Evening. I want to say upfront that I've spent twenty-two years in seismology, the last eleven specifically in ambient noise analysis, and I've never called a program like this before. I'm calling because I genuinely don't know what else to do with what I found. I work at a monitoring station in eastern Nevada. We're about 340 kilometers from the nearest urban center, and that distance isn't incidental. That's why the station was placed there. Anthropogenic noise, the signal generated by humans just existing, traffic, industry, the low-frequency bleed from millions of machines operating simultaneously, it contaminates everything when you're too close. You need that separation to hear the planet without us in the way. My field is background seismic noise. Not earthquakes. Most people think seismology is earthquakes, and it is, but there's a whole parallel discipline built around what the Earth does when nothing dramatic is happening. When you run a seismometer and there are no events, no large storms, no nearby human activity, you still get signal. The Earth hums. Continuous free oscillations propagating through the mantle and crust. The dominant period clusters between 2 and 7 millihertz. We call it the Earth's hum. It's real, globally measured, and the source mechanisms are still not fully understood. That background. That quiet, apparently structureless signal. That's what I've spent the better part of a decade studying. The weekend this started was the first Saturday of November 2019. My colleague Dr. Chen was presenting at a conference in Portland. So it was just me and the instruments. I'd planned to run routine quality checks on the long-period channels and get through some backlogged processing. I brought a thermos of coffee and a library book I never ended up touching. Before I got to any of that, I pulled up the cross-spectral output from the previous eight weeks and something in it made me stop.
Cross-spectral analysis is a way of examining how different frequency components in the noise relate to each other across time. Normally the results confirm what you expect: ocean-generated microseisms dominating the 5 to 10 second band, atmospheric loading in the longer periods, and then the Earth's hum underneath all of it, essentially flat in spectral density, no particular organization. What I saw instead was a component at approximately 0.031 hertz that had no business being coherent. That's a period of about 32 seconds. Right at the edge of what we call the microseismic notch, a frequency range where the noise floor is naturally low and source attribution is genuinely ambiguous. The component was narrow-band. Very narrow. Much narrower than any known mechanism would produce. And it held coherence across the full eight-week window, which ruled out any transient artifact. It had been there, continuous, the entire time. My first read was that it was instrumental. A resonance in the sensor housing, or electronic feedback. I've seen those before. I pulled up the station health logs and went through the calibration records. Everything was nominal. Vault temperature stable within a quarter degree. No power fluctuations, no sensor drift. The instrument was fine. My second read was interference from a human source operating continuously nearby. So I went through our regional anthropogenic inventory. The 32-second period doesn't match industrial sources. It doesn't match power grid harmonics, traffic signatures, pumping operations, none of it. I went through the full list. I remember sitting back and just looking at the spectral plot. The peak was clean. Symmetric. It had the shape you only get from something genuinely periodic, not approximate. Whatever this was, it was precise in a way that bothered me.
The obvious move was to check the network. We share real-time data with roughly forty other stations globally through the consortium, and I pulled the previous eight weeks from six of them. Two in western Europe, one in southern Australia, one in Japan, one in South Africa, one in Alaska. Thousands of kilometers apart, different geological settings, different instrument types, completely different local noise environments. All six showed the same peak at 0.031 hertz. I want to explain why that matters, because it's the thing that changed everything for me. For a signal to appear coherently at the same frequency across stations on opposite sides of the planet, it has to be global. It has to propagate through the deep Earth, through the mantle or the core, not through the shallow crust where local sources would dominate. The only phenomena that do that are major earthquakes, or the Earth's known free oscillation modes, or something unidentified. This didn't fit the free oscillation spectrum. Those modes are well characterized. 0.031 hertz isn't one of them. And it was too narrow-band anyway. Free oscillations decay. They ring down over days after large earthquakes. This was continuous. I went back into the archive data. Our station has clean records back to the early 1990s, and the consortium has digitized older records from some stations going back to the late 1970s. I spent most of that afternoon pulling and reprocessing it. The signal was in all of it. Every record I checked, back to the oldest data available. Always present. Always at that exact 32-second period. Not constant amplitude, there were variations, but the frequency never drifted. Not by a single bin. Then I looked at the amplitude history across the full timeline, how the signal strength had changed from the 1970s to 2019. And that's when I started to feel like I was looking at something I maybe wasn't supposed to be looking at.
[ Story continues in the full game... ]