Faster Than Real Time Tsunami Warning with Associated Hazard Uncertainties

Physical size of tsunamigenic earthquakes of the northwestern Pacific

While it is unlikely that man will ever devise a means of preventing tsunamis-more commonly known in the West by the misnomer "tidal waves"-scientists today are constantly adding to our knowledge of their nature, sources, and causes, and, through the development of highly refined technical instrumentation, are making possible early-warning systems that already, even in these early stages, are proving of significant value.In October, 1969, an International Symposium on Tsunamis and Tsunami Research was held at the East-West Center, University of Hawaii. Participants included representatives from Austalia, Canada, Germany, Japan, New Zealand, U.S.S.R., and the U.S.A. Sponsored jointly by the International Union of Geodesy and Geophysics Tsunami Committee and the East-West Center, the major seminars were devoted to (1) Seismic Source and Energy Transfer, (2) Tsunami Instrumentation, (3) Tsunami Propagation and Run-up.This volume contains more than 30 of the papers presented at that Symposium. Recent developments and future plans for the Pacific Tsunami Warning System are included

The Vanuatu Trench hosts tsunamigenic earthquakes exceeding magnitude 7. For these source locations, current tsunami early warning systems in Aotearoa - New Zealand are based on earthquake point-source parameters and as tsunamis propagate, initial forecasts are refined by DART (Deep-ocean Assessment and Reporting of Tsunamis) sea-level analysis. This study, which is part of the R-CET (The Rapid Characterization of Earthquakes and Tsunami) project led by GNS Science, aims to seismologically characterize the spatio-temporal behavior of the regional tsunamigenic ruptures in near real-time. This transition from basic point sources to 4-D rupture propagation could enhance the accuracy of initial tsunami threat maps and hazard response. A new array of broadband seismic stations, designated as the R-CET array, has been strategically deployed in New Zealand to support this analysis for events in the South Pacific.In this proof-of-concept study, we applied a beamforming array seismological technique to analyze the R-CET array recordings from a recent tsunamigenic earthquake (Mw 7.7) in the Vanuatu region, the Loyalty Islands, on May 19, 2023. We use sliding window fk-analysis beamforming, which can simultaneously measure backazimuth and slowness. Incorporating the sliding window, in conjunction with the fk diagrams, helps to observe temporal azimuthal changes, which facilitates tracking the rupture length and direction over time. Preliminary results showed the azimuthal and temporal variation of the rupture is in alignment with post-processing estimates of the finite fault solution for this event reported by USGS. We test the utility of this analysis in tsunami forecasts by comparing threat maps generated from the beamforming source and initial response maps used in real-time response on the day. We further compare our results to actual measured coastal cancellation gauges (tide gauges) and DART observations. We show that this analysis has the potential to improve initial tsunami forecasts prior to the onset of tsunami waves at deep ocean tsunamimeters. We further present the technique as an enticing path to meet UN Ocean Decade tsunami warning goals within our framework of ensemble and time-dependent forecasting.

Quantifying kinetic uncertainty in turbulent combustion simulations using active subspaces

To assist the analysis of tsunami hazards for Qatar coastal areas were using numerical model. By Tsunamis waves created from submarine earthquakes of magnitude of (Mw) 8.6 and 9.0 in Richard scale along the Makran Subduction Zone (MSZ) as well as coastal landslides with soil volume of 1.25 to 2.0 km3 along Iranian coast inside the Arabian Gulf is considered. TUNAMI-N2KISR model (Al-Salem) was applied in this study to predict the tsunami propagation and magnitude of Tsunami induced wave heights. The model adopts to solve shallow water equations describing nonlinear long-wave theory. The model also incorporate tidal effect inside the Arabian Gulf as a tsunami travel time from Makran Subduction to Qatar coastline takes more than 9 hours with the tidal range of about 1.6 m during Spring Tide event. For coastal landslides, tsunami generation was simulated using a two-layer numerical model, developed by solving nonlinear long-wave equations. Two-layer model was used to determine initial wave deformation generated by a landslide case. Then TUNAMI-N2KISR was use to simulate tsunami wave propagation. Tsunami waves from landslide scenario arrived after 2.5 - 3 hr with maximum tsunami amplitudes along coasts of Ras laffan-Qatar were 0.8 to 1.0 m. Incorporation of ocean tide is found to impose some small effect on tsunami amplitude at Qatar coastline and nearby areas for the Mw 9.0 earthquake due to small tidal range in this area. In addition, it is found that the tsunami arrival time has become shorter.

