A statistical approach to crowdsourced smartphone-based earthquake early warning systems

TL;DR

This paper presents a statistical method for detecting earthquakes using data from a dynamic, heterogeneous network of smartphones, aiming to improve early warning systems by controlling false alarms and analyzing detection performance.

Contribution

It introduces a novel statistical approach that manages network variability and sensor heterogeneity for earthquake detection in crowdsourced smartphone networks.

Findings

Effective detection of earthquakes in real-world smartphone networks.

Controlled false alarm probability in earthquake detection.

Analysis of detection delay relative to earthquake magnitude.

Abstract

The Earthquake Network research project implements a crowdsourced earthquake early warning system based on smartphones. Smartphones, which are made available by the global population, exploit the Internet connection to report a signal to a central server every time a vibration is detected by the on-board accelerometer sensor. This paper introduces a statistical approach for the detection of earthquakes from the data coming from the network of smartphones. The approach allows to handle a dynamic network in which the number of active nodes constantly changes and where nodes are heterogeneous in terms of sensor sensibility and transmission delay. Additionally, the approach allows to keep the probability of false alarm under control. The statistical approach is applied to the data collected by three subnetworks related to the cities of Santiago de Chile, Iquique (Chile) and Kathmandu (Nepal). The detection capabilities of the approach are discussed in terms of earthquake magnitude and detection delay. A simulation study is carried out in order to link the probability of detection and the detection delay to the behaviour of the network under an earthquake event.