The seconds between an earthquake and a useful warning
Recent ShakeAlert reports reveal why earthquake detection time, phone delivery and useful warning are different measures of the same system.
An earthquake can be detected correctly without triggering a warning on your phone. An official report can publish a number of seconds without establishing that anyone received that much advance notice. Those distinctions are visible in the newest entry on ShakeAlert’s public message-summary page when checked for this article: a small earthquake west of Fortuna, California, posted on October 23. The interesting result is not a dramatic countdown. It is a trace of the machinery behind one. See the current summary archive.
A message after the earthquake began
The October 23 summary identifies an ANSS magnitude 3.4 earthquake 49.1 miles west of Fortuna. It began at 10:38 p.m. Pacific on October 22, or October 23 in UTC. ShakeAlert estimated magnitude 3.6 and issued its initial message 15.9 seconds after origin time. Five stations contributed to the final message. The report says the event was below the Wireless Emergency Alert magnitude threshold; the four listed cities have no warning-time value because their predicted shaking was weak or imperceptible.
That is a processing result, not a claim that coastal residents received 15.9 seconds of warning. The system spent that interval getting from the earthquake’s beginning to a message. The archive also cautions that summaries below estimated magnitude 4.0 are generally not reviewed for quality or accuracy. This provisional, small-event record is useful for understanding the pipeline, not grading its performance in a destructive earthquake.
Early warning is not prediction
An early-warning system starts with an earthquake already in progress. Faster P waves reach instruments ahead of slower waves that can produce stronger shaking farther away. Electronic communication can outrun the traveling disturbance, but it cannot begin with information that no sensor has yet observed. The USGS explanation of the physical sequence separates the rupture, wave arrivals, processing center and delivery partner. Each sits at a different place in the chain.
This creates a geographic constraint. A person close to the source may encounter shaking before the chain finishes; someone farther away may have more time, but may also experience weaker shaking. “Farther away” is therefore not equivalent to “better protected,” and a useful system cannot be evaluated by selecting only its largest possible countdown. For an engineering assessment, the important pairing is time available at a particular site and the shaking expected at that same site. The archival ocean-bottom instruments shown above represent another timescale: researchers prepare equipment to study earthquakes and aftershocks. A post-event research deployment is not evidence that an offshore early-warning network was already operating during a later earthquake.
A station is more than its sensor
A ground-motion measurement needs a physical reference. A seismometer measures motion using the response of an internal mass; modern instruments record electronic changes rather than drawing a pen trace on paper. A seismogram is the resulting record, not the instrument itself. That distinction, explained in the USGS instrument FAQ, matters when interpreting photographs of field equipment: the box that houses electronics is only one part of the measurement system.
The archival California station at the top of this story makes another point visible. The sensor depends on power and telemetry, not simply an algorithm in a distant data center. As an engineering inference, reducing processing time is of limited value if the data needed for that calculation arrive late. Reliability has to include the whole route from ground coupling to an available data stream. A solar panel, battery or communications connection may be less recognizable than a phone alert, but those ordinary components belong in the explanation of how an alert becomes possible.
The first solution is allowed to change
The current ShakeAlert algorithm description identifies three complementary approaches. EPIC uses early P-wave observations to estimate an event’s location, time and magnitude. FinDer uses the spatial pattern of peak acceleration to estimate rupture characteristics. GFAST-PGD uses satellite-positioning measurements of ground displacement for very large earthquakes. A solution aggregator combines source estimates, and a ground-motion module translates them into expected shaking distributions. A decision module determines whether to publish.
This is not one sensor declaring an immutable number. EPIC requires triggers from at least four stations, along with a nearby-station consistency condition. Estimates are revised as information accumulates. The documented alert-pause feature also limits delivery regions to a 100-kilometer radius during the first five seconds of an alert sequence before allowing unrestricted regions. That feature addresses early uncertainty and distant over-alerting; it is not a five-second guarantee to recipients. The technical design exposes the tradeoff: using early evidence gains time, while waiting can improve the picture of what is actually happening.
Why satellite positioning joins the seismic network
GNSS is the broader satellite-navigation family that includes GPS. For earthquake monitoring, a receiver can help measure how far the ground moves, rather than only how quickly it shakes. USGS announced the integration of real-time GNSS data in June 2024 to improve characterization of very large earthquakes. The addition complements seismic observations; it does not replace the fast initial detection stage or turn an earthquake into a predictable event.
The practical distinction is between speed and scale. The first portion of a rupture is not a complete description of the entire earthquake. A second type of measurement can provide information that the earliest seismic estimate lacks. For readers, that is a reason to expect updating information, not to assume that a later magnitude automatically proves the earlier alert was careless. It also changes what a useful network description should contain. The number of instruments is only the starting point; their locations, data quality and availability matter to the estimate. A deployment target and a verified operating network are different facts.
Magnitude describes the event. Intensity describes a place.
Magnitude is a measure of earthquake size; shaking intensity varies by location. The distinction is central to ShakeAlert’s delivery-threshold guidance, which explicitly says that different technical-partner routes need not activate for the same earthquake. A threshold based only on the event’s magnitude cannot tell a reader whether their particular location belongs inside the predicted shaking area.
