LIGHTNING DETECTION NETWORKS FOR COMMERCIAL AVIATION TECHNOLOGY, METEOROLOGY AND APPLICATIONS

Transcription

1 LIGHTNING DETECTION NETWORKS FOR COMMERCIAL AVIATION TECHNOLOGY, METEOROLOGY AND APPLICATIONS Charlie Liu 1, James Anderson 2 and Stan Heckman 3 1.Vice President, International, Earth Networks, janderson@earthnetworks.com 2 Dr. Chief Computer Scientist, Earth Networks, cliu@earthnetworks.com 3. Dr. Stan Heckman, Chief Lightning Scientist, Earth Networks, sheckman@earthnetworks.com Keywords: Lightning, In-Cloud, Time-of-Arrival, Severe Weather, Total Lightning 1 Abstract The ability to detect and accurately pinpoint areas of existing and/or early-stage convection is critical to improving situational awareness and safety in the aviation industry. Given thunderstorms by definition include lightning and the vast majority of lightning is typically in-cloud (IC) as opposed to cloud-to ground (CG), it is necessary to have a network in place that detects total lightning activity (i.e., both IC and CG). Total lightning has been demonstrated to correlate well with storm dynamics and both case studies and statistical analyses suggest that total lightning is related to the presence or high likelihood of flight hazards such as hail, icing and turbulence. Additionally, lightning itself also poses a direct safety hazard to airborne aircraft as lightning strikes can cause engine failure, disrupt and damage aircraft electrical systems, lead to smoke and on-board fires, and temporarily affect flight crew vision. Total lightning information is also an important complement to existing weather radars as it can be highly useful in identifying areas of convection beyond radar ranges, locations where radar beams are blocked (i.e., mountainous areas), as well as early stage convective regions where precipitation returns do not yet appear significant. Given that CG activity typically represents a small fraction of the lightning occurring in the atmosphere, CG detection networks are not sufficient to meet the mission requirements of organizations tasked with providing critical information to aviation interests. The emergence of total lightning availability offers significant opportunities to enhance aviation flight safety. To this end the National Transportation Safety Board (NTSB) published a safety recommendation stating that the Federal Aviation Administration should consider several enhancements to its operations including; incorporation of total lightning data into weather displays at both air route traffic control centers and terminal radar approach control facilities, as well as into products supplied to pilots in the cockpit [17]. This paper reviews available total lightning network technology and potential applications for enhancing aviation safety. Case studies are reviewed and derivative products, such as proxy radar estimates, VIL and eco-tops are described. 2 Introduction Meteorology The lifecycle of a thunderstorm convection cell can be described by the classical tripole model, in which the main negative charge is located in the center of the cell, the main positive charge is in the cloud-top ice crystals, and a smaller positive charge is in the lower section of the cell, below the negative charge [9]. The initial electrification of the central and top parts may give rise to cloud flashes with intense enough charging producing ground lightning [9]. Severe thunderstorms, which may generate lightning, high wind, hail and tornadoes have certain characteristics in the lightning flashes, such as high IC flash rates in the storm formation stage. Severe storms may have either exceptionally low negative CG flash rates, or have exceptionally high positive CG flash rates; the greater volume of strong updrafts during a severe thunderstorm results in more charging overall, leading to greater numbers of IC flashes and positive CG flashes [5,10]. Past studies have shown that the CG flash rate has no correlation with tornadogenesis and that using CG lightning flash patterns exclusively to detect tornado formation is not practical [7]. This general finding seems to carry over to all damaging thunderstorms. There is no system known that can reliably predict intense thunderstorms using CG flash data alone. A study focused on severe thunderstorms in Florida using the lightning detection and ranging network (LDAR) total lightning data confirmed a distinguishing feature of severe storms, i.e., the systematic total lightning and abrupt increase in total lightning rate precursor to severe weather of all kinds wind, hail and tornados [11]. Previous studies on the Lightning Jump Concept using the Lightning Jump Algorithm have demonstrated the correlation between severe weather occurances and the sharp increases in lightning jump prior to and during the weather incidents. Rapid increase in total lightning activity (both CG and IC) were observed in advance of severe weather[13, 16]. The results of the study linked the timing of Lightning Jumps to corresponding severe weather occurances like hail and damaging wind. See Figure 1. A pure CG lightning detection system, due to the lack of IC detection capability, is not adequate for predicting severe storm development. The convection-cell structure of a thunderstorm is often visible in a weather radar image, and it can also be identified in lightning flash clusters when the rates 1

