The term "extreme event" is frequently applied in policy, media, and colloquial settings. The Pipeline and Hazardous Materials Safety Administration (PHMSA) Mega Rule presents requirements for natural gas transmission pipeline operators to inspect their pipelines following an "extreme weather event" or "natural disaster." However, these terms are poorly defined, and stakeholders have expressed concern that identifying the onset/cessation of an extreme event is unclear, muddying compliance. Here, we present a framework and thresholds for defining an "extreme event" as major streamflow, rainfall, or an earthquake that produces large-scale geomorphic change, rather than "associated hazards" like landslides and channel scour that directly produce ground deformation. Thresholds for streamflow and rainfall are defined in terms of frequency, where the 1% exceedance storm or flood (i.e., the 100-year event) is considered "extreme." Earthquakes are classified as "extreme" with absolute ground shaking thresholds over which the probability of pipeline damage is considered non-negligible. The definitions of "extreme event" outlined in this paper were formulated within the context of pipeline integrity management but are potentially applicable to other buried or aboveground infrastructure.
Climate change is spurring the transition from power generated by fossil fuels to power generated by wind, solar, and nuclear systems. These developments are often constructed at suboptimal sites that face geotechnical and geologic hazards requiring mitigation for operational resiliency. This paper presents the results of an extensive ground improvement program applied to a 41-turbine wind power project in the Mississippi River Delta. The site is subject to significant seismic demand from the nearby New Madrid Seismic Zone (NMSZ) and contains loose alluvial soil that is susceptible to soil liquefaction when loaded to the design Peak Ground Accelerations (PGA) of up to 0.44g. The ground improvement system was required to mitigate liquefaction-induced settlements of up to 5 in., which would render the turbines inoperable. Further, the design seismically induced shear stresses would result in significant soil shear strength reduction and lead to bearing instability. This paper discusses the design methods used to provide a robust and economical solution and summarizes the results of the post-installation cone penetration test (CPT) soundings used to validate performance. This paper is of particular importance because it provides a design framework for the effective treatment of Central United States soil liquefaction when loaded to NMSZ motions, an effort that resulted in greatly improved geotechnical performance and resiliency.
Plastics have become an essential part of human life; however, the increased dependence on single-use disposable plastic items has led to the ubiquity of microplastics (MPs) in our environment. As plastics in the environment degrade into MPs, they can spread through various means, entering the ground and causing implications for the health and well-being of the surrounding environment as well as for the properties of the soil in which the contaminants can reside. A current gap exists in the geotechnical engineering community's understanding of the effects of MPs on the dynamic properties of soils. The influence of MPs on soil behavior under dynamic loading is assessed through a series of constant volume strain-controlled cyclic simple shear tests in conjunction with bender elements (BE). This study examines the behavior of medium-dense specimens of clean Ottawa sand and Ottawa sand mixed with 10% MPs by dry mass. The MPs utilized in this investigation are composed of polyethylene terephthalate (PET) with a particle diameter range of 50 to 100 microns. Equivalent pore pressure ratio curves vs. increasing cyclic strain percentages were developed for both clean Ottawa sand and contaminated Ottawa sand. Monotonic simple shear tests are performed on the liquefied specimens following cyclic testing. The post-liquefaction stress-strain curves of clean and contaminated sand are also developed and compared. The results determined from this study improve the assessment of the liquefaction potential of highly contaminated soils in seismic-prone areas such as dense urban communities, which can reduce the potential loss of lives and resources.
