Estimation of Hydraulic Conditions of Tsunami from Deposits: Inverse
Model using Deep-Learning Neural Network
Abstract
Tsunami deposits provide information for estimating the magnitude and
flow conditions of paleo-tsunamis, and the inverse model has potential
for predicting hydraulic conditions of tsunamis from their deposits.
Majority of the previously proposed models are based on oversimplified
assumptions and share limitations in applicability to original tsunami
hydrodynamics in transportation and deposition settings. We present a
new inverse model that serves as a modified version of the previously
proposed model FITTNUSS, which incorporates nonuniform and unsteady
transport of suspended sediment and turbulent mixing. The present model
uses the deep neural network (DNN) for the inversion method. In this
method, forward model calculation was repeated at random initial flow
conditions to produce artificial training data sets that represent
depositional characteristics such as thickness and grain size
distribution. Subsequently, the DNN was trained for establishing the
general inverse model based on artificial data sets derived from the
forward model. Tests conducted using independent artificial data sets
indicated that the established DNN can reconstruct the original flow
conditions from the characteristics of the deposits. Finally, the model
was applied to a data set of 2011 Tohoku-Oki tsunami deposits. Jackknife
resampling was applied for estimating the precision of the result. The
estimated results of the flow velocity and maximum flow depth were
approximately 5.4±0.140 m/s and 4.11±0.152 m, respectively after the
uncertainty analysis. The DNN showed promising results for
reconstruction of the event from natural data set, which would help in
estimating the hydraulic conditions of paleo-tsunamis based on realistic
settings of tsunami deposits.