Biondo Biondi How to measure an earthquake through the internet

Show video

today on the future of everything the future of earthquake sensing so many places on earth are subject to the risk of earthquakes which traditionally have been measured with the richter magnitude scale bigger richter values means more and longer shaking uh and it ranges from one to for the barely perceptible to seven eight nine which means significant violence shaking and in many cases buildings falling down there are more modern measurements uh but all try to quantify both the amount and duration of the shaking or seismic activity now earthquakes can happen anywhere but the most active earthquake regions are in asia from japan to central asia and the west coasts of the north and south america and of course the damage from earthquakes includes not only the direct damage from shaking but also the subsequent fires landslides on the land tsunami waves at sea many of us remember the earthquake in sumatra in 2004 that triggered a tsunami causing 227 000 deaths of course the japanese earthquake in 2011 triggered both a tsunami and the breakdown of the fukushima nuclear reactor um a little bit closer to where i live san francisco suffered great loss of life and property in 1906 uh and in 1989 uh the san francisco area was again hit by the loma prieta earthquake at the time if you will allow me i was a medical student i i was stationed in the icu of the veterans administration hospital in palo alto uh there were several violent jolts and then the beds started rolling around the icu uh it turns out that we had forgotten to lock down the wheels this was actually quite serious because several patients became disconnected from their breathing machines uh we of course rapidly reconnected them it was clear that the building was severely damaged and so we had to evacuate everybody to the lawn outside and then ultimately to nearby buildings and it went well and we didn't lose anybody in that uh in that in that hospital during the earthquake so there is great interest in predicting earthquakes but this has proved very difficult even however an early indication of an oncoming earthquake might provide precious seconds of warning that would allow people to seek safety there's also great interest in monitoring uh baseline seismic activity so can we can understand the amount of activity that is usual and when that activity may become notable or dangerous professor biando biandi is a professor of geophysics at stanford university his group studies seismic activity through many measurement technologies including imaging motion detection and other new methods biando recently earthquake data has come from fiber optic cables the one the same ones i believe that provide internet and telecommunications how did this come about and how useful is this data well uh thanks so much and thanks for inviting me i'm happy to be here and i do share some of the memories that you mentioned on my creator i was a grad student up on the fourth floor mitchell building on stanford campus and that's definitely marked my thinking uh then on i went uh and became a seismic imager so different than an earthquake seismologist in which we have some permanent one here at stanford i use seismic waves to image the subsurface and the one of the differences between cyclic imager and earthquake seismologists is that we need to sample the wave fields well enough to be able to use that wave field to image the subsurface so how the fiber optic seismology came about is when many conjunctional things one of the really of the starting point is when uh google fiber installed fibers in the house that i was building on stanford campus and i saw how easy and what kind of a physical system was of the fiber used for telecom then i learned from a friend of mine that you could use those fibers to record seismic data with unprecedented sampling of the order of one meter instead of several kilometers of conventional earthquake seismology putting all that together the telecommunication infrastructure that is uh everywhere and is in particular is around the pacific so we have a high-tech ocean and on both sides and we are also the area of earth that is most uh really uh subject to earthquakes danger together with my uh homeland in italy but unfortunately the uh periodically is shaken up so i thought can we actually use those fibers that have been deployed for telecommunication to record seismic wavefields and better imagery earth and better understanding earthquakes so let me let me let me just stop you so for people who don't think about this i don't think it's immediately obvious that little um glass fibers that that carry light would be useful for motion detection so can you just take us through a little bit of that i guess it's physics oh that's absolutely uh the so as uh many technology was initially developed uh by the military for submarine sensing so just a a good credit to the government that puts a lot of money and funds in basic research and then finds a very useful applications and the basic physics is that you do have a laser on one end of the fiber and you have a receiver at the same end and the impurity in the fiber reflect back some of the lights that is emitted by the laser now very complex processing of what is sends back measure basically how the fiber cable gets infinitive so longer or infinite extremely shorter we call that in physics a strain but is really the cable getting shorter and longer and when if seismic waves pass through the cable it gets the cable to get longer and shorter so that is the basic mechanism in which you can measure seismic waves using a fiber cable are you able so uh some of these cables as i understand it can be pretty long are you able to localize wear in the