Driven by dangerous Santa Ana winds and fueled by dry vegetation, the 2025 Eaton and Palisades wildfires in California caused historic levels of devastation, ultimately becoming the second and third most destructive fires in California history. Burning at the same time and drawing from the same resources, these fires burned a combined total of 16,251 structures. The first several hours of an emerging wildfire are a crucial period for fire officials to assess potential damage and develop a timely and appropriate response. A method to quickly generate accurate estimates of structural damage is essential to providing this crucial rapid response to wildfires. In this paper, we present a machine learning approach for automated assessment of structural damage caused by wildfires. By leveraging multiple data sources in model development (satellite-based building footprints, expert-labeled post-fire damage points, fire perimeters, and aerial thermal imagery) and innovative data processing techniques, the approach can be used to identify various levels of structural damage from just aerial thermal imagery during operational use. The resulting system offers an effective approach for rapid and reliable assessment of burned structures, suitable for operational wildfire damage assessment. Results on the Eaton and Palisades Fires demonstrate the effectiveness of this method and its applicability to real-world scenarios.

AWP-ODC is an open-source dynamic rupture and wave propagation code which solves the 3D velocity-stress wave equation explicitly by a staggered-grid finite-difference method with fourth-order accuracy in space and second-order accuracy in time. The code is memory-bandwidth, with excellent scalability up to full machine scale on CPUs and GPUs, tuned on CLX, with support for generating vector folded finite difference stencils using intrinsic functions. AWP-ODC includes frequency-dependent anelastic attenuation Q(f), small-scale media heterogeneities, support for topography, Drucker-Prager visco-plasticity, and a multi-yield-surface, hysteretic (Iwan) nonlinear model using an overlay concept. Support for a discontinuous mesh is available for increased efficiency. An important application of AWP-ODC is the CyberShake Strain-Green-Tensor (SGT) code used for probabilistic hazard analysis in CA and other regions. Here, we summarize implementation and verification of some of the widely-used capabilities of AWP-ODC, as well as validation against strong motion data and recent applications for future earthquake scenarios. We show for a M7.8 dynamic rupture ShakeOut scenario on the southern San Andreas fault that while simulations with a single yield surface reduces long period ground motion amplitudes by about 25% inside a wave guide in greater Los Angeles, multi-surface Iwan nonlinearity further reduces the values by a factor of two. In addition, we show assembly and calibration of a 3D Community Velocity Model (CVM) for central and southern Chile as well as Peru. The CVM is validated for the 2010 M8.8 Maule, Chile, earthquake up to 5 Hz, and the validated CVM is used for scenario simulations of megathrust scenario events with magnitude up to M9.5 in the Chile-Peru subduction zone for risk assessment. Finally, we show simulations of 0-3 Hz 3D wave propagation for the 2019 Mw 7.1 Ridgecrest earthquake including a data-constrained high-resolution fault-zone model. Our results show that the heterogeneous near-fault low-velocity zone inherent to the fault zone structure significantly perturbs the predicted wave field in the near-source region, in particular by more accurately generating Love waves at its boundaries, in better agreement with observations, including at distances 200+ km in Los Angeles.

The impact of non-linear soil behaviour on seismic hazard in low-to-moderate seismicity areas is often neglected; however, it may become relevant for long return periods. In this study, we used fully non-linear 1-D simulations to estimate the site-specific non-linear soil response in the low seismicity area, using the city of Lucerne in Switzerland as an example. The constitutive model considers the development of pore pressure excess and requires calibration of complex soil models, including the soil dilatancy parameters. In the absence of laboratory measurements, we mainly used the cone penetration test data to estimate the model variables and perform inversion for the dilatancy parameters. Our findings, using Swiss building code-compatible input ground motions, suggest a high probability of strong non-linear behaviour and the possibility of liquefaction at high ground motion levels in the case study area. While the non-linearity observations from strong-motion recordings are not available in Lucerne, the comparison with empirical data from other sites and other methods shows similarity with our predictions. Moreover, we show that the site response modelled is largely influenced by the strong pore pressure effects produced in thin sandy water-saturated layers. In addition, we demonstrate that the variability of the results due to the input motion and the soil parameters is significant, but within reasonable bounds.

