Abstract Predicting the timing and location of lightning initiation is a challenging but important task for operational meteorology, as many people are killed or injured each year by lightning. The advent of the GOES-‐16 satellite offers the potential for new answers to this challenge. This study takes advantage of newly available high-‐ resolution infrared imagery from GOES-‐16, along with lightning data from ground-‐ based detection systems, in order to determine any features or trends that might help predict the onset of lightning in the cloud. Using an operational NWS AWIPS workstation, cloud temperature data for over 60 thunderstorm cells in four regions of the CONUS were collected at 5-‐minute intervals and entered into spreadsheets, along with 5-‐minute total lightning counts. The data collection period for each storm was 30 minutes before lightning initiation to 30 minutes after. Preliminary results indicate that lightning initiation in storms over the US Northeast and Midwest occurs with cloud temperatures in the range -‐20 °C to -‐40 °C, averaging about -‐30 °C. However for storms in the Southeast, Great Plains and Rockies the corresponding temperatures were colder, averaging about -‐40 °C to -‐45 °C. The period of most rapid cooling typically begins a few minutes before lightning initiation but also continues for some time after. Further efforts to relate lightning initiation temperature to environmental factors have thus far been generally inconclusive. Aaron Haegele Meteorology, Geology Minor Dr. M. Majcen, Dr. C. Kauffman, Dr. G. Gould, Bill Meloy Keywords: Lightning, Remote Sensing, GOES-‐16, Cloud-‐top Temperature Acknowledgements I would like to thank my Thesis Advisor Dr. Mario Majcen and my Thesis Committee members consisting of Dr. Chad Kauffman, Dr. Greg Gould, and Bill Meloy. I would also like to express my gratitude towards the University Honors Program (UHP) at California University of Pennsylvania for the unparalleled opportunities given to me. I would also like to extend my gratitude to the staff of the National Weather Service Office in Philadelphia/Mt. Holly NJ for aiding in my research during my internship in the summer of 2017. Finally I would like to recognize the NOAA Ernest F. Hollings Undergraduate Scholarship Program for providing me with tuition assistance to support me during my studies at California University of Pennsylvania as well as a paid research internship in the summer of 2017. Haegele 1 1. Introduction Lightning is one of the most spectacular weather phenomena, yet our scientific knowledge of exactly how and when it will occur remains an elusive research problem. Each year lightning strikes the United States about 25 million times and is responsible for striking 400+ people causing about 47 fatalities and hundreds of injuries in the U.S. alone (NOAA 2015). Annually, lightning causes numerous economic impacts including damages of $1 billion in insured losses (NOAA 2010), $2 billion in aviation operating costs (Golding 2005), thirty percent of all power outages in the U.S. (Crum and Forster 2014), and 24,600 fires, both house and wildfire, averaging about $407 million in damages and an additional 16 fatalities (Ahrens 2013). The need to better understand inter-‐related lightning processes and be able to predict the onset of lightning is critical to ensuring the safety of both life and property. a. Lightning as a Risk/Danger Lightning and related threats pose serious risks/dangers to people, various business enterprises, and the environment in the United States and around the globe. In the United States, lightning strikes are annually responsible for about 47 fatalities and hundreds of injuries, with the greatest concentration in Florida (NOAA 2015). Despite the fact that the U.S. population has increased by about a factor of four since the early twentieth century, lightning fatalities have decreased from more than 400 to less than 50. During these years, large improvements in the quality of buildings, workplace environment, and overall knowledge and education of weather Haegele 2 amongst the public and scientists have lead to a decrease in fatalities. Lightning fatality incidents have shifted from agriculture and other outdoor work to largely outdoor recreation incidents (Holle 2016). Globally, lightning fatality records are much less complete. One study by Cardoso et al. (2011) estimates 6000 global fatalities and another study by Holle and Lopez (2003) estimates as much as 24,000 annual fatalities. The amount of annual lightning related fatalities differ greatly between developed and developing countries. Lesser-‐developed countries have less accessible lightning safe structures, lack of awareness or education about lightning safety, and high rate of labor-‐ intensive manual agriculture (Holle 2016). Many developing countries lie near the equator and the Inter-‐Tropical Convergence Zone (ITCZ), making the population prone to frequent lightning threats. The combination of frequent lightning threats alongside poor infrastructure and lifestyles prone to the dangers of lightning (i.e. agriculture based economy focused on manual labor, poor education, and lack of scientific knowledge) cause many developing countries to have the greatest amounts of annual lightning fatalities (Holle 2016). In the United States about 0.3 people in 1 million die from lightning strikes. Similarly in European countries the figure is about 0.2 per million, however in Zimbabwe it is about 20 per million and in Malawi it is 84 per million (Mulder et al. 2012). Lightning is responsible for 24,600 fires each year (including house and wildfires) causing about $407 million in damage, upwards of $1 billion in insured losses and an additional 16 fatalities. Lightning-‐caused house fires represent about 5% of all residential claims, averaging about 1 claim for every 57 lightning strikes. Haegele 3 In 1996, the National Park Service alone reported 517 lightning caused fires with a total cost of damages of over $19.5 million. (Crum and Forster 2014). Wildfires started by lightning burn an average of 5.5 million acres annually (NOAA 2012). Lightning-‐caused wildfires tend