我们能实时预测海啸吗？.pdf

1、Can We Predict Tsunamis in Real Time?Omar GhattasOden Institute for Computational Engineering&SciencesDept of Mechanical EngineeringDept of Earth&Planetary Sciences(by courtesy)The University of Texas at Austin1Stefan Henneking Sreeram Venkat2025 Gordon Bell Prize team2Cascadia subduction zone3Last

2、major earthquake and orphan tsunami:January 26,1700Paleoseismic evidence indicates a 250500 years recurrence intervalImage credit:Rob DeGraff/Flickr,20084Subduction earthquake tsunami generationVideo credit:Alaska Earthquake Center5Cascadia subduction zone tsunami(USGS model)Image credit:Washington

3、State Dept.of Natural ResourcesProposed sensor network6Image credit:D.Schmidt et al.,2019Current forecasting models rely on shallow water equations for tsunami propagationEfficient in the far-field;does not make use of near-field pressure transients from hydroacoustic wavesOur digital twin approachU

4、se near-field pressure observations to infer the seafloor motion and forward predict the tsunami propagationEmploy high-fidelity,full-physics model(coupled 3D acousticgravity wave)Quantify uncertainty via Bayesian inferenceSolve inverse problem in real time(order of seconds)7Digital twin for tsunami

5、 forecastingAcoustic wave equation in mixed velocity-pressure form+=0,1+=0,in(0,T)Boundary conditionsSea surface(coupling with surface gravity wave)=,/=,on!#$%&(0,T)Seafloor velocity(boundary source)=on()*)+(0,T)Lateral absorbing boundary(outgoing waves)=/,on%(!)#(0,T)Homogeneous initial conditionsI

6、mplemented with MFEM using high-order finite elements and RK4 time-stepping8Acousticgravity wave propagation forward modelBathymetry-adapted meshing910Given:Observations)(!of acoustic pressure at sensor locationsLinear parameter-to-observable map:,(,)governed by forward acousticgravity wave propagat