MyShake应用:将你的手机变成地震仪

MyShake: Turning Smartphones into Seismometers

The MyShake application, developed at the University of California Berkeley Seismological Laboratory, demonstrates a novel approach to seismic monitoring: harnessing the collective sensing power of millions of smartphones to supplement — and in data-sparse regions potentially replace — traditional seismic station networks. Understanding how MyShake works reveals both the ingenuity of the approach and the engineering challenges involved in deploying distributed seismic sensing at global scale.

The Accelerometer at the Core

Every modern smartphone contains a micro-electromechanical systems (MEMS) 加速度计测量地面运动加速度的传感器，对地震工程至关重要。现代强震加速度计能够记录大地震附近的剧烈震动。 — a tiny chip-scale device that measures acceleration in three orthogonal axes. These accelerometers were originally designed for screen orientation and step counting, but they also respond to ground shaking. The sensitivity of modern MEMS accelerometers is sufficient to detect moderate earthquakes (M 4.0+) at distances of tens of kilometers, though they lack the sensitivity of purpose-built 地震仪用于探测并记录地震波引起的地面运动的仪器。现代数字地震仪可探测到小于一纳米的位移。 instruments at small magnitudes.

How MEMS Differs from Traditional Sensors

A traditional broadband 地震仪用于探测并记录地震波引起的地面运动的仪器。现代数字地震仪可探测到小于一纳米的位移。 uses a hanging mass suspended by springs in an evacuated vault, capable of resolving ground motions as small as picometers. MEMS accelerometers have noise floors millions of times higher, limiting their effective detection threshold. However, what MEMS devices lack in individual sensitivity they compensate for in numbers — millions of devices creating a globally distributed 地震观测网由若干地震台站协同组成、持续监测地震活动的系统。全球地震台网（GSN）拥有150多个台站，提供全球范围的观测覆盖。 at essentially zero marginal cost per sensor.

The On-Device Detection Algorithm

MyShake runs a lightweight machine learning classifier continuously on the phone's processor, analyzing three-axis 加速度计测量地面运动加速度的传感器，对地震工程至关重要。现代强震加速度计能够记录大地震附近的剧烈震动。 data in real time. The classifier was trained on thousands of waveform examples to distinguish seismic shaking from human activities — walking, driving, dropping the phone — which can produce accelerations far larger than earthquake ground motion. When the on-device classifier detects a candidate seismic signal, it uploads a summary to the MyShake server without sending the complete raw waveform, conserving battery and data bandwidth.

Network-Level Earthquake Confirmation

A single phone detecting a candidate signal is insufficient for reliable event confirmation — the on-device classifier has a non-trivial false alarm rate because everyday human activities can mimic seismic waveforms. The MyShake back-end server applies a network-level algorithm that looks for spatially and temporally correlated detections across multiple devices. When dozens of phones within a geographic cluster report candidate signals within seconds of each other, the network classifier confirms an earthquake and estimates the 震中地震发生在地下的震源正上方对应的地表位置，新闻报道中通常将其作为地震发生的位置。 location and 震级量化地震所释放总能量的单一数值。震级每增加一个整数单位，释放的能量约增加31.6倍。. This two-tier approach dramatically reduces false alarm rates.

Contribution to 地震预警（EEW）一种在强震到达前探测地震并向人员和系统发送警报的系统，可提供数秒至数十秒的预警时间，足以采取自我保护行动。 Systems

The MyShake network has been integrated into California's ShakeAlert预警系统由美国地质调查局及多所大学合作运营的美国地震预警系统，覆盖美国西海岸（加利福尼亚州、俄勒冈州、华盛顿州），通过无线紧急警报发送提醒。 early warning system as a supplementary data source. In earthquake-prone regions with dense MyShake installations, crowd-sourced phone detections can provide an additional layer of ground truth that reduces the time to event confirmation. In developing countries where traditional 地震观测网由若干地震台站协同组成、持续监测地震活动的系统。全球地震台网（GSN）拥有150多个台站，提供全球范围的观测覆盖。 infrastructure is sparse — much of South Asia, Central America, and sub-Saharan Africa — MyShake-style networks represent a low-cost pathway to basic early warning capability.