The Washington, DC, research photograph above illustrates a related distinction: measuring a building’s ground motion is not the same activity as sending a West Coast warning. USGS explains that instrumenting structures helps engineers assess their response and improve earthquake-resistant design. A sensor in a building can supply evidence about motion at that structure, while an early-warning product estimates what is approaching. Neither makes the other redundant. The building, the ground beneath it and the incoming waves together determine consequences that a single magnitude headline cannot fully describe.
One source message, several paths to people
The USGS threshold chart currently available lists magnitude 5.0 and greater for WEA, paired with predicted intensity IV or greater. It lists magnitude 4.5 and greater for phone operating systems and apps, with intensity conditions that differ by route. The graphic includes both iOS-based and Android operating-system delivery. Its current file is dated January 30, 2026, even though the surrounding media page retains an older publication date. That is why an old platform comparison should not be treated as a current compatibility guide.
The USGS enrollment page separates the scientific message from the systems that deliver alerts. It also warns against confusing early warning with the Earthquake Notification Service. ENS can report an earthquake after it happens, but it is not designed to provide protective lead time. An email about a ShakeAlert performance report is therefore not evidence that the same email channel can warn someone before shaking. A delivery route must be judged by its intended function, not by the presence of “earthquake” in the message.
September’s Washington report shows the updating process
A second recent record makes the distinction between initial and final estimates concrete. The September 29 Maplewood-area summary lists an ANSS magnitude 4.3 earthquake in Washington, about seven miles west of Maplewood. ShakeAlert’s initial estimate was magnitude 3.6, with the first message 9.7 seconds after origin. Its peak and final estimate reached 4.0 at 20.4 seconds. The final message used 153 stations. Again, the report says the event was below the WEA threshold.
Those figures should not be turned into a simplistic contest with the offshore Fortuna event. Different depth, location and network geometry mean different information reaches the system at different times. Nor does a much larger final station count describe how much evidence existed at the first decision. The useful comparison is procedural: both reports preserve a sequence of estimates and identify where the public-alert threshold was not met. They document the system’s decisions more precisely than a headline claiming that it “warned everyone” or “warned no one.”
Reliability includes recognizing signals that are not an earthquake
USGS’s January 13, 2026 technical update on a December 4, 2025 false alert is an important counterweight to success stories. That alert concerned a nonexistent magnitude 5.9 earthquake near Carson City, Nevada. The agency’s preliminary analysis described five problematic station inputs: three associated with malfunctioning communications, one with a seismometer malfunction and one with local urban noise. The existing safeguards did not prevent the false interpretation. Delivery partners then sent alerts through several channels.
The update said the investigation would inform software changes to recognize false triggers while retaining rapid detection. It did not establish that false alerts had become impossible. The engineering lesson is narrower and more useful: agreement among multiple inputs must be evaluated alongside how those inputs can fail. Several bad observations do not necessarily represent several independent earthquakes. This is a dated example of a documented system failure, not an allegation of a new October malfunction or a claim that every unexpected alert has the same cause.
Automated protection has its own clock
The USGS communication and technical-engagement strategy distinguishes detection and processing, delivery, and protective action. It describes uses such as slowing trains, opening firehouse doors and controlling utility infrastructure. These are examples of possible or implemented applications across the wider system, not proof that every train, door or valve on the West Coast is connected. A science message and a functioning local installation are separate achievements.
For a hypothetical rail application, the relevant endpoint would not be the instant a server logged an alert. It would include the command reaching the control system and the equipment responding under the operator’s design. Likewise, a school announcement would need to reach occupied spaces in a form people can understand. These are analytical examples, not reports of specific deployments. The distinction changes what a meaningful test asks: did the intended protective response happen in the available interval, and what evidence shows that it did? Counting messages alone cannot answer both questions.
Measure usefulness without asking it to do the impossible
When a future performance claim advertises a number of seconds, three questions make it interpretable. What starts the clock: earthquake origin, source-message publication or phone display? What stops it: first felt motion, predicted S-wave arrival or a specified level of shaking? And whose location is being described? The public summaries explicitly exclude the extra time that licensed operators need to deliver alerts to end users. Their timing is valuable evidence, but it is not a substitute for recipient-side measurements.
No alert is also an ambiguous observation on its own. A threshold may not have been crossed, a person may have been outside the target area, or the chain may not have delivered a useful warning in time. Conversely, receiving an alert does not establish that strong shaking must occur at that exact location. The official limitations graphic captures those possibilities without suggesting that users should diagnose an event while it unfolds.
The immediate public message in the USGS guidance remains brief: if shaking is felt or an alert arrives, Drop, Cover, and Hold On. The larger engineering story belongs before and after those seconds. Historical field measurements and building instrumentation help improve understanding and resilience over years; early warning tries to use a much smaller interval within an event. It is a useful additional layer, not a promise that software can stop the ground from moving.
Comments
Post a Comment