Climatology of the Harmonic Frequency Separation of Ionospheric Alfvén Resonances at Eskdalemuir Observatory, UK

Abstract

We extracted the harmonic frequency separation (Δf) of Ionospheric Alfvén Resonances (IAR) observed in the Eskdalemuir induction coil magnetometer data for the 9 year data set of 2013–2021. To obtain Δf values, we used a machine learning technique that identifies the harmonics and from this we calculated the average separation. To investigate the climatology of the IAR, we have modeled the Δf of the IAR for the data set using a time of flight calculation with model Alfvén velocity profiles. When analyzing Δf from the model and data, we found that in general they follow the same trends. The modeled Δf and Δf from the data both show an inverse correlation with foF2, which confirms that the frequencies of the IAR are controlled by electron density. It follows that Δf is greater around midnight and during the winter months, due to the decrease in plasma mass density. Variability is also reflected when comparing yearly trends in Δf with the sunspot number; higher frequencies are observed and modeled at low sunspot number. It is difficult to examine trends with instantaneous geomagnetic activity as IAR are not visible in spectrograms when geomagnetic activity is high. We find cases where the difference in measured and modeled Δf is significant, suggesting that the model does not capture short term variations in plasma mass density that influence the IAR during these days. We plan to undertake further modeling of Δf on shorter timescales.

Modeling and Observations of the Effects of the Alfvén Velocity Profile on the Ionospheric Alfvén Resonator

“…This demonstrates that the time-of-flight calculation of Δf accurately describes the separation of higher harmonics, but not the frequency of the fundamental, resulting in an offset of the harmonics and Δf, which was noted by Hodnett et al (2023). This offset is not constant, which is a result of the non-uniformity of the ionosphere, which we next analyze.…”

“…Overall, we find 12,488 time bins in the data where five harmonics have been identified by the U-net, across the 9-year data set. This is fewer than the number of points used in the study from Hodnett et al (2023), as previously we used a less constrained approach and found times where at least four harmonics were identified to obtain Δf, rather than five. These instances where at least five harmonics were identified occurred over 644 separate days.…”

“…The harmonics are identified by a neural network called a U-net (Ronneberger et al, 2015). We used the U-net developed by Marangio et al (2020) and the method to extract frequencies is detailed in Hodnett et al (2023). The input consists of grayscale spectrograms scaled to 256 × 256 pixels, where the time range is limited to "nighttime" periods (16:00-08:00 UT) and the frequency range is fixed between 0.5 and 6.5 Hz.…”

“…Alfvén velocity profiles can be modeled for the IAR at Eskdalemuir using the IRI (Bilitza et al, 2017) to obtain plasma mass density, and the IGRF (Alken et al, 2021) to obtain magnetic field strength. This is the same method used by Hodnett et al (2023), where such Alfvén velocity profiles were employed to model Δf using a time-of-flight approach and compared with observations. Whilst capturing the general behavior of the frequency of the IAR, such an approach cannot provide information on any variability of the individual frequencies of the harmonics which may result from the non-uniformity in V A .…”

“…As such, we have modeled the IAR harmonics (fundamental, second, third, fourth, and fifth) of the electric field at Eskdalemuir by solving a one-dimensional wave equation using the same non-uniform Alfvén velocity profiles as Hodnett et al (2023). Previous modeling work has shown that the vertical structure of the ionosphere is the most important consideration for the IAR (Lysak et al, 2013).…”