Most of what we know about Earth’s interior comes from observing seismic waves, which travel at different speeds depending on whether rock is solid, semi-molten, waterlogged, fractured, or just having a bad day. Gather enough data from enough seismic events, and you can piece together a picture of what’s lurking beneath our feet. Earthquakes provide natural data; explosives provide intentional data. But a team at Penn State suggests a middle ground: thunderstorms. Yes, those noisy sky tantrums can cause 'thunderquakes' - seismic signals that are, for various physical reasons, a complete mess. The team says they’ve finally built a model that makes sense of this chaos, and used it to reconstruct the terrain under their own campus.

Why are thunderquakes so hideously complex? It starts with lightning, which creates thunder by forming superheated plasma bubbles along its path - like a string of beads. Each bead can generate an acoustic shock wave, leading to a chain of expanding waves that trace the lightning’s erratic path. These waves interfere with each other, and when they hit the ground, they don’t just stop - they hit soft soil, hard rock, and human infrastructure, each affecting energy transmission. Some energy becomes Rayleigh waves that skitter along the surface; the rest plunge deeper. To extract useful info, you need to model what a thunderclap’s seismic waves should look like - which means modeling all that chaos. Since every thunderquake is unique, any model is an approximation. The fear that approximations wouldn’t be good enough probably kept scientists away for years.

The team used SPECFEM3D Cartesian, a 3D seismic wave reconstruction software, which forced compromises. The atmosphere is treated as a 3.6 km-thick homogeneous layer (which it definitely isn’t near a storm), and updates run slower than waves at the Earth-air interface - so they stretched the top 20 meters of Earth to cover 200 meters. Plenty of reasons to think the model would fail - so they tested it against real thunderstorms passing over campus.

Here’s the fun part: fiber-optic cables can act as seismometers, and Penn State has a 4 kilometer fiber line dedicated to seismic sensing running under campus, conveniently located in a region with regular summer thunderstorms. Two years of data yielded 458 well-resolved thunderquakes, each confirmed using records from the US’s National Lightning Detection Network (a thing that apparently exists). The quakes showed multiple signals arriving from different altitudes, consistent with that string-of-beads structure. Upon hitting the ground, each bubble generated 'a high-energy impulsive wavelet followed by a decaying wave train dominated by surface-wave content lasting one to two seconds.'

Using this data, the team identified four 'weak zones' where seismic signals slow down - areas of sediments, fractured rock, or high water content. Penn State sits on a karst formation, where water has slowly altered limestone bedrock, potentially creating weak spots. They confirmed these four sites have something unusual using radar, engineering surveys, boreholes, and independent seismic data.

Despite all the approximations, the model produces accurate reconstructions, and thunderquakes offer advantages: they’re relatively frequent, and they’re best for imaging near-surface zones where our infrastructure lives. So this is a case where the model is wrong, but also useful. Science Advances, 2026. DOI: 10.1126/sciadv.aeg8096.