An Analysis of Anomalous Winter and Spring Tornado Frequency by Phase of the El Nino/Southern Oscillation, the Global Wind Oscillation, and the Madden-Julian Oscillation.
More tornadoes occur in the United States (US) per year than in any other country , and these tornadoes are capable of producing incredible economic and human loss [2-5]. Despite their relatively common occurrence, there is notable intra- and interannual variability in the number of US tornadoes [6-10]. Due to the danger and variability of tornado activity, there is a need to improve seasonal outlook capabilities. One approach is to identify relationships between seasonal tornado count and climate oscillations, such as the El Nino/Southern Oscillation (ENSO) (e.g., [11-19]).
Numerous studies over the past couple of decades have analyzed the relationship between tornadoes and ENSO. An early study by Monfredo  reported that strong and violent tornadoes were more common during the La Nina (LN) phase of ENSO and less common during the El Nino (EN) phase. Cook and Schaefer  later reported that winter tornado outbreaks were stronger and more frequent during the neutral phase (N) of ENSO, followed by the LN then EN phases. The most recent studies report that tornado activity is heightened during the LN phase [12-14, 17-19]. Recent studies also illustrate that the seasons with the most tornadoes tend to be classified as LN. Lee et al. , for example, analyzed the number of tornadoes rated 3 or higher on the Fujita or Enhanced Fujita damage scales (hereafter referred to as E(F)) and reported that the five most extreme tornado outbreak years were characterized as persistent LN events or LN events that were transitioning to a different phase. Allen et al.  similarly reported that most of the seasons with a tornado count >100% of climatology occurred during the LN phase, whereas most that were <75% of climatology occurred during the EN phase.
The relatively consistent reporting of heightened tornado activity with the LN phase of ENSO elicits the possibility of predicting the likelihood of below- or above-normal tornado activity based on ENSO conditions. Elsner et al.  developed a statistical model to evaluate variability in tornado activity in which ENSO was the most important predictor. Their model also indicated that tornado activity was heightened in the Midwest and Southeast regions of the US during LN and in the Great Plains region during EN. Allen et al.  and Lepore et al.  developed extended logistic regression models using ENSO to predict the likelihood of below-normal, normal, or above-normal tornado activity in spring. Their models similarly showed that tornado activity is increased during the LN phase, particularly in the south central US.
Anomalous seasons occur. Some LN seasons have fewer than expected tornadoes and some EN seasons have more than expected. Lee et al. , for example, noted that an anomalously large number of tornadoes occurred in the 2015-2016 EN winter. Elsner et al.  also noted the occurrence of anomalous seasons and attributed them to additional unknown factors. While previous studies have noted these anomalous seasons, none have provided detailed descriptions and analyses of them. More research, including work focusing on the anomalous seasons, is needed to improve our understanding of the tornado-ENSO link and to refine statistical models.
A case study approach was taken here to describe and analyze the two EN winters and springs with the most tornadoes, and the two LN winters and springs with the fewest tornadoes. The states of the Madden-Julian Oscillation (MJO) and Global Wind Oscillation (GWO) are of particular interest here. Both of these oscillations vary over subseasonal timescales (i.e., shorter than the timescale over which ENSO varies) and have been linked to variations in tornado activity [20-23]. Seasonal and daily composites are also generated to document the synoptic patterns that accompanied the anomalous seasons.
The specific objectives of this study are to
(1) identify and describe the two EN winters and springs with the most tornadoes and the two LN winters and springs with the fewest tornadoes;
(2) analyze GWO and MJO activity during these anomalous seasons;
(3) document the synoptic patterns that accompany these anomalous seasons.
2. Data and Methods
Tornado data were taken from the Storm Prediction Center's Severe Weather Database (specifically, the 1950-2016_actual_tornadoes.csv file) . This dataset was a subset to the 18,251 E(F)1+ tornadoes that occurred over the period December 1978-November 2016 in the contiguous US. The time series of E(F)1+ tornadoes is less influenced by detection and reporting changes over time than is the E(F)0+ series, which includes a notable upward trend [25, 26]. The beginning of the period of record is consistent with recent studies [12, 13] and corresponds to the first year of the North American Regional Reanalysis (NARR) product , which is used to generate seasonal and daily atmospheric composites. Seasonal tornado counts were generated for winter (December of previous year, January, and February (DJF)) and spring (March, April, and May (MAM)). DJF and MAM are the focus of this study because previous studies illustrate that tornado counts and ENSO are related in these seasons [12-14, 17-19].