Climate change poses unprecedented challenges to railway infrastructure through increased frequency of extreme geotechnical events and intense rainfall patterns, necessitating enhanced monitoring capabilities for ballast stability. This research enhances railway resilience by integrating Electrical Resistivity Imaging (ERI) with open-source Internet of Things (IoT) sensors to complement existing ballast monitoring systems. The synergy between IoT technology and ERI creates new opportunities for continuous assessment of water accumulation, subgrade deformation, and potential washout conditions in railway ballast structures. While ERI applications in ballast monitoring have been limited by contact resistance challenges, our cost-effective open-source sensor array design addresses these constraints while providing continuous monitoring capabilities across varying geotechnical conditions. The system creates a comprehensive digital twin of ballast moisture conditions and subsurface stability, delivering real-time insights into water retention patterns, ground movement, and potential ponding locations that precede washout failures. Field implementations demonstrated the system's effectiveness in identifying critical water accumulation zones and subsurface anomalies under extreme weather and geotechnical events, enabling proactive maintenance planning. This climate-adaptive approach enhances maintenance strategies by providing continuous data streams that complement routine inspections, strengthening infrastructure resilience against both hydrological and geotechnical challenges. The integration of IoT with ERI expands the toolkit available for railway asset management, offering an economically viable solution for continuous ballast monitoring under diverse ground conditions. This methodology provides railway operators with additional tools for climate adaptation, supporting informed decision-making for maintenance interventions and contributing to more resilient and sustainable rail infrastructure systems.
Internal erosion in levees is a primary embankment failure mechanism that can sometimes be caused by the presence of conduits designed to support water drainage and mitigate ponding or flooding. According to the US Army Corps of Engineers (USACE), failures often arise due to erosion along a pipe, pressurized pipe leakage, erosion into a pipe, seepage at the pipe outlet, or due to ponding. Current internal erosion inspection methods rely on closed-circuit television (CCTV) and sonar, which require site-specific equipment and travel. As a proactive alternative, ongoing efforts to develop a culvert inspection system that uses AI/ML on internal images of the pipes are promising, but more data, especially data that showcases indicators of failure, is needed to train useful algorithms. The USACE and its stakeholders are using physical modeling of culvert failure modes to identify these predominant indicators of collapse. This study investigates the internal erosion into a culvert failure mode. Using centrifuge modeling to simulate complex erosion behaviors, four tests examining the failure mode at different locations within the levee are discussed. The methodology combines visual symmetry modeling and image edge detection to show how void location affects erosion channel progression and surface propagation. Findings from these tests provide critical insights into erosion rates, facilitating improved detection and mitigation strategies and contributing to embankment safety across the US.
Vegetation, particularly plant root systems, offers an environmentally sustainable solution to improve the stability of slopes by enhancing the mechanical stabilization of soil through increased shear strength, in addition to ecological benefits. However, predicting the behavior of root-reinforced soils remains challenging due to the complex variability in root characteristics (e.g., root diameter, length, orientation, and root content within the soil). To address these challenges, this study takes a multi-step approach. First, it develops a comprehensive database for root-reinforced soil characteristics based on experimental and empirical data. The data analysis shows that root area ratio (RAR) is the parameter most influencing shear strength increase, quantified in terms of additional root cohesion (c(r)). Then, a shallow neural network (SNN) model is trained using the database to predict additional root cohesion (c(r)), achieving an R-2 = 0.72 on the testing data set. Finally, the predicted additional root cohesion is integrated into a finite element model analysis comprising 114 models with different slope configurations subjected to various rainfall conditions. The study reveals that root-reinforced soils significantly enhance slope stability. The effect of RAR on slope stability is presented in terms of the improvement of the Factor of Safety in two summary charts for different slope geometries and rainfall conditions.
