Seismo Blog

It Worked ... and It Did Not!

Categories:   Southern California  |  Earthquake Faults and Faulting  |  Earthquake Early Warning

July 10, 2019 

Shake-Alert performed well, but the public was left in the dark.

The magnitude 7.1 quake under the Mojave Desert on Independence Day weekend was the first test of the West Coast wide earthquake early warning system in a really strong shaker – and this system dubbed Shake-Alert performed very well. The public, however, got a completely different impression. While the Los Angeles basin was rattled by the seismic waves coming from the Mojave quake, downtown high rises swayed and several movie theaters were evacuated as a precaution. TV anchors and Twitter headlines claimed: "Earthquake early warning system failed!" Disappointed Angelinos, who felt the rattling but had not gotten any warning at all, vented their frustration on social media – a mood which eventually swept through the whole State and even made headlines on national news. So: How can scientists celebrate the good performance of their system, while everybody else feels left out? Where is the disconnect?

Let's look at the facts. The original research and development of Shake-Alert was performed jointly by researchers at the Berkeley Seismological Laboratory (BSL), Caltech in Pasadena and the Swiss Federal Institute of Technology in Zürich. This fully automated system is designed to give people up to few dozen seconds of warning time before the seismic waves rock one's location. In most cases, this warning time is long enough for people to take protective measures (drop, cover and hold on). After several years of testing researchers from the Universities of Oregon and Washington joined the team and the United Stated Geological Survey (USGS) became the lead agency. A prototype production system is now in place for the US West Coast. Financed by the USGS and several State agencies, among them Cal OES, the system is currently being built out to include several hundred new and upgraded seismic stations in California, Oregon and Washington.

Despite relying only on the current bare minimum of stations, the system created the first alert for the 7.1 quake in just eight seconds after the quake's rupture had started on the evening of July 5th. This short time is even more remarkable, because the hypocenter of the quake was located more than ten miles below the Earth's surface. That means, it took the first seismic waves at least five seconds to reach the nearest seismic stations at the surface, before they could even be detected. Immediately after this detection, the system calculated the location of the epicenter with the tiny error margin of less than two miles.

As more seismic stations in Southern California detected the seismic waves, Shake-Alert estimated the magnitude of the quake at 6.3. This is of course significantly lower than the quake's actual magnitude of 7.1, but it is inherently difficult to estimate a quake's ultimate energy (as expressed by the magnitude) from just the first few seconds of seismic recordings.

Based on these initial calculations, a warning was generated, which would have given someone in downtown Los Angles 49 seconds time to take preemptive action (see playback of the warning in the accompanying figure). However, as Shake-Alert is still in a prototype production state, this warning was only received by several dozen "beta users" like police departments and emergency centers. It was not sent to the public at large.

There are several reasons for not making the alerts fully public as of yet. Like hurricane, tornado or tsunamis warnings, the official alert should be issued not by a University lab like the BSL but by a US government agency. The United States Geological Survey has been assigned to do just that, but their experts are still looking for the best way to send an earthquake early warning to the public.

Other countries with similar systems have long solved this problem of alert dissemination. Mexico relies on warning sirens in public places and radio messages. In Japan, alerts flash as banner headlines across TV screens, computers deliver messages received via the internet, cell phones beep and in the Tokyo subway and on other rail lines, monitors show the alerts.

As the backbone of the Shake-Alert system has been up and running for more than a year, the City of Los Angeles felt that it could not leave its residents in the dark any longer. Hence, earlier this year it rolled out its own cell phone warning app called "Shake-Alert-LA". This app uses the information generated by the early warning computers to send messages to the phones of each resident in Los Angeles County, who has downloaded the app.

However, this LA specific app did not trigger after having received the Shake-Alert warning for the 7.1 Mojave earthquake. The reason can be found in the lower left corner of the display in the figure. Under "Expected Intensity" it displays the Roman numeral III, an indication of weak shaking. As rattling of intensity III will not cause damage and may even be missed by many people, the trigger threshold for the LA specific dissemination app was set to "IV", the next higher step on the intensity scale. Hence the LA app kept silent although the seismic waves of the Mojave quake rolled through LA county with intensity IV, being felt by many. The reason that Shake-Alert predicted only intensity III for the LA basin was the glitch mentioned above. The warning algorithms underestimated the magnitude of the quake by 0.8 magnitude units. Had the computers gotten the magnitude right, the prediction for LA would have been "expected intensity IV" and the app would have sent out warnings.

There are several lessons to be learned from the performance of the various components of the current Shake-Alert system:
- however difficult it may be, the initial magnitude calculations need to be improved,
- the threshold for triggering the LA specific app need to be adjusted, and most importantly
- it is time for the government to make a decision of how to bring earthquake early warning to the public at large. Researchers at the BSL have developed a tool for that, dubbed My Shake, which the blogger will describe in one of the next posts. (hra167)