Multiple indices represent ENSO conditions. Some represent mostly the oceanic component of ENSO (e.g., the Nino 1.2, 3, 3.4, and 4 regions and the Oceanic Nino Index (ONI)), some represent mostly the atmospheric component (e.g., the Southern Oscillation Index (SOI)), and some combine both (e.g., the Multivariate ENSO Index (MEI) and Bivariate ENSO Index (BEST)). The classification of seasons as LN or EN will likely change in some cases, depending on the index. In this study, ONI data were obtained from the Climate Prediction Center  to represent ENSO conditions, which is consistent with recent efforts to predict the likelihood of an active or inactive season based on the state of ENSO [12, 13]. ONI is also one of the commonly used indices by the Climate Prediction Center to diagnose ENSO conditions and to place current events into historical perspective. It is given as anomalies of 3-month running means of sea surface temperature in the Nino 3.4 region (5[degrees]N-5[degrees]S, 120[degrees]W-170[degrees]W). Series of ONI for DJF and MAM were extracted and merged with the seasonal tornado counts. Once merged, seasons were classified as LN if ONI [less than or equal to]-0.5 and EN if ONI >0.5. Seasons with -0.5 < ONI < 0.5 were classified as neutral (N). The two most active EN DJFs and MAMs (i.e., with the most tornadoes) and two least active LN DJFs and MAMs (i.e., with the fewest tornadoes) were extracted, yielding eight seasons as the focus of this study.
Climate oscillations other than ENSO have been linked to tornado activity. Those that vary over subseasonal timescales may provide insight into the intraseasonal distribution of tornadoes in the anomalous seasons. Two such oscillations are the MJO and GWO. The MJO is a tropical wave of enhanced convection that propagates the globe from west to east at approximately 5m x [s.sup.-1] . It is linked to variability in tornado activity in the midlatitudes through changes in global relative atmospheric angular momentum (AAM), which arise from tropical convective forcing and teleconnected Rossby wave propagation in the midlatitudes [20,23, 30-32]. Other factors, such as friction and mountain torque, affect AAM in addition to tropical convection [30-32]. The GWO combines these factors and, therefore, provides a comprehensive representation of AAM and midlatitude circulation patterns [31, 32]. Daily GWO data were obtained from the Earth System Research Laboratory's GWO dataset . The attributes of the GWO used in this study are AAM anomalies, AAM time tendency, and GWO phase. The relationship between AAM anomalies and AAM time tendency determines the phase of GWO, which run from 1 to 8. Phases 2 and 3 represent anomalously low AAM, whereas phases 6 and 7 represent anomalously high AAM. Phases 1, 4, 5, and 8 represent transition states. Daily MJO data were obtained from the Australian Bureau of Meteorology's Real-Time Multivariate (RMM) MJO dataset [34, 35]. The attributes used here are the first two principal components of empirical orthogonal functions that consist of a meridional average of the 200 mb zonal wind, the 850 mb zonal wind, and outgoing longwave radiation between 15[degrees]N and 15[degrees]S (referred to as RMM1 and RMM2) and the MJO phase. Similar to GWO, the MJO index has eight phases, but they represent the location of enhanced tropical convection. Phase 1 of the MJO indicates that the enhanced convection is located near eastern Africa and the subsequent phases represent its eastward progression.
AAM anomalies, AAM time tendency, RMM1, and RMM2 were used to construct GWO and MJO phase space diagrams for each of the anomalous seasons. Daily tornado counts were generated (where a day is defined as 0000-2359 UTC) and plotted in the phase space diagrams along with the progression of GWO and MJO. Daily tornado counts were also assessed across the phases of MJO and GWO, and Kruskal-Wallis tests were used to determine if the mean rank of the tornado counts significantly varied across the phases of the oscillations. This nonparametric test was chosen because of the skewed nature of daily tornado counts.
The synoptic patterns associated with the anomalous seasons were characterized using gridded composites from NARR . The following variables were chosen to be comparable to composites in previous research (e.g., [12, 20, 21]): upper-level flow was characterized using 300 mb geopotential heights; low-level flow was characterized using 850 mb geopotential heights; low-level moisture was characterized using 850 mb specific humidity; and atmospheric stability was characterized using surface-based convective available potential energy (CAPE). Surface-based CAPE was used because Gensini et al.  illustrate that surface-based parcels are more accurate than 100 mb mixed layer parcels, largely due to errors in low-level moisture fields. The seasonal mean composites are averages of the variables for the dates under consideration (e.g., mean 300 mb geopotential height in DJF 2016 is generated by averaging the 300 mb geopotential height for the days in DJF 2016). The seasonal anomaly composites represent the difference between the mean composites and the 1981-2010 climatology. Daily composites are averages of the 3 hr NARR data. All composites were generated with NOAA's Earth System Research Laboratory's online plotting tool [37, 38]. Once generated, the Network Common Data Form (NetCDF) of the composites was imported into ArcMAP  for plotting and display.
3. Results and Discussion
3.1. Identification and Description of the Anomalous Seasons. The relationships between the number of E(F)1+ tornadoes and ONI in DJF and MAM are depicted in Figure 1. The negative relationship reported by others, whereby tornado frequency tends to be greater during the LN phase and lesser during the EN phase [12-14,17-19], is visually apparent, but the correlation is not statistically significant in this subset. The central tendencies in Table 1 also illustrate the negative association by showing that the mean and median tornado counts of the LN seasons exceed those of the EN seasons.