Unmanned aerial vehicle (UAV) surveying has emerged as a promising technique for measuring surface topography of visible ground surfaces. Extreme events loading in low-lying coastal environments have the potential to either significantly erode existing soil or to deposit material that has been eroded from other locations. As such, extreme events can potentially induce significant morphological change in low-lying coastal areas. The use of accurate remote surveying tools like drone mapping both before and after extreme events offers significant potential advantages for tracking permanent morphological changes in a fast and cost-effective manner; this approach can consequently be very useful within an asset management framework. When used for this application, the accuracy level of drone mapping becomes critically important, and operator experience indicates that measurement accuracy can be heavily tied to the approach used within the drone surveying framework for georeferencing. Using control points to calibrate the photos captured by drones is a powerful but not well-optimized technique for georeferencing. Factors such as the number and arrangement of control points can potentially play a significant role in determining the accuracy of a drone survey. This study surveyed a relatively small area of 50 by 50 m(2) that contained 25 evenly distributed points plus 5 additional random points. An aerial drone mapping survey of these checkpoints was performed, with confirmation surveying of the points' locations being performed using GNSS-RTK surveying measurements. The ITwin Modeler software package was used to construct three-dimensional models from the recorded drone data. Blind location assessment of the five randomly placed checkpoints was performed, and the measured results were compared to the GNSS-RTK measurements. Two-dimensional evaluation results show that square patterns with double the number of control points are more precise and accurate, especially at higher frontal overlaps in the drone imagery stereographic process. Furthermore, the evaluation of two triangular patterns showed that the pattern with control points in the corners is more precise than the other patterns. Three-dimensional evaluation results showed that the corner points also play a significant role in the precision of the checkpoints. Overall, the configuration of control points was shown to be as important as the frontal overlap in the image acquisition process with respect to the overall accuracy of the survey.
Geofencing and targeted online advertising have become standard tools in day-to-day life. However, there is an opportunity to expand these tools further to improve localized risk communication in communities vulnerable to extreme events. This paper investigates how these technologies can support emergency managers and local emergency service agencies in delivering timely and accurate information to at-risk populations to enhance preparedness, response, and recovery. Geofencing requires setting virtual boundaries around hazard areas, allowing deploying agencies to send real-time alerts and guidance to individuals' devices when they enter specific areas. Targeted online ads complement these efforts by raising awareness of seasonal hazards, promoting actionable steps, and providing preparedness resources. Utilizing case studies from communities affected by recent extreme events, geo-fenced messaging, and ad campaigns can effectively reach diverse demographics (local or transient).
Throughout the United States, storms and floods cause widespread damage, disruption, hardship, and economic burdens when transportation systems are disrupted. Extreme rainfall threatens infrastructure in riverine environments through mechanisms such as flooding, landslides, and erosion. Both intense short-duration and long-duration rainfall events are especially challenging in mountainous terrain where road and rail corridors often encroach on rivers and concentrate floodwater flow and increase velocity. Both historically and recently, extreme flood events throughout the nation have resulted in dramatic landscape changes with consequential damage to infrastructure, including bridges, rail and roadway embankments, utilities, as well as residences, businesses, and other facilities. Mountainous areas, because of the topography, often lack redundant networks, increasing the need for resilient transportation corridors to ensure the ability to rapidly respond and recover after flood events. Transportation agencies are exploring ways to improve resilience, identify risk, prioritize investments, and systematically explore ways to lessen the severity of impacts or shorten recovery times to restore transportation corridors to serviceable conditions. While technical strategies to reduce risk exist, they are often not implemented in a systematic, and well-funded, way. These slopes are not explicitly included in transportation asset management plans, so State Departments of Transportation (DOTs) have a wide range of approaches to handling their monitoring and repair. Multidisciplinary efforts will be essential to develop vulnerability assessments and integrated solutions addressing both complex technical challenges and asset management planning and funding through risk-based approaches.
Landslide susceptibility assessment (LSA) is essential for identifying hazard-prone regions. While machine learning (ML) has been widely used in LSA, conventional ML models often overlook underlying physical processes, limiting their robustness, especially in data-scarce or geologically complex areas. This study proposes a physics-informed machine learning (PIML) framework that integrates landslide physics and failure probabilities to enhance susceptibility predictions. The framework was applied to a 2013 rainfall-triggered landslide event in Niang-niang-pa, Gansu, China. The Simplified Transient Infiltration Model (PRL-STIM) was used to calculate the factor of safety (FoS), while the probability of failure (PoF) was derived using the first-order reliability method (FORM). These values guided ML models to ensure scientifically consistent predictions. Spatial cross-validation was employed to assess model generalizability. Results show that while the baseline ML model achieved an AUC of 0.68 on unseen regions, it exhibited significant physical inconsistencies. In contrast, the PIML model demonstrated superior physical consistency and improved generalization by 13% (AUC = 0.77). The proposed PIML framework provides a more reliable solution for LSA than either purely data-driven or purely physics-based approaches.