to be larger, burning on average 402 acres, as opposed to human-‐caused wildfires fires, which burn 45 acres (Ahrens 2013). Lightning related threats affect U.S. industry, especially the aviation and energy enterprises. Each year, lightning and related thunderstorm risks are responsible for $2 billion in aviation operating costs and passenger delays. Thunderstorm weather hazards are responsible for an average of 10 million minutes of delay annually for U.S. airports (Golding 2005). Lightning poses a serious safety risk to airport personnel that work outdoors servicing gate-‐side aircraft and maintaining airport grounds. Safety procedures trigger ramp closures and suspend outdoor work including baggage handling, food and fuel supplying, mechanical servicing and other related work. There are many uncertainties in the decision making process for lightning related weather delays, including lightning detection and forecasting. NCAR (National Center for Atmospheric Research) is working toward a better characterization of the true lightning hazard that is needed as a basis for improving the safety of outdoor personnel and minimizing avoidable operational inefficiencies. Ongoing research is combining multiple sources of relevant information (e.g., radar and various lightning data) for a robust diagnosis of lightning threats in a project called BoltAlert. Most importantly a nowcasting component is developed that enables recognition of lightning threats prior to impact allowing for proactive actions (NCAR 2017). Haegele 4 Lightning is a frequent cause of power outages to the United States electrical grid. Lightning strikes can hit electrical equipment directly or indirectly by striking nearby objects such as trees that may fall onto utility infrastructure. Lightning is responsible for about 30% of U.S. power outages totaling about $1 billion in costs. Other energy enterprises such as nuclear power plants and offshore oil rigs are often at frequent risks due to lightning (Crum and Forster 2014). b. Lightning Processes Despite its global abundance, lightning and the electrification processes, which govern initiation, are not fully understood. This is largely because lightning formation processes are complex and difficult to observe at all scales in nature. Lightning discharges span 15 orders of magnitude in scale, from atomic-‐scale electron transfer to thunderstorm dynamics tens or hundreds of kilometers in size (Harris et al. 2010). Lightning initiation is caused by the electrification process within a thunderstorm cloud, therefore the beginning of the electrification process begins with cloud dynamics and properties of a thunderstorm. Lightning is caused by differential charge distributions within the thunderstorm cloud and the ground surface. In simplest terms, moisture, atmospheric instability, and a lifting mechanism are the ingredients for thunderstorm development. Thunderstorms go through life stages including growth, development, electrification, and dissipation. Higher amounts of moisture (generally dewpoints greater than 55 °F), allow for sufficient latent heat release leading to greater instability. Atmospheric instability Haegele 5 allows air from low levels of the atmosphere to rise into upper levels, supporting deep convection and thunderstorm cloud development. Various features including frontal boundaries, dry lines, outflow boundaries, upslope flow, low-‐pressure systems, differential heating, and low level warm air or moisture advection may trigger lifting mechanisms (NWS 2004). During the growth stage, rising moist unstable air leads to convective growth of cumulus clouds due to strong updrafts. Continued growth occurs as the cumulus cloud structure develops into a towering cumulus with an anvil top. At the development stage, the towering cloud structures (containing mainly liquid particles) rise above the freezing level and critical temperature levels such as -‐15°C where glaciation occurs. Cloud droplets grow in size until the updrafts are no longer able to suspend the droplets, at which time precipitation occurs. Meanwhile, cool dry air flows downward in the cloud structure creating a downdraft. At this point the cloud structure is effectively called a cumulonimbus cloud as it contains an updraft, downdraft, and precipitation. Once precipitation processes within the cloud have formed, different phased (water, mixed, super-‐cooled and ice) droplets create electric charges within the cloud, known as the electrification stage. Eventually the downdraft in the cloud structure becomes stronger than the updraft causing a net loss of energy to the structure and therefore is known as the dissipation stage (Jensenius 2012). For purposes of this study, the dynamics and processes leading up to and during the electrification stage will be presented in most detail. Many microphysical processes occur prior to yielding lightning within strongly developed Haegele 6 cumulonimbus clouds. Lightning occurs in cold clouds containing water, supercooled water, and ice. Initially, a convective cumulus cloud contains only water droplets. However strong updrafts carry unstable moist air and water droplets above the freezing level. Precipitation processes usually begin to occur once cloud tops reach the freezing level. Cloud droplets remain liquid (supercooled) until the cloud tops ascend to temperatures around the -‐15°C where freezing known as glaciation occurs. Droplets, under strong updraft environments, may continue to ascend into levels from -‐15°C to -‐25°C. In this region supercooled droplets freeze to form larger ice crystals, graupel, or hail. These larger frozen particles are heavy and therefore either descend through the cloud or are suspended by the updraft. Lighter water droplets and ice crystals from lower levels in the cloud are ascended by the updraft and collide with the heavier frozen particles in central part of the cloud structure (Harris et al. 2010). The collision of ascending liquid droplets and suspended or descending ice particles is largely believed to be the reason for differential charge generation and distribution within a thunderstorm cloud. Collision events are