ERAD 2010 - THE SIXTH EUROPEAN CONFERENCE ON RADAR IN METEOROLOGY AND HYDROLOGY

The structure and behaviour of supercell storms in the State of ,

G. Held1, A. M. Gomes1, K. P. Naccarato2 1Instituto de Pesquisas Meteorológicas, Universidade Estadual Paulista, , S.P.

2 Grupo de Eletricidade Atmosférica, Instituto de Pesquisas Espaciais, São José dos Campos, S.P. Gerhard Held

1. Introduction

In the State of São Paulo, located in the south-east of Brazil, severe convective storms frequently cause local flooding in towns and/or devastation of crops and property due to hail or wind storms (microbursts, tornadoes), resulting in many millions of US Dollar damage annually, loss of lives and leave many persons injured (Gomes et al., 2000; Held et al., 2001; Held and Nachtigall, 2002). The Bauru-based Meteorological Research Institute (IPMet) has been observing and analyzing the three-dimensional structure of thunderstorm and rain events in the central and western region of the State of São Paulo, using S-band Doppler radars, since 1992. The first supercell storm recorded by the Bauru radar occurred on 14 May 1994 (Gomes et al., 2000). However, no similar event was observed during the following years and therefore it was believed, that supercell and especially tornadic storms were rather rare events in Southeast Brazil, since very few of those reported had been observed within radar range. However, this changed when the first severe tornadoes, most of them associated with supercells, were observed by IPMet´s radars in May 2004 (F1 and F2, according to the Fujita scale), followed by an F3 in May 2005 (Held et al., 2005, 2006a, 2006b). These latter events also prompted the analysis of lightning flashes relative to the storm areas, from 2004 onwards, and in collaboration with the Lightning Research Group (ELAT) of the Institute for Space Research (INPE).

2. Data and Methods

IPMet’s radars are located in the central and western State of São Paulo, viz. in Bauru and Presidente Prudente, which is 240 km west of it (275 azimuth; Figure 1). Both have a 2° beam width and a range of 450 km for surveillance (0 or 0,3 PPI every 30 or 15 min), covering the entire State of São Paulo, but when operated in volume-scan mode every 15 or 7,5 minutes it is limited to 240 km, with a resolution of 1 km radially (250 m since March 2006) and 1° in azimuth, recording reflectivities, spectral width and radial velocities. The Brazilian Lightning Detection Network (BrasilDat) currently comprises 47 sensors in total (most are outside the area shown in Figure 1), with a detection efficiency of 80-90 % (CG = Cloud-Ground strokes only) and a location accuracy of 0,5 - 2,0 km (Pinto Jr., 2003). Reliable lightning observations are only available from 1999 onwards. The analysis of the various cases was performed with Sigmet/IRIS/Analysis, the propriety Software of IPMet’s radars, and NCAR’s (National Center for Atmospheric Research) TITAN (Thunderstorm Identification, Tracking, Analysis and Nowcasting; Dixon and Wiener, 1993) Software, which had been implemented at IPMet and adapted for local requirements in 2005/2006. TITAN produces a variety of important parameters for a chosen reflectivity and volume threshold throughout the lifetime of storms, such as Area, Volume, Precipitation Flux, VIL (Vertically Integrated Liquid water content), Maximum Reflectivity, Hail Metrics, speed and direction of propagation, etc, per volume scan, as well as cell tracking, including splits and mergers of cells. It also has the facility to superimpose flashes with the radar echoes, including a separation into positive and negative strokes. For this analysis TITAN was running with a resolution of 1 km in the horizontal and 750 m in the vertical. A reflectivity threshold of 40 dBZ with a volume of >16 km3 was chosen for tracking the cells.

