HOST: The last lightning talk is from Diego Arcas of the Tsunami Group who will tell you about their research.

DIEGO ARCAS: All right. Thank you. So, yeah my name is Diego Arcas. I'm going to be telling you about the work that we do at the NOAA Center for Tsunami Research here at PMEL. So at NCTR where are we
focus on methods and tools to forecast tsunamis and protect life and property. As David mentioned earlier, we do that, we do short-term forecasting and long-term forecasts. But we really put an emphasis on short-term forecasting, which is predicting the impact of a tsunami after it has been generated, before it makes landfall so that we can warn the population. 

You know actually has a natural responsibility to issue tsunami warnings. And so we work in support of the Weather Service offices.

So to tell you about the work that we've been doing in the last five years, let me tell you where we were five years ago. So five years ago we had developed a tsunami forecast system, we have developed and delivered to the Tsunami Warning Center that forecast system, that combined tsunami detection systems developed here at PMEL, with numerical methods, so that tsunami inundation forecasts could be issued for at-risk communities.

And so what you see in the image here, is the deep water forecast that the system issued during the 2011 Japan event. The [INAUDIBLE] and maximuml water elevation reached the Pacific Ocean, by the way. The back and front angles here show the locations of the Dart Buoys resistance that we had in place at the time. And, the right panel here shows the predicted inundation in Kahului, Hawaii. That was the location, the US location, that experienced the highest [INAUDIBLE] inundation during [INAUDIBLE].

So, we were, over the last 15 years, we had a number of smaller tsunami events, that have allowed us to test the performance of the system. And we've been in general pretty satisfied with how the system performed, particularly in the intermediate and far field, and so, we go into [INAUDIBLE] far field areas that are approximately more than 1 1/2 hours tsunami travel time away from the earthquake generation
area of the tsunami.

For areas that were closer to the generation area, we were struggling to be able to produce a flooding forecast in time before the tsunami arrived at the coast. And so in the last five years, this is what we call the near-field problem. and in the last five years, we've actually focused on trying to minimize or reduce the near-field problem in the forecast. 

And, we have done that by incorporating three new technologies to the forecast that system. The first technology is the fourth, the fourth generation of Dart buoys, Dart 4Gs, that were developed here in the lab. And they minimize the latency doing just observations. The second technology is the
use of real-time geodetic GNSS systems, that these systems are used by, are used by seismologists to make a very rapid estimation of the seismic deformation. 

And then the last technology that we have incorporated is the high performance computing accelerators in the form of graphics processing units to run our numerical models and reduce simulation time. I'm going to tell you a little bit more in detail about each one of these technologies. So, the fourth generation of Dart buoys, I mentioned that they reduce the latency in just two tsunami observations.

What you see here, is the location of the Dart buoys and the proximity [INAUDIBLE] in the 2011 Japan event. And the black time series here is the actual signal that was recorded by the Dart. You can see that there is some signal here, which is actually seismic pulse and then the hydrodynamic tsunami signal is this one circle in red there. So, seismic waves travel a different much faster than hydrodynamic waves. 

So, by the time the hydrodynamic signal arrives at the Dart buoy here, the seismic waves have already gone by. What we have here is a hydrodynamic signal that's pain of any seismic noise, which is the way we want to it. So the problem with this signal, is that it limits how close we can deploy the Dart to the source. 

If we were to move this Dart closer to the source, there wouldn't be enough time for the seismic and the hydrodynamic signals to separate completely, and when we get this signal they will to be overlapped. And so, this is the fourth generation of Dart buoys resolved. 

So they can actually, they have a much higher sampling frequency, together with an electronic filter, that even if the two signals, the hydrodynamic and the seismic, are overlapping, they can filter out the seismic noise and report the hydrodynamic signal, which allows us to place the Dart buoy much closer to the source and reduce latency in detection time. 

So what you see here on this panel, is actually that green triangles is the current location of the previous Dart 2 systems of the West Coast of the United States and the red triangles here is an array of Dart 4Gs like we have proposing in terms of using the detection of a Cascadia event in less than five minutes. So the other technology that we have been using is the real-time geodetic GPS system. So these are-- GPS is very similar to what you have in your car but much more precise. 

And so they can record the ground displacement during an earthquake. What you see on the left is the horizontal displacement of the ground experienced in Japan during the 2011 event, and then the right panel shows the vertical displacement that was at the ground underwent during the event. So, it turns out that seismologists can look at these vectors and deformation fields, the vector fields, and they can very rapidly tell us the type of mechanism that the earthquake has generated. 

So if it 's a strikes-slip mechanism that's a very non-tsunamigenic. It could be a thrust fault mechanism that's a much more efficient tsunamigenic event. And but not only that, in less than five minutes they can give us what we call a finaite fault solution, which is, a distribution of deformation along the trench under the ocean, which is information that we need to put into our numerical models in order to make a forecast for this month. 

So, the last technology that I wanted to show you is the use of high performance computing accelerators. So, the animation that is running on the left panel, it's running very slow but it's accelerated actually six times so that we can fit it into space, is actually the current parallelized CPU fastest computation that we can perform of our simulation in the Pacific Ocean. 

The right panel is the current GPU computational speed that we have proportional in time to the one in CPU, so it runs about 20 times faster. So, it allows us to run computations, what we used to take us two minutes to compute, we can do it now in seconds and this technology has already been delivered to the Tsunami Warning Centers. 

The Dart 4G technology we have several prototypes in the water already and the geodetic GPS systems, we have a [INAUDIBLE] with NASA to deliver the technology to the Tsunami Warning Centers in a couple of years. That's all I have, thank you.