most numerous in the mixed phase region of the cloud and are proportional to the intensity of the updraft. The smaller ascending liquid droplets and ice crystals become positively charged while the larger suspended or descending frozen particles become negatively charged. The updraft carries the positive charges toward the top of the cloud to temperature levels exceeding -‐40°C, while the heavier ice particles are either suspended in the middle or fall toward the lower part of the cloud due to gravity and the downdraft (Gremillion and Orville 1999). A thin layer of positive charges Haegele 7 develops at the base of the cloud structure due to warmer air temperatures and liquid precipitation properties. Figure 1.0 shows a typical charge distribution within a thunderstorm cloud. The Earth’s surface is naturally negatively charged. However the Earth’s surface is influenced by charges within overlying clouds. The abundant amount of accumulated negative charges within the mixed phase region of clouds induces a proportional positive charge upon the ground surface (and anything touching the ground) beneath the thunderstorm cloud. The electric field and gradient increases until the insulating properties of the air break down, resulting in electric discharge in the form of a lightning flash or strike. The resulting lightning discharges negative charges to one or more locations on the ground (negative flash). Lightning can also be a positive flash where there is a discharge between positive charges within a cloud and negative charges on the ground. Positive cloud-‐to-‐ ground (CG) lightning strikes are much less common than negative strikes; 90-‐95% of all CG lightning is negatively charged (Harris et al. 2010). Haegele 8 Figure 1.0 Charge Distributions within Thunderstorm Cloud and Ground Surface (ALDIS 2013) Electric discharges do not have to connect differential charges from the cloud to the ground surface. In fact, cloud-‐to-‐ground (CG) lightning is less frequent than inter-‐cloud and intra-‐cloud (IC) lightning. Inter-‐cloud lightning is caused by a discharge from negative charges in one cloud to positive charges in another cloud whereas intra-‐cloud lightning is caused by a discharge from negative charges to positive charges within the same cloud. Intra-‐cloud lightning flashes tend to briefly neutralize charge between the upper positive and mid-‐lower level negative charge regions within the cloud. IC lightning is more abundant and usually precedes CG lightning in a thunderstorm (Medici et al. 2017). A study by Melick et al. (2015) determined that the total lightning across the contiguous U.S. is composed of 85-‐ 91% IC lightning with the small remainder composed of CG. Figure 1.1 explains the Haegele 9 different types of lightning (e.g. cloud-‐to-‐ground, inter-‐cloud and intra-‐cloud) and their respective charge distributions. Figure 1.1 Different types of lightning and respective charge distributions (Beroual and Fofana 2016) c. Forecasting & Detection Operational meteorologists use a combination of radar, satellite imagery, and even recently developed numerical weather models to forecast the onset of lightning. Over the years much research regarding relationships between different types of lightning and radar and satellite signatures have been explored. Many of these studies have made significant contributions to operational meteorology and the understanding of lightning processes. Radar studies (using WSR 88-‐D) in the Haegele 10 late 1980s and 1990s relating radar reflectivity to temperature heights have proven useful in determining CG lightning activity. From these studies it has been widely accepted that lightning frequently occurs when the 40 dBz echo reaches approximately the -‐10 °C height and also that cloud tops usually have to exceed 9 km for CG lightning to occur. Michimoto (1991) discovered in their study that the first CG lightning discharge occurred approximately five minutes after the 30 dBz echo reached the -‐20 °C height, therefore providing a possible proxy to the onset of CG lightning (Gremillion and Orville 1999). More recently a research group from University of Alabama-‐Huntsville (UAH) examined summertime lightning activity and found the best index for CG flash initiation is a 25 dBz echo at -‐20°C, and a 25 dBz echo at the -‐15 °C for IC flash initiation (Mecikalski et al. 2013). Geostationary satellite studies since the 1990s have shown significant usefulness for nowcasting the development of severe storms and estimating storm intensity with lightning flash rates. Infrared (IR) interest fields from GOES satellites have proven to be able to identify locations of cloud-‐top glaciation within convective systems and also identify locations of strong updrafts, two proxies for lightning initiation. Various satellite-‐measured parameters including cloud-‐top height, temperature, and phase have been used to gain understanding of possible relationships to lightning initiation (Mecikalski et al. 2013). Numerical Weather Models such as the Weather Research and Forecasting Model (WRF) contains derived variables such as the potential electrical energy (Ep) and lightning potential index (LPI). The two variables use similar microphysical Haegele 11 processes to predict the possible production of CG and IC lightning in convection, allowing forecasts (Lynn et al. 2012). Lightning detection, until the advent of GOES-‐R, has been observed by numerous ground-‐based networks sensitive to electromagnetic pulses emitted by lightning strikes. Cloud-‐to-‐ground lightning detecting networks were the first invented and became operationally used starting in 1989 with the implementation of the National Lightning Data Network (NLDN) operated by Vaisala Inc. There are numerous ground-‐based networks but the two used in U.S. operational meteorology are the NLDN and a newer system developed by Earth Networks called Earth Networks Total Lightning Network (ENTLN). Most ground-‐based networks were originally only able to detect CG strikes and not IC. The ENTLN lead the way in detecting total lightning and differentiating between CG and IC flashes (Thompson et al. 2014). Lightning Mapping