A brief review of the problem of short-term (operative) tsunami forecasting is presented. A method of short-term tsunami forecasting based on the sea level data at remote points is described. The forecast is based only on the seismological information about the coordinates of the earthquake’s epicenter. The application of this method for a retrospective forecast of the 1996 Andreanov tsunami demonstrates a satisfactory coincidence with the predicted and observed tsunami waveforms. The results are compared with the modeling prediction of the same event by other authors. The suggested method of the operative tsunami forecast can serve a basis for the development of a unique programming complex for operative forecasting and can be realized within regional and local tsunami warning services. The method can be used in tsunami warning systems at the stage of their development.

While there is high certainty that chronic coastal hazards like flooding and erosion are increasing due to climate change induced sea‐level rise, there is high uncertainty surrounding the timing, intensity, and location of future hazard impacts. Assessments that quantify these aspects of future hazards are critical for adaptation planning under a changing climate and can reveal new insights into the drivers of coastal hazards. In particular, probabilistic simulations of future hazard impacts can improve these assessments by explicitly quantifying uncertainty and by better simulating dependence structures between the complex multivariate drivers of hazards. In this study, a regional‐scale probabilistic assessment of climate change induced coastal hazards is conducted for the Cascadia region (Northern Washington to Northern California), USA during the 21st century. Three co‐produced hazard proxies for beach safety, erosion, and flooding are quantified to identify areas of high hazard impacts and determine hazard uncertainty under three sea‐level rise scenarios. A novel chronic coastal hazard hotspot indicator is introduced that identifies areas that may experience significant increases in hazard impacts compared to present day conditions. We find that beaches near the California‐Oregon border and in Northern Washington have larger hazard impacts and hazard uncertainty due to their morphologic setting. Erosional hazards, relative to beach safety and coastal flooding, will increase the most in Cascadia during the 21st century under all sea‐level rise scenarios. Finally, we find that hazard uncertainty associated with wave and water level variability exceeds the uncertainty associated with sea‐level rise for most of the 21st century.

Surrogate models are statistical or conceptual approximations for more complex simulation models. In this context, it is crucial to propagate the uncertainty induced by limited simulation budget and surrogate approximation error to predictions, inference, and subsequent decision-relevant quantities. However, quantifying and then propagating the uncertainty of surrogates is usually limited to special analytic cases or is otherwise computationally very expensive. In this paper, we propose a framework enabling a scalable, Bayesian approach to surrogate modeling with thorough uncertainty quantification, propagation, and validation. Specifically, we present three methods for Bayesian inference with surrogate models given measurement data. This is a task where the propagation of surrogate uncertainty is especially relevant, because failing to account for it may lead to biased and/or overconfident estimates of the parameters of interest. We showcase our approach in three detailed case studies for linear and nonlinear real-world modeling scenarios. Uncertainty propagation in surrogate models enables more reliable and safe approximation of expensive simulators and will therefore be useful in various fields of applications.

We propose a high-order stochastic-statistical moment closure model for efficient ensemble prediction of leading-order statistical moments and probability density functions in multiscale complex turbulent systems. The statistical moment equations are closed by a precise calibration of the high-order feedbacks using ensemble solutions of the consistent stochastic equations, suitable for modeling complex phenomena including non-Gaussian statistics and extreme events. To address challenges associated with closely coupled spatiotemporal scales in turbulent states and expensive large ensemble simulation for high-dimensional systems, we introduce efficient computational strategies using the random batch method (RBM). This approach significantly reduces the required ensemble size while accurately capturing essential high-order structures. Only a small batch of small-scale fluctuation modes is used for each time update of the samples, and exact convergence to the full model statistics is ensured through frequent resampling of the batches during time evolution. Furthermore, we develop a reduced-order model to handle systems with really high dimensions by linking the large number of small-scale fluctuation modes to ensemble samples of dominant leading modes. The effectiveness of the proposed models is validated by numerical experiments on the one-layer and two-layer Lorenz '96 systems, which exhibit representative chaotic features and various statistical regimes. The full and reduced-order RBM models demonstrate uniformly high skill in capturing the time evolution of crucial leading-order statistics, non-Gaussian probability distributions, while achieving significantly lower computational cost compared to direct Monte-Carlo approaches. The models provide effective tools for a wide range of real-world applications in prediction, uncertainty quantification, and data assimilation.

