CAAD — Interactive Research Paper

Can two sensor networks locate an earthquake faster than one?

We deployed two independent seismic monitoring networks across Kenya's Rift Valley, injected a magnitude 4.2 earthquake into the simulation, and measured how comparing the two networks — against each other — reveals what neither could see alone. Here's what happened.

The Experiment
Two ways to listen to the same ground

We simulated 120 minutes of seismic activity across the Kenyan Rift Valley using a synthetic dual-network configuration. At minute 45, we injected a magnitude 4.2 earthquake near Longonot. An aftershock followed at minute 75. The question: how much more can you learn when two independent observers disagree?

Observer A — Primary Seismometer Network

High-precision broadband seismometers. Research-grade instruments with excellent sensitivity. But only 60% spatial coverage — sparse deployment across the rift. Measures what it sees precisely, but misses what it can't reach.

vs

Observer B — Accelerometer Network

Lower-cost MEMS accelerometers. Wider deployment with 90% coverage. Noisier measurements with higher instrument error. Sees more of the region, but each reading is less precise.

18 stations
120 min monitoring window
M4.2 earthquake at t=45
M2.5 aftershock at t=75
60% primary coverage
90% secondary coverage
7 Rift Valley stations
Explore the Data
Scrub through two hours. Watch the gap.

The gap between what the two networks report isn't random — it's information. When both agree, you have confidence. When they disagree, the shape of the disagreement tells you what's happening underground. Drag the timeline to see.

00:00 / 120 min
Background microseismic
Gap (RMSE)
--
between networks
Gap Trend
--
growing or shrinking?
Signal vs Noise
--
instrument calibration
Wave Propagation
--
stations with elevated readings
First Deviation
--
where the gap moved most
Magnitude Est.
--
from network convergence
Observer Gap Over Time
RMSE between primary seismometers and accelerometer network
Signal vs Noise Decomposition
Is the gap caused by instrument calibration (noise) or real ground motion (signal)?
Rate of Change
How fast is the gap growing? Spikes = seismic event onset.
The Result
What two observers give you that one can't

A single network detects the earthquake. Two networks understand it.

Any properly calibrated seismometer picks up a magnitude 4.2 event. That's table stakes. But a single network can't tell you whether an anomalous reading is instrument drift or real ground motion. It can't identify which areas of the rift have monitoring blind spots. And it can't track wave propagation by watching HOW the disagreement between networks moves across stations.

The gap between two independent observers is structured information

By comparing the primary seismometer network against the accelerometer network, we could separate instrument noise from real seismic signal (signal rose 18x during the main event), pinpoint the first station to show deviation (Longonot-N — closest to the injected epicentre), track wave propagation across 12 stations, and identify 40% of the rift where only one network had coverage — the monitoring blind spots that matter most during an event.

Capability Traditional (single network) Convergent (two observers)
Detects seismic event Yes (t=46 min) Yes (t=46 min)
Locates epicentre Roughly — triangulation only Yes — first deviation station + gap shape
Distinguishes instrument noise from ground motion No Yes — signal/noise decomposition
Tracks wave propagation across stations No Yes — propagation count over time
Identifies monitoring blind spots No Yes — where only one network has coverage
Separates main event from aftershock Yes — amplitude only Yes — plus propagation pattern differences
Works with imperfect, noisy sensors Assumes calibrated instruments Models and measures imperfection

The same method works beyond seismology

This isn't a seismic-specific tool. The same principle — measuring the gap between independent observers — applies to health systems (field data vs reported data), financial anomalies (price vs fundamentals), traffic networks (ground truth vs sensors), and any domain where you have two independent ways to measure the same thing. The gap is always structured information.

Want to apply this to your monitoring network?

CAAD is being built as a service. If you have multiple sensor networks, data sources, or measurement systems that should agree but sometimes don't — the gap between them is information you're not using.

Get in touch