Arrays (LMA) provide 3-‐D mapping of lightning channels over a specific area. Up to thousands of points can be mapped for an individual lightning flash to reveal its location and structure. LMAs are used extensively in research to investigate lightning characteristics and how it’s related to updrafts, precipitation and severe storm processes. National research LMAs currently exist in Oklahoma, Texas, Northern Alabama, Washington D.C. and the Kennedy Space Center (NSSL 2012). The Geostationary Lightning Mapper (GLM) on board GOES-‐R satellites is an incredibly powerful lightning-‐monitoring instrument that is capable of detecting and differentiating between lightning types at 90% accuracy. The GLM uses optical Haegele 12 technology to detect visible radiances from clouds and collects information such as the frequency, location, and extent of lightning discharges to identify intensifying thunderstorms and tropical cyclones (Goodman et al. 2013). Studies have shown lightning activity trends are related to severe storm development and can be used as predictors of tornados, excessive winds, and hail events (Schultz et al. 2011, Thompson et al. 2014, Medici et al. 2017). Earlier studies in the 1990s claimed that rapid increases in lightning flash rates occur before the onset of severe weather. These abrupt increases in rates have been acceptably named “lightning jumps” within the scientific community. These studies used CG lightning data due to the ease of availability, high detection efficiency, and broad coverage across the United States and other regions. With the advent of newer lightning detection technologies, more recent studies have demonstrated that trends in total and IC lightning are more robustly correlated to severe weather occurrences than previously stated CG trends (Medici et al. 2017). A study by Schultz et al. (2011) found that average lead times prior to severe weather occurrence were higher using total lightning as compared with CG lightning (20.6 min vs. 13.5 min). 2. Project Motivation Previous studies such as Molinie et al. (2004), Mecikalski et al. (2010) Harris et al. (2010), Sieglaff et al. (2011), and Mecikalski et al. (2013) have provided insight into the usefulness of relationships between satellite imagery and convection/lightning initiation. These studies investigated cloud top temperatures Haegele 13 and cooling rates in connection to the onset of thunderstorm development (convection) or lightning in order to better predict the timing and location of either convective storm initiation or lightning initiation. Mecikalski et al. (2010) used results to formulate an algorithm called Satellite Convection Analysis and Tracking (SATCAST) system, which uses spatial temporal and spectral information from Geostationary Operational Environmental Satellites (GOES) to identify, track, and monitor growing convective clouds in their pre-‐convective state to nowcast convective initiation. These studies used older and now outdated satellite imagers including GOES-‐ 8, 10, and 12 and Meteosat-‐9. Mecikalski et al. (2010) in regards to newer and more capable forthcoming imagers stated, “Therefore, developing an understanding that guides the proper use of the ‘new’ IR channel information to improve applications such as CI & LI nowcasting will be needed.” GOES-‐R, the newest series of satellites (GOES-‐16 launched in November 2016, operational in December 2017) provides new instruments capable of better understanding convection and lightning development processes. The onboard Advanced Baseline Imager (ABI) improves every product from previous GOES imagers and introduces a host of new products. The new ABI views Earth with 16 spectral bands (compared to 5 on previous satellites), including two visible, four near-‐infrared and ten infrared channels. In addition to the 16 available channels, 25 derived products and RGB composites are newly available. The GOES-‐R series satellites provide 3x more spectral information, 4x better spatial resolution and are Haegele 14 5x faster than previous GOES imagers (NOAA-‐NASA 2017). Figure 2.0 compares the GOES-‐R Series ABI to the current (now previous) GOES. Figure 2.0 Comparison Table of GOES-‐R ABI capabilities vs. previous GOES (NOAA-‐NASA 2017) All of these previous studies also used either ground-‐based lightning networks that were capable of measuring cloud-‐to-‐ground strikes only or used lightning mapping arrays capable of detecting cloud flashes, but very limited in spatial coverage. The GOES-‐R series satellites carry a revolutionary instrument called the Geostationary Lightning Mapper (GLM), an optical transient detector capable of detecting momentary changes in an optical scene indicating the presence of lightning. The GLM is the first operational lightning mapper in geostationary orbit and is able to measure and discern IC and CG lightning flashes. GLM observable trends in IC, CG, and total lightning provide critical information to forecasters, Haegele 15 allowing them to focus on initial thunderstorm development and intensifying severe storms before these storms produce damaging winds, hail or even tornados (NOAA-‐ NASA 2017). The advent of GOES-‐16 (GOES-‐R series) provides potential for new answers regarding the difficult operational meteorology challenge of lightning prediction. With newly available instruments and capabilities, this study aims to determine any features or trends in GOES-‐16 ABI imagery, specifically cloud-‐top temperature, that might help predict the onset of lightning in the cloud. With GOES-‐16 launching in November 2016 and the study period being May-‐July 2017, this study serves as an initial research assessment of the usefulness and capabilities of GOES-‐16 products in providing new insight for operational meteorology. 