2 are high enough. But the lightning cells based on CG flashes can only show the mature stage of a convection cell [8], and they can t be used for early severe storm warning. The Huntsville, Alabama, National Weather Service office utilizes total lightning information from the North Alabama Lightning Mapping Array (NALMA) to diagnose convective trends; this lightning data has led to greater confidence and lead time in issuing severe thunderstorm and tornado warnings [3]. In one study, the IC lightning precursor provided a valuable short-term warning for microburst hazard at ground level [9]. The lightning cells identified from the total lightning data would be able to track the whole lifecycle of a storm. A study based on data from the Lightning Detection Network in Europe (LINET) achieved an important step in tracking lightning cells using total lightning data [2]. Figure 1: The graph above shows the results from a lightning jump concept study demonstrating the Lightning Jump Algorithhm (LJA) used in the Geostionary Lightning Mapper (GLM) network. The top graph details two instances of lightning jumps recorded during a severe thunderstorm on May 22, 1997 that reported 1 hail. The bottom image is a wind differential velocity graph showing the correlation between the lightning jumps and corresponding severe weather conditions. It has also been noted in these and other studies that the analysis of total lightning can enable more precise location of the greatest convective activity within a storm, since IC flash rates increase the most dramatically in the areas of greatest convection [15]. 3 Lightning Detection and Severe Weather Monitoring Technology methodology. We focus briefly on surveying these national scale networks, since they cover the greatest area and in some cases reliably detect large proportions of the IC flash activity and are thus more applicable for aviation. There are three main providers of such systems globally Earth Networks, Nowcast and Viasala, Inc. All three detection network providers have deployed extensive regional or global networks. Reliable intercomparisons of the performance of these networks can be difficult to find. Table 1 shows some basic characteristics regarding each network relevant for aviation culled from various websites and research reports. Network Provider Sensor Models Deployments Method of Detection Frequency Range of Detection IC-CG Detection Altitude of IC reported Data currently being used for aviation applications Severe Weather Nowcasting Products Earth Networks Nowcast (LINET) Viasala N/A N/A Various TLS200, LS7002 Globally with Dense regional Global and dense regional networks stand alone networks regional networks Time of Arrival Time of Arrival Various TOA, Magnetic Direction Finding and Interferometry VLF to HF VLF/LF VLF/LF and VHF Yes, waveform based Yes, 3D detection based Yes Yes N/A Yes Yes Yes Yes validated against NWS alerts and warnings Yes Table 1: Lightning Detection Network Characteristics Yes. Various N/A There are various studies that provide some indication of network performance. A full survey of them cannot be provided here. Studies include those comparing LINET to the LLDN in Poland [6] and various comparisons of networks to satellite observations or very local networks [2]. The U.S. National Weather Service has also done a large scale comparison of the two network datasets it uses, the ENTLN and NLDN. A search of the proceedings from the American Meteorological Society yields a large number of comparisons [18]. Currently, there are many different types of lightning detection systems and networks in operation today. These range from relatively simple and single point sensors which operate on a variety of principles to sophisticated national or global scale lightning detection networks, which generally speaking operate primarily on a time-of-arrival based 2

3 latitude, longitude and altitude. Strokes are then clustered into a flash if they are within 700 milliseconds and 10 kilometers. A flash that contains at least one return stroke is classified as a CG flash. Other Time-of-Arrival based systems use a similar method for locating lightning pulses and grouping them into flashes. Figure 2: The graph on the left shows the waveforms from an IC pulse, across multiple sensors in ENTLN. The graph on the right shows the waveforms from a return stroke (CG). The Earth Networks Total Lightning Network (ENTLN) utilizes a wide-band time of arrival based sensor. The deployment of this sensor network and the improvement in the detection efficiency, especially in IC flash detection on a national or continental scale, made it practical to track and predict severe weather in real-time over large areas. By combining advanced lightning detection technologies with modern electronics, an Earth Networks Lightning Sensor (ENLS) can acquire detailed signals emitted from both IC and CG flashes and continuously sends information to a central data processing system. An ENLS is composed of an antenna, a global