The two most inactive LN DJF and MAM seasons are marked with blue diamonds in Figure 1, and the two most active ENs are marked with red diamonds. The two most active EN DJFs, which occurred in 2016 and 1983, are especially noticeable. There were 133 E(F)1+ tornadoes in both seasons. These counts are 183% greater than the mean number of tornadoes in EN DJFs and 359% greater than the median. The two most active EN MAM seasons were 1982 and 1983, when 340 and 251 E(F)1+ tornadoes occurred, respectively. MAM 1982 had approximately 79% more E(F)1+ tornadoes than the mean EN MAMs (76% greater than the median). MAM 1983 exceeded the mean and median counts of EN MAM seasons by approximately 30%.
The two DJF seasons with the fewest E(F)1+ tornadoes were 1985 and 1986, with 9 and 21 tornadoes, respectively. DJF 1985 had approximately 80% fewer tornadoes than the mean and median LN DJF, and DJF 1986 had 64% and 54% fewer than the mean and median counts, respectively. The least active LN MAM seasons were 1985 when 179 E(F)1+ tornadoes occurred and 2000 when 178 E(F)1+ occurred. These seasons had approximately 43% fewer tornadoes than the mean LN MAM tornado count and approximately 24% fewer than the median count.
As with most seasons, the tornadoes in the anomalous seasons were not uniformly distributed (Figure 2). In DJF 2016, for example, 50% of the tornadoes occurred on 4 days, each with 10+ tornadoes. There were 6 days in DJF 1983 with 10+ tornadoes that account for 68% of the tornadoes in that season. The tornadoes in the anomalous MAM seasons were spread over a larger number of days, but there were still clusters of activity. Twelve days in MAM 1982 and eight in MAM 1983 had 10+ tornadoes. The tornadoes on these days account for 60% and 52% of the seasonal count, respectively.
3.2. Role of GWO and MJO in the Anomalous Seasons. Previous studies illustrate that tornado activity is heightened during certain phases of the GWO and MJO. Gensini and Marinaro  reported that daily tornado anomalies in spring (March-June) are greatest on GWO phase 1 and 2 days when AAM is negative. Moore  similarly reported a tendency for tornado frequency to be greater in MAM seasons when GWO phase 2, 3, and 4 days are more common. Moore  also showed this to be true in DJF. Barrett and Gensini  reported that tornado days in April are most common on phase 6 and 8 days of the MJO and less common on phase 3, 4, and 7 days. They also reported that tornado days are most common with phases 5 and 8 and less common with phases 2 and 3 in May. Thompson and Roundy  reported that violent tornado outbreaks in MAM are most common on MJO phase 2 days and least common on phase 8 days. GWO and MJO vary on subseasonal timescales. They are, therefore, capable of modulating tornado activity within a given season and may provide insight into some of the subseasonal periods of suppressed and heightened tornado activity during these anomalous seasons (as seen in Figure 2).
AAM anomalies were positive throughout both of the anomalously active EN DJFs (Figures 3(a) and 3(c)). The mean AAM anomaly was 2.6kg x [m.sup.2] x [s.sup.-2] in DJF 1983 and 1.9 kg x [m.sup.2] x [s.sup.-2] in DJF 2016 (Table 1). AAM anomalies were also positive throughout the active EN MAM 1983 season, when the mean was 1.8kg x [m.sup.2] x [s.sup.-2] (Figure 3(g), Table 1). Tornadoes in these seasons, therefore, occurred on GWO phase 5-8 days when AAM was anomalously high and the amplitude was most often >1. Positive anomalies were expected in these seasons because AAM tends to be heightened during the EN phase of ENSO , but it is unexpected, based on previous studies linking enhanced tornado activity to anomalously low AAM [21, 22], that all of the days with tornadoes in DJF 1983, DJF 2016, and MAM 1983 had anomalously high AAM. Despite the positive AAM anomalies throughout these seasons, the tendency of AAM was volatile, and tornadoes occurred on days when AAM tendency was increasing and decreasing.
In MAM 1982, there were periods when AAM was anomalously low, which resulted in a lower seasonal mean of 0.1 kg x [m.sup.2] x [s.sup.-2] (Figure 3(e), Table 1). In this season, numerous tornadoes occurred on GWO phase 1-4 in addition to 5-8 days. However, there is not a significant difference in daily tornado counts across the phases of GWO in this season or any of the others (Table 2).
