Density vs. sensitivity

The global infrasound network that watches for nuclear tests — the IMS — is a marvel of sensitivity: about 60 stations, spaced ~2000 km apart, can hear a bolide on the far side of the planet. But a network built to hear the whole Earth is, by design, deaf to the neighbourhood. HERD makes the opposite bet: not fewer, more sensitive ears, but a hundred times more of them — cheap and close together.

A network built for the planet, not the parish

The International Monitoring System (IMS) is one of humanity's finest listening machines: about 60 infrasound stations, spread roughly evenly across the globe to detect an atmospheric nuclear explosion anywhere on Earth1. Its detection capability is carefully modelled and depends on how far apart the stations sit and on the state of the atmosphere34, and its data now feed civilian science far beyond treaty verification2. It heard the 2013 Chelyabinsk meteor and the 2022 Hunga Tonga eruption on the far side of the planet13.

But a network designed to hear the whole Earth is, by construction, sparse. Between two stations ~2000 km apart, a small eruption, a debris flow, or a weak coastal signal can rise and fade completely unheard. Sensitivity at global range and awareness of the neighbourhood are different problems.

Cheap ears are finally good enough

For decades, 'real' infrasound meant expensive instruments. The last fifteen years changed that. Low-cost loggers like the Gem5 and the infraBSU sensor6, the mobile INFRA-EAR platform7, and robust, well-calibrated low-cost designs8 now deliver usable data — and independent laboratories have measured exactly how good the cheap packages are9. Low-cost small-aperture arrays already improve monitoring in the field, for example in the Azores10. The physics of catching a pressure wave hasn't changed; the price per node has collapsed.

Density buys what sensitivity cannot

Three things appear only when sensors are close together. First, localization: you find where an event is and how fast a front moves by comparing arrival times across many nearby sensors — the classic PMCC array method11 — so more, tighter-spaced ears mean sharper answers. Second, local events: debris flows, avalanches and small eruptions radiate signals that fade within tens of kilometres and never reach a distant station12. Third, coverage of data-poor regions that the sparse global grid simply skips.

The proof that numbers win

This isn't a hunch. In 2025, Google turned millions of ordinary Android phones into the largest earthquake-detection system on Earth14 — the exact logic of taking not accuracy but the sheer number of cheap ears. Crowdsourced Raspberry Shake & Boom observations measurably expanded the monitoring record of the 2022 Hunga Tonga eruption15. Citizen seismo-acoustic sensors16 and inexpensive MEMS barometers17 are already in millions of hands. HERD's bet is to organize them.