3. Case Studies

In this paper, all known cases of supercell occurrences will be discussed, but this does not imply that there are not more cases, which have not yet been detected in the vast amount of 18 years of Doppler radar observations. The days which have been studied so far are: 14 May 1994 (one supercell lasting >5 hours), 17 October 1999 (hail swath extending for 200 km during 4 hours), 21/22 May 2004 (one supercell lasting >5,5 hours), 25 May 2004 (two supercells moving on parallel tracks for >8,5 hours and >5 hours, respectively, but only the shorter one produced an ERAD 2010 - THE SIXTH EUROPEAN CONFERENCE ON RADAR IN METEOROLOGY AND HYDROLOGY

F2 tornado), 24 May 2005 (two supercells moving on parallel tracks for >3 hours and almost 4 hours, respectively, but only one produced an F3 tornado) and 24 July 2007 (hail swath extending for about 300 km during 5 hours). It is noteworthy, that all, but one, of these cases occurred during the month of May, which is the austral autumn, and thus a transition period of the atmosphere from summer conditions, with predominantly moist tropical air being advected from Amazônia, to the winter regime, when occasional baroclinic systems (cold fronts), coming from south-west, affect the western and central interior of the State. A similar seasonality was observed for the occurrence of confirmed tornados, which have a maximum frequency in autumn, followed by a secondary spring maximum (Held et al., 2006a). All times in the case studies are in Local Time (LT = UT-3h), unless otherwise stated (TITAN uses UT).

FIG. 1. IPMet’s Radar Network (BRU = Bauru; PPR = Presidente Prudente), showing 240 and 450 km range rings. Lightning sensors are marked as red stars.

3.1 14 May 1994

A frontal system, extending from an Atlantic Ocean Low across the southern States of Brazil, and an associated cut-off low over the State of São Paulo induced strong convergence at the surface, which triggered and maintained a supercell storm with radar reflectivities exceeding 60 dBZ for five hours along its atypical southward path. It caused extensive damage to property (US $ 11 million) and life, killing three people and injuring 130 along its path of destruction. Numerical meso-scale modeling of the vorticity budget for this day indicated that the divergence term was dominating the low-level vorticity tendency and that the horizontal advection of vorticity was working against the cyclonic vortex evolution (Silva Dias et al., 1996). During the early afternoon on this day, radar observations already indicated the development of isolated precipitation cells in the central and far north-eastern area of the State of São Paulo. The convective activity intensified from 15:00 onwards with the majority of storms developing during the late afternoon and early evening. Maximum reflectivities exceeded 55 dBZ and echo tops were on average above 8 km. In contrast to the supercell, they generally moved in south-easterly directions with average speeds of 50 km.h-1, producing severe hail falls and strong surface winds along their paths. Significant damage to houses and people, as well as great damage to coffee plantations in the central and southern parts of the State, was reported as a result of these storms. Several of these severe storms have been described in detail by Gomes et al. (2000), but the emphasis in this discussion will be on the supercell storm only. Radar surveillance indicated very strong convective activity over southern at 16:30, moving southwards along an unusual north - south trajectory at an average speed of 62 km.h–1 (Gomes et al., 2000). At least one cell of this long-lasting storm complex could be classified as a supercell storm, because radar reflectivities exceeding 60 dBZ were recorded for five hours. Figure 2a shows the tracks of all cells for a 24-hour period, starting 09:00 on 14 May 1994, the majority being long-lived, but not in the same category as this superell, visible on the far right. When the reflectivities of this particular cell were analyzed in conjunction with the radial velocity fields, rotation signatures that could be related to the observed severity were identifiable. The supercell storm was first detected at 5,5 km (48 dBZ) at 21:47. For most of its five-hour life cycle the maximum reflectivity exceeded 60 dBZ and echo tops varied between 10 and 13 km. Figures 2b and 2c show the reflectivities and radial velocities of this storm at 23:02, respectively, clearly indicating an anticyclonic rotation, coincidental with the echo core (“couplet”). Considering that the storm was more than 150 km north-east of the radar, the rotation was evident between 0.3° and 1.7° elevation, corresponding to 4–6 km height, but most prominent at 4 km. It should be noted that in the south-eastern part of the rotation core radial velocities of up to +10 m.s-1 (outwards) were observed, while on the opposite side the radial velocities varied between –9 and –14 m.s-1 (towards ERAD 2010 - THE SIXTH EUROPEAN CONFERENCE ON RADAR IN METEOROLOGY AND HYDROLOGY the radar). The centers of velocity maxima were about 6 km apart (Figure 2c), thus implying a local shear of –3,5x10- 3s-1. Such values are generally associated with meso-cyclones or meso-anticyclones and are precursors for the formation of tornadoes (Nielsen-Gammon et al., 1995).

a) b) c)

FIG. 2. 14 May 1994: (a) Cell tracks during the 24-hour period from 09:00 to 15 May 1994, 09:00; (b) Reflectivity of the supercell at 23:02 and (c) Corresponding radial velocity field.