Wildland fire modeling tools can ingest high resolution 3D vegetation models as inputs. However, data used to build the surface fuels in these models is often at a 30-meter resolution, which does not necessarily provide sufficient detail for accurate modeling of fires. Terrestrial laser scans are increasingly being used to collect detailed vegetation data that could be integrated with new approaches to fuel and fire modeling, but manual segmentation of scans is not scalable beyond a small number of scans. There is a need to automatically segment these high resolution point clouds as they are collected in the field, such that they may be leveraged by fuel and fire models for wildland fire response and mitigation and other applied climate science. This paper summarizes our early work on a labeling, visualization and machine learning pipeline for detailed segmentation of fuels. Specific contributions are: (1) a labeling approach involving 3 dimensional segmentation of point clouds using a point cloud processing engine; (2) a visualization approach using a computer graphics engine; and (3) early results from a deep learning modeling approach for fuel segmentation by category (live and dead) and size class (1, 10, 100 and 1000 hour fuels).

Long-standing fire suppression policies, global warming and human influence at the urban-wildland interface are fueling a global wildfire crisis. Prescribed burns are increasingly being recognized as an essential procedure to reduce fuel (biomass) overgrowth and mitigate the size and severity of uncontrolled wildfires. Next-generation three-dimensional (3D) fire behavior simulations, which can help land managers to identify risks and improve planning for successful prescribed burns, depend on accurate 3D fuel structure models. We introduce TrueTrees, a workflow that integrates tree-level observations from airborne lidar surveys into FastFuels 3D fuel models. The workflow is optimized and distributed to allow processing of point cloud data for typical burn units within minutes. TrueTrees is implemented into a prototype of BurnPro3D, a user-friendly science-driven decision support platform for prescribed burn planners.

ABSTRACT We have implemented and verified a parallel-series Iwan-type nonlinear model in a 3D fourth-order staggered-grid velocity–stress finite-difference method. The Masing unloading and reloading behavior is simulated by tracking an overlay of concentric von Mises yield surfaces. Lamé parameters and failure stresses pertaining to each surface are calibrated to reproduce the stress–strain backbone curve, which is controlled by the reference strain assigned to a given depth level. The implementation is successfully verified against established codes for 1D and 2D SH-wave benchmarks. The capabilities of the method for large-scale nonlinear earthquake modeling are demonstrated for an Mw 7.8 dynamic rupture ShakeOut scenario on the southern San Andreas fault. Although ShakeOut simulations with a single yield surface reduces long-period ground-motion amplitudes by about 25% inside a waveguide in greater Los Angeles, Iwan nonlinearity further reduces the values by a factor of 2. For example, inside the Whittier Narrows corridor spectral accelerations at a period of 3 s are reduced from 1g in the linear case to about 0.8 in the bilinear case and to 0.3–0.4g in the multisurface Iwan nonlinear case, depending on the choice of reference strain. Normalized shear modulus reductions reach values of up to 50% in the waveguide and up to 75% in the San Bernardino basin at the San Andreas fault. We expect the implementation to be a valuable tool for future nonlinear 3D dynamic rupture and ground-motion simulations in models with coupled source, path, and site effects.

Timely prediction of debris flow probabilities in areas impacted by wildfires is crucial to mitigate public exposure to this hazard during post-fire rainstorms. This paper presents a machine learning approach to amend an existing dataset of post-fire debris flow events with additional features reflecting existing vegetation type and geology, and train traditional and deep learning methods on a randomly selected subset of the data. The developed methods achieve AUC (area under the receiver operational characteristic curve) values of 0.93 (random forest) and 0.92 (neural network) on the test set, representing a significant improvement over a logistic regression model currently used (AUC 0.79). The paper also overviews a distributed, Kubernetesbased big data processing pipeline to efficiently retrieve features in areas impacted by new fires, and deploy the methods for real-time prediction of debris flow hazards.

