average rating, 0.67. This result is surprising since supercellular CC systems (and all supercellular morphologies) were associated with larger average numbers of tornadoes, and the supercellular CC morphology also was responsible for the strongest tornado (the Greensburg, Kansas, storm, which was rated as an EF5), and six EF3s (the largest number of EF3s produced by any morphology). These results imply that supercellular CC systems also produce many F0 tornadoes and may also reflect the small sample size of PS and BE systems. It is also possible that some weak tornadoes that occurred in the PS and BE systems were mistakenly reported as wind damage due to the frequent occurrence of severe wind occurring throughout them. Thus these systems may have produced some EF0 and EF1 tornadoes that did not appear in Storm Data, reducing the average intensity value computed in this study. However, the p value for the Wilcoxon test between supercell CC systems and PS systems was 0.3648, and thus the difference was not significant. The differences were also not significant between PS systems and the other supercell systems, nor between BE systems and the supercell systems (p values ranged from 0.0673 to 0.6132). It should also be noted that of the 207 supercellular systems identified, 28% produced at least one tornado (not shown). This result is very similar to the 26% rate found in a study of over 5000 MDA detections from 54 different radar sites nationwide during the period 1992–99 (Trapp et al. 2005a), and supports recent findings that tornadoes are relatively rare, even when storms contain persistent rotation.

Fig. 14.

Average number of reports per system for (a) tornadoes, (b) hail, (c) wind, and (d) flooding by morphology.

Fig. 14.

Average number of reports per system for (a) tornadoes, (b) hail, (c) wind, and (d) flooding by morphology.

Table 3.

As in Table 2, but for average numbers of reports per system for the various types of severe weather and for each type of morphology.

As in Table 2, but for average numbers of reports per system for the various types of severe weather and for each type of morphology.
As in Table 2, but for average numbers of reports per system for the various types of severe weather and for each type of morphology.
Table 4.

The p values for various tests of statistical significance using the Wilcoxon test. Values in boldface indicate those deemed not significant using a threshold p value of 0.05.

The p values for various tests of statistical significance using the Wilcoxon test. Values in boldface indicate those deemed not significant using a threshold p value of 0.05.
The p values for various tests of statistical significance using the Wilcoxon test. Values in boldface indicate those deemed not significant using a threshold p value of 0.05.

Supercellular systems produced the most reports, on average, of all three size ranges of hail, and especially for hail greater than 1 in. in diameter (Fig. 14b). The most productive morphology was supercellular BL systems, which produced the greatest average number of all three size ranges of hail per system and, thus, for total hail reports. The average number of reports of hail per system for the supercellular systems generally was at least double the rate for linear systems and roughly an order of magnitude larger than the rate for nonsupercellular cellular systems (a statistically significant difference; see Table 4) and NL events.

BE systems produced the greatest average number of severe wind reports per system (Fig. 14c), which is consistent with G08 and not surprising (e.g., Klimowski et al. 2003). However, supercellular BL and supercellular CC systems produced the second and third greatest averages, producing more severe wind reports per system than TS systems, events found in G08 to be the second largest producer. The difference between BE systems and supercellular BL systems was not significant (p = 0.0764 for total wind reports), however. Likewise, the difference between supercellular BL systems and TS systems was not significant (p = 0.1742), although it was significant between BE and TS systems (p = 0.0108). Perhaps the higher averages for supercellular BL and supercellular CC systems is related to wind gusts produced by rear-flank downdrafts associated with the supercells in those systems (e.g., Markowski 2002; Finley and Lee 2008). Wind reports were roughly an order of magnitude less common with NL systems, and even less likely with nonsupercellular cellular events (Fig. 14c, Table 3).

As with tornadoes, hail, and wind reports, flood reports were more common per event in supercellular systems than in nonsupercellular cellular events (Fig. 14d). This result is perhaps more surprising than that for tornadoes, hail, and wind, and suggests that the supercells may possess greater rain rates, or cover larger areas than the nonsupercellular cellular events. The analysis of radar data from 2007 did suggest that many supercells were larger in size than the nonsupercells, and often persisted longer, both factors that could lead to larger accumulations of rainfall and possible flooding. Nonetheless, all cellular morphologies were less likely to produce flooding than nonlinear or linear systems, and the morphologies most likely to produce flooding remained BE, PS, TS, and NL, (although the difference between NL and the next lowest average, that from supercellular CC systems, was not significant, with p = 0.7941) as shown earlier in Fig. 10.

