sign in sign up
<<
Home , Cape Verde hurricane, Rainband, Supercell, Tropical cyclone

Comparative Analysis of Environment in Hurricanes Harvey and Irma

IVY MACDANIEL Affilitation/Fall 2020: Austin Peay State University [email protected]

Significant Opportunities in Atmospheric Research and Science

Science Mentor: Christopher Rozoff, Dereka Carroll, and Jonathan Vigh Writing/Communication Mentor: Chris Davis Coach: Kristen Aponte

July 31, 2020

Abstract

Although severe forecasting has improved over , small rotating , some- called miniature , often pose challenges with tracking and issuing warnings. Being smaller than typical supercells, those are more difficult to discern on radar, especially at greater distances. Those storms are also found in tornado-producing tropical , and tornadoes produced by tropical cyclones, while they tend to be weak, can still pose a threat to and properties in the Southeast U.S. The most likely tropical to produce tornadoes is a formerly intense that weakens at . Tornadoes within Hurricanes Harvey and Irma from the 2017 Atlantic were examined in this study. In Harvey’s case, the majority of tornadoes formed inland from southeast to . Mesoanalysis showed moderately high convective available potential en- ergy (CAPE) and more moderate vertical shear than was found in Irma. Irma, however, produced the majority of tornadoes near the shoreline. Environmental analysis showed higher CAPE over than over the peninsula during Irma’s landfall. Yet, the shear was higher over land. However, supercells met their demise over the land when the environment became unfavorable. While it was still unclear why the supercells died over land, a key observation showed that there was a significant decrease of CAPE over land compared to offshore despite favorable shear. However, supercells in post-tropical Harvey formed and sustained themselves far inland while being driven by buoyancy. Thus, additional investigation in the future would be necessary.

This work was performed under the auspices of the Significant Opportunities in Atmospheric Research and Science Program. SOARS is managed by the University Corporation for Atmospheric Research and is funded by the Na- tional Science Foundation, the National Center for Atmospheric Research, the National Oceanic and Atmospheric Administration, and the University of at Boulder. 1 Introduction

Over the decades, the research has helped save lives thanks to advanced technology and improved warnings. However, the research has been based heavily based on classic supercells with a limited scope for miniature supercells regardless whether they were part of tropical cyclones (TCs) or not. Yet, miniature supercells, especially those in TCs, should not be dismissed since they are capable of producing tornadoes that pose threat to lives and properties. The occurrence of tornadoes within tropical cyclones also complicates the warning process because of the presence of multiple hazards at the same time. Early research into TC tornadoes from hurricanes and (Novlan and Gray, 1974) revealed key fundamentals about the location and general occurrence of TC tornadoes within TCs. The Novlan and Gray study examined US hurricanes from 1948 to 1972 and Japanese typhoons from 1950 to 1971. Tornadoes were found most likely to occur during amid the increasing vertical as observed in some hurricanes and typhoons. Here, vertical wind shear refers to the variation of the horizontal wind with height. Overall, TC tornadoes were found to be weaker than their Great Plains counterparts. However, 10% of TC-related deaths from that period were attributed to tornadoes. In observed TCs, tornadoes usually form in the front right sector of TCs (oriented relative to the move- ment of the TC) (Novlan and Gray, 1974). The most prolific tornado-producing TCs are dissipating, but formerly intense TCs. They have on average three times the filling rate, (as measured by a change in the minimum central sea-level pressure of roughly 30 mb/12 hr), compared with non-tornadic TCs that weak- ened at a more moderate (10 mb/12 hr). The shear was also much higher on average for tornado-producing TCs than non-tornadic TCs, with the average vertical shear of tornado-producing environments being greater than 40 kt from surface to 1524 m (5,000 feet). Non-tornadic TCs, however, tend to show minimal to no shear for same layer of . It also suggested that dry air intrusion plays a role in TC tornadoes as observed in Hurricanes Jeanne and Ivan (Baker et al., 2008). Ivan experienced a dry air intrusion and produced several tornadoes. Jeanne, however, did not experience a dry air intrusion and did not produce any tornadoes. Research has shown that there are distinctive differences between TC and classical Great Plains en- vironments (Novlan and Gray, 1974). Though most of the Great Plains and TC tornadoes are produced by supercell storms, long lived thunderstorms with rotating updrafts, the environments of tornadic storms within TCs are different from the tornadic Great Plains storms. Usually, the Great Plains environment fea- tures high convective available potential (CAPE) and moderate vertical shear over a deep layer. TC environments may have lower overall CAPE but strong vertical wind shear at low . The majority of TC tornadoes were spawned by supercells where low level shear was usually generated in the during the landfall as a result of surface friction (Green et al., 2011). Surface friction dramatically slows the near the ground, and also changes the flow direction inward toward low pressure, whereas winds 500-1000 m above the ground are less affected. Therefore, vertical wind shear in the very lowest levels of the atmosphere increases significantly upon landfall. Approximately 79% of all tornadoes produced by TCs were produced by right moving supercells (Ed- wards et al., 2012), meaning that they moved to the right of the vector wind shear averaged over a deep layer. Miniature supercells, characteristic of most TC tornado-producing storms, are smaller than classi- cal supercells over the Great Plains. Average top heights are 24,000 to 32,000 feet for miniature storms (NWS Louisville, n.d.). Classical supercells are known to produce intense severe weather usually have 40-50,000 feet cloud tops (up to 16 km). While the tornadoes produced in tropical cyclones are weak, roughly 7% are rated at EF-2 or greater, where EF refers to the and ranges from 0 (minimal) to 5 (catastrophic) damage (Edwards et al., 2012). Tools for non-TC supercells can be used for their TC counterparts (Baker et al., 2008), and these include parameters derived from atmospheric soundings and surface meso-analyses that depict fine-scale structures in the surface winds. Common tool for severe is sounding which can be used in TCs. A couple exampes of those forecasting tools