ABSTRACTWe developed a 3D elastic wave propagation solver that supports topography using staggered curvilinear grids. Our method achieves comparable accuracy to the classical fourth-order staggered grid velocity–stress finite-difference method on a Cartesian grid. We show that the method is provably stable using summation-by-parts operators and weakly imposed boundary conditions via penalty terms. The maximum stable timestep obeys a relationship that depends on the topography-induced grid stretching along the vertical axis. The solutions from the approach are in excellent agreement with verified results for a Gaussian-shaped hill and for a complex topographic model. Compared with a Cartesian grid, the curvilinear grid adds negligible memory requirements, but requires longer simulation times due to smaller timesteps for complex topography. The code shows 94% weak scaling efficiency up to 1014 graphic processing units.

Empirical transfer functions (ETFs) between seismic records observed at the surface and depth represent a powerful tool to estimate site effects for earthquake hazard analysis. However, conventional modeling of site amplification, with assumptions of horizontally polarized shear waves propagating vertically through 1D layered homogeneous media, often poorly predicts the ETFs, particularly, in which large lateral variations of velocity are present. Here, we test whether more accurate site effects can be obtained from theoretical transfer functions (TTFs) extracted from physics-based simulations that naturally incorporate the complex material properties. We select two well-documented downhole sites (the KiK-net site TKCH05 in Japan and the Garner Valley site, Garner Valley Downhole Array, in southern California) for our study. The 3D subsurface geometry at the two sites is estimated by means of the surface topography near the sites and information from the shear-wave profiles obtained from borehole logs. By comparing the TTFs to ETFs at the selected sites, we show how simulations using the calibrated 3D models can significantly improve site amplification estimates as compared to 1D model predictions. The primary reason for this improvement in 3D models is redirection of scattering from vertically propagating to more realistic obliquely propagating waves, which alleviates artificial amplification at nodes in the vertical-incidence response of corresponding 1D approximations, resulting in improvement of site effect estimation. The results demonstrate the importance of reliable calibration of subsurface structure and material properties in site response studies.

ABSTRACT We use deep learning to predict surface-to-borehole Fourier amplification functions (AFs) from discretized shear-wave velocity profiles. Specifically, we train a fully connected neural network and a convolutional neural network using mean AFs observed at ∼600 KiK-net vertical array sites. Compared with predictions based on theoretical SH 1D amplifications, the neural network (NN) results in up to 50% reduction of the mean squared log error between predictions and observations at sites not used for training. In the future, NNs may lead to a purely data-driven prediction of site response that is independent of proxies or simplifying assumptions.