Citizen Science and Research Data

Beyond early warning, MyShake accumulates a massive dataset of seismic recordings from around the world. Researchers use this archive to study 地震群在数天至数月内发生于局部区域、且无明显主导主震的一系列地震，常与火山活动或流体注入相关。 sequences, calibrate attenuation models in regions lacking traditional stations, and test new magnitude estimation algorithms. The app also includes a citizen science component inviting users to contribute felt intensity reports, analogous to the USGS “你感觉到了吗？”（DYFI）美国地质调查局的一项计划，收集震后公众的烈度报告，形成基于公众参与的烈度分布图，任何感受到地震的人都可以提交报告。 program, creating a parallel stream of macroseismic data.

Battery and Privacy Considerations

Running a continuous sensor application raises legitimate concerns about battery consumption and privacy. MyShake is designed to minimize battery impact by activating the accelerometer only when the phone is stationary — movement detected by the gyroscope pauses seismic monitoring since a moving phone cannot distinguish seismic ground motion from user activity. Location data used for event confirmation is anonymized and aggregated on the server; individual device locations are not logged or shared. The app transmits only compact detection reports, not continuous waveform streams.

Expanding the Global 地震观测网由若干地震台站协同组成、持续监测地震活动的系统。全球地震台网（GSN）拥有150多个台站，提供全球范围的观测覆盖。

Traditional 地震观测网由若干地震台站协同组成、持续监测地震活动的系统。全球地震台网（GSN）拥有150多个台站，提供全球范围的观测覆盖。 expansion is constrained by the cost of instruments (tens of thousands of dollars each), installation logistics, power supply requirements, and ongoing maintenance. A single MyShake user's smartphone provides sensing capability in a location that might otherwise be unmonitored. In subduction zone countries like Indonesia, the Philippines, and Chile, where 地震仪用于探测并记录地震波引起的地面运动的仪器。现代数字地震仪可探测到小于一纳米的位移。 station density is insufficient for optimal 地震预警（EEW）一种在强震到达前探测地震并向人员和系统发送警报的系统，可提供数秒至数十秒的预警时间，足以采取自我保护行动。 system performance, MyShake-class crowdsourced networks fill critical gaps.

Limits of Smartphone Seismology

Smartphone seismology faces fundamental constraints that prevent direct equivalence with professional instrumentation. The dynamic range of MEMS accelerometers is insufficient to record both the weak early P-wave phase and the strong S-wave shaking in the same waveform without clipping. Phones placed on soft furnishings decouple from ground motion, reducing sensitivity. Phones in pockets or bags show highly variable coupling. The heterogeneity of deployment environments makes precise calibration impossible, limiting the precision of magnitude estimates derived from smartphone data alone. These limitations mean that MyShake is best understood as a force multiplier for existing networks, not a replacement.

Summary

MyShake demonstrates that the 加速度计测量地面运动加速度的传感器，对地震工程至关重要。现代强震加速度计能够记录大地震附近的剧烈震动。 chips embedded in billions of smartphones constitute a latent seismic network of unprecedented geographic reach. By combining on-device machine learning for preliminary detection with network-level spatiotemporal correlation for confirmation, the system achieves reliable earthquake detection and 地震预警（EEW）一种在强震到达前探测地震并向人员和系统发送警报的系统，可提供数秒至数十秒的预警时间，足以采取自我保护行动。 capability at near-zero infrastructure cost. As smartphone penetration increases globally, crowd-sourced 地震观测网由若干地震台站协同组成、持续监测地震活动的系统。全球地震台网（GSN）拥有150多个台站，提供全球范围的观测覆盖。 density will continue to grow in regions where traditional monitoring remains sparse.