1 are sounding and mesoanalysis. According to reports by from the National Hurricane Center, Category 4 produced 52 tornadoes from Texas to Tennessee (Blake and Zelinsky, 2018). Irma, on the other hand, produced 25 tornadoes along the Florida and coasts (Cangialosi et al., 2018). As a hurricane, Irma was a Category 5 that tracked through islands and along the Florida coast. Both Harvey and Irma were selected from a pool of candidate TCs. The whole scope of the project this and onward was to study miniature supercells and their environments within tropical cyclones and outside tropical cyclones. Some key objectives were examining different tropical cyclones to examine environment the supercells were in, taking detailed observations of supercells, and examining nontropical environments. It could be expanded to study how miniature supercells behave in tropical cyclones and in various non-tropical scenarios. Two possible expansions were miniature supercells with history of significant severe (EF-2+ tornadoes, 75+ mph/121 kph, or 2+”/5 cm hailstones) and features of supercells in various sectors of tropical cyclones at different stages. Therefore, the findings described in the paper thus serve as a preliminary study that can be expanded on through future studies.

2 Methods

After investigating different potential TC cases, Harvey and Irma were determined to be best cases because they were prolific tornado producers and also strong TCs. The pool of candidate TCs for this study was major Atlantic hurricanes from 2017 to 2019. Those that did not make landfall in U.S. or were eliminated. Originally, the cases were Harvey, Irma, Maria, Michael, and Dorian, but they were reduced to Harvey and Irma due to time constraints of the project. Harvey and Irma were prolific tornado producers, but they spawned tornadoes in different regions, primarily inland for Harvey and near the immediate coast for Irma. This difference warranted investigation primarily based on observations. There is no single comprehensive source of observations for tropical cyclone tornadoes. Therefore, I had to piece together information from various sources. The (SPC) Severe Weather Event Archive and Preliminary Reports contained the information needed to identify pinpoint the time and location of each tornado. This was useful for identifying which upper air soundings could be considered as representative of the environment of the storms. Upper air soundings were obtained from SPC and the University of Wyoming archives. For offshore information, I considered the FLIGHT+ dataset from Dr. Jonathan Vigh at NCAR (Vigh et al., 2020). These data contain observations from reconnaissance aircraft. Further However, further exploration of these data will be left for future investigation. To identify the times and locations of rotating thunderstorms, I used radar archive information from the NOAA National Center for Environmental Information (NCEI) Toolkit for radar images. From these data I selected two particular storms from Irma’s principal , and tracked these storms as they evolved. At the locations of the storms I estimated environmental parameters CAPE and vertical wind shear in the lower from surface to 3 km. These parameters are important for the dynamics of rotating thunderstorms, especially the lowest kilometer of the atmosphere. Supercells, even miniature versions, require CAPE to sustain their updrafts. The shear helps separate updraft from downdraft, thus allowing the storms to sustain longer than ordinary cell . This happens through the creation of vertical pressure gradients that force air upward preferentially on the flank of the updraft and help the updraft propagate either to the left or right of the wind shear vector. The NOAA Climate Toolkit provided archive of radar data from radar sites which was useful for examining internal structure and rainbands for supercells with aid of archived tornado watches and reports from the SPC. North American Regional reanalysis (NARR) gridded data were interpolated in MATLAB with and longitudes of selected supercells from Irma’s principal rainband (MATLAB, 2019). The NARR represents a best estimate of the atmospheric state from a wide range of the available observations. Analyses are available every 3 hours. This is long compared to the duration of supercell storms in most cases.