positioning system (GPS) receiver, a GPS-based timing circuit, digital signal processors (DSP), and on-board storage and internet communication equipment. The ENLS is unique compared to other existing sensor technologies, with detection frequency ranging from 1HZ to 12MHZ. The lowest frequencies (below 1 KHz) are used for long range detection (CG). The middle frequencies (1KHz to 1MHz) are used for locating return strokes. The highest frequencies (1MHZ to 12 MHz) are used to detect and locate in-cloud pulses. Overlapping frequency ranges are used at two Analog to Digital Converters (ADC) to cover several orders of maginitude in frequency. The sensor records whole waveforms of each flash and sends them back, in compressed data packets, to the central server. Instead of using only the peak pulses, the whole waveforms are used in locating the flashes and differentiating between IC and CG strokes. The signal information enhances the detection efficiency and location accuracy of the system. Sophisticated digital signal processing technologies are used on the server side to ensure high-quality detections, to eliminate false locations, and to reduce noise typically associated with detecting electromagnetic energy. When lightning occurs, electromagnetic energy is emitted in all directions. Every ENTLN sensor that detected the waveforms records and sends the waveforms to the central lightning detection server via the Internet. The precise arrival times are calculated by correlating the waveforms from all the sensors that detected the strokes of a flash. The waveform arrival time and signal amplitude can be used to determine the peak current of the stroke and its exact location including A high-density network covers the contiguous United States, Alaska, Hawaii, the Caribbean basin, Australia, parts of Western Europe, Japan, Brazil and other geographies. The weather station and lightning sensor at site locations are plugged into a data logger, which sends independent weather and lightning data via the Internet to central servers. This is a unique feature of the ENTLN, where it allows for the collocation of an automated weather station. A global lightning network for long-range CG lightning detection utilizing low frequency data is also deployed but is not used in this study. An example of IC detection efficiency is shown in Figure 3 below. Figure 3: (a) An Earth Networks Lightning Sensor (ENLS) with other weather instruments mounted on a typical mast; (b) The Earth Networks Weather Station Network with more than 8,000 surface weather stations; and (c) ENTLN sensors in North America. Detection Efficiency and Classification Error IC is the most common type of lightning representing 5 to 10 times as many flashes than CG. [12] Lightning detection efficiency is a vital gauge of the total lightning performance. When lightning occurs, electromagnetic energy is emitted in all directions; many ENTLN sensors detect and record the waveforms, and then send the waveforms to a central server via the Internet. The precise arrival times are calculated by correlating the waveforms from all the sensors that detected the strokes of a flash. The waveform arrival time and signal amplitude are used to determine the stroke type (IC or CG), polarity, peak current, and exact location including latitude, longitude, and altitude. Strokes are then clustered into a flash if they are within 700 milliseconds and 10 kilometres. A flash that contains at least one return stroke is classified as a CG flash, otherwise it is classified as an IC flash. [12] ENTLN has a dense senor network of over 900 total lightning sensors. The density of the network allows lighting to be detected by a mesh of sensor and increases the accuracy of 3

4 detection. ENTLN has a detection efficiency (up to 95%+) in the US, Midwest and East where most storms occur. ENTLN classifies IC flashes with > 95% accuracy. See Figure 4. It is essential to detect a large proportion of IC flashes to accurately identify and track severe storms. Figure 4: IC Detection Efficiency over North America Lightning Cell Tracking and Dangerous Thunderstorm Alert Lightning detection networks that reliably detect a large proportion of IC flashes can be utilized to predict, detect and track areas of severe weather that can adversely impact aviation operation. Earth Networks has developed and tested a variety of methods for doing this. A lightning cell is a cluster of flashes with a boundary as a polygon determined by the flash density value for a given period. The color of the polygon represents a predefined number lightning flashes per minute thresholds that are used to help determine the severity of a storm cell. The polygon is calculated every minute with a six-minute data window. The cell tracks and directions can be determined by correlating the cell polygons over a period of time. By counting the flashes in the cell, it is possible to estimate the lightning flash rate (flashes/min), cell speed and direction and total cell area. See Figure 5. Figure 5: Lightning Cell Track. Figure A is a red lightning cell track polygon. Red indicates that >50 flashes per minute are being detected. Figure B is the cell track detail derived