We estimate ground motions in the Pacific Northwest urban areas during M9 subduction scenario earthquakes on the Cascadia megathrust by simulating wave propagation from an ensemble of kinematic source descriptions. Velocities and densities in our computational mesh are defined by integrating the regional Cascadia Community Velocity Model (CVM) v1.6 (Stephenson et al. P-and S-wave velocity models incorporating the Cascadia subduction zone for 3D earthquake ground motion simulations—update for open-file report 2007–1348, US Geological Survey, 2017) including the ocean water layer with a local velocity model of the Georgia basin (Molnar, Predicting earthquake ground shaking due to 1D soil layering and 3D basin structure in SW British Columbia, Canada, 2011), including additional near-surface velocity information. We generate six source realizations, each consisting of a background slip distribution with correlation lengths, rise times and rupture velocities consistent with data from previous megathrust earthquakes (e.g., 2011 M 9 Tohoku or 2010 M 8.8 Maule). We then superimpose M ~ 8 subevents, characterized by short rise times and high stress drops on the background slip model to mimic high-frequency strong ground motion generation areas in the deeper portion of the rupture (Frankel, Bull Seismol Soc Am 107(1):372–386, 2017). The wave propagation is simulated using the discontinuous mesh (DM) version of the AWP finite difference code. We simulate frequencies up to 1.25 Hz, using a spatial discretization of 100 m in the fine grid, resulting in surface grid dimensions of 6540 × 10,728 mesh points. At depths below 8 km, the grid step increases to 300 m. We obtain stable and accurate results for the DM method throughout the simulation time of 7.5 min as verified against a solution obtained with a uniform 100 m grid spacing. Peak ground velocities (PGVs) range between 0.57 and 1.0 m/s in downtown Seattle and between 0.25 and 0.54 m/s in downtown Vancouver, while spectral accelerations at 2 s range between 1.7 and 3.6 m/s2 and 1.0 and 1.3 m/s2, respectively. These long-period ground motions are not significantly reduced if plastic Drucker-Prager yielding in shallow cohesionless sediments is taken into account. Effects of rupture directivity are significant at periods of ~ 10 s, but almost absent at shorter periods. We find that increasing the depth extent of the subducting slab from the truncation at 60 km in the Cascadia CVM version 1.6 to ~ 100 km increases the PGVs by 15% in Seattle and by 40% in Vancouver.

Numerical models of earthquake ground motions are able to capture multi-dimensional wave propagation phenomena at multiple scale lengths, and help to understand surface ground motions as the coupled nonlinear response of source, path and site effects. High-performance computing applications like the AWP-ODC finite difference code, which employs 1000's of GPUs on modern supercomputers to accelerate calculations, have resulted in a significant increase in the resolution of such large-scale 3D wave propagation simulations. The code accounts for plastic yielding using a Drucker-Prager yield condition, and has been used to simulate nonlinear effects in both the fault damage zone and shallow crust. For example, a fully nonlinear simulation of a M 7.7 earthquake on the southern San Andreas fault was recently performed for a maximum frequency of 4 Hz using AWP. Dynamic rupture simulations with inelastic deformation in the fault zone, also performed using AWP, were able to reproduce surface deformation patterns observed after the 1992 M 7.3 Landers earthquake. Unfortunately, the high computational cost still limits such physics-based ground motion prediction to frequencies in the lower band of the spectrum relevant for buildings (0 – 10 Hz). In order to resolve the wavefield across the full spectrum relevant for engineering while keeping computational requirements within reach we have implemented a discontinuous mesh (DM) in AWP-ODC. We test the DM method by simulating the Mw 5.1 La Habra earthquake for frequencies up to 4 Hz, using a grid spacing of 20 m in the fine grid and a minimum shearwave velocity of 500 m/s. Synthetics obtained with the DM are accurately reproducing reference solutions computed using a uniform mesh in the time and frequency domain. 1Dept. of Geological Sciences, San Diego State University, San Diego, CA 92182 (email: droten@mail.sdsu.edu) Roten, D., Olsen K.B., Nie S., Day, S.M., High-frequency nonlinear earthquake simulations on discontinuous finite difference grids. Proceedings of the 11th National Conference in Earthquake Engineering, Earthquake Engineering Research Institute, Los Angeles, CA. 2018. High-Frequency Nonlinear Earthquake Simulations on Discontinuous Finite Difference Grid D. Roten2, K.B. Olsen1, S. Nie1 and S.M. Day1

We describe a set of benchmark exercises that are designed to test if computer codes that simulate dynamic earthquake rupture are working as intended. These types of computer codes are often used to understand how earthquakes operate, and they produce simulation results that include earthquake size, amounts of fault slip, and the patterns of ground shaking and crustal deformation. The benchmark exercises examine a range of features that scientists incorporate in their dynamic earthquake rupture simulations. These include implementations of simple or complex fault geometry, off‐fault rock response to an earthquake, stress conditions, and a variety of formulations for fault friction. Many of the benchmarks were designed to investigate scientific problems at the forefronts of earthquake physics and strong ground motions research. The exercises are freely available on our website for use by the scientific community.