The sometimes large differences between supercellular systems and nonsupercellular systems are summarized in Table 2. Supercellular systems produced severe weather more frequently than other types of systems. Despite consisting of only 23% of all systems and 34% of all severe systems, supercellular systems produced over half of all severe weather reports, more than two-thirds of all tornadoes and hail reports, and produced a substantial number of wind reports compared to the nonsupercellular systems.

The total number of reports produced by nonsupercellular cellular systems was tiny compared to that of the supercellular systems. The 33 tornadoes produced by the three nonsupercellular cellular morphologies were only 14% of the 234 tornadoes produced by the supercellular systems. The nonsupercellular cellular systems produced only about 10% as many hail and wind reports as the supercellular cellular systems and produced only six reports of hail greater than or equal to 2 in. in diameter and eight reports of wind greater than or equal to 65 kt (not shown). Those numbers compare to 164 and 153 reports of hail at least 2 in. in diameter and wind gusts of at least 65 kt, respectively, for supercells. Results differed less for flooding but nonsupercellular cellular systems still produced less than half of the reports per case associated with supercellular events.

In terms of average numbers of severe reports per event, supercellular systems far exceeded nonsupercellular cellular events, as well as all other systems (Table 3), for every type of report except flooding. As was alluded to earlier, several configurations of comparisons were tested for statistical significance, and those results are presented in Table 4. It is clear that the differences in the average number of reports per system between the supercellular systems and the nonsupercellular systems for nearly all types of severe weather were statistically significant, with many p values being exceptionally small. The higher average for flooding in nonsupercellular systems likely reflects the much larger sizes of the top flood-producing systems, BE and PS. The average tornado rating (on the EF scale) of all supercellular systems was 0.56, twice that of all other systems, and far larger than the 0.07 average for nonsupercellular cellular systems. Thus, it is clear that supercellular systems in 2007 were more likely to produce severe weather than their nonsupercellular counterparts, supporting our hypothesis.

4. Conclusions and discussion

The present study expanded the work done by G08, in which all convective events that occurred within a 10-state domain that included the Midwest and Great Plains between April and August 2007 were classified according to their dominant morphology, with supercells added as a morphology. To allow the classification of supercells, a different dataset had to be used from that used in G08, and this also allowed a comparison between storms occurring in 2007 and those occurring in 2002. Systems had to meet specific radar criteria to be classified into the nine morphologies defined by G08. All severe reports, which were obtained from NCDC’s Storm Events Database, were attributed to the dominant morphology that characterized each system during its lifetime. Then, using MDA output via level III storm attribute data and the mesocyclone product from NCDC, supercells were separated from their nonsupercellular counterparts based on the existence of a persistent mesocyclone in a recognizable cellular element from one of the cellular morphologies (IC, CC, and BL).

It was found that, in general, similar trends were present in the 2007 data compared to those in 2002 presented in G08, with NL, IC, and CC morphologies most common. NL events were noticeably less common in 2007, however, than in 2002. Cellular systems produced the most severe reports in both years, especially CC events. LS systems in both years contributed the fewest number to the total severe reports and were relatively rare in occurrence, along with PS events. Monthly trends were also generally similar in both years, although storms and severe weather reports were more frequent in August 2007 than in August 2002. The biggest differences in the average number of reports for tornadoes, hail, and wind between the two datasets occurred with BL, PS, LS, and BE systems, and for flooding with BL, NL, and BE events. Some of the differences are likely related to small sample sizes for PS, LS, and BE morphologies in one or both of the years.

The classification of cellular systems as supercells or nonsupercells revealed dramatic differences in the frequency of storm reports from each morphology. Supercellular systems produced much more severe weather of all types than did nonsupercellular cellular morphologies, and most differences were statistically significant, as determined by a Wilcoxon rank sums test. In addition, for tornadoes and hail, the three supercellular morphologies all were associated with more storm reports per system than any other morphology. For severe wind, BE systems were the most prolific producers of reports, but CC and BL supercells were next most prolific. For flooding, however, PS and TS systems were among the morphologies producing the most reports per system in both years, with BL also ranking high in 2002 and BE and NL ranking high in 2007. All of these morphologies tend to include large areas of rainfall, which should enhance the flood risk, as discussed in G08. The division of the cellular morphologies into those with rotation and those without also revealed that the nonsupercellular cellular morphologies generally produced the fewest severe storm reports per system and had the highest rates of nonsevere systems, less than nonlinear events and the five linear morphologies.