This paper examines the application of mechanics-guided machine learning (MGML) technique to model seismic energy dissipation of rocking foundations. Numerical simulations are carried out using mechanics-based models available in OpenSees to model the behavior of rocking structure-foundation-soil systems during earthquake loading. The output of OpenSees simulations for normalized seismic energy dissipation in soil (NED) is fed as an additional input feature to machine learning (ML) algorithms to develop MGML models. Shapley Additive Explanations (SHAP values) are used to decipher and interpret MGML model predictions. It is found that the MGML models improve the accuracy of predictions of NED by as much as 27% when compared to their purely data-driven ML model counterparts. The inclusion of mechanics in MGML models ensures that they not only extract information from experimental data but also learn the mechanics of the problem and produce results that are consistent with the domain knowledge.
This study proposes a model that constructs a hypothetical railway network to identify impassable railway sections using seismic fragility functions for each structure and predicts freight volume loss caused by earthquakes through a freight volume estimation model. The proportion of passengers boarding and alighting at each station within the railway network is calculated using an OD (Origin-Destination) Matrix; this allows for quantitatively evaluating the freight volume at each station based on population distribution. In addition, predicting freight transportation loss in the railway network during natural disasters such as earthquakes serves as a key indicator for determining the prioritization of network restoration. To evaluate the applicability of the proposed model, a railway network was constructed based on a segment of a subway line in South Korea, and freight volume loss was estimated by using a hypothetical earthquake scenario. When the probability of damage exceedance for the extensive damage state in the developed fragility function exceeded 30%, it was determined that freight movement through the corresponding section of the proposed railway model was unfeasible, resulting in a freight volume loss of approximately 58.42% compared to the original freight volume. Based on the results, this study is expected to contribute to the development of a more efficient model capable of quantitatively assessing freight volume loss in the railway network and evaluating socioeconomic damage.
The geology of the State of Qatar comprises carbonate and evaporitic rocks, with abundant karst features occurring in a manner that is hard to predict. This gives rise to a geohazard risk during land development or infrastructure construction, and the government of Qatar has identified that existing geodatabases and geological and hydrogeological maps require improvement to mitigate this risk. Dedicated tools are needed to inform the hydrogeological conditions across the country and related geohazards such as subsidence, the presence of karst, or flooding caused by rising groundwater. Therefore, the Ministry of Municipality (MM) launched the Qatar Hydrogeological Assessment Project with the aim of developing a national set of hydrogeological models, geohazard maps, guidelines, and related products, with a focus on major existing urban areas and future development areas. Thorough assessment of karst geohazards is being performed, including a karst features inventory, geologic study of karstification, and the mapping of karst geohazards using a multicriteria approach. Based on this assessment, karst geohazard risk mitigation and management plans will be developed. The geohazards and risk maps, as well as the guidelines and operational recommendations, will be made accessible through a web portal. In parallel, an educational program on karst and geohazards is under development for different stakeholders: the general public, the technical community, and decision-makers. This ongoing work confirms the widespread presence of karst features across the Qatar peninsula and the occurrence of caves formed by dissolution in the Umm Bab carbonate and cave collapse features of deeper origin. The type of caves and the spatial variations in the subsurface geology are identified as important criteria for the ongoing karst subsidence geohazard assessment.
The transition from ergodic to non-ergodic ground motion models (GMMs) accounts for spatially varying systematic source, site, and path effects. This study evaluates different strategies for estimating non-ergodic terms and their interactions using a robust Turkish data set. Results indicate that the sequence of systematic effect estimation significantly influences tradeoffs among non-ergodic terms, affecting mean and correlation length estimates but not the final non-ergodic standard deviation. To address these trade-offs, an iterative approach is proposed to identify a distance threshold where path effects minimally impact other systematic effects, improving site and source effect estimation. The findings emphasize that non-ergodic GMM development extends beyond statistical inference, requiring careful treatment of systematic effects.
