A study of National Lightning Detection Network responses to natural lightning based on ground truth data acquired at LOG with emphasis on cloud discharge activity

Abstract

The U.S. National Lightning Detection Network (NLDN) detection efficiency (DE) and classification accuracy (CA) for cloud discharge (IC) activity (identified here by a sequence of non‐return‐stroke‐type electric field pulses not accompanied by channels to ground) were evaluated using optical and electric field data acquired at the LOG (Lightning Observatory in Gainesville), Florida. Our ground truth “IC events” include 26 “isolated IC events” (complete IC flashes), 58 “IC events before first return stroke,” and 69 “IC events after first return stroke.” For the total of 153 IC events, 33% were detected by the NLDN, and the classification accuracy was 86%. For complete IC flashes, the detection efficiency and classification accuracy were 73% and 95%, respectively, and the average number of NLDN‐reported cloud pulses was 2.9 per detected event. For 24 preliminary breakdown pulse trains in CG flashes, the detection efficiency and classification accuracy were 46% and 82%, respectively. We have additionally estimated the DE and CA for return strokes in CG flashes. Irrespective of stroke order and polarity, the DE was 92% (339/367), and the CA was also 92% (312/339). The DEs for negative first and subsequent strokes were 98% and 90%, respectively.

Comparisons of Lightning Rates and Properties From the U.S. National Lightning Detection Network (NLDN) and GLD360 With GOES‐16 Geostationary Lightning Mapper and Advanced Baseline Imager Data

“…The summary line at the bottom indicates that only about half of each system's flashes are matched by the other system, so there is a population of flashes that each system detects that the other system misses. In the case of NLDN, with regard to total lightning, this is expected: The NLDN cloud flash DE has been validated against LMAs (Murphy & Nag, ) and found to be between 50% and 60%, while the CG flash DE of NLDN is around 95% (Zhu et al, , and references therein). If cloud flashes make up approximately 75% of all lightning flashes, then the overall total lightning DE of NLDN should be expected to be approximately 0.75(0.55)+0.25(0.95) = 0.65 The GLM, on the other hand, was originally expected to detect 70–90% of total lightning (Goodman et al, ).…”

Time Evolution of Satellite‐Based Optical Properties in Lightning Flashes, and its Impact on GLM Flash Detection

“…Any LIS flash that was space‐and‐time correlated with an NLDN‐defined CG flash is defined as a LIS CG flash. Given the high (>90%) CG flash detection by the NLDN (Zhang et al, 2015; Zhu et al, 2016), all other LIS flashes are defined as LIS IC flashes, even though a great portion of them had no temporally and spatially matched NLDN reports. In addition, the NLDN CG data were also used to classify the LMA flashes (IC vs. CG) employed in this study to evaluate GLM performance in the vicinity of KSC in 2018–19.…”

Characteristics of Radio Emissions Associated With Terrestrial Gamma‐Ray Flashes

“…After a network‐wide upgrade in 2013, the NLDN IC flash detection efficiency is about 50% (Murphy & Nag, ). The detection efficiency for negative first strokes in the Gainesville, Florida, region was reported to be about 98%; the IC flash and CG return‐stroke classification accuracy was reported to be 95% and 92%, respectively (Zhu et al, , ). The location accuracy for CG strokes is expected to be about a few hundred meters (Nag et al, ).…”