Citation
Characterization of Florida Lightning with Emphasis on the Preliminary Breakdown Process, Bipolar Lightning, and Lightning Interaction with the 257-m Tower

Material Information

Title:
Characterization of Florida Lightning with Emphasis on the Preliminary Breakdown Process, Bipolar Lightning, and Lightning Interaction with the 257-m Tower
Creator:
Zhu, Yanan
Place of Publication:
[Gainesville, Fla.]
Florida
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (439 p.)

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Electrical and Computer Engineering
Committee Chair:
RAKOV,VLADIMIR ALEK SANDROVICH
Committee Co-Chair:
UMAN,MARTIN A
Committee Members:
MOORE,ROBERT C
HAGER,WILLIAM WARD
JORDAN,DOUGLAS M

Subjects

Subjects / Keywords:
breakdown -- lightning -- physics
Electrical and Computer Engineering -- Dissertations, Academic -- UF
Genre:
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Electrical and Computer Engineering thesis, Ph.D.

Notes

Abstract:
Various characteristics of Florida lightning were examined and the results were found to be consistent with previous studies. An automated data processing algorithm was developed for studying preliminary breakdown (PB) pulse trains in negative cloud-to-ground (CG) lightning. For more than 3000 high-intensity (> 50 kA first-stroke peak current) flashes, the time interval between PB and return stroke (RS) was found to decrease with increasing RS peak current. The largest PB pulse peak exhibited positive correlation with the RS peak current. It appears that the high-intensity negative CG is characterized by shorter (and by inference faster) stepped leaders and more pronounced PB pulse trains. A highly unusual bipolar cloud-to-ground lightning flash was observed to start with a negative stroke followed by a positive stroke and bipolar continuing current, all in the same channel. This is the first observation of positive stroke in the channel of preceding negative stroke after not unduly-long (70 ms) time interval. The 2-D speed profile, for the positive leader in the previously-conditioned channel was examined for the first time. Electric field and high-speed video records of two flashes that contained a total of 8 strokes terminated on a 257-m tower were obtained. All these strokes exhibited very similar and unusually narrow bipolar electric field waveforms with damped oscillatory tails. By using an engineering model of lightning striking a tall object, the measured electric field waveforms were reproduced using as the channel-base current a narrow pulse followed by a steady-current tail. The effects of the input parameters on model-predicted fields were examined. The National Lightning Detection Network (NLDN) detection efficiency and classification accuracy for cloud discharge activity were evaluated (for the first time) using optical and electric field data acquired at LOG. A similar evaluation was also performed for natural CG flashes and the results were compared with previous evaluations based on rocket-triggered lightning data. The performance characteristics of the Earth Network Total Lightning Network (ENTLN) were evaluated by using as ground-truth both natural lightning data (for the first time) recorded at LOG and rocket-triggered lightning data obtained at Camp Blanding. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis:
Thesis (Ph.D.)--University of Florida, 2017.
Local:
Adviser: RAKOV,VLADIMIR ALEK SANDROVICH.
Local:
Co-adviser: UMAN,MARTIN A.
Electronic Access:
RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2018-06-30
Statement of Responsibility:
by Yanan Zhu.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Embargo Date:
6/30/2018
Classification:
LD1780 2017 ( lcc )

Downloads

This item has the following downloads:


Full Text

PAGE 1

CHARACTERIZATION OF FLORIDA LIGHTNING WITH EMPHASIS ON THE PRELIMINARY BREAKDOWN PROCESS BIPOLAR LIGHTNING, AND LIGHTNING INTER A C TION WITH A 257 M TOWER By Y ANAN ZHU A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2017

PAGE 2

2017 Yanan Zhu

PAGE 3

To my mother, Ping Liu, and my father, Ruiming Zhu

PAGE 4

4 ACKNOWLEDGMENTS I would like to thank Dr. Vladimir Rakov for giving me the opportunity to do lightning research in the Lightning Lab at the University of Florida. I appreciate his inspiration and patient guidance in my research work. I would like to thank Dr. Martin Uman, Dr. Robert Moore, Dr. Douglas Jordan, and Dr. William Hager for serving on my supervisory committee. I would like to thank Shreeharsh Mallick for making great contributions to upgrade the Lightning Observatory in Gainesville (LOG) and the Golf Course site (GC). I would like to thank Manh Tran for operating the LOG and providing the LOG data for my research work. I would like to thank Vijaya Somu, Keith Rambo, and Sam Brake for helping with maintaining the GC site. I would like to thank Dr. Weitao Lu for providing optical data and useful comments for my research. I would like to thank Shuji Fujimaru and Daniel Kotovsky who are always willing to help and give suggestions. I would like to thank John Pilkey for providing LMA data for the fast st epped leader events. I also thank Terry Ngin, John Pilkey, William Gamerota, Jamie Caicedo, Daniel Kotovksy, Felipe Carvalho, Robert Wilkes, Brian Hare Dr. Jordan, and Dr. Uman for acquiring and providing the rocket triggered lightning data for the ENTLN performance evaluation study. I would like to thank Amitabh Nag and William Brooks of Vaisala Inc. for providing NLDN data. I would also like to thank Christopher Sloop, Stan Heckman, Charlie Liu, and Michael Stock of Earth Networks for providing ENTLN dat a and other support needed to complete the ENTLN performance evaluation study I would also extend my deepest thanks to my beautiful fiance and my wonderful parents for their forever supports and encouragement.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ ............... 4 LIST OF TABLES ................................ ................................ ................................ ........................... 8 LIST OF FIGURES ................................ ................................ ................................ ....................... 10 ABSTRACT ................................ ................................ ................................ ................................ ... 24 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .................. 26 1.1 Cloud to Ground Lightning ................................ ................................ .......................... 26 1.2 Rocket Triggered Lightning ................................ ................................ ......................... 28 1.3 Preli minary Breakdown ................................ ................................ ................................ 30 2 RESEARCH FACILITIES ................................ ................................ ................................ ..... 33 2.1 Lightning Observatory in Gainesville (LOG) ................................ ............................... 33 2.2 Golf Course Site in Starke (GC) ................................ ................................ ................... 36 2.3 Camp Blanding Facility (CB) ................................ ................................ ....................... 37 3 CHARACTERIZATION OF FLORID A NEGATIVE CLOUD TO GROUND LIGHTNING ................................ ................................ ................................ .......................... 40 3.1 Literature Review ................................ ................................ ................................ .......... 40 3.2 Data ................................ ................................ ................................ ............................... 41 3.3 Results ................................ ................................ ................................ ........................... 42 3.3.1 Multiplicity and Percentage of Single Stroke Flashes ................................ ...... 42 3.3.2 Interstroke Interval and Flash Durati on ................................ ............................ 43 3.3.3 First to Subsequent Return Stroke Field Peak Ratio ................................ ......... 46 3.3.4 Variation of Lightning Parameters from One Storm to Anothe r ...................... 48 3.4 Summary ................................ ................................ ................................ ....................... 49 4 ELECTRIC FIELD SIGNATURES OF PRELIMINARY BREAKDOWN AND AUTOMATED ANALYSIS FOR HIGH INTENSITY EVENTS ................................ ........ 51 4.1 Literature Review ................................ ................................ ................................ .......... 51 4.2 Factors Affecting Detectability of PB Pulse Trains ................................ ...................... 54 4.2.1 Signal/Noise Ratio of Recording System ................................ .......................... 54 4.2.2 Type of Storm ................................ ................................ ................................ ... 55 4.2.3 Prospective Return Stroke Peak Current ................................ .......................... 56 4.2.4 Distance from the Strike Point Location to the Observation Point .................. 59

PAGE 6

6 4.3 Automated Analysis of Preliminary Breakdown and Retu rn Stroke Processes in High Intensity Negative Lightning Discharges ................................ ............................ 59 4.3.1 Data ................................ ................................ ................................ ................... 59 4.3.2 Methodology ................................ ................................ ................................ ..... 60 4.3.3 Characteristics of PB Pulse Trains in High Intensity Negative Flashes ........... 66 4.3.4 Correlation between Parameters of Preliminary Breakdown and Return St roke Processes ................................ ................................ ................................ 69 4.4 Summary ................................ ................................ ................................ ....................... 77 5 A SUBSEQUENT POSITIVE STROKE DEVELOPING IN THE CHANNEL OF PRECEEDING NEGATIVE STROKE AND CONTAI NING BIPOLAR CONTINUING CURRENT ................................ ................................ ................................ .... 79 5.1 Literature Review ................................ ................................ ................................ .......... 79 5.2 Instrumentation ................................ ................................ ................................ ............. 80 5.3 Observations and Analysis ................................ ................................ ............................ 81 5.3.1 General Description ................................ ................................ .......................... 81 5.3.2 Characteristics of the Negative and Positive L eaders ................................ ....... 87 5.4 Discussion ................................ ................................ ................................ ..................... 89 5.5 Summary ................................ ................................ ................................ ....................... 92 6 OPTICAL AND ELECTRIC F IELD SIGNATURES OF LIGHTNING INTERACTION WITH THE 257 M TOWER IN FLORIDA ................................ ............... 94 6.1 Literature Review ................................ ................................ ................................ .......... 94 6.2 Observations and Analysis ................................ ................................ ............................ 99 6.2.1 General Description of two Flashes Terminated on the 257 m and 60 m Towers in Florida ................................ ................................ ............................ 101 6.2.2 NLDN Responses to the 2 57 m and 60 m Tower Strokes .............................. 103 6.2.3 High Speed Camera and Electric Field Data for Eight Negative Strokes of Flashes 1593 and 1594 ................................ ................................ .................... 104 6.3 Summary ................................ ................................ ................................ ..................... 108 7 MODELING OF LIGHTNING INTERACTION WITH THE TOWER ............................. 111 7.1 Literature Review ................................ ................................ ................................ ........ 111 7.2 Model Description ................................ ................................ ................................ ....... 114 7.3 Sensitivity Analysis ................................ ................................ ................................ ..... 117 7.4 Modeling of Lightning Events Termin ated on the 257 m Tower ............................... 124 7.4 Summary ................................ ................................ ................................ ..................... 130 8 NATIONAL LIGHTNING DETECTION NETWORK RESPONSES TO NATURAL LIGHTNING BASED ON GROUND TRUTH DATA ACQUIRED AT LOG ................. 132 8.1 Literature Review ................................ ................................ ................................ ........ 132 8.2 Data and Methodology ................................ ................................ ................................ 134 8.3 Analysis and Discussion ................................ ................................ ............................. 143 8.3.1 Detection Efficiency and Classification Accuracy of IC Events .................... 143

PAGE 7

7 8.3.2 D etection Efficiency and Classification Accuracy of CG Strokes .................. 145 8.4 Summary ................................ ................................ ................................ ..................... 146 9 EVALUATION OF ENTLN PERFORMANCE CHARACTERISTIC S BASED ON THE NATURAL AND ROCKET TRIGGERED LIGHTNING DATA ACQUIRED IN FLORIDA ................................ ................................ ................................ ............................. 149 9.1 Literature Review ................................ ................................ ................................ ........ 149 9.2 Data and Methodol ogy ................................ ................................ ................................ 151 9.3 Analysis and Discussion ................................ ................................ ............................. 155 9.3.1 Natural Lightning ................................ ................................ ............................ 155 9.3.2 Rocket Triggered Lightning ................................ ................................ ........... 157 9.4 Summary ................................ ................................ ................................ ..................... 162 10 SUMMARY OF RESULTS AND RECOMMENDATIONS FOR FUTURE RESEARCH ................................ ................................ ................................ ......................... 164 10.1 Summary of Results ................................ ................................ ................................ .... 164 10.2 Recommendations for Future Research ................................ ................................ ...... 167 APPENDIX A TWO STATION MEASUREMENTS OF ROCKET TRIGGERED LIGHTNING ELECTRIC FIELD WAVEFORMS (2013 2016) ................................ ............................... 169 B TWO STATION MEASUREMENTS OF ELECTRIC FIELD WAVEFORMS OF NATURAL NEGATIVE CLOUD TO GROUND LIGHTNING ................................ ....... 351 C TWO STATION MEASUREMENTS OF ELECTRIC FIELD WAVEFORMS OF NATURAL POSITIVE CLOUD TO GROUND LIGHTNING ................................ .......... 395 D HIGH SPE ED VIDEO RECORDS OF THE TWO FLASHES TERMINATED ON THE 257 M TOWER ................................ ................................ ................................ ........... 417 LIST OF REFERENCES ................................ ................................ ................................ ............. 429 BIOGRAPHICAL SKETCH ................................ ................................ ................................ ....... 439

PAGE 8

8 LIST OF TABLES Table page 3 1 Summary of multiplicity of negative flashes and percentage of single stroke flashes in different regions ................................ ................................ ................................ ............. 44 3 2 Summary of geometric mean interstroke intervals and flash durations ............................. 46 3 3 Summary of statistics on the ratio of the first to subsequent return strok e field peaks ..... 47 3 4 Variation of lightning parameters from one storm to another ................................ ........... 50 4 1 Percentages of flashes with detectable PB pulses before and after filtering for 17 storms ................................ ................................ ................................ ................................ 55 4 2 Characterization of PB pulses in 3077 negative flashes each containing a single PB pulse train. ................................ ................................ ................................ .......................... 65 4 3 Summary of short PB RS interval events ................................ ................................ .......... 70 5 1 Comparison of bipolar flashes with a positive stroke following the negative stroke channel ................................ ................................ ................................ ............................... 92 6 1 NLDN data on 8 negative strokes in flashes 1593 and 1594 terminated on the 257 m tower. ................................ ................................ ................................ ............................... 102 6 2 NLDN data on the first, positive stroke of bipolar flas h 1594, which terminated on the 60 m tower, located 3.6 km from the 257 m tower. ................................ .................. 102 6 3 Electric field waveform parameters for the seven negative strokes terminated on the 257 m tower. ................................ ................................ ................................ .................... 107 6 4 Characteristics of narrow bipolar electric field waveforms observed in different studies ................................ ................................ ................................ .............................. 110 7 1 Current equations for transmissi on line type return stroke models ................................ 112 7 2 Model input parameters used for computing electric field waveforms shown in Figure 7 11 ................................ ................................ ................................ ....................... 126 7 3 Model input parameters used for computing electric field waveforms shown in Figure 7 13 and NLDN reported peak current for the 7 strokes ................................ ..... 128 8 1 Summary of the Ground Truth Dataset for IC Events ................................ ..................... 137 8 2 Summary of the Ground Truth Dataset for CG Strokes ................................ .................. 141

PAGE 9

9 8 3 Summary of the NLDN Detection Efficiency (DE) and Cla ssification Accuracy (CA) for IC Events ................................ ................................ ................................ .................... 144 8 4 Summary of the NLDN Detection Efficiency (DE) and Classification Accuracy (CA) for CG Strokes ................................ ................................ ................................ ................. 145 8 5 NLDN DE and CA for CG strokes obtained in different studies ................................ .... 148 9 1 Summary of ground truth datasets for natural and rocket triggered lightning acquired in Florida and used in this study ................................ ................................ ...................... 152 9 2 Summary of the ENTLN performance characteristics evaluated using natural lightning data ................................ ................................ ................................ ................... 156 9 3 Summary of t he estimated values of ENTLN stroke DE and CA for different types of strokes in natural lightning ................................ ................................ ............................... 156 9 4 Summary of the ENTLN performance characteristics evaluated using rocket triggered light ning data ................................ ................................ ................................ .... 158 9 5 Comparison of ENTLN performance characteristics evaluated for four different processors using rocket triggered lightning data ................................ ............................. 162 A 1 Inventory of two station (LOG GC) waveforms of rocket triggered flashes from 2013 to 2016 ................................ ................................ ................................ .................... 169 B 1 Inventory of two station (LOG GC) field measurements for 10 natural negat ive cloud to ground lightning ................................ ................................ ................................ 351 C 1 Inventory of two station (LOG GC) field measurements for 10 natural positive cloud to ground lightning ................................ ................................ ................................ 395

PAGE 10

10 LIST OF FIGUR ES Figure page 1 1 Four types of cloud to ground lightning. ................................ ................................ ........... 26 1 2 A nearly complete sequence of processes of a typical downward negative flash. ............ 28 1 3 Sequence of events in classical triggered lightning. ................................ .......................... 29 2 1 Overview of Lightning Observatory in Gainesville (LOG), Florida. ................................ 33 2 2 Photographs of Lightning Observatory in Gainesville. ................................ ..................... 34 2 3 Three station and two station trigg er schemes. ................................ ................................ 36 2 4 Overview of the Golf Course (GC) site. ................................ ................................ ............ 37 2 5 Overview of the ICLRT. Buildings and measurement stations are lab eled. ...................... 38 3 1 Histogram of the number of strokes per flash (multiplicity). ................................ ............ 43 3 2 Histogram of the interstroke interval. ................................ ................................ ................ 45 3 3 Histogram of the flash duration ................................ ................................ ........................ 45 3 4 Histogram of the ratio of the first to subsequent return stroke field peaks ........................ 47 3 5 Normalized electric field peaks for strokes of different order. ................................ .......... 48 4 1 Schematic representation of four types of lightning (left) and the ex pected electric field waveforms (right). ................................ ................................ ................................ ..... 51 4 2 Locations of 478 flashes (first strokes only) recorded at LOG and reported by the NLDN. ................................ ................................ ................................ ............................... 53 4 3 Comparison of the electric field waveforms before and after filtering. ............................. 54 4 4 Percentage of flashes with detectable PB pulse trains versus peak current. ...................... 57 4 5 Percentage of flashes with detectable PB pulse trains versus distance. ............................. 58 4 6 Flowchart of the automated data processing algorithm. ................................ .................... 61 4 7 An example of the output figure produced by the auto mated data processing algorithm ................................ ................................ ................................ ............................ 64 4 8 NLDN reported locations of the 5498 negative first strokes within 50 to 500 km of LOG. ................................ ................................ ................................ ................................ .. 65

PAGE 11

11 4 9 Histogram of NLDN reported peak currents for the 3496 negative first strokes in flashes with detectable PB pulse trains. ................................ ................................ ............. 66 4 10 Histogram of PB/RS pulse peak ratio for the 3077 flashes. ................................ .............. 67 4 11 Histogram of PB RS interval for the 3077 flashes. ................................ ........................... 68 4 12 Histogram of PB pulse train duration for the 3077 flashes. ................................ ............... 68 4 13 Histogram of bipolar pulse width in the 3077 flashes. ................................ ...................... 69 4 14 PB RS interval versus NLDN reported RS peak current for 3077 negative first strokes. ................................ ................................ ................................ ............................... 71 4 15 PB RS interval versus NLDN reported RS peak current for 3363 neg ative first strokes. ................................ ................................ ................................ ............................... 72 4 16 Peak of the largest PB pulse normalized to 100 km versus NLDN reported RS peak current for the 3077 negative first strokes. ................................ ................................ ........ 73 4 17 Peak of the largest PB pulse normalized to 100 km versus NLDN reported RS peak current for the 3363 negative first strokes. ................................ ................................ ........ 74 4 18 Peak of the largest PB pulse n ormalized to 100 km versus PB RS interval for the 3077 negative first strokes. ................................ ................................ ................................ 75 4 19 Peak of the largest PB pulse normalized to 100 km versus PB RS interval for the 3363 negative first stroke s. ................................ ................................ ................................ 76 5 1 Composite Phantom images of the negative first stroke (left panel labeled CG) and the positive second stroke (right panel labeled +CG). ................................ ....................... 83 5 2 Electric field and dE/dt records of the bipolar flash. ................................ ......................... 84 5 3 The frame to frame 2D speeds of the negative stepped leader (L1) and the following positive leader (L2) ver sus height of the leader tip above ground. ................................ .... 88 5 4 Sequence of events that lead to a bipolar flash with the negative second stroke initiated in a decayed branch of positive leader. ................................ ................................ 89 6 1 Current derivative, current, electric field, and magnetic field records of the 2 nd stroke of the flash strking the 553 m CN tower on August 19 th 2015.. ................................ ....... 96 6 2 to ................................ ..... 97 6 3 Current and electric field records of a negative lightning stroke terminated on the 100 m Gaisberg to wer. ................................ ................................ ................................ ....... 98

PAGE 12

12 6 4 Electric field waveforms of a typical LBE recorded by nine stations at distances ranging from 37.5 to 236.1 km in winter in Japan. ................................ ............................ 98 6 5 A subsequent return stroke terminated on the Tokoyo Skytree showing the narrow bipolar signature in electric field records. ................................ ................................ .......... 99 6 6 Simulated electric field waveforms of LB Es at 100 km. ................................ ................. 100 6 7 Locations of strike points reported by the NLDN for 8 negative strokes terminated on the 257 m tower. ................................ ................................ ................................ .............. 103 6 8 First video frames showing the channel of each of the eight negative strokes terminated on the 257 m tower. ................................ ................................ ....................... 105 6 9 Electric field waveforms of 8 negative strokes terminated on the 257 m towe r. ............ 106 6 10 Measurements (definitions) of electric field waveform parameters. ............................... 107 7 1 Measured current waveform (black line) assoc iated with the first return stroke in a downward flash recorded at the Santis tower. ................................ ................................ 114 7 2 Transmission line representation of lightning strike to a tall object. ............................... 116 7 3 Typical current waveform of subsequent stroke (left panel) and the computed electric field (right panel) at a distance of 10 km from the tower. ................................ ............... 119 7 4 Similar to Figure 7 3, but for symmetric Gaussian waveform. ................................ ....... 119 7 5 The asymmetric Gaussian waveforms with different rise times and fall times and their corresponding simulated electric field waveforms. ................................ ................. 121 7 6 Electric field waveforms computed for different reflection coefficients at the tower t g ). ................................ ................................ ......................... 121 7 7 Electric field waveforms computed for different heights of the strike object. ................ 123 7 8 Electric field waveforms computed for different return stroke speeds. .......................... 123 7 9 Electric field waveforms computed for different return stroke models. .......................... 124 7 10 The current waveforms used for computing the electric field wavefo rms shown in Figure 7 11. ................................ ................................ ................................ ...................... 125 7 11 The measured and computed electric field waveforms. The corresponding I sc for each event is shown in Figure 7 10. ................................ ................................ ......................... 126 7 12 The current waveforms (with continuing current components) used for computing the electric field waveforms shown in Figure 7 13. ................................ ........................ 127

PAGE 13

13 7 13 The measured and computed e lectric field waveforms. The corresponding I sc for each event is shown in Figure 7 12. ................................ ................................ ......................... 129 8 1 Example of an isolated IC event. ................................ ................................ ..................... 136 8 2 Examples of PB pulse trains in negative (left panels) and positive (right panels) CG flashes. ................................ ................................ ................................ ............................. 137 8 3 An example of regular pulse burst (RPB) that occurred in the later stage of a K chang e. ................................ ................................ ................................ ............................. 138 8 4 Histogram of durations of 153 IC events. ................................ ................................ ........ 139 8 5 Flow chart used to determine the detection efficiency and classificati on accuracy for IC events. ................................ ................................ ................................ ......................... 140 8 6 An example of ground truth data for a two stroke CG flash. ................................ .......... 142 9 1 Locations of LOG (gray squar e), CB (yellow square) and ENTLN sensors (red circles) in the Florida region. ................................ ................................ ........................... 150 9 2 Histogram of peak currents for 171 return strokes in 18 flashes triggered using the rocket and wire techniq ue at Camp Blanding, Florida. ................................ ................... 153 9 3 Flowchart showing the methodology to determine the detection efficiency and classification accuracy for natural cloud to ground lightning. ................................ ........ 155 9 4 Histograms of absolute (upper panels) and signed (lower panels) peak current estimation errors for old (left panels) and new (right panels) processors. ....................... 159 9 5 Scatterplots of peak current estimated by the ENTLN vs. ground truth peak current measured at CB for old (left panel) and new (right panel) processors. ........................... 160 9 6 Histograms of location error for the old (left panel) and new (right panel) processors. .. 160 9 7 Plots of ENTLN reported locations for old (left panels) and new (right panels) processors. ................................ ................................ ................................ ........................ 161 A 1 Two station electric field waveforms of flash UF 13 31. ................................ ................ 171 A 2 Two station electric field waveforms of the RS1 of flash UF 13 31. .............................. 172 A 3 Two station electric field waveforms of the RS2 of flash UF 13 31. .............................. 173 A 4 Two station electric field waveforms of the RS3 of flas h UF 13 31. .............................. 174 A 5 Two station electric field waveforms of the RS4 of flash UF 13 31. .............................. 175 A 6 Two station electric field waveform s of the RS5 of flash UF 13 31. .............................. 176

PAGE 14

14 A 7 Two station electric field waveforms of the RS6 of flash UF 13 31. .............................. 177 A 8 Two station ele ctric field waveforms of the RS7 of flash UF 13 31. .............................. 178 A 9 Two station electric field waveforms of the RS8 of flash UF 13 31. .............................. 179 A 10 Two station electric field waveforms of the RS9 of flash UF 13 31. .............................. 180 A 11 Two station electric field waveforms of the RS10 of flash UF 13 31. ............................ 181 A 12 Two station electric field waveforms of the RS11 of flash UF 13 31. ............................ 182 A 13 Two station electric field waveforms of flash UF 13 33. ................................ ................ 183 A 14 Two station electric field waveforms of the RS1 of flash UF 13 33. .............................. 184 A 15 Two station electric field waveforms of the RS2 of flash UF 13 33. .............................. 185 A 16 Two station electric field waveforms of the RS3 of flash UF 13 33. .............................. 186 A 17 Two station electric field waveforms of the RS4 of fla sh UF 13 33. .............................. 187 A 18 Two station electric field waveforms of the RS5 of flash UF 13 33. .............................. 188 A 19 Two station electric field wavef orms of the RS6 of flash UF 13 33. .............................. 189 A 20 Two station electric field waveforms of flash UF 13 34. ................................ ................ 190 A 21 Two station electric field waveforms of the RS1 of flash UF 13 34. .............................. 191 A 22 Two station electric field waveforms of the RS2 of flash UF 13 34. .............................. 192 A 23 Two station electric field waveforms of the RS3 of flash UF 13 34. .............................. 193 A 24 Two station electric field waveforms of the RS4 of flash UF 13 34. .............................. 194 A 25 Two station electric field waveforms of flash UF 14 01. ................................ ................ 195 A 26 Two station electric field waveforms of the RS1 of flash UF 14 01. .............................. 196 A 27 Two station electric field waveforms of flash UF 14 05. ................................ ................ 197 A 28 Two station electric field waveforms of the RS1 of flash UF 14 05. .............................. 198 A 30 Two station electric field waveforms of the RS3 of flash UF 14 05. .............................. 200 A 31 Two station electric field waveforms of the RS4 of flash UF 14 05. .............................. 201 A 32 Two station electric field waveforms of the RS5 of flash UF 14 05. .............................. 202

PAGE 15

15 A 33 Two station electric field waveforms of flash UF 1 4 06. ................................ ................ 203 A 34 Two station electric field waveforms of the RS1 of flash UF 14 06. .............................. 204 A 35 Two station electric field waveforms of the RS2 of flash UF 14 06. .............................. 205 A 36 Two station electric field waveforms of flash UF 14 07. ................................ ................ 206 A 37 Two station electric field w aveforms of the RS1 of flash UF 14 07. .............................. 207 A 38 Two station electric field waveforms of the RS2 of flash UF 14 07. .............................. 208 A 39 Two st ation electric field waveforms of the RS3 of flash UF 14 07. .............................. 209 A 40 Two station electric field waveforms of flash UF 14 08. ................................ ................ 210 A 41 Two station electric field waveforms of the RS1 of flash UF 14 08. .............................. 211 A 42 Two station electric field waveforms of the RS2 of flash UF 14 08. .............................. 212 A 43 Two station electric field waveforms of the RS3 of flash UF 14 08. .............................. 213 A 44 Two station electric field waveforms of the RS4 of flash UF 14 08. .............................. 214 A 45 Two station electric field waveforms of flash UF 14 11. ................................ ................ 215 A 46 Two station electric field waveforms of the RS1 of flash UF 14 11. .............................. 216 A 47 Two station electric field waveforms of the RS2 of flash UF 14 11. .............................. 217 A 48 Two station electric field waveforms of the RS3 of flash UF 14 11. .............................. 218 A 49 Two station electric field waveforms of the RS4 of flash UF 14 11. .............................. 219 A 50 Two station electric field waveform s of the RS5 of flash UF 14 11. .............................. 220 A 51 Two station electric field waveforms of the RS6 of flash UF 14 11. .............................. 221 A 52 Two station e lectric field waveforms of the RS7 of flash UF 14 11. .............................. 222 A 53 Two station electric field waveforms of the RS8 of flash UF 14 11. .............................. 223 A 54 Two station electric field waveforms of flash UF 14 12. ................................ ................ 224 A 55 Two station electric field waveforms of the RS1 of flash UF 14 12. .............................. 225 A 56 Two station electric field waveforms of the RS2 of flash UF 14 12. .............................. 226 A 57 Two station electric field waveforms of the RS3 of flash UF 14 12. .............................. 227

PAGE 16

16 A 58 Two station electric field waveforms of the RS4 of flash UF 14 12. .............................. 228 A 59 Two station electric field waveforms of the RS5 of flash UF 14 12. .............................. 229 A 60 Two station electric field waveforms of the RS6 of flash UF 14 12. .............................. 230 A 61 Two station electric field waveforms of the RS7 of fl ash UF 14 12. .............................. 231 A 62 Two station electric field waveforms of flash UF 14 35. ................................ ................ 232 A 63 Two station electric field waveforms of th e RS1 of flash UF 14 35. .............................. 233 A 64 Two station electric field waveforms of the RS2 of flash UF 14 35. .............................. 234 A 65 Two station electric field waveforms of the RS3 of flash UF 14 35. .............................. 235 A 66 Two station electric field waveforms of the RS4 of flash UF 14 35. .............................. 236 A 6 7 Two station electric field waveforms of flash UF 14 36. ................................ ................ 237 A 68 Two station electric field waveforms of the RS1 of flash UF 14 36. .............................. 238 A 69 Two station electric field waveforms of the RS2 of flash UF 14 36. .............................. 239 A 70 Two station electric field waveforms of flash UF 14 43. ................................ ................ 240 A 71 Two station electric field waveforms of the RS1 of flash UF 14 43. .............................. 241 A 72 Two station electric field waveforms of the RS2 of flash UF 14 43. .............................. 242 A 73 Two station electric field waveforms of the RS3 of flash UF 14 43. .............................. 243 A 74 Two station electric field waveforms of the RS4 of flash UF 14 43. .............................. 244 A 75 Two station electric field waveforms of the RS5 of flash UF 14 43. .............................. 245 A 76 Two station electric field waveforms of the RS6 o f flash UF 14 43. .............................. 246 A 77 Two station electric field waveforms of the RS7 of flash UF 14 43. .............................. 247 A 78 Two station electric field waveforms of flash UF 14 51. ................................ ................ 248 A 79 Two station electric field waveforms of the RS1 of flash UF 14 51. .............................. 249 A 80 Two station elec tric field waveforms of the RS2 of flash UF 14 51. .............................. 250 A 81 Two station electric field waveforms of the RS3 of flash UF 14 51. .............................. 251 A 82 Two station electric field waveforms of the RS4 of flash UF 14 51. .............................. 252

PAGE 17

17 A 83 Two station electric field waveforms of the RS5 of flash UF 14 51. .............................. 253 A 84 Two station electric field waveforms of the RS6 of flash UF 14 51. .............................. 254 A 85 Two station electric field waveforms of the RS7 of flash UF 14 51. .............................. 255 A 86 Two station electric field waveforms of the RS8 of flash UF 14 51. .............................. 256 A 87 Two station electric field waveforms of flash UF 14 52. ................................ ................ 257 A 88 Two station electric field waveforms of the RS1 of flash UF 14 52. .............................. 258 A 89 Two station electric field waveforms of the RS2 of flash UF 14 52. .............................. 259 A 90 Two station electric field waveforms of the RS3 of flash UF 14 52. .............................. 260 A 91 Two station electric field wavefor ms of the RS4 of flash UF 14 52. .............................. 261 A 92 Two station electric field waveforms of the RS5 of flash UF 14 52. .............................. 262 A 93 Two station electric field waveforms of flash UF 14 53. ................................ ................ 263 A 94 Two station electric field waveforms of the RS1 of flash UF 14 53. .............................. 264 A 95 T wo station electric field waveforms of the RS2 of flash UF 14 53. .............................. 265 A 96 Two station electric field waveforms of the RS3 of flash UF 14 53. .............................. 266 A 97 Two station electric field waveforms of the RS4 of flash UF 14 53. .............................. 267 A 98 Two station electric field waveforms of the RS5 of flash UF 14 53. .............................. 268 A 99 Two station electric field waveforms of flash UF 15 11. ................................ ................ 269 A 100 Two station electric field waveforms of the RS1 of flash UF 15 11. .............................. 2 70 A 101 Two station electric field waveforms of the RS2 of flash UF 15 11. .............................. 271 A 102 Two station electric field waveforms of flash UF 15 12. ................................ ................ 272 A 103 Two station electric field waveforms of the RS1 of flash UF 15 12. .............................. 273 A 104 Two station electric field waveforms of the RS2 of flash UF 15 12. .............................. 274 A 105 Two station electric field waveforms of the RS3 of flash UF 15 12. .............................. 275 A 106 Two station electric fi eld waveforms of the RS4 of flash UF 15 12. .............................. 276 A 107 Two station electric field waveforms of the RS5 of flash UF 15 12. .............................. 277

