Iyan E. Mulia

Tsunami data assimilation of Cascadia seafloor pressure gauge records from the 2012 Haida Gwaii earthquake

We use tsunami waveforms recorded on a dense array of seafloor pressure gauges offshore Oregon and California from the 2012 Haida Gwaii, Canada, earthquake to simulate the performance of two different real-time tsunami-forecasting methods. In the first method, the tsunami source is first estimated by inversion of recorded tsunami waveforms. In the second method, the array data are assimilated to reproduce tsunami wavefields. These estimates can be used for forecasting tsunami on the coast. The dense seafloor array provides critical data for both methods to produce timeliness (>30 min lead time) and accuracy in both timing and amplitude (>94% confidence) tsunami forecasts. Real-time tsunami data on dense arrays and data assimilation can be tested as a possible new generation tsunami warning system.

Initial tsunami source estimation by inversion with an intelligent selection of model parameters and time delays

We propose a method for accurately estimating the initial tsunami source. Our technique is independent of the earthquake parameters, because we only use recorded tsunami waveforms and an auxiliary basis function, instead of a fault model. We first use the measured waveforms to roughly identify the source area using backward propagated travel times, and then infer the initial sea surface deformation through inversion analysis. A computational intelligence approach based on a genetic algorithm combined with a pattern search was used to select appropriate least squares model parameters and time delays. The proposed method significantly reduced the number of parameters and suppressed the negative effect of regularization schemes that decreased the plausibility of the model. Furthermore, the stochastic approach for deriving the time delays is a more flexible strategy for simulating actual phenomena that occur in nature. The selected parameters and time delays increased the accuracy, and the models ability to reveal the underlying physics associated with the tsunami-generating processes. In this paper, we applied the method to the 2011 Tohoku-Oki tsunami event and examined its effectiveness by comparing the results to those using the conventional method.

Estimate of tsunami source using optimized unit sources and including dispersion effects during tsunami propagation: The 2012 Haida Gwaii earthquake

We apply a genetic algorithm (GA) to find the optimized unit sources using dispersive tsunami synthetics to estimate the tsunami source of the 2012 Haida Gwaii earthquake. The optimal number and distribution of unit sources gives the sea surface elevation similar to that from our previous slip distribution on a fault using tsunami data, but different from that using seismic data. The difference is possibly due to submarine mass failure in the source region. Dispersion effects during tsunami propagation reduce the maximum amplitudes by up to 20% of conventional linear long wave propagation model. Dispersion effects also increase tsunami travel time by approximately 1 min per 1,300 km on average. The dispersion effects on amplitudes depend on the azimuth from the tsunami source reflecting the directivity of tsunami source, while the effects on travel times depend only on the distance from the source.

Optimum Sea Surface Displacement and Fault Slip Distribution of the 2017 Tehuantepec Earthquake (Mw 8.2) in Mexico Estimated From Tsunami Waveforms

17 The 2017 Tehuantepec earthquake (Mw 8.2) was the first great normal fault event 18 ever instrumentally recorded to occur in the Middle America Trench. The earthquake 19 generated a tsunami with an amplitude of 1.8 m (height=3.5 m) in Puerto Chiapas, Mexico. 20 Tsunami waveforms recorded at coastal tide gauges and offshore buoy stations were used to 21 estimate the optimum sea surface displacement without assuming any fault. Our optimum sea 22 surface displacement model indicated that the maximum uplift of 0.5 m is located near the 23 trench and the maximum subsidence of 0.8 m on the coastal side near the epicenter. We then 24 estimated the fault slip distribution that can best explain the optimum sea surface 25 displacement assuming ten different fault geometries. The best model suggests that a 26 compact region of large slip (3 – 6 m) extends from a depth of 30 km to 90 km, centered at a 27 depth of 60 km. 28 29

Retrieval of missing values in water temperature series using a data-driven model

A measurement buoy with attached sensors has been deployed at our study area to monitor hydrodynamics, water properties, and water quality conditions. High-resolution temporal data have been collected and streamed into an online system that is accessible in nearly real-time. However, in certain circumstances the sensors may fail to provide continuous and high quality data. This results in gaps or corrupted values. The aim of this study was to reconstruct the faulty values. This paper proposes a method based on a data-driven model, using an Artificial Neural Network combined with a Genetic Algorithm to generate a synthetic data series. The generated data can be used as a patch for the incomplete measured data. Additional improvements were achieved by removing seasonal patterns from the original time series using a wavelet decomposition prior to the data-driven model training process. Comparisons with a standard missing-data imputation method using the Kohonen self-organizing map were made to further asses the performance of the proposed data-driven model. The algorithm was applied to water temperature data, but the same approach is applicable to other parameters of interest.

Optimal Design for Placements of Tsunami Observing Systems to Accurately Characterize the Inducing Earthquake

Recently, there are numerous tsunami observation networks deployed in several major tsunamigenic regions. However, guidance on where to optimally place the measurement devices is limited. This study presents a methodological approach to select strategic observation locations for the purpose of tsunami source characterizations, particularly in terms of the fault slip distribution. Initially, we identify favorable locations and determine the initial number of observations. These locations are selected based on extrema of empirical orthogonal function (EOF) spatial modes. To further improve the accuracy, we apply an optimization algorithm called a mesh adaptive direct search to remove redundant measurement locations from the EOF-generated points. We test the proposed approach using multiple hypothetical tsunami sources around the Nankai Trough, Japan. The results suggest that the optimized observation points can produce more accurate fault slip estimates with considerably less number of observations compared to the existing tsunami observation networks.