2 3 Results

Hurricane Harvey produced numerous tornadoes near the time of landfall on August 25, 2017, but produced the majority of tornadoes over several days following its initial landfall. Harvey is well known for its slow movement and erratic path. It actually moved offshore again on August 28 and made landfall a second time in on August 30. Irma produced most of its tornadoes over eastern Florida near its time of landfall on the southwest coast of Florida. Irma decayed rapidly after landfall and did not produce many tornadoes on subsequent days because its circulation became too weak. Doppler radar is probably the best tool we have to view the characteristics of supercell storms. Both Harvey and Irma featured supercells whose cloud top heights were roughly 7620 to 9140 m (25-30,000 feet) estimated from radar. This fact, plus their small horizontal extend of only a few kilometers (Figs. 1-3) places the storms into the “miniature” category. Being miniature supercells, those storms were more difficult to discern on radar with more diminutive appearance compared to classical Plains supercells with hook echos, especially at greater distances from the radar where even the lowest level scan was still too high to capture all of the low-level rotation signature. For an example, Figure 1 shows a couple of individual supercells from Harvey when it was in its later stage of producing tornadoes over and Tennessee. Despite the image being taken while one of those supercells was producing a tornado, the classical “” (literally a radar echo that curls back on itself) is absent. Shallow supercells often required more attention to details since the signatures of rotation can be vague for storms more than 50-100 km from the radar. Additionally, a supercell (Fig. 2) in Irma’s principal rainband showed similarly vague features. It was one of multiple supercells observed offshore in the rainband. However, a broader view of an entire “garden” of supercells can be seen in Fig. 3 which reveals more than 20 miniature supercells at the same time. The mean flow in this image is very strong (30 m/s or more) from the east. These storms are moving quickly from right to left. Judging from reflectivity, these storms reach their maximum intensity within 30-40 km from the coast, and clearly have a greater intensity offshore. Many of the storms have a curved shape. Though smaller, they still have noticeable features similar to classical supercells, such as distinct hook echoes. Once onshore, the storms move rapidly inland and decay. Evidence of remnant storm structure can be seen in some of the more moderate (yellow) areas toward the left side of the image. From the Doppler velocity, which is a measure of the wind component toward or away from the radar (Fig. 4), several velocity couplets were noted, appearing as darker circles, sometimes green in the center. Multiple clear couplets in velocity confirmed that there was indeed a wave of incoming miniature supercells from offshore. These couplets denote a large horizontal variation of velocity within the storms in the sense of cyclonic (counterclockwise) rotation. One can infer velocity couplets of roughly 50 m/s in magnitude occurring on the scale of only a couple of kilometers. These signatures are not tornadoes themselves, but they are indicative of storms that can produce tornadoes (or tornadic , as they are known offshore). Two supercells were selected for this study with and longitude data. Supercell A was noted to leave higher CAPE (Fig. 5) at 19 z with a sudden decrease at 1920z with a brief increase at 1950z before being deprived of CAPE it needed to sustain its updraft. The starting CAPE for Supercell A was approximately 1,180 j/kg, and it decreased to 1,000 j/kg before momentarily increasing to 1120 j/kg. After the second increase, it dropped off as Supercell A moved onshore. In meantime, the shear (0-3 km helicity) increased in the same period from 140 m2 s-2 to 200 m2 s-2 onshore. Supercell B was noted for its similar pattern of leaving higher CAPE (Fig. 7) into lower CAPE onshore. From 1900z to 1940z, the CAPE increased from 1,000 to 1,100 j/kg, but after 1940z, the CAPE dropped from 1,100 j/kg to below 800 j/kg when Supercell B met its demise. However, the shear increased rapidly during the period from 150 m2 s-2 to above 195 m2 s-2 once Supercell B moved onshore and decayed. There appears to be a ”Goldilocks zone” of favorable environment alongside the shoreline where tracks