PAGE 18

18 A 108 Two station electric field waveforms of flash UF 15 15. ................................ ................ 278 A 109 Two station electric field waveforms of the RS1 of flash UF 15 15. .............................. 279 A 110 Two station electric field waveforms of the RS2 of flash UF 15 15. .............................. 280 A 111 Two station electric field waveforms of the RS3 of flash UF 15 15. .............................. 281 A 112 Two station electric field waveforms of the RS4 of flash UF 15 15. .............................. 282 A 113 Two station electric field waveforms of the RS5 of flash UF 15 15. .............................. 283 A 114 Two station electric field waveforms of the RS6 of flash UF 15 15. .............................. 284 A 115 Two station electric field waveforms of the RS7 of fla sh UF 15 15. .............................. 285 A 116 Two station electric field waveforms of the RS8 of flash UF 15 15. .............................. 286 A 117 Two station electric field wav eforms of flash UF 15 16. ................................ ................ 287 A 118 Two station electric field waveforms of the RS1 of flash UF 15 16. .............................. 288 A 119 Two station elect ric field waveforms of the RS2 of flash UF 15 16. .............................. 289 A 120 Two station electric field waveforms of the RS3 of flash UF 15 16. .............................. 290 A 121 Two station electric field waveforms of the RS4 of flash UF 15 16. .............................. 291 A 122 Two station electric field waveforms of the RS5 of flash UF 15 16. .............................. 292 A 123 Two station electric field waveforms of the RS6 of flash UF 15 16. .............................. 293 A 124 Two station electric field waveforms of the RS7 of flash UF 15 16. .............................. 294 A 125 Two station electric field waveforms of the RS8 of flash UF 15 16. .............................. 295 A 126 Two station electric field waveforms of the RS9 of flash UF 1 5 16. .............................. 296 A 127 Two station electric field waveforms of the RS10 of flash UF 15 16. ............................ 297 A 128 Two station electric field waveforms of the RS11 of flash UF 15 16. ............................ 298 A 129 Two station electric field waveforms of the RS12 of flash UF 15 16. ............................ 299 A 130 Two statio n electric field waveforms of the RS13 of flash UF 15 16. ............................ 300 A 131 Two station electric field waveforms of the RS14 of flash UF 15 16. ............................ 301 A 132 Two station electric field waveforms of flash UF 15 20. ................................ ................ 302

PAGE 19

19 A 133 Two station electric field waveforms of the RS1 of flash UF 15 20. .............................. 303 A 134 Two station electric field waveforms of the RS2 of flash UF 15 20. .............................. 304 A 135 Two station electric field waveforms of flash UF 15 25. ................................ ................ 305 A 136 Two station electric field waveforms of the RS1 of flash UF 15 25. .............................. 306 A 137 Two station electric field waveforms of the RS2 of flash UF 15 25. .............................. 307 A 138 Two station electric field waveforms of the RS3 of flash UF 15 25. .............................. 308 A 139 Two station electric field waveforms of the RS4 of flash UF 15 25. .............................. 309 A 140 Two station electric field waveforms of flash UF 15 26. ................................ ................ 310 A 141 Two station electric field waveforms of the RS1 of flash UF 15 26. .............................. 311 A 142 Two station electric field waveforms of the RS2 of flash UF 15 26. .............................. 312 A 143 Two station e lectric field waveforms of the RS3 of flash UF 15 26. .............................. 313 A 144 Two station electric field waveforms of the RS4 of flash UF 15 26. .............................. 314 A 145 Two station electric field waveforms of the RS5 of flash UF 15 26. .............................. 315 A 146 Two station electric field waveforms of the RS6 of flash UF 15 26. .............................. 316 A 147 Two station electric field waveforms of the RS7 of flash UF 15 26. .............................. 317 A 148 Two station electric field waveforms of the RS8 of flash UF 15 26. .............................. 318 A 149 Two station electric field waveforms of flash UF 15 38. ................................ ................ 319 A 150 Two station electric field waveforms of the RS1 of flash UF 15 38. .............................. 320 A 151 Two station electric field waveforms of the RS2 of flash UF 15 38. .............................. 321 A 152 Two station electric field waveforms of the RS3 of flash UF 15 38. .............................. 322 A 153 Two station electric field waveforms of the RS4 of flash UF 15 38. .............................. 323 A 154 Two station electric field waveforms of the RS5 of flash UF 15 38. .............................. 324 A 155 Two station electric field waveforms of flash UF 15 39. ................................ ................ 325 A 156 Two sta tion electric field waveforms of the RS1 of flash UF 15 39. .............................. 326 A 157 Two station electric field waveforms of the RS2 of flash UF 15 39. .............................. 327

PAGE 20

20 A 158 Two station electric field waveforms of the RS3 of flash UF 15 39. .............................. 328 A 159 Two station electric field waveforms of the RS4 of flash UF 15 39. .............................. 329 A 160 Two station electric field waveforms of the RS5 of flash UF 15 39. .............................. 330 A 161 Two station electric field waveforms of the RS6 of flash UF 15 39. .............................. 331 A 162 Two station electric field waveforms of flash UF 15 41. ................................ ................ 332 A 163 Two station electric field waveforms of the RS1 of flash UF 15 41. .............................. 333 A 164 Two station electric field waveforms of the RS2 of flash UF 15 41. .............................. 334 A 165 Two station electric field waveforms o f the RS3 of flash UF 15 41. .............................. 335 A 166 Two station electric field waveforms of the RS4 of flash UF 15 41. .............................. 336 A 167 Two station el ectric field waveforms of the RS5 of flash UF 15 41. .............................. 337 A 168 Two station electric field waveforms of flash UF 15 42. ................................ ................ 338 A 169 T wo station electric field waveforms of the RS1 of flash UF 15 42. .............................. 339 A 170 Two station electric field waveforms of the RS2 of flash UF 15 42. .............................. 340 A 171 Two station electric field waveforms of the RS3 of flash UF 15 42. .............................. 341 A 172 Two station electric field waveforms of the RS4 of flash UF 15 42. .............................. 342 A 173 Two station electric field waveforms of the RS5 of flash UF 15 42. .............................. 343 A 174 Two station electric field waveforms of the RS6 of flash UF 15 42. .............................. 344 A 175 Two station electric field waveforms of flash UF 15 43. ................................ ................ 345 A 176 Two station electric field waveforms of the RS1 of flash UF 15 43. .............................. 346 A 177 Two station electric field waveforms of the RS2 of flash UF 15 43. .............................. 347 A 178 Two station electric field wavef orms of the RS3 of flash UF 15 43. .............................. 348 A 179 Two station electric field waveforms of flash UF 16 04. ................................ ................ 349 A 180 Two station electri c field waveforms of the RS1 of flash UF 16 0 ................................ 350 B 1 Two station electric field waveforms of flash 00389. ................................ ..................... 352 B 2 Two station ele ctric field waveforms of RS1 of flash 00389. ................................ ......... 353

PAGE 21

21 B 3 Two station electric field waveforms of RS2 of flash 00389. ................................ ......... 354 B 4 Two stati on electric field waveforms of RS3 of flash 00389. ................................ ......... 355 B 5 Two station electric field waveforms of flash 00390. ................................ ..................... 35 6 B 6 Two statio n electric field waveforms of RS1 of flash 00390. ................................ ......... 357 B 7 Two station electric field waveforms of RS2 of flash 00390. ................................ ......... 358 B 8 Two station electric field waveforms of RS3 of flash 00390. ................................ ......... 359 B 9 Two station electric field waveforms of flash 00467. ................................ ..................... 360 B 10 Two station electric field waveforms of the RS1 of flash 00467. ................................ ... 361 B 12 Two station electric field waveforms of the RS3 of flash 00467. ................................ ... 363 B 13 Two station electric field waveforms of flash 00468. ................................ ..................... 364 B 14 Two station electric field waveforms of the RS1 of flash 00468. ................................ ... 365 B 15 Two station electric field waveforms of the RS2 of flash 00468. ................................ ... 366 B 16 Two station electric field waveforms of the RS3 of flash 00468. ................................ ... 367 B 17 Two station electric field waveforms of the RS4 of flash 00468. ................................ ... 368 B 18 Two station electric field waveforms of flash 00491. ................................ ..................... 369 B 19 Two station electric field waveforms of the RS1 of flash 00491. ................................ ... 370 B 20 Two station electric field waveforms of the RS2 of flash 00491. ................................ ... 371 B 21 Two station electric field waveforms of the RS3 of flash 00491. ................................ ... 372 B 22 Two station electric field waveforms of the RS4 of flash 00491. ................................ ... 373 B 23 Two station electric field waveforms of flash 00532. ................................ ..................... 374 B 24 Two station electric field waveforms of the RS1 of flash 0053 2. ................................ ... 375 B 25 Two station electric field waveforms of the RS2 of flash 00532. ................................ ... 376 B 26 Two station electric field waveforms of the R S3 of flash 00532. ................................ ... 377 B 27 Two station electric field waveforms of the RS4 of flash 00532. ................................ ... 378 B 28 Two station electric field wa veforms of flash 00537. ................................ ..................... 379

PAGE 22

22 B 29 Two station electric field waveforms of the RS1 of flash 00537. ................................ ... 380 B 30 Two station electric fie ld waveforms of the RS2 of flash 00537. ................................ ... 381 B 31 Two station electric field waveforms of flash 00564. ................................ ..................... 382 B 32 Two station electri c field waveforms of the RS1 of flash 00564. ................................ ... 383 B 33 Two station electric field waveforms of the RS2 of flash 00564. ................................ ... 384 B 34 Two station electric field waveforms of the RS3 of flash 00564. ................................ ... 385 B 35 Two station electric field waveforms of the RS4 of flash 00564. ................................ ... 386 B 36 Two station electric field waveforms of the RS5 of flash 00564. ................................ ... 387 B 37 Two station electric field waveforms of flash 00569. ................................ ..................... 388 B 38 Two station electric field waveforms of the RS1 of flash 00569. ................................ ... 389 B 39 Two station electric field waveforms of the RS2 of flash 00569. ................................ ... 390 B 40 Two station electric field waveforms of the RS3 of flash 00569. ................................ ... 391 B 41 Two station electric field waveforms of the RS4 of flash 00569. ................................ ... 392 B 42 Two station electric field waveforms of flash 00620. ................................ ..................... 393 B 43 Two station electric field waveforms of the RS1 of flash 00620 ................................ .... 394 B 44 Two station electric field waveforms of flash 00436. ................................ ..................... 396 B 45 Two station electric field waveforms of the RS1 of flash 00436. ................................ ... 397 B 46 Two station electric field waveforms of flash 00438. ................................ ..................... 398 B 47 Two station electric field waveforms of the RS1 of flash 00438. ................................ ... 399 B 48 Two station electric field waveforms of flash 00453. ................................ ..................... 400 B 49 Two station electric field waveforms of the RS1 of flash 00453. ................................ ... 401 B 50 Two station electric field waveforms of flash 00456. ................................ ..................... 402 B 51 Two station electric field waveforms of the RS1 of flash 00456 ................................ ... 403 B 52 Two station electric field waveforms of flash 00477. ................................ ..................... 404 B 53 Two station electric field waveforms of the RS1 of flash 00477. ................................ ... 405

PAGE 23

23 B 54 Two station electric field waveforms of flash 00577. ................................ ..................... 406 B 55 Two station electric field waveforms of the RS1 of f lash 00577. ................................ ... 407 B 56 Two station electric field waveforms of flash 00580. ................................ ..................... 408 B 57 Two station electric field waveforms of the RS1 of flash 00580. ................................ ... 409 B 58 Two station electric field waveforms of flash 00590. ................................ ..................... 410 B 59 Two station electric field waveforms of th e RS1 of flash 00590. ................................ ... 411 B 60 Two station electric field waveforms of flash 01162. ................................ ..................... 412 B 61 Two station electric field waveforms of the RS1 of flash 01162. ................................ ... 413 B 62 Two station electric field waveforms of the RS2 of flash 01162. ................................ ... 414 B 63 Two station electric field waveforms of flash 01163. ................................ ..................... 415 B 64 Two station electric field waveforms of the RS1 of flash 01163. ................................ ... 416 C 1 Frames showing th e first stroke of flash 1593. ................................ ................................ 418 C 2 First two frames showing the second of flash 1593. ................................ ........................ 419 C 3 Last two frames showing the secon d stroke of flash 1593. ................................ ............. 420 C 4 Frames showing the third stroke of flash 1593. ................................ ............................... 421 C 5 First two frames showing the fourth stroke of flash 1593. ................................ .............. 422 C 6 Last two frames showing the fourth stroke of flash 1593. ................................ ............... 423 C 7 Frames showing the fifth stroke of flash 1 593. ................................ ................................ 424 C 8 First two frames showing the sixth stroke of flash 1593. ................................ ................ 425 C 9 Last two frames showing the sixth stroke of flash 15 93. ................................ ................. 426 C 10 Frames showing the second stroke of flash 1594. ................................ ........................... 427 C 11 Frames showing the third stroke of flash 1594. ................................ ............................... 428

PAGE 24

24 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy CHARACTERIZATION OF FLORIDA LIGHTNING WITH EMPHASIS ON THE PRELIMINARY BREAKDOWN PROCESS BIPOLAR LIGHTNING, AND LIGHTNING INTER A CTION WITH A 257 M TOWER By Yanan Zhu December 2017 Chair: Vladimir A. Rakov Cochair: Martin A. Uman Major: Electrical and Computer Engineering Various characteristi cs of Florida lightning were examined and the results were found to be consistent with previous studies A n auto mated data processing algorithm was developed for studying preliminary breakdown (PB) pulse trains in negative cloud to ground (CG) lightning F or m ore than 3000 high intensity ( 50 kA first stroke peak current ) flashes, t he time interval between PB and return stroke (RS) was found to decrease with increasing RS peak current The largest PB pulse peak exhibited positive correlation with the RS pe ak current. It appears that the high intensity negative CG is characterized by shorter (and by inference faster) stepped leaders and more pronounced PB pulse trains. A highly unusual bipolar cloud to ground lightning flash was observed to start with a nega tive stroke followed by a positive stroke and bipolar continuing current all in the same channel. This is the first observation of positive stroke in the channel of preceding negative stroke after not unduly long (70 ms) time interval. The 2 D speed profi le for the po sitive leader in the previously conditioned channel was examined for the first time

PAGE 25

25 E lectric field and high speed video records of two flashes that contained a total of 8 strokes terminated on a 257 m tower were obtained. All these strokes e xhibited very similar and unusually narrow bipolar electric field waveforms with damped oscillatory tails. By using an engineering model of lightning striking a tall object the measured electric field waveforms were reproduced using as the channel base cu rrent a narrow pulse followed by a steady current tail The effects of the input parameters on model predicted fields were examined. T he National Lightning Detection Network (NLDN) detection efficiency and classification accuracy for cloud discharge activi ty were evaluated (for the first time) using optical and electric f ield data acquired at LOG. A similar evaluation was also performed for natural CG flashes and the results were compared with previous evaluations based on rocket triggered lightning data. T he performance characteristics of the Earth Network Total Lightning Network (ENTLN) were evaluated by using as ground truth both natural lightning data (for the first time) recorded at LOG and rocket triggered lightning data obtained at Camp Blanding.

PAGE 26

26 C HAPTER 1 INTRODUCTION 1.1 Cloud to Ground Lightning Based on the polarity of charge effectively lowered to the ground and the direction of initial leader propagation, four types of lightning can be identified. They are downward negative lightning, upward n egative lightning, downward positive lightning, and upward positive lightning, as illustrated in Figure 1 1. Figure 1 1. Four types of cloud to ground lightning. Adapted from Rakov and Uman [2003]

PAGE 27

27 Downward negat ive lightning accounts for about 90% percent of cloud to ground lightning. A complete sequence of processes involved in a typical downward negative cloud to ground lightning will be introduced next and the corresponding illustration is shown in Figure 1 2 Thunderstorm charge structure usually can be simplified as a tripole model, which contains a main positive charge region, a main negative charge region, and a lower positive charge region. They are labeled P, N, and LP respectively in Figure 1 2. Alth ough not shown in Figure 1 2, usually a negative charge region called upper screening layer is induced at the top of the thundercloud. Preliminary breakdown (PB) process is commonly viewed as the initiation of lightning. More detailed information on prelim inary breakdown process will be given in section 1. 3 Preliminary breakdown process serves to provide the condition for the formation of a stepped leader. The latter is a hot plasma channel which moves downward in a discrete manner (in the case of CG) E ach luminous step typically moves tens of meters in one microsecond or less, then there is a pause for 20 to 50 s, then another step is formed. Negative charges are effectively lowered from the negative charge region and distributed along the path of the stepped leader. When the stepped leader is close to ground, the positively charged upward leader will be initiated from the ground or protruding objects on the ground. When the upward leader makes contact with the branched downward negative leader, the mo st luminous process termed the return stroke starts, during which negative charges deposited on the stepped leader channel flow into the ground, resulting a current wave of tens of kA amplitude that travels from ground up the channel at the one third to on e half of the speed of light. If a flash ceased after the first return stroke, this flash is called a single stroke flash, otherwise a continuously propagating leader (as opposed to stepped leader), known as dart leader, will traverse the downward the firs t stroke

PAGE 28

28 channel and lead to a subsequent return stroke. About 80% negative cloud to ground flashes have multiple (usually 3 5) leader return stroke sequence. Figure 1 2. A nearly complete sequence of processes of a typical downward negative flash. Adap ted from Uman [1987] 1.2 Rocket Triggered Lightning Lightning c an be triggered by launching a small rocket with a thin metallic trailing wire toward th technique and section.

PAGE 29

29 Figure 1 3. Sequence of events in classical triggered lightni ng. Adapted from Rakov et al. [1998] In Figure 1 3, an illustration of six stages in the classical triggering process is shown. A rocket is launched w hen the quasi static electric field measured at the ground level reaches some critical value. Also, it is usually launched during the dissipating stage of the thunderstorm that characterized by a low flash rate, since the occurrence of nearby natural lightning could reduce the electric field needed for a successful triggering. The rocket ascends at the speed of about 150 produced at the wire tip and typically extend a few meters ahead of the ascending rocket [ Biagi et al. 2011, 2012] These precursors abort within several meters since the enhanced electric field at the tip of the wire is in sufficient for sustained propagation. When the rocket reaches the height of 200 300 m, the enhanced electric field near the tip of the wire is strong enough to initiate a positive leader propagates upward toward the cloud at a speed of about 10 5 m/s, as sh own in the

PAGE 30

30 second stage in Figure 1 3. The increasing UPL current will vaporize the triggering wire which will be quickly replaced by a plasma channel The UPL will continue to extend toward the cloud When the UPL reaches the negative charge region, init ial continuous current (ICC) characterized by slowly varying current with duration of several hundreds of milliseconds and average current of the order of 100 A. Usually, the lightning channel is straight below the wire top while the channel above the wir e top is tortuous. The precursor s UPL, and ICC form the initial stage (IS). After the end of the IS stage, the lightning channel decays and the current measured at the channel base falls to nearly zero. One or more downward dart leader/return stroke sequ ences may traverse the remnants of the decayed lightning channel established between the cloud and the triggering facility. The dart leader/return stroke sequences in rocket triggered lightning are similar to the dart leader/return stroke sequences in natu ral lightning. 1.3 Preliminary Breakdown Preliminary breakdown (PB) or initial breakdown process seems to be the most mysterious process of lightning. How lightning (or preliminary breakdown channel) is initiated and processes within the thundercloud that create the high electric field needed for lightning to be initiated remain some of the biggest unsolved problems in lightning physics [ Dwyer and Uman 2013] Clarence and Malan [1957] suggested that PB is a vertical discharge betw een the main negative charge region and the lower positive charge region. However, Krehbiel et al. [1979] from eight station electric field measurements, found that breakdown events before the stepped leader involved considerable horizontal extent. Rakov and Uman [2003] concluded that the PB process can be viewed as a sequence of channels extending in seemingly random directions from the cloud source with one of them evolving i nto the stepped leader to the ground. More recently studies by Stolzenburg et al. [ 2013, 2014] who used high speed video records of preliminary

PAGE 31

31 breakdown, indicated that the luminosity bursts were correlated with the amplitude of the PB pulses and suggested that each PB pulse is caused by a substantial current surge traversing a channel segment that is hundreds of meters long. Wilkes et al. [2016] found that characteristics of luminosity pulses of preliminary (initial) breakdown processes in CG and IC are different and suggested that physics of the initiation process in each case either may inherently be different or may be the same but in different cloud environments. Preliminary breakdown of negative cloud to ground lightning is characterized by a sequence of bipolar pulses typically occurring a few tens of milliseconds before the firs t return stroke with duration on the order of 1 ms and pulse width of 20 40 s [ Rakov et al. 1996] .The amplitude of PB pulse can be comparable to or even exceed the amplitude of the following return stroke (RS) pulse [ Brook 1992; Nag and Rakov 2009a] Nag and Rakov [2008] examined electric field records of negative cloud to ground lightning acquired in Gainesville, FL, i n 2006 and found that only 18% of them had detectable PB pulses. However, Baharudin et al. [2012] and Marshall et al. [2014] found that 100% of flashes in each study had detectable PB pulse train s Marshall et al. [2014] found that the noise level of recording system, record length, and distance to the observation point, can reduce the detectability of PB pulses. The amplitudes of preliminary breakdown pulses were found to be significantly higher at high er latitude (temperate) region s than the ir counterparts at lower latitude (tropical or subtropical) region s [ Gomes et al. 1998; Nag and Rakov 2009b; Marshall et al. 2014] Gomes et al. [1998] and Nag and Rakov [2009b] attributed this difference to a more significant lower positive charge r egion at higher latitude region s Brook [1992] found that PB pulses produced in winter storms were more intense than i n summer storms. Also, the PB RS interval was found to be a factor of 4 shorter for winter storms.

PAGE 32

32 The shortest PB RS interval was 2.5 ms, observed in winter. He attributed the disparity to the difference of precipitation mixes in summer and winter storms. However, Heavner et al. [2002] and [20 14] reported PB RS interval s as short as a few milliseconds for summer storms and the inferred leader speed was on the order of 10 6 m/s, which is an order of magnitude greater than typical value of stepped leader. l. [2014] suggested that unusually strong negative charge sources account for such fast leader s.

PAGE 33

33 CHAPTER 2 RESEARCH FACILIT IES 2.1 Lightning Observatory in Gainesville (LOG) The Lightning Observatory in Gainesville (LOG), Florida is currently located on the roof of the five story New Engineering Building on the campus of the University of Florida, which is 45 km from Camp Blanding. Sensors currently used at LOG include electric field (E) antennas, electric field derivative (dE/dt) antennas, magnetic fiel d derivative (dB/dt) antennas, and an x ray detector. Signals from the sensors are transmitted to digitizing oscilloscopes in the LOG cupola using fiber optic links. Trigger times are GPS time stamped on a computer. Figure 2 1. Overview of Lightning Obse rvatory in Gainesville (LOG), Florida. Adapted from Mallick [2013] and Tran [2015]

PAGE 34

34 Figure 2 2. Photographs of Lightning Observatory in Gainesville a) South side a nd b) North side. Adapted from Mallick [2013] Two high speed ca meras, Phantom V310 and HHC X2, were installed at LOG in 2012 and 2013, respectively. Also, the Total sky Lightning Channel Imager (TLCI) was installed at LOG in 2013 as part of our collaboration with the Chinese Academy of Meteorological Sciences. Additio nally, Lightning Attachment Process Observation System (LAPOS) was installed in LOG in 2016 as part of our collaboration with the Gifu University Japan An overview of LOG is shown in Figure 2 1. Photographs of LOG are shown in Figure 2 2. Two experimen ts single station experiment and multi station experiment are performed at LOG. Single station experiment is designed to study local natural lightning. Single station system can record electromagnetic fields, optical images and x rays/gamma rays produced by natural lightning over Gainesville area. The electric field measuring systems in single station experiment include a low gain and a high gain electric field measuring systems and an electric field derivative (dE/dt) measuring system. The low gain elec tric field measuring system includes a circular flat plate antenna, installed nearly flush with the roof surface, followed by an amplifier with an RC time constant of 10 ms. The high gain electric field measuring system includes a similar but elevated flat plate antenna followed by a different amplifier having a higher gain and a shorter RC time constant o f

PAGE 35

35 to 10 MHz and 360 Hz to 10 MHz for the low gain and high gain systems, respectively. The upper frequency response of the dE/dt measuring system is 10 MHz. Muti station experiment is aimed at recording electromagnetic radiation produced by rocket triggered lightning (RTL) at CB and natural lightning over the CB area. It can be further divided into three station RTL experiment and two station natural lightning exp eriment. When a n RTL or onsite natural lightning occurs at ICLRT, a signal will be sent via a dedicated phone line to trigger the oscilloscope Yokogawa DL 850 at LOG, from where a signal is sent via Internet by an IP addressed digital input and output devi ce (ipIO) to trigger the oscilloscope Yokogawa DL 850 at Golf course site (GC) in Starke, which is a station 3 km west of CB. For the three station experiment, electric field, the electric field derivative (dE/dt) and magnetic field derivative (dB/dt) prod uced by RTL or on site natural lightning at CB were recorded at both LOG and GC from 2013 to 2015. When the electric field produced by natural lightning over CB area exceeds a preset value, the Lecroy 7100A oscilloscope at GC is triggered and a signal will be sent by ipIO to LOG via Interent to trigger the Lecroy 7100A oscilloscope at LOG. For two station experiment, electric field and dE/dt produced by natural lightning over the Starke area were recorded from 2013 to 2014. With the upgrade in the summer of 2015 (the new oscilloscope Lecroy HDO 6034), GC can additionally record magnetic field derivative (dB/dt) produced by natural lightning over the CB area. Three station and two station trigger schemes are shown in Figure 2 3. The multi station electric fie ld measuring system at LOG includes an elevated circular flat plate antenna followed by an integrator/amplifier with a decay time constant of 10 ms. The frequency bandwidth was 16 Hz to 10 MHz. Signals from the antenna were relayed to an oscilloscope via a fiber optical link. The vertical resolution of the Lecroy 7100A oscilloscope

PAGE 36

36 (dedicated for recording natural lightning from 2013 to 2014) was 8 bits and the sampling rate was 50 Msamples/s (sampling interval was 20 ns). The record length was 2 s. For the oscilloscope Yokogawa DL 850, which is dedicated for recording rocket triggered lightning, the vertical resolution was 12 bit and the sampling rate was 100 Msamples/s (sampling interval was 10 ns). The record length was set to be 5 s. Figure 2 3. Thre e station and two station trigger schemes. The ipIO trigger scheme is designed by S. Mallick. 2.2 Golf Course Site in Starke (GC) As one site of multi station experiment, the Golf Course site (GC) at Starke is 3 km west of the ICLRT at Camp Blanding (CB). Two circular flat plate antennas and one square loop antenna were used at GC to measure the electric field, electric field derivative, and magnetic field derivative produced by rocket triggered lightning and onsite natural lightning at CB, and close natura l lightning over the CB area. All the signals from the antennas are transmitted via

PAGE 37

37 fiber optical links to the digitizing oscilloscopes (Yokogawa DL850 and Lecroy 7100A) in the instrumentation truck (Figure 2 4b) All the recorded events are GPS time stamp ed on the computer in the truck. The field measuring system used at GC is very similar to the one used for multi station experiment at LOG except that RC time constant used for electric field measurement at GC is 1.36 s, while it is 10 ms for the one used at LOG. Additionally, after the upgrade in the summer of 2015, the new scope (Lecroy HDO 6034) allows the GC to record two orthogonal magnetic field derivatives for close natural lightning. For the oscilloscope Lecroy HDO 6034, the vertical resolution was 12 bit and the sampling rate was 50 Msamples/s (sampling interval was 20 ns). The record length was set to be 1 s. Figure 2 4 shows the overview and a photograph of GC. In order to avoid flood issue at GC during the thunderstorm season, limerock layer was installed to elevate the ground level by several inches in the summer of 2013. Figure 2 4. Overview of the Golf Course (GC) site. a) Layout of GC site b) A photograph of GC site (2013). Photo courtesy of author 2.3 Camp Blanding Facility (CB) Rocket triggered lightning experiments have been conducted at Camp Blanding (CB), Florida, since 1993. It is also referred to as the International Center for Lightning Research and

PAGE 38

38 Testing (ICLRT) at CB, although it presently also includes LOG and GC The satell ite image of the ICLRT is shown in Figure 2 5. The buildings, measurement stations, and the rocket launcher are labeled. The CB facility occupies an area of roughly 1 km 2 and includes a network of about one hundred measurements of lightning electric/magne tic fields and their derivatives, energetic radiation (x rays and gamma rays). A ddition ally lightning channels are optically imaged using a variety of high speed video and still cameras. The launcher controls and the data acquisition systems are located i nside the launch control trailer. Fiber optic links are used to transmit data from the sensors to the data acquisition systems. An eight station Lightning Mapping Array (LMA) network, a continuously operating system used to locate VHF radiation sources ass ociated with lightning discharges, was installed in 2011 around CB. N atural lightning occurring on or very near the site are also recorded by CB facilities Detailed descriptions of CB facilities can be found in Gamerota [2014] Figure 2 5. Overview of the ICLRT. Buildings and measurement stations are labeled. Taken from Gamerota [2014]

PAGE 39

39 T he lightning channel base current was measured at the ground launcher using a 1 current viewing resistor (CVR) with bandwidth from DC to 8 MHz. The signal from the CVR was transmitted to four separate sets of electronics (channels with different measuring ranges), which could collectively measure currents ranging from 1 mA to 60 kA [ Ngin et al. 2014] The channel base curren ts are recorded on several oscilloscopes, each configured with different record lengths and sample intervals. When the current magnitude exceeded 6 kA, a GPS time stamp was produced, which was used as the ground truth timing of the stroke. When the return stroke current did not exceed the 6 kA threshold, the GPS timing of the stroke was determined from another current measuring system operated at CB by NASA

PAGE 40

40 CHAPTER 3 CHARACTERIZATION OF FLORIDA NEGATIVE CLOUD TO GROUND LIGHTNING 3.1 Literature Review It is well known that lightning flash density and polarity dramatically vary with geographical location and season. However, it is still not certain if similar dependences exist for other lightning parameters. Clearly, lightning parameters can vary from one storm to another [ Biagi et al. 2007; Saraiva et al. 2010] which may influence statistics, particularly in the case of small sample size. Before attributing any variation of lightning parameters to regional or meteorological peculiarities, one should rule out the in fluence of measuring and data processing techniques used in different locations, as well as methodology and limited sample size. Many characteristics of negative cloud to ground lightning in Florida were studied by the University of Florida Lightning Resea rch Group in the early 1990s [e.g., Rako v and Uman 1990; Thottappillil et al. 1992; Rakov et al. 1994] For 76 negative flashes recorded during 3 storms in Florida in 1979, Rakov et al. [1994] examined the number of strokes per flash (multiplicity) and percentage of single stroke flashes. They used electric field and optical data and found that the percentage of single stroke flashes was 17% and the arithmetic mean flash multiplicity was 4.6. Thottappillil et al. [1992] found that 15 (33%) of 46 multiple stroke flashes had one or mor e subsequent return strokes with distance normalized initial electric field peak greater than that of the corresponding first return stroke, and that the interstroke interval preceding these greater than first subsequent strokes were more than 1.7 times lo nger than the average preceding interstroke interval for all subsequent strokes in their dataset.

PAGE 41

41 Nag et al. [2008] examined the ratio of electric field peaks of first and subsequent return strokes b ased on the data acquired in Florida, Austria, Brazil, and Sweden. They found that the electric field peak of first return strokes was on average 1.7 to 2.4 times larger than its counterpart for subsequent return strokes. For 239 negative cloud to ground f lashes in Florida, the arithmetic and geometric means of first to subsequent return stroke field peak ratio were 2.1 and 1.7, respectively. In this chapter, we will examine characteristics of negative cloud to ground lightning flashes using their electric field waveforms acquired at the Lightning Observatory in Gainesville (LOG), Florida in the summers of 2013 and 2014. Flash multiplicity, interstroke interval, flash duration, and first to subsequent stroke field peak ratio are determined for 478 flashes co ntaining 2188 strokes and compared with previous results obtained in Florida. 3.2 Data The dataset of 478 negative cloud to ground flashes in this study was acquired at LOG by using two station (LOG Golf Course site) trigger scheme. The Golf Course site (G C) is located in Starke, about 4 5 km from LOG. When electric field at GC exceeds the preset threshold, the measuring instrumentation at GC is triggered, and a trigger pulse is sent to LOG over the Internet by using an IP addressed digital input and output device. Due to this trigger scheme, the majority of lightning flashes recorded at LOG were relatively close to GC. The distances from LOG to lightning strike points were in the range of 16 330 km and the geometric mean distance was 55 km. Over 73% of event s were in the 20 60 km range.