3. Data & Methods a. GOES-‐R Satellite Imagery Using a standard National Weather Service (NWS) AWIPS II (Advanced Weather Interactive Processing System) workstation, GOES-‐16 ABI infrared (IR) channels were examined to determine the strongest signal for cloud-‐top cooling. IR channels are preferred because they are not limited by the amount of sunlight and therefore provide accurate measurements both during the day and night. Previous literature (Harris et al. 2010, Mecikalski et al. 2013) found the legacy GOES ABI “clean window IR” channel to be best. After reviewing all 10 IR available channels from GOES-‐16, it was determined the best signals were found in “long wave window” channels 11, 13, 14, 15, with channel 13 (10.3 μm) “clean window” Haegele 16 producing the strongest signal. Channel 13 was thus used for this study as the cloud-‐ top temperature measurement for all interested cells. Figure 3.0 shows the 10 IR channels responses to cloud-‐top temperatures in respect to lightning initiation (t=0). Although all 10 IR channels were investigated, the four longwave IR channels (shown in Figure 3.0 as blue lines) yield the most accurate cooling trend for cloud-‐ top temperatures. Figure 3.0 Temperature trends around lightning initiation for the 10 GOES-‐16 IR channels. The four long-‐wave window channels (blue) show the greatest response. Channel 13 was used for this study Following similar previous studies (Molinie et al. 2004, Sieglaff et al. 2011, Harris et al. 2010), this study focused on only isolated thunderstorm cells. Any embedded or multi-‐cell thunderstorms such as mesoscale convective systems Haegele 17 (MCS), squall lines and other linear features were not included for data quality purposes. Multi-‐cell and other embedded systems were proven to disturb lightning flash data and made it difficult to decipher which cell was responsible for the production of lightning. Isolated single cells made this process much easier and more accurate. The temporal scale for this study was 30 minutes prior and after lightning initiation (defined at t=0). GOES-‐16 ABI imagery is available for the Continental United States (CONUS) (5000km x 3000 km) every 5 minutes and at either 1 minute or 30 seconds for smaller viewing regions known as “mesoscale sectors” (1000km x 1000km). The mesoscale viewing sectors are able to be re-‐positioned to focus on various interest areas across the CONUS and even the hemisphere. At the time of this study, GOES-‐16 was still in preliminary stage and therefore mesoscale sector imagery often had data blackouts and/or were not available due to frequent viewing jumps during instrumentation tests. This study took place from May-‐July 2017 and during this time period the mesoscale sector was often positioned over tropical storms and/or severe storms of special interest. However, the sector was re-‐positioned frequently, making data collection for 1 hour very unreliable. For these reasons, the CONUS 5-‐ minute imagery was used because it provided the most constant and reliable data for cloud-‐top temperature measurements. Figure 3.1 shows a comparison of the cloud top temperatures using the legacy GOES-‐13 15-‐minute (green triangles) and newly available GOES-‐16 5-‐minute (red squares) and 1-‐minute (blue diamonds) imagery during a preliminary investigation. Note the increase in data points with Haegele 18 the GOES-‐16 imagery and therefore the more detailed cooling trend as compared to the legacy GOES-‐13 imagery every 15 minutes. Figure 3.1 Comparison example of 1-‐min and 5-‐min (GOES-‐16) and legacy 15-‐ min (GOES-‐13) cloud-‐top temperatures for one storm. b. Lightning Data Earth Networks Total Lightning Network (ENTLN) intra-‐cloud (IC) pulses were used for the lightning initiation data. Preliminary investigations showed that IC pulses preceded both cloud flashes and CG strikes and were best proxy signal for initial cloud electrification and lightning initiation. ENTLN boasts greater than 89% total lightning detection efficiency for CONUS (Liu and Heckman 2011). Combined with linked temporal scales to the GOES-‐16 ABI imagery and the unavailability of Haegele 19 GOES-‐16 GLM data, ENTLN is the best alternate IC lightning detection method for the study. ENTLN provides 1-‐min, 5 minute, and 15-‐ minute lightning data. The 5-‐ minute data were used because it allowed for real-‐time comparison of lightning data and the GOES-‐16 ABI data. Figure 3.2 shows an example of sampled cloud-‐top temperatures and various lightning counts (using AWIPS II) for a cell near Baltimore, MD. Note the equal time stamps for both the ENTLN and GOES-‐16 data. Also note the offset in location of the cloud-‐top temperatures and lightning due to parallax from the satellite-‐viewing angle. Figure 3.2 Sampled Cloud-‐Top Temperature of a cell located near Baltimore, MD and time correlated lightning data Haegele 20 c. Soundings Investigations into further understanding lighting initiation relationships beyond cloud-‐top temperatures led to examining soundings for possible atmospheric parameter relationships. A Rapid Refresh (RAP) model analysis sounding corresponding to the approximate time and location of each cell was obtained from a freely available source at: http://mtarchive.geol.iastate.edu. RAP soundings provide spatially and temporarily efficient sounding parameters easily displayed in BUFKIT software. METAR (Meteorological Terminal Aviation Routine Weather Report) stations were displayed on AWIPS II during the sampling of cloud-‐ top temperatures and lightning data, making it easy to determine close proximity sounding stations. This study selected sounding parameters including CAPE, Lifted Index, Equilibrium Level, Precipitable Water, LCL, LFC, WBZ and 500-‐700mb Lapse Rates. Figure 3.3 shows an example of a sounding used in the study using BUFKIT. Note the measured sounding parameters (in orange) in the black table to the left. These values were then used for analyzing relationships with cloud-‐top temperature at lightning initiation and sounding indices. Haegele 21 Figure 3.3 Example of a sounding used in BUFKIT and measured indices d. Procedures Cloud-‐top temperatures using GOES-‐16 ABI Channel 13 (10.3 μm) CONUS 5-‐ minute imagery were recorded for isolated cells 30 minutes prior and after the first IC pulse(s) as determined by corresponding ENTLN 5-‐minute lightning data. Selected cells from June-‐July 2017 were classified into four CONUS regions: Northeast, Southeast, Midwest, and Plains-‐Rockies. Figure 3.4 shows the area of study displaying all cells divided into corresponding regions. Blue circles mark Northeast cells, green mark Midwest cells, purple mark Southeast cells, and red mark Plains-‐Rockies cells. In total, 60 cells and corresponding data were collected. Haegele 22 Figure 3.4 Map of area study displaying all cells for each region Linear regressions and simple statistics were used to determine trends in cloud-‐top temperatures in relationship to lightning initiation and total lightning across the various regions. Further investigation into possible relationships between lightning initiation and atmospheric parameters (i.e. CAPE, LI, LCL, LFC, WBZ, etc.) were then examined using RAP model analysis soundings for nearest (in relation to the cell position) and time approximate METAR stations. Linear regressions and simple statistics were used to determine relationships between lightning initiation and sounding indices. 