Earthquake Early Warning (EEW) aims to provide warning that shaking from an earthquake is expected before it actually arrives. EEW can be used to protect vulnerable infrastructures using automated shutdown procedures, and to save lives if the general public is notified about imminent strong ground shaking and trained to respond appropriately. EEW is not earthquake prediction, as the earthquake has already begun. In normal cases, EEW can only provide up to a handful of seconds warning time.
Two characteristics associated with the speed at which earthquake information travels make EEW possible. First, an earthquake emits different types of seismic waves: Smaller amplitude seismic P-waves are the fastest, traveling at roughly twice the speed of the more damaging S-waves and surface waves that follow behind. Second, seismic waves are far slower than the electromagnetic waves used in modern communication systems to transfer information. In an effective EEW system, the first P-waves from an earthquake are recorded across a dense seismic network as soon as waves reach the surface (or the shoreline). The data is then immediately transmitted to a processing hub at the speed of light, and information from multiple stations is combined to predict ground shaking. In the systems developed at ETH, the predicted ground motion is based on locating and characterising earthquake. Based on the predicted ground shaking, the EEW system then can broadcast warnings to the affected areas. Targeted warnings allow the public for example to seek cover and machines to automatically move to a save mode (e.g. elevators).
An early warning system should ideally give as much warning as possible to areas that will experience damaging shaking. In general, the closer a site is to the rupture, the stronger is the shaking that can be expected, and the shorter is the warning time. This is particularly true for small and moderate earthquakes, up to about a magnitude of 6.5. Very close to the rupture, where shaking will be strongest and damage hence greatest, it may not be possible to provide a warning before the strong shaking begins. This is because the time difference between the weaker P- and the stronger S-waves becomes very short, and even the P-waves can be damaging. If an advanced warning can be provided, the warning time will only be seconds or even fractions of seconds. Warning times can be longer at greater distances, though the shaking will be generally less intense.
There are two exceptions to this rule: First, for large earthquakes with magnitudes of about 6.5 or more, the rupture length and duration can be significant. Thus, at regions far from the epicentre but still close to the rupture, more advanced warning of strong shaking is possible. This is a key argument for building early warning systems in California along the San Andreas fault. The second exception concerns areas with strong local site effects. Mexico City for example, is built on a deep sedimentary basin that strongly amplifies earthquake ground motions. Therefore, even moderate shaking induced by a strong earthquake on the Pacific Coast may trigger heavy damage in Mexico City. As the city is about 300 km away from the Pacific subduction zone, a major source of earthquakes, there can be an exceptional long warning time.
As a simple example consider an earthquake of magnitude 7.5 for which the radius of the area that will experience strong ground shaking is roughly 55km, assuming a point source rupture for simplicity. The S-waves, which mark the onset of the most energetic seismic waves, travel at a speed of about 3.5km/s, and P-waves, which carry the first information about the earthquake, travel at a speed of about 6.5 km/s. For an earthquake that happened at a depth of 10km, a seismometer directly at the epicentre could first detect the earthquake from P-waves around 1.5s after it occurred. At this point, the S-wave arrives only around 1.3s later, with very limited scope to provide advanced warning. The S-waves though will take over 15s to reach a site 55km from the earthquake. Considering operational delays of around 3 s (to collect seismic data from the wider network at the hub, predict the ground shaking, and send the alert), a warning could be issued around 10 s before the onset of strong ground shaking for sites at 55km, though right above the epicentre the warning would be late.
EEW is not a recent concept. A simple EEW system was outlined in 1868 by the Californian physician J. D. Cooper in an article published by the San Francisco Daily Bulletin. Cooper proposed the installation of an array of seismic detectors from 10 to 100 km away from San Francisco. When large ground motions triggered a sensor, a signal would be telegraphed to San Francisco and would automatically ring a bell. Despite the simplicity and plausibility of this idea, the first EEW system was not realized until the era of digital seismometry began in the late 1980s. In Japan, the high-speed Shinkansen trains were automatically slowed down when a strong earthquake was detected in the proximity of the railway tracks. Following the 1985 earthquake that devastated Mexico City, a public system for the city became operational in 1993. Today, EEW is becoming more normal: apart from Japan and Mexico, operational or demonstration EEW systems that send alerts to test users or even to the general public are available in countries across the world, including Romania, Taiwan, Turkey, Italy, Switzerland, Chile, Nicaragua, China, and the West Coast of the USA.
In Switzerland we expect an earthquake with a magnitude of about 6 to occur every 50 to 150 years. For earthquakes of this size, the zone around the earthquake’s epicentre, which experiences strong shaking, is only approximately 20 to 30km wide. As a result, there is very little time between the first detection of the P-wave and the arrival of the destructive S-waves, and possible warning times are short at best. Automated processes like putting systems in a safe mode (e.g. shutting down machinery or electrical equipment) are suited best to such short warning times. But even when arriving during or shortly after the onset of shaking, early warning will increase situational awareness and in some cases allow people to take precautionary actions. A train driver may be warned of potential landslides or officials can initiate emergency procedures. EEW can thus be seen as one end of the spectra of the SED’s alerting scheme, which we attempt to improve continuously.
All stations in the Swiss national network include strong motion sensors and we target a data latency of 1 s. With more than 150 stations in the network, we typically provide first source parameter estimates of an earthquake within 8 to 12 s of its occurrence across the entire country. The monitoring infrastructure in Switzerland is essentially EEW-capable, though it can always be improved to provide faster and more reliable alerts.
The SED is involved in developing and testing different EEW algorithms in Switzerland and around the world. We focus on building operational software that allows us to account for complex EEW scenarios while at the same time making warnings faster and more accurate (see below). We also build tools for disseminating EEW to end users.
Over the last decade, the Seismic Monitoring Group at the SED has developed three algorithms that can provide rapid estimates of earthquake magnitude, location and fault rupture extent, with parameter estimates typically becoming available within a few seconds from the event origin.
The group uses the SeisComP Seismic Monitoring Platform for automatic monitoring of seismicity in Switzerland, and for curating the national earthquake catalogue. SeisComP is popular for regional seismic monitoring across the world. Our strategy is to integrate our EEW algorithms into SeisComP, allowing a single framework to be used at all timescales and enabling other groups to easily install and test EEW in their own networks.
Further details and information about the methods and software used for different early warning products are linked below: