California’s seismic network is 95% installed. Here’s what the last 5% means for coverage.

California’s earthquake early warning system runs on sensors. More than 1,115 of them, spread across the state, feeding real-time ground motion data to ShakeAlert so it can detect earthquakes and push alerts before strong shaking arrives.

As of late 2025, more than 95% of that network is in the ground. The remaining stations are scheduled for completion by December 2026. That last 5% matters more than the number suggests.

Where the gaps are

The stations already installed are concentrated where most Californians live and work: the Bay Area, Greater Los Angeles, the Central Valley, San Diego. Urban areas have the densest coverage, with sensors spaced roughly 10 km (about 6 miles) apart. Near major fault traces, the target is even tighter: stations within 5 km of mapped faults, because proximity to the source is what determines how fast ShakeAlert can detect a quake.

The last 5% of stations are going into less populated areas. Rural stretches of the Eastern Sierra. Remote sections of the Northern Coast. Parts of the desert. These are places where sensor spacing is currently wider than the 20 km target for seismic source zones, and much wider than the 40 km floor for general coverage.

That matters because earthquakes don’t follow population density. The Ridgecrest sequence in 2019 happened in the middle of the desert and was felt across half the state. The recent M5.7 in Nevada hit just outside ShakeAlert’s reach. Sparse sensor coverage in remote areas means longer detection times and shorter warnings for everyone downstream.

What denser coverage actually changes

Warning time in ShakeAlert is a function of two things: how far you are from the earthquake, and how fast the system detects it. You can’t change the first one. But you can change the second by putting sensors closer to where earthquakes start.

According to the USGS Open-File Report 2021-1026, ShakeAlert’s detection-to-alert time ranges from about 4 to 20 seconds depending on earthquake depth, fault type, and local station density. In dense urban networks, alerts go out in 8 seconds or less. In areas with wider spacing, it takes longer, and that eats into the warning window.

Research on station density improvements shows that closing the gap from 50 km spacing down to 20 km can add roughly 5 seconds of warning time. That may not sound like much in everyday terms. For a hospital mid-surgery, a fire station with apparatus bay doors open, or a manufacturer running a hazmat process, 5 seconds is the difference between an automated shutdown completing and a preventable accident.

The funding question

This progress happened because California invested. The state’s Earthquake Early Warning Program, managed by CalOES, has funded station installation, backhaul communications, and ongoing maintenance alongside federal USGS contributions. The California Integrated Seismic Network (CISN) operates the sensors that feed ShakeAlert in the state.

But the network isn’t just a build-once project. Sensors need power, connectivity, and maintenance. The FY2027 federal budget proposes cutting ShakeAlert funding by 72%, which puts long-term operations at risk even as the physical buildout approaches completion. California has signaled it will continue state-level investment, but the federal picture is uncertain.

What this means for facility operators

If your facility is in a major metro area, you already benefit from the densest part of the network. The question for you is whether your building is connected to ShakeAlert in a way that acts on the warning automatically. A phone alert is better than nothing. An integrated system that triggers gas shutoffs, opens bay doors, pauses elevators, and alerts occupants is better than a phone alert.

If your facility is in a less dense area, the network buildout through December 2026 may meaningfully improve your warning time. It’s worth understanding what ShakeAlert coverage looks like in your specific location and what automated early warning solutions can do with even a few extra seconds of lead time.

The sensors are going in. Whether your facility can act on what they detect is a separate problem, and one worth solving before December.