PAGE 42

42 The more detailed information on the multi station electric field measuring system at LOG can be found in Section 2.2 Pretrigger time (time interval between the beginning of the record and the first RS) was not fixed because of the IP triggering scheme. The pretrigger times were in the range of 46 1879 ms. The average pretrigger time was 612 ms and over 97% of records had >100 ms pretrigger times. Additional information about LOG is found in Rakov et al. [2014] and in Section 2.1 and that about GC in Section 2.2 All the electric field waveforms analyzed in this study were smoothed (filtered) by using a 50 point (1 averag ing window in order to imp rove the signal/noise ratio. It was determined by trial and error that the 50 point window for our data is optimal in that it allows a significant reduction of noise, while keeping the distortion minimal. The reduction of field peaks caused by filtering w as determined to be less than 5%. NLDN data were used to confirm that the first stroke of each flash was not missed by our system due to insufficient pretrigger time (assuming that first strokes are unlikely to be missed by the NLDN). 3.3 Results 3.3.1 Mul tiplicity and Percentage of Single Stroke Flashes Out of 478 flashes containing 2188 strokes recorded during 17 storms, 57 (12%) were single stroke flashes. The average number of strokes per flash was 4.6 and the geometric mean was 3.7. A histogram of mult iplicity is shown in Figure 3 1. In the previous study of Rakov et al. [1994] 76 flashes recorded during 3 storms on average had 4.6 strokes per flash, and the percentage of single stroke flash was 17%. Our results are consistent with the prev ious findings. Information on multiplicity and percentage of single stroke flashes in Florida and in other

PAGE 43

43 regions is summarized in Table 3 1, from which it follows that these two parameters are probably not significantly influenced by location. Figure 3 1. Histogram of the number of strokes per flash (multiplicity). 3.3.2 Interstroke Interval and Flash Duration Figure 3 2 shows a histogram of the interstroke interval. The interstroke intervals were measured between the return stroke field peaks. The ari thmetic mean (AM) and geometric mean (GM) of all the interstroke intervals are 80 ms and 53 ms, respectively. Thottappillil et al. [1992] reported that GM of 199 interstroke intervals (46 flashes) was 57 ms. Rakov et al. [2014] reported that the GM of 270 interstroke intervals (76 flashes) was 60 ms. Our results are comparable with the corresponding values from the previous studies in Florida.

PAGE 44

44 Figure 3 3 shows a histogram of the flash duration f or multiple stroke flashes only. We define here the flash duration as the time interval between the electric field peaks of the first stroke and the last subsequent stroke. The GM duration for 421 multiple stroke flashes is 223 ms, which is close to 216 ms and 229 ms, which are GM flash durations of negative cloud to ground flashes observed in Arizona and Sao Paulo, respectively [ Saraiva et al. 2010] We are not aware of previous flash duration measurements in Florida. Information on interstroke intervals and flash duration s in Florida and in other regions is summarized in Table 3 2, from which no significant variation fr om one region to another is seen. Table 3 1. Summary of multiplicity of negative flashes and percentage of single stroke flashes in different regions Reference and Region Average Number of Strokes per Flash (Multiplicity) Percentage of Single Stroke Flash es Sample Size Kitagawa et al. [1962] New Mexico 6.4 13% 83 Rakov et al. [1994] Florida 4.6 17% 76 Cooray and Jayaratne [1994] Sri Lanka 4.5 21% 81 Cooray and Prez [1994] Sweden 3.4 18% 137 Sa raiva et al. [2010] Arizona 3.9 19% 209 Ballarotti et al. [2012] Brazil 4.6 17% 883 Baharudin et al. [2014] Malaysia 4.0 16% 100 Present Study, Florida 4.6 12% 478

PAGE 45

45 Figure 3 2. Histogram of the interstroke interval. Figure 3 3. Histogram of the flash duration

PAGE 46

46 Table 3 2. Summary of geometric mean interstroke intervals and flash durations (sample sizes are given in the parentheses) 3.3.3 First to Subsequent Return Stroke Field Peak Ratio Figure 3 4 shows a histogram of the ratio of first to subsequent electric field peaks. For 1693 subsequent strokes (e xcluding saturated records), the ratio ranges from 0.3 to 28 with a GM of 2.4 and an AM of 3.1, which are somewhat higher than their counterparts reported for Florida and other regions by Nag et al. [20 08] (Table 3 3). Higher ratios in this study are possibly a result of filtering, which allowed us to detect smaller amplitude strokes (many of them had NLDN reported currents below 10 kA). It also could be a result of the fact that the electric field peak s of subsequent strokes got more attenuation along the path (longer in our study) of propagation since they contain more high frequency component. Out of 421 multiple stroke flashes, 144 (34%) had at least one subsequent stroke whose field peak was greater than that of the first stroke, which is very close to 33% (15 of 46) reported by Thottappillil et al. [1992]

PAGE 47

4 7 Figure 3 4. Histogram of the ratio of the first to subsequent return stroke field peaks Table 3 3. Summary of statistics on the ratio of the fi rst to subsequent return stroke field peaks Reference Region Arithmetic Mean Geometric Mean Sample Size Nag et al. [2008] Florida 2.1 1.7 239 Austria 2.3 1.6 247 Brazil 2.4 1.9 909 Sweden 2.4 1.9 258 Present Study Florida 3.1 2.4 1693 Shown in Figure 3 5 are electric field peaks of subsequent strokes that are normalized to their corresponding first return stroke peaks and plotted versus stroke order. For strokes of order 2 to 10 (when sampl e sizes are greater than 20), the geometric mean of normalized electric field peak shows a relatively weak tendency to decrease with increasing stroke order. The range of variation of GM normalized electric field peak for strokes of order 2 to 15 is from 0 .50 (stroke ord er 2) to 0.23 (stoke order 13).

PAGE 48

48 Figure 3 5. Normalized electric field peaks for strokes of different order. Ranges of variation (vertical bars) and sample sizes (at the top of vertical bars) for subsequent strokes of different order are g iven. 3.3.4 Variation of Lightning Parameters f rom O ne Storm to A nother Statistics on flash multiplicity, interstroke interval, flash duration, and ratio of first to subsequent electric field peak for 17 individual storms are presented in Table 3 4. Some s ignificant differences were observed. For instance, the average multiplicity for the storm on 09/06/2013 was 3.5 (N=17) while its counterpart for the storm on 08/02/2014 was 6.1 (N=24). The GM interstroke interval and GM flash duration for the storm on 05/ 15/2014 were 38 ms (N=48) and 128 ms (N=16), respectively, 2 4 times smaller than 71 ms (N=75) and 554 ms (N=14) for the storm on 07/05/2014. The average ratio of first to subsequent return stroke electric field peak for the storm on 07/25/2014 was 5.5 (N= 61), 2.8 times larger than 2.2 for the

PAGE 49

49 storm on 06/08/2014 (N=114). However, the majority of storms had similar statistics for all the c onsidered lightning parameters. 3.4 Summary The characteristics of negative cloud to ground lightning were examined by a nalyzing the electric field waveforms of 478 flashes recorded during 17 storms in Florida. The percentage of single stroke flashes is 12% and the average flash multiplicity is 4.6. The arithmetic mean (AM) and geometric mean (GM) of interstroke intervals a re 80 ms and 52 ms. The AM and GM of flash durations are 329 ms and 223 ms. The ratios of first to subsequent stroke field peaks for 1693 subsequent strokes range from 0.3 to 28 with an AM of 3.1 and a GM of 2.4. The GM normalized electric field peaks of s ubsequent strokes show a slowly descending trend with increasing stroke order. Out of 421 multiple stroke flashes, 144 (34%) had at least one subsequent stroke whose field peak was greater than that of the first stroke. Significant differences were observe d in lightning parameters for different storms, however, the majority of storms had similar lightning parameter statistics. No significant disparities were found between characteristics of Florida negative cloud to ground lightning obtained in this study a nd their counterparts from previous studies.

PAGE 50

50 Table 3 4. Variation of lightning parameters from one storm to another

PAGE 51

51 CHAPTER 4 ELECTRIC FIELD SIGNATURES OF PRELIMINARY BREAKDOWN AND AUTOMATED ANALYSIS FOR HIGH INTENSITY EVENTS 4.1 Literature Review Nag and Rakov [2008] examined electric field records of negative cloud to ground flashes acquired in Gainesville, Florida, in 2006 and found that 18% of them had detectable preliminary breakdown (PB) pulse trains. However, from more recent studies of PB pulse trains in negative lightning in Florida, Baharudin et al. [2012] and Marshall et al. [ 2014] found that 100% o f flashes in each study had detectable PB pulse trains. Possible reasons for the discrepancy, including differences in noise level, record length, and distance, were discussed by Marshall et al. [2014] Figure 4 1. Schematic representation of four types of lightning (left) and the expected electric field waveforms (right). The charge configuration in each of the scenarios represents only its verti cal profile. Adapted from Nag and Rakov [2009b]

PAGE 52

52 The amplitud es of preliminary breakdown pulses were found to be significantly higher at high er latitude (temperate) region s than the counterparts at lower latitude (tropical or subtropical) region s [ Gomes et al. 1998; Nag and Rakov 2009b; Marshall et al. 2014] Gomes et a l. [1998] and Nag and Rakov [2009b] attributed this differenc e to a more significant lower positive charge region at higher latitude region s Nag and Rakov [2009b] proposed that the preliminary breakdown pulses is a manifestation of the interaction of a downward extending negative leader channel with the lower positive charge region (LPCR) Depending on the magnitude of LPCR, four different scenarios of lightning development were inferred to exist. The expected electric field signatures for these scenarios are shown Figure 4 1. Brook [1992] found that PB pulses produced in winter storms were more intense than in summer storms. Also, he found the time interval between the peaks of the first PB pulse and the first return stroke pulse (PB RS inte rval) in winter storms to be a factor of four shorter than in summer storms. The shortest PB RS interval in his study was 2.5 ms, observed in winter. He attributed the disparity to the difference in precipitation mixes in summer and winter thunderclouds. H owever, many researchers observed PB RS intervals as short as a few milliseconds in summer storms [ Heavner et al. 2002; Frey et al. 2005; Ko 2014; Zhu et al. 2015; Kotovsky et al. 2016] For the short PB RS interval events, Heavner et al. [2002] inferred leader speeds of the order of 10 6 m/s, which is an order of magnitude greater than typical speeds of negative stepped leaders. Nag and Rakov [2009b] and [2014] suggested that such unusually fast stepped leaders are produced by unusually strong negative charge sources. According to Nag and Rakov [2009b] only 5% of stepped leaders in Florida have durations shorter than 5 ms. The events with short PB RS intervals were recently found by

PAGE 53

53 Kotovsky et al. [2016] to be associated with long lasting disturbances in the upper meso sphere and lower ionosphere. In this chapter, the dataset of 478 negative flashes used in Chapter 3 was used to study the factors affecting detectability of PB pulse trains in negative cloud to ground lightning in Florida and the results are presented in s ection 4.2 The locations of these 47 8 flashes are shown in Figure 4 2. In section 4.3 distant (50 to 500 km) electric field waveforms of PB pulses in negative cloud to RS peak curr ents are analyzed by using an automated data processing algorithm. The following parameters are examined: PB RS interval, the ratio of peaks of the largest PB pulse and the corresponding first return stroke pulse (PB/RS pulse peak ratio), range normalized peak of the largest PB pulse, and PB pulse train duration. Figure 4 2. Locations of 478 flashes (first strokes only) recorded at LOG and reported by the NLDN. The expansion of the black box area in the left panel is shown in the right panel.

PAGE 54

54 4.2 Facto rs Affecting Detectability of PB Pulse Trains 4.2.1 Signal/Noise Ratio of Recording System The criteria for identifying the PB pulse trains described by Nag and Rakov [2008] were used in this study. For the raw data, the percentage of flashes with PB pulse train was 22%. However, after applying moving average filtering to the data, the percentage increased to 46%, which means that the signal/noise ratio significantly affects the detectability of PB pulse trains. Figure 4 3 shows an example of records before and after filtering. Figure 4 3. Comparison of the electric field waveforms before and after filtering. The top panel shows the waveform before filtering, in which no PB pulses can be seen. The same waveform after filtering, shown in the middle panel, exhibits a readily detectable P B pulse train around 5 ms. The bottom panel shows an expansion of the PB pulse train seen in the middle panel.

PAGE 55

55 4.2.2 Type of Storm The 478 flashes are grouped by storm in Table 4 1. For each storm, the percentage of flashes with detectable PB pulse train is given. It is clear from Table 4 1 that there is a significant variation of detectability from one storm to another, both before and after filtering. For example, before filtering, the storm on 07/03/2014 had only 6% of flashes with detectable PB pulse t rain, while this percentage for the storm on 07/10/2014 was 53%. After filtering, the percentages of flashes with detectable PB pulse train for these two storms increased to 28% and 100%, respectively, with the difference still being large. Table 4 1. Pe rcentages of flashes with detectable PB pulses before and after filtering for 17 storms Storm ID mm/dd/yyyy Number of Flashes Recorded Percentage of flashes with detectable PB Pulses Train Before Filtering After Filtering 08/17/2013 25 20% 60% 08/22/ 2013 16 38% 56% 08/31/2013 88 13% 30% 09/06/2013 17 18% 47% 05/15/2014 17 41% 59% 05/25/2014 19 21% 37% 05/28/2014 21 29% 62% 05/29/2014 21 52% 81% 06/08/2014 31 16% 42% 07/03/2014 96 6% 28% 07/05/2014 16 31% 69% 07/07/2014 15 33% 47% 07/10/2014 15 53% 100% 07/25/2014 20 15% 55% 08/02/2014 24 42% 58% 08/15/2014 20 45% 55% 08/23/2014 17 18% 35% Total 478 22% 46%

PAGE 56

56 Cooray and Jayaratne [2000] found that the PB pulses produced by lightning in Sweden were much more intense than those in Sri Lanka. They attributed the difference to stronger lower positive charge region in thunderstorms in Sweden. Based on the hypothesis proposed by Nag and Rakov [2009b] that PB puls es are the manifestation of interaction between a negative stepped leader and the lower positive charge region we speculate that storms with higher percentage of flashes with detectable PB pulse train may have a more significant lower positive charge reg ion. We also observed that flashes with detectable PB pulses tend to occur close to each other temporally (cluster in time), which might indicate that PB intensity depends on the cloud charge structure that changes during the storm life cycle. 4.2.3 Prospe ctive Return Stroke Peak Current We found that flashes with higher first return stroke peak currents (only current magnitudes are considered here) are more likely to have detectable PB pulse train s The detectability of PB for all the 478 flashes is plotte d versus first stroke peak current in Figure 4 4a, which shows a generally increasing trend. In order to reduce the effect of different distances to the LOG for different flashes, which will be examined in the next section, we chose 204 flashes in the rela tively narrow 40 50 km distance range (which has the largest sample size compared to other ranges) to see the effect of the peak current more clearly. One can see from Figure 4 4b that the PB pulse train detectability tends to increase with the increasing prospective return stroke peak current. If we combine some adjacent bins in Figure 4 4b, the percentages of flashes with detectable PB pulse train will be 44% for the 0 40 kA range (N=133), 63% for the 40 80 kA range (N=59), and 92% for the >80 kA range (N =12). Thus, flashes with higher first return stroke peak currents are generally more likely to have detectable PB pulse trains, although this trend can be countered by the distance dependence (discussed in the next section). In other

PAGE 57

57 words, a less intense event at a smaller distance can have the same probability of PB detection as a more distant event of higher intensity. Figure 4 4. Percentage of flashes with detectable PB pulse trains versus peak current. a) The percentage of flashes with detectable PB pulse trains versus peak current of first return stroke reported by the NLDN for all 478 flashes. The denominator is the sample size and the numerator is the number of flashes wi th detectable PB pulse trains. b)

PAGE 58

58 Figure 4 5. Percentage of flashes with detectable PB pulse trains versus distance. a) The percentage of flashes with detectable PB pulse trains versus distance from the strike point to the observat ion point for a ll 478 flashes. b) The same as Figure 4 5a, but for 107 flashes with first stroke peak currents ranging from 30 to 40 kA.

PAGE 59

59 4.2.4 Distance from the Strike Point Location to t he Observation Point The effect of distance from the strike point location of flash to the observation point on detectability of PB pulse trains is examined in this section. The PB detectability for all the 478 flashes is plotted versus distance in Figure 4 5a from which no clear dependence on distance can be seen, probably because of th e effect of first stroke peak current discussed in the previous section. In order to reduce that effect, we chose flashes with first stroke peak currents in the relatively narrow 30 40 kA range (which has the largest sample size compared to other ranges) t o examine the PB detectability variation versus distance. The results are shown in Figure 4 5b. Although the sample size in the first two bins and the last several bins are rather small, similar to the analysis of the effect of prospective peak current on detectability of PB pulse trains, a decreasing trend is evident. If we combine some adjacent bins, we can find that the PB detectability for flashes in the distance ranges 10 30 km (N=6), 30 60km (N=80), and more than 60 km (N=21) are 83%, 46%, and 19%, re spectively. Also, one can see from Figure 4 2 that most flashes with detectable PB pulse train before filtering (black dots) are close to the LOG. Therefore, as expected, the detectability of PB pulse trains decreases significantly with increasing distance from the strike point location to the observation point. However, this trend can be countered by a higher probability of recording larger current events from larger distances, when measurements are performed at a single station with a fixed trigger thresh old. 4.3 Automated Analysis of Preliminary Breakdown and Return Stroke Processes in High Intensity Negative Lightning Discharges 4.3.1 Data All the electric field records in this section were acquired at the Lightning Observatory in Gaine sville (LOG), Flor ida, in 2014. In the late summer of 2014, the multi station electric field measuring system was operated in single station mode. The record length is 2s with 700 ms

PAGE 60

60 pretrigger time. The system was triggered when the electric field change produced by lightn ing exceeded a fixed threshold. This triggering scheme resulted in some bias toward a larger fraction of higher intensity events in the records from larger distances. Unfortunately, this bias did not allow us to meaningfully compare, for the same azimuth, the stroke intensity over land and over salt water, since the over salt water events were farther away from the LOG than the over land events. The electric field measuring system at LOG was mostly triggered by first strokes in cloud to ground (CG) flashes, although triggering by other lightning processes occasionally occurred. Data from the U.S. National Lightning Detection Network (NLDN) were used to provide the location and first RS peak current for each flash. The median values of NLDN location error and absolute peak current estimation error were found by Mallick et al. [2014] to be 334 m and 14%, based on a comparison of NLDN responses to rocket triggered lightning with ground truth dat a. 4 .3 2 Methodology In order to process LOG electric field data more efficiently, an automated procedure was developed to match NLDN data and LOG electric field records (all GPS time stamped), process the electric field data, detect the PB pulse trains an d corresponding first RS pulses, and measure parameters of those trains, as well as the PB RS time interval. The automated procedure is described below and illustrated in Figure 4 6 Step 1. Find LOG records containing NLDN reported negative first strokes For each LOG electric field record, we define a 3 ms time window centered at the GPS time stamp of the record. Then we search the NLDN negative flash dataset containing information only on the first stroke in each flash (subsequent strokes in multiple stro ke flashes are not included in this dataset) in the 50 to 500 km range of distances from the LOG to see if the NLDN detected a first stroke in that 3 ms window. If no NLDN reported first stroke is found, the

PAGE 61

61 electric field record is discarded. If the NLDN reported a first stroke within the 3 ms time window (within 1.5 ms of the LOG trigger) and within 50 to 500 km of LOG, we go to step 2. During step 1, LOG records triggered by subsequent strokes, cloud discharges, and close (<50 km) lightning events (whic h may contain significant non radiation field components that make the automated detection of pulses difficult), as determined by the NLDN, are excluded from this study. Figure 4 6. Flowchart of the automated data processing algorithm. LOG, Lightning Ob servatory in Gainesville, Florida; NLDN, U.S. National Lightning Detection Network; PB, preliminary breakdown; RS, return stroke. Step 2. Exclude LOG records that are saturated or contain other strokes prior to the first stroke of interest If an NLDN repor ted first stroke is found in step 1, we further check, in a 60 ms window saturated or (2) one or more strokes are found up to 50 ms prior to the first return stroke. This is

PAGE 62

62 accomplished by using another NLDN dataset which includes information on both first and subsequent strokes in each flash. The 50 ms pre return stroke interval was selected because 95% of PB RS intervals in Florida negative CGs are less than 50 ms [ Baharudin et al. 2012; Zhu et al. 2015] Occurrence of other return strokes (e.g ., from a different flash) less than 50 ms prior to the first return stroke would not allow us to adequately examine the PB RS time interval. If the record is saturated or there are other return strokes within 50 ms prior to the first return stroke of inte rest, the record is discarded. Otherwise, we go to step 3. Step 3. Reduce noise At this point, ideally, we have a 60 first return stroke at t = 0 and possibly cloud pulses before and after the first return stroke. Additionally, we allowed the occurrence of subsequent strokes (using the NLDN temporal and spatial criteria to group strokes into flashes) between 0 and +10 ms, which would not interfere with our analysis. The 60 ms records were passed through the moving average and power line noise removal filters to remove 60 Hz power li ne noise and to improve the signal/noise ratio. The power line noise removal filter generates a 60 Hz sinusoidal waveform, based on the phase and amplitude of the power line signal found from the 60 ms record, and subtracts that sinusoidal waveform from th e original record. The moving average window was 0.6 s (31 points). After reducing noise we go to step 4. Step 4. Identify and characterize individual pulses From the beginning of the 60 ms record to 1.4 ms prior to the first return stroke (from 60 ms to all bipolar pulses wider than 5 s with amplitudes greater than three times the residual noise level. Bipolar pulses were defined as composites of two consecutive monopo lar pulses of

PAGE 63

63 opposite polarity separated by a time interval shorter than one tenth of the width of the first pulse. The criteria used at this step were optimized by the trial and error method to achieve the best discrimination of lightning generated pulse for the pulse search window was needed to exclude pulses generated by the stepped leader near ground; also, no PB RS intervals shorter than 1.4 ms were reported for Florida lightning by Baharudin et al. [2012] and Zhu et al. [2015] At the end of this step, we have amplitudes and widths of all pulses in the search window. Then we go to step 5. Step 5. Identify PB pulse trains (if any) and measure parameters of the train W e consider as a pulse train a sequence of at least three detected bipolar pulses. Pulses that are separated from the last pulse of the train by more than 2 ms are considered not belonging to that train, which is the same criterion as that used in Nag and Rakov [2008] and Zhu et al. [2015] If no PB pulse train is identified, the flash is labeled as having no detectable PB pulses and not included in further analysis. If one or more PB pulse trains are identified, the flash is considered as a flash with detectable PB pulses. In the following, we consider only pulses that belong to PB pulse trains. If a record contains one or more detected PB pulse trains, the automated dat a processing code will output a plot showing the 60 ms filtered electric field record, including the first return stroke pulse and all detected PB pulses, each marked by a box, along with all the pertinent information (given in a table above the plot) abou t this event, including the filename, date, location, NLDN reported peak current, and all the PB pulse train parameters (including the PB RS interval). An example of the output plot is shown in Figure 4 7 From August to October 2014, during 31 storms, tho usands of electric field records were obtained and processed using the automated algorithm described above. For a total of 5498 flashes with NLDN reported first return strokes, 3496 (64%) had detectable PB pulse trains. The

PAGE 64

64 NLDN reported locations of all t he 5498 first strokes are shown in Figure 4 8 Since (1) the algorithm excluded all flashes that occurred within 50 km of LOG and (2) the average distance to LOG for the 5498 accepted flashes was 235 km, our dataset is biased toward flashes with higher pea k currents. Specifically, the arithmetic mean (AM) and geometric mean (GM) of NLDN reported peak currents are 119 kA and 104 kA, respectively, for all the 5498 events, and 124 kA and 109 kA, respectively, for 3496 events with detectable PB pulse trains, wh ich are considerably higher than the GM value of 30 kA for first strokes in negative lightning [ Rakov and Uman 2003] A histogram of NLDN reported peak currents for the 3496 events with detectable PB pulse trains is shown in Figure 4 9 In this study, we exclu ded 133 events with multiple PB pulse trains and considered only high intensity events with the NL DN reported first 122 kA. Figure 4 7. An example of the output figure produced by the automated data processing algorithm, a) An output figure and a table cont aining pertinent information about the flash; b) Expansion of the initial part of the PB pulse train shown in (a) with the detected PB pulses marked by boxes. The d.u. stands for digitizer units.

PAGE 65

65 Figure 4 8. NLDN reported locations of the 5498 negative first strokes within 50 to 500 km of LOG. Table 4 2 Characterization of PB pulses in 3077 negative flashes each containing a single PB pulse train. Parameters AM GM Min Max Sample size PB/RS pulse peak ratio 0.15 0.13 0.02 0.81 3077 PB RS interval ( ms) 8.8 7.5 1.7 49.9 3077 PB pulse train duration (ms) 2.7 2.2 0.19 26.8 3077 Bipolar pulse width (s) 25 21 5 170 43010 NLDN reported peak current (kA) 134 122 14 432 3077 Distance to LOG (km) 246 222 51 495 3077

PAGE 66

66 Figure 4 9. Histogram of NLDN reported peak currents for the 3496 negative first strokes in flashes with detectable PB pulse trains. The arithmetic mean (AM) and geometric mean (GM) distances to LOG for the 3496 flashes are 229 and 201 km, respectively. 4.3. 3 Characteristics of PB Pulse Trains in High Intensity Negative Flashes We found that 133 (3.8%) of 3496 flashes with detectable PB pulses had multiple PB pulse trains. In order to avoid ambiguity, these 133 events were excluded from the analysis presented in this section. Fur ther, to limit our analysis to high intensity events, we also excluded 286 (8.2%) events with NLDN reported peak current <50 kA, so that the sample size here is reduced to 3077 with the total number of bipolar pulses being 43,010. Statistics for the PB RS interval, PB/RS pulse peak ratio, PB pulse train duration, and bipolar pulse width, as well as for the NLDN reported current and distance are given in Table 4 2

PAGE 67

67 The AM PB/RS pulse peak ratio of 0.15 in Table 4 2 is comparable to the AM values of 0.20 to 0.29 reported in the literature [ Baharudin et al. 2012; Marshall et al. 2014; Zhu et al. 2015] The AM PB RS interval of 8.8 ms in Table 4 2 is more than a factor of two smaller than previously reported, th is disparity being likely due to the very high AM peak current of 134 kA (GM = 122 kA) in our dataset. The AM and GM values of PB pulse train duration in Table 4 2 are somewhat longer than their counterparts, 2.2 and 1.7 ms, respectively, given in Zhu et al. [2015] The AM and GM bipolar pulse widths in Table 4 2 are 25 s and 21s, resp ectively, which are consistent with typical widths of 20 40 s reported by Rakov et al. [1996] Histograms of the parameters of PB pulses summarized in Table 4 2 are shown in Fig ures 4 10 to 4 13 Figure 4 10. Histogram of PB/RS pulse peak ratio for the 3077 flashes.

PAGE 68

68 Figure 4 11. Histogram of PB RS interval for the 3077 flashes. Figure 4 12. Histogram of PB pulse train duration for the 3077 flashes.

PAGE 69

69 Figure 4 13. Histo gram of bipolar pulse width in the 3077 flashes. 4.3. 4 Correlation between Parameters of Preliminary Breakdown and Return Stroke Processes In this section, besides showing the correlation plots for the dataset (N=3077) of high intensity ( 50 kA ) flashes, similar plots were shown for the dataset (N=3363) without excluding the flashes with first stroke peak currents ranging from 0 to 50 kA. Figure 4 14 a shows a scatter plot of PB RS interval vs. NLDN reported RS peak current. Clearly, the PB RS interval tend s to decrease with increasing peak current. The best fit curve, y = 1560 x 1.12 Spearman correlation coefficient between the PB RS interval and the NLDN reported peak value less than 0.001; this indicates a strong negative correlation which is statistically significant at the 99.9% confidence level. Note that Spearman correlation coefficient is a non parametric measure of rank correlation [ MacDonald

PAGE 70

70 2008] which i s used to assess monotonic relationships (not necessarily linear). S imilar plot is also shown in Figure 4 15 for 3363 negative first strokes, in which the 0 50 kA events were not excluded. A stronger correlation could be found. Zhu et al. [2014] reported nine negative flashes (Table 4 3) with short PB RS intervals, all of which exhibited high (AM = 1 31 kA) NLDN reported peak currents. For five of the nine events, the corresponding Lightning Mapping Array (LMA) data were available and showed that the first LMA source heights ranged from 4.8 to 6 km. Based on these observed flash initiation heights and corresponding PB RS intervals (stepped leader durations), Zhu et al. [2014] estimated the averag e 1D stepped leader speed to be 1.2 10 6 m/s, which (although an underestimate) is almost an order of magnitude higher than 2 10 5 m/s thought (e.g., Rakov and Uman [2003] Chapter 4) to be typical for negative stepped leaders. Thus, the strong correlation seen in Figure s 4 14 a and b suggests that negative flashes with faster stepped leaders tend to have higher first RS peak currents. Jordan et al. [1992] found that the subsequent (dart or dart stepped) leader speed and return stroke peak current are positively correlated for both natural and rocket triggered lightning. It is likely that this trend also holds f or first strokes initiated by stepped leaders. Table 4 3. Summary of short PB RS interval events Flash ID Time Interval between PB and First Return Stroke (ms) First Return Stroke Peak Current Reported by NLDN (kA) Inferred Leader Speed* (m/s) 839 3.5 222 1.6110 6 854 4.5 133 1.310 6 881 5.9 82 0.7810 6 882 6.0 102 1138 4.4 129 1203 4.0 150 1.2310 6 1204 3.6 172 1.2810 6 1205 4.3 110 1215 5.0 128 GM 4.5 131 1.2110 6 Estimated as v=H/ T PB RS, where H is the altitude of the first LMA so urce. This estimate should be considered as a lower bound because the actual (3D) channel length should be considerably larger than H.

PAGE 71

71 Figure 4 14. PB RS interval versus NLDN reported RS peak current for 3077 negative first strokes. a) Scatterplot of PB RS interval versus NLDN reported RS peak current for 3077 negative first strokes within 50 to 500 km of LOG. The best fit curve, y = 1560 x 1.12 arman correlation coefficient; b) AM values of PB RS interv al versus NLDN reported RS peak current for individual 50 kA bins. The standard errors in mean values are shown by vertical bars and the corresponding sample sizes are given above them.

PAGE 72

72 Figure 4 15 PB RS interval versus NLDN reported RS peak current for 3363 negative first strokes. a) Scatterplot of PB RS interval versus NLDN reported RS peak current for 3363 negative first strokes. within 50 to 500 km of LOG. The best fit curve, y=203.4* x 0.52 8.49, is shown by solid line. R is the Spe arman correlat ion coefficient. b) AM values of PB RS versus NLDN reported RS peak current interval for individual 50 kA bins. The standard errors in mean values are shown by vertical bars with the corresponding sample size shown above.

PAGE 73

73 Figure 4 1 6 Peak of the larges t PB pulse normalized to 100 km versus NLDN reported RS peak current for the 3077 negative first strokes. a) Scatterplot of the peak of the largest PB pulse normalized to 100 km (in digitizer units) versus NLDN reported RS peak current for the 3077 negativ e first strokes within 50 to 500 km of LOG. R is the Spe arman correlation coefficient; b) AM values of normalized PB pulse peak versus NLDN reported RS peak current for individual 50 kA bins. The standard errors in mean values are shown by vertical bars an d the corresponding sample sizes are given above them.

PAGE 74

74 Figure 4 17. Peak of the largest PB pulse normalized to 100 km versus NLDN reported RS peak current for the 3363 negative first strokes. a) Scatterplot of the peak of the largest PB pulse normalize d to 100 km (in digitizer units) versus NLDN reported RS peak current for the 3363 negative first strokes within 50 to 500 km of LOG. R is the Spearm an correlation coefficient. b) AM values of normalized PB pulse peak versus NLDN reported RS peak current f or individual 50 kA bins. The standard errors in mean values are shown by vertical bars with the corresponding sample size shown above.

PAGE 75

75 Figure 4 1 8 Peak of the largest PB pulse normalized to 100 km versus PB RS interval for the 3077 negative first stro kes. a) Scatterplot of the peak of the largest PB pulse normalized to 100 km (in digitizer units) versus PB RS interval for the 3077 negative first strokes within 50 to 500 km of LOG. R is the Spe arman correlation coefficient; b) AM values of normalized PB pulse peak versus PB RS interval for individual 5 ms bins. The standard errors in mean values are shown by vertical bars and the corresponding sample sizes are given above them.

PAGE 76

76 Figure 4 19. Peak of the largest PB pulse normalized to 100 km versus PB R S interval for the 3363 negative first strokes. a) Scatterplot of the peak of the largest PB pulse normalized to 100 km (in digitizer units) versus PB RS interval for the 3363 negative first strokes within 50 to 500 km of LOG. R is the Spe arman correlation coefficient. b) AM values of normalized PB pulse peak versus PB RS interval for individual 5 ms bins. The standard errors in mean values are shown by vertical bars with the corresponding sample size shown above.

