1. Why Did a Newspaper's 1868 Earthquake Warning Idea Not Become Real?
On November 3, 1868, a newspaper in San Francisco published a proposal by Dr. J.D. Cooper. His plan was to detect an earthquake near its source and use the telegraph to automatically ring a bell as a warning [3]. An overview by Yutaka Nakamura introduces this as the "first concept" [3]. The plan was never realized.
About 100 years later, in 1972, Dr. Motohiko Hino and others proposed a "10-second advance detection system" for Tokyo [3]. Since the author of the overview works for a company related to UREDAS, it is good to read the development history as "how it is introduced" rather than absolute fact. The idea of "warning before the shaking" has been in people's minds for over 100 years. Where does that "before" come from?
2. How Do Faster Weak P-Waves Arrive Before Strong S-Wave Shaking?
Earthquake waves include P-waves, which arrive first, and S-waves, which arrive later and cause strong shaking. P-waves travel at about 7 km per second, and S-waves at about 4 km per second [1]. This speed difference creates the time available for warnings.
For example, at a location 28 km (about 17 miles) from the epicenter, simple calculations show P-waves arrive in about 4 seconds and S-waves in about 7 seconds, leaving a difference of about 3 seconds. At 140 km (about 87 miles), P-waves take about 20 seconds and S-waves about 35 seconds, widening the difference to about 15 seconds. These are calculated values based on speed alone; actual usable time is shorter because detection and calculation take time.
This is similar to the time difference between seeing lightning and hearing thunder. However, there is a key difference. With earthquakes, machines can analyze the first wave on the spot and send a warning ahead to distant locations [1].
3. How Did the UREDAS System Warn Japanese Trains Ahead of Earthquakes?
UREDAS is a system that was successfully developed by the Japanese National Railways in 1983 and put into practical use on the Tokaido Shinkansen in 1992 [3]. Its feature is using just one observation point to detect the first motion of P-waves from three directions to make a judgment.
It estimates the earthquake's magnitude from the period of the first motion, the distance to the epicenter from the amplitude and magnitude, and the direction and depth from how the wave enters. Then it predicts the area likely to suffer damage and issues a warning [3]. Initially, estimating the period to determine magnitude took about 3 seconds. After the Great Hanshin-Awaji Earthquake in 1995, this 3-second delay became a major issue. In 1997, a new warning method was devised, and "Compact UREDAS," which could issue a warning 1 second after detection, was created by JR East in 1998 [3].
Going from 3 seconds to 1 second, UREDAS shows how the idea of getting ahead was actually used in the field of railways and gradually became faster. However, documents do not show any direct connection between UREDAS and the current earthquake early warning system beyond the common use of P-waves.
4. How Do Nationwide Seismometers Calculate an Earthquake Starting From One Close Station?
According to the Japan Meteorological Agency (JMA), there are about 690 JMA seismometers and intensity meters nationwide. Data from the National Research Institute for Earth Science and Disaster Resilience's observation network (about 1,000 locations nationwide) is also used [1]. The numbers change by year, so these figures are from the time of writing.
When an earthquake occurs, seismometers near the epicenter detect the P-waves. Using data from multiple observation points, the epicenter, magnitude, and expected shaking intensity are calculated. The processing is automatic and described as instantaneous [1].
There are two reasons why the epicenter and magnitude can be estimated quickly from data from a small number of observation points. First, computer performance has improved. Second, methods have become available to estimate the epicenter and magnitude from P-wave data from a single observation point [1]. This network allows calculations to begin using data from stations close to the epicenter, combining information from multiple points to issue warnings.
5. Why Do Earthquake Early Warnings Not Reach Places Close to the Epicenter?
The general public earthquake early warning system started on October 1, 2007. The JMA began trial provision in February 2004 and started providing warnings to advanced users in August 2006 [4].
During the Niigataken Chuetsu-oki Earthquake on July 16, 2007 (Magnitude 6.8, epicenter depth about 17 km), the JMA issued the first report 3.8 seconds after detection. However, this 3.8 seconds is the time "from detection to the first report," not the time remaining before shaking. It did not reach areas close to the epicenter in time, though surrounding areas received warnings seconds to tens of seconds in advance [6]. At that time, this was an announcement for users of the advance provision.
The JMA publishes these limits as usage precautions. From the announcement to strong shaking, it takes seconds to at most tens of seconds. For shallow inland earthquakes, warnings to places close to the epicenter cannot, in principle, arrive before the strong shaking. There is also an error of about ±1 intensity level in the predicted intensity [2]. Knowing that "there are places where warnings do not arrive" is an important step in using this system well. So, what improvements have been made to increase the places where warnings arrive and reduce errors?
6. How Did Newer Systems Handle Multiple Earthquakes and Measure Shaking Directly?
Previous systems had a weakness. They sometimes processed separate earthquakes occurring almost simultaneously as a single earthquake. This meant the correct epicenter and magnitude were not found, leading to cases of predicting excessively high intensity [5].
To address this, the IPF method for distinguishing multiple earthquakes was introduced in December 2016. In March 2018, the PLUM method was introduced. The PLUM method predicts future shaking based on shaking actually observed, without finding the epicenter [5]. The first is a "method to distinguish," and the second is a "method to skip the step of finding the epicenter." These are two different answers to the same problem.
In the about 10 years since general provision began, warnings were issued 11,343 times, including forecasts. For earthquakes where intensity 4 or higher was observed or predicted, the predicted intensity was within ±1 intensity level of the observed intensity in about 80% of cases, according to the column [5].
7. How Can You Calculate the Time Gap Between P-Waves and S-Waves at Different Distances?
Using a calculator, let's calculate the arrival times and the difference for distances from the epicenter, assuming P-waves travel at 7 km per second and S-waves at 4 km per second. Try changing the distance to 14 km, 56 km, 140 km, and so on, and make a table.
For example, at 14 km (about 9 miles), P-waves take 2 seconds and S-waves take 3.5 seconds, so the difference is 1.5 seconds. You should see with numbers that the shorter the distance, the smaller the difference. This lets you see for yourself why earthquake early warnings "sometimes do not arrive in time for places close to the epicenter." Note that the actual usable time is shorter by the amount of time needed for detection and calculation.
If you want to read the original sources, the JMA's "How Earthquake Early Warnings Work" [1] and "Characteristics, Limits, and Usage Precautions" [2] are available in Japanese. For the history of development, Yutaka Nakamura's overview [3] is a key resource.