Previous studies have shown that plastic yielding in crustal rocks in the fault zone may impose a physical limit to extreme ground motions. We explore the effects of fault-zone non-linearity on peak ground velocities (PGVs) by simulating a suite of surface-rupturing strike-slip earthquakes in a medium governed by Drucker–Prager plasticity using the AWP-ODC finite-difference code. Our simulations cover magnitudes ranging from 6.5 to 8.0, three different rock strength models, and average stress drops of 3.5 and 7.0 MPa, with a maximum frequency of 1 Hz and a minimum shear-wave velocity of 500 m/s. Friction angles and cohesions in our rock models are based on strength criteria which are frequently used for fractured rock masses in civil and mining engineering. For an average stress drop of 3.5 MPa, plastic yielding reduces near-fault PGVs by 15–30% in pre-fractured, low strength rock, but less than 1% in massive, high-quality rock. These reductions are almost insensitive to magnitude. If the stress drop is doubled, plasticity reduces near-fault PGVs by 38–45% and 5–15% in rocks of low and high strength, respectively. Because non-linearity reduces slip rates and static slip near the surface, plasticity acts in addition to, and may partially be emulated by, a shallow velocity-strengthening layer. The effects of plasticity are exacerbated if a fault damage zone with reduced shear-wave velocities and reduced rock strength is present. In the linear case, fault-zone trapped waves result in higher near-surface peak slip rates and ground velocities compared to simulations without a low-velocity zone. These amplifications are balanced out by fault-zone plasticity if rocks in the damage zone exhibit low-to-moderate strength throughout the depth extent of the low-velocity zone (\(\sim\)5 km). We also perform dynamic non-linear simulations of a high stress drop (8 MPa) M 7.8 earthquake rupturing the southern San Andreas fault along 250 km from Indio to Lake Hughes. Non-linearity in the fault damage zone and in near-surface deposits would reduce peak ground velocities in the Los Angeles basin by 15–50%, depending on the strength of crustal rocks and shallow sediments. These results show that non-linear effects may be relevant even at long periods, in particular in earthquakes with high stress drop and in the presence of a low-velocity fault damage zone.

Kinematic source inversions of major ( M ≥7) strike‐slip earthquakes show that the slip at depth exceeds surface displacements measured in the field, and it has been suggested that this shallow slip deficit (SSD) is caused by distributed plastic deformation near the surface. We perform dynamic rupture simulations of M 7.2–7.4 earthquakes in elastoplastic media and analyze the sensitivity of SSD and off‐fault deformation (OFD) to rock quality parameters. While linear simulations clearly underpredict observed SSD and OFDs, nonlinear simulations for a moderately fractured fault damage zone predict a SSD of 44–53% and OFDs of 39–48%, consistent with the 30–60% SSD and 46 ± 10% (1 σ ) OFD reported for the 1992 M 7.3 Landers earthquake. Both SSD and OFDs are sensitive to the quality of the fractured rock mass inside the fault damage zone, and surface rupture is almost entirely suppressed in poor quality material.

The omission of nonlinear effects in large-scale 3D ground motion estimation, which are particularly challenging due to memory and scalability issues, can result in costly misguidance for structural design in earthquake-prone regions. We have implemented nonlinearity using a Drucker-Prager yield condition in AWP-ODC and further optimized the CUDA kernels to more efficiently utilize the GPU's memory bandwidth. The application has resulted in a significant increase in the model region and accuracy for state-of-the-art earthquake simulations in a realistic earth structure, which are now able to resolve the wavefield at frequencies relevant for the most vulnerable buildings (> 1 Hz) while maintaining the scalability and efficiency of the method. We successfully run the code on 4,200 Kepler K20X GPUs on NCSA Blue Waters and OLCF Titan to simulate a M 7.7 earthquake on the southern San Andreas fault with a spatial resolution of 25 m for frequencies up to 4 Hz.