Summary Prudent decision making in subsurface assets requires reservoir uncertainty quantification. In a typical uncertainty-quantification study, reservoir models must be updated using the observed response from the reservoir by a process known as history matching. This involves solving an inverse problem, finding reservoir models that produce, under simulation, a similar response to that of the real reservoir. However, this requires multiple expensive multiphase-flow simulations. Thus, uncertainty-quantification studies employ optimization techniques to find acceptable models to be used in prediction. Different optimization algorithms and search strategies are presented in the literature, but they are generally unsatisfactory because of slow convergence to the optimal regions of the global search space, and, more importantly, failure in finding multiple acceptable reservoir models. In this context, a new approach is offered by estimation-of-distribution algorithms (EDAs). EDAs are population-based algorithms that use models to estimate the probability distribution of promising solutions and then generate new candidate solutions. This paper explores the application of EDAs, including univariate and multivariate models. We discuss two histogram-based univariate models and one multivariate model, the Bayesian optimization algorithm (BOA), which employs Bayesian networks for modeling. By considering possible interactions between variables and exploiting explicitly stored knowledge of such interactions, EDAs can accelerate the search process while preserving search diversity. Unlike most existing approaches applied to uncertainty quantification, the Bayesian network allows the BOA to build solutions using flexible rules learned from the models obtained, rather than fixed rules, leading to better solutions and improved convergence. The BOA is naturally suited to finding good solutions in complex high-dimensional spaces, such as those typical in reservoir-uncertainty quantification. We demonstrate the effectiveness of EDA by applying the well-known synthetic PUNQ-S3 case with multiple wells. This allows us to verify the methodology in a well-controlled case. Results show better estimation of uncertainty when compared with some other traditional population-based algorithms.

Response surface approximations (RSA) are often used as inexpensive replacements for computationally expensive computer simulations. Once a RSA has been computed, it is cheap to evaluate this “meta-model” or surrogate, and thus the RSA is often used in a variety of contexts, including optimization and uncertainty quantification. Usually, some method of sampling points over the input domain is used to generate samples of the input variables. These samples are run through the computer simulation. A response surface approximation is then generated based on the sample points. This report presents a study investigating the dependency of the response surface method on the sampling type. The purpose of this study was to address the question: Does a particular RSA type perform better (in terms of a better fit) if a particular sampling method is used? The RSA types examined were kriging, polynomial regression, and multivariate adaptive regression splines (MARS). The sampling types examined were Latin Hypercube, Halton, Hammersley, Centroidal Voronoi Tesselation (CVT), and standard Monte Carlo sampling. The example problems were a 5-dimensional version of Rosenbrock’s function and the Paviani function. RSA of the three response surface types were developed based on the five sampling methods. Performance was compared using ANOVA techniques.

On April 25–26, 1992, a swarm of large earthquakes rattled northern California, injuring 98 people and causing $66 million dollars in damage. The swarm began with a Ms 7.1 thrust earthquake that generated a small tsunami recorded at tide stations in California, Oregon, and Hawaii. This earthquake/tsunami event provided importalt evidence that the Cascadia Subduction Zone is seismically active and that tsunamis are likely from earthquakes occurring along the zone. This new evidence also led the National Oceanic and Atmospheric Administration (NOAA), the U.S. agency responsible for tsunami warnings, to evaluate the state of tsunami preparedness for communities potentially affected.NOAA conducted workshops on tsunami hazard assessment, tsunami warnings, and public education over a 15-month period. Over 50 tsunami scientists, emergency planners, and educators participated and produced technical reports with findings and recommendations. The most significant findings that emerged from these workshops include: 1. Technology exists to produce tsunami inundation maps for emergency preparedness. 2. Technology exists to issue tsunami warnings within 5 minutes for earthquakes occurring along U.S. coastlines. 3. Technology exists to detect tsunamis in the offshore ocean. 4. Tsunami education for local and distant tsunami is deficient for west coast decision makers and residents. A summary of the three workshops is presented in this report, along with 12 recommendations that participants felt would mitigate U.S. tsunami hazards.KeywordsTsunami WaveCoastal CommunityFederal Emergency Management AgencyTsunami HazardTsunami WarningThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