PAGE 77

77 Scatterplot s of the peak of the largest P B pulse normalized to 100 km versus NLDN reported peak current are shown in Figure s 4 1 6 a and 4 17a for datasets with peak currents from 50 to 450 kA and peak currents from 0 to 450 kA, respectively Normalization was performed assuming that the electric f ield peak in the 50 to 500 km distance range varies as the inverse inclined distance from the elevated PB source to the LOG. The inclined distance was roughly estimated using the NLDN reported horizontal distance and the assumed source height of 6 km. One can see that the maximum normalized PB pulse peak tends to increase with increasing RS peak current. For the datasets with peak currents from 50 to 450 kA, t he Spearman correlation coefficient between the maximum normalized PB pulse peak and NLDN reported peak current is 0.48 and it is statistically significant at the 99.9% confidence level. It is difficult to discern any trend in Figure 4 1 6 a where the individual data points are shown. However, there is a clear trend (Figure 4 1 6 b) for the AM normalized PB pulse peak corresponding to individual 50 kA bins to increase with increasing RS peak current. Scatterplot s of the normalized PB pulse peak vs. PB RS interval are shown in Figure s 4 1 8 a and 4 19a The normalized PB pulse peak tends to decrease with increa sing PB RS interval and the corresponding Spearman coefficient are f or the dataset with peak currents from 50 to 450 kA and 0.6 for the dataset with peak currents from 0 to 450 kA It is statistically significant at the 99.9% confidence level. Similar trends were previously observed in Florida [ Marshall et al. 2014] as well as in Japan [ Wu et al. 2013] and China [ Wang et al. 2016] 4.4 Summary For 47 8 negative cloud to ground lightning flashes, factors that might affect detectability of PB pulse trains were examined. By using the moving average filtering, the percentage of flashes with detectable PB pulse trains increased from 22% to 46%. Thus, the de tectability of PB

PAGE 78

78 pulse train is significantly affected by the signal to noise ratio of the recording system. Further, PB pulse train detectability can vary from one storm to another. The PB pulse trains of flashes with higher peak currents of the first re turn stroke and smaller distances to the observation point are more likely to be detected. Using an automated data processing algorithm, we have examined the characteristics of PB pulse trains and the following first return strokes (RSs) in negative cloud to ground lightning flashes in Florida. Out of 5498 flashes within 50 to 500 km of LOG, 3496 (64%) had PB pulse trains that were detected by the automated algorithm. For the 3077 flashes with one detectable stroke peak current, the arithmetic (geometric) mean values of peak current, PB pulse train duration, PB RS interval, PB/RS pulse peak ratio, and bipolar pulse width were 134 (122) kA, 2.7 (2.2) ms, 8.8 (7.5) ms, 0.15 (0.13), and 25 (21) s, respectively. The PB RS interval was found to decrease with increasing NLDN reported first (statistically significant at the 99.9% confidence level). Since shorter PB RS intervals were found to correspond to faster leaders [ Zhu et al. 2014] the latter result suggests that negative flashes with faster stepped leader s tend to have higher first stroke peak currents. The largest range normalized PB pulse peak exhibited positive correlation with the RS peak current, with Spearman correlation coefficient of 0.48 (statistically significant at the 99.9% confidence level). T hus, it appears that the high (and, by inference, faster) stepped leaders and more pronounced PB pulse trains. The range normalized PB pulse peak tended to decrease with increasing the PB RS interval (Spearman

PAGE 79

79 CHAPTER 5 A SUBSEQUENT POSITIVE STROKE DEVELOPING IN THE CHANNEL OF PRECEEDING NEGATIVE STROKE AND CONTAINING BIPOLAR CONTINUING CURRENT 5 .1 L iterature Review Correlated optical and electric field records of natural downward cloud to ground bipolar flashes are rarely reported. Based on video observations and multi station field measurements, Jerauld et al. [2009] gave the fir st well documented description of a natural downward bipolar flash containing two initial positive strokes with strike points separated by about 800 m, followed by four negative strokes that traversed the same channel as the second positive stroke. Fleenor et al. [2009] reported four bipolar flashes that all started with a positive stroke followed by one or two negative strokes. Out of five subsequent nega tive strokes in the four flashes, two followed the pre existing but decayed channel of the first (positive) stroke. Saba et al. [2013] different polarity occurred in the same channel, although there could have been other strokes in differe nt channels. Each of their bipolar flashes started with a positive stroke and all the second (negative) strokes were initiated by optically imaged recoil leaders in decayed upper level branches of the first downward positive leader. Saraiva et al. [2014] observed one single channel downward bipolar flash and one multi channel downward bipolar flash. Each of these two started with a positive stroke and all the negative subsequent strokes occurred as a result of recoil leaders, in the manner described above. Chen et al. [2015] reported a downward bipolar flash with a first (positive) stroke followed by five negative strokes, al l occurring in the same channel terminated on a 90 m tall structure. Tian et al. [2016] reported a downward bipolar f lash in which the first (positive) stroke was followed by three negative strokes along the same channel. Similarly to Saba et al. [2013] and Saraiva et al. [2014] they foun d that the three subsequent

PAGE 80

80 negative strokes were initiated by recoil leaders. There is only one case found in the literature, in which a subsequent positive stroke developed in the channel of preceding negative stroke. It was recently reported by Xue at a l. [2015] who attributed the polarity change to some intracloud process that altered the cloud charge structure during the relatively long time interval (136 ms) between the two strokes. In this chapter we present cor related electric field and high speed video records of a natural four stroke cloud to ground flash having a negative first stroke followed by a positive stroke in the same channel whose 2D length was 4.2 km. As follows from the literature review given abov e, this scenario is highly unusual. The second (positive) stroke was followed by bipolar continuing current (with the initial positive charge transfer to ground followed by negative charge transfer to ground), so that the flash exhibited the features of bo th Type 1 and Type 3b bipolar discharges identified by Rakov [2003] The third and fourth strokes in our flash were negative and followed a newly created channel, different from the one of the first and second strokes. The atmospheric electricity sign convention, according to whi ch the downward directed electric field change vector produced by a negative return stroke is positive, is used throughout this chapter 5 .2 Instrumentation The bipolar flash (labeled 2117) was recorded at 01:55:20 UT on Aug. 22 nd (at 21:55:20 local time o n Aug. 21 st ), 2014 at the Lightning Observatory in Gainesville (LOG), Florida, by two high speed video cameras, Phantom V310 and Megaspeed HHC X2, and by electric field measuring systems. Additionally, the event was recorded by the Total sky Lightning Chan nel Imager (TLCI) installed at LOG as part of our collaboration with the Chinese Academy of Meteorological Sciences. The electric field measuring systems include the low gain and high gain electric field measuring systems and the electric field derivative (dE/dt) measuring system.

PAGE 81

81 The record length for the field measuring systems was 1 s with 200 ms pretrigger time. The Phantom V310 was operated at 3200 frames per second (fps) with 80 s exposure time (232.5 s deadtime) and resolution of 1280800 pixels. T he HHC X2, equipped with a fish eye lens to provide a wider field of view, was operated at 1000 fps with 1 ms exposure time (essentially no deadtime) and resolution of 832600 pixels. The electric field records and high speed video records were GPS time st amped. The synchronization accuracy between the Phantom records and electric field records was better than 1.3 s. The spatial resolution of Phantom records used in this study was 5 m. The TLCI had a fish eye lens and was operated at 40 fps with 25 ms expo sure time [ Lu et al. 2014] Outputs of all three optical instruments are generally consistent with each other. Only Phantom images are presented in this chapter Additionally, U.S. National Lightning Detection Network (NLDN) data, including locations and peak current estimates for lightning strokes, were used in this study. 5 .3 Observations and Analysis 5 .3.1 General Description According to the NLDN, flash 2117 contained four strokes with peak currents, in the order of occurrence, of 101 k A, 16 kA, 20 kA, and 32 kA. Based on the locations of strike points provided by the NLDN, the distances between the 1st and 2nd strokes, 2nd and 3rd strokes, and 3rd and 4th strokes were 0.14 km, 1.87 km, and 0.12 km, respectively. The corresponding inte rstroke intervals (measured between return stroke field peaks) were 70 ms, 210 ms, and 65 ms. The distance between the strike point of the first (negative) stroke and LOG was 5 km. This distance was used for estimating all the heights, as well as 2D distan ces and speeds presented in this chapter The Phantom high speed video camera captured all the strokes of this flash. The Megaspeed HHC X2 high speed video camera and TLCI captured only the first and second strokes, with the third and fourth strokes being outside their fields of view.

PAGE 82

82 From all the video records, it was unambiguously determined that the second stroke followed the same channel as the first stroke, except for the bottom 115 m, where the channels of the first two strokes were found, from the P hantom record, to be slightly different from each other. It appears that the second stroke, positive leader, after extending from 3.9 km (the upper (negative) st roke, deviated from that path and contacted ground 40 m from the first stroke termination point. About 210 ms after the second return stroke, the third, negative stroke occurred. It exhibited branching and created an entirely new channel whose termination point was about 1.87 km from that of the second stroke. The fourth stroke was not branched and followed the main channel of the third one. Leaders of the third and fourth strokes had propagation speeds that were characteristic of stepped and dart leaders, respectively. In the field of view of Phantom camera, no clear relation was seen between the former two strokes and the latter two strokes, but they all do satisfy the spatial and temporal stroke grouping into flash criteria used by the NLDN [ Cummins et al. 1998] The third and fourth strokes are not further discussed in this chapter Composite images of t he first two leader return stroke sequences are shown in Figure 5 1. As noted above, within about 115 m of the ground, the first and second strokes followed slightly different paths and formed separate ground terminations, which were 40 m apart (2 D distan ce estimated from the Phantom camera record). The separate terminations are clearly seen in the magnified superposition of the bottom portions of the channels of the first and second strokes shown in the inset in Figure 5 1. As discussed below, the bottom part of the second stroke channel could be actually created by the first stroke leader. Based on the Phantom camera

PAGE 83

83 record, the durations of continuing currents (CCs) following the first (negative) stroke and the second (positive) stroke were 5 ms and 122 ms, respectively. Figure 5 1. Composite Phantom images of the negative first stroke (left panel labeled CG) and the positive second stroke (right panel labeled +CG). The images were produced using all the frames (except for one and three saturated one s for the negative and positive strokes, respectively) corresponding to the leader, return stroke, and continuing current processes. The top of the imaged channel was about 3.9 km above ground. The inset shows the magnified composite image of the bottom po rtions of the channels of the first and second strokes. T1 and T2 mark the ground terminations of the first and second strokes, respectively. T he predominantly horizontal branches seen in the upper left corner of the right panel were repeatedly illuminated before and during the development of downward positive leader, as well as during and after the continuing current following the positive return stroke. These branches were apparently associated with concurrent in cloud discharge activity and were possibly connected to the faintly luminous channel segment in the upper right corner (to the right from the main channel) of the right panel, although the connection was obscured by cloud debris. The faintly luminous channel segment (maybe indiscernible in the rep roduction) probably had intermittent connection to the main channel to ground.

PAGE 84

84 Figure 5 2. Electric field and dE/dt records of the bipolar flash. Panels a) to c) show simultaneous low gain electric field, high gain electric field, and dE/dt records of the initial part of the bipolar flash, including the preliminary breakdown (PB), the first and the second leader (L) /return stroke (RS) sequences and bipolar continuing current (CC). Duration of the bipolar CC was estimated from the Phantom camera record In panels d) to g), the l ow gain electric field (panels d) and e)) and ele ctric field derivative (panels f) and g)) waveforms of the negative first and positive second return strokes are shown on a 200 s time scale. All records have been obtained at a di stance of 5 km from the lightning channel.

PAGE 85

85 Figures 5 2a to 5 2c show simultaneous low gain electric field, high gain electric field, and dE/dt records of the two strokes of interest. The time interval between the negative first return stroke and the positi ve second return stroke is 70 ms, which is comparable to the typical interstroke interval in negative flashes of 60 ms [e.g., Rakov and Uman 2003, p. 7] The leader durations (measured in the low gain electric field record) for the first and second strokes are similar, 9.5 ms and 9.4 ms, respectively. The net electric field change produced by the leader of the first stroke has the same polarity as the corresponding return stroke field change, while for the second stroke the net electric field changes due to the leader and return stroke have opposite polarities. This disparity is likely to be caused by different horizontal displacements of the negative and positive charge sources relative to the observer and the channel to ground, as discussed by Rakov et al. [1990] with the positive charge being considerably farther from both the observer and channel to ground than the negative one. If so and in view of another polarity change during the continuing current after RS2, the cloud charge structure was far from Rakov and Uman [2003, Ch. 3] and references therein), which could be possibly related to the fact that the thunderstorm was in its dissipating s tage. Based on the high gain electric field record, the first preliminary breakdown (PB) pulse, which had the same polarity as the negative return stroke pulse, occurred 9.5 ms before the negative first return stroke onset, which is consistent with the st epped leader duration obtained by measuring the duration of leader electric field waveform in the low gain record. The expansions of the negative and positive return stroke electric field and dE/dt waveforms are shown in Figures 5 2d to 5 2g. Note that the first (negative) return stroke exhibited a double peak electric field waveform (Figure 5 2d), which might be indicative of two channel terminations on ground created by the forked first stroke leader (e.g., Thottappillil et al. 1992;

PAGE 86

86 Rakov and Uman 1994; Ballarotti et al. 2005 optical records. It is possible that the ground termination labeled T2 in the inset of Figure 5 1 was one of the terminations formed by the first (negative) stroke leader, but appeared only in the first frame containing the return stroke, which was saturated, and was too faint to be imaged in the following frames. If so, the apparent deviation of the second stroke leader from the first stroke channel at height of 115 above ground level could actually be the second stroke leader following one of the decayed paths to ground created by the forked first stroke leader. T he RC decay time constants of the low gain and high gain electric field measuring systems were 10 ms and 440 s, respectively, so that only field changes occurring on the time scale shorter than 1 ms or so (for example, Figures 5 2d and 5 2e) were faithful ly reproduced by the low gain electric field measuring system and only shorter than 44 s or so by the high gain electric field measuring system. The instrumental decay can be compensated for using the method proposed by Rubinstein et al. [2012] We have used this method in determining the CC polarity reversal point in our low gain field record (Figure 5 2 a). In practice, the late part of the signals that have decayed to the level that is close to the noise may not be accurately reproduced, so determination of the moment that polarity change occur is a rough estimation. Our approach was as follows. We assum ed that 1) the electric field change produced by CC is essentially electrostatic, 2) the electrostatic field change is proportional to the charge transfer Q, and 3) the corresponding current I is given by dQ/dt. Under these assumptions, the CC polarity cha nge should correspond to the zero rate of change (negative maximum) of the bipolar CC electric field signature, after its compensation for instrumental decay. Using the compensated electric field waveform, in which the negative maximum occurred considerabl y later than seen in Figure 5 2a, we found that the initial 44 ms long portion of the CC following the second return stroke was

PAGE 87

87 positive (positive charge transported to ground) and the following 78 ms long portion was negative (negative charge transported to ground). The average luminosity of the main channel during the positive CC was 3.7 times higher than that during the negative CC, which is likely indicative of about 3.7 times higher current of the positive CC. For comparison, average currents associate d with upward negative leaders (positive charge transported to ground) initiated from tall objects are considerably higher than their counterparts associated with upward positive leaders. For example, the average current for negative charge transfer to gro und at the Gaisberg tower (Austria) was 113 A [ Diendorfer et al. 2011] while for positive charge transfer to ground it was 707 A [ Zhou et al. 2012] It is likely that upward negative leaders are more heavily branched inside the cloud than the upward positive ones, which makes the upward negative leaders more efficient in collecting the cloud charges and funne ling them to the channel to ground. 5 .3.2 Characteristics of t he Negative and Positive Leaders One can see from the left panel of Figure 5 1 that the stepped leader (labeled L1 in Figure 2a) initiating the negative first stroke, like most negative stepped leaders, was branched. Its 26 frame to frame speeds range from 2.610 5 m/s to 13.110 5 m/s with an arithmetic mean of 4.710 5 m/s, which is equal to the average speed calculated by dividing the entire 2D length of the channel by the time it took the leade r to traverse that channel. This average speed is somewhat higher than the typical negative stepped leader speed of 210 5 m/s given in Rakov and Uman [2003, Ch apter 4] possibly because the stroke was of higher than typical intensity. Variation of the frame to frame speed of the negative stepped leader versus height of the leader tip above ground is shown in red in Figure 5 3. The speed varied irregularly between 2.6 and 6.410 5 m/s at heights ranging from 3600 m to 1000 m above ground and then significantly increased in the last few frames before the leader attachment to ground.

PAGE 88

88 The positive leader (labeled L2 in Figure 5 2a) initiating the second stroke exhibited no low level branches. It was fainter tha n L1 and became clearly visible only when it was at the height of 3.2 km above ground. For comparison, the typical cloud base height during summer thunderstorms in Florida is about 1.5 km above ground. Variation of the frame to frame 2D speed vs. height fo r the positive leader is shown in blue in Figure 5 3. Similar to L1, L2 accelerated as it was approaching the ground. The minimum, maximum, and mean values of 14 frame to frame speeds are 4.810 5 m/s, 1210 5 m/s, and 7.210 5 m/s, respectively. Figure 5 3 The frame to frame 2D speeds of the negative stepped leader (L1) and the following positive leader (L2) versus height of the leader tip above ground. Both leaders developed in the same channel and significantly accelerated below 2 km above ground.

PAGE 89

89 5 .4 Discussion example, Ch. 5 of Rakov and Uman [2003] and references therein), believe that subsequent strokes never transport positive charge to groun d along the previously created channel. For example, for the case of channel created by a negative stroke, Saba et al. [2013] positive subsequent return stroke following a negative return stroke is likely not viable given that recoil leaders (RL) would be required to retrace the horizontal channels created by neg ative leaders. To date, RLs forming on decayed negative leader br The mechanism proposed by Saba et al. [2013] for natural bipolar cloud to ground flashes, in which negative stroke follows the path of positive stroke, is illustrated in Figure 5 4. In this suggested mechanism, the leader develops bidirectionally with the postive end propagating toward the ground and heavily branched n egative end developing inside the cloud. Altough the positve leader developed nearly vertical near ground, it has some horizontal braches near the cloud base which cut off from the main channel as the positive leader is approaching the ground After the c essation of the postive stroke, t he recoil leader is initiated in decayed horizontal positive channel and propagates birdirectionally. When their negative end gets connection to the decayed main channel, a subsequent negative return stroke is initiated. Figure 5 4. Sequence of events that lead to a bipolar flash with the negative second stroke initiated in a decayed branch of positive leader. RL is the bidirectional recoil leader. Adapted from Saba et al. [2013]

PAGE 90

90 However, this generalizing statement is not supported by our observations and those of Xue at al. [2015] who presented correlated high speed video and electric fie ld records of a two stroke bipolar cloud to ground flash with the positive second stroke retracing the channel of the negative first stroke. Additionally, the occurrence of recoil leader like processes that formed in decayed negative leader branches was re cently reported by Montany et al. [2014] and Stolzenburg et al. [2015] In our view, a recoil leader is an electrical breakdown in warm, low density air which the decayed lightning channel branches are filled with. T he polarities of its ends should be determined by the direction of local electric field or electric field change vector. It is not clear how the conditions for occurrence of this breakdown can be influenced by the sign of charge previously hosted by the br anches. The average speed of L2 is about 1.5 times higher than that of L1. For negative lightning, it is expected (e.g., Rakov and Uman 2003 Ch. 4) that a leader traversing previously created channel is considerab ly (1 to 2 orders of magnitude) faster than the stepped leader that created the channel. For positive strokes traversing previously created channels, such information is not available. Prior to this writing, the speed of positive leader that traversed the pre existing channel (regardless of preceding stroke polarity) was reported only for one event by Xue et al. [2015] They optically observed a bipolar flash with only two video frames being recorded for each the negat ive stepped leader and the subsequent positive leader, with both leaders following the same channel. The corresponding frame to frame 2D speeds were 3.610 5 m/s and 1.710 6 m/s and the interstroke interval was 136 ms. The typical speeds of negative dart st epped and dart leaders are (1 2)10 6 m/s and (1 2)10 7 m/s, respectively [ Rakov and Uman 2003, Ch.4] and, hence, the subsequent, positive leader speed reported by Xue et al [2015] is similar to that expected for negative dart stepped leaders. In our case, the 7.210 5 m/s value is

PAGE 91

91 lower than expected for negative subsequent leaders in the previously formed channel. As noted above, the interstroke interval between the fi rst two strokes in our bipolar flash was not too long (in contrast with the event reported by Xue et al [2015] ), 70 ms, which is not much different from the median value of 60 ms for interstroke intervals in negative flashes. Further, the magnitude of NLDN reported first stroke peak current was very large (101 kA). Thus, the lightning channel traversed by the positive leader might be expected to be warm enough to allow faster than observed leader propagation. Saba et al. [2010] reported that the aver age speeds of 29 positive leaders (it is not specified if all of them were developing in virgin air) ranged from 0.24 to 11.810 5 m/s with the arithmetic mean of 2.7610 5 m/s, which is about 2.5 times lower than the average speed of the subsequent positive leader observed in our study. Considering that positive flashes normally contain a single stroke or have subsequent strokes following newly created channels, the majority of the 29 positive leaders studied in Saba et al. [2010] probably propagated through virgin air, and hence, are not d irectly comparable to our positive leader traversing the previously conditioned channel. Additional data on positive leaders developing in warm air are needed to help interpret our observations. One possible explanation of the relatively low speed of our positive subsequent leader is the relatively low (but rather typical for negative subsequent strokes) magnitude of corresponding return stroke peak current, only 16 kA vs. 101 kA for the preceding negative stroke. Indeed, Jordan et al. [1992] reported positive correlation between the subsequent leader speed and return stroke peak current in both natural and triggered negative lightning. Also, Zhu et al. [2015] found a tendency for negative first return strokes with higher peak current to be preceded by faster stepped leaders. Another possible factor is the height above sea level, 2496 m

PAGE 92

92 in the work of Xue et al. [2015] vs. 52 m in our study. Detailed comparison of our flash with that of Xue et al. [2015] is given in Table 5 1. Clearly, further research of this rare phenomenon is needed. Table 5 1. Compari s on of bipolar flashes with a positive stroke following the negative stroke channel Xue et al. [2015] This study Location Qinghai, China Florida, U.S. Season Summer Summer Altitude above sea level (m) 2496 52 Storm stage Dissipation Dissipation field of view (km) unknown 3.9 Interstroke interval (ms) 136 70 Pe ak current of R1 (kA) unknown 101 Peak current of R2 (kA) unknown 16 R1/R2 peak current ratio 1.6* 6.3 Leader speed of L1 (10 5 m/s) ** 3.6 4.7 Leader speed of L2 (10 5 m/s) ** 17 7.2 L2/L1 speed ratio 4.7 1.5 *Peak currents are not available in the s tudy of Xue et al. [2015] so that the electric field peak ratio is used instead. **Each of the leader speeds in the study of Xue et al. [2015] was the frame to frame spee d based on two frames only, while our speeds are average values based on 27 and 15 frames for the negative leader and the positive leader, respectively. 5 .5 Summary Using simultaneous high speed video camera records and electric field measurements, we exa mined a bipolar flash that started with a negative stroke with a peak current of 101 kA, which was followed by a positive stroke with a peak current of 16 kA. From the leader electric

PAGE 93

93 field waveforms, the positive charge source is inferred to be located considerably farther from both the observer and the channel to ground than the negative charge source. The second ( positive) stroke was followed by a bipolar continuing current The average luminosity of the main channel during the positive CC was 3.7 time s higher than that during the negative CC, which is likely indicative of significantly higher current of the positive CC. It is likely that upward negative leaders are more heavily branched inside the cloud than the upward positive ones, which makes the up ward negative leaders more efficient in collecting the cloud charges and funneling them to the channel to ground. In our view, a recoil leader is an electrical breakdown in warm, low density air which the decayed lightning channel branches are filled with The polarities of its ends should be determined by the direction of local electric field or electric field change vector. It is unlikely that the conditions for occurrence of this breakdown can be influenced by the sign of charge previously hosted by the branches. The first two strokes (including the continuing current) followed the same channel to ground, whose imaged 2D length was 4.2 km, except for the bottom 115 m, where the paths of the two strokes were slightly different. As of this writing, there i s only one previously documented case of positive leader following the path of preceding negative stroke. We presented the first leader speed versus height profiles for such an unusual sequence. The average leader speeds for the first (negative) and second (positive) strokes were 4.710 5 m/s and 7.210 5 m/s, respectively. The speed of the positive leader traversing the previous stroke channel after not unduly long (70 ms) interstroke interval is lower than typical speeds of negative leaders following previo usly formed channels. The speeds of both the negative leader and the positive leader increased as they approached the ground.

PAGE 94

94 CHAPTER 6 OPTICAL AND ELECTRIC FIELD SIGNATURES OF LIGHTNING INTERACTION WITH THE 257 M TOWER IN FLORIDA 6 .1 Literature Review Several research groups reported narrow bipolar electric field pulses (initial half cycle durations ranging from 5 to 15 s or so), most of which, if not all, being associated with lightning strikes to tall towers. Pavanello et al. [2007] observed that for lightning striking the 553 m tall CN tower (Canada), the ele ctric and magnetic field waveforms measured at 16.8 km and 50.9 km exhibited a very short first zero crossing time of about 5 s followed by a narrow (<5 s) opposite polarity overshoot ( Figure 6 1). They attributed these unusual field signatures to the tr ansient process excited by lightning along the tower. The CN tower observations were performed in summer. Ishii and Saito [2009] reported bipolar electric field waveforms produced to waveforms are shown in Fig ure 6 2. Ishii and Saito [2009] suggested that GC discharges involved an upward leader making contact with a horizontal in cloud channel, as illustrated in the left panel of Fig ure 6 2 b. In this scenario, the source is located in the cloud, as opposed to being located at ground level in the case of normal cloud to ground (CG) lightning (right panel of Figure 6 2b ). Pichler et al. [2010] presented simultaneous current and electric field records of 73 subsequent return strokes terminated on the 100 m Gaisberg tower (Austria) and found that the arithmetic mean zero crossing time of the electric field waveforms of those strokes was 11.2 s, which is mo involving tall objects. Examples of current and electric field waveforms of a stroke terminated on the Gaisberg tower are shown in Fig ure 6 3. Most of the Gaisberg tower events occ urred during the cold season. Wu et al [2014] using a low frequency lightning location network, observed

PAGE 95

95 in Japan. They speculated that LBEs involved tall grounded objects and occurred when the n egative charge layer in the cloud was very close to the top of the strike object. The LBE label was also used for the discharge giving rise to the observed electric field waveforms, examples of which are shown in Fig ure 6 4. Ishii et al. [2013] and Saito et al. [2015] reported unusually narrow electric field waveforms produced by return strokes in upward and downward lightning discharges to the 634 m high Tokyo Skytree i n Japan, observed in spring and summer. Observed and calculated electric field waveforms for one subsequent return stroke recorded at three stations, along with their causative current waveform, are shown in Fig ure 6 5. Note that the current waveform meas ured at the tower is similar to those typical for normal negative return strokes (relatively fast rise to peak and relatively slow decay). Chen et al. [2015a] simulated far field waveforms (Fig ure 6 6) characteristic of LBEs observed by Wu et al. [2014] assuming that they occurred in relatively short channels (500 10 00 m) attached to strike objects of 100 to 300 m in height. They used the bouncing wave model for two different positions of the source representing the RS like and ICC pulse like processes and found that the electric field waveforms observed by Wu et al. [2014] could be reproduced in either case, but only when the injected current waveform was a symmetric Gaussian pulse. Note that Chen et al. [2015a] used the field sign convention that is opposite to the one used by Wu et al. [2014] In this chapter we present two lightning flashes containing a total of eight negative strokes that terminated on a 257 m tower and produced unusually narrow bipolar electric field waveforms with damped oscillatory tails. High speed video camera records were also obtained for these two flashes. The observed electric field waveforms exhibit some similarities to the

PAGE 96

96 events revi m tower was obscured by trees, but the azimuth of the 8 strokes and NLDN data clearly indicate that all of them terminated on the tower. Figure 6 1. Current derivative, current, electric field, and magn etic field records of the 2 nd stroke of the flash strking the 553 m CN tower on August 19 th 2015. The current was measured at the height of 474 m. Adapted from Pavanello et al. [2007]

PAGE 97

97 Figure 6 2. The observed and simulated to a) Electric field waveforms simultaneously recorded at three different stations in winter in Japan and attributed to a GC discharge. b) Models of GC stroke initiated by an upward positive leader (left) and normal return stroke initiated by a downward negative leader (right). In the latter case the souce is pl aced at the channel base and the channel is vertical, while in the former case the source is placed at the junction point between a 2 km vertical, grounded channel section and a 250 m horizontal, in clo ud section. c) Electric field waveform calculated at 1 00 km using the mo del shown in the left panel of b). Ground reflection occurred at A and reflection from the upper end of the vertical channel occurred at B. The calculated electric field waveform sho wn in c) resembles the GC stroke wavef orm measured at 10 km shown in a) (middle trace). Adapted from Ishii and Saito [2009]

PAGE 98

98 Figure 6 3. Current and electric field records of a negative lightning stroke ter minated on the 100 m Gaisberg tower. The current was measured at the top of the tower. The peak current of this event is 15.9 kA. Adapted from Pichler et al. [2010] Figure 6 4. Electric field waveforms of a typical LBE recorded by nine stations at distances ranging from 37.5 to 236.1 km in winter in Japan. The waveform recorded at 236.1 km is shown on the expanded time scale in the inset (the second, smaller pulse between 400 and 500 s is the ionospheric r eflection). The typical initial half cycle width was 15 s. Adapted from Wu et al. [2014]

PAGE 99

99 Figure 6 5. A subsequent return stroke terminated on the Tokoyo Skytree showing the narrow bipolar signature in electric field records. a) Observed and calculated current waveforms for a subsequent return stroke of an upward negative flash terminated on the 634 m high Tokyo Skytree in May of 2012 (Flash 2012 1). The current was measured at the height of 497 m. b) d) Corresponding observed and calculated electric field waveforms at three distances ra nging from 27 to 101 km. The initial half cycle width of electric field waveforms is about 5 s, which is considerably smaller than the duration of current waveform sh own in a). The measured current peak is 16 kA. Adapted from Ishii et al. [2013] 6 2 Observations and Analysis The two flashes presented in this chapter were recorded at the Lightning Observatory in Gainesville (LOG), Florida, by electric field measuring systems and Megaspeed HHC X2 high speed video camera in the s ummer of 2014. The detailed information of the field measuring systems at LOG can be found in Section 2.2. In this study, the record length for the field measuring systems was 1 s with 200 ms pretrigger. The Megaspeed HHC X2 camera, equipped with a fish ey e lens to provide a wider field of view (about 185 o ), was operated at 1000 frames

PAGE 100

100 per second (fps) with 1 ms exposure time (essentially no deadtime) and resolution of 832600 pixels. The record length of the camera was 1.2 s with 200 ms pretrigger. No proc essing of optical images was done to remove the fish eye effect. Since the imaged channels were near the lens center, the distortion was not significant. All the records were GPS time stamped The field measuring system was synchronized with the high speed video camera with precision better than 1 ms. The measuring system and high speed camera were triggered when the electric field exceeded a preset threshold. More detailed information on LOG can be found in Section 2.2 and the review paper by Rakov et al. [2014] Figure 6 6. Simulated electric field waveforms of LBEs at 100 km P anels a) and c) for RS like proces s and panels b) and d) for ICC pulse like process H, h, r, and v are the lightning channel length, strike object height, horizontal distance from the lightning strike point, and current wave speed, respectively. g T t are the reflection coefficients at the ground level, at the channel top, and at the strike object top, respectively. Physics sign convention (a downward directed electric field or field change vector is considered to be negative) is used in this figure. Adapted from C hen et al. [2015a]

PAGE 101

101 6 2 .1 General Description of t wo Flashes Terminated on the 257 m and 60 m Towers in Florida Flash 1593 was recorded at 12:10:58 UT (at 08:10:58 local time) on July 16 th 2014. It was an upward negative flash whose upward positive leader initiated from the 257 m high tower (located 8.8 km from the LOG). This upward leader was clearly seen in six consecutive frames (for about 6 ms) moving at an average 2D frame to frame speed of 3.210 5 m/s. It was optically detectable up to a height of 1.9 km. Flash 1593 contained 6 negative strokes, all of which terminated on the 257 m tower. This flash occurred after and possibly was initiated by a nearby intracloud discharge. Flash 1594 was a downwa rd bipolar flash that occurred 8 minutes after flash 1593. Natural downward bipolar flashes are rare with only several observations being found in the literature [ Fleeno r et al. 2009; Jerauld et al. 2009; Nag and Rakov 2012; Saba et al. 2013; Chen et al. 2015b; Xue et al. 2015; Tian et al. 2016; Zhu et al. 2016b] During the 8 minute interval between flashes 1593 and 1594, only one cloud discharge was reported by the NLDN within 40 km of the LOG. The first stroke of flash 1594 was positive and terminated on the 60 m tower located 3.6 km from the 257 m tower and 8.9 km from the LOG. It was followed by two negative strokes that terminated on the same 257 m tower as the 6 strokes of flash 1593. All of the 8 negative strokes that terminated on the 257 m tower exhibited very similar electric field waveforms, characterized by a narrow bipolar pulse with a damped oscillatory tail. The electric field waveform of the positi ve stroke terminated on the 60 m tower was unipolar and exhibited initial, predominately radiation field peak followed by a large electrostatic ramp.

