More Like this

1. The RELAMPAGO Lightning Mapping Array: Overview and Initial Comparison with the Geostationary Lightning Mapper

   https://doi.org/10.1175/JTECH-D-20-0005.1  

   Lang, Timothy J.; Ávila, Eldo E.; Blakeslee, Richard J.; Burchfield, Jeff; Wingo, Matthew; Bitzer, Phillip M.; Carey, Lawrence D.; Deierling, Wiebke; Goodman, Steven J.; Medina, Bruno Lisboa; et al (August 2020, Journal of Atmospheric and Oceanic Technology)

   Abstract During November 2018–April 2019, an 11-station very high frequency (VHF) Lightning Mapping Array (LMA) was deployed to Córdoba Province, Argentina. The purpose of the LMA was validation of the Geostationary Lightning Mapper (GLM), but the deployment was coordinated with two field campaigns. The LMA observed 2.9 million flashes (≥ five sources) during 163 days, and level-1 (VHF locations), level-2 (flashes classified), and level-3 (gridded products) datasets have been made public. The network’s performance allows scientifically useful analysis within 100 km when at least seven stations were active. Careful analysis beyond 100 km is also possible. The LMA dataset includes many examples of intense storms with extremely high flash rates (>1 s−1), electrical discharges in overshooting tops (OTs), as well as anomalously charged thunderstorms with low-altitude lightning. The modal flash altitude was 10 km, but many flashes occurred at very high altitude (15–20 km). There were also anomalous and stratiform flashes near 5–7 km in altitude. Most flashes were small (<50 km2 area). Comparisons with GLM on 14 and 20 December 2018 indicated that GLM most successfully detected larger flashes (i.e., more than 100 VHF sources), with detection efficiency (DE) up to 90%. However, GLM DE was reduced for flashes that were smaller or that occurred lower in the cloud (e.g., near 6-km altitude). GLM DE also was reduced during a period of OT electrical discharges. Overall, GLM DE was a strong function of thunderstorm evolution and the dominant characteristics of the lightning it produced. 

   more » « less
2. Transformer‐based DNA methylation detection on ionic signals from Oxford Nanopore sequencing data

   https://doi.org/10.15302/J-QB-022-0323  

   Wang, Xiuquan; Ahsan, Mian Umair; Zhou, Yunyun; Wang, Kai (September 2023, Quantitative Biology)

   Transformer is an algorithm that adopts self‐attention architecture in the neural networks and has been widely used in natural language processing. In the current study, we apply Transformer architecture to detect DNA methylation on ionic signals from Oxford Nanopore sequencing data. We evaluated this idea using real data sets (Escherichia colidata and the human genome NA12878 sequenced by Simpsonet al.) and demonstrated the ability of Transformers to detect methylation on ionic signal data. BackgroundOxford Nanopore long‐read sequencing technology addresses current limitations for DNA methylation detection that are inherent in short‐read bisulfite sequencing or methylation microarrays. A number of analytical tools, such as Nanopolish, Guppy/Tombo and DeepMod, have been developed to detect DNA methylation on Nanopore data. However, additional improvements can be made in computational efficiency, prediction accuracy, and contextual interpretation on complex genomics regions (such as repetitive regions, low GC density regions). MethodIn the current study, we apply Transformer architecture to detect DNA methylation on ionic signals from Oxford Nanopore sequencing data. Transformer is an algorithm that adopts self‐attention architecture in the neural networks and has been widely used in natural language processing. ResultsCompared to traditional deep‐learning method such as convolutional neural network (CNN) and recurrent neural network (RNN), Transformer may have specific advantages in DNA methylation detection, because the self‐attention mechanism can assist the relationship detection between bases that are far from each other and pay more attention to important bases that carry characteristic methylation‐specific signals within a specific sequence context. ConclusionWe demonstrated the ability of Transformers to detect methylation on ionic signal data. 

   more » « less
3. Bipolar Lightning Flashes With Multiple Ground Terminations Including Tall Towers

   https://doi.org/10.1029/2025JD045328  

   Chen, S; Rakov, V A; Zhu, Y; Ding, Z (February 2026, Journal of Geophysical Research: Atmospheres)

   Abstract Three large‐scale (tens of kilometers in horizontal extent) multiple‐stroke cloud‐to‐ground (CG) flashes were examined in detail. The flashes occurred in two summer thunderstorms in Florida and were bipolar in that some of the strokes (leader/return‐stroke sequences) in a given flash transported to ground positive charge, while others transported negative charge. Positive strokes tended to have higher NLDN‐reported peak currents and all terminated essentially on the ground surface, each time forming a new channel. Negative strokes tended to terminate on tall towers and all that did so produced characteristic wideband E‐field signatures that exhibited oscillations after the initial peak. Additionally, we examined a negative flash composed of seven strokes all of which terminated on a tall tower. There were a total of three tall (451–497 m) towers involved, with one of them (497‐m tall) being strike object for each of the four flashes examined here. The maximum GLM group energy for positive strokes was one to two orders of magnitude higher than for negative strokes. The oscillating E‐field signatures are indicative of bouncing current waves that are manifestations of the transient response of a tall object to the lightning‐caused excitation at or near its top (e.g., Rakov, 2001,https://doi.org/10.1109/15.974646); it is due to impedance discontinuities at the lightning attachment point and at the ground. The period of E‐field oscillations can be linked to the tower height. The oscillations were detectable both in the ground wave and in the first skywave. 