cable stretching is happening that's because the speed of light is so high so you can basically measuring the timing of the reflections or you can localize with basically one meter pushes you over all over one meter and uh with just one interrogator we just recorded data a months ago uh under monterey bay and under san jose and uh with 50 kilometers you can interrogate a 50 kilometers of fiber with john just one instrument oh wow so you get 50 000 channels running at all time just with one instrument okay so i i had interrupted you you were telling me the story first that you had you had you had fiber installed in your home which gave you this kind of idea then you realized that the network is pretty available um and so um what kind of measure how does this does this revolutionize the seismic measurement or is it is it a cheaper faster way to get the same data that you always could get or is it actually opening up new types of data that you hadn't been able to get before it does open to a new type of data i would not have got excited if it was just uh or cheaper and more and basically because what i mentioned in the beginning to use the wayfield to image the subsurface you need to sample them densely enough so you go from the conventional seismometer even in the bay area that is uh kind of together with la and tokyo probably the urban area that are better instrument around the world you do have a size material five ten kilometers here we are talking about having a sensor every meter and uh since fibers is running everywhere for thanks to the telecommunication uh use and is increasing exponentially partially even by 5g by all the netflix and the zoom that we are doing and so conceivably so when i started it that was uh six years ago i call it the billion sensor away project that was a little ambitious and i always kind of denuded and say i hope that before i retire i see that happening and affirmative is coming uh before the thai retiree don't need to uh to be retiring at ninety and uh the the idea is that if you now sample their way fields under cities where has been historically very much of a problem to put instruments and uh manage them uh with that kind of one meters uh sampling we can really discover and see things in the wayfields that before we were not able to do it yeah so so a couple this is very exciting and i just i'm trying to get these questions a little bit orderly so forgive me but i guess the i understand from doing a little bit of homework that there's plenty of capacity in other words you don't need to use the actual live wire that my internet is coming over because there are available fibers in these bundles that are not being used and is it my understanding is you take advantage of these unused ones is that right that is correct uh there is a is sometimes called dark fiber or the cost of the fiber is fairly minimal we are talking about dollar a meter uh what is expensive is deployed so for economical region a company that deployed fibers they over deploy in case that time break or for future uh growth so there is plenty of dark fibers available around around the world even under the oceans that is actually quite exciting as well right and so then my next question was can you give me any kind of understanding of you said it's very dense in the in the san francisco bay area maybe la maybe tokyo and other so how far am i from a fiber uh probabilistically at this moment i i don't have a feeling for the density of fibers in the world well well you probably most likely you uh we are talking through a fiber i know that i'm talking for a fiber but probably you too uh but uh the if you walk in the street let's say of san francisco you're probably 10 meters away 20 meters away maximum okay so i really should imagine a pretty dense network of these fibers and in fact because of what you're saying they're going to be most dense in the areas where there's the most people so in a funny way it may be disappointing for some measurements where you need to be out in the middle of the desert but for safety and for things like that it's actually matching population density pretty well would be my guess that's exactly right and that makes a possibility to monitor not only earthquakes one thing that we found out so i should say that thanks to the collaboration of i.t service department at stanford we deployed the first urban dark fiber network five years ago under stanford campus and uh so we had the basically recording understand for campus more or less for five years and now we are recording actually in mello park in palo alto as i said a month ago we recorded the end of the city of san jose and we discovered that there are many more signals some of them we consider the noise maybe in the future can be utterly useful but uh bottom line a city creates a lot of seismic noise and you can use that seismic noise in particular made by uh cows and trucks and buses and trains to image the subsurface continuously down under easily 100 meters so one of the things that you can do is monitor the subsurface not only the earthquake but how the subsurface changes so sinkholes or landslides or other uh potential hazard could be uh monitored in real time and also what's happening above the ground so you see cows going it's already used as a way of monitoring traffic in but you can even think that it may have a role in a driverless transportation of the future and as well above ground you can monitor overpasses bridges and so on are there fibers already going over many bridges or would that have to be additional infrastructure no but the data in san jose we actually ran it was over two bridges one was a overpass over 101 and we could basically using the noise of 101 we could uh measure the standing waves in the overpass and