AAM is often relatively low during the LN phase of ENSO . It is, therefore, not surprising that AAM was anomalously low throughout most of the inactive LN seasons (Figure 4). The mean AAM anomaly was -1.5-kg x [m.sup.2] x [s.sup.-2] in MAM 2000 and -0.9 kg x [m.sup.2] x [s.sup.-2] in DJF 1985 (Table 1). AAM fluctuated between negative and positive anomalies in MAM 1985 and DJF 1986, which led to higher mean values of 0.1 kg x [m.sup.2] x [s.sup.-2] in each season (Figures 4(c) and 4(e); Table 2). Similar to the active EN seasons, there is not a significant difference in daily tornado counts across the GWO phases in any of the inactive LN seasons. Also, similar to the tornadoes in the active EN seasons, those in the inactive LN seasons occurred during periods of increasing and decreasing AAM (Figure 4).
MJO varied more than GWO throughout the anomalous seasons (Figures 3 and 4). The progression from phase 1 through 8 is apparent, with multiple oscillations in most seasons. Tornadoes concentrate on certain MJO phase days more so than with GWO phases, which led to significant differences in the mean number of tornadoes per day across the phases. In the two active EN DJFs, for example, the mean and mean ranks of the tornado counts were greatest with phases 1 and 6 of the MJO (Table 2). A Kruskal-Wallis test and subsequent post hoc comparisons indicate that the mean rank of phase 6 is significantly greater than the mean ranks of phases 4, 7, and 8; remaining comparisons yielded insignificant differences (see the subscript below Table 2). The percentage of days with tornadoes was also greatest with phase 6--tornadoes occurred on 10 of the 19 (53%) phase 6 days. In the two EN MAMs, the mean and mean ranks ofthe tornado counts are greatest with phase 4 (Table 2). The statistical tests indicated that the mean rank of phase 4 is significantly greater than that of phase 1 (see the subscript below Table 2). The percentage of days with tornadoes was also greatest with phase 4 (23 of 31 (74%) phase 4 days had tornadoes). The percentage of days with tornadoes was also high with phases 7 and 8 (69% and 68%, resp.). There were not any significant differences in daily tornado counts across the phases of MJO in the inactive LN seasons (Table 2).
3.3. Atmospheric Composites of the Anomalous Seasons. The seasonal composites of 300 mb geopotential height, 850 mb geopotential height and specific humidity, and CAPE are shown in Figures 5-8. Anomalously high geopotential heights at 300 mb were present across most of the contiguous US in DJF 2016 and across the north central and northeast US in DJF 1983 (Figures 5(a) and 5(b)). The presence ofhigher than normal heights is more similar to the LN composite reported by Allen et al.  rather than their EN composite. Upper-level ridges and troughs were present over the western and central US, respectively, in both of the active DJFs. Anomalously high heights were also present during the MAM 1982 seasons, but only over the southern and eastern US (Figure 6(a)). Low height anomalies were present over the contiguous US during MAM 1983 (Figure 6(b)), which is consistent with the EN composite reported by Allen et al. . The patterns seen in the composites of low-level moisture are inconsistent across the active seasons. For example, anomalously high specific humidity was present in the southeastern US during MAM 1982, but the humidity in this region was anomalously low during MAM 1983 (Figures 6(c) and 6(d)). The humidity was also anomalously low in southeast Texas and Louisiana during both DJF seasons, but was higher to the north and east where most of the tornadoes occurred (Figures 5(c) and 5(d)). CAPE was near normal or anomalously high over most of the eastern US during the DJF 1983, DJF 2016, and MAM 1982 active EN seasons (Figures 5(e), 5(f), and 6(e)). Similar to some of the patterns of 300 mb geopotential height, these patterns of elevated CAPE are more similar to the CAPE composites shown by Allen et al.  in association with the LN phase rather than the EN phase.
The patterns of 300 mb geopotential height varied between the two LN DJFs (Figures 7(a) and 7(b)). A dipole pattern was present in DJF 1985, with anomalously low heights spanning the northern Great Plains southwestward to the Southwest and high heights across the Pacific Northwest and Southeast US. The US was split in DJF 1986, with anomalously high heights across the western region and low heights across the eastern. The low-level humidity anomaly patterns also varied between the DJFs (Figures 7(c) and 7(d)). The upper-level height and low-level humidity patterns were more similar in the two inactive LN MAMs--near normal or anomalously high heights most of the US and anomalously humid conditions across most of its eastern half (Figures 8(a)-8(d)). Anomalously low CAPE was present across the Southeast US in DJF 1985 when the fewest tornadoes occurred (Figure 7(e)). Anomalously high CAPE was present across the portions of the US with the most tornadoes in the remaining seasons (Figures 7(f), 8(e), and 8(f)).