Future work should include expanding the areal coverage of the study to that of the entire continental United States to develop a climatology of severe weather and convective events for all portions of the country, especially because differences in reporting strategies among National Weather Service offices and radar coverage would likely influence the results. In addition, future studies should expand the time period to include all portions of the year, add additional years, add additional morphologies [such as TL/AS from Schumacher and Johnson (2005)], and allow systems from all morphologies (not just cellular ones) to be eligible to contain supercells. However, if the last task is undertaken, substantial thought must be given to the value of classifying an entire system as supercellular based on the presence of just one or two supercells, particularly if the majority of severe storm reports are associated with the individual supercells. It might be advantageous in such an analysis to focus on the morphologies of individual elements and not on entire systems, a philosophical change from G08 and the current study.

Acknowledgments

Thanks are given to Nathan Snook and Elise Johnson for their advice and input on the methodology used in G08. Thanks also to Daryl Herzmann for supplying storm attribute data and offering advice on data sources, and to Robert Lee for his advice on the definition of a supercell. Thanks to Jon Hobbs for offering advice on proper tests of statistical significance to apply to the data. The paper was improved by the constructive comments of three anonymous reviewers. This research was partially supported by National Science Foundation Grants ATM-0537043 and ATM-0848200. Much of the research was performed by the first author as part of a required senior meteorology undergraduate thesis at Iowa State University guided by Tsing-Chang Chen.

