Automatic detection of ionospheric Alfvén resonances using signal and image processing techniques

Abstract

Abstract. Induction coils permit the measurement of small and very rapid changes of the magnetic field. A new set of induction coils in the UK (at L = 3.2) record magnetic field changes over an effective frequency range of 0.1-40 Hz, encompassing phenomena such as the Schumann resonances, magnetospheric pulsations and ionospheric Alfvén resonances (IARs). The IARs typically manifest themselves as a series of spectral resonance structures (SRSs) within the 1-10 Hz frequency range, usually appearing as fine bands or fringes in spectrogram plots and occurring almost daily during local night-time, disappearing during the daylight hours. The behaviour of the occurrence in frequency (f ) and the difference in frequency between fringes ( f ) varies throughout the year. In order to quantify the daily, seasonal and annual changes of the SRSs, we developed a new method based on signal and image processing techniques to identify the fringes and to quantify the values of f , f and other relevant parameters in the data set. The technique is relatively robust to noise though requires tuning of threshold parameters. We analyse 18 months of induction coil data to demonstrate the utility of the method.

Observation of Ionospheric Alfvén Resonances at 1–30 Hz and Their Superposition With the Schumann Resonances

“…We will continue to collect more induction coil measurements to confirm these findings. Future work will also examine correlation between f 0 F 2 and the Δ f variation in more detail across the 5 years, for example, using the image processing techniques from Beggan ().…”

“…In addition, the peak‐to‐peak difference between the IAR fringes also changes with season and is largest in winter at 0.5–0.7 Hz but smaller in summer and the equinoxes (cf. Beggan, ).…”

Automatic detection of Ionospheric Alfvén Resonances in magnetic spectrograms using U-net

“…Using the methodology of Beggan (2014), panels (b-d) show a prior attempt to automatically identify and delineate IARs based on signal and image processing techniques alone. This approach relies heavily on peak detection of the IARs as they rise above the general background level and the joining of these peaks using image dilation and erosion algorithms with fixed thresholds based on manual experimentation.…”

“…We note that no IARs are present in panel (d), though the algorithm does pick out the 1 Hz harmonics. At the time of their study, Beggan (2014) had around 14 months of magnetic data available, which was used to create statistics of the behavior and occurrence of IARs. However, the method was not wholly successful so further improvements were sought, particularly to remove the reliance on manually set thresholds.…”

“…Several methods based on signal and image processing techniques have been developed for the semi-automated detection (i.e. labelling) of IARs and computation of parameters of interest (Odzimek et al, 2006;Beggan, 2014). However, not only do these methods require the tuning of several thresholds and parameters, but they are also prone to noise detection regardless of whether IARs are present or not.…”