PAGE 102

102 Table 6 1. NLDN data on 8 negative strokes in flashes 1593 and 1594 terminated on the 257 m tower. Flas h Type Stroke ID Peak Current (kA) Preceding Interstroke Interval (ms) NLDN Classification (C for cloud events and G for cloud to ground events) Semi Major Axis Length (m) Semi Minor Axis Length (m) Number of Reporting Sensors Distance between the NLDN Re ported Location and the 257 m Tower (m) Upward Negative Flash 1593 1 7.6 C 200 200 4 110 1593 2 5.7 18 C 200 200 5 50 1593 3 20.7 64 G 200 100 5 40 1593 4 6.5 56 G 200 200 4 80 1593 5 6.6 18 C 200 200 4 60 1593 6 6.7 14 C 200 200 5 90 Downwa rd Bipolar Flash* 1594 2 6.5 148 C 200 200 5 70 1594 3 10.2 20 C 200 200 5 140 Mean 8.8 42 200 200 5 80 Information on the first, positive stroke of this bipolar flash is given in Table 6 2. Table 6 2. NLDN data on the first, positive stroke of b ipolar flash 1594, which terminated on the 60 m tower, located 3.6 km from the 257 m tower. Flash Type Stroke ID Peak Current (kA) Preceding Interstroke Interval (ms) NLDN Classification (C for cloud events and G for cloud to ground events) Semi Major Axi s Length (m) Semi Minor Axis Length (m) Number of Reporting Sensors Distance between the NLDN Reported Location and the 60 m Tower (m) Downward Bipolar Flash* 1594 1 193 G 100 100 20 30 Information on the second and third strokes of this bipolar flas h is given in Table 6 1

PAGE 103

103 Figure 6 7. Locations of strike points reported by the NLDN for 8 negative strokes terminated on the 257 m tower. Blue squares are the repo rted strike points for 6 strokes of flash 1593 and green squares are the reported strike points for 2 strokes of flash 1594. The yellow square is the location of the tower and the ground truth location of the 8 strokes. SMA stands for the semi major axis l ength and r is the NLDN location error defined as the distance from the NLDN reported location to the ground truth location (yellow square). Note that the yellow square, whose position is determined by the geographical coordinates of the tower (available f rom http://www.cellreception.com/towers/ details.php?id=1029807), is about 25 m from the location of the tower base seen in the Google Earth map. The reason for this offset is unknown. 6 2 .2 NLDN Responses to the 257 m and 60 m Tower Strokes The NLDN information for the two flashes is summarized in Table 6 1 (negative strokes) and Table 6 2 (positive stroke). The 8 negative strokes had NLDN reported peak currents ranging from 5.7 to 20.7 kA with a geometric mean of 8 kA and an arithmetic mean of 8.8 kA. Out of the 8 negative strokes, 6 were misclassified by the NLDN as cloud discharges. The peak currents for the 2 correctly classified events were 20.7 and 6.5 kA. Warner et al. [2012] reported (leader/return stroke sequences and ICC pulses) in upward flashes initiated from towers in South

PAGE 104

104 Dakota They found that the peak to zero times and pulse durations of electric field pulses produced by misclassified events were smaller than those for the correctly classified ev ents. The NLDN reported locations for the 8 negative strokes and the 257 m tower location are shown in Figure 6 7. The distances between NLDN reported locations and the tower location ranged from 40 m to 140 m, all being less than 200 m, the median error ( assumed to be equal to the semi major axis length of the location error ellipse) reported by the NLDN for each located event. For the first, positive stroke of the bipolar flash 1594, the NLDN reported peak current was 193 kA and the distance between the NLDN reported location and the 60 m tower was 30 m. The positive stroke is not further discussed in this chapter. 6. 2 .3 High Speed Camera and Electric Field Data for Eight Negative Stro kes of Flashes 1593 and 1594 First frames showing the channels (including the upper parts normally hidden inside the cloud) of the eight negative return strokes from the two flashes are shown in Figure 6 8. Multiple frames showing the evolution of luminous channels (not discussed in this chapter) are presented for each event in Appendix C. From the high speed video camera data, the time interval between the initiation of upward positive leader and the first return stroke in flash 1593 was 184 ms. There was no initial continuous current detectable in either optical or electric field records after the upward positive leader. The channel length for individual strokes progressively increased (extended upward) with increasing stroke order. The corresponding elect ric field waveforms on a 30 s time scale are shown in Figure 6 9. They are very similar to each other. Except for 1593 R4 having a shoulder in its field wavefront, all the eight strokes had very fast rise and fall times (<2 s) of the field initial half c ycle. The field waveforms of all eight strokes show damped oscillatory tails. The similarity of field waveforms for different strokes in two different flashes suggest that they are largely determined by the transient response of the tower. The field

PAGE 105

105 wavefo rm parameters are defined in Figure 6 10, in the same manner as in Wu et al. [2014] and summarized in Table 6 3. Electric field waveforms on a longer time scale (not discussed in this paper) are presented in Appendix C. Figure 6 8. First video frames showing the channel of each of the eight negative strokes terminated on the 257 m tower. Table 6 4 compares the field waveform parameters obtained in the prese nt study with the parameters of similar field waveforms found in the literature [ Pavanello et al. 2007; Ishii and Saito 2009; Pichler et al. 201 0; Ishii et al. 2013; Wu et al. 2014] Our events have the narrowest average electric field initial half cycle width (T p ) of 2.44 s. It is significantly smaller than its counterparts reported by Ishii and Saito [2009] Pichler et al. [2010] and Wu et al. [2014] but only a factor of two smaller than those reported by Pavanello et al. [2007] and Ishii et al. [2013] The NLDN reported peak currents for our events are considerably smaller than those reported by Ishii and Saito [2009] and Wu et al. [2014] but are rather similar to those

PAGE 106

106 reported by Pavanello et al. [2007] Pichler et al. [2010] and Ishii et al. [2013] Note that direct current measurements are available only for the events presented by Pavanello et al. [2007] Pichler et al. [2010] and Ishii et al. [2013] For the events presented in this chapter, as well as for those studied by Ishii and Saito [2009] and Wu et al. [2014] the peak currents were estimated from measured fields. The ratio of the first (positive) and the second (negative) field half cycles for our events is about a factor of 4 to 5 larger than those for LBEs studied by Wu et al. [2014] and GC events reported by Ishii and Saito [2009] but very close to those reported by Pavanello et al. [2007] and Ishii et al. [2013] Further observations and associated modeling are needed to better understand the nature of narrow bipolar field waveforms observed both in winter and in summ er with strike object heights ranging from 100 m or less to 634 m. Modeling of our events is presented in Chapter 7. Figure 6 9. E lectric field waveforms of 8 negative strokes terminated on the 257 m tower. The electric field waveform of 1593 R3 is saturated. Atmospheric electricity sign convention (a downward directed electric field or field change vector is considered to be positiv e) is used in this figure and in all other figures in this chapter showing electric field waveforms, except for Figure 6 6.

PAGE 107

107 Figure 6 10. Measurements (definitions) of electric field waveform parameters. Table 6 3. Electric field waveform parameters fo r the seven negative strokes terminated on the 257 m tower. a T p of 1593 to 5 s, which is the time interval between first and second zero crossing b May be biased due to exclusion of the largest event (1593 R3) c The ratio of the st andard deviation to the arithmetic mean Stroke ID T p (s) T 10 90 (s) T 10 10 (s) T r (s) T f (s) T r /T f T n (s) T p /T n A p /A n A p A n 1593 1 2.82 0.98 3.43 1.24 1.58 0.78 0.61 4.60 19.5 1.17 0.06 1593 2 2.25 0.80 3.22 0.93 1.32 0.70 0.97 2.32 4.1 1.03 0.25 1593 3 a 0.51 0.19 1593 4 3.91 2.45 5.08 2.71 1.20 2.26 1.17 3.34 3.3 1.2 0.36 1593 5 1.92 0.55 3.06 0.72 1.20 0.60 1.14 1.68 2.7 1.39 0.51 1593 6 2.04 0.59 3.12 0.79 1.25 0.63 1.08 1.89 3.0 1.37 0.44 1594 2 2.10 0 .62 3.33 0.85 1.25 0.68 1.23 1.71 2.7 1.30 0.48 1594 3 2.02 0.57 2.95 0.82 1.20 0.68 1.14 1.78 2.9 2.06 0.71 AM b 2.44 0.94 3.46 1.15 1.29 0.90 0.98 2.47 5.5 1.36 0.38 Coefficient of variation c 0.29 0.73 0.21 0.61 0.11 0.66 0.28 0.45 1.14 0.24 0.51

PAGE 108

108 6. 3 Summary Simultaneous electric field and high speed video camera records of two flashes, labeled 1593 and 1594, were acquired at the Lightning Observatory in Gainesville (LOG), Florida, in the summer of 2014. Fla sh 1593 was an upward negative flash whose upward positive leader initiated from a 257 m high tower after a nearby intracloud discharge and propagated with an average 2D frame to frame speed of 3.210 5 m/s. It contained 6 downward leader/return stroke sequ ences, all of which terminated on the tower following the upward positive leader path. Flash 1594 was a downward bipolar flash, the first stroke of which was positive and terminated on a 60 m tower, while two subsequent strokes were negative and terminated on the same 257 m tower as the six strokes of flash 1593. The first negative stroke of 1594 created a new path to the 257 m tower and, hence, was likely stepped. The peak current reported by the NLDN for the 8 negative strokes ranged from 5.7 to 20.7 kA, with a geometric mean of 8 kA and an arithmetic mean of 8.8 kA. Out of the 8 negative strokes, 6 were misclassified by the NLDN as cloud discharges. All the 8 negative strokes that terminated on the 257 m tower (located 8.8 km from LOG) exhibited very simi lar narrow (2.44 s on average) bipolar electric field pulses with damped oscillatory tails. From comparison with similar narrow bipolar electric field waveforms observed in different studies, we found that our events have the narrowest average electric f ield pulse width. It is significantly smaller than its counterparts reported by Ishii and Saito [2009], Pichler et al. [2010], and Wu et al. [2014], but only a factor of two smaller than those reported by Pavanello et al. [2007] and Ishii et al. [2013]. Th e NLDN reported peak currents for our events are considerably smaller than those reported by Ishii and Saito [2009], and Wu et al. [2014], but are rather similar to those reported by Pavanello et al. [2007], Pichler et al. [2010], and Ishii et al. [2013]. The ratio of the amplitudes of first and the second field half cycles for our events is

PAGE 109

109 about a factor of 4 to 5 larger than those for LBEs studied by Wu et al. [2014] and GC events reported by Ishii and Saito [2009], but very close to those observed by Pa vanello et al. [2007] and Ishii et al. [2013]. It is not clear if all the lightning events producing narrow bipolar electric field waveforms discussed in this chapter are of the same nature or not. Further research is needed.

PAGE 110

110 Table 6 4. Characteristics of narrow bipolar electric field waveforms observed in different studies References Location and season Strike object height (m) Sample size Tp (s) T 10 90 (s) T 10 10 (s) Peak current (kA) Tr/Tf Tp/Tn Ap/An Pavanello et al. [2007] Toronto, Canada, summer 553 1 Around 5 5.5 2 a 4.4 a Ishii and Saito [2009] Hokuriku region, Japan, winter unknown 21 AM=12.3 AM=53 >70 1.3 b Pichler et al. [2010] Salzburg, Austria, mostly at cold season 100 73 AM=11.2 AM=1.8 AM=10.6 Wu et al. [2014] Western Japan, winter unknown 356 AM=15.1 AM=68.8 >1 for 92.4% of events Median around 1 Median around 1 Ishii et al. [2013] Tokyo, Japan, summer 634 1 4.5 c 16 1.6 c 4 c This study Florida, USA, summer 257 7 AM=2.4 AM=0.9 AM=3.5 AM=7.1 AM=0.9 AM=2.47 AM=5.5 a Measured in Figure 4g of Pavanello et al. [2007] b Measured in Figure 2 of Ishii and Saito [2009] c Measured in Figure 13b of Ishii et al. [2013]

PAGE 111

111 CHAPTER 7 MODELING OF LIGHTNING INTERACTION WITH THE TOWER 7 .1 Literature Review Rakov and Uman [1998] categorized lightning return stroke models into four classes based on the types of governing equations. Th e first class models are gas dynamic models, which involve the solution of gas dynamic equations (conservation of mass, momentum, and energy) coupled to two equations of state. In contrast with the other three classes of return stroke models, i t is primari ly concerned with the radial evolution the lightning channel. T he gas dynamic model developed by Paxton et al. [ 1986] could output the profiles of temperature, mass density, pressure, and conductivity versus channel radius at different time s The second class models are electromagnetic models, in wh ich the lightning channel s are often represented by loaded or unloaded conducting wire s E lectromagnetic models, using numerical techniques, such as Finite Difference Time Domain (FDTD) method or Method of Moment (MoM), solve equations to produce the current distribution along the lightning channel, which could be used to calculate the radiated electric and magnetic fields. The third class models are distributed circuit model s in which the lightning channel is considered as an RLC transmission li ne, where R, L, C are the series resistance, series inductance, and shunt capacitance, all per unit, respectively. By ation s the profiles of current and voltage along the channel as a function of time and height can be obtaine d lightning channel have the transverse electromagnetic (TEM) field structure. The fourth class models are engineering mode ls, which are equations relating the longitudinal current along the lightning channel at any height and any time to the current at the channel base. Engineering models can be further grouped in two categories: traveling current source type models and trans mission line

PAGE 112

112 type models. In the traveling current source type models, the return stroke may be viewed as be ing generated at the upward moving return stroke front and then propagating downward, while in the transmission line type models a current source i s placed at the channel base to inject current into the channel. The current wave propagates upward without attenuation for the original transmission line model [ Uman and McLain 1969] and with specified height dependent attenuation terms for different modified transmission line models [ Rakov and Dulzon 1987, 1991; Nucci et al. 1988] The current equations for the four widely used transmission line type models are summarized in Table 7 1. Those models included (1) the transmission line (TL) model, (2) the modified TL model with linear current decay with height (MTLL), (3) the modified TL model with exponential curre nt decay with height (MTLE), and (4) the modified transmission line model with parabolic current decay with height (MTLP). Table 7 1. Current equations for transmission line type return stroke models Type Equation TL [Uman and McLain, 1969] MTLL [Rakov and Dulzon, 1987] MTLE [Nucci et al., 1988] MTLP [Rakov and Dulzon, 1991; Maslowski and Rakov, 2006] As discussed in Chapter 6, lightning striking tall objects can produce electric field waveforms with first zero crossing times rangin g from 2 s to 15 s or so [ Pavanello et al. 2007; Ishii and Saito 2009; Pichler et al. 2010; Wu et al. 2014; Zhu et al. 2016c] which are significantly smaller than the typical values ranging from 30 to 50 s [ Rakov and Uman 2003, chapter 4] Some researchers attempted to model such narrow electri c field waveforms. Saito et al. [2015a, 2015b] obtained multiple station measurements of very narrow (first zero crossing time is about 2.5 s) bipolar electric field waveforms produced by a stroke in

PAGE 113

113 upward lightning striking the Tokyo Skytree (634 m) in Japan. The corresponding current measured at the height of 497 m showed fast rise followed by a slow decay, which is the charac teristic of a normal return stroke. Numerical Electromagnetics Code (NEC 4) based on the method of moments was used for numerical analysis by Saito et al. [2015a, 2015b] In the model, lightning channel and tower were represented as thin wires and the voltage source was placed at the top of the tower. T he observed electric field was reproduced by using the measured current wavefor m and channel geometry reconstructed from high speed camera images. The channel was not vertical, and the authors noted that the channel tilt was an important factor in achieving a good match with measured field waveforms. Chen et al. [2015] used a bouncing wave model to simulate far field waveforms of large bipolar events (LBEs) observed by Wu et al. [2014] In the model, the lightning channel was assumed to be very short (hundreds of meters) and the tower height ranged from 100 to 300 m. Reflections at the channel top, tower top, and tower bottom were considered. They found that the LBE like waveform can be reproduced only when the injected current was characterized with a symmetric Gaussian pulse. Also, they found that the simulated electric field waveforms for RS like events (current injected at the top of the tower) and ICC pulse like processes (current injected at the top of the channel) were very similar. Azadifar et al. [2017] measured a narrow bipolar current waveform recorded at the height of 82 m of the 124 m Santis Tower in Switzerland, which was associated with the first stroke in a downward negative 4 stroke flash str iking the tower. The peak current was as high as 102 kA. T he recorded current waveform shown in Figure 7 1 is very similar to a symmetric Gaussian pulse except that it was followed by a small opposite polarity overshoot. Neglecting the presence of the t ower and using the MTLE model, Azadifar et al. [2017] calcul ated the

PAGE 114

114 corresponding electric field waveform at a distance of 100 km that showed the characteristics similar to those of LBEs observed by Wu et al. [2014] Figure 7 1. Measured current waveform (black line) associated with the first return stroke in a downward flash recorded at the Santi s tower The red line is the analytical represent Adapted from Azadifar et al. [2017] In this chapter, we will try to reproduce the narrow bipolar electric field waveforms produced by lightning striking the 257 m tower i n Florida and presented in Chapter 6 Effects of each model para meter on the vertical electric field waveform will be investigated. 7 .2 Model D escription The engineering return stroke models were originally introduced for lightning strik es to flat ground. For lightning striking tall towers, reflections at tower extremi ties have to be taken into consideration. Rachidi et al. [2002] based on a distributed current source representation of the lightning channel, extended several engineering lightning return stroke models to include the effect of t ransient processes in tower. Baba and Rakov [ 2005] used the lumped voltage source (illustrated in Figure 7 2) in order to extend the TL model to the case of lightning strike to the

PAGE 115

115 tall object. The approach proposed by Baba and Rakov [2005] will be used here for all four models The distributions of current along the tower and along the light ning channel are given by equations 7 1 and 7 2. Along the tower, ( 7 1) Along the lightning channel, ( 7 2) where n is the number of reflections occurring between the top and bottom of the tower and h is the height of the tower. v is the speed of current wave traveling upward in the lightning channel. is the lightning short circuit current, which is defined by Baba and Rakov [2005] as the lightning current that would be measured at an ideally ground ed object of negligible height. and are the current reflection coefficients at the tower bottom and tower top, respectively, which could be expressed as ( 7 3) ( 7 4) where , are the characteristic impedance of the tower, grounding impedance, and the equivalent impedance of the lightning channel

PAGE 116

116 Figure 7 2. Transmission line representation of lightning strike to a tall object. The derived expression for curr ent along the tower (equation 7 1) is same as its counterpart in Rachidi et al. [2002] However, the expression for current along the channel (equation 7 2) slightly differs from its counterpart derived by Rachidi et al. [2002] In the latter case the ground reflected current wave transmitted into the lightning channel was set to propagate upward along the lightning channel at the speed of light, while Baba and Rakov [2005] assumed the corresponding value of speed to be v < c, which is more physic al (does not allow the higher order waves to catch up with the original upward moving front) As noted above, the model proposed by Baba and Rakov [2005] will be used in this study. Several assumptions were used in this model. The reflection coefficients at the top and bottom of the towers are assu med to be constants. The lightning channel and the tower are simplified as vertical conductors above a

PAGE 117

117 perfectly conducting ground. The lightning channel extending from the tower top to the cloud is assumed to be very long so that there is no need to consi der the reflection at the channel top. The expression (equation 7 5) for electric field produced by infinitesimal vertical dipole in cylindrical coordinate system given by Thottappillil et al. [1997] and integration over the channel length will be used to compute the vertical electric field. The code developed in this study to compute the fields was first validated with the example given by Uman et al. [1975 Fig. 3 ] (7 5 ) The last term in equati on 7 5 is no zero when the current waveform has a discontinuity at its front. 7 .3 Sensitivity Analysis In this section, we will first vary the value of each model parameter (e.g. rise and fall times of current waveform, reflection coefficients, tower heigh t) to study their effects on the resultant electric field waveforms. In Section 7.4, we will try to reproduce the narrow bipolar electric field waveform associated with all (except one for which the measured electric field waveform was saturated ) of the Fl orida tower strokes presented in Chapter 7. Current measurements for L BE like events were rarely reported. For a stroke terminated on the Tokyo Skytree that produc ed narrow bipolar electric field waveform s at multiple stations the measured current (Figure 6 5) was very similar to the typical current waveform of subsequent return stroke [ Saito et al., 2015a, 2015b] However, for a nother presumed LBE like event observed at the Santis tower, the measured current waveform was like a Gaussian pulse with a

PAGE 118

118 small opposite polarity overshoot (Figure 7 1). Both types of current waveform we re used here as I sc to simulate the resultant verti cal electric field. Figure 7 3 shows the typical current waveform of a subsequent stroke (left panel) and the resultant electric field (right panel) at 10 km from the tower. The current waveform is expressed by the Heidler function. Parameters for the Heid ler function (given in the left panel ) were the same as specified by Rakov [2003] except for the peak of the current, which was reduced here to 10 kA from 50 kA. The tower was assumed to be 300 m in height and the current reflection coefficients at the top and the bottom of the tower were assumed to be 0.5 and 1 respectively One half of the speed of light was used as the return stroke speed. T he t ransmission line return stroke model wa s used. It appears from Figure 8 2, that the use of current waveform typical of subsequent stroke does not allow reproduction of the narrow bipolar electric fi eld waveform. However, the periodic field variation caused by current bouncing between the tower ends is clearly seen in Figure 7 3 In Figure 7 4 the symmetric Gaussian waveform was used as the input current. The function for the Gaussian waveform is giv en in Equation 7 6 where a and t 1 determine the magnitude and time of the pulse peak. The g 1 and g 2 determine the steepness of the rising and falling edges (i.e., rise and fall times). As s een from the right panel of Figure 7 4 the narrow bipolar field waveform followed by the oscillatory tail is produced. In this case, the Gaussian waveform was symmetric since g 1 is equal to g 2 The electric field contributions from the tower and from the lighting channel are shown in the right panel in Figure 7 4. The electric field waveform produced by current along the tower is a narrow bipolar pulse while that produced by current along the lightning channel is a unipolar pulse that is very similar to the current pulse. It follows that the presence of tower causes th e total electric field to become narrower because of the opposite polarity ( negative ) overshoot in the electric field contribut ion from the tower. It can also be observed that the initial

PAGE 119

119 half cycles of the field contributions from the tower and from the l ightning channel have comparable magnitudes ( 7 6 ) Figure 7 3. Typical current waveform of subsequent stroke (left panel) and the computed electric field (right panel) at a distance of 10 km from the tower. The time axis in right panel is shifted by d/c, where d is the distance from tower to th e observation point and c is the speed of light. Figure 7 4. Similar to Figure 7 3, but for symmetric Gaussian waveform. In the right panel, the electric field contributions from the tower and from the lightning channel are indicated.

PAGE 120

120 The effects of ris e and fall times of Gaussian waveform were investigated as well. The results are shown in Figure 7 5 With decreasing the rise time of the current waveform, the width of the initial half cycle of the electric field waveform become s smaller and the peaks of both the initial half cycle and the first opposite polarity overshoot become greater. The magnitude of the first opposite polarity overshoot and the following oscillation s become larger as the sharpness of the falling edge increases In the following, th e effects of reflection coefficients, strike object height, return stroke speed, and return stroke model will be tested by using the current waveform shown in Figure 7 4 (left panel) Except for the variable to be tested, all other parameters remain the sa me, as specified in Figure 7 4 The reflection coefficients at the tower top t ) and bottom g ) determine the magnitude s of current waves bouncing along the tower and transmitted into the lightning channel, which in turn affect s the electric field waveform as shown in Figure 7 6 As expected, the magnitudes of the initial half cycle and the following oscillation s are increasing with in creasing (in absolute value) t or g due to more energy being trapped within the tower t =0 (characteristic impedances of the tower and the lightning channel are the same ), there is no reflection at the tower top and oscillations in electric field waveform disappear (blue curve in Figure 8 g =1 (perfect ly conducting ground), there is full reflection, and the reflection g decreases.

PAGE 121

121 Figur e 7 5. The asymmetric Gaussian waveforms with different rise times and fall times and their corresponding simulated electric field waveforms. Figure 7 6. Electric field waveforms computed for different reflection coefficients at the tower t ) and g ).

PAGE 122

122 The effects of the height of the strike object and return stroke speed on the electric field waveform can be seen in Figures 7 7 and 7 8 respectively For taller strike objects, the electric field is more enhanced and the width of the initial cycle and period of the following oscillation s are larger. For shorter strike objects the overshoot and oscillations become less pronounced and eventually disappear (blue curve in Figure 7 7 ) The return stroke speed controls the ratio of magnitud es of the initial and second half cycles. For larger return stroke speed s the magnitude of the initial half cycle of the electric field waveform becomes larger while the magnitude of the overshoot becomes smaller. The return stroke speed has little effec t on the following oscillation s Four return stroke models, TL, MTLE, MTLL, and MTLP, were compared and the results are shown in Figure 7 9 The attenuation constant was set to 1 km for the MTLE model. For all return stroke models, the peak of the initial half cycle of the electric field waveforms which is attained within a few microseconds, is almost the same. The peak of the opposite polarity overshoot appear s to increase with increasing the current attenuation with height along the channel (from no att enuation for the TL model to the strongest for the MTLE model) Note that in this study we used only the transmission line type return stroke model s that were developed the 8 er attenuation of current wi th height predicted by the existing models A n attenuation of more than 30% in the lower portion of lightning channel s was inferred from luminosi ty pulses produced by return strokes in rocket triggered lightning [e.g. Wang et al. 1999; Carvalho et al. 201 5] A stronger attenuation of current along the lightning c hannel (also changing channel geometry from predominantly vertical to predominantly horizontal) could affect the contribution from the

PAGE 123

123 lightning channel (yellow curve in the right panel of Figure 7 4) to the total electric field waveform Figure 7 7. Electric field waveforms computed for different heights of the strike object. Figure 7 8. Electric field waveforms computed for different return stroke speeds.

PAGE 124

124 Figure 7 9. Electric field wa veforms computed for different return stroke models. 7.4 Modeling of Lightning Events Terminated on the 257 m Tower As described in Chapter 6, we obtained optical and electric field records for 8 strokes terminated on the 257 m tower in Florida E lectric f ield waveforms for 7 strokes (excluding one stroke with saturated field measurement) were computed by using the model described above. A symmetric Gaussian pulse s shown in Figure 7 10 was used as the short circuit current (I sc ). Asymmetric Gaussian pulse s w ere used to represent current waveform s in modeling preliminary break down pulses and compact intracloud discharges [ e.g., Shao and Heavner 2006; Nag and Rakov 2010a, 2010b; Karunarathne et al. 2014] Also, in Section 7.3, we used asymmetric Gaussian pulse to re produce the opposite polarity overshoot in the measured electric field waveform, which could not be achieved using the typical ret urn stroke current waveform. The computed and measured electric field waveforms for 7 strokes at 8.8 km are shown in Figure 7 1 1 The MTLL return stroke model was used for all the 7 strokes The model parameters used to

PAGE 125

125 compute electric fields waveforms sh own in Figure 7 11 are summarized in Table 7 2. The model parameters were determined by using trial and error method to achieve the best matching between the computed and measured electric field waveforms. The first step is to roughly match the positive pe ak and the rising edge of the electric field waveform by varying the rise time and magnitude of the current waveform. After that, efforts were made to roughly match the opposite polarity overshoot by choosing appropriate values of reflection coefficients, g 2 and return stroke speed. When the rough matching was achieved, some further small tunings would be made to polish the matching. Figure 7 10. The current waveforms used for computing the electric field waveforms shown in Figure 7 11.

PAGE 126

126 Table 7 2. Mode l input parameters used for computing electric field waveforms shown in Figure 7 11 a (kA) g 1 (10 5 ) g 2 (10 5 ) t g v 1593 1 4.0 35 4.1 0.3 1 0.90c 1593 2 3.6 35 6.5 0.3 1 0.90c 1593 4 4.2 35 9.0 0.3 1 0.90c 1593 5 5.0 40 9.0 0.3 1 0.85c 1593 6 4.7 35 9.0 0.3 1 0.95c 1594 2 4.7 40 9 .0 0.3 1 0.8 0 c 1594 3 7.2 30 10 0.3 1 0.9 0 c Figure 7 11. The measured and computed electric field waveforms. The corresponding I sc for each event is shown in Figure 7 10.

PAGE 127

127 In the process of matching (by trial and error) the model prediction s to measurements it was found that the width of the opposite polarity overshoot is mainly determined by the fall time of the current and the height of strike object (fixed at 257 m in this case). However, a shorter fall tim e results in larger oscillation s (Figure 7 5 b). In order to match the peak of the opposite polarity overshoot t = 0.3) at the tower top had to be chosen for all the 7 events Because of such a smaller value of t current in the tower is significantly reduced which results in no observable oscillation s after 15 s (Figure 7 1 1 ) in the computed electric field waveform in contrast with the measurements . Figure 7 12. The current waveforms (with continuing current compon ents) used for computing the electric field waveforms shown in Figure 7 13.

PAGE 128

128 In Figure 7 11, there are offsets roughly ranging from 1 to 6 V/m between the tails of computed and measured electric field waveforms, which result from the insufficiently large el ectrostatic component in the computed electric field waveforms. In order to better reproduce the electrostatic field component we added continuing current like components to the input current waveforms, as shown in Figure 7 12. Calculations of fields were repeated with the modified current waveforms, and the results are shown in Figure 7 13 with the model parameters being summarized in Table 7 3 For the initial half cycle and the opposite polarity overshoot, there is a good agreement between the computed electric field waveforms and their orresponding measured electric field waveforms. On the other hand, the following oscillatory tail is not well produced (it is less pronounced in the computed waveforms) although the offset see n in Figure 7 11 is largely eliminated in Figure 7 13. Table 7 3. Model input parameters used for computing electric field waveforms shown in Figure 7 13 and NLDN reported peak current for the 7 strokes a (kA) g 1 (10 5 ) g 2 (10 5 ) t g v I cc (A) NLDN reported peak current (kA) 1593 1 4.0 35 4.1 0.3 1 0.90c 700 7.6 1593 2 3.6 35 6.5 0.3 1 0.85c 500 5.7 1593 4 5.0 35 9.0 0.3 1 0.55c 1500 6.5 1593 5 5.2 40 9.0 0.3 1 0.80c 500 6.6 1593 6 5.1 35 9.0 0.3 1 0.80c 700 6.7 1594 2 4.8 35 9.0 0.3 1 0.80c 150 6.5 1594 3 7.2 30 10 0.3 1 0.90c 650 10.2

PAGE 129

129 Figure 7 13. The measured and computed electric field waveforms. The corresponding I sc for each event is shown in Figure 7 12. As seen from Figure 7 13, the time interval between the peaks (period) of the oscillatory tail in the measured electric field waveform is not constant. The varying period of oscillations could be caused by the presence of guy wires or/and impedance discontinuities within the tower, which are not reproduced in our simplified model. It is also conceivable that the lightning channel could be involved. The NLDN reported peak current for the seven strokes are also listed in Table 7 3 for comparison with the peak currents of I sc For all the events, the NLDN over est imated the peak current and the overestimation range from 27% to 90% with an average of 45%.

PAGE 130

130 7.4 Summary The model proposed by Baba and Rakov [2005] was used to compute the electric field waveforms produced by lightning strike to tower. Effects of each model parameter on the electric field wavefo rm were examined. It was found that the narrow bipolar electric field waveform for lightning strikes to tower can be produced by using asymmetric Gaussian pulse as channel base current, rather than the typical return stroke current waveform. The rise and f all times had significant impacts on the electric field wave form With decreasing the rise time of the current waveform, the width of the initial half cycle of the electric field waveform becomes smaller and the peaks of both the initial half cycle and the first opposite polarity overshoot become greater. The magnitude of the first opposite polarity overshoot and the following oscillations become larger as the sharpness of the falling edge increases. The magnitudes of the initial half cycle and the followin g oscillations are increasing with increasing reflection coefficients at tower top and bottom. For the taller strike objects, the electric field is more enhanced and the width of the initial cycle and period of the following oscillations are larger. As the return stroke speed become larger, the ratio of magnitudes of the initial and second half cycles become larger. Compared to other model parameters, return stroke model had relatively small effect on the electric field waveforms. We also tried to reproduc e the narrow bipolar electric field waveforms for lightning striking the 257 m tower in Florida. Using asymmetric Gaussian pulses with continuing current like components superposed on the tails, t he computed electric field waveform s matched well t he corres ponding measured waveform s for the initial half cycle and the opposite polarity overshoot as well as for the late time electrostatic field offset However, due to the selection of the small reflection coefficient at the tower top, the current in the tower is significantly reduced, which

PAGE 131

131 results in no observable oscillations. Compared to the peak currents of I sc the NLDN reported peak currents for the 7 strokes striking tower were 45% higher on average.

