アプリで地震を予測できるか?

The Myth: Smartphone Apps Can Predict Earthquakes

The app economy has transformed how people interact with every domain of life, and earthquake awareness is no exception. A proliferating ecosystem of apps claims to warn users of incoming earthquakes, predict quakes from unusual patterns, or crowdsource early detection. Some apps claim to detect "electromagnetic precursors" or "ionospheric disturbances" that allegedly precede earthquakes. Others promise to combine your phone's accelerometer data with crowd signals to provide minutes of warning. The promises are appealing, especially in an era when technology seems to make the impossible routine. But earthquake prediction remains beyond current science, and most apps claiming predictive capability are not what they seem.

The Fundamental Distinction: Prediction vs. Early Warning

The most important concept for evaluating earthquake apps is the distinction between 地震予知と地震予測の違い予知は将来の地震の正確な時刻・場所・規模を特定することを指すが、これは現在のところ不可能である。予測は、ある期間における地震発生の可能性を確率的に見積もるものである。 and early warning. These are fundamentally different capabilities.

Earthquake prediction — knowing in advance that a specific magnitude earthquake will occur at a specific location within a specific time window — is currently impossible with any reliable accuracy. No app, algorithm, or technology can do this, despite many claims to the contrary.

Earthquake 緊急地震速報（EEW）地震を検知し、強い揺れが到達する前に人々やシステムに警報を送るシステム。数秒から数十秒の猶予を提供し、身を守る行動をとるのに十分な時間となる。 is something entirely different and genuinely real. It detects P-waves from an earthquake that has already started, calculates the likely shaking intensity at distant locations, and sends alerts that arrive before the damaging S-waves and surface waves. The window is seconds to a few tens of seconds — enough to take cover, stop trains, pause surgeries, open firehouse doors. This is real-time physics, not prediction.

Legitimate Apps: [[ShakeAlert]] and Government-Backed Systems

[[ShakeAlert]] is the USGS-developed early warning system for the western United States, operational in California, Oregon, and Washington. It uses a network of hundreds of seismic sensors to detect P-wave arrivals, rapidly estimate earthquake parameters, and issue warnings to subscribers. The My Shake app (University of California Berkeley) and the FEMA/USGS ShakeAlert Earthquake Early Warning apps deliver these legitimate alerts to smartphones.

These apps genuinely work within their physical limitations: they provide real-time warning of ongoing earthquakes, with lead times proportional to how far you are from the epicenter. They cannot warn about future earthquakes they haven't detected yet. This is analogous to how a smoke detector works — it detects fire that is already happening, it doesn't predict when you will next accidentally leave something on the stove.

Pseudoscientific Apps and Their Red Flags

A separate category of apps makes claims that go well beyond what physics and seismology support. Common red flags include:

Apps that claim to predict earthquakes days or weeks in advance based on electromagnetic signals, atmospheric ionization, or animal behavior reports. No peer-reviewed science supports these mechanisms as reliable precursors, and any pattern-finding from these inputs is almost certainly spurious (see confirmation bias discussion elsewhere).

Apps that aggregate crowdsourced "feeling dizzy" or "animals acting strange" reports and use machine learning to predict earthquakes. While machine learning is powerful, it cannot extract predictive signal from data that contains no reliable signal. Training a model on weak correlations produces confident-sounding predictions that fail in prospective testing.

Apps that exploit the novelty of smartphone magnetometers and accelerometers to claim detection of "pre-seismic electromagnetic pulses." While phones do have these sensors, they are designed for consumer use cases (compass, step counting) and are not calibrated seismometers. The noise floor of a smartphone accelerometer is far too high to detect genuine pre-seismic signals even if such signals existed reliably, which they don't.

The 加速度計地面の揺れの加速度を測定するセンサーで、耐震工学において重要な役割を果たす。現代の強震動加速度計は、大地震近傍の激しい揺れを記録できる。 Reality Check

Modern smartphones do contain 加速度計地面の揺れの加速度を測定するセンサーで、耐震工学において重要な役割を果たす。現代の強震動加速度計は、大地震近傍の激しい揺れを記録できる。 sensors sensitive enough to detect strong earthquake ground motion — a smartphone on a table during a M4.0 earthquake will record detectable vibration. This capability has been used legitimately for crowdsourced strong motion networks (MyShake's seismic detection mode, for example) that help characterize earthquake shaking intensity in real time for events that have already occurred.

But there is a vast gulf between detecting ground motion from an ongoing earthquake (legitimate and useful) and detecting any signal that predicts a future earthquake (currently impossible). The former is an engineering problem with a solution; the latter is a scientific frontier with no demonstrated breakthrough.

Why False Prediction Apps Are Harmful

Apps that claim earthquake prediction capability cause real harm beyond the obvious financial exploitation of scared users. They generate constant false alarms, which create panic and erode trust in legitimate warning systems. When a real earthquake warning arrives through シェイクアラート（ShakeAlert）USGSと大学パートナーが運用する、アメリカの緊急地震速報システム。西海岸（カリフォルニア州・オレゴン州・ワシントン州）をカバーし、緊急速報メールを通じて警報を送信する。 after users have received dozens of false alarms from prediction apps, they may dismiss the legitimate alert. The "boy who cried wolf" problem is not hypothetical — studies of warning system effectiveness consistently find that false alarm rates critically undermine public response.

They also provide false reassurance when predictions don't materialize — users may lower their actual 地震への備え家具の固定、連絡計画の作成、非常用物資の維持、訓練の実施など、地震の被害を最小限に抑えるための継続的な計画・準備の過程。 because they believe an app is "watching" for danger.

The Current State of Earthquake Forecasting Science

Legitimate earthquake 地震予知と地震予測の違い予知は将来の地震の正確な時刻・場所・規模を特定することを指すが、これは現在のところ不可能である。予測は、ある期間における地震発生の可能性を確率的に見積もるものである。 science focuses on probabilistic hazard assessment over long time periods, as discussed in the USGS hazard map program. Short-term operational forecasting — elevated probability windows of days to weeks following significant earthquakes — is possible because 余震本震と同じ断層領域で本震の後に続く、より規模の小さい地震。余震活動は数週間から数年続くことがあり、最大の余震は通常本震よりマグニチュードが1.0〜1.2小さい。 sequences are predictable in statistical terms using 大森公式（オモリ則）時間経過に伴う余震発生頻度の減衰を示す経験則で、余震の発生率は本震からの経過時間にほぼ反比例して減少する。 and related models. After a M6.0 earthquake, scientists can estimate that there is approximately a 5% chance of a larger aftershock in the following week. This is useful information for emergency managers, but it is not earthquake prediction in the folk sense, and no app has improved on these statistical models through any novel data source.

How to Use Earthquake Technology Wisely

Install government-backed 緊急地震速報（EEW）地震を検知し、強い揺れが到達する前に人々やシステムに警報を送るシステム。数秒から数十秒の猶予を提供し、身を守る行動をとるのに十分な時間となる。 apps appropriate to your region (ShakeAlert/MyShake for the US West Coast, PLUM/NHK for Japan, SASMEX for Mexico). Understand what they actually do — detect ongoing earthquakes, not predict future ones. Treat any app claiming prediction capability with strong skepticism and check whether its methodology has been peer-reviewed in reputable seismological journals. Use the Seismic Risk Checker for probabilistic hazard information based on real seismological data.