3.2 17 October 1999

A similar frontal system as on 14 May 1994, with a surface trough extending over the south-south-eastern States of Brazil, resulted in a line of precipitation over the State of São Paulo. A closed cyclonic vortex was positioned over eastern Mato Grosso do Sul, associated with strong convergence from the surface up to 200 hPa over northern Paraná and central São Paulo, resulting in high instability. Again, severe hail falls from several individual storms were noted and extensive damage to property and life was reported. Radar observations indicated an extensive area of stratiform precipitation with several intense cells forming along the leading edge of the squall line. The most intense cells are labeled A and B and the positions of the 40 dBZ contour envelopes are shown in Figure 3 for a 195-minute period, including the 1-hour forecast, which was shown to be correct. The hail-producing Cell A moved at 60 km.h-1 to the north-east, with reflectivities mostly in excess of 55 dBZ (max. 62-70 dBZ) and echo tops (10 dBZ) between 13 and 16 km height. These characteristics are associated with strong vertical air flows, resulting in a destructive hail path along its

trajectory, lasting for 4 hours. Radial velocities in the vicinity FIG. 3. TITAN cell track image of 17 October 1999 at 11:22: annotation indicates velocities (km.h-1), the thin of the maximum reflectivity core indicated a signature of yellow ellipses show the past track every 15 min and the red strong convergence at about 4 km, reaching its most intense ellipses represent a 1-hour forecast. phase at 14:01, when large hailstones fell in the town of Tibiriçá.

3.3 21/22 May 2004

A frontal system, approaching from the south-west during early morning, traversed the entire western and central State of São Paulo in the course of 21 May 2004, and continued after midnight across the southern parts of Minas Gerais. Due to great instability, widespread severe thunderstorms occurred, with the supercell trailing the leading edge of the storm complex slightly on its western flank for at least 7 hours. Confirmation was received, that it caused significant damage along its track in the northern part of the State (Riberão Preto and Batatais), but most likely also during its continuation in Minas Gerais. The first echo of this supercell was detected to the west-north-west of shortly before 18:00, and TITAN tracked its 40 dBZ contour from 18:23 until 01:08 (Figure 4a), but the cell continued beyond the 240 km range with undiminished intensity, as confirmed by the 0,3º surveillance PPIs. It moved at 55 - 60 km.h-1 towards north-east. It is noteworthy, that during its >6 h 45 min lifetime the supercell received a boost as it traversed the Tietê River at 21:07, most likely ingesting additional moisture through its updraft at the leading edge (light blue echo in Figure 4a). ERAD 2010 - THE SIXTH EUROPEAN CONFERENCE ON RADAR IN METEOROLOGY AND HYDROLOGY

This has also been observed in other cases, e.g., cell T3 on 24 May 2005, which followed the Tietê River before spawning the tornado (Figure 6a).

(a) (b)

FIG. 4. 21/22 May 2004: (a) Composite radar image, showing the tracks of 40 dBZ centroids of the supercell between 18:23 - 01:08 LT. Not all simultaneous tracks are shown. (b) Storm Time History of the most intense part of the supercell track from 22:30 - 00:08 LT.

The Storm Time History during the second part of the cell track (21:07 - 01:08; Figure 4b) revealed a maximum reflec-tivity of 65 dBZ for most of the time, a Probability of Hail (POH; Waldvogel et al., 1997) of 80–100 % and an FOKR Index (Foote et al., 2005) of 3. The VIL was generally between 55-70 kg.m-2, with a maximum of 85 kg.m-2 at 23:37.