that you can think about that there is a way of very cheaply continuously monitoring bridges and uh uh and see eventually that's really that's really exciting because i know even from other guests on this uh on this show that there's an issue about aging infrastructure in the united states and the idea that we may have already placed in the sensors that we need to figure out which bridges are perhaps bending too much or not bending enough uh in response to normal stresses uh that could be a huge yeah and i would like to put a plug for a collaboration with a a younger faculty in civil engineering that we have we the fiber was passing across a smaller bridge that was owned by the city so it was easier to instrument compared to the overpass that is caltrans uh and uh they searched uh hergo put geophones and ran back and forth from the bridge with an instrumented car we have not analyzed the data yet but there is a potential of uh really this collaboration with civil engineers between geophysicists and civil engineers to have effective and efficient monitoring yeah so this is so tell me i so as i understand your comments the uh the stanford area and then the a little bit of the greater area surrounding stanford because that's your home institution you've already kind of instrumented and you've gotten the fibers that you need um what are you learning about seismic activity i mean i'm sure people are and this is we we consider ourselves a fairly seismic active uh area absolutely um what are you seeing in the data oh we see events that we cannot explain yet so let's first from uh where we need the really to learn more but also we see events that were not in the usgs catalogue because they were too weak and too far away from the instruments that usgs uses to catalog uh events and that is where i do think that has really the opportunity the bigger earthquakes they tend to be well felt and studied uh using uh uh the uh conventional instruments but the smaller shaking local ones that can tell you a lot about the local geology and where potentially faults that you're not aware of uh that's how they really having this dense network may make a a big difference so as we think about deploying this uh more generally uh in beyond the local region it seems to me that there's a and i don't know if this is true there's going to be a uh administrative challenge uh you made it you made a couple of comments about how a a publicly owned bridge was a little bit easier than the caltrain which is a state organization or maybe even federal so is it true that you're going that there's a little bit of an administrative nightmare trying to figure out who to ask permission to have some of those fibers so you can access them for these seismic measurements that's absolutely true and i consider my myself a scientist so i'm not necessarily good about doing that but i'm learning slowly for example the experiment in san jose we got the san jose city attention by the fact that we may be able to create a detailed map of the location of a fiber in a physical space so that will allow them to troubleshooting the problems in their own fiber infrastructure by the city much more effectively than they have been doing with the current technology so for us is not our goal however it might be very much a something that will induce cities to get not only cities but also telecom provider potentially to provide access so it's always a give and take right right this is the future of everything i'm russ altman we'll have more with professor biando biandi next about seismic sensing here on sirius xm welcome back to the future of everything i'm russ altman i'm speaking with professor biando biandi about seismic sensing fiber optics uh and the future so uh you know in our previous discussion we heard about this amazing ability to detect uh uh seismic activity uh and also that it was very high resolution that even at the kilometer level over a 50 kilometer uh uh fiber optic but i didn't ask you how much data are we talking about is this does this count as a lot of data for you and your colleagues or is it just routine no it does count even uh if uh my area do typically we call peter byte of data per survey uh this is still more than what we are accustomed to deal with so a petabyte which just to remind everybody that's a thousand terabytes that's correct um over what period of time or what what what is the scope of that petabyte that it will give you yes also they they experiment that we did with the newest generation of interrogator months ago uh collected about 20 terabytes a day okay and that was for 50 000 sensors so if you want to go to a 50 million uh then you you have to be 20 petabytes and then if you want to or 50 billions that probably that's beyond my lifespan uh then you go to a 50 petabyte sorry exabyte a day so this is huge and this is just only for relatively small array right right and whenever we hear these numbers we know that i'm going to guess that there are young students in your lab applying the techniques of ai and deep networks to try to understand this data i'm just kind of assuming that's right because another of the issues is that we have so much signal so typically is the earthquake signal versus the traffic signal yes traffic signal can be used to monitor traffic to use the seismic waves generated by cars and tracks to image the subsurface but it's nice when we try to detect a week or an earthquake and do that in real time we really need to complement our understanding of the physics that is basically elastic wave propagation with the statistics of this all this kind of news sources of data and signal so my students has built a convolutional neural network running on gpus that uh towards the goal of doing this analysis in real time now