REFERENCES

REFERENCES
Baldwin
,
M. E.
,
J. S.
Kain
, and
S.
Lakshmivarahan
,
2005
:
Development of an automated classification procedure for rainfall systems.
Mon. Wea. Rev.
,
133
,
844
862
.
Bluestein
,
H. B.
, and
M. H.
Jain
,
1985
:
Formation of mesoscale lines of precipitation: Severe squall lines in Oklahoma during the spring.
J. Atmos. Sci.
,
42
,
1711
1732
.
Burgess
,
D. W.
, and
L. R.
Lemon
,
1990
:
Severe thunderstorm detection by radar.
Radar in Meteorology, D. Atlas, Ed., Amer. Meteor. Soc., 619–647
.
Doswell
III,
C. A.
, and
D. W.
Burgess
,
1988
:
On some issues of United States tornado climatology.
Mon. Wea. Rev.
,
116
,
495
501
.
Doswell
III,
C. A.
, and
D. W.
Burgess
,
1993
:
Tornadoes and tornadic storms: A review of conceptual models.
The Tornado: Its Structure, Dynamics, Hazards, and Prediction, Geophys. Monogr., No. 79, Amer. Geophys. Union, 161–172
.
Duke
,
J. W.
, and
J. A.
Rogash
,
1992
:
Multiscale review of the development and early evolution of the 9 April 1991 derecho.
Wea. Forecasting
,
7
,
623
635
.
Finley
,
C. A.
, and
B. D.
Lee
,
2008
:
Mobile mesonet observations of an intense RFD and multiple gust fronts in the May 23 Quinter, Kansas tornadic supercell during TWISTEX 2008.
Preprints, 24th Conf. on Severe Local Storms, Savannah, GA, Amer. Meteor. Soc., P3.18. [Available online at http://ams.confex.com/ams/pdfpapers/142133.pdf]
.
Forbes
,
G. S.
,
1981
:
On the reliability of hook echoes as tornado indicators.
Mon. Wea. Rev.
,
109
,
1457
1466
.
Gallus
Jr.,
W. A.
,
N. A.
Snook
, and
E. V.
Johnson
,
2008
:
Spring and summer severe weather reports over the Midwest as a function of convective mode: A preliminary study.
Wea. Forecasting
,
23
,
101
113
.
Glickman
,
T.
,
Ed.
2000
:
Glossary of Meteorology.
2nd ed. Amer. Meteor. Soc., 855 pp
.
Grams
,
J. S.
,
W. A.
Gallus
Jr.
,
L. S.
Wharton
,
S. E.
Koch
,
A.
Loughe
, and
E. E.
Ebert
,
2006
:
The use of a modified Ebert–McBride technique to evaluate mesoscale model QPF as a function of convective system morphology during IHOP 2002.
Wea. Forecasting
,
21
,
288
306
.
Jirak
,
I. L.
,
W. R.
Cotton
, and
R. L.
McAnelly
,
2003
:
Satellite and radar survey of mesoscale convective system development.
Mon. Wea. Rev.
,
131
,
2428
2449
.
Johns
,
R. H.
, and
C. A.
Doswell
III
,
1992
:
Severe local storms forecasting.
Wea. Forecasting
,
7
,
588
612
.
Klimowski
,
B. A.
,
M. J.
Bunkers
,
M. R.
Hjelmfelt
, and
J. N.
Covert
,
2003
:
Severe convective windstorms over the northern high plains of the United States.
Wea. Forecasting
,
18
,
502
519
.
Markowski
,
P. M.
,
2002
:
Hook echoes and rear-flank downdrafts: A review.
Mon. Wea. Rev.
,
130
,
852
876
.
Miller
,
D. J.
, and
R. H.
Johns
,
2000
:
A detailed look at extreme wind damage in derecho events.
Preprints, 20th Conf. on Severe Local Storms, Orlando, FL, Amer. Meteor. Soc., 4.3. [Available online at http://ams.confex.com/ams/pdfpapers/15800.pdf]
.
Mitchell
,
E. D.
,
S. V.
Vasiloff
,
G. J.
Stumpf
,
A.
Witt
,
M. D.
Eilts
,
J. T.
Johnson
, and
K. W.
Thomas
,
1998
:
The National Severe Storms Laboratory tornado detection algorithm.
Wea. Forecasting
,
13
,
352
366
.
Moller
,
A. R.
,
C. A.
Doswell
III
,
M. P.
Foster
, and
G. R.
Woodall
,
1994
:
The operational recognition of supercell thunderstorm environments and storm structures.
Wea. Forecasting
,
9
,
327
347
.
Parker
,
M. D.
,
2007
:
Simulated convective lines with parallel stratiform precipitation. Part I: An archetype for convection in along-line shear.
J. Atmos. Sci.
,
64
,
267
288
.
Parker
,
M. D.
, and
R. H.
Johnson
,
2000
:
Organizational modes of midlatitude mesoscale convective systems.
Mon. Wea. Rev.
,
128
,
3413
3436
.
Schumacher
,
R. S.
, and
R. H.
Johnson
,
2005
:
Organization and environmental properties of extreme-rain-producing mesoscale convective systems.
Mon. Wea. Rev.
,
133
,
961
976
.
Smull
,
B. F.
, and
R. A.
Houze
,
1987
:
Rear inflow in squall lines with trailing stratiform precipitation.
Mon. Wea. Rev.
,
115
,
2869
2889
.
Storm
,
B. A.
,
M. D.
Parker
, and
D. P.
Jorgensen
,
2007
:
A convective line with leading stratiform precipitation from BAMEX.
Mon. Wea. Rev.
,
135
,
1769
1785
.
Stumpf
,
G. J.
,
A.
Witt
,
E. D.
Mitchell
,
P. L.
Spencer
,
J. T.
Johnson
,
M. D.
Eilts
,
K. W.
Thomas
, and
D. W.
Burgess
,
1998
:
The National Severe Storms Laboratory Mesocyclone Detection Algorithm for the WSR-88D.
Wea. Forecasting
,
13
,
304
326
.
Trapp
,
R. J.
,
G. J.
Stumpf
, and
K. L.
Manross
,
2005a
:
A reassessment of the percentage of tornadic mesocyclones.
Wea. Forecasting
,
20
,
680
687
.
Trapp
,
R. J.
,
S. A.
Tessendorf
,
E.
Savageau Godfrey
, and
H. E.
Brooks
,
2005b
:
Tornadoes from squall lines and bow echoes. Part I: Climatological distribution.
Wea. Forecasting
,
20
,
23
34
.
Trapp
,
R. J.
,
D. M.
Wheatley
,
N. T.
Atkins
,
R. W.
Przybylinski
, and
R.
Wolf
,
2006
:
Buyer beware: Some words of caution on the use of severe wind reports in postevent assessment and research.
Wea. Forecasting
,
21
,
408
415
.
Verbout
,
S. M.
,
H. E.
Brooks
,
L. M.
Leslie
, and
D. M.
Schultz
,
2006
:
Evolution of the U.S. tornado database: 1954–2003.
Wea. Forecasting
,
21
,
86
93
.
Weiss
,
S. J.
,
J. A.
Hart
, and
P. R.
Janish
,
2002
:
An examination of severe thunderstorm wind report climatology: 1970–1999.
Preprints, 21st Conf. Severe Local Storms, San Antonio, TX, Amer. Meteor. Soc., 11B.2. [Available online at http://ams.confex.com/ams/pdfpapers/47494.pdf]
.

Footnotes

Corresponding author address: Jeffrey D. Duda, 3134 Agronomy, Iowa State University, Ames, IA 50011. Email: jdduda@iastate.edu