3.4 25 May 2004

Several storms, associated with areas of strong convective activity, created by the passage of a baroclinic system with strong convective instability and vertical wind shear, had been observed by the radars on 25 May 2004 since early morning. Two supercells developed, one of which (C1) traversed almost the entire State during its >8,5 hour life cycle (Held et al., 2005, 2006a, 2006b). Since these storms occurred during the southern hemisphere autumn, the cells were not amongst the most intense in terms of radar reflectivity (50-60 dBZ) and their echo tops rarely exceeded 12 km, but they exhibited extremely strong radial velocities and rotational shear (up to –5.0x10-2s-1), which initiated a cyclonic vortex in the center of the cells, spawning the tornadoes. The 40 dBZ cell tracks of C1 and T1 are shown in Figure 5a, while the Time-History of the severe storm parameters of C1 can be seen in Figure 5b. The supercell storm (C1) had almost identical characteristics as its tornadic partner cell (T1), except for a Weak-Echo- Region. The tornado-spawning cell lasted >5 hours.

(a) (b)

FIG. 5. 25 May 2004: (a) Composite radar image, showing the tracks of 40 dBZ centroids of supercell C1 and tornadic cell T1 (Palmital). Times of first and last detection in LT. Not all simultaneous tracks are shown. Red centroids indicate the confirmed tornado. (b) Storm Time History of supercell C1 from 10:45 – 19:15 LT.

The supercell revealed much greater severity parameters (VIL=70,6 kg.m-2, MAX-Z=60 dBZ, VOL = 500 to >1000km3 sustained for four hours; Figure 5b), than the tornadic cells, but no reports of damage or the formation of a tornado were received. A second tornado was observed at around 17:00, about 40 km south-east of Bauru, being spawned from a storm which had initially developed at 15:00 about 105 km west-north-west of the BRU radar, ERAD 2010 - THE SIXTH EUROPEAN CONFERENCE ON RADAR IN METEOROLOGY AND HYDROLOGY moving south-eastwards at speeds of 55–65 km.h-1. TITAN identified its 40 dBZ centroid at 15:38 and tracked it for 3,5 hours (Held et al., 2006a). 3.5 24 May 2005 The synoptic situation on this day was very similar to the case exactly one year ago ((Held et al., 2006a). Widespread pre-frontal rain occurred during most of this day in the southern half of the State of São Paulo. At least one isolated cell embedded in the general precipitation, spawned a tornado (T3), killing one person in Capivari and causing great damage in , while another cell with cyclonic convergence (C2) created an unusually strong and damaging windstorm in the town of Iaras. Their tracks are shown in Figure 6a. The tornadic cell T3 had VIL values of 5-12 kg.m-2 during its first half of life time, but during the second half it developed a double peak of VIL, the first one after a sharp increase to 40,5 kg.m-2 about 15 min prior to the first touch-down at 17:00 (Figure 6b). During the next 20 min it dropped to 11,4 kg.m-2, but increased again to 20,3 kg.m-2, dropping to 5 kg.m-2 as the tornado faded out. The maximum reflectivity was around 55 dBZ during the first 1,5 hours, increasing to 57,5 dBZ with the 40 dBZ reaching up to 10,6 km shortly before 17:00. Thereafter, Max Z dropped together with the other parameters to 49,4 dBZ and increased during the second peak to 54,5 dBZ (40 dBZ at 10 km). The volume of the 40 dBZ centroid fluctuated between 90 and 150 km3 during the first part of the track, but increased to 500 - 700 km3 during the peak period (Figure 6b).

(a) (b)

FIG. 6. 24 May 2005: (a) Composite radar image, showing the tracks of 40 dBZ centroids of supercell C2 and tornadic cell T3. Times of first and last detection in LT. Red centroids indicate the confirmed tornado activity of cell T3 and the severe windstorm associated with cell C2. (b) Storm Time History of tornadic cell T3 during the second half of its life time (16:08 - 18:00 LT).