from algorithms that measure flash count. Flash Density polygon Storm Cell and Track. The flash data is streamed from a lightning manager service to the cell tracker as soon as a flash is located. The cell tracker keeps flashes in a moving time window of six minutes. Two gridding processes are executed every minute, using a snapshot of the flash data in that time window. The first gridding is on a coarse grid to quickly locate areas of interest and the second gridding is operated on a much finer grid using density functions to find the closed contours. To simplify the calculations, a convex polygon, which is the cell polygon at the time, is generated from each of the closed contours. In most cases, the cell polygon is similar to the previous minute polygon, so the correlation between the two polygons is straightforward. But in the case of sharp rise of the flash rate, or cell split or merger, the correlation of subsequent cells is not obvious. Special care is taken to link the cell polygons and produce a reasonable path of the moving cells. When a storm cell regroups after weakening, based on the trajectory of the cell and the time-distance of two polygons, a continuous cell path may be maintained. Once a lightning cell is located and tracked, the total flash rates, including IC and CG, are calculated. By monitoring the flash rates and the rate changes, the severe storm cells or 4

5 the ones to potentially become severe, can be identified. An alert polygon called a Dangerous Thunderstorm Alert (DTA) is issued when lightning exceeds a 25 flashes /per minute threshold for an identified cell. The alert indicates an increased threat of heavy rain, lightning, hail, convective winds and tornadic activity. Figure 6 shows an example DTA being issued for a storm cell with a total flash rate of flash/per minute. can be issued at ti. The threshold of total lightning rate may vary in different regions or different type of storms. To simplify the study, a threshold of 25 flashes per minute was chosen. Combining the information from the cells, such as the moving speed and direction and size of the cell, a warning area ahead of the storm cell can be determined. The cell may reenergize and repeat the process again and trigger more alerts. Some cells may disappear quickly and some may keep going for hours. Some storms may contain mostly CG flashes, although they are not usually severe in terms of high wind, hail or tornadoes. CG strokes can cause serious property damages and be threats to people. The alert polygon covers the distance that a cell will travel in 45 minutes with the speed demonstrated at the moment when the alert is generated. The alert polygon is updated every 15 minutes to reflect the updated path of the cell. Sufficient surface weather station density is needed to provide wind gust and rain rate data in real-time along the storm cell path. R-time weather data provides additional information for the dangerous thunderstorm alerts. Figure 6: An alert polygon can be created for the area 45 minutes ahead of the moving cell. Figure 7 shows the schematic cell history, the total lightning rate has a sudden jump at t o and the severe weather follows at ts after the rate peaks at t p. In a microburst, the pattern may show up once, while in a super cell thunderstorm the pattern can repeat many times during the lifetime of the cell. One concern in previous studies [3] is the issue when artificial trends in lightning data are strictly related to efficiency or range issues. For such reason, flash data instead of stroke data are used in the cell tracking; the latter may be affected more by the detection efficiency. The thresholds can be adjusted when the detection efficiency becomes known for different regions. Further study will be conducted in this area. Subsequently, the performance of DTA s were validated against National Weather Service (NWS) Warnings and Local Store Reports (LSR s). The study finding show that in comparison to official alerts and ground truth from LSR s the DTA s provide similar alert quality, measured with false alarm rates and other metrics, as the NWS alerts, while also providing greater warning lead time [18]. LINET also highlights severe weather alerting capabilities, however no validation studies were found. 4 Case Studies and Applications in Aviation Figure 7: Total lightning rate graph with t o = jump time; t p = peak total lightning rate time; t s = time of severe weather; t i = issuing time of Dangerous Thunderstorm Alert (DTA); t s t i represents the lead time of the alert When a cell is identified and the total lightning rate jumps passing the threshold, a dangerous thunderstorm alert (DTA) Studies have shown a presence and increase in-cloud lightning and Lightning Jump prior to convective events. [11, 13, 14] These confirm a consistent correlation between severe weather events, total lightning detection and advance warning to such severe weather. The U.S. National Transportation and Safety Board (NTSB) used ENTLN data to analyse multiple cases of inflight convective events (turbulence, thunderstorms, etc.) between 2010 to It was determined during the study that each flight case had substantial amounts of lightning detected by ENTLN prior to each of the flight incidents. See Figure 8. After a thorough review, the NTSB recommended in its May 2012 formal Safety Recommendation to the FAA the utilization of incloud and cloud-to-ground (total) lighting detection in weather displays at air route traffic- and terminal radar 5