4. Results a. Cloud-‐Top Temperature at Lightning Initiation Cloud-‐top temperatures at lightning initiation vary from region to region in storms over the over the United States. Table 4.0 shows the minimum, mean, and Haegele 23 maximum cloud-‐top temperatures at lightning initiation as well as standard deviation. Northeast and Midwest storms reflect similar temperature profiles with cloud temperatures in the range of -‐20°C to -‐40°C, averaging about -‐30°C. Cloud-‐top temperatures for Northeast storms range from -‐19.8°C to -‐40.7°C, averaging -‐30.0°C. Midwest storms occur at temperature ranges from -‐19.7°C to –42.0°C, averaging -‐30.1°C. Storms in the Southeast, Great Plains and Rockies share similar temperature profiles with colder temperatures and broader ranges than those in Northeast and Midwest storms. Lightning initiation in Southeast storms tend to occur in clouds with temperatures ranging from -‐17.7°C to -‐55.3°C, averaging -‐38.5°C. Lightning initiation occurs in the Great Plains and Rockies in cloud temperatures ranging from -‐19.6°C to -‐55.5°C, averaging -‐42.7°C. Figure 4.0 shows the trends in average cloud-‐top temperature for cells in each region from 30 minutes prior to lightning initiation (t=-‐30), through lightning initiation (t=0), to 30 minutes after lighting initiation (t=30). Figure 4.1 is a box and whisker diagram to visually display the distribution of cloud-‐top temperatures at lightning initiation for each region. Northeast Midwest Southeast Plains-‐ Rockies Max Temp (°C) -‐19.8 -‐19.7 -‐17.7 -‐19.6 Mean Temp (°C) -‐30.0 -‐30.6 -‐38.5 -‐42.7 Min Temp (°C) -‐40.7 -‐42.0 -‐55.3 -‐55.5 Standard 6.78 7.08 11.45 9.98 Deviation Table 4.0 Regional Maximum, Mean, Minimum Temperatures at Lightning Initiation (LI) and Standard Deviation Haegele 24 Figure 4.0 1-‐hr Regional Average Cloud-‐Top Temperature Trends Centered at Lightning Initiation (LI) Figure 4.1 Regional Cloud-‐Top Temperatures at Lightning Initiation Box and Whisker Diagram Haegele 25 b. Cloud-‐top Temperature Cooling Rates During the 1-‐hour period, centered at lightning initiation, various cooling trends occur in cloud-‐top temperature across the four regions. Table 4.1 shows starting (t= -‐30 min) temperatures, ending (t= 30 min) temperatures, total cooling, maximum cooling rate (°C/5-‐minutes), and the maximum cooling rate time in relation to lightning initiation. Cells in the Southeast have the greatest total cooling of -‐56.7°C, followed by cells in the Midwest with total cooling of -‐50.0°C, Great Plains and Rockies with total cooling of -‐46.1°C, and cells in the Northeast have the least amount of total cooling of -‐41.4°C. Maximum cooling rates (°C/5-‐minutes) are greatest in Southeast cells at -‐12.6, followed by Midwest cells at -‐8.9, Northeast cells at -‐8.3, and cells in the Great Plains and Rockies have the weakest cooling rates at -‐7.3. Cells in all regions show greatest cooling rates at or near the time of lightning initiation except for cells in the Great Plains and Rockies where maximum cooling rates occur ten minutes prior to lightning initiation. Figure 4.2 shows average regional cloud-‐top 5-‐minute cooling trends from 30 minutes prior to 30 minutes after lightning initiation. Start (°C) End (°C) Northeast Midwest Southeast Plains-‐ Rockies -‐8.2 -‐0.6 -‐2.2 -‐9.3 -‐49.6 -‐50.6 -‐58.9 -‐55.4 Total Cooling (°C) -‐41.4 -‐50.0 -‐56.7 -‐46.1 Max Cooling Rate (°C/5min) -‐8.3 -‐8.9 -‐12.6 -‐7.3 Time in Relation to LI (min) 0 0 0 -‐10 Table 4.1 Regional Cloud-‐Top Temperature Endpoints and Cooling Trends Haegele 26 Figure 4.2 Regional Average Cloud-‐Top Temperature Cooling Trends c. Cloud-‐Top Temperature at LI and Total IC Lightning Linear regressions of cloud-‐top temperatures at LI and storm total IC lightning yielded moderately correlated relationships. Figure 4.3 shows these regressions and r-‐values. In Northeast cells there was a moderate negative relationship between LI temperature and total storm IC with an r-‐value of 0.64. Thus, cooler temps at LI tend to lead to higher amounts of total storm IC lightning. Cells in the Southeast and Midwest had similar moderate negative relationships with r-‐values of 0.65 and 0.64 respectively. Cells in the Great Plains and Rockies had a slightly weaker, but still moderate, positive relationship with an r-‐value of 0.54. In this region, unlike the others, cooler temps at LI did not correlate with greater amounts of total IC lightning. Haegele 27 Figure 4.3 Linear Regressions of Regional Cloud-‐Top Temperature & Total IC Lightning d. Sounding Indices Linear regressions of cloud-‐top temperatures at lightning initiation and sounding indices of special interests (i.e. CAPE, PW, WBZ, LCL, LFC, LI, EL, and 500-‐ 700 mb lapse rate) yield various results with no region-‐to-‐region correlation signatures. Most regressions yielded very weak (r < 0.3) correlations except for a few “stand out” moderate correlations between cloud-‐top temperature and sounding indices. Figure 4.4 shows the only four correlations with r > 0.50. Cloud-‐ top temperatures at lightning initiation in Southeast cells had moderate negative relationships between precipitable water (PW) and also wet bulb zero (WBZ) heights with respective r-‐values of 0.68 and 0.58. Cloud-‐top temperatures