The research, carried out by Danielle G. Markowski and Stephen R. Kaeppler of Clemson University, repurposed Coastal Ocean Dynamics Applications Radars, or CODARs, as ionospheric sounders. These radars normally sit on shorelines and measure surface currents by reflecting high-frequency radio waves off the sea. But their frequency-modulated continuous wave transmissions, operating between roughly 4 and 5 megahertz, also travel upward and bounce off the ionosphere. By treating each CODAR transmitter as one half of a bistatic oblique sounder — with the receiver located at the Clemson Atmospheric Research Laboratory in South Carolina — the team could extract the group delay of each echo and convert it into a virtual height for the reflecting layer, sampled every two minutes nearly continuously from 2020 through 2023.

Twelve transmitters spread from Texas to the Northeast were grouped into four geographic regions, with ground distances from the receiver ranging from about 590 to nearly 1,600 kilometers. Because the fixed transmission frequencies suffer strong daytime absorption in the D-region of the ionosphere, the analysis was restricted to nighttime hours, specifically 0100 to 0900 Universal Time, which corresponds to roughly 8 p.m. to 4 a.m. local time. For each night, the researchers subtracted a reconstructed 12-hour-period waveform to remove diurnal trends, applied a Gaussian filter, and then ran a Lomb-Scargle periodogram analysis to identify any remaining dominant oscillation periods, flagging only those with false alarm probabilities below 0.1 percent — meaning the chance that the detected peak was mere noise was vanishingly small.

The result was striking. Across the September-to-December windows of four consecutive years, the single most common dominant period in the virtual height data was not the familiar 30-minute to 3-hour range associated with traveling ionospheric disturbances, but the much longer 4-to-5-hour band. In total, 535 of 1,348 usable nights — an average occurrence rate of 39 percent — showed this oscillation, and it appeared in every region and every year of the study. A secondary peak at 2 to 3 hours, which fits the standard definition of a large-scale traveling ionospheric disturbance, or LSTID, showed up on 630 nights, a 48 percent occurrence rate. In other words, the newly identified 4-to-5-hour oscillation happens nearly as often as the classic LSTID signature, yet it sits in a period range that falls between the established categories: longer than the sub-2-hour disturbances usually classified as TIDs, but not a clean subharmonic of the solar day like the 24-, 12-, 8-, and 6-hour oscillations typically attributed to atmospheric tides.

To understand what these oscillations were doing, the team went beyond single-station detection. Using a cross-correlation technique developed in an earlier study, they reconstructed the waveforms from the dominant period and measured the time delays between stations. A day qualified for further analysis only if at least three transmitters had data, the oscillation period fell in the 4-to-5-hour band, and the correlation coefficients between traces exceeded 0.5. If all traces showed a time delay of zero minutes, the structure was classified as non-propagating — a standing oscillation affecting a wide area simultaneously. If any trace showed a nonzero delay, the disturbance was propagating, and the geometry of the delays could be used to compute its horizontal phase speed and direction of travel.

The attrition was severe. Out of more than a thousand flagged nights, only 29 had the 4-to-5-hour oscillation appearing simultaneously and coherently at three or more stations. Of those, 9 were propagating disturbances and 20 were non-propagating. The propagating cases told a clear story. Every single southward-propagating event followed a geomagnetically active night, with a nightly-averaged Kp index of 3 or higher and maximum auroral electrojet indices exceeding 750 nanotesla in the ten hours before the disturbance. Their phase speeds fell between 300 and 800 meters per second — squarely within the accepted range for LSTIDs — and their average wavelength was approximately 5,000 kilometers, consistent with disturbances driven by energy injected from the auroral oval during geomagnetic storms. By contrast, every westward or northwestward propagating case occurred during geomagnetically quiet conditions, and in all but one of those cases, eastward-moving thunderstorm systems covered the entire observation region, suggesting that tropospheric weather may have launched atmospheric gravity waves that seeded the ionospheric oscillations.