   more » « less
4. Continental‐Scale Lightning Observations at High Frequency

   https://doi.org/10.1029/2025GL116075  

   Pu, Yunjiao; Cummer, Steven A; Jia, Yongze (October 2025, Geophysical Research Letters)

   Abstract We initiated high‐frequency (HF, 3–30 MHz) lightning observations to remotely study lightning processes using skywave propagation. HF lightning skywaves were detected at distances up to 3,300 km (potentially farther) during both day and night, with over 64% of events having peak currents below 50 kA. HF measurements revealed distinct temporal features of negative cloud‐to‐ground (CG) leaders, narrow bipolar events, and negative breakdown following positive return strokes. Unlike VLF and LF methods, which primarily capture current pulses along lightning channels, HF observations detect both transient and prolonged leader emissions. This provides more equal sensitivity to both in‐cloud (IC) and CG events, potentially benefiting IC/CG classification and studies of IC‐associated phenomena such as terrestrial gamma‐ray flashes. Ionospheric conditions may influence HF signal propagation but can be mitigated by observing the full HF band. These findings demonstrate the potential of HF observations for advancing lightning physics and storm monitoring over long distances. 

   more » « less
5. Large-scale, automated detection of gigantic jets using machine learning and sensor fusion

   Boggs, L; Cohen, M; Mach, D; Cummer, S; Trostel, J; Burke, J; Smith, J; Peterson, M (December 2024, AGU fall meeting 2024)

   Gigantic jets (GJs) are a type of transient luminous event (TLE) which also includes sprites, elves, halos, and blue jets [Pasko2010, doi: 10.1029/2009JA014860]. However, GJs are unique in that they directly couple electric charge reservoirs in the troposphere (i.e. thunderclouds) with the lower ionosphere, allowing significant amounts of charge (100s of C) to flow between these regions. We do not understand how this affects the ionosphere and global electric circuit. Past observations are very limited, resulting from ground-based cameras getting lucky enough to capture an event while looking over a distant thunderstorm [Liu et al. 2015, doi: 10.1038/ncomms6995]. Additionally, GJ-producing storms are normally accompanied by substantial areas of stratiformclouds obscuring the view, and they tend to occur more often over the ocean. To solve this problem of limited detection capability, we have developed a pipeline that utilizes machine learning and sensor fusion of multiple sensing modalities (optical, VLF, ELF). Our pipeline can detect GJs over nearly a hemisphere and operate 24/7, potentially revolutionizing how GJs are detected and paving the way for other TLE and unique lightning event detection. Our pipeline begins by performing detection with data from the Geostationary Lightning Mapper (GLM), which is a staring optical imager in geostationary orbit that detects the 777.4 nm (OI) triplet from lightning leaders [Goodman et al. 2013, doi: 10.1016/j.atmosres.2013.01.006]. Gigantic jets have unique signatures in the GLM data from past studies [Boggs et al. 2019, doi: 10.1029/2019GL082278]. We have developed a supervised, ensemble machine learning classifier that detects potential gigantic jets in the GLM data. Considering we have an imbalanced dataset, we use data imbalance techniques such as Synthetic Minority Oversampling Technique (SMOTE) when training the classifier. Next, we combine data from multiple sensing modalities to vet the candidate GJs from the classifier in multiple stages. The first stage filters the candidate GJs to the stereo GLM region [Mach and Virts, 2021, doi: 10.1175/JTECH-D-21-0078.1], and calculates the stereo altitudes for all the events. GJs have stereo altitude sources consistently between 15-25 km altitude from the leader escaping the cloud top [Boggs et al. 2022, doi: 10.1126/sciadv.abl8731]. Next, we match the events spatiotemporally to GLD360 data to remove cloud-to-ground events. Subsequently, we use a statistical GOES ABI model (developed at GTRI) to filter out events that have differing parent storms from our truth database. Finally, we use a multi-parameter extremely low frequency (ELF) vetting model (developed by Duke) to filter out the remaining non-GJ events. After a few complete detection and vetting cycles, we have found tens of new events with a high degree of confidence. With further development of our pipeline and deployment to the entire GLM field-of-view (not limited to stereo region), we anticipate hundreds of new detections per year, significantly more than ground-based cameras, allowing for comprehensive studies relating gigantic jets to the other atmospheric phenomena 

   more » « less