at lightning initiation in Northeast cells had moderately negative correlation, r = 0.55, Haegele 28 with the height of the level of free convection (LFC). Cloud-‐top temperatures at lightning initiation in Great Plains and Rockies cells had moderately negative correlation, r = 0.51, with the convective available potential energy (CAPE). Figure 4.4 Regressions of Cloud-‐Top Temperature at Lighting Initiation and Sounding Parameters 5. Discussion Thunderstorm cloud-‐top temperature relative to lightning initiation (LI) varies considerably from region to region and even from storm to storm within the studied regions. However, storms in the Northeast and Midwest showed quite similar temperature trends especially from around the time of LI to 30 minutes after. Average cloud-‐top temperatures at LI for the Northeast were -‐30.0°C and -‐30.6°C for Midwest cells. Minimum (warmest) and maximum (coldest) cloud-‐top temperatures were also quite similar for cells in these two regions. The maximum Haegele 29 temperature for lightning initiation is -‐19.8°C in the Northeast and -‐19.7°C in the Midwest. Minimum temperatures at LI were -‐40.7°C in the Northeast and -‐42.0°C in the Midwest. Cells in the Northeast and Midwest also have less spread in temperatures at LI. Storms in the Southeast, Great Plains and Rockies had colder cloud-‐top temperatures at LI and also much larger spread than storms in the Northeast and Midwest. Average cloud-‐top temperature for Southeast cells were -‐38.5°C and in the Great Plains and Rockies -‐42.7°C. Maximum temperatures at LI were -‐17.7°C in the Southeast and -‐19.6°C. Minimum temperatures at LI were nearly identical for both regions with -‐55.3°C for Southeast cells and -‐55.5°C for cells in the Great Plains and Rockies. Despite variation in average and even minimum cloud-‐top temperatures at LI for various regions, the maximum cloud-‐top temperatures at LI were very similar for all regions. In all regions the warmest temperatures at LI were -‐17.7 to -‐19.8°C. Maximum temperatures at LI for cells in the Northeast, Midwest, and the Southeast were 0.1-‐0.2°C different from one another and 1.9-‐2.1°C different from maximum temperatures at LI for cells in the Southeast. These results reinforce the fact that glaciation in cloud tops generally begins around -‐15 to -‐20°C and, therefore, since ice particles play a significant role in charge distribution within a cloud, LI does not occur earlier or at warmer temperatures. The overall maximum cloud-‐top cooling and cooling rate occurred in storms in the Southeast, especially along the Gulf Coast. Maximum cooling rates occurred at or very near the time of lightning initiation for cells in the Northeast, Southeast, and Haegele 30 Midwest while cells in the Great Plains and Rockies had maximum cooling rates about 10-‐minutes prior to lightning initiation. There was moderate correlation between cloud-‐top temperature at LI and the amount of IC lightning over the next 30 minutes in storms over all four regions. Storms in the Northeast, Southeast, and Midwest had greater amounts of cloud lightning when lightning initiation occurred at colder temperatures. Therefore the cloud-‐top temperature at lightning initiation can possibly determine the intensity of storms, in terms of IC lightning. Opposingly, storms in the Great Plains and Rockies had greater amounts of cloud lightning at warmer temperatures. However, more samples from this region would help determine the strength of the correlation. As shown in Figure 4.3, it appears the direction of the trend line is highly influenced by two data points near -‐20°C that show very high lightning totals. Larger sample sizes would help determine if this opposing relationship, compared to the other regions, is valid or not. Further efforts to relate lightning initiation temperature to atmospheric parameters measurable by soundings have been generally inconclusive. There were no significant region-‐to-‐region correlations between RAP sounding parameters and cloud-‐top temperatures and lightning activity. There were some “stand out” moderate correlations including cloud-‐top temperature at LI and precipitable water (PW) and also wet bulb zero (WBZ) heights for storm in the Southeast. Colder cloud-‐ top temperatures coincided with greater PW values. PW values ranged from approximately 1.5 to 2.25” with values around 2-‐2.25” coinciding with colder cloud-‐ top temperatures and lesser amounts of cloud lightning. Wet bulb zero heights for Haegele 31 storms in the Southeast occurred from approximately 10,000 to 13,600 feet with higher WBZ heights coinciding with colder temperatures and lesser amounts of cloud lightning. In the northeast, storms had level of free convection (LFC) heights ranging from approximately 2500 to 6500 feet with higher LFC heights generally coinciding with colder cloud-‐top temperatures at lightning initiation and lesser amounts of cloud lightning. In the Great Plains and Rockies, storms in environments with more convective available potential energy (CAPE) coincided with colder cloud-‐top temperatures and higher amounts of cloud lightning. Subsequent work with larger sample sizes, including GOES-‐16 ABI derived products and 1-‐minute imagery, and especially optical lightning data from the Geostationary Lightning Mapper (GLM), would be beneficial in confirming results described herein. Increasing the sample size to include more cells for each region is critical in the accuracy of the statistics determined in the results. Including storms occurring throughout the year or a least including some seasonal variation would be beneficial in determining if there are any inter-‐seasonal signatures as well in cloud-‐ top temperatures and lightning activity. Increasing temporal resolution by using 1-‐ minute satellite imagery may allow for greater detection of cloud-‐top temperature signatures. The most anticipated supplement would be incorporating GLM data. Using GLM data, would allow for validation of ground-‐based lightning detection systems, such as Earth Networks used in this study, and also validate the enhanced usefulness of lightning detection that the GLM