PAGE 132

132 CHAPTER 8 NA TION AL LIGHTNING DETECTION NETWORK RESPO NSES TO NATURAL LIGHTNING BASED ON GROUND TRUTH DATA ACQUIRED AT LOG 8 .1 Literature Review The U.S. National Lightning Detection Network (NLDN) has more than 100 sensors installed in the contiguous USA with the typical separation distance of 300 350 km [ Cummins and Murphy 2009] Both the time of arrival (TOA) and magnetic direction finding (MDF) techniques are used. The NLDN reports both cloud (IC) and cloud to ground (CG) lightning discharges, which are classified based on the magnetic field waveform criteria. In general terms, pulses wider than a certain threshold are interpreted as being produced by return strokes (RSs) in s are attributed to cloud flashes and labeled parameter classification method was implemented in the course of the 2013 upgrade. Since any CG flash involves some cloud discharge activity (notably the preliminary breakdown process) CG flashes. Due to a large variation of pulse parameters for either cloud or ground discharges, cloud The detection efficiency is usually defined as the percentage of total ground truth events that were detected by the lightning locating system. For the CG lightning, estimation of the stroke detection efficiency is straightforwa rd, since each stroke involves a cloud to ground channel, which can be observed in optical records. However, for cloud discharges, the detection efficiency is more difficult to define since they mainly occur inside the cloud and do not have readily identif iable features. Rakov [2013] state

PAGE 133

133 Biagi et al. [2007] using video camera records, studied the per formance characteristics of the NLDN in Southern Arizona, Oklahoma, and Texas. The ground flash detection efficiency was found to be 93% in Southern Arizona and 92% in Texas/Oklahoma, with the corresponding stroke detection efficiency being 76% and 85%. Fleenor et al. [2009] who additionally used electric field records from Los Alamos Sferic Array (LASA), conducted a similar field campaign in the region of Colorado Kansas Nebraska (U.S. Central Great Plains). They found, based on the LASA field waveforms, that 54% of NLDN reported CG strokes were actually cloud pulses. Cummins and Murphy [2009] found that the NLDN classification accuracy var ies from region to region and that for regions with higher frequency of positive lightning the classification accuracy tends to be lower. Also, rocket triggered lightning data have been used to evaluate the NLDN performance characteristics in the Florida r egion [ Jerauld et al. 2005; Nag e t al. 2011; Mallick et al. 2014] For 2004 2014, Mallick et al. [2014] found the ground flash and stroke detection efficiencies to be 94 % and 75 % respectively. The strokes in rocket tr iggered lightning are similar to regular subsequent strokes in natural lightning. Hence, the 75% stroke detection efficiency value cited above should be an underestimate for natural lightning, since the first strokes in natural lightning tend to be larger than subsequent ones. Information about NLDN responses to cloud discharge activity is rather limited compared to cloud to ground lightning and may be outdated due to system upgrades (particularly the latest one completed in 2013 [ Nag et al. 2014] ). As reported by Cummins and Murphy [2009] in 2006 the NLDN cloud flash detection efficiency was in the range of 10 20%. Wilson et al. [2013] reported that the NLDN typically detected 1 3 cloud pulses per flash prior to the 2013 upgrade. Zhang et al. [2015] who used video and VHF light ning mapping array (LMA) observations, reported that the NLDN cloud flash detection efficiency in 2012 was 29% and

PAGE 134

134 increased to 41% in 2013, after the upgrade. From a more recent study based on using LMA data as reference, Murphy and Nag [2015] reported the cloud f lash detection efficiency in 2014 to be in the 50 60% range. Nag et al. [2010] found that the NLDN detection efficiency and classification accuracy for 157 compact intracloud discharges (CIDs) were 96% and 95%, respectively. Note that CIDs produce VLF/LF field pulses that are comparable in magnitude to higher intensity return stroke pulses. Th e focus of this chapter is on the NLDN detection efficiency (DE) and classification accuracy (CA) of cloud discharge activity based on the ground truth dataset containing 153 IC events (identified by sequences of electric field pulses not accompanied by ch annels to ground) recorded at the Lightning Observatory in Gainesville (LOG), Florida. Additionally, a ground truth dataset of 367 CG strokes recorded at LOG will be used to evaluate the NLDN DE and CA for CGs after the 2013 upgrade. In this upgrade, the p revious IMPACT (Improved Accuracy through Combined Technology) and LS7001 sensors were replaced by Vaisala's LS7002 sensors with enhanced sensitivity to low amplitude signals. By using pulse onset corrections, the LS7002 can better determine the arrival ti me of electromagnetic pulse, which improves the location accuracy. Further, as noted above, a multi parameter classification method was implemented. Also, a new algorithm, called burst processing, is presently used to locate individual pulses in the pulse train. More detailed information on this upgrade can be found in the works of Buck et al. [2014] and Nag et al. [2014] 8 2 Data and Methodology Simultaneous electric field, electric field derivative (dE/dt), and high speed (HS) video camera records were used in this study. All the records were obtained at the LOG, Florida, in the summer of 2014. The experimental setup in this study was same as the one that was introduced in Section 6.2.

PAGE 135

135 pulses produced by either IC or CG flash that (1) h ad waveshapes clearly different from those characteristic of close return strokes and (2) were not associated with channels to ground in the corresponding HS video camera record (the camera had about 185 o wide field of view). The characteristic features of close RS electric field waveform include the initial (predominantly radiation) peak and the following electrostatic ramp. In order to be counted as a pulse in a given sequence, the pulse had to meet two requirements: 1) the amplitude of the pulse exceeds twice the noise level and 2) the time separation from the preceding pulse is less than 200 ms. We assumed that the interpulse interval in an IC event was unlikely to exceed 200 ms since the total cloud flash duration is usually less than some hundreds of m illiseconds [ Rakov and Uman 2003, chapter 8 ]. Particularly in the case of CG flashes, there could be multiple IC events in a single 1 s record, when they contained multiple strokes. In the latter case, pulse sequences occurring between the return strokes and after the last stroke were treated as individual IC events after the s record. Our ground 8 1) inc complete IC flashes and 127 IC events that occurred in 76 CG flashes (70 negative and 6 events after first RS fter the last stroke). Krider et al. [1975] and Rakov et al. [1996] and CA for these two types of IC events were computed separately, in addition to the three main

PAGE 136

136 Examples of isolated IC events, PB pulse trains, and regular pulse bursts are shown in Figures 8 1, 8 2, and 8 3, respectively. Figure 8 1. Example of an isolated IC event. No channel to ground was observed in the corresponding high speed video camera record. The overall record of the event is shown in panel a). Four NLDN detected cloud pulses, labeled C1 to C4, clustered in the initial portion of the event, who se expansion is shown in panel b). Expansions of the four NLDN detec ted pulses are shown in panels c) to f). The pulses were located by the NLDN at distances of 22 to 28 km from the LOG.

PAGE 137

137 Table 8 1. Summary of the Ground Truth Dataset for IC Events Event Type Isolated IC Events IC Events Before First RS IC Events After First RS All IC Events PB Pulse Trains Regular Pulse Bursts (RPBs) Sample Size 26 58 69 153 24 19 Geometric Mean Duration (ms) 50 4 23 69 64 2.7 1.3 Fig ure 8 2. Examples of PB pulse trains in negative (left panels) and positive (right panels) CG flashes. The top panels show the PB pulse trains in their entirety. The bottom panels show the largest pulses in those two trains on an expanded (100 s) t ime scale. The pulses shown in c) and d) were located by the NLDN at 20 and 30 km from LOG, respectively, and were both misclassified as CGs.

PAGE 138

138 Figure 8 3. An example of regular pulse burst (RPB) that occurred in the later stage of a K change. No RPBs were recorded by t he NLDN. As seen in Table 8 64 ms. For 24 PB pulse trains it was 2.7 ms and 1.3 ms for 19 RPBs. The IC event duration was limited by the electric field record length, which was 1 s. The histogram of event duration is shown in Figure 8 4. Sources of most (85%) of the cloud pulses were reported by the NLDN to be at distances less than 30 km from the LOG. Note, however, that no ground truth information on source locations was

PAGE 139

139 available and, hence, the NLDN location accuracy for IC events could not be evaluated. Also, we do not know the d istribution of source intensities for the events in our ground truth dataset. Figure 8 4. Histogram of durations of 153 IC events. Our methodology was as follows (Figure 8 5). We first identified in our 1 s long electric field records the start and the return stroke type pulses, for which no channel to ground was observed by our HS video camera). The onset of the first pulse and the end of the last pulse (each exceeding twice the noise level) in the pulse sequence 8 1a). Then we searched NLDN data within that time window (between the start and the end of the IC event) and within 40 km of the LOG. If the NLDN reported n o pulses corresponding to

PAGE 140

140 ed as the fraction of NLDN Figure 8 5. Flow chart used to determine the detection efficiency and classification accuracy for IC events. NLDN responses to individual pul ses in IC events were not evaluated in this study, since the number of ground truth events would be less certain than in the case of more readily identifiable multi pulse IC events (pulse sequences). Such identification is particularly straightforward for PB pulse trains (Figure 8 2) and RPBs (Figure 8 3). Note that the NLDN data

PAGE 141

141 used in this study contained information for individual cloud pulses and CG strokes, and that the NLDN did not group individual pulses into flashes or other multi pulse events; thi s was done by us in the ground truth data and then NLDN responses (or lack of them) to those flashes/events were determined. Note also that our pulse grouping algorithm, described at the beginning of this section, is different from that used by Murphy and Nag [2015] Table 8 2. Summary of the Ground Truth Dataset for CG Strokes Stroke Type Negative First Strokes Negative Subsequent Strokes All Negative Strokes Positive First Strokes Positive Subsequent Strokes All Positive Strokes Total Sample Size 84 257 341 21 5 26 367 The ground truth dataset for CGs includes 367 strokes recorded by both the electric field measuring systems and HS video camera. The channel to ground was unambiguously documented for each of those strokes. Similar to IC events, no independe nt information on actual source location and its intensity was available. We believe that our requirement of simultaneous capturing of CGs by both optical and electric field measuring systems serves to reduce any potential bias. Most of the CG strokes were reported by the NLDN within 20 km of LOG. A summary of the CG dataset is given in Table 8 2. The 367 strokes were from 112 negative, 20 positive, and 2 bipolar CG flashes. The percentage of positive flashes was 15%, higher than average for summer thunders torms. Note that the number of negative first strokes (84) is smaller than the number of negative flashes (112) due to the fact that some first return strokes were not included in the ground truth dataset since they were outside the field of view of our HS video camera, even though they were identified in our electric field data. An example of ground truth negative CG flash (both video and electric field records) is shown in Figure 8 6. Out of the 367

PAGE 142

142 strokes, 39 were from single stroke flashes and the ot her 328 strokes were from multiple stroke flashes. Single stroke flashes constituted 20% of CGs and 85% of +CGs. Figure 8 6. An example of ground truth data for a two stroke CG flash. The overall flash ele ctric field record is shown in a) and expansi ons for the two stro kes (RS1 and RS2) are shown in b) and d) on 600 s and 350 s scales, respectively. The corresponding video image s (single frames) are shown in c) and e). For the first stroke (RS1), the second frame is shown because the first frame is saturated. The NLDN reported that the RS1 occurred at 1.7 km and had peak current of 78 kA, while the second stroke (RS2) was missed. The peak current of RS2 was estimated to be 3.4 kA based on its field peak relative to that of RS1.

PAGE 143

143 By using a 2 ms tim e window (1 ms relative to the GPS time of ground truth stroke), we identified all the NLDN reported events (if any) in that time window and within 40 km of the LOG. If no events were reported by the NLDN in the search window, we regarded this stroke as a missed event. If a G pulse was reported in the window, we regarded this stroke as a correctly classified event. If a C pulse was reported in this window or the reported G pulse was assigned incorrect ( opposite) polarity, we regarded this stroke as a miscl assified event. For the rare cases of multiple pulses reported by the NLDN in the 2 ms window, we used the pulse whose timing was closest to that of the ground truth stroke. It is worth making a comment regarding our selected 40 km search radius. Strictly speaking, we cannot rule out the situation when the source that produced the electric field pulse in our record was farther than 40 km from LOG. The trigger threshold of our electric field measuring system was empirically set to provide triggering by ligh tning events within 20 km or so. We have chosen the 40 km search radius to cover the expected source locations with a outside the 40 km search radius. If th is did happen and the event was reported by the NLDN, the detection efficiency estimated in our study wo uld be somewhat underestimated. 8 3 Analysis and Discussion 8 3 .1 Detection Efficiency and Classification Accuracy of IC Events Out of the total of 153 IC events, 26 were isolated IC events that could be viewed as complete IC flashes, one example of which is shown in Figure 8 1, 58 were IC events before first return stroke, and 69 were IC events after first return stroke. The overall detection efficiency and classification accuracy of IC events were 33% and 86%, respectively. More detailed results for the DE, CA, and the average numbers of NLDN reported cloud pulses per detected IC event (minimum number of pulses is one) are given in Table 8 3. DE for isol ated IC events was 73%,

PAGE 144

144 which is 2 3 times higher than that for the other two IC event categories. The DE for cloud flashes reported by Murphy and Nag [2015] was about 50 60%, which is somewhat lower than the 73% found for our complete IC flashes, but higher than t he 33% for all IC events in our study. The average number of NLDN reported cloud pulses per detected IC event was found to be 2.1. Table 8 3. Summary of the NLDN Detection Efficiency (DE) and Classification Accuracy (CA) for IC Events For complete IC flashes, the average number of NLDN reported cloud pulses was 2.9 per detected event, a nd the maximum number of reported pulses was 12, these numbers being higher than their counterparts for the other two IC event categories. Classification accuracies for isolated IC events, IC events before first RS, and IC events after first RS are 95%, 88 %, and 73%, respectively. Note that our sample sizes are not very large, especially for isolated IC events, so further studies are needed to reduce statistical uncertainties. Due to very small pulse amplitudes, none of the 19 regular pulse bursts (in both IC and CG flashes) was detected by the NLDN. Out of the 24 preliminary breakdown pulse trains in CG flashes, 11 (46%) were detected and 9 (82%) of the 11 were correctly classified as cloud events. The two misclassified events are relatively high intensity PB pulse trains, one of which preceded

PAGE 145

145 the negative return stroke and the other one occurred before the positive return stroke. Those misclassified PB pulse trains are shown in Figures 8 2a and 8 2b. The largest pulses (shown in Figures 8 2c and 8 2d) we re incorrectly reported by the NLDN as a 45 kA CG stroke and a 30 kA +CG stroke, respectively. Table 8 4. Summary of the NLDN Detection Efficiency (DE) and Classification Accuracy (CA) for CG Strokes Stroke Type DE CA Negative first strokes 98% (82/84) 96% (79/82) Negative subsequent strokes 90% (231/257) 90% (208/231) Positive first strokes 100% (21/21) 95% (20/21) Positive subsequent strokes 100% (5/5) 100% (5/5) All negative strokes 92% (313/ 3 41) 92% (287/ 3 13) All positive strokes 100% ( 2 6/26) 9 6% ( 2 5/26) All first strokes 98% (103/105) 9 6 % (9 9 /103) All subsequent strokes 90% (236/262) 90% (213/23 6 ) All strokes combined 92% (339/ 36 7) 92% (312/339) 8 3 .2 Detection Efficiency and Classification Accuracy of CG Strokes Out of the 367 positive an d negative CG strokes, 28 were missed by the NLDN. For the 339 detected strokes, 312 were correctly classified as CGs with correct polarity and 27 were misclassified as cloud pulses. No CG strokes were reported with incorrect polarity. The resultant stroke detection efficiency (DE) is 92% and classification accuracy (CA) is 92%. Our results for DE and CA for different categories of CG strokes are summarized in Table 8 4. One can see from the table that both DE and CA of +CGs are higher than those of CGs, a nd that DE and CA of first strokes are higher than those of subsequent strokes. Both DE and CA for the only strokes in single stroke flashes are 100% (N=39), while for first strokes (N=78) in multiple stroke

PAGE 146

146 flashes they are 97% and 93%, respectively. For all strokes (first and subsequent strokes combined) in multiple stroke flashes, the DE and CA are 92% and 91%, respectively. NLDN DE and CA for CG strokes obtained in different studies are summarized in Table 8 5. Our results for negative subsequent strok es, DE = 90% and CA = 90%, can be compared with their counterparts (75% and 96%) for negative strokes in rocket and wire triggered lightning [ Mallick et al. 2014] which are thought to be similar to subsequent strokes in natural lightning. For 231 NLDN detected negative subsequent strokes in this study, the GM NLDN reported peak current was 17 kA vs. 12 kA for 290 strokes in rocket triggered lightning studied by Mallick et al. [2014]. Thus, the higher DE in our study can be, at least in part, associated with higher peak currents in our ground truth dataset. Another possible reason for this discrepancy is the fact that the NLDN was upgraded between 2004 2012 (the study of Mallick et al. [2014 ]) and 2014 (this study ). Note that the classification accuracy for CG strokes found in this study cannot be generalized to the entire NLDN since it is known to vary by region and by storm [e.g., Cummins and Murphy 2009] 8. 4 Summary The NLDN detection efficiency (DE) and classification accuracy (CA) for cloud discharge activity (IC events) and CG strokes in Florida were estimated by using the electric field and optical data acquired at LOG. For 153 ground truth IC events, the DE and CA were 33% (50/153) and 86% (43/50), respectively. The average number of NLDN reported cloud pulses per detected IC event was 2.1. Comp ared to IC events associated with CG flashes, isolated IC events (complete IC flashes) were found to have higher DE (73%), CA (95%), and average number of NLDN reported cloud pulses (2. 9 ). Out of the 24 preliminary breakdown pulse trains

PAGE 147

147 in CG flashes, 11 (46%) were detected and 9 (82%) of the 11 were correctly classified as cloud events. None of the 19 regular pulse bursts was detected. For CG strokes, the DE and CA were 92% (339/367) and 92% (312/339), respectively. Both DE and CA for +CGs are higher tha n those of CGs, and DE and CA for first strokes are higher than those for subsequent strokes. The DE for negative subsequent strokes was 90% (GM peak current=17 kA), which is appreciably higher than the 75% estimated based on the rocket and wire triggered lightning data (GM peak current=12 kA). The CA for negative subsequent strokes in our study was 90%, which is somewhat lower than the 96% estimated using the triggered lightning data. Note that the results of the present study correspond to the Florida r egion and to the NLDN configuration and settings that existed in the summer of 2014.

PAGE 148

1 48 Table 8 5. NLDN DE and CA for CG strokes obtained in different studies Reference Jerauld et al. [2005] Biagi et al. [2007] Biagi et al. [2007] Fleenor et al. [2009] Nag et al. [2011] Mallick et al. [2014] This study Type of lightning Triggered Natural Natural Natural Triggered Triggered Natural Time period 2001 2003 2003 2004 2003 2004 2005 2004 2009 2004 2012 2014 Region Florida Arizona Texas Oklahoma Colorado Kansas Nebraska Florida Florida Flo rida Number of strokes 159 (negative subsequent*) 3620 (positive and negative, first and subsequent) 882 (positive and negative, first and subsequent) 547 (positive and negative, first and subsequent) 139 (negative subsequent) 326 (negative subsequent) 36 7 (positive and negative, first and subsequent) Stroke DE 60% 68% 77% 84% 76% 75% 92% Stroke CA 44% 96% 92% All triggered lightning strok es are classified as subsequent

PAGE 149

149 CHAPTER 9 EVALUATION OF ENTLN PERFORMANCE CHARACTERISTICS BASED ON THE NATURAL AND ROCKET TRIGGERED LIGHTNING DATA ACQUIRED IN FLORIDA 9 .1 Literature Review The Earth Networks Total Lightning Detection Network (ENTLN) consists of more than 1500 wideband (1 Hz to 12 MHz) sensors deploy ed in more than 40 countries around the world, including North and South America, Europe, Africa, Asia, and Australia. More than 900 hundred sensors are presently installed in the contiguous United States. The sensors record electric field waveforms produc ed by lightning and send them to the central server via the Internet. By using the time of arrival technique, the ENTLN can report location and time of each lightning produced pulse it detects. For each pulse, the polarity and type of discharge (either CG or IC) are determined based on the electric field pulse polarity (initial half cycle for bipolar pulses) and waveshape. All waveforms from ENTLN sensors are saved and can be used for reprocessing in the future. More information on the ENTLN can be found in Liu and Heckman [2011] Mallick et al. [2015] evaluated the ENTLN performance by using as ground truth rocket triggered lightning data acquired in Florida in 2009 to 2012. Two different ENTLN datasets were evaluated in that study. The first dataset was prod uced by the old processor that was in service in June 2009 to August 2012. The second dataset was produced by rerunning the same raw data (saved electric field waveforms) through the new processor that was put in service in November 2012. They found that t he stroke detection efficiency and classification accuracy were 49% and 40% for the old processor and 67% and 48% for the new one. For the new processor, the medians of the location error and peak current estimation error were found to be 760 m and 19%, re spectively. in June 2014. Compared to previous ENTLN processors, this one had a decreased time window

PAGE 150

150 for searching for pulses, which served to increase the number of events seen by the ENTLN. The August 2015. This latter (currently operating) processor features a new lightning classification algorithm, which uses multiple wavefo rm parameters to distinguish between CGs and ICs. Figure 9 1. Locations of LOG ( gray square ) CB (yellow square) and ENTLN sensors (red circles) in the Florida region. In this paper, the ENTLN performance in the Florida region is evaluated by using n atural cloud to ground lightning data recorded at the Lightning Observatory in Gainesville (LOG), Florida, in 2014 and 2015. Additionally, rocket triggered lightning data acquired at Camp

PAGE 151

151 Blanding (CB), Florida, during the same time period were used for ev aluation. The locations of LOG and CB (45 km apart) are shown in Figure 9 1. Also shown are the locations of ENTLN sensors deployed in the Florida region. The total number of sensors in Figure 9 1 is 84. Similar to the work of Mallick et al. [2015] two datasets produced using two processors were evaluated. The same originally recorded field waveforms were used as in put to the two processors. The flash detection efficiency (DE), flash classification accuracy (CA), stroke DE, and stroke CA were examined for both natural lightning and rocket triggered lightning, while the location errors and peak current estimation erro rs could be estimated only for rocket triggered lightning. The results are important for proper interpretation of ENTLN data that are used in a variety of meteorological and geophysical studies and amount to the calibration of the network against the groun d truth data. The developed methodology can be applied to other lightning locating systems 9 2 Data and Methodology Simultaneous electric field, electric field derivative (dE/dt), and high speed (HS) video camera records, obtained at LOG, Florida from 201 4 and 2015, were used as ground truth natural lightning data in this study. The experimental setup at LOG in this study was same as the one that was introduced in Section 6.2. For rocket triggered lightning, channel base currents measured at Camp Blanding and their GPS timing were used as ground truth data in this study. Also, the position of the rocket launcher (lightning termination point on ground ) was known precisely and was used as ground truth for estimating location errors. The lightning triggering p rocess is described in Rakov and Uman [2003, Chapter 7] More information of channel base current measurements at CB can be found in Section 2.3. For this study, the high current channel (with highest mea suring range) was used to record currents up to 60 kA with a resolution of 3 A,

PAGE 152

152 sampling interval of 10 ns, and record length of 2 s. Records from this channel were used to measure the peak current of each return stroke in this study. The ground truth dat asets for both natural lightning and rocket triggered lightning are summarized in Table 9 1. In the summers of 2014 and 2015, electric fields for a total of 219 natural cloud to ground lightning flashes (175 negative 39 positive, and 5 bipolar flashes) co ntaining 608 strokes were recorded at LOG. The channel to ground was unambiguously documented for each of those strokes, although the termination point could be obscured by trees. No ground truth information on the strike point nor on peak current is avail able for natural lightning. Note that strokes with characteristic CG electric field waveforms but occurring outside of the field of view of our camera are not included in this study. Table 9 1. Summary of ground truth datasets for natural and rocket trigg ered lightning acquired in Florida and used in this study Year Natural lightning Rocket triggered lightning Number of flashes Number of strokes Number of flashes Number of strokes 2014 134 367 18 78 2015 85 241 18 97 Total 219 608 36 175 A total of 36 flashes containing 175 negative strokes were triggered at Camp Blanding in the summers of 2014 and 2015. Channel base current records are available for 171 strokes. For one flash containing 4 strokes that were obtained from an altitude trigger [ Lalande et al. 1998; Rakov et al. 1998] due to the breakage of the wire, no current record is available since the lightning channel did not attach to the instrumented launcher. These 4 strokes are not included in the evaluation of either current esti mation errors or location errors. The histogram of ground

PAGE 153

153 truth peak currents for the 171 strokes in triggered lightning is shown in Figure 9 2. As noted in Section 2.2, the position of the rocket launcher was used as the ground truth location of the chann el termination point on the ground Figure 9 2. Histogram of peak currents for 171 return strokes in 18 flashes triggered using the rocket and wire technique at Camp Blanding, Florida. The methodology to determine the detection efficiency and classific ation accuracy for natural CGs used here has been developed by Zhu et al. [2016] and is shown in the form of a flowchart in Figure 9 3. Each ground truth natural CG had clearly identified cloud to ground channel in the high speed video record and characteristic feat ures of cloud to ground stroke in the corresponding electric field waveform. Since most our strokes were within 20 km of LOG, the return stroke electric field waveforms had an initial (radiation) peak followed by an

PAGE 154

154 electrostatic ramp. The timing of the r eturn stroke was determined by using the time of the pulse peak relative to the trigger time of the record, which was GPS time stamped. By using a 2 ms time window (1 ms relative to the GPS time of ground truth stroke), we identified all the ENTLN reporte d events (if any) in that time window and within 40 km of the LOG. If no events were reported by the ENTLN in the search window, we regarded this stroke as a missed event. If a CG was reported in the window, we regarded this stroke as a correctly classifie d event. If a cloud pulse was reported in this window or the reported stroke was assigned incorrect (opposite) polarity, we regarded this stroke as a misclassified event. For the rare cases of multiple pulses reported by the ENTLN in the 2 ms window, we us ed the pulse whose timing was closest to that of the ground truth stroke. A similar methodology was used for the rocket triggered lightning dataset, except that the channel base current records were used and the search area was centered at CB. Stroke DE is the percentage of ground truth strokes that were detected by the ENTLN. Flash DE is defined as the percentage of ground truth flashes in which at least one stroke was detected. The stroke CA is the percentage of ENTLN detected ground truth strokes that we re correctly reported as CGs. Flash CA is the percentage of ENTLN detected ground truth flashes in which at least one stroke was correctly classified as a CG. As noted above, two different ENTLN processors were evaluated in this study. The ENTLN datasets for evaluation were produced by running the old/new processor with previously saved waveforms as input, as if each of those processors was in service during the time period when our ground truth data were collected

PAGE 155

155 Figure 9 3. Flowchart showing the met hodology to determine the detection efficiency and classification accuracy for natural cloud to ground lightning. 9 3 Analysis and Discussion 9 3 .1 Natural L ightning Following the procedure outline in Figure 9 3, flash DE, flash CA, stroke DE, and stroke C A were estimated using ground truth data for 219 natural lightning flashes containing 608 strokes for old and new ENTLN processors. The results are summarized in Table 9 2. For the old processor, the ENTLN flash DE, flash CA, stroke DE, and stroke CA were found to be 99%, 91%, 97%, and 68%, respectively, and for the new processor they were 99%, 97%, 96%, and 91%. Note significant improvement in the flash/stroke CA, while the flash/stroke DEs remain essentially the same. The average number of strokes per fl ash for ground truth natural lightning dataset is 2.8, which is smaller than the 4.6 reported for negative CG flashes in Florida by Rakov and Uman [1990]

PAGE 156

156 Table 9 2. Summary of the ENTLN perfo rmance characteristics evaluated using natural lightning data Processor Old New Number of flashes 219 219 Number of strokes 608 608 Flash DE 99% 99% Flash CA 91% 97% Stroke DE 97% 96% Stroke CA 68% 91% Table 9 3. Summary of the estimated values of ENTLN stroke DE and CA for different types of strokes in natural lightning Stroke type Number of strokes Stroke DE Stroke CA GM Peak current for ENTLN detected strokes (kA) Old processor New processor Old processor New processor Old processor New pro cessor First negative strokes 139 99% 99% 86% 96% 31 30 First positive strokes 40 98% 98% 85% 90% 47 47 Subsequent negative strokes 419 96% 95% 60% 91% 18 18 Subsequent positive strokes 10 100% 100% 60% 60% 26 25 All first strokes 179 99% 99% 86% 95% 33 34 All subsequent strokes 429 96% 95% 60% 90% 18 18 All negative strokes 558 97% 96% 67% 92% 20 21 All positive strokes 50 98% 98% 80% 84% 41 41 All strokes combined 608 97% 96% 68% 91% 21 22

PAGE 157

157 Stroke DE and stroke CA for different types of strokes in natural lightning are summarized in Table 9 3. Also given in Table 9 3 is the geometric mean (GM) peak current reported by the ENTLN for detected strokes (including misclassified ones). Both DE and CA for first strokes (positive and negative strokes co mbined) are higher than those for subsequent strokes (positive and negative strokes combined), which is likely due to the higher peak current for the first strokes. For the new processor, the GM peak currents for ENTLN detected first strokes and subsequent strokes are 34 kA and 18 kA, respectively. The GM ENTLN reported peak currents for misclassified strokes were 11 kA (N=189) for the old processor and 8 kA (N=52) for the new processor, while the corresponding values for correctly classified strokes were 3 0 kA (N=399) and 24 kA (N=532). It appears that strokes with higher peak current (inferred from measured electric field peak) are more likely to be both detected and correctly classified by the ENTLN. One possible reason could be that return strokes with h igher peak current have wider field waveforms at the measurement threshold level. Also, field waveforms of strokes with higher peak currents are less affected by noise, so that more accurate waveform characteristics can be obtained, which should improve cl assification accuracy 9 3 .2 Rocket Triggered L ightning Flash DE, flash CA, stroke DE, and stroke CA were examined for 36 rocket triggered lightning flashes containing 175 strokes. The results are summarized in Table 9 4. The ENTLN detected all the flashes for both old and new processors. For the new processor, only one flash was misclassified vs. three for the old processor Compared to the old processor, the new processor stroke DE increased from 94% to 97% and the stroke CA increased from 42% to 86%. It is known that negative strokes in rocket triggered lightning are similar to regular subsequent negative strokes in natural cloud to ground lightning [ Rakov and Uman 2003, chapter 7] The new processor stroke DE (97%) for rocket triggered lightning is slightly higher than that (95%)

PAGE 158

158 for subsequent negative strokes in natural lightning, as seen in Table 9 3. The new processor stroke CA for rocket triggered lightning is 86%, which is lower than 91% for the subsequent negative strokes in natural lightning. As seen from Table 9 4, the GM ground truth peak current for detected strokes is about a factor of 5 greater than that for undetected strokes. Similarly, the GM ground truth peak current for correctly clas sified strokes is about twice higher than that for misclassified strokes Table 9 4. Summary of the ENTLN performance characteristics evaluated using rocket triggered lightning data Processor Old New Number of flashes 36 36 Number of strokes 175 175 GM channel base peak current (kA) 11.6 11.6 Number of detected flashes 36 36 Number of correctly classified flashes 33 35 Number of detected strokes 169 169 Number of correctly classified strokes 71 145 Flash DE 100% 100% Flash CA 92% 97% Stroke DE 97% 97% Stroke CA 42% 86% GM ground truth peak current for undetected strokes (kA) 2.6 2.4 GM ground truth peak current for detected strokes (kA) 12.2 12.3 GM ground truth peak current for misclassified strokes (kA) 9.8 6.5 GM ground truth peak curren t for correctly classified strokes (kA) 16.5 13.6 Median absolute current estimation error 15% 15% Median location error (m) 205 215 For rocket triggered lightning strokes, peak current estimation errors and location errors were also examined. The hist ograms for absolute (unsigned) peak current estimation error (|I ENTLN I CB |/I CB ), and signed peak current estimation error ((I ENTLN I CB )/I CB ), are shown in Figure 9 4. Figure 9 5 shows the scatterplots of the absolute current estimated by the ENTLN versu s ground truth peak current measured at CB, from which one can also see that the majority of misclassified events had peak currents < 20 kA for the old processor and < 10 kA for the new

PAGE 159

159 processor. Histograms for location error and ENTLN reported stroke loc ations are shown in Figures 9 6 and 9 7, respectively. The median location errors for old and new processors are 205 m and 215 m, respectively Note that in Figures 9 4 to 9 7, the type ( CG, IC, or +IC) for each event was designated by the ENTLN and that the ground truth type for all the events is negative return stroke ( CG) Figure 9 4. Histograms of absolute (upper panels) and signed (lower panels) peak current estimation errors for old (left panels) and new (right panels) processors.

PAGE 160

160 Figure 9 5. Scatterplots of peak current estimated by the ENTLN vs. ground truth peak current measured at CB for old (left panel) and new (right panel) processors. Note that the type ( CG, IC, or +IC) for each event was designated by the ENTLN and that the ground tr uth type for all the events is negative return stroke ( CG). Figure 9 6. Histograms of location error for the old (left panel) and new (right panel) processors.