Coastal cities are increasingly vulnerable to climate change impacts such as sea-level rise, storm surges, and extreme weather events. In the Netherlands, dune systems serve as critical coastal defense structures. This study evaluates the geomorphological and ecological performance of foredune systems in Katwijk by integrating high-resolution GNSS RTK surveys, airborne LiDAR, drone-based multispectral imagery, and historical elevation data sets. All spatial data were standardized to RD New and EPSG 4289, and analyses were conducted using a 0.5 m x 0.5 m fishnet grid. NDVI thresholds were applied to classify vegetation health, with values >= 0.50 indicating healthy vegetative cover and <=-1 indicating no vegetation. Getis-Ord Gi* hotspot analysis was independently applied to elevation change and NDVI rasters, and spatial overlays were used to assess statistically significant co-occurrence. Results show that the engineered dike-in-dune system, located within a built-up area, exhibited overlapping hotspots of sediment accretion (up to +0.5m/year) and healthy vegetation of +0.5 across more than 60% of the area, with >99% confidence. Proximity analysis identified 15 land use features within 25 m of vegetated zones, including infrastructure such as paved paths, beach houses, and lighting. Despite these pressures, the system demonstrated high geomorphic and ecological resilience. In contrast, the adjacent natural dune, bordered primarily by cropland and dense vegetation, showed over 70% cold spots of concurrent elevation loss and NDVI decline, suggesting degradation linked to reduced sediment supply and lack of stabilization. These findings underscore a strong spatial correlation between vegetation health and geomorphic stability, with marram grass thriving in zones of sustained sediment deposition. The study demonstrates that engineered nature-based systems can effectively support both dune formation and ecological resilience even in urbanized contexts, offering spatially explicit, statistically validated insights for adaptive coastal management in a changing climate.
This study evaluates a three-dimensional underground station structure consisting of six basement levels to determine the maximum evacuation time using the Dijkstra algorithm. By adding emergency pathways, the maximum evacuation times for each case were recalculated, and the optimal additional evacuation pathway to reduce the maximum evacuation time was proposed. A total of three cases were analyzed with an additional evacuation pathway. Notably, the reduction in walking speed due to crowd density was considered, and movement speeds were adjusted across three levels of congestion. In Case No. 1, where an emergency staircase connecting B6 and B1 was added, the maximum evacuation time was reduced by 36.4% compared to the baseline case (No. 0) without any additional evacuation pathway. In Case No. 2, the addition of an emergency staircase connecting B2 and B1 resulted in an 18.0% reduction in evacuation time. In Case No. 3, where an additional exit was introduced, no change was observed in the maximum evacuation time. These results highlight the importance of appropriately constructing emergency pathways to ensure that all occupants can evacuate within the golden time.
This study proposes a physics-informed machine learning (PIML) framework for geotechnical engineering applications that combines the strengths of traditional physics-based methods and data-driven machine learning (ML) models. Bearing capacity prediction of shallow foundations is used as a case study to demonstrate the applicability of the proposed framework. The PIML framework integrates Vesic's bearing capacity method as an embedded differentiable physics module to improve model generalization and robustness. A benchmark data set was created using finite element analysis (FEA), consisting of 3,500 samples with randomly generated values for soil properties, foundation geometries, and loading conditions. To evaluate the effectiveness of the proposed framework, the data set was divided into non-overlapping clusters based on input features, followed by cross-cluster validation. Results show that both the baseline neural network (NN) and the PIML model outperform Vesic's method. Notably, the PIML model showed superior generalization performance and faster convergence, as evidenced by higher R-2 and lower RMSE values. The proposed PIML framework presents a hybrid approach that effectively addresses the limitations of standalone physics-based and ML methods, providing a scalable and interpretable tool for geotechnical and other engineering challenges under various environmental conditions.