The application of numerical models for flood and inundation modelling has become widespread in the past decades as a result of significant improvements in computational capabilities. Computational approaches to flood forecasting have significant benefits compared to empirical approaches which estimate statistical patterns of hydrological variables from observed data. However, there is still a significant computational cost associated with numerical flood modelling at high spatio-temporal resolutions. This limitation of numerical modelling has led to the development of statistical emulator models, machine learning (ML) models designed to learn the underlying generating process of the numerical model. The data-driven approach to ML involves relying entirely upon a set of training data to inform decisions about model selection and parameterisations. Deep learning models have leveraged data-driven learning methods with improvements in hardware and an increasing abundance of data to obtain breakthroughs in various fields such as computer vision, natural language processing and autonomous driving. In many scientific and engineering problems however, the cost of obtaining data is high and so there is a need for ML models that are able to generalise in the ‘small-data’ regime common to many complex problems. In this study, to overcome extrapolation and over-fitting issues of data-driven emulators, a Physics-Informed Neural Network model is adopted for the emulation of all two-dimensional hydrodynamic models which model fluid according the shallow water equations. This study introduces a novel approach to encoding the conservation of mass into a deep learning model, with additional terms included in the optimisation criterion, acting to regularise the model, avoid over-fitting and produce more physically consistent predictions by the emulator.

The NEAM Tsunami Hazard Model 2018 (NEAMTHM18) is a probabilistic hazard model for tsunamis generated by earthquakes. It covers the coastlines of the North-eastern Atlantic, the Mediterranean, and connected seas (NEAM). NEAMTHM18 was designed as a three-phase project. The first two phases were dedicated to the model development and hazard calculations, following a formalized decision-making process based on a multiple-expert protocol. The third phase was dedicated to documentation and dissemination. The hazard assessment workflow was structured in Steps and Levels. There are four Steps: Step-1) probabilistic earthquake model; Step-2) tsunami generation and modeling in deep water; Step-3) shoaling and inundation; Step-4) hazard aggregation and uncertainty quantification. Each Step includes a different number of Levels. Level-0 always describes the input data; the other Levels describe the intermediate results needed to proceed from one Step to another. Alternative datasets and models were considered in the implementation. The epistemic hazard uncertainty was quantified through an ensemble modeling technique accounting for alternative models’ weights and yielding a distribution of hazard curves represented by the mean and various percentiles. Hazard curves were calculated at 2,343 Points of Interest (POI) distributed at an average spacing of ∼20 km. Precalculated probability maps for five maximum inundation heights (MIH) and hazard intensity maps for five average return periods (ARP) were produced from hazard curves. In the entire NEAM Region, MIHs of several meters are rare but not impossible. Considering a 2% probability of exceedance in 50 years (ARP≈2,475 years), the POIs with MIH >5 m are fewer than 1% and are all in the Mediterranean on Libya, Egypt, Cyprus, and Greece coasts. In the North-East Atlantic, POIs with MIH >3 m are on the coasts of Mauritania and Gulf of Cadiz. Overall, 30% of the POIs have MIH >1 m. NEAMTHM18 results and documentation are available through the TSUMAPS-NEAM project website (http://www.tsumaps-neam.eu/), featuring an interactive web mapper. Although the NEAMTHM18 cannot substitute in-depth analyses at local scales, it represents the first action to start local and more detailed hazard and risk assessments and contributes to designing evacuation maps for tsunami early warning.

An efficient surrogate modelling framework is proposed for full-field predictions of stresses and cracks in composite material microstructures. The framework comprises two sequential convolutional neural networks (CNNs), predicting the elastic stress fields and the local crack maps, respectively. Training and test data are created from high-resolution fracture simulations of randomly generated representative volume elements (RVEs), including geometric variabilities such as fibre volume fraction and porosity. This work shows that the inclusion of a self-attention layer within the network enables the model to capture relevant local and global features, which are important in determining the heterogeneous stress distribution and crack patterns. The performance of the trained CNN models is evaluated with unseen data. The CNN models speed up the full-field predictions by 3 ∼ 4 orders of magnitude compared to the physics-based model. The surrogate model’s accuracy and efficiency are key enables for applications such as multiscale simulation, microstructure optimisation and uncertainty quantification.