Some of the seasonal composites illustrate reasonable patterns. Anomalously high CAPE is shown in both of the active EN DJFs, for example, in Figures 5(e) and 5(f). Most of the eastern US was also anomalously humid in these seasons (Figures 5(c) and 5(d)). Other seasonal composites are unexpected and do not capture the environment on the days when many tornadoes occurred. For example, the negative CAPE anomalies across the eastern US in MAM 1983 are unexpected, given that anomalously many tornadoes occurred in this season (Figure 6(f)). Examination of CAPE on 1-2 May 1983 and 18-20 May 1983, when 39 and 51 tornadoes occurred, respectively, illustrates that elevated CAPE spread from the Gulf of Mexico northward into the eastern US (Figure 9). The negative anomalies of low-level humidity across the south central US in DJF 2016 also did not represent well the environments that were present when many of the tornadoes occurred (Figure 5(d)). Closed shortwave troughs with upstream southerly moisture advection into the US were present on 23 December 2015 and 23-24 February 2016, when 18 and 36 tornadoes occurred, respectively (Figure 10). These cases illustrate that anomalously active seasons can have seemingly unfavorable seasonal composites.
Previous studies have established a relationship between tornado and ENSO in DJF and MAM, generally with more tornadoes during the LN phase and fewer during the EN phase [12-14, 17-19]. This study was focused on the seasons that do not fit this relationship--EN seasons with many tornadoes and LN seasons with few tornadoes. Specifically, the two EN DJFs and MAMs with the most tornadoes and the two LN DJFs and MAMs with the fewest tornadoes were described and analyzed. The most anomalous seasons were DJF 1983 and 2016, both of which were active EN seasons with 133 E(F)1+ tornadoes. They were, therefore, 183% (359%) above the mean (median) EN DJF. These seasons illustrate that large numbers of tornadoes are possible even during EN seasons when such large numbers might be unexpected. The other seasons were far less anomalous.
GWO does not explain the anomalous nature of the seasons. Climatological studies show that tornado activity in DJF and MAM tends to be heightened during GWO phases 1-4 when AAM is anomalously low [21,22], but nearly all of the tornadoes in the active EN seasons occurred on GWO phase 5-8 days when AAM was anomalously high. Furthermore, daily tornado count did not significantly vary across the phases of GWO in any of the seasons. The concentration of tornadoes on high AAM days (GWO phase 5-8 days) during EN seasons, as suggested here, would undoubtedly weaken the statistical relationship between tornadoes and GWO that was reported by others, whereby tornadoes are most common on GWO phase 1-4 days when AAM is anomalously low [21, 22]. Analyzing the tornado-GWO relationship by ENSO phase might amend this relationship and provide additional insight into the interactions between tornado activity, ENSO, and GWO. Another consideration is that the GWO data used in this study are based on globally integrated AAM. As noted by Gensini and Allen , this may confound the results. Therefore, it would also be worthwhile to reassess the relationship between tornado activity and GWO with AAM calculated on a hemispheric or latitudinal basis.
Unlike with GWO, daily tornado count did significantly vary across the phases of MJO, but only in the active EN seasons. The mean and mean rank of the tornado counts was greatest with phase 4 in MAM. Clustering of tornadoes on MJO phase 4 days in MAM is unique because climatological studies [20, 23] that aggregated data over a larger number of seasons reported heightened tornado activity with phase 2, 5, and 7 of the MJO in boreal spring months. While only representative of MAM 1982 and 1983, the clustering of tornadoes on MJO phase 4 days in these EN seasons highlights the possibility for a large number of tornadoes when subseasonal climate oscillations are in unfavorable phases. The mean and mean rank of the tornado counts was also greatest with phase 6 of the MJO in the anomalously active EN DJFs. Additional study is needed to determine if the tendency for DJF tornadoes to cluster on phase 6 days is common across a larger number of seasons.
This study highlights EN seasons with anomalously many tornadoes and LN seasons with anomalously few tornadoes. It also highlights days in these seasons with many tornadoes when the subseasonal GWO and MJO were in phases previously unassociated with tornado activity. These results suggest that conditions at the synoptic scale and smaller contributed most to the anomalously active seasons (i.e., favorable conditions emerged despite the seemingly unfavorable states of the climate patterns in the active EN DJFs). Active tornado days in seasons when climate patterns are more favorable are also often driven by synoptic- and subsynoptic-scale conditions, but the tornado-favorable conditions are likely to emerge more often in these seasons. This study only considered the two most anomalously active EN MAMs and DJFs and the two most anomalously inactive LN MAMs and DJFs. Additional study of a larger number of seasons is needed. Study of additional EN seasons is particularly warranted to see if the weakened association between tornado activity and GWO that was seen here is common in other EN seasons. A better understanding of the ways in which ENSO, GWO, and MJO interact across seasons to influence tornado activity will likely improve seasonal and subseasonal outlooks.
Conflicts of Interest
The authors declare that there are no conflicts of interest regarding the publication of this paper.
 National Centers for Environmental Information, "U.S. Tornado climatology," January 2018, https://www.ncdc.noaa.gov/ climate-information/extreme-events/us-tornado-climatology.
 S. Changnon, "Tornado losses in the United States," Natural Hazards Review, vol. 10, no. 4, pp. 145-150, 2009.
 K. M. Simmons and D. Sutter, Economic and Societal Impacts of Tornadoes, American Meteorological Society, Boston, MA, USA, 2011.
