Seasonal Variation of the D‐Region Ionosphere: Very Low Frequency (VLF) and Machine Learning Models

Abstract

Previous machine learning model of D-region extended to cover many days and include nighttime • Daytime ionospheric height varies by 1.5-2 km from winter to summer, with ∼1 h of memory • Daily ionospheric D-region height flutters by 3-4 km within and across nights, with only minutes of memory Supporting Information:

A D‐Region Ionospheric Imaging Method Using Sferic‐Based Tomography

“…As these waves propagate, they collect information about the ionosphere along their propagation path, which has enabled researchers to infer electron density altitude profiles within the D region. These studies have typically relied on either VLF transmitters (McRae & Thomson, 2000; Richardson & Cohen, 2021; N. Thomson, 1993; N. R. Thomson & McRae, 2009; N. R. Thomson et al., 2007) or broadband emissions from lightning known as radio atmospherics, commonly shortened to sferics (Cheng et al., 2006; Cummer et al., 1998; Han et al., 2011; Lay et al., 2014; Maurya et al., 2012; J. C. McCormick & Cohen, 2021; Shao et al., 2012). Most of the prior work has focused on inferring a two‐parameter model for electron density, h’ and β from Equation , first introduced by Wait and Spies (1964) where h’ represents the height of the ionosphere and β represents the rate of change in electron density with altitude.…”

“…As these waves propagate, they collect information about the ionosphere along their propagation path, which has enabled researchers to infer electron density altitude profiles within the D region. These studies have typically relied on either VLF transmitters (McRae & Thomson, 2000;Richardson & Cohen, 2021;N. Thomson, 1993; N. R. Thomson & McRae, 2009; N. R. Thomson et al, 2007) or broadband emissions from lightning known as radio atmospherics, commonly shortened to sferics (Cheng et al, 2006;Cummer et al, 1998;Han et al, 2011;Lay et al, 2014;Maurya et al, 2012;J.…”

Unifying VLF Transmitter and Sferic Modeling Efforts Via Tomography

“…The steps in this process can be summarized as follows: The path average electron density profiles are computed for both the VLF‐transmitter‐to‐receiver and sferic‐to‐receiver paths according to McCormick et al. (2018) and Richardson and Cohen (2021), respectively. A system of equations is set up to compute discrete cosine transform (DCT) basis function coefficients representing electron density. Weights are computed to inform the model of how heavily to trust different paths' electron density profiles. The system of equations is solved, one altitude at a time, to produce 3D electron density maps. …”

“…On the VLF transmitter side, there have been many studies about various approaches to solving the problem, with differing limitations to each. In this work, a recently developed machine learning-based method will be used due to the flexibility and time resolution available (Richardson & Cohen, 2021). Although the entire details of this method are covered by Richardson and Cohen (2021), a short description is provided here.…”

“…In this work, a recently developed machine learning-based method will be used due to the flexibility and time resolution available (Richardson & Cohen, 2021). Although the entire details of this method are covered by Richardson and Cohen (2021), a short description is provided here. The method begins by training a simple, single-layer neural network using synthetically generated training data.…”