The 20 non-propagating cases pointed to a different and equally intriguing set of drivers. All of them occurred during geomagnetically quiet nights, with Kp values at or below 3.5 and auroral electrojet indices below 250 nanotesla. When the researchers compared the CODAR-derived virtual heights with automatically scaled ionosonde measurements of foF2, the critical frequency of the F2 layer, and hmF2, the height of maximum electron density, they found that the oscillations fell into recognizable patterns. In some cases, the electron density declined through the evening and then dropped sharply near local midnight, with the virtual height rising in inverse proportion — a signature consistent with the midnight collapse, a well-known midlatitude phenomenon in which the F-layer height suddenly decreases around local midnight as meridional winds reverse and plasma is redistributed. Recent statistical work at Arecibo has shown that the midnight collapse occurs in roughly 94 percent of nighttime measurements there, underscoring how common it is.

In other cases, the pattern matched the post-midnight enhancement, an increase in electron density around local midnight that appears in the CODAR data as a corresponding dip in virtual height. This phenomenon is thought to arise when downward plasma flux along geomagnetic field lines from higher altitudes compensates for, and occasionally exceeds, the recombination losses that normally deplete the nighttime ionosphere; electric fields and equatorward neutral winds may also contribute. Notably, in none of the 20 non-propagating cases did both foF2 and hmF2 remain constant through the analysis window — either the density, the height, or both were changing, which is exactly what the CODAR virtual height oscillations were recording. Taken together, over half of the non-propagating cases showed strong similarities to these two established midlatitude phenomena, suggesting that what looks like a mysterious wave may in part be the familiar breathing of the nighttime ionosphere as it collapses and recovers.

The study also revealed something important about the spatial coherence of the ionosphere itself. The correlation length of the 4-to-5-hour oscillations across the southeastern United States was generally smaller than the 500-to-1,600-kilometer scales previously reported for midlatitude winter conditions, meaning that coherent structure spanning thousands of kilometers is the exception rather than the rule. That rarity is itself informative: it aligns with the expectation that only disturbances with very large scale sizes — such as auroral-driven LSTIDs, whose curvature follows the auroral oval — can remain correlated across continental distances. The researchers also acknowledged limitations: the fixed CODAR frequencies restrict observations to nighttime, the strict three-station requirement eliminated many usable days, and the absence of Texas data on some nights made it difficult to distinguish westward from southward propagation unambiguously.

For the broader community, the findings carry a practical warning and an invitation. On the practical side, ionospheric oscillations of this magnitude and frequency — appearing on nearly 40 percent of fall nights — can shift the reflection heights of high-frequency radio waves in ways that matter for over-the-horizon radar, shortwave broadcasting, and amateur radio propagation forecasting. On the scientific side, the authors encourage future studies to stop filtering out oscillations longer than 3 hours, since doing so may have hidden an entire class of ionospheric variability in plain sight. Whether the 4-to-5-hour pulse ultimately proves to be a tide, a terminator-driven wave, a weather-launched gravity wave, or the combined fingerprint of the midnight collapse and post-midnight enhancement, it is now a documented, repeatable feature of the eastern United States nighttime ionosphere — detected, fittingly, by radars that were never meant to look up at all.

Subject of Research: Observations of 4–5 h oscillation periods in virtual height during the fall seasons of 2020–2023 across the eastern United States

Article Title: Observations of 4–5 h oscillation periods in virtual height during the fall seasons of 2020–2023 across the eastern United States

Article References: Markowski, D. G., & Kaeppler, S. R. (2026). Observations of 4–5 h oscillation periods in virtual height during the fall seasons of 2020–2023 across the eastern United States. Frontiers in Astronomy and Space Sciences, 13, Article 1886813. https://doi.org/10.3389/fspas.2026.1886813

Image Credits: AI Generated

DOI: 10.3389/fspas.2026.1886813

Keywords: Observations, oscillation, periods, virtual, height, during, fall, seasons, across, eastern, United, States