Everyday · California Life · Central Valley
Your earthquake alert is not predicting an earthquake. It is telling you one has already started.
ShakeAlert detects an earthquake that is already underway and races the message ahead of the shaking. Warning time depends on distance from the source, which is why the people closest to it get the least — and sometimes none.
The phone goes off and the floor is still. For a second it feels like the device knows something that has not happened yet.
It does not. The earthquake has already begun — somewhere else. What the alert is doing is racing a message against the shaking, and the message travels faster.
That single fact explains everything else about the system: why the warning is measured in seconds, why the people nearest the epicentre get the least of it, and why an alert that arrives during the shaking is not evidence of failure.
What is actually happening in those seconds
Warnings delivered in California come from ShakeAlert, a system operated by the Governor’s Office of Emergency Services and the U.S. Geological Survey. It analyses data from seismic networks, calculates a preliminary magnitude, and estimates which areas will feel shaking.
Sensors near the rupture register the earliest ground motion. That data reaches a processing centre, algorithms estimate strength and expected shaking, and an alert is pushed out electronically — to phones, and to the machinery that can act on it without a human.
The system is exploiting one simple asymmetry: an electronic signal outruns seismic waves through rock. There is no forecasting involved. USGS is explicit that ShakeAlert is not earthquake prediction.
rupture begins
→ nearby sensors register it
→ algorithms estimate magnitude and affected area
→ alert transmitted electronically
→ stronger shaking arrives where you are
Distance is the whole budget
Because the warning is a race, the amount of time it can buy is set by how far you are from the source. Someone far away may get useful seconds. Someone close may get almost nothing, and may feel shaking before the phone reacts at all.
Seismologists call this the late-alert zone, and it is a property of physics rather than a defect in the software. Cal OES makes the same point in plainer terms: the primary reason people near an epicentre do not get the alert before the shaking is proximity.
So no alert before shaking does not mean the system failed. It usually means you were close to it.
Not every earthquake sets off your phone, and that is deliberate
Different channels carry different thresholds, and mixing them up is the source of most complaints that “the app didn’t work.”
The thresholds are a design choice with an obvious rationale. A system that shouted at everyone for every detectable tremor would train an entire state to ignore it, and the one alert that mattered would arrive to a population that had already learned to swipe it away.
Three channels, and you may already have one
Californians reach this system through more than one route, and the routes are not identical.
- Android handsets. In partnership with Google, California’s earthquake early warning is built into millions of Android phones used in the state without a separate app, using the same data feed as the state system.
- MyShake. The state’s app, developed at UC Berkeley and funded by Cal OES. It uses the phone’s general location so that a warning can reach handsets in the affected area; Cal OES advises setting location to always on, and alert audio is available in six languages — English, Spanish, Traditional Chinese, Tagalog, Korean and Vietnamese.
- Wireless Emergency Alerts. The system that also carries Amber and wildfire alerts, at the higher magnitude bar.
For a household, redundancy is reasonable rather than excessive. The channels have different thresholds and different failure modes, and none of them costs anything.
What ten seconds is actually for
Not running outside. Not standing in a doorway — that instruction belongs to a style of house most Californians do not live in.
Drop. Cover. Hold on.
Get low, cover your head and neck, and hold on to something sturdy. That is the instruction attached to both feeling shaking and receiving an alert.
A handful of seconds is enough to move away from glass, to stop doing something that becomes dangerous mid-shake — carrying a pan, standing on a ladder, holding a knife — to get under a table, or to get a child under one.
And it is enough for the systems that do not need a human at all. The same alert can slow a train, halt a lift at the nearest floor, or protect equipment. Much of the value of earthquake early warning never passes through anybody’s attention.
The proposition is not prediction. It is action.
The Central Valley is not exempt from the question
There is a quiet assumption in the middle of the state that earthquakes are a coastal problem. It is a reasonable inference from where the famous ones happen and a poor basis for a household decision, because the cost of being wrong is asymmetric: enabling an alert takes two minutes, and the alternative is finding out during the event.
The alerting system is statewide, and the actions that matter locally — securing a water heater and tall furniture, knowing where the gas shutoff is, keeping shoes and a torch by the bed, agreeing where the family meets — are the same everywhere and cost nothing to have already done.
What this page deliberately does not do is tell a Fresno reader what their own local hazard is. That is a question for California’s published hazard mapping against a specific address, not for a general article, and a site that assigns a risk level to your street without reading the map is guessing.
How to verify this yourself
- California’s Earthquake Warning California page carries the operating arrangement between Cal OES and USGS, the M4.5 alerting threshold, the M2.5 display threshold, the Android integration and the six alert languages.
- The USGS ShakeAlert programme explains the detection and estimation chain, and states plainly that it is not prediction.
- The MyShake FAQ from the UC Berkeley Seismology Lab sets out the alerting threshold, the notification permissions the app needs, and the accuracy trade-off — early warnings are built from the first signals of a few sensors and are designed for speed at the expense of precision.
Thresholds and app behaviour are changed from time to time by the agencies that run them. Where this page and the state’s own page disagree, the state’s page is current and this one is not.
Also on this desk
- The Governor’s 30 September deadline — the other California calendar most people only meet at the last minute.
- The FAIR Plan’s dwelling increase — what a standard policy does and does not cover.
- Heat standards in Valley workplaces.