Characteristics of Long Continuing Currents Observed in Broadband ELF Measurements

Continuing current (CC) is a slowly varying lightning current that can follow a return stroke (RS) in cloud‐to‐ground lightning flashes and typically lasts from a few milliseconds to several hundred milliseconds. The necessary conditions and generation mechanism of CC formation have been studied for decades, motivated by the increased risk of physical lightning damage due to the long‐lasting current and large charge transfers. Nowadays, there is a growing interest in the study of CC, mainly because of the important role it plays in the natural ignition of forest fires. The Krakow ELF group operates a pair of broadband magnetic antenna (sampling frequency: 3004.81 Hz, antenna bandwidth: 0.02 Hz – 1.1 kHz) in an electromagnetically very quiet environment in Hylaty, south-eastern Poland, which is very suitable for the recording of CCs. In this contribution, we introduce our semi-automated procedure for detecting and characterizing CCs in this measurement data and describe some characteristics of the long CCs (>40 ms) identified by our method on three selected days (3-5 July, 2025). Our algorithm first searches for peaks in the magnetic data that represent the ELF manifestation of RSs, then estimates the beginning and end of the waveform based on classic signal processing techniques. The time passed between the RS peak and the end of the waveform is considered to be the CC duration. In order to increase the reliability of the data system, we manually discard ambiguous cases and correct the estimated CC durations. Over the three selected days, a total of 7,052 RSs have been detected, of which 349 (~5%) were followed by a clear and long lasting CC signature. 90% of the CCs were observed in the daytime, 36.1% of them lasted longer than 100 ms, but only 6.6%/0.6% lasted longer than 200/300 ms. Part of the CCs can be well described as an exponential decay, but there are also a number of more complicated waveforms with M-component signatures and prolonged, slightly fluctuating parts. Interestingly, in approximately 5% of the cases, the RS is preceded by some initial activity (current flow) lasting longer than 10 ms. Next, we plan to use WWLLN/ENTLN and MTG data to automatically identify the source lightning discharge of the detected events, as well as to employ machine learning techniques to make the CC detection more effective. We expect that our method will enable us to study lightning CCs in a much larger data set than ever before.