 K. R. Knupp, T. A. Murphy, T. A. Coleman et al., "Meteorological overview of the devastating 27 April 2011 tornado outbreak," Bulletin of the American Meteorological Society, vol. 95, no. 7, pp. 1041-1062, 2014.
 National Weather Service, "Natural hazards statistics," January 2018, http://www.nws.noaa.gov/om/hazstats.shtml.
 H. E. Brooks, G. W. Carbin, and P. T. Marsh, "Increased variability of tornado occurrence in the United States," Science, vol. 346, no. 6207, pp. 349-352, 2014.
 M. K. Tippett, "Changing volatility of U.S. annual tornado reports," Geophysical Research Letters, vol. 41, no. 19, pp. 6956-6961, 2014.
 L. Guo, K. Wang, and H. B. Bluestein, "Variability of tornado occurrence over the continental United States since 1950," Journal of Geophysical Research: Atmospheres, vol. 121, no. 12, pp. 6943-6953, 2016.
 T. W. Moore, "On the temporal and spatial characteristics of tornado days in the United States," Atmospheric Research, vol. 184, pp. 56-65, 2017.
 T. W. Moore, "Annual and season tornado trends in the contiguous United States and its regions," International Journal of Climatology, vol. 38, no. 3, pp. 1582-1594, 2017.
 M. K. Tippett, A. H. Sobel, and S. J. Camargo, "Association of U.S. tornado occurrence with monthly environmental parameters," Geophysical Research Letters, vol. 39, no. 2, article L02801, 2012.
 J. T. Allen, M. K. Tippett, and A. H. Sobel, "Influence of the El Nino/Southern Oscillation on tornado and hail frequency in the United States," Nature Geoscience, vol. 8, no. 4, pp. 278-283, 2015.
 C. Lepore, M. K. Tippett, and J. T. Allen, "ENSO-based probabilistic forecasts of March-May U.S. tornado and hail activity," Geophysical Research Letters, vol. 44, no. 17, pp. 9093-9101, 2017.
 A. R. Cook, L. M. Leslie, D. B. Parsons et al., "The impact of El Nino-Southern Oscillation (ENSO) on winter and early spring U.S. tornado outbreaks," Journal of Applied Meteorology and Climatology, vol. 56, no. 9, pp. 2455-2478, 2017.
 W. Monfredo, "Relationships between phases of the El Nino-Southern Oscillation and character of the tornado season in the south-central United States," Physical Geography, vol. 20, pp. 413-421, 1999.
 A. R. Cook and J. T. Schaefer, "The relation of El Nino-Southern Oscillation (ENSO) to winter tornado activity," Monthly Weather Review, vol. 136, no. 8, pp. 3121-3137, 2008.
 S.-K. Lee, R. Atlas, D. B. Enfield, C. Wang, and L. Hailong, "Is there an optimal ENSO pattern that enhances large-scale atmospheric processes conducive to major tornado outbreaks in the US?," Journal of Climate, vol. 26, no. 5, pp. 1626-1642, 2013.
 J. B. Elsner, T. H. Jagger, and T. Fricker, "Statistical models for tornado climatology: long and short-term views," PLoS One, vol. 11, no. 11, Article ID e0166895, 2016.
 S.-K. Lee, A. T. Wittenberg, D. B. Enfield, S. J. Weaver, C. Wang, and R. Atlas, "US regional tornado outbreaks and their links to spring ENSO phases and North Atlantic SST variability," Environmental Research Letters, vol. 11, no. 4, article 044008, 2016.
 D. B. Thompson and P. E. Roundy, "The relationship between the Madden-Julian Oscillation and US violent tornado outbreaks in the spring," Monthly Weather Review, vol. 141, no. 6, pp. 2087-2095, 2013.
 V. A. Gensini and A. Marinaro, "Tornado frequency in the United States related to global relative angular momentum," Monthly Weather Review, vol. 144, no. 2, pp. 801-810, 2016.
 T. W. Moore, "Annual and seasonal tornado activity in the United States and the global wind oscillation," Climate Dynamics, vol. 50, no. 11-12, pp. 4323-4334, 2017.
 B. S. Barrett and V. A. Gensini, "Variability of central United States April-May tornado day likelihood by phase of the Madden-Julian Oscillation," Geophysical Research Letters, vol. 40, no. 11, pp. 2790-2795, 2013.
 Storm Prediction Center, "Severe weather database files (1950-2016)," January 2018, http://www.spc.noaa.gov/wcm/.
 S. M. Verbout, H. E. Brooks, L. M. Leslie, and S. M. Schultz, "Evolution of the U.S. tornado database: 1954-2003," Weather and Forecasting, vol. 21, no. 1, pp. 86-93, 2006.
 K. E. Kunkel, T. R. Karl, H. Brooks et al., "Monitoring and understanding trends in extreme storms," Bulletin of the American Meteorological Society, vol. 94, no. 4, pp. 499-514, 2013.