if you were lucky the frequencies of the cars and the buses would be very different than the frequencies but are we lucky or is it a very nice because uh they uh typically the weights generated by traffic goes up to 15 20 hertz that is the band of local earthquakes and on the top of that there is just the deformation of the ground that in the order of hertz that is uh clearly earthquake uh bands so they do overlapping frequency but they do have a very different special uh pattern that's his advantage of having a sensor every meter right you take advantage of a special pattern spatial and temporal pattern to distinguish what is you consider signal and what is noise and what basically we are doing you know i'm not surprised about that answer because we live literally right next to a train and we often get confused when they're shaking in the house we sometimes cannot distinguish whether it's a freight train or an earthquake so we have to use our ears because if it's a freight train that's very quiet it's an earthquake and if it's a loud freight train it's a freight train so so i wanted to ask about other applications of this um i know that in your writings you've talked this this might have some implications for building codes and our knowledge of like the strata upon which we're putting major civil infrastructure that's correct as i said using mostly traffic noise but also noise generated by ocean waves they propagated all over in the subsurface we can using this network potentially every meter or every few meters we can build the high resolution maps of the subsurface for the first uh the civil engineering like to have a measurement of 30 meters that is very standard but for a bigger building you probably need to go deeper for the foundation so i don't want to get into that but many people will argue that the problem that the millennium tower in san francisco is really caused by not sufficient uh studies of the soil where they built the foundation so was a geotech problem and that's kind of this is just one example yeah you can potentially have much better understanding of the soil where you're building your private home as well infrastructure and high rises there's another famous example in your home country i believe it's in pisa that's exactly right back in 1200 they didn't do their geoattack fight and it's hard to criticize them it was 800 or 900 years yeah that's right um i know that you wrote a paper about coven 19 and that really popped out into my attention because i might not guess that a seismic person was doing anything related to covert 19 but can you tell me about that work oh sure first i should give credit to my postdoc nate lindsay for doing the work and the uh the basic idea is the traffic monitoring so we monitor traffic we have the fiber along uh send a hello road that connects the same for the hospital all the way up to 280 to uh freeway 280. and we monitor the traffic in and out of a hospital and around 280 before and after they uh first covered the shutdown in march of uh 2020 and we could see clear patterns uh and unfortunately the hospital emergency room uh traffic didn't decrease as much as other parts of the traffic so we basically monitor the traffic in this very local area um i'm staying from house because so this is great so this is a very compelling kind of uh demonstration that the uh social and uh an economic activity of a region is actually showing up as a differential seismic signal that's exactly right and uh in a broader sense let me put a plug for what the uh new school focus on the climate and sustainability at stanford yes one of the innovative part of the school is what they called an accelerator so that is also touches uh what you said about interaction with governments and and the communities i think that is one of the many things that i hope that the new school accelerator will enable is bringing some of this innovative technology of in general sensing of earth beyond seismic uh to uh really to solve uh connect with social and health problems and try to address in a positive um futuristic way of looking at the future way the problem of our society yeah that's very exciting and and in our in our final two minutes um i wanted to ask about earthquake prediction because everybody wants to know about it it seems to be one of the holy grail problems in our facing especially every time we get one of these big earthquakes where many people die or there's a large loss of property how should we think about earthquake prediction and what should be like reasonable expectations of the public in terms of our understanding of earthquakes in the next decade or two or three well i i think that uh probably even in the next few decades will be still a statistical exercise of the statistics will get better and better so i absolutely there is no to my knowledge and i'm not an earthquake seismology to really say an earthquake is going to happen in half an hour from now on that fault and so what i i'm hoping that this kind of new seismic data may enable to have a better statistics and a better statistical forecast of what is going to happen another thing that it can happen is very early sensing of earthquake and we have already a first version of that in california that started one or two years ago in which a lot of lives can be saved even if you do have a real time warning of community and that's again the probably the next generation of that system most likely will have the fibro seismology component uh it's too early to say but uh that's probably will have an impact there well yeah well that's great to hear and uh thank you for listening to the future of everything i'm russ altman if you missed any of this episode listen anytime on demand with the sirius xm app

2021-09-06

Show video