In terms of storm statistics, cell C2 (Figure 6a) was very much average. During the windstorm at Iaras the VIL was <5 kg.m-2 and only increased to 15,7 kg.m-2 towards the end of its life time. Maximum reflectivities were around 50 dBZ, but the height of the 40 dBZ contour rose to 7,0 - 8,5 km only during the second half. 3.6 24 July 2007 An intense formation of instability areas, associated with frontal activity and the presence of troughs in the middle and higher levels of the atmosphere, was observed between 24 and 25 of July 2007, which was responsible for the severe episodes that have produced hailstorms across the northern areas of the State of São Paulo. Extreme values for rainfall were recorded in different areas of the State on these particular days, exceeding their expected monthly value. Amongst the scattered storms persisting on this day, at least one storm with supercell characteristics was identified by TITAN, using a threshold of 40 dBZ (Figure 7a), which caused extensive damage to orange plantations and resulted in a loss of 30% of the production, with direct implications for the coming year’s harvest. The SSS index (Storm Structure Severity; Visser, 2001) has shown signatures typical of an extreme severe episode, where the structures for severe base, severe top and severe volume were observed during most of its life span (Figure 7b). It lasted for 5 hours, starting at midday in the west of the state, moving at 50 – 60 km.h-1 and producing an intermittent hail swath.

ERAD 2010 - THE SIXTH EUROPEAN CONFERENCE ON RADAR IN METEOROLOGY AND HYDROLOGY

(a) (b)

FIG. 7. (a) Storm track (light-green) showing cells, identified by a 40 dBZ threshold, that have produced severe hail during the 24 July 2007 event. (b) Distribution of SSS (Storm Structure Severity; Visser, 2001) index at 15:05 LT.

3.7 Lightning Analysis

Analysis of lightning records, superimposed on radar images indicated a preferential location of CGs around or ahead of the core of tornadic cells, while in the supercells the CGs were observed within and around the core and with greater frequency. Lightning activity almost ceased shortly before the touch-down of the tornadoes (Figure 8), which is in agreement with observations of tornadoes and supercell storms in Oklahoma (Rison et al., 2005). No significant differences of lightning parameters (peak current, multiplicity, polarity) were found for the tornadic and non-tornadic cells. However, flash polarity seems to be a good discriminator between mature convective cells and stratiform rain regions, with the latter producing only few and mostly positive flashes, while even slowly decaying convective cells may still generate large numbers of flashes, but an increasing portion is positive.

(a) (b)

FIG. 8. Position of CG flashes (+ negative; + positive) relative to the echo cores of the supercells during a 7,5 min interval. (a) 25 May 2004: supercell C1 and tornadic storm T1; (b) 24 May 2005: supercell C2 and tornadic storm T3.

4. Conclusion

The analysis of the six case studies revealed, that the tornadic cells, as well as other supercell storms, had maximum radar reflectivities of 50-60 dBZ in the core of the cells during most of their life cycle, moved very rapidly (>50 km.h-1) and were characterized by strong rotational shear of up to –5,0x10-2s-1. Deploying NCAR´s TITAN (Thunderstorm Identification, Tracking, Analysis and Nowcasting) Software (threshold 40 dBZ, minimum volume 16 km3), much greater severity parameters (VIL=70,6 kg.m-2, MAX-Z=60 dBZ, sustained VOL= 500 to >1000km3 for 4 hours) were yielded for the supercell in May 2004, than for any of the tornadic cells. The temporal evolution of VIL values showed a rapid decrease close to the time of the observed destructive winds at ground level (e.g., tornado touch-down), while the highest values of VIL were not necessarily observed close to that time, but generally at a later stage of the cell. Based on observations of cloud-to-ground (CG) flashes from the Brazilian Lightning Detection Network (BrasilDat), the intensive lightning activity of the cells prior to the tornado touch-down, dropped to very low stroke frequencies during the mature tornado stage, after which it increased again. In contrast, the supercell storms, which did not spawn a tornado, had relatively constant flash rates. Considering the occurrence of confirmed tornadoes and supercells during the past 18 years, one could conclude, that there is a prevalence of synoptic situations favoring the development of severe tornadoes in this region during ERAD 2010 - THE SIXTH EUROPEAN CONFERENCE ON RADAR IN METEOROLOGY AND HYDROLOGY the austral autumn, despite the storms being less intense in terms of echo tops and radar reflectivity, than during summer.