3 of Supercells A and B were interpolated. Along the shore, there was an overlap of moderate CAPE of 800 to 1,200 j/kg (Fig. 9) and 0-3 km helicity ranging from 100 to 200 m2 s-2 (Fig. 10). In the case of CAPE, most of the Florida peninsula except for southern tip saw low CAPE ranging from 0 to 600 j/kg. The area of higher CAPE at southern tip was to the northeast of Irma’s center. The helcity was far higher on land except for the southern tip adjacent to Irma’s center which is an area of lowest helicity. The helicity was more moderate offshore, however. Time of the day may have played a role since there was a notable decrease of CAPE in a few hours from 18z (Fig 11) to 21z (Fig 12) while the shear remained largely unchanged from 18z (Fig 13) to 21z (Fig 14). It should be pointed out that the the range of latitude was 26°to 28°N and the longitudes range was -80°to -70°E where most of supercells in the rainband were. Some of early analysis of Harvey in comparison of Irma indicated that Harvey’s tornadoes (Fig. 15) in wider range of CAPE in an environment of slightly lower helicity values than those formed in Irma (Fig. 16). In Harvey, majority of tornadoes formed in environment with SRH less than 300 m2 s-2. Overall, tornadoes in Irma formed under lower CAPE ranging from 500 to 1,500 j/kg with lone outliner in 2,500 j/kg CAPE.

4 Discussion

While the supercell thunderstorms investigated in this study could be classified as miniature, and often produced weaker tornadoes compared with classical Great Plains counterparts, but their large numbers still imply a significant hazard. Consistent with classical supercells, substantial CAPE is important for main- taining a strong updraft. Unlike classical supercells, CAPE in landfalling tropical cyclones is comparatively small. The higher CAPE in Irma’s principal rainband was found offshore, and this fact, combined with low-level shear increasing closer to the coast, narrowed the region of supercell environmental conditions to the coastal zone. Given the inferred importance of high SST for maintaining at least a minimal amount of CAPE, it is curious that the supercells thrived over water in daytime (18-21z). Normally, we would not expect a strong diurnal cycle of CAPE over water. Perhaps the increase of CAPE is more related to the changing structure of the approaching hurricane than to the diurnal cycle itself. This hypothesis remains for future investigation. However, both Harvey and Irma were quite different in tornado production since Harvey was not pro- lific until it became post-tropical remnant. Irma produced several supercells and tornadoes during Florida landfall. This study demonstrates that landfalling hurricanes of similar intensity can produce very different patterns of tornadic storms. Also, monitoring offshore environment is essential for making forecasts relating to tornado potential that pose threats to wind farms, oil rigs, and rarely, boaters. A TC remnant should not be ignored, especially when it encounters a favorable environment and contribute to tornadoes. More work is required for Harvey, however. There was no supercell tracking data and NARR files for Harvey due to time constraints. FLIGHT+ data was originally planned for analysis this summer, but it would be attempted next year. A much more detailed analysis would be necessary for Harvey compared to Irma because the tornadoes related to Harvey extended over several days and occurred in widely differing environments. One example was a cluster of multiple tornadoes in southeastern Texas when Harvey made its initial landfall. Harvey eventually produced a series of multiple tornadoes over next several days during its slow meander from Texas to S.E. US. The scatterplots of tornado reports in environmental conditions indicated that Harvey’s tornadoes were reported in areas with higher overall buoyancy in somewhat favorable shear while Irma’s were reported in areas experiencing Irma’s powerful dynamics despite lackluster CAPE over land. However, it should be pointed out that Irma’s supercells formed in higher CAPE environments offshore and migrated to Florida peninsula where there were lower CAPE, but lower CAPE is usually offset by high shear. While briefly investigated, a nocturnal supercell that produced long-track Nashville EF-3 and Cookeville EF-4 tornadoes is being considered one of the potential non-TC cases for future work. Also, Harvey and

4 Irma had been part of this project, the original plan had Hurricanes Maria, Michael, and Dorian included. These storms may be included in the future.