6 control approach centers, as well as in technology used in the cockpit [17]. Figure 8: Case study highlights from the NTSB May 2012 Safety Recommendation Another case, which shows the value of in-cloud lightning detection, was an event at the Baltimore/Washington International Thurgood Marshall Airport, in Hanover, Maryland, USA, on September 12, Early that afternoon, a storm cell formed with heavy lightning, resulting in a lightning strike to the control tower that injured an employee. ENTLN detected the first in-cloud lightning 47 minutes prior to the reported time of the strike. The National Aviation Weather Program (NAWP) released its 10-year accident reduction report for It defines specific flight hazard categories and weather factor that effect both general aviation and commercial. The largest of the hazard category is turbulence and convection. See Figure 9. Figure 9: NAWP 10 year accident reduction report hazard category and weather factors assessment matrix Government aviation agencies are also integrating total lighting into their operations. Most notable is the United States Air Force Weather Agency, which utilizes data from the ENTLN as well as StreamerRT SM a weather visualization tool, to monitor total lightning and track severe weather in real time, supporting Air Force operations around the world. Total lightning data and visualization tools are also used at NASA Goddard Space Flight Center s Wallops Flight Facility to aid personnel in decision making and helping to ensure range safety during rocket launches and aircraft operations. Numerous airports and airlines also utilize these services or similar ones provided by other commercial weather companies and lightning detection network data providers. 6

7 On February 8, 2015, TAM Airlines flight JJ3307 took off from Rio de Janiero airport. Shortly after take-off the flight encountered hail associated with known thunderstorms in and around the fight path. See Figure 10. The flight was forced to make an emergency landing. The plane sustained significant damage to its radome and cockpit windshields. Flight Operations was unaware of the hail that caused the damaged and triggered the emergency landing. ENTLN data and lightning detection data sets from other providers are also available as a data feed for easy integration into aviation support applications. For example, Earth Networks provides MeteoStar with global lightning information for integration into its Flight Explorer system (Figure 12) product, an environmental analysis and display system. The integration of total lightning data enables MeteoStar to provide its clients enhanced visibility into dangerous lightning and severe storm events for improved situational awareness. The bottom line: Aviation professionals must consider integrating total lightning information into aviation operations for improved safety inflight and on the ground. Managing Convective Events Figure 10: TAM flight JJ3307 flight into thunderstorms. February 8, In commercial aviation, Orlando International Airport in central Florida (USA) is the travel hub for many of the largest vacation destinations in the US. The climate in Florida is particularly conducive to convective events. Airport officials and airline staff implemented a complete early warning system consisting of an automated weather station equipped with a lightning detection sensor, visualization and an indoor alerting tool. The system is available to provide personnel over 20 airlines with insight into weather and lightning information for informed decision making and improved on-time performance. Numerous other airports in multiple countries have implemented the same technology, including Miami International Airport, Madrid International Airport Barajas, and Dirección Nacional de Aeronáutica Civil in Paraguay. Case study evidence warrants the use of total lightning information, particularly in-cloud lightning, as an effective tool for improved visibility and increased lead-time in the prediction of convective events as recommended by the NTSB to the FAA. Utilizing real-time total lightning and precise storm cell location and tracking information has proven to improve safety in the air and on the ground, as well as enable efficient scheduling and operational continuity during severe weather events. Figure 12: ENTLN data integrated into MeteoStar s Flight Explorer system provides enhanced visibility for routing of aircraft out of harm s way 5 Conclusions Figure 11: StreamerRT image showing total lightning, cell tracks and DTAs (purple polygons) over Florida's Gulf coast Reductions in weather related aviation incident have marginally improved overall aviation safety over the last couple of decades. Improved weather forecasting and observational data has aided this effort. The use of total lightning information, particularly in-cloud lightning, can further improve visibility and increased lead time in the prediction of severe weather events as recommended by the NTSB to the FAA. The Earth Networks Total Lightning Network (ENTLN) enables total lightning detection used for real-time severe weather nowcasting and alerting applicable to all aspects of aviation operations. LINET and Viasala also 7