 F. Mesinger, G. DiMego, E. Kalnay et al., "North American Regional Reanalysis," Bulletin of the American Meteorological Society, vol. 87, no. 3, pp. 343-360, 2006.
 Climate Prediction Center, "Cold & warm episodes by season," January 2018, http://www.cpc.ncep.noaa.gov/products/ analysis_monitoring/ensostuff/ONI_v5.php.
 R. A. Madden and P. R. Julian, "Description of global-scale circulation cells in the tropics with a 40-50 day period," Journal of Atmospheric Sciences, vol. 29, no. 6, pp. 1109-1123, 1972.
 K. M. Weickmann, S. Khalsa, and J. Eischeid, "The atmospheric angular momentum cycle during the tropical Madden-Julian Oscillation," Monthly Weather Review, vol. 120, no. 10, pp. 2252-2263, 1992.
 K. M. Weickmann and E. Berry, "A synoptic-dynamic model of sub-seasonal atmospheric variability," Monthly Weather Review, vol. 135, no. 2, pp. 449-474, 2007.
 K. M. Weickmann and E. Berry, "The tropical Madden-Julian oscillation and the global wind oscillation," Monthly Weather Review, vol. 137, no. 5, pp. 1601-1614, 2009.
 Earth System Research Laboratory, "GWO dataset," February 2018, https://www.esrl.noaa.gov/psd/map/clim/gwo.data.txt.
 M. Wheeler and H. H. Hendon, "An all-season real-time multivariate MJO index: development of an index for monitoring and prediction," Monthly Weather Review, vol. 132, no. 8, pp. 1917-1932, 2004.
 Australian Bureau of Meteorology, "Madden-julian oscillation (MJO)," February 2018, http://www.bom.gov.au/climate/ mjo/.
 V. A. Gensini, T. L. Mote, and H. E. Brooks, "Severe-thunderstorm reanalysis environments and collocated radiosonde observations," Journal of Applied Meteorology and Climatology, vol. 53, no. 3, pp. 742-751, 2014.
 National Oceanic and Atmospheric Administration, "Daily average NCEP NARR composites," February 2018, https:// www.esrl.noaa.gov/psd/cgi-bin/data/narr/plotday.pl/.
 National Oceanic and Atmospheric Administration, "NARR monthly/seasonal climate composites," February 2018, https:// www.esrl.noaa.gov/psd/cgi-bin/data/narr/plotmonth.pl.
 ESRI, ArcMAP Desktop, Release 10.5.1, Environmental Systems Research Institute, Redlands, CA, USA, 2017.
 J. O. Dickey, S. L. Marcus, and T. M. Chin, "Thermal wind forcing and atmospheric angular momentum: origin of the Earth's delayed response to ENSO," Geophysical Research Letters, vol. 34, no. 17, article L17803, 2007.
 V. A. Gensini and J. T. Allen, "U.S. hail frequency and the global wind oscillation," Geophysical Research Letters, vol. 45, no. 3, pp. 1611-1620, 2018.
Todd W. Moore (iD), Jennifer M. St. Clair (iD), and Tiffany A. DeBoer (iD)
Department of Geography and Environmental Planning, Towson University, Towson, MD 21252, USA
Correspondence should be addressed to Todd W. Moore; firstname.lastname@example.org
Received 25 February 2018; Revised 29 May 2018; Accepted 7 June 2018; Published 16 July 2018
Academic Editor: Anthony R. Lupo
Caption: Figure 1: Number of E(F)1+ tornadoes in (a) DJF and (b) MAM and the concurrent Oceanic Nino Index (ONI). Vertical dashed blue and red lines are placed at ONI = -0.5 and 0.5 and demarcate LN and EN, respectively. The blue and red diamonds represent the two most inactive and active LNs and ENs, respectively. Spearman's [rho] rank correlation coefficients and [rho] values are provided in the top-right corners of the graphs.
Caption: Figure 2: Daily tornado count of the two EN DJFs (a, b) and MAMs (c, d) with the most tornadoes and of the two LN DJFs (e, f) and MAMs (g, h) with the fewest tornadoes.
Caption: Figure 3: Phase space diagrams for the GWO (a, c, e, and g) and MJO (b, d, f, and h) during the two EN DJFs and MAMs with the most tornadoes.
Caption: Figure 4: Phase space diagrams for the GWO (a, c, e, and g) and MJO (b, d, f, and h) during the two LN DJFs and MAMs with the fewest tornadoes.
Caption: Figure 5: (a, b) 300 mb geopotential heights (contours (m)) and anomalies (color (m)), (c, d) 850 mb geopotential heights (contours (m)) and specific humidity anomalies (color (g x [kg.sup.-1])), and (e, f) CAPE anomalies (J x [kg.sup.-1]) for the two anomalously active EN DJFs.
