Edit: Two new items may shed more light on this question:

The Nature Communications paper Prompt gravity signal induced by the 2011 Tohoku-Oki earthquake presents a "...report on the search for a prompt gravity signal during the rupture of the 2011 Mw 9.0 Tohoku-Oki earthquake in data recorded by a superconducting gravimeter in the underground Kamioka Observatory and five nearby broadband seismometers from the Japanese network F-net."

They conclude that with a certainty of about 99% that a prompt gravity signal was detected. The change in position of large masses of crust resulted in a tiny change in gravity even 500 km away and this was within detectable limits of this extremely sensitive instrument.

Since gravity's influence "travels" at the speed of light it is suggested this might be an avenue to be explored for a potential early warning system.

Considering that electrical and optical signals (e.g. the internet or radio or a more dedicated communications system) travel about 2/3 the speed of light in wires and fiber, and even conventional internet pings are of the order of 100 milliseconds or better, is there any suggestion or discussion anywhere else besides this paper that super-sensitive gravimeters would be in any way better and/or cheaper than a network of cheap sensors?

In either case the idea is an early warning for something "big" rather than accurate measurements. Although the paper and analysis is interesting and elegant - is there any serious discussion or speculation that something like this - a prompt gravity signal - could somehow be actually useful as an early warning system?


2 Answers 2


If you have a sufficient number of sensors, e.g., land-based or ocean-bottom seismometers, in the vicinity of where strong earthquakes typically happen, then it will take a few seconds for seismic waves to reach them, and a few hundred milliseconds for the electric signal to get to a place where a warning could be triggered.

On the other hand, the gravimeter would provide essentially instant warning. So you save a few seconds. At the same time, it takes many many seconds for earthquake waves to travel to the places where the warning would matter. In fact, the earthquake rupture itself may be many many seconds long. And it will take minutes for tsunamis to reach shore.

In other words, what you gain in advance warning time is not actually very large, and it may not be worth the effort. I would also not be surprised if it takes tens of seconds or minutes to actually see the change in gravity in noisy gravity measurements, and/or to take the same amount of time to just process the data.

  • $\begingroup$ Thanks for your perspective, it makes a lot of sense. I'm also wondering if this is a new idea, or if it has been considered previously, or if this is pretty much a new idea prompted by the opportunity that the 2011 Tohoku-Oki earthquake provided by happening so close to this facility. $\endgroup$
    – uhoh
    Commented Nov 26, 2016 at 0:13
  • 1
    $\begingroup$ I'd like to add the the "data process" part: this is a huge field of research, to understand where are the most damaged area from the seismic signal. During an earthquake, you don't always know where the damage is because of communications failure and overall chaos. Proper analyses can take weeks to months, and research is focused on attempting to make good enough estimates in minutes to hours. Several milliseconds do not matter in this case. $\endgroup$
    – Gimelist
    Commented Nov 26, 2016 at 2:17
  • $\begingroup$ I've edited the question, added two new links. $\endgroup$
    – uhoh
    Commented Dec 2, 2017 at 16:45

The Tohoku earthquake registered on both the superconducting gravimeter (SG) and broadband seismic networks. The SGs are accelerometers with roughly 0.1 nanometer/second^2 sensitivity at 1 sample per second. The broadband seismometers are about 10 times less sensitive, but usually store at 40 sps or 100 sps. And the broadband seismometers are three axis, so you get the benefit of three independent signals. The most practical way to begin using and calibrating a gravimeter array is to use time of flight methods (the speed of gravity and speed of electromagnetism are identical, not close, identical) globally. Very sensitive detectors at picometer/second^2 are practical and relatively low cost now. Operating those at 1 MegaSamplePerSecond (Msps) gives roughly 300 meter resolution. And you would be modeling and tracking the seismic waves at the surface and interior, and also motions at the source. So you can image the source region and use the gravity data to constrain the seismic (acoustic) models of the energy, direction and timing of the event. The global seismic arrays are fairly stable and sophisticated. There are "Big N" experiments with thousands of three axis seismic sensor that run up to 5000 sps. I have tried to test many of the new ones to see if they are potential gravimeter array detectors.

The broadband seismometer usually report velocity (meters/second), but take the first time derivative and it is an "accelerometer". My rule for groups is "if you can track the sun and moon with your "accelerometer" then you can call it a "gravimeter". There are MEMS "gravimeters" now by that definition, where groups take methods used to make cell phone MEMS three axis accelerometers, and beef up the read out. Those MEMS gravimeters drift, but you use the sun moon tidal signal (a huge signal that is most of what the SG records look like) and calibrate them continuously. It is the most elegant thing I have ever worked on to see that two numbers (an offset and scale factor) are enough to fit the complex signal at a SG or gravimeter station for a month of data. The reason it works so simple is that gravitational acceleration is almost pure Newtonian GM/r^2 -if you use the precise positions of the sun moon earth and use vector calculations to the station. The "trick" is that you have to subtract the acceleration of the sun and moon on the center of the earth, add the rotational acceleration for latitude, and use JPL or Python to get the positions of the Earth Sun Moon every second.

You could possibly use three LIGO "detectors" to try tracking earthquakes. But their strain data is only 16384 sps (about 18 km) and they are huge and really tedious to use their data. And they did not remove the "Newtonian noise" from their methods well. They do not share their earth and solar system data at all. And there are atom interferometer, atomic clock gravitational sensors, Bose Einstein gravimeters, electrochemical gravimeters, electron interferometers gravimeters, and many sensitive small ways to measure atomic scale displacements over time. You can use an atomic force sensor and record the data at up to Msps or even Gsps now. There are so many ways to measure gravity for this sort of thing now, I almost gave up trying to track them all.

I paid attention to that earthquake because my son and his wife were living in north Tokyo and had just moved there from Sendai, near where the earthquake hit. I studied gravitational detection at UMD College Park. Joe Weber wanted his detector to be used for communication. Robert Forward his student and collaborator wanted gravitational methods to be practical. I call it "gravitational engineering" because of them.

There are lots of ways to measure the speed of gravity now. So that is my second hurdle. "If your array can measure speed of gravity precisely" then it it can be called "a gravitational imaging detector array". That is the first thing you want to do when you calibrate your array - check to make sure it measure the speed of gravity and use some earthquakes to calibrate. You can also use cars on the highway if you track and classify them with precise locations and times. Surf waves with 3D models can also work. I got into this because I used the SGs to measure the speed of gravity. it was ambiguous because of "single axis" so that is why I spent close to a year to go through ALL the IRIS.edu seismometers and check decades of records to find quiet sites where the sun moon tidal signal is clear on all three axes. Those only require 6 numbers, linear regression for each axis. Really very beautiful how well Newton and basic vector calculations work for everyday things. When you do time of flight gravity sensing, it gets complex again, because all the frequency of "gravity signals" are accessible now - up to about 100 GHz. But people talk about x-ray and gamma ray frequencies for gravitational signals. Most of the "quantum detectors" and many "entanglement" detectors and communications systems are likely very good high frequency gravitational sensors. It starts with tracking the sun and moon vector tidal signal, and then "measure the speed of gravity". After that it is just "gravitational engineering" with first year vector calculus and off the shelf low noise amplifiers and high bit high sampling rate ADCs.

Richard Collins, The Internet Foundation


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.