8 provide lightning detection network data that detects IC lightning to varying degrees. This study provided further facts about the relationship between the total lightning rate and severe weather; CG flash rate in a storm does not have a clear correlation with the severe weather activities, but IC flash rate and the rate jumps can provide early indicators of severe thunderstorms capable of producing hail, high wind or tornadoes. Most severe convective storms can generate high IC flash rate and high IC/CG flash-rate ratios. By tracking the lightning cells in a storm and monitoring the total lightning flash rates, it is possible to issue Dangerous Thunderstorm Alerts by Earth Networks (DTAs) with a lead time of up to 30 minutes before ground-level severe weather develops. ENTLN is a total lightning detection network that can detect both CG and IC flashes efficiently, and can be used to provide advance warning of severe weather. The cell tracking and Dangerous Thunderstorm Alerts (DTAs) can be used as an automated severe storm prediction tool, which can be used to augment radar, computer model data and observations to issue reliable severe weather warnings. Further study is needed to determine the appropriate Dangerous Thunderstorm Alert (DTA) threshold for various storm characteristics and relative geographic differences in ENTLN detection efficiency. Further study is also needed to correlate the Dangerous Thunderstorm Alert by Earth Networks (DTA) to NWS severe thunderstorm warning to evaluate lead time and accuracy. 6 Acknowledgements The authors would like to thank members of the Earth Networks management team, Robert S. Marshall and Christopher D. Sloop, for initiating the lightning cell tracking project and providing key inputs. We would also like to thank additional members of the Earth Networks team, Benjamin Beroukhim, Mark Hoekzema, Steve Prinzivalli and James West for using the cell tracking system and providing valuable feedback. 8

9 7 References [1] Betz, H. D., and Schmidt, K., Oettinger, W. P., and Montag, B., 2008: Cell-tracking with lightning data from LINET. Adv. Geosci., 17: [2] Betz, H. D., and Schmidt, K., Oettinger, W. P., and Wuerl, A.: Ground Truth from Lightning Detection Network LINET in Europe for the Interpretation of Thunderstorm-Related Data Collection by Space Born Sensors. EUMETSAT, file:///c:/users/janderson/downloads/pdf_conf_ P_S5_30_BETZ_V.pdf. [3] Darden, Christopher B., and Nadler, David J., Carcione, Brian C., Blakeslee, Richard J., Stano, Geoffrey T., and Buechler, Dennis E., 2010: Utilizing total lightning information to diagnose convective trends. Bulletin of the American Meteorological Society, 91: [4] Lang, T. J., and Rutledge, S. A., 2002: Relationships between convective storm kinematics, precipitation, and lightning. Mon. Wea. Rev., 130: [5] Lang, T. J., and Rutledge, S.A., Dye, J. E., Venticinque, M., Laroche, P., and Defer, E., 2000: Anomalously low negative cloud-to-ground lightning flash rates in intense convective storms observed during STERAO-A. Mon. Wea. Rev., 128: [6] Loboda, M., Betz, H. D., Baranski, P., Wiszniowski, J., and Dziewit, Z. New Lightning Detection Networks in Poland LINET and LLDN. The Open Atmospheric Science Journal, 2009, 3, [7] Perez, A. H., and Wicker, Louis J., and Orville, Richard E., 1997: Characteristics of Cloud-to-Ground lightning associated with violent tornadoes. Weather and Forecasting, 12: [8] Tuomi, T. J., and Larjavaara, M., 2005: Identification and analysis of lightning in thunderstorms. Meteorol. Soc., 131: [9] Williams, E. R., 1989: The tripole structure of thunderstorms, J. Geophys. Res., 94(D11), 13: ,167. [10] Williams, E. R., and Weber, M.E., and Orville, R.E., 1989: The relationship between lightning type and convective state of thunderclouds. J. Geophys. Res., 94(D11), [11] Williams, E.R., and Boldi, B., Matlin, A., Weber M., Hodanish, S., Sharp, D., Goodman, S., Raghavan, R., and Buechler, D., 1999: The behavior of total lightning activity in severe Florida thunderstorms. Atmospheric Research, 51: [12] ing/types, accessed June 2015 [13] Christopher J. Schultz, Walter A Peterson, Lawrence D. Carey, Preliminary Development and Evaluation of Lightning Jump Algorithms for Real-Time Detection of Severe Weather, available online from: MC2237.1/12/2009 [14] Scott Rudlosky, Ph.D. Evaluating Ground-Based Lightning Using TRMM/LIS Observations, NOAA/NESDIS/STAR, United states, 2012, pp. 4 [15] Alberto Tapia, James A. Smith, and Michael Dixon, estimation of Convective Rainfall from Lightning Observations, available online from: CO%3B2 [27/01/1997] [16] Steven J Goodman, Patrick N. Gatlin, A Total Lightning Trending Algorithm to Identify Severe Thunderstorms, University of Alabam, in Huntville, NOAA/NESDIS, available online from: CHA [07/14/2009] [17] ecommendation.pdf, accessed July 2015 [18] Paper html, accessed July