PAGE 161

161 Figure 9 7. Plots of ENTLN reported locations for old (left panels) and new (right panels ) processors. The lower panels are 2 km by 2 km expansions of the upper panels. The ground truth location (the position of the rocket launcher) is at the origin of coordinates [0,0]. In Table 9 5, the ENTLN performance characteristics evaluated using rock et triggered lightning data in this study are compared with the results found in Mallick et al. [2015] Note that

PAGE 162

162 a total of four processors are compared in that table. One can see from Table 9 5 that the upgrade in November 2012 significantly decreased the current estimation error, which resulted from calibration of ENTLN peak current evaluation formula against dir ectly measured currents for rocket triggered lightning [ Mallick et al. 2013] Further, the upgrade in June 2014 greatly improved the detection efficiency and location accuracy. Finally, the upgrade in August 2015 notably increased the classification accuracy, due to the implementation of new multi parameter classification algorithm Table 9 5. Comparison of ENTLN performance characteristics evaluated for four different processors using rocket triggered lightning data Reference Mallick et al. [2015] This study Processor service time period 2009 to N ov. 2012 Nov. 2012 to June 2014 June 2014 to Aug. 2015 Aug. 2015 to present Time period of ground truth data collection 2009 2012 2009 2012 2014 2015 2014 2015 Number of flashes 57 57 36 36 Number of strokes 245 245 175 175 Flash DE 77% 89% 100% 100% Flash CA 92% 97% Stroke DE 49% 67% 97% 97% Stroke CA 39% 46% 42% 86% Median location error (m) 631 760 205 215 Median absolute current estimation error 51% 19% 15% 15% 9 4 Summary The performance characteristics of the ENTLN were evaluated by us ing as ground truth natural lightning data recorded at LOG and rocket triggered lightning recorded at CB in 2014 and 2015. For 219 natural CG lightning flashes containing 608 strokes and for the new

PAGE 163

163 processor, the flash DE, flash CA, stroke DE, and stroke CA were 99%, 97%, 96%, and 91%, respectively, while they were 99%, 91%, 97%, and 68% for the old processor. The stroke DE and stroke CA for first strokes are higher than those for subsequent strokes. For 36 rocket triggered lightning flashes containing 175 strokes and for the old processor, the flash DE, flash CA, stroke DE, and stroke CA were 100%, 92%, 97%, and 42%, respectively, while their counterparts for the new processor were 100%, 97%, 97%, and 86%. The median values of location error and absolute peak current estimation error were 205 m and 15% for the old processor and 215 m and 15% for the new processor. For both natural and triggered lightning, strokes with higher peak currents (inferred or directly measured, respectively) were more likely to be detected and correctly classified by the ENTLN. Note that the results of the present study correspond to the Florida region and might not be generalized to the entire ENTLN

PAGE 164

164 CHAPTER 10 SUMMARY OF RESULTS AND RECOMMENDATION S FOR FUTURE RESEARCH 10 .1 Summ ary of Results 1. Based on the electric field records of 478 negative cloud to ground lightning flashes acquired at Lightning Observator y in Gainesville (LOG), Florida, w e found that the average number of strokes per flash is 4.6 and the percentage of sing le stroke flash es is 12%. The geometric means of interstroke interval, flash duration, and first to subsequent stroke field peak ratio are 52 ms, 223 ms, and 2.4, respectively. It was found that the detectability of preliminary breakdown pulse trains is af fected by the signal/noise ratio, distance, type of storm, and return stroke peak current The dependences on storm type and peak current were studied for the first time. 2. The s tudy on preliminary breakdown process was extended to 5498 negative high inte nsity ( 50 kA first stroke peak currents ) cloud to ground flashes using an original automated data processing algorithm. For 3077 flashes with detectable PB pulse trains, t he arithmetic mean values of PB pulse train duration, PB RS interval, and PB/RS puls e peak ratio were 2.7 ms, 8.8 ms, and 0.15, respectively. The PB RS interval was found to decrease with increasing RS peak current (Spearman correlation coefficient was statistically significant and equal to 0.80). The range normalized PB pulse peak exhib ited statistically significant positive correlation with the RS peak current, with the Spearman correlation coefficient being 0.48. Thus, it appears that the high intensity negative lightning is characterized by shorter (and, by inference, faster) stepped leaders and more pronounced PB pulse trains 3. Using simultaneous high speed video camera records and electric field measurements, we examined a bipolar flash that started with a negative stroke with a peak current of 101 kA, which was followed by a posi tive stroke with a peak current of 16 kA. From the leader electric

PAGE 165

165 field waveforms, the positive charge source is inferred to be located considerably farther from both the observer and the channel to ground than the negative charge source. The second ( pos itive) stroke was followed by a bipolar continuing current. The average luminosity of the main channel during the positive CC was 3.7 times higher than that during the negative CC, which is likely indicative of about 3.7 times higher current of the positiv e CC. It is likely that upward negative leaders are more heavily branched inside the cloud than the upward positive ones, which makes the upward negative leaders more efficient in collecting the cloud charges and funneling them to the channel to ground. Th e first two strokes (including the continuing current) followed the same channel to ground, whose imaged 2D length was 4.2 km, except for the bottom 115 m, where the paths of the two strokes were slightly different. As of this writing, there is only one p reviously documented case of positive leader following the path of preceding negative stroke. We presented the first leader speed versus height profiles for such an unusual sequence. The average leader speeds for the first (negative) and second (positive) strokes were 4.710 5 m/s and 7.210 5 m/s, respectively. The speed of the positive leader traversing the previous stroke channel after not unduly long (70 ms) interstroke interval is lower than typical speeds of negative leaders following previously formed channels. The speeds of both the negative leader and the positive leader increased as they approached the ground. 4 Simultaneous electric field and high speed video camera records of two flashes, terminating on a 257 m tower were obtained at LOG One of the flashes was an upward negative flash whose upward positive leader initiated from the 257 m tower, located 8.8 km from LOG. It contained six leader/return stroke sequences, all of which developed along the upward leader path and terminated on the 257 m tower. The other flash was a three stroke downward bipolar flash whose first stroke was positive and terminated on a 60 m tower. The two subsequent

PAGE 166

166 strokes were negative and terminated on the same 257 m tower as the six strokes of the first flash All stro kes terminated on the 257 m tower exhibited very similar and unusually narrow bipolar electric field waveforms with damped oscillatory tails. Characteristics of those electric field waveforms are examined and compared to similar observations found in the l iterature. 5 By using the an engineering model developed by Baba and Rakov [2005] the electric field waveform produced at LOG by 7 of the 8 stroke s (for one stroke the electric field signature was saturated) terminated on the 257 m tower was computed using a narrow pulse followed by a steady curr ent tail as the channel base current. Also, the effects of each model parameter (current waveform, return stroke speed, and return stroke model ) on the electric field waveform were investigated. 6. The NLDN detection efficiency and classification accuracy for cloud discharge activity (IC events) and natural CG strokes in Florida were estimated (for the first time) by using the electric field and optical data acquired at LOG. For 153 ground truth IC events, the DE and CA were 33% (50/153) and 86% (43/50), re spectively. The average number of NLDN reported cloud pulses per detected IC event was 2.1. Compared to IC events associated with CG flashes, isolated IC events (complete IC flashes) were found to have higher DE (73%), CA (95%), and average number of NLDN reported cloud pulses (2.9). For 366 CG strokes, the DE and CA were both 92%. Both DE and CA for positive CGs were higher than those of negative CGs, and DE and CA for first strokes were higher than those for subsequent strokes. The DE for negative subsequ ent strokes was 90% (GM peak current=17 kA), which is appreciably higher than the 75% estimated based on the rocket and wire triggered lightning data (GM peak current=12 kA). The CA for negative subsequent strokes in our study was 90%, which is somewhat lo wer than the 96% estimated using the triggered lightning data.

PAGE 167

167 7. The performance characteristics of the ENTLN were evaluated by using as ground truth natural lightning data (for the first time) recorded at LOG and rocket triggered lightning recorded at CB in 2014 and 2015. For 219 natural CG lightning flashes containing 608 strokes and for the new processor, the flash DE, flash CA, stroke DE, and stroke CA were 99%, 97%, 96%, and 91%, respectively, while they were 99%, 91%, 97%, and 68% for the old process or. The stroke DE and stroke CA for first strokes are higher than those for subsequent strokes. For 36 rocket triggered lightning flashes containing 175 strokes and for the old processor, the flash DE, flash CA, stroke DE, and stroke CA were 100%, 92%, 97% and 42%, respectively, while their counterparts for the new processor were 100%, 97%, 97%, and 86%. The median values of location error and absolute peak current estimation error were 205 m and 15% for the old processor and 215 m and 15% for the new pro cessor. For both natural and triggered lightning, strokes with higher peak currents (inferred or directly measured, respectively) were more likely to be detected and correctly classified by the ENTLN. 1 0 .2 Recommendations for Future Research 1. Us e of LMA and high speed video cameras at CB to study the fast stepped leader events and the charge structure of the parent thunderstorm. 2. Characteri zation of two station (LOG and GC) electric field waveforms of different types of natural lightning ( CG + CG) with a view toward testing return stroke models 3. Further analysis of the b idirectional leader development in the two tower flashes terminated on the 257 m tower and look for additional data (high speed video current, and field records) on lightning strikin g towers to improve our understanding of the lightning interaction with tall object s. 4. Modeling of narrow electric field waveforms associated with lightning strikes to different towers for which measured currents are available C onsider different rates of current attenuation

PAGE 168

168 along the lightning channel possible reflection at the channel top and no vertical channel geometry 5 Comparison of wideband electric field waveforms produced by return strokes in natural and rocket triggered lightning, which has important implications for evaluating the performance characteristics of lightning locating systems.

PAGE 169

169 APPENDIX A TWO STATION MEASUREMENTS OF ROCKET TRIGGERED LIGHTNING ELECTRIC FIELD WAVEFORM S (2013 2016) E lectric field waveforms produced by rocket trigg ered lightning at Camp Blanding in 2013 2016 and measured at two stations (LOG GC) are shown in this appendix Two station field measurements are available for a total of 29 flashes containing 153 strokes. All strokes we re negative. Flash ID and number of return strokes are listed in Table A 1. For each stroke, the complete field record of the flash is shown first and followed by expansions o n 1 ms and 200 s time scales for each stroke For all rocket triggered lightning flashes the distances to GC and LOG are 3 km and 45 km, respectively. The LOG electric field s produced by return stroke s in rocket triggered seem to have a faster decay after the peak and sho r ter zero crossing time compared with those f or negative strokes in natural lightning also recor ded at LOG (see Appendix B), which is consistent with the findings reported by Mallick and Rakov [2014] Table A 1. Inventory of two s tation (LOG GC) field measurements for rocket triggered lightning flashes from 2013 to 2016 Flash ID Number of Return Strokes UF 13 31 11 UF 1 3 33 6 UF 13 34 4 UF 14 01 1 UF 14 05 5 UF 14 06 2 UF 14 07 3 UF 14 08 4 UF 14 11 8

PAGE 170

170 Table A 1. Continued Flash ID Number of Return Strokes UF 14 12 7 UF 14 35 4 UF 14 36 2 UF 14 43 7 UF 14 51 8 UF 14 52 5 UF 14 53 5 UF 15 1 1 2 UF 15 12 5 UF 15 15 8 UF 15 20 2 UF 15 25 4 UF 15 26 8 UF 15 38 5 UF 15 39 6 UF 15 41 5 UF 15 42 6 UF 15 43 5 UF 16 04 1 Total Number of Flashes: 2 9 Total Number of Strokes: 15 3

PAGE 171

171 Figure A 1. Two station electric field wav eforms of flash UF 13 31.

PAGE 172

172 Figure A 2. Two station electric field waveforms of the RS1 of flash UF 13 31. T he timing offset between GC and LOG records is an artifact caused by electronics.

PAGE 173

173 Figure A 3. Two station electric field waveforms of the RS2 of flash UF 13 31.

PAGE 174

174 Figure A 4. Two station electric field waveforms of the RS3 of flash UF 13 31.

PAGE 175

175 Figure A 5. Two station electric field waveforms of the RS4 of flash UF 13 31.

PAGE 176

176 Figure A 6. Two station electric field waveforms of the RS5 of flash UF 13 31.

PAGE 177

177 Figure A 7. Two station electric field waveforms of the RS6 of flash UF 13 31.

PAGE 178

178 Figure A 8. Two station electric field waveforms of the RS7 of flash UF 13 31.

PAGE 179

179 Figure A 9. Two station electric field waveforms of the RS8 of flash UF 13 31.

PAGE 180

180 F igure A 10. Two station electric field waveforms of the RS9 of flash UF 13 31.

PAGE 181

181 Figure A 11. Two station electric field waveforms of the RS10 of flash UF 13 31.

PAGE 182

182 Figure A 12. Two station electric field waveforms of the RS11 of flash UF 13 31.

PAGE 183

183 Figure A 13. Two station electric field waveforms of flash UF 13 33.

PAGE 184

184 Figure A 14. Two station electric field waveforms of the RS1 of flash UF 13 33.

PAGE 185

185 Figure A 15. Two station electric field waveforms of the RS2 of flash UF 13 33.

PAGE 186

186 Figure A 16. Two statio n electric field waveforms of the RS3 of flash UF 13 33.

PAGE 187

187 Figure A 17. Two station electric field waveforms of the RS4 of flash UF 13 33.

PAGE 188

188 Figure A 18. Two station electric field waveforms of the RS5 of flash UF 13 33.

PAGE 189

189 Figure A 19. Two station elect ric field waveforms of the RS6 of flash UF 13 33.

PAGE 190

190 Figure A 20. Two station electric field waveforms of flash UF 13 34.

PAGE 191

191 Figure A 21. Two station electric field waveforms of the RS1 of flash UF 13 34.

PAGE 192

192 Figure A 22. Two station electric field waveform s of the RS2 of flash UF 13 34.

PAGE 193

193 Figure A 23. Two station electric field waveforms of the RS3 of flash UF 13 34.

PAGE 194

194 Figure A 24. Two station electric field waveforms of the RS4 of flash UF 13 34.

PAGE 195

195 Figure A 25. Two station electric field waveforms of fl ash UF 14 01.

PAGE 196

196 Figure A 26. Two station electric field waveforms of the RS1 of flash UF 14 01.

PAGE 197

197 Figure A 27. Two station electric field waveforms of flash UF 14 05.

PAGE 198

198 Figure A 28. Two station electric field waveforms of the RS1 of flash UF 14 05.

PAGE 199

199 Fi gure A 29. Two station electric field waveforms of the RS2 of flash UF 14 05.

PAGE 200

200 Figure A 30. Two station electric field waveforms of the RS3 of flash UF 14 05.

PAGE 201

201 Figure A 31. Two station electric field waveforms of the RS4 of flash UF 14 05.

PAGE 202

202 Figure A 32. Two station electric field waveforms of the RS5 of flash UF 14 05.

PAGE 203

203 Figure A 33. Two station electric field waveforms of flash UF 14 06.

PAGE 204

204 Figure A 34. Two station electric field waveforms of the RS1 of flash UF 14 06.

PAGE 205

205 Figure A 35. Two station e lectric field waveforms of the RS2 of flash UF 14 06.

PAGE 206

206 Figure A 36. Two station electric field waveforms of flash UF 14 07.

PAGE 207

207 Figure A 37. Two station electric field waveforms of the RS1 of flash UF 14 07.

PAGE 208

208 Figure A 38. Two station electric field wav eforms of the RS2 of flash UF 14 07.

PAGE 209

209 Figure A 39. Two station electric field waveforms of the RS3 of flash UF 14 07.

PAGE 210

210 Figure A 40. Two station electric field waveforms of flash UF 14 08.

PAGE 211

211 Figure A 41. Two station electric field waveforms of the RS1 of flash UF 14 08.

PAGE 212

212 Figure A 42. Two station electric field waveforms of the RS2 of flash UF 14 08.

PAGE 213

213 Figure A 43. Two station electric field waveforms of the RS3 of flash UF 14 08.

PAGE 214

214 Figure A 44. Two station electric field waveforms of the RS4 of flas h UF 14 08.

PAGE 215

215 Figure A 45. Two station electric field waveforms of flash UF 14 11.

PAGE 216

216 Figure A 46. Two station electric field waveforms of the RS1 of flash UF 14 11.

PAGE 217

217 Figure A 47. Two station electric field waveforms of the RS2 of flash UF 14 11.

PAGE 218

218 Figu re A 48. Two station electric field waveforms of the RS3 of flash UF 14 11.

PAGE 219

219 Figure A 49. Two station electric field waveforms of the RS4 of flash UF 14 11.

PAGE 220

220 Figure A 50. Two station electric field waveforms of the RS5 of flash UF 14 11.

PAGE 221

221 Figure A 51 Two station electric field waveforms of the RS6 of flash UF 14 11.

PAGE 222

222 Figure A 52. Two station electric field waveforms of the RS7 of flash UF 14 11.

PAGE 223

223 Figure A 53. Two station electric field waveforms of the RS8 of flash UF 14 11.

PAGE 224

224 Figure A 54. Two station electric field waveforms of flash UF 14 12.

PAGE 225

225 Figure A 55. Two station electric field waveforms of the RS1 of flash UF 14 12.

PAGE 226

226 Figure A 56. Two station electric field waveforms of the RS2 of flash UF 14 12.

PAGE 227

227 Figure A 57. Two station electric f ield waveforms of the RS3 of flash UF 14 12.

PAGE 228

228 Figure A 58. Two station electric field waveforms of the RS4 of flash UF 14 12.

PAGE 229

229 Figure A 59. Two station electric field waveforms of the RS5 of flash UF 14 12.

PAGE 230

230 Figure A 60. Two station electric field wa veforms of the RS6 of flash UF 14 12.

PAGE 231

231 Figure A 61. Two station electric field waveforms of the RS7 of flash UF 14 12.

PAGE 232

232 Figure A 62. Two station electric field waveforms of flash UF 14 35.

PAGE 233

23 3 Figure A 63. Two station electric field waveforms of the RS1 of flash UF 14 35.

PAGE 234

234 Figure A 64. Two station electric field waveforms of the RS2 of flash UF 14 35.

PAGE 235

235 Figure A 65. Two station electric field waveforms of the RS3 of flash UF 14 35.

PAGE 236

236 Figure A 66. Two station electric field waveforms of the RS4 of fla sh UF 14 35.

PAGE 237

237 Figure A 67. Two station electric field waveforms of flash UF 14 36.

PAGE 238

238 Figure A 68. Two station electric field waveforms of the RS1 of flash UF 14 36.

PAGE 239

239 Figure A 69. Two station electric field waveforms of the RS2 of flash UF 14 36.

PAGE 240

240 Fig ure A 70. Two station electric field waveforms of flash UF 14 43.

PAGE 241

241 Figure A 71. Two station electric field waveforms of the RS1 of flash UF 14 43.

PAGE 242

242 Figure A 72. Two station electric field waveforms of the RS2 of flash UF 14 43.

PAGE 243

243 Figure A 73. Two sta tion electric field waveforms of the RS3 of flash UF 14 43.

PAGE 244

244 Figure A 74. Two station electric field waveforms of the RS4 of flash UF 14 43.

PAGE 245

245 Figure A 75. Two station electric field waveforms of the RS5 of flash UF 14 43.

PAGE 246

246 Figure A 76. Two station el ectric field waveforms of the RS6 of flash UF 14 43.

PAGE 247

247 Figure A 77. Two station electric field waveforms of the RS7 of flash UF 14 43.

PAGE 248

248 Figure A 78. Two station electric field waveforms of flash UF 14 51.

PAGE 249

249 Figure A 79. Two station electric field wavef orms of the RS1 of flash UF 14 51.

PAGE 250

250 Figure A 80. Two station electric field waveforms of the RS2 of flash UF 14 51.

PAGE 251

251 Figure A 81. Two station electric field waveforms of the RS3 of flash UF 14 51.

PAGE 252

252 Figure A 82. Two station electric field waveforms of the RS4 of flash UF 14 51.

PAGE 253

253 Figure A 83. Two station electric field waveforms of the RS5 of flash UF 14 51.

PAGE 254

254 Figure A 84. Two station electric field waveforms of the RS6 of flash UF 14 51.

PAGE 255

255 Figure A 85. Two station electric field waveforms of the RS 7 of flash UF 14 51.

PAGE 256

256 Figure A 86. Two station electric field waveforms of the RS8 of flash UF 14 51.

PAGE 257

257 Figure A 87. Two station electric field waveforms of flash UF 14 52.

PAGE 258

258 Figure A 88. Two station electric field waveforms of the RS1 of flash UF 14 5 2.

PAGE 259

259 Figure A 89. Two station electric field waveforms of the RS2 of flash UF 14 52.

PAGE 260

260 Figure A 90. Two station electric field waveforms of the RS3 of flash UF 14 52.

PAGE 261

261 Figure A 91. Two station electric field waveforms of the RS4 of flash UF 14 52.

PAGE 262

262 Fi gure A 92. Two station electric field waveforms of the RS5 of flash UF 14 52.

PAGE 263

263 Figure A 93. Two station electric field waveforms of flash UF 14 53.

PAGE 264

264 Figure A 94. Two station electric field waveforms of the RS1 of flash UF 14 53.

PAGE 265

265 Figure A 95. Two st ation electric field waveforms of the RS2 of flash UF 14 53.

PAGE 266

266 Figure A 96. Two station electric field waveforms of the RS3 of flash UF 14 53.

PAGE 267

267 Figure A 97. Two station electric field waveforms of the RS4 of flash UF 14 53.

PAGE 268

268 Figure A 98. Two station e lectric field waveforms of the RS5 of flash UF 14 53.

PAGE 269

269 Figure A 99. Two station electric field waveforms of flash UF 15 11.

PAGE 270

270 Figure A 100. Two station electric field waveforms of the RS1 of flash UF 15 11.

PAGE 271

271 Figure A 101. Two station electric field wa veforms of the RS2 of flash UF 15 11.

PAGE 272

272 Figure A 102. Two station electric field waveforms of flash UF 15 12.

PAGE 273

273 Figure A 103. Two station electric field waveforms of the RS1 of flash UF 15 12.

PAGE 274

274 Figure A 104. Two station electric field waveforms of the RS2 of flash UF 15 12.

PAGE 275

275 Figure A 105. Two station electric field waveforms of the RS3 of flash UF 15 12.

PAGE 276

276 Figure A 106. Two station electric field waveforms of the RS4 of flash UF 15 12.

PAGE 277

277 Figure A 107. Two station electric field waveforms of the RS5 of flash UF 15 12.

PAGE 278

278 Figure A 108. Two station electric field waveforms of flash UF 15 15.

PAGE 279

279 Figure A 109. Two station electric field waveforms of the RS1 of flash UF 15 15.

PAGE 280

280 Figure A 110. Two station electric field waveforms of the RS2 of flash UF 15 15.

PAGE 281

281 Figure A 111. Two station electric field waveforms of the RS3 of flash UF 15 15.

PAGE 282

282 Figure A 112. Two station electric field waveforms of the RS4 of flash UF 15 15.

PAGE 283

283 Figure A 113. Two station electric field waveforms of the RS5 of flash UF 15 15.

PAGE 284

284 Figure A 114. Two station electric field waveforms of the RS6 of flash UF 15 15.

PAGE 285

285 Figure A 115. Two station electric field waveforms of the RS7 of flash UF 15 15.

PAGE 286

286 Figure A 116. Two station electric field waveforms of the RS8 of flash UF 15 15.

PAGE 287

287 Fi gure A 117. Two station electric field waveforms of flash UF 15 16.

PAGE 288

288 Figure A 118. Two station electric field waveforms of the RS1 of flash UF 15 16.

PAGE 289

289 Figure A 119. Two station electric field waveforms of the RS2 of flash UF 15 16.

PAGE 290

290 Figure A 120. T wo station electric field waveforms of the RS3 of flash UF 15 16.

PAGE 291

291 Figure A 121. Two station electric field waveforms of the RS4 of flash UF 15 16.

PAGE 292

292 Figure A 122. Two station electric field waveforms of the RS5 of flash UF 15 16.

PAGE 293

293 Figure A 123. Two s tation electric field waveforms of the RS6 of flash UF 15 16.

PAGE 294

294 Figure A 124. Two station electric field waveforms of the RS7 of flash UF 15 16.

PAGE 295

295 Figure A 125. Two station electric field waveforms of the RS8 of flash UF 15 16.

PAGE 296

296 Figure A 126. Two stati on electric field waveforms of the RS9 of flash UF 15 16.

PAGE 297

297 Figure A 127. Two station electric field waveforms of the RS10 of flash UF 15 16.

PAGE 298

298 Figure A 128. Two station electric field waveforms of the RS11 of flash UF 15 16.

PAGE 299

299 Figure A 129. Two station electric field waveforms of the RS12 of flash UF 15 16.

PAGE 300

300 Figure A 130. Two station electric field waveforms of the RS13 of flash UF 15 16.

PAGE 301

301 Figure A 131. Two station electric field waveforms of the RS14 of flash UF 15 16.

PAGE 302

302 Figure A 132. Two station electric field waveforms of flash UF 15 20.

PAGE 303

303 Figure A 133. Two station electric field waveforms of the RS1 of flash UF 15 20.

PAGE 304

304 Figure A 134. Two station electric field waveforms of the RS2 of flash UF 15 20.

PAGE 305

305 Figure A 135. Two station electric field waveforms of flash UF 15 25.

PAGE 306

306 Figure A 136. Two station electric field waveforms of the RS1 of flash UF 15 25.

PAGE 307

307 Figure A 137. Two station electric field waveforms of the RS2 of flash UF 15 25.

PAGE 308

308 Figure A 138. Two station electric field waveforms of th e RS3 of flash UF 15 25.

PAGE 309

309 Figure A 139. Two station electric field waveforms of the RS4 of flash UF 15 25.

PAGE 310

310 Figure A 140. Two station electric field waveforms of flash UF 15 26.

PAGE 311

311 Figure A 141. Two station electric field waveforms of the RS1 of flash UF 15 26.

PAGE 312

312 Figure A 142. Two station electric field waveforms of the RS2 of flash UF 15 26.

PAGE 313

313 Figure A 143. Two station electric field waveforms of the RS3 of flash UF 15 26.

PAGE 314

314 Figure A 144. Two station electric field waveforms of the RS4 of flash UF 1 5 26.

PAGE 315

315 Figure A 145. Two station electric field waveforms of the RS5 of flash UF 15 26.

PAGE 316

316 Figure A 146. Two station electric field waveforms of the RS6 of flash UF 15 26.

PAGE 317

317 Figure A 147. Two station electric field waveforms of the RS7 of flash UF 15 26

PAGE 318

318 Figure A 148. Two station electric field waveforms of the RS8 of flash UF 15 26.

PAGE 319

319 Figure A 149. Two station electric field waveforms of flash UF 15 38.

PAGE 320

320 Figure A 150. Two station electric field waveforms of the RS1 of flash UF 15 38.

PAGE 321

321 Figure A 15 1. Two station electric field waveforms of the RS2 of flash UF 15 38.

PAGE 322

322 Figure A 152. Two station electric field waveforms of the RS3 of flash UF 15 38.

PAGE 323

323 Figure A 153. Two station electric field waveforms of the RS4 of flash UF 15 38.

PAGE 324

324 Figure A 154. Two station electric field waveforms of the RS5 of flash UF 15 38.

PAGE 325

325 Figure A 155. Two station electric field waveforms of flash UF 15 39.

PAGE 326

326 Figure A 156. Two station electric field waveforms of the RS1 of flash UF 15 39.

PAGE 327

327 Figure A 157. Two station ele ctric field waveforms of the RS2 of flash UF 15 39.

PAGE 328

328 Figure A 158. Two station electric field waveforms of the RS3 of flash UF 15 39.

PAGE 329

329 Figure A 159. Two station electric field waveforms of the RS4 of flash UF 15 39.

PAGE 330

330 Figure A 160. Two station electri c field waveforms of the RS5 of flash UF 15 39.

PAGE 331

331 Figure A 161. Two station electric field waveforms of the RS6 of flash UF 15 39.

PAGE 332

332 Figure A 162. Two station electric field waveforms of flash UF 15 41.

PAGE 333

333 Figure A 163. Two station electric field wavefor ms of the RS1 of flash UF 15 41.

PAGE 334

334 Figure A 164. Two station electric field waveforms of the RS2 of flash UF 15 41.

PAGE 335

335 Figure A 165. Two station electric field waveforms of the RS3 of flash UF 15 41.

PAGE 336

336 Figure A 166. Two station electric field waveforms o f the RS4 of flash UF 15 41.

PAGE 337

337 Figure A 167. Two station electric field waveforms of the RS5 of flash UF 15 41.

PAGE 338

338 Figure A 168. Two station electric field waveforms of flash UF 15 42.

PAGE 339

339 Figure A 169. Two station electric field waveforms of the RS1 of fl ash UF 15 42.

PAGE 340

340 Figure A 170. Two station electric field waveforms of the RS2 of flash UF 15 42.

PAGE 341

341 Figure A 171. Two station electric field waveforms of the RS3 of flash UF 15 42.

PAGE 342

342 Figure A 172. Two station electric field waveforms of the RS4 of flash UF 15 42.

PAGE 343

343 Figure A 173. Two station electric field waveforms of the RS5 of flash UF 15 42.

PAGE 344

344 Figure A 174. Two station electric field waveforms of the RS6 of flash UF 15 42.

PAGE 345

345 Figure A 175. Two station electric field waveforms of flash UF 15 43.

PAGE 346

346 Fig ure A 176. Two station electric field waveforms of the RS1 of flash UF 15 43.

PAGE 347

347 Figure A 177. Two station electric field waveforms of the RS2 of flash UF 15 43.

PAGE 348

348 Figure A 178. Two station electric field waveforms of the RS3 of flash UF 15 43.

PAGE 349

349 Figure A 179. Two station electric field waveforms of flash UF 16 04.

PAGE 350

350 Figure A 180. Two station electric field waveforms of the RS1 of flash UF 16 0

PAGE 351

351 APPENDIX B TWO STATION MEASUREMENTS OF ELECTRIC FIELD WAVEFORMS OF NATURAL NEGATIVE CLOUD TO GROUND LIGHTNING The two station (LOG GC) measurements of electric field waveforms for 10 natural negative cloud to ground lightning near Camp Blanding are presented in this appendix Flash ID data, and number of return strokes are listed in Table B 1. For each stroke, the complete field record of the flash is shown first and followed by expansions for each stroke o n 1 ms and 200 s time scales The distances from each of the stations to the lightning strik e point were determined from NLDN reported location of the strok e. Table B 1. Inventory of two station (LOG GC) field measurements for 10 natural negative cloud to ground lightning Flash ID Date Number of Return Strokes 00389 08/07/2015 3 00390 08/07/2015 3 00467 08/07/2015 3 00468 08/07/2015 4 00491 08/07/ 2015 4 00532 08/07/2015 4 00537 08/07/2015 2 00564 08/07/2015 5 00569 08/07/2015 4 00620 08/13/2015 1 Total Number of Flashes: 10 Total Number of Strokes: 33

PAGE 352

352 Figure B 1. Two station electric field waveforms of flash 00389.

PAGE 353

353 Figure B 2. Two station electric field waveforms of RS1 of flash 00389.

PAGE 354

354 Figure B 3. Two station electric field waveforms of RS2 of flash 00389.

PAGE 355

355 Figure B 4. Two station electric field waveforms of RS3 of flash 00389.

PAGE 356

356 Figure B 5. Two station electric field wa veforms of flash 00390.

PAGE 357

357 Figure B 6. Two station electric field waveforms of RS1 of flash 00390.

PAGE 358

358 Figure B 7. Two station electric field waveforms of RS2 of flash 00390.

PAGE 359

359 Figure B 8. Two station electric field waveforms of RS3 of flash 00390.

PAGE 360

360 Figur e B 9. Two station electric field waveforms of flash 00467.

PAGE 361

361 Figure B 10. Two station electric field waveforms of the RS1 of flash 00467.

PAGE 362

362 Figure B 11. Two station electric field waveforms of the RS2 of flash 00467.

PAGE 363

363 Figure B 12. Two station electric field waveforms of the RS3 of flash 00467.

PAGE 364

364 Figure B 13. Two station electric field waveforms of flash 00468.

PAGE 365

365 Figure B 14. Two station electric field waveforms of the RS1 of flash 00468.

PAGE 366

366 Figure B 15. Two station electric field waveforms of the RS2 of flash 00468.

PAGE 367

367 Figure B 16. Two station electric field waveforms of the RS3 of flash 00468.

PAGE 368

368 Figure B 17. Two station electric field waveforms of the RS4 of flash 00468.

PAGE 369

369 Figure B 18. Two station electric field waveforms of flash 00491.

PAGE 370

370 Figure B 19. Two station electric field waveforms of the RS1 of flash 00491.

PAGE 371

371 Figure B 20. Two station electric field waveforms of the RS2 of flash 00491.

PAGE 372

372 Figure B 21. Two station electric field waveforms of the RS3 of flash 00491.

PAGE 373

373 Figure B 22. Two statio n electric field waveforms of the RS4 of flash 00491.

PAGE 374

374 Figure B 23. Two station electric field waveforms of flash 00532.

PAGE 375

375 Figure B 24. Two station electric field waveforms of the RS1 of flash 00532.

PAGE 376

376 Figure B 25. Two station electric field waveforms of the RS2 of flash 00532.

PAGE 377

377 Figure B 26. Two station electric field waveforms of the RS3 of flash 00532.

PAGE 378

378 Figure B 27. Two station electric field waveforms of the RS4 of flash 00532.

PAGE 379

379 Figure B 28. Two station electric field waveforms of flash 00537.

PAGE 380

380 Figure B 29. Two station electric field waveforms of the RS1 of flash 00537.

PAGE 381

381 Figure B 30. Two station electric field waveforms of the RS2 of flash 00537.

PAGE 382

382 Figure B 31. Two station electric field waveforms of flash 00564.

PAGE 383

383 Figure B 32. Two station electric field waveforms of the RS1 of flash 00564.

PAGE 384

384 Figure B 33. Two station electric field waveforms of the RS2 of flash 00564.

PAGE 385

385 Figure B 34. Two station electric field waveforms of the RS3 of flash 00564.

PAGE 386

386 Figure B 35. Two station electric field waveforms of the RS4 of flash 00564.

PAGE 387

387 Figure B 36. Two station electric field waveforms of the RS5 of flash 00564.

PAGE 388

388 Figure B 37. Two station electric field waveforms of flash 00569.

PAGE 389

389 Figure B 38. Two station electric field waveforms of the RS1 of fla sh 00569.

PAGE 390

390 Figure B 39. Two station electric field waveforms of the RS2 of flash 00569.

PAGE 391

391 Figure B 40. Two station electric field waveforms of the RS3 of flash 00569.

PAGE 392

392 Figure B 41. Two station electric field waveforms of the RS4 of flash 00569.

PAGE 393

393 Figu re B 42. Two station electric field waveforms of flash 00620.

PAGE 394

394 Figure B 43. Two station electric field waveforms of the RS1 of flash 0062

PAGE 395

395 APPENDIX C TWO STATION MEASUREMENTS OF ELECTRIC FIELD WAVEFORMS OF NATURAL POSITIVE CLOUD TO GROUND LIGHTNING T he two station (LOG GC) measurements of electric field waveforms for 10 natural positive cloud to ground lightning near Camp Blanding are presented in this appendix. Flash ID, data, and number of return strokes are listed in Table C 1. Except for one flash containing two strokes, all the flashes were single stroke flashes. For each stroke, the complete field record of the flash is shown first and followed by expansions for each stroke on 1 ms and 200 s time scales. The distances from each of the stations t o the lightning strike point were determined from NLDN reported location of the stroke Table C 1. Inventory of two station (LOG GC) field measurements for 10 natural positive cloud to ground lightning Flash ID Date Number of Return Strokes 00436 08/07 /2015 1 00438 08/07/2015 1 00453 08/07/2015 1 00456 08/07/2015 1 00477 08/07/2015 1 00577 08/07/2015 1 00580 08/07/2015 1 00590 08/07/2015 1 01162 08/19/2015 2 01163 08/19/2015 1 Total Number of Flashes: 10 Total Number of Strokes: 11

PAGE 396

396 Figure B 44. Two station electric field waveforms of flash 00436.

PAGE 397

397 Figure B 45. Two station electric field waveforms of the RS1 of flash 00436.

PAGE 398

398 Figure B 46. Two station electric field waveforms of flash 00438.

PAGE 399

399 Figure B 47. Two station electri c field waveforms of the RS1 of flash 00438.

PAGE 400

400 Figure B 48. Two station electric field waveforms of flash 00453.

PAGE 401

401 Figure B 49. Two station electric field waveforms of the RS1 of flash 00453.

PAGE 402

402 Figure B 50. Two station electric field waveforms of flash 00456.

PAGE 403

403 Figure B 51. Two station electric field waveforms of the RS1 of flash 00456.

PAGE 404

4 04 Figure B 52. Two station electric field waveforms of flash 00477.