Caption: Figure 6: (a, b) 300 mb geopotential heights (contours (m)) and anomalies (color (m)), (c, d) 850 mb geopotential heights (contours (m)) and specific humidity anomalies (color (g x [kg.sup.-1])), and (e, f) CAPE anomalies (J x [kg.sup.-1]) for the two anomalously active EN MAMs.
Caption: Figure 7: (a, b) 300 mb geopotential heights (contours (m)) and anomalies (color (m)), (c, d) 850 mb geopotential heights (contours (m)) and specific humidity anomalies (color (g x [kg.sup.-1])), and (e, f) CAPE anomalies (J x [kg.sup.-1]) for the two anomalously inactive LN DJFs.
Caption: Figure 8: (a, b) 300 mb geopotential heights (contours (m)) and anomalies (color (m)), (c, d) 850 mb geopotential heights (contours (m)) and specific humidity anomalies (color (g x [kg.sup.-1])), and (e, f) CAPE anomalies (J x [kg.sup.-1]) for the two anomalously active LN MAMs.
Caption: Figure 9: CAPE (J x [kg.sup.-1]) on (a) 1-2 May 1983 and (b) 18-20 May 1983, when 39 and 51 tornadoes occurred, respectively.
Caption: Figure 10: 850 mb geopotential heights (contours (m)) and specific humidity (color (g-[kg.sup.-1])) on (a) 23 December 2015 and (b) 23-24 February 2016, when 18 and 36 tornadoes occurred, respectively.
Table 1: Minimum, mean, and maximum values of global atmospheric angular momentum (AAM; kg x [m.sup.2] x [s.sup.-2]) anomalies during the active EN and inactive LN seasons. Seasons Minimum Mean Maximum Active EN seasons DJF 1983 1.3 2.6 4.4 DJF 2016 0.4 1.9 4.0 MAM 1982 -1.3 0.1 1.4 MAM 1983 0.1 1.8 3.5 Inactive LN seasons DJF 1985 -2.0 -0.9 0.1 DJF 1986 -1.1 0.1 1.3 MAM 1985 -1.0 0.1 1.1 MAM 2000 -2.7 -1.5 -0.4 Table 2: Mean (mean rank) of daily tornado count by GWO and MJO phase. GWO phase 1 2 3 4 Active EN DJF 0.0 (70.0) -- -- -- Active EN MAM 3.8 (100.5) 2.9 (86.8) 7.5 (106.9) 3.2 (81.1) Inactive LN DJF 0.1 (89.4) 0.1 (90.5) 0.1 (86.4) 0.1 (87.2) Inactive LN MAM 2.3 (115.4) 3.6 (100.0) 1.4 (88.8) 1.0 (79.8) MJO phase Active EN DJFa 3.5 (113.2) 0.4 (90.0) 1.9 (101.4) 0.7 (81.2) Active EN MAMb 1.1 (58.0) 2.5 (85.4) 2.6 (87.3) 5.9 (116.1) Inactive LN DJF 0.4 (100.3) 0.0 (84.0) 0.1 (88.0) <0.1 (87.3) Inactive LN MAM 2.0 (96.6) 2.2 (87.3) 2.7 (92.3) 0.5 (70.1) GWO phase 5 6 7 8 Active EN DJF -- 1.5 (83.3) 1.3 (92.8) 2.3 (96.3) Active EN MAM 3.6 (83.9) 3.3 (89.5) 2.7 (98.3) 2.8 (88.2) Inactive LN DJF 0.0 (84.0) 0.5 (100.5) 0.8 (106.5) 0.1 (90.2) Inactive LN MAM 2.1 (93.6) 1.8 (111.0) 2.6 (79.8) 1.8 (84.8) MJO phase Active EN DJFa 1.1 (80.6) 4.0 (119.9) 1.3 (80.5) 0.3 (80.6) Active EN MAMb 1.2 (72.5) 1.5 (83.3) 4.5 (101.9) 3.1 (97.8) Inactive LN DJF 0.1 (91.4) 0.2 (91.7) 0.5 (92.7) 0.1 (87.9) Inactive LN MAM 2.3 (103.0) 2.6 (96.7) 2.0 (95.6) 1.7 (94.5) (a) A Kruskal-Wallis test indicates that the mean rank of tornado count varies across show that the mean rank of tornado count is greater in phase 6 than in phases 4, 7, and 8. (b) A Kruskal-Wallis test indicates that the mean rank of tornado count varies across the phases of MJO ([chi square] = 17.9; df=7; p = 0.012). Post hoc comparisons show that the mean rank of tornado count is greater in phase 4 than in phase 1.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Research Article|
|Author:||Moore), Todd W.; St. Clair, Jennifer M.; DeBoer, Tiffany A.|
|Publication:||Advances in Meteorology|
|Date:||Jan 1, 2018|
|Previous Article:||Spatial Distribution and Temporal Trend of Tropospheric N[O.sub.2] over the Wanjiang City Belt of China.|
|Next Article:||Global and Regional Remote Sensing Precipitation Estimation, Evaluation, and Applications.|