References

Dixon, M. and Wiener G., 1993: TITAN: Thunderstorm Identification, Tracking, Analysis & Nowcasting - A radar- based methodology, J. Atmos. Oceanic Technol., 10, p.785-797. Foote G.B., Krauss T.W. and Makitov V., 2005. Hail metrics using conventional radar. Proceedings, 16th Conference on Planned and Inadvertent Weather Modification, January 2005, AMS, Boston. In: www.ams.confex.com/ams/pdfpapers/86773.pdf Gomes A.M., Held G, Lima M.A. e Calheiros R.V., 2000: Estudo da Evolução de Tempestades Severas na Área Central do Estado de São Paulo por Radar Doppler. Anais, XI Congresso Brasileiro de Meteorologia, SBMET, , 16-20 Outubro 2000, Paper MR00006, p.1921-1929. Held G., Gomes, A.M. e Nachtigall L.F., 2001: Severe Storm Signatures Observed by Doppler Radar in the State of São Paulo, Brazil: From Synoptic Scale to Microburst. Preprints, 30th International Conference on Radar Meteorology, Munich, Germany, 19-24 July 2001, p.248-250. Held G. and Nachtigall L.F. 2002: Flood-producing Storms in Bauru during February 2001. Proceedings, XII Congresso Brasileiro de Meteorologia (CD ROM), SBMET, Foz de Iguaçu, 4-9 August 2002, Paper SP5_35, p.3155-3163. Held G., Gomes A.M., Naccarato K.P., Pinto O. Jr., Nascimento E. de L., Correia A.A. and Marcelino I.P.V.O., 2005: Analysis of Tornado Characteristics in the State of São Paulo for the Improvement of an Automatic Alert System (Paper P3R.9). Preprints, 32nd Conference on Radar Meteorology, CD, AMS, Albuquerque, USA, 24-29 October 2005, 10pp. Held G., Gomes A.M, Naccarato K.P., Pinto Jr. O, and Nascimento E., 2006a: The Structure of Three Tornado- Generating Storms Based on Doppler Radar and Lightning Observations in the State of São Paulo, Brazil. Proceedings, 8th International Conference on Southern Hemisphere Meteorology and Oceanography, Foz do Iguaçu, 24-28 April 2006, p.1787-1797. Held G., Gomes A.M., Naccarato K.P. and Pinto O. Jr., 2006b: Analysis of severe thunderstorms in the State of São Paulo, Brazil, using “TITAN” to identify the position of positive and negative ground strokes relative to radar echoes. Proceedings, International Conference on Grounding and Earthing (GROUND’2006) & 2nd International Conference on Lightning Physics and Effects (2nd LPE), Maceió, AL, 26-29 November 2006, p.535-540. Nielsen-Gammon J.W. and Read E.W.L., 1995: Detection and interpretation of left-moving severe thunderstorms using the WSR-88D: A case study. Weather and Forecasting, 10, p.127-140. Pinto Jr., O., 2003: The Brazilian lightning detection network: a historical background and future perspectives. Proceedings, VII International Symposium on Lightning Protection, Curitiba, Brazil, 17-21 November 2003, p.3-5. Rison W., Krehbiel P., Thomas R.J. and Hamlin T., 2005: Three-dimensional lightning mapping observations as a compliment to radar observations of storms, Preprints, 32nd Conference on Radar Meteorology, Poster JP3J.23, AMS, Albuquerque, USA, 24-29 October 2005. Silva Dias M.A., Carvalho L.M.V. and Gomes A.M., 1996: Um caso de tempestade em ambiente ciclônico em grande e mesoescala. Proceedings, IX Congresso Brasileiro de Meteorologia, SBMET, Campos do Jordão, SP, Novembro de 1996, Vol. 2, p.1316-1320. Visser P.J.M., 2001: The Storm-Structure-Severity method for the identification of convective storm characteristics with conventional weather radar. Meteorological Applications, 8, p.1-10 (doi:10.1017/S1350482701001013). Waldvogel A., Federer B. and Grimm P., 1979: Criteria for the Detection of Hail Cells, J. Appl. Meteor., 18, p.1521- 1525.