PAGE 405

405 Figure B 53. Two station electric field waveforms of the RS1 of flash 00477.

PAGE 406

406 Figure B 54. Two station electric field waveforms of flash 00577.

PAGE 407

407 Figure B 55. Two station electric field waveforms of the RS1 of flash 00577.

PAGE 408

408 Figure B 56. Two station electric field waveforms of flash 00580.

PAGE 409

409 Figure B 57. Two station electric field waveforms of th e RS1 of flash 00580.

PAGE 410

410 Figure B 58. Two station electric field waveforms of flash 00590.

PAGE 411

411 Figure B 59. Two station electric field waveforms of the RS1 of flash 00590.

PAGE 412

412 Figure B 60. Two station electric field waveforms of flash 01162.

PAGE 413

413 Figure B 61. Two station electric field waveforms of the RS1 of flash 01162.

PAGE 414

414 Figure B 62. Two station electric field waveforms of the RS2 of flash 01162.

PAGE 415

415 Figure B 63. Two station electric field waveforms of flash 01163.

PAGE 416

416 Figure B 64. Two station electric field waveforms of the RS1 of flash 01163

PAGE 417

417 APPENDIX D HIGH SPEED VIDEO RECORDS OF THE TWO FLASHES TERMINATED ON THE 257 M TOWER This section shows the images from high speed video records of the two flashes terminated on the 257 m tower. More information of th e two flashes are given in Chapter 6. The frame rate of the camera is 1000 fps with 1 ms exposure time. Frame number is given in the upper right corner of each panel. F rame number marked by star indicates the strongest enhancement of the image. It seems th at all strokes in flash 1593 were initiated by bi directional leader s with the positive end s propagating upward with many branches and the negative ends sharing the same channel that terminated on the 257 m tower.

PAGE 418

418 Figure C 1. Frames showing the first stroke of flash 1593. Frame number is at the upper right corner. The star indicates the strongest enhancement of the image. The frame rate is 1000 fps and the exposure time is 1 ms.

PAGE 419

419 Figure C 2. First two frames showing the second of flash 1593.

PAGE 420

420 Figu re C 3. Last two frames showing the second stroke of flash 1593.

PAGE 421

421 Figure C 4. Frames showing the third stroke of flash 1593.

PAGE 422

422 Figure C 5. First two frames showing the fourth stroke of flash 1593.

PAGE 423

423 Figure C 6. Last two frames showing the fourth s troke of flash 1593.

PAGE 424

424 Figure C 7 Frames showing the fifth stroke of flash 1593.

PAGE 425

425 Figure C 8. First two frames showing the sixth stroke of flash 1593.

PAGE 426

426 Figure C 9. Last two frames showing the sixth stroke of flash 1593.

PAGE 427

427 Figure C 10. Frames showing the second stroke of flash 1594.

PAGE 428

428 Figure C 1 1 Frames showing the third stroke of flash 1594.

PAGE 429

429 LIST OF REFERENCES Azadifar, M., G. Diendorfer, M. Paolone, D. Pavanello, F. Rachidi, V. A. Rakov, M. Rubinstein, and W. Schulz (2017), A Large Bipolar Event Recorded at the Sntis Tower, in 4th International Symposium on Winter Lightning (ISWL) Joetsu, Niigata ken, Japan. Baba, Y., and V. A. Rakov (2005), On the use of lumped sources in lightning return stroke models, J. Geophys. Res. Atmos. 110 (3), 1 10, doi:10.1029/2004JD005202. Baharudin, Z. A., N. A. Ahmad, M. Fernando, V. Cooray, and J. S. Mkel (2012), Comparative study on preliminary breakdown pulse trains observed in Johor, Malaysia and Florida USA, Atmos. Res. 117 111 121, doi:10.1016/j.atmosres.2012.01.012. Baharudin, Z. A., N. A. Ahmad, J. S. Mkel, M. Fernando, and V. Cooray (2014), Negative cloud to ground lightning flashes in Malaysia, J. Atmos. Solar Terrestrial Phys. 108 61 67, doi :10.1016/j.jastp.2013.12.001. Ballarotti, M. G., M. M. F. Saba, and J. Pinto (2005), High speed camera observations of negative ground flashes on a millisecond scale, Geophys. Res. Lett. 32 (23), 1 4, doi:10.1029/2005GL023889. Ballarotti, M. G., C. Medeiro s, M. M. F. Saba, W. Schulz, and O. Pinto (2012), Frequency distributions of some parameters of negative downward lightning flashes based on accurate stroke count studies, J. Geophys. Res. Atmos. 117 (D6), n/a n/a, doi:10.1029/2011JD017135. Biagi, C. J., K L. Cummins, K. E. Kehoe, and E. P. Krider (2007), National Lightning Detection Network (NLDN) performance in southern Arizona, Texas, and Oklahoma in 2003 2004, J. Geophys. Res. 112 (D5), D05208, doi:10.1029/2006JD007341. Biagi, C. J., M. A. Uman, J. D. Hill, and D. M. Jordan (2011), Observations of the initial, upward propagating, positive leader steps in a rocket and wire triggered lightning discharge, Geophys. Res. Lett. 38 (24), 1 5, doi:10.1029/2011GL049944. Biagi, C. J., M. A. Uman, J. D. Hill, V. A Rakov, and D. M. Jordan (2012), Transient current pulses in rocket extended wires used to trigger lightning, J. Geophys. Res. Atmos. 117 (D7), D07205, doi:10.1029/2011JD016161. Brook, M. (1992), Breakdown electric fields in winter storms, Res. Lett. Atmo s. Electr. 12 47 52. Buck, T. L., A. Nag, and M. J. Murphy (2014), Improved Cloud to Ground and Intracloud Lightning Detection with the LS7002 Advanced Total Lightning Sensor, in WMO Technical Conference on Meteorological and Environmental Instruments an d Methods of Observation Saint Petersburg, Russia. Carvalho, F. L., M. A. Uman, D. M. Jordan, and T. Ngin (2015), Lightning current and

PAGE 430

430 luminosity at and above channel bottom for return strokes and M components, J. Geophys. Res. Atmos. 120 (20), 10,645 10 ,663, doi:10.1002/2015JD023814. Chen, L., Q. Zhang, W. Hou, and Y. Tao (2015a), On the field to current conversion factors for large bipolar lightning discharge events in winter thunderstorms in Japan, J. Geophys. Res. Atmos. 120 (14), 6898 6907, doi:10.10 02/2015JD023344. Chen, L., W. Lu, Y. Zhang, and D. Wang (2015b), Optical progression characteristics of an interesting natural downward bipolar lightning flash, J. Geophys. Res. Atmos. 120 (2), 708 715, doi:10.1002/2014JD022463. Clarence, N. D., and D. J. Malan (1957), Preliminary discharge processes in lightning flashes to ground, Q. J. R. Meteorol. Soc. 83 (356), 161 172, doi:10.1002/qj.49708335603. Cooray, V., and K. P. S. C. Jayaratne (1994), Characteristics of lightning flashes observed in Sri Lanka in the tropics, J. Geophys. Res. Atmos. 99 (D10), 21051 21056, doi:10.1029/94JD01519. Cooray, V., and R. Jayaratne (2000), What directs a lightning flash towards ground?, Sri Lankan J. Phys. 1 1 10, doi:10.4038/sljp.v1i0.165. Cooray, V., and H. Prez (1994 ), Some features of lightning flashes observed in Sweden, J. Geophys. Res. 99 (D5), 10683, doi:10.1029/93JD02366. Cummins, K. L., and M. J. Murphy (2009), An overview of lightning locating systems: History, techniques, and data uses, with an in depth look at the U.S. NLDN, IEEE Trans. Electromagn. Compat. 51 (3), 499 518, doi:10.1109/TEMC.2009.2023450. Cummins, K. L., M. J. Murphy, E. A. Bardo, W. L. Hiscox, R. B. Pyle, and A. E. Pifer (1998), A Combined TOA/MDF Technology Upgrade of the U.S. National Light ning Detection Network, J. Geophys. Res. 103 (D8), 9035, doi:10.1029/98JD00153. Diendorfer, G., H. Zhou, and H. Pichler (2011), Review of 10 years of lightning measurement at the Gaisberg Tower in Austria, in The 3rd International Symposium on Winter Light ning (ISWL) pp. 185 190, Sapporo, Japan. Dwyer, J. R., and M. A. Uman (2013), The physics of lightning, Phys. Rep. 534 (4), 147 241, doi:10.1016/j.physrep.2013.09.004. Fleenor, S. A., C. J. Biagi, K. L. Cummins, E. P. Krider, and X. M. Shao (2009), Charac teristics of cloud to ground lightning in warm season thunderstorms in the Central Great Plains, Atmos. Res. 91 (2 4), 333 352, doi:10.1016/j.atmosres.2008.08.011. Frey, H. U., S. B. Mende, S. A. Cummer, A. B. Chen, R. R. Hsu, H. T. Su, Y. S. Chang, T. Ada chi, H. Fukunishi, and Y. Takahashi (2005), Beta type stepped leader of elve producing lightning, Geophys. Res. Lett. 32 (13), 1 4, doi:10.1029/2005GL023080. Gamerota, W. (2014), The Dart Stepped Leader in Rocket Triggered Lightning, University of

PAGE 431

431 Florida. Gomes, C., V. Cooray, and C. Jayaratne (1998), Comparison of preliminary breakdown pulses observed in Sweden and in Sri Lanka, J. Atmos. Solar Terrestrial Phys. 60 (10), 975 979, doi:10.1016/S1364 6826(98)00007 8. Heavner, M. J., D. A. Smith, A. R. Jacobs on, and R. J. Sheldon (2002), LF/VLF and VHF lightning fast stepped leader observations, J. Geophys. Res. 107 (D24), 4791, doi:10.1029/2001JD001290. Ishii, M., and M. Saito (2009), Lightning electric field characteristics associated with transmission line faults in winter, IEEE Trans. Electromagn. Compat. 51 (3), 459 465, doi:10.1109/TEMC.2009.2025496. Ishii, M., M. Saito, T. Miki, D. Tanka, T. Shindo, A. Asakawa, H. Motoyama, Y. Suzuhigashi, and H. Taguchi (2013), Reproduction of Electromagnetic Field Wave forms and Tower Currents Associated with Return Strokes Struck Tokyo Skytree, in XII International Symposium on Lightning Protection (SIPDA) pp. 96 101, Belo Horizonte, Brazil. Jerauld, J., V. A. Rakov, M. A. Uman, K. J. Rambo, and D. M. Jordan (2005), An evaluation of the performance characteristics of the U.S. National Lightning Detection Network in Florida using rocket triggered lightning, J. Geophys. Res. 110 (D19), D19106, doi:10.1029/2005JD005924. Jerauld, J. E., M. A. Uman, V. A. Rakov, K. J. Rambo, D. M. Jordan, and G. H. Schnetzer (2009), Measured electric and magnetic fields from an unusual cloud to ground lightning flash containing two positive strokes followed by four negative strokes, J. Geophys. Res. Atmos. 114 (19), 1 17, doi:10.1029/2008JDO1 1660. Jordan, D. M., V. P. Idone, V. A. Rakov, M. A. Uman, W. H. Beasley, and H. Jurenka (1992), Observed dart leader speed in natural and triggered lightning, J. Geophys. Res. Atmos. 97 (D9), 9951 9957, doi:10.1029/92JD00566. Karunarathne, S., T. C. Marsh all, M. Stolzenburg, and N. Karunarathna (2014), Modeling initial breakdown pulses of CG lightning flashes, J. Geophys. Res. Atmos. 119 (14), 9003 9019, doi:10.1002/2014JD021553. Kitagawa, N., M. Brook, and E. Workman (1962), Continuing currents in cloud to ground lightning discharges, J. Geophys. Res, 67 67 (2), 637 647, doi:10.1029/JZ067i002p00637. unusually short pulse sequences occurring prior to the first strokes of negative cloud to ground lightning flashes, Geophys. Res. Lett. 41 6192, doi:10.1002/2014GL060913. Kotovsky, D. A., R. C. Moore, Y. Zhu, M. D. Tran, V. A. Rakov, J. T. Pilkey, J. A. Caicedo, B. Hare, D. M. Jordan, and M. A. Uman (2016), Initia l breakdown and fast leaders in lightning discharges producing long lasting disturbances of the lower ionosphere, J. Geophys. Res. Sp. Phys. 121 (6), 5794 5804, doi:10.1002/2015JA022266.

PAGE 432

432 Krehbiel, P. R., M. Brook, and R. A. McCrory (1979), An analysis of t he charge structure of lightning discharges to ground, J. Geophys. Res. 84 (C5), 2432, doi:10.1029/JC084iC05p02432. Krider, E. P., G. J. Radda, and R. C. Noggle (1975), Regular radiation field pulses produced by intracloud lightning discharges, J. Geophys. Res. 80 (27), 3801 3804, doi:10.1029/JC080i027p03801. Lalande, P., A. Bondiou Clergerie, P. Laroche, A. Eybert Berard, J. P. Berlandis, B. Bador, A. Bonamy, M. A. Uman, and V. A. Rakov (1998), Leader properties determined with triggered lightning techniqu es, J. Geophys. Res. Atmos. 103 (D12), 14109 14115, doi:10.1029/97JD02492. Liu, C., and S. Heckman (2011), The application of total lightning detection and cell tracking for severe weather prediction, in 91st AmericanMeteorological Society Annual Meeting Seattle. Lu, W., Y. Ma, Y. Zhang, J. Yang, W. Yao, D. Zheng, Q. Meng, and Y. Zhang (2014), Total Sky Lightning Event Observation System and Method, MacDonald, J. H. (2008), Handbook of Biological Statistics Sparky House Publishing. Mallick, S. (2013), Obs ervation of Natural and Rocket Triggered Lightning at Three Research Stations in Florida and with Large Scale Locating Networks, Ph. D. Proposal, University of Florida. Mallick, S., and V. A. Rakov (2014), Characterization of far electric field waveforms p roduced by rocket triggered lightning, in 2014 International Conference on Lightning Protection, ICLP 2014 pp. 1527 1534, Shanghai, China. Mallick, S., V. A. Rakov, J. D. Hill, W. R. Gamerota, M. A. Uman, S. Heckman, C. D. Sloop, and C. Liu (2013), Calibr ation of the ENTLN against rocket triggered lightning data, in 2013 International Symposium on Lightning Protection, (XII SIPDA) pp. 39 46, Belo Horizonte, Brazil. Mallick, S. et al. (2014), Performance characteristics of the NLDN for return strokes and p ulses superimposed on steady currents, based on rocket triggered lightning data acquired in Florida in 2004 2012, J. Geophys. Res. Atmos. 119 (7), 2013JD021401, doi:10.1002/2013JD021401. Mallick, S. et al. (2015), Performance characteristics of the ENTLN e valuated using rocket triggered lightning data, Electr. Power Syst. Res. 118 15 28, doi:10.1016/j.epsr.2014.06.007. Marshall, T., W. Schulz, N. Karunarathna, S. Karunarathne, M. Stolzenburg, C. Vergeiner, and T. Warner (2014), On the percentage of lightn ing flashes that begin with initial breakdown pulses, J. Geophys. Res. Atmos. 119 (2), 445 460, doi:10.1002/2013JD020854.

PAGE 433

433 Maslowski, G., and V. A. Rakov (2006), A study of the lightning channel corona sheath, J. Geophys. Res. Atmos. 111 (14), 1 16, doi:10. 1029/2005JD006858. Montany, J., O. Van Der Velde, E. Williams, N. Pineda, D. Romero, and G. Sol (2014), Observation and properties of a sub cloud bidirectional leader, in XV International Conference on Atmospheric Electricity (ICAE) Norman, Oklah oma, U.S.A. Murphy, J. M., and A. Nag (2015), Cloud lightning performance and climatology of the U.S. based on the upgraded U.S. National Lightning Detection Network, in Seventh Conference on the Meteorological Applications of Lightning Data Phoenix, AZ. Nag, A., and V. a. Rakov (2010a), Compact intracloud lightning discharges: 1. Mechanism of electromagnetic radiation and modeling, J. Geophys. Res. 115 (D20), D20102, doi:10.1029/2010JD014235. Nag, A., and V. a. Rakov (2010b), Compact intracloud lightning discharges: 2. Estimation of electrical parameters, J. Geophys. Res. 115 (D20), D20103, doi:10.1029/2010JD014237. Nag, A., and V. A. Rakov (2008), Pulse trains that are characteristic of preliminary breakdown in cloud to ground lightning but are not follow ed by return stroke pulses, J. Geophys. Res. 113 (D1), D01102, doi:10.1029/2007JD008489. Nag, A., and V. A. Rakov (2009a), Electric field pulse trains occurring prior to the first stroke in negative cloud to ground lightning, IEEE Trans. Electromagn. Compa t. 51 (1), 147 150, doi:10.1109/TEMC.2008.2005488. Nag, A., and V. A. Rakov (2009b), Some inferences on the role of lower positive charge region in facilitating different types of lightning, Geophys. Res. Lett. 36 (5), L05815, doi:10.1029/2008GL036783. Nag A., and V. A. Rakov (2012), Positive lightning: An overview, new observations, and inferences, J. Geophys. Res. 117 (D8), D08109, doi:10.1029/2012JD017545. Nag, A., V. a. Rakov, W. Schulz, M. M. F. Saba, R. Thottappillil, C. J. Biagi, A. Oliveira Filho, A. Kafri, N. Theethayi, and T. Gotschl (2008), First versus subsequent return stroke current and field peaks in negative cloud to ground lightning discharges, J. Geophys. Res. 113 (D19), D19112, doi:10.1029/2007JD009729. Nag, A., V. A. Rakov, D. Tsalikis, and J. A. Cramer (2010), On phenomenology of compact intracloud lightning discharges, J. Geophys. Res. 115 (D14), D14115, doi:10.1029/2009JD012957. Nag, A. et al. (2011), Evaluation of U.S. National Lightning Detection Network performance characteristics u sing rocket triggered lightning data acquired in 2004 2009, J. Geophys. Res. 116 (D2), 1 8, doi:10.1029/2010JD014929. Nag, A., M. J. Murphy, K. L. Cummins, A. E. Pifer, and J. A. Cramer (2014), Recent Evolution of the U.S. National Lightning Detection Netw ork, in 23nd International Lightning

PAGE 434

434 Detection Conference (ILDC) Tucson, Arizona, USA. Ngin, T., M. A. Uman, J. D. Hill, R. C. Olsen, J. T. Pilkey, W. R. Gamerota, and D. M. Jordan (2014), Does the lightning current go to zero between ground strokes? Is t here a current Geophys. Res. Lett. 41 (9), 3266 3273, doi:10.1002/2014GL059601. Nucci, C. A., C. Mazzetti, F. Rachidi, and M. Ianoz (1988), On lightning return stroke models for LEMP calculations, in 19th International Conference on Lightning Pr otection Graz, Austria. Pavanello, D. et al. (2007), On return stroke currents and remote electromagnetic fields associated with lightning strikes to tall structures: 2. Experiment and model validation, J. Geophys. Res. Atmos. 112 (13), 1 12, doi:10.1029/ 2006JD007959. Paxton, A. H., R. L. Gardner, and L. Baker (1986), Lightning return stroke. A numerical calculation of the optical radiation, Phys. Fluids 29 (8), 2736, doi:10.1063/1.865514. Pichler, H., G. Diendorfer, and M. Mair (2010), Some parameters of correlated current and radiated field pulses from lightning to the Gaisberg tower, IEEJ Trans. Electr. Electron. Eng. 5 (1), 8 13, doi:10.1002/tee.20486. Rachidi, F., V. A. Rakov, C. A. Nucci, and J. L. Bermudez (2002), Effect of vertically extended strike object on the distribution of current along the lightning channel, J. Geophys. Res. Atmos. 107 (23), 1 6, doi:10.1029/2002JD002119. Rakov, V. A. (2003a), A Review of Positive and Bipolar Lightning Discharges, Bull. Am. Meteorol. Soc. 84 (6), 767 776, doi: 10.1175/BAMS 84 6 767. Rakov, V. A. (2003b), A review of the interaction of lightning with tall objects, Recent Res. Dev. Geophys. 5 58 71. Rakov, V. A. (2013), Electromagnetic Methods of Lightning Detection, Surv. Geophys. 34 (6), 731 753, doi:10.1007/s 10712 013 9251 1. Rakov, V. A., and A. A. Dulzon (1987), Calculated electromagnetic fields of lightning return stroke, Tekh. Elektrodinam 1 87 89. Rakov, V. A., and A. A. Dulzon (1991), A modified transmission line model for lightning return stroke field calculations, in 9th International Symposium Electromagnetic Compatibility pp. 229 235, Zurich, Switzerland. Rakov, V. A., and M. A. Uman (1990), Some properties of negative cloud to ground lightning flashes versus stroke order, J. Geophys. Res. Atmos. 95 (D5), 5447 5453, doi:10.1029/JD095iD05p05447. Rakov, V. A., and M. A. Uman (1994), Origin of lightning electric field signatures showing two return stroke waveforms separated in time by a millisecond or less, J. Geophys. Res. 99 (D4), 8157, doi:10.1029/9 4JD00165.

PAGE 435

435 Rakov, V. A., and M. A. Uman (1998), Review and evaluation of lightning return stroke models including some aspects of their application, IEEE Trans. Electromagn. Compat. 40 (4), 403 426, doi:10.1109/15.736202. Rakov, V. A., and M. A. Uman (2003) Lightning: Physics and Effects Cambridge University Press, New York. Rakov, V. A., M. A. Uman, D. M. Jordan, and C. A. Priore (1990), Ratio of leader to return stroke electric field change for first and subsequent lightning strokes, J. Geophys. Res. Atm os. 95 (D10), 16579 16587, doi:10.1029/JD095iD10p16579. Rakov, V. A., M. A. Uman, and R. Thottappillil (1994), Review of lightning properties from electric field and TV observations, J. Geophys. Res. Atmos. 99 (D5), 10745 10750, doi:10.1029/93JD01205. Rako v, V. A., M. A. Uman, G. R. Hoffman, M. W. Masters, and M. Brook (1996), Bursts of pulses in lightning electromagnetic radiation: observations and implications for lightning test standards, IEEE Trans. Electromagn. Compat. 38 (2), 156 164, doi:10.1109/15.4 94618. Rakov, V. A. et al. (1998), New insights into lightning processes gained from triggered lightning experiments in Florida and Alabama, J. Geophys. Res. Atmos. 103 (D12), 14117 14130, doi:10.1029/97JD02149. Rakov, V. A., S. Mallick, A. Nag, and V. B. Somu (2014), Lightning Observatory in Gainesville (LOG), Florida: A review of recent results, Electr. Power Syst. Res. 113 (October 2015), 95 103, doi:10.1016/j.epsr.2014.02.037. Rubinstein, M., J. Bermdez, V. Rakov, F. Rachidi, and A. Hussein (2012), Comp ensation of the Instrumental Decay in Measured Lightning Electric Field Waveforms, IEEE Trans. Electromagn. Compat. 54 (3), 685 688, doi:10.1109/TEMC.2012.2198482. Saba, M. M. F., W. Schulz, T. A. Warner, L. Z. S. Campos, C. Schumann, E. P. Krider, K. L. C ummins, and R. E. Orville (2010), High speed video observations of positive lightning flashes to ground, J. Geophys. Res. 115 (D24), D24201, doi:10.1029/2010JD014330. Saba, M. M. F., C. Schumann, T. A. Warner, J. H. Helsdon, W. Schulz, and R. E. Orville (2 013), Bipolar cloud to ground lightning flash observations, J. Geophys. Res. Atmos. 118 (19), 11098 11106, doi:10.1002/jgrd.50804. Saito, M., T. Miki, T. Shindo, H. Motoyama, and M. Ishii (2015a), Reproduction of Electromagnetic Field Waveforms of Return S trokes Hitting Tokyo Skytree, in XIII International Symposium on Lightning Protection (SIPDA) pp. 89 95, Balnerio Cambori, Brazil. Saito, M., T. Miki, T. Shindo, H. Motoyama, M. Ishii, T. Sonehara, H. Taguchi, A. Tajima, and A. Fujisawa (2015b), Reprodu ction of Electromagnetic Field Waveforms of Subsequent Return Strokes Hitting Tokyo Skytree over Lossy Ground, IEEJ Trans. Power Energy 135 (7), 472 478, doi:10.1541/ieejpes.135.472.

PAGE 436

436 Saraiva, A. C. V, M. M. F. Saba, O. Pinto, K. L. Cummins, E. P. Krider, a nd L. Z. S. Campos (2010), A comparative study of negative cloud to ground lightning characteristics in S??o Paulo (Brazil) and Arizona (United States) based on high speed video observations, J. Geophys. Res. Atmos. 115 (11), 1 9, doi:10.1029/2009JD012604. Saraiva, A. C. V. et al. (2014), High speed video and electromagnetic analysis of two natural bipolar cloud to ground lightning flashes, J. Geophys. Res. Atmos. 119 (10), 6105 6127, doi:10.1002/2013JD020974. Shao, X. M., and M. Heavner (2006), On the VLF/ LF radiation pulse shapes at the initial milliseconds of lightning discharges, in 2006 17th International Zurich Symposium on Electromagnetic Compatibility pp. 402 404, IEEE, Zurich, Switzerland. Stolzenburg, M., T. C. Marshall, S. Karunarathne, N. Karuna rathna, L. E. Vickers, T. A. Warner, R. E. Orville, and H. D. Betz (2013), Luminosity of initial breakdown in lightning, J. Geophys. Res. Atmos. 118 (7), 2918 2937, doi:10.1002/jgrd.50276. Stolzenburg, M., T. C. Marshall, S. Karunarathne, N. Karunarathna, and R. E. Orville (2014), Leader observations during the initial breakdown stage of a lightning flash, J. Geophys. Res. Atmos. 119 (21), 12,198 12,221, doi:10.1002/2014JD021994. Stolzenburg, M., T. C. Marshall, S. Karunarathne, N. Karunarathna, and R. E. O rville (2015), Transient luminosity along negative stepped leaders in lightning, J. Geophys. Res. Atmos. 120 (8), 3408 3435, doi:10.1002/2014JD022933. Thottappillil, R., V. A. Rakov, M. A. Uman, W. H. Beasley, M. J. Master, and D. V. Shelukhin (1992), Ligh tning subsequent stroke electric field peak greater than the first stroke peak and multiple ground terminations, J. Geophys. Res. 97 (D7), 7503, doi:10.1029/92JD00557. Thottappillil, R., V. A. Rakov, and M. A. Uman (1997), Distribution of charge along the lightning channel: Relation to remote electric and magnetic fields and to return stroke models, J. Geophys. Res. Atmos. 102 (D6), 6987 7006, doi:10.1029/96JD03344. Tian, Y., X. Qie, G. Lu, R. Jiang, Z. Wang, H. Zhang, M. Liu, Z. Sun, and G. Feng (2016), Ch aracteristics of a bipolar cloud to ground lightning flash containing a positive stroke followed by three negative strokes, Atmos. Res. 176 177 222 230, doi:10.1016/j.atmosres.2016.02.023. Tran, M. D. (2015), Lightning Properties From High Speed Video an d X Ray Observations Synchronized with Electric Field Measurements, Ph. D. Proposal, University of Florida. Uman, M. A. (1987), The Lightning Discharge Academic Press, Inc., New York. Uman, M. A., and D. K. McLain (1969), Magnetic field of lightning retur n stroke, J. Geophys. Res. 74 (28), 6899 6910, doi:10.1029/JC074i028p06899. Uman, M. A., D. K. McLain, and E. P. Krider (1975), The electromagnetic radiation from a finite antenna, Am. J. Phys. 43 (1), 33, doi:10.1119/1.10027.

PAGE 437

437 Wang, D., V. A. Rakov, M. A. Uman, N. Takagi, T. Watanabe, D. E. Crawford, K. J. Rambo, G. H. Schnetzer, R. J. Fisher, and Z. I. Kawasaki (1999), Attachment process in rocket triggered lightning strokes, J. Geophys. Res. Atmos. 104 (D2), 2143 2150, doi:10.1029/1998JD200070. Wang, Y., X. Qie, D. Wang, M. Liu, D. Su, Z. Wang, D. Liu, Z. Wu, Z. Sun, and Y. Tian (2016), Beijing Lightning Network (BLNET) and the observation on preliminary breakdown processes, Atmos. Res. 171 121 132, doi:10.1016/j.atmosres.2015.12.012. Warner, T. A., K. L Cummins, and R. E. Orville (2012), Upward lightning observations from towers in Rapid City, South Dakota and comparison with National Lightning Detection Network data, 2004 2010, J. Geophys. Res. Atmos. 117 D19019, doi:10.1029/2012JD018346. Wilkes, R. A., M. A. Uman, J. T. Pilkey, and D. M. Jordan (2016), Luminosity in the initial breakdown stage of cloud to ground and intracloud lightning, J. Geophys. Res. Atmos. 121 (3), 1236 1247, doi:10.1002/2015JD024137. Wilson, N., J. Myers, K. Cummins, M. Hutchin son, and A. Nag (2013), Lightning attachment to situ peak current measurements, in Vienna. Wu, T., Y. Takayanagi, T. Funaki, S. Yoshi da, T. Ushio, Z. I. Kawasaki, T. Morimoto, and M. Shimizu (2013), Preliminary breakdown pulses of cloud to ground lightning in winter thunderstorms in Japan, J. Atmos. Solar Terrestrial Phys. 102 91 98, doi:10.1016/j.jastp.2013.05.014. Wu, T., S. Yoshida T. Ushio, Z. Kawasaki, Y. Takayanagi, and D. Wang (2014), Large bipolar lightning discharge events in winter thunderstorms in Japan, J. Geophys. Res. Atmos. 119 (2), 555 566, doi:10.1002/2013JD020369. Xue, S., P. Yuan, J. Cen, Y. Li, and X. Wang (2015), Spectral observations of a natural bipolar cloud to ground lightning, J. Geophys. Res. Atmos. 120 (5), 1972 1979, doi:10.1002/2014JD022598. Zhang, D., K. L. Cummins, and A. Nag (2015), Assessment of cloud lightning detection by the U.S. National Lightning Detection Network using video and lightning mapping array observations, in Seventh Conference on the Meteorological Applications of Lightning Data Phoenix, Arizona, USA. Zhou, H., G. Diendorfer, R. Thottappillil, H. Pichler, and M. Mair (2012), Characteri stics of upward positive lightning flashes initiated from the Gaisberg Tower, J. Geophys. Res. Atmos. 117 (6), 1 13, doi:10.1029/2011JD016903. Zhu, Y., V. A. Rakov, S. Mallick, M. D. Tran, J. Pilkey, and M. A. Uman (2014), Preliminary Breakdown Pulse Train s in Electric Field Records of Negative Cloud to Ground Lightning, in XV International Conference on Atmospheric Electricity Norman, Oklahoma, U.S.A.

PAGE 438

438 Zhu, Y., V. A. Rakov, S. Mallick, and M. D. Tran (2015), Characterization of negative cloud to ground lig htning in Florida, J. Atmos. Solar Terrestrial Phys. 136 8 15, doi:10.1016/j.jastp.2015.08.006. Zhu, Y., V. A. Rakov, M. D. Tran, and A. Nag (2016a), A Study of National Lightning Detection Network Responses to Natural Lightning Based on Ground Truth Dat a Acquired at LOG with Emphasis on Cloud Discharge Activity, J. Geophys. Res. Atmos. 2016JD025574, doi:10.1002/2016JD025574. Zhu, Y., V. A. Rakov, M. D. Tran, and W. Lu (2016b), A Subsequent Positive Stroke Developing in the Channel of Preceding Negative Stroke and Containing Bipolar Continuing Current, in 33rd International Conference on Lightning Protection (ICLP) Estoril, Portugal. Zhu, Y., V. A. Rakov, and M. D. Tran (2016c), Optical and electric field signarues of lightning interaction with a 257 m t all tower in Florida, Electr. Power Syst. Res. doi:10.1016/j.epsr.2016.08.036.

PAGE 439

439 BIOGRAPHICAL SKETCH Yanan Zhu was born in Nanjing, Jiangsu, China in 1990. He received a Bachelor of Science degree in electronic information engineering in 2012 from Nanjin g University of Information Science & Technology. In the fall of 2012, he started to work as a research assistant under the guidance of Dr. Vladimir Rakov in the lightning lab at the University of Florida. He was responsible for the operation of the Golf C ourse site in Starke from 2012 to 2017 He earne d his Ph.D. in electrical and computer engineering from the University of Florida in December 2017 He has authored six peer viewed j ournal papers, and has been a co author on two others.