A home seismometer for everyone. Precise. Open. Always on.

Individually, a single sensor has a noise floor around 70 µg/√Hz; the T1 stacks four of them into one array, bringing the effective floor down to roughly 35 µg/√Hz low enough to pick up shaking above M2.0 nearby and M4+ regionally. And accuracy is also a network property: when dozens of nodes confirm the same wave arrival times, we triangulate to scientific-grade precision.

Very little the sensor sends compact MQTT packets: periodic health metrics when idle and a brief burst of readings during an event. Typical usage is under 5 MB per day, about as much as a single email with an attachment.

You choose it yourself. During setup you place your station on the map, and we recommend picking a point 50 to 100 m away from your home rather than your exact address. Position accuracy matters for early warning, because wave arrival times are compared between stations, but an offset that small does not affect the calculation and keeps the public marker off your front door.

No. Nodes in stable regions are scientifically valuable as "quiet references" they help characterize network noise. And the network becomes stronger when coverage reaches beyond the obvious fault lines.

No we supplement them. Our data is fully open, so researchers, institutions and anyone else can use GeoShake feeds alongside official sources. We're a complementary layer, not a replacement.

Median end-to-end latency from detection to your phone is about 2.3 seconds in well-covered regions. That usually translates to a 5–20 second head start depending on your distance from the epicenter.

For P-wave early warning, network density matters more than individual sensor sensitivity. Instead of a single $10,000 geophone, a dense swarm of $20-class MEMS sensors covers far more ground for the same budget the only sustainable way to locate an epicenter faster. Raw precision is a network property once you have enough nodes.

Partly, and we'd rather be precise than sound impressive. The firmware is open source under GPL-3.0 including the signal-processing reference it is bit-locked to, so anyone can audit our detection maths. The enclosure designs are open under CC BY-SA 4.0, and our seismic data is openly available. The PCB design and the cloud backend are proprietary, so the device as a whole is not open source hardware. Everything we call open is on GitHub, sources and all.

Raspberry Shake is a scientific archive-focused station closed-ish hardware/software, recording full waveforms at high fidelity. GeoShake is designed for early warning: open source, low cost, P-wave first. We correlate across the whole network 500+ sensors and growing and send lightweight MQTT event messages rather than heavy continuous waveforms. The two are complementary Raspberry Shake is one high-accuracy station; GeoShake is a wide, fast, cheap swarm. They don't replace each other.

This is the project's hardest problem cheap sensors alone are noisy. Our answer is a correlation algorithm. If only your device is shaking, it's noise. If ten different devices within a 5 km radius report similar signals within milliseconds, it's a seismic wave. Consensus across the mesh is what separates a passing truck from a tremor.

Mount it on a fixed, rigid surface a load-bearing column or beam is ideal. A device on a wobbly table or soft carpet is useless. An early warning system also has to run 24/7: keep it plugged in and keep the Wi-Fi connected. The network is only as healthy as its nodes.

Traditional seismological networks (SeedLink) stream full waveforms for archival and are heavy. GeoShake is speed-focused. With MQTT we send only the "event detected" trigger plus summary metrics (PGA, RMS). In scenarios where a thousandth of a second matters for P-wave detection, a small header wins.

False positives are a real risk which is why GeoShake is an experimental community project and does not replace the official earthquake data published by national authorities. To minimize false alarms our algorithms use high thresholds and require multi-device consensus before anyone's phone buzzes.

Very little. A GeoShake node draws roughly 0.7–1 W on average far less than a modem or an LED bulb. Running 24/7 for a month, the impact on your electricity bill is negligible.