promises. Haegele 32 References Ahrens, M., 2013: Lightning Fires and Lightning Strikes. National Fire Protection Association (NFPA). NFPA No. USSS1, 31pp. ALDIS, 2013: Lightning Discharge. Accessed 01 February 2018, https://www.aldis.at/en/lightning-‐research/lightning-‐physics/lightning-‐ discharge/ Beroual A., and I. Fofana, 2016: Lightning Discharge. Discharge in Long Air Gaps Modelling and Applications, IOP Publishing Ltd, 5/1-‐5/19. Cardoso, I., O. Pinto Jr., I. R. C. A. Pinto, and R. L. Holle, 2011: A new approach to estimate the annual number of global lightning fatalities. Preprints, 14th Int. Conf. on Atmospheric Electricity, Rio de Janeiro, Brazil, IUGG/IAMAS International Commission on Atmospheric Electricity, 4 pp. Crum & Forster, 2014: Lightning. United States Fire Insurance Company (USFIC) Loss Control Technical Guide, 32pp. Golding W. L., 2005: Lightning Strikes on Commercial Aircraft: How the Airlines are Coping. Journal of Aviation/Aerospace Education & Research., 15, 41-‐50, https://doi.org/10.15394/jaaer.2005.1511. Goodman, S.J., and Coauthors, 2013: The GOES-‐R Geostationary Lightning Mapper (GLM). Atmopsheric Research, 125-‐126, 34-‐49, https://doi.org/10.1016/j.atmosres.2013.01.006. Gremillion M.S., and R.E. Orville, 1999: Thunderstorm Characteristics of Cloud-‐to-‐ Ground Lighting at the Kennedy Space Center, Florida: A Study of Lightning Initiation Signatures as Indicated by the WSR-‐88D. Wea. Forecasting, 14, 640-‐649. Harris, R.J., J.R. Mecikalski, W.M. MacKenzie, P.A. Durkee, and K.E. Nielsen, 2010: The Definition of GOES Infrared Lightning Initiation Interest Fields. J. Appl. Meteor. Climatol.,49, 2527–2543, https://doi.org/10.1175/2010JAMC2575.1 Holle, R.L., and R. E. López, 2003: A comparison of current lightning death rates in the U.S. with other locations and times. Pre-‐ prints, Int. Conf. on Lightning and Static Electricity, Black-‐ pool, United Kingdom, Royal Aeronautical Society, P103-‐34. Holle, R.L., 2016: A Summary of Recent National-‐Scale Lightning Fatality Studies. Wea. Climate Soc., 8, 35-‐42, https://doi.org/10.1175/WCAS-‐D-‐15-‐0032.1. Haegele 33 Jensenius, J.S., 2012: Understanding Lightning: Thunderstorm Electrification. NOAA, Accessed 01February 2018, http://www.lightningsafety.noaa.gov/science/science_ele Liu, C., S. Heckman, 2011: Using Total Lightning Data in Severe Storm Prediction: Global Case Study Analysis from North America, Brazil and Australia. Proc. IEEE International Symposium on Lightning Protection 2011, Fortaleza, Brazil, Institute of Electrical and Electronics Engineers, 20-‐24, https://doi.org/10.1109/SIPDA.2011.6088433 Lynn, B.H., Y. Yair, C. Price, G. Kelman, and A.J. Clark, 2012: Predicting Cloud-‐to-‐ Ground and Intracloud Lightning in Weather Forecast Models. Wea. Forecasting, 27, 1470–1488, https://doi.org/10.1175/WAF-‐D-‐11-‐00144.1. Mecikalski, J.R., W.M. MacKenzie, M. Koenig, and S. Muller, 2010: Cloud-‐Top Properties of Growing Cumulus prior to Convective Initiation as Measured by Meteosat Second Generation. Part I: Infrared Fields. J. Appl. Meteor. Climatol., 49, 521–534, https://doi.org/10.1175/2009JAMC2344.1. Mecikalski, J.R., X. Li, L.D. Carey, E.W. McCaul, and T.A. Coleman, 2013: Regional Comparison of GOES Cloud-‐Top Properties and Radar Characteristics in Advance of First-‐Flash Lightning Initiation. Mon. Wea. Rev., 141, 55–74, https://doi.org/10.1175/MWR-‐D-‐12-‐00120.1. Medici, G., K.L. Cummins, D.J. Cecil, W.J. Koshak, and S.D. Rudlosky, 2017: The Intracloud Lightning Fraction in the Contiguous United States. Mon. Wea. Rev., 145, 4481–4499, https://doi.org/10.1175/MWR-‐D-‐16-‐0426.1. Melick, C.J., P. Marsh, A. Dean, I.L. Jirak, and S.J. Weiss: 2015: Lightning Characteristics and Relationship to Preliminary Local Storm Reports. Preprints, 40th Natl. Wea. Assoc. Annual Meeting, Oklahoma City, OK, Natl. Wea. Assoc. Molinie, G., and A.R. Jacobson, 2004: Cloud-‐to-‐ground lightning and cloud top brightness temperature over the contiguous United States. Journal of Geophysical Research, 109, D13106.1-‐D13106.16, https://doi.org/10.1029/2003JD003593. Mulder, M.B., L. Msalu, T. Caro, and J. Salerno, 2012: Remarkable Rates of Lightning Strike Mortality in Malawi. Public Library of Science, 7, https://doi.org/10.1371/journal.pone.0029281 NCAR, 2017: BoltAlert® : Predicting Lightning Threats. Accessed 01 February 2018, https://ral.ucar.edu/sites/default/files/public/media/files/bolt-‐alert-‐ral-‐ broch-‐2017-‐10.pdf Haegele 34 NOAA, 2010: Thunderstorms, Tornadoes, Lightning… A Preparedness Guide. National Weather Service, 18pp, http://www.lightningsafety.noaa.gov/resources/ttl6-‐10.pdf NOAA, 2012: Lightning and Fires. Accessed 01 February 2018, http://www.lightningsafety.noaa.gov/fire.shtml NOAA, 2015: Lightning Safety. Accessed 01 February 2018, http://www.lightningsafety.noaa.gov/ NOAA, NASA, 2017: Advanced Baseline Imager (ABI). Accessed 01 February 2018, https://www.goes-‐r.gov/spacesegment/abi.html NOAA, NASA, 2017: Geostationary Lightning Mapper (GLM). Accessed 01 February 2018, https://www.goes-‐r.gov/spacesegment/glm.html NSSL, 2012: Lightning Detection. Accessed 01 February 2018, https://www.nssl.noaa.gov/education/svrwx101/lightning/detection/ NWS, 2004: Training Page: Thunderstorm Ingredients. Accessed 01 February 2018, https://www.weather.gov/source/zhu/ZHU_Training_Page /thunderstorm_stuff/Thunderstorms/thunderstorms.htm#INDEX. Schultz, C.J., W.A. Petersen, and L.D. Carey, 2011: Lightning and Severe Weather: A Comparison between Total and Cloud-‐to-‐Ground Lightning Trends. Wea. Forecasting, 26, 744–755, https://doi.org/10.1175/WAF-‐D-‐10-‐05026.1 Sieglaff, J.M., L.M. Cronce, W.F. Feltz, K.M. Bedka, M.J. Pavolonis, and A.K. Heidinger, 2011: Nowcasting Convective Storm Initiation Using Satellite-‐Based Box-‐ Averaged Cloud-‐Top Cooling and Cloud-‐Type Trends. J. Appl. Meteor. Climatol., 50, 110–126, https://doi.org/10.1175/2010JAMC2496.1 Thompson, K.B., M.G. Bateman, and L.D. Carey, 2014: A Comparison of Two Ground-‐ Based Lightning Detection Networks against the Satellite-‐Based Lightning Imaging Sensor (LIS). J. Atmos. Oceanic Technol., 31, 2191–2205, https://doi.org/10.1175/JTECH-‐D-‐13-‐00186.1