Unmanned Aerial Vehicles (UAVs) have gained widespread popularity as an efficient means for remote sensing, with useful applications in geotechnical engineering, including asset management and reality modelling. When performing UAV surveying, the precision of survey measurements that are inferred is critical for assessing changes in soil morphology due to extreme events such as landslides or coastal erosion during severe storms. The georeferencing methods that are selected for UAV surveying have the potential to significantly affect the outcome and accuracy of reality modelling using current tools and software packages. Ground Control Points (GCPs) referenced to a conventional survey plane are the most common approach for georeferencing and orthophoto determination. It follows logically that the configuration and number of GCPs significantly affect the accuracy of the UAV survey. In this study, three analyses were performed on the photos extracted from two recorded videos of UAV flights conducted at different flight elevations. The semi-small planar area shown in each photo was georeferenced completely, and just two checkpoints were chosen to compare the effect of the positions of these points on the resulting UAV survey accuracy. Also, considered is the effect of frontal overlap of the extracted UAV images in the respective direction of the UAV flight path. The mean absolute error for two-dimensional, frontal, side, and altitude discrepancies was calculated under three different scenarios. Overall, the accuracy was observed to fluctuate for locations at larger magnitudes of frontal overlap, i.e., more than 85%. For the corner points, results show that the Mean absolute error for 640 photos (92.8% of frontal overlap) is larger than the corresponding error values for 320 and 1,280 photos (85.7% and 96.4% of frontal overlap, respectively) for a 30-m height flight. Although the planar errors slightly increase with higher frontal overlaps, they remain within the same accuracy range of the Ground Sample Distance scale. Additionally, the altitude error also remains consistent. The mean absolute error variation for lower-height flights is similar for the corner point assessments. However, the use of inside points has an increasingly higher error, which can be caused by the lower side overlap. Based on the results of this study, the location of the checkpoints was found to be one of the most critical factors for assessing the accuracy of a given georeferencing pattern.
Approaches to quantifying flood risk and event-specific damages typically rely on asset-specific depth-damage curves to characterize asset fragility. Current literature and practice lack sufficient information to develop tunnel-specific depth-damage curves for road tunnels, which are high-value critical infrastructure assets that typically traverse bodies of water and floodplains. Relying on an extant fragility function, tunnel geometry, and principles of fluid mechanics, we present and implement a volume-oriented approach for developing tunnel-specific depth-damage functions for road tunnels. We apply this approach to characterize the fragility of major road tunnels in Baltimore, Maryland, demonstrating the influence of tunnel geometry on estimated fragility. This volume-oriented approach can be readily applied to any tunnel (road, rail, etc.), thereby enabling infrastructure managers to characterize present and future flood risk to this class of high-value critical infrastructure assets. Adequate flood risk assessment of tunnel assets is critical to quantifying the flood risk reduction benefits of adaptation measures.
Rainfall-induced landslides are a significant geological hazard, causing severe economic losses and casualties. Accurate forecasting of these events is particularly challenging due to the complex interactions of spatial and temporal factors governing slope stability. The Iverson model, which uses the Richards equation to describe water infiltration in unsaturated soils, is a widely adopted framework for analyzing rainfall-induced landslides. However, its reliance on traditional numerical methods limits its scalability and efficiency, particularly for complex boundary conditions and transient behaviors near slope failure. To address these limitations, we propose a physics-informed neural network (PINN) enhanced with transfer learning (TL-PINN) to solve the Iverson model for landslide forecasting. The TL-PINN employs a transfer learning strategy, freezing specific hidden layers to improve performance in capturing the dynamics of slope stability under rainfall. This TL-PINN method predicts critical parameters such as the pressure head distribution, factor of safety (FS), and failure timing by efficiently solving the Richards equation. Validation through a case study demonstrates the ability of TL-PINN to handle discontinuous top-boundary conditions, sharp gradients, and nonlinear phenomena, outperforming traditional PINNs. The results highlight the potential of TL-PINN to advance the application of the Iverson model, providing valuable insights for analyzing and forecasting rainfall-induced landslides.