5 Conclusion

This study has examined the phenomenon of miniature supercell storms that produce tornadoes in land- hurricanes. The analysis in this paper focused on , a major hurricane that spawned many tornadoes on the East Coast of Florida near the time of landfall. A secondary objective of this study was to compare the tornadic storms in Irma with those in Hurricane Harvey, a much different storm that produced tornadoes over many days. The primary finding is a large number of supercell storms preferen- tially offshore in Irma. These storms were fueled by the combination of significant conditional instability and strong low-level vertical wind shear, with the most favorable region being near the coast extending a few 10’s of kilometers offshore. The spatial variation of both CAPE and low-level shear were large. CAPE appeared to be enhanced by the warm offshore, which maintained instable conditions presumably through transfer of heat and moisture from the ocean surface into the lower atmosphere. Heavy rainfall over land appeared to diminish conditional instability by cooling the lowest levels of the atmosphere. There- fore, while the low-level wind shear was largest over land, the instability was insufficient to allow strong . Preliminary analysis of Harvey revealed that numerous tornadoes formed near the coast at the time of landfall, as with Irma. But Harvey continued for many days as either a weak tropical storm or tropical depression, and spawned many tornadoes, and most of these tornadoes were farther inland than in the case of Irma. However, Harvey was apparently more driven by buoyancy than Irma given that it spent days on land when it was spawning tornadoes. Future work will focus on the changing environmental conditions for tornadoes in Harvey over the period of several days following Harvey’s landfall. The long-term goal of this research is to understand the difference between landfalling tropical cyclones that produce many tornadoes versus few tornadoes in terms of environmental characteristics. It is also possible that this research will need to consider the structure of the rainbands within which these tornadoes occur. If a strong linkage to environmental characteristic can be obtained, better capability for useful warnings about tornado risk associated with landfalling hurricanes is possible. To elborate on future work, Harvey needed more attention and analysis in similar manner as Irma did in the short-term, so Harvey would be top priority. For Harvey, Irma, and other potential TC cases, incomplete MATLAB scripts for FLIGHT+ and data needed to be completed before data can be analyzed for in-depth environmental condition in tropical cyclones. Though NARR files had been used briefly for Irma, NARR files could be applied to tropical and non-tropical cases. In addition, non-tropical miniature supercells would be examined, and those with a history of intense tornadoes would be investigated in depth. The author added that investigating features and environments of miniature supercells with a history of significant severe would offer more answers to how supercells function in lower CAPE and higher shear similar to TC environment.

5 References

Baker, A. K., M. D. Parker, and M. D. Eastin, 2008: Environmental Ingredients for Supercells and Tornadoes Within . Weather Forecasting, 24, 223–244.

Blake, E. S., and D. A. Zelinsky, 2018: Hurricane Harvey. National Hurricane Center Tropical Cyclone Report, National Hurricane Center, 77 pp.

Cangialosi, J. P., A. S. Latto, and R. Berg, 2018: Hurricane Irma. National Hurricane Center Tropical Cyclone Report, National Hurricane Center, 111 pp.

Edwards, R., D. R. Andrew, R. L. Thompson, and B. T. Smith, 2012: Convective Modes for Significant Severe Thunderstorms in the Contiguous . Part III:Tropical Cyclone Tornadoes. 1507–1519.

Green, B. W., Z. Fuqing, and P. Markowski, 2011: Multiscale Processes Leading to Supercells in Landfalling Outer Rainbands of (2005). Weather Forecasting, 26, 828–847.

MATLAB, 2019: 9.7.01190202 (R2019b). The Mathworks Inc.

Novlan, D. J., and W. M. Gray, 1974: Hurricane-Spawned Tornadoes. Monthly Weather Review, 102, 476– 488.

NWS Louisville, K., n.d.: Mini supercell thunderstorms. Accessed July 22, 2020.

Vigh, J., and Coauthors, 2020: Flight+: The extended flight level dataset for tropical cyclones (version 1.3). Tropical Cyclone Data Project, National Center for Atmospheric Research, Research Applications Laboratory, Boulder, Colorado. Accessed June 4, 2020.

6 Figure 1: Radar image of couple supercells in southwest Tennessee taken from KOHX (Nashville, TN). One on the left was producing a weak tornado in Olive Hill, TN.

Figure 2: Close up of an individual supercell in Irma’s rainband taken from KJAX (Jacksonville, FL)

7 Figure 3: Reflectivity of multiple supercells in Irma’s rainbands taken from KMLB (, FL)

Figure 4: Radial velocity of multiple supercells in Irma’s rainbands taken from KMLB (Melbourne, FL)

8 Figure 5: Plot of CAPE over time for Supercell A

9 Figure 6: Plot of helicity over time for Supercell A

10 Figure 7: Plot of CAPE over time for Supercell B

11 Figure 8: Plot of helicity over time for Supercell B

12 Figure 9: CAPE grid interpolated with two supercell tracks over general Florida peninsula

13 Figure 10: Helicity grid interpolated with two supercell tracks over general Florida peninsula

14 Figure 11: 21z 3-hour average of CAPE from NARR files

15 Figure 12: 21z 3-hour average of CAPE from NARR files

16 Figure 13: 18z 3-hour average of helicity from NARR files

17 Figure 14: 21z 3-hour average of helicity from NARR files

18 Figure 15: Scatter plot of reported tornadoes in corresponding SRH and CAPE produced by Hurricane Harvey

19 Figure 16: Scatter plot of reported tornadoes in corresponding SRH and CAPE produced by Hurricane Irma

20