Estimating spatial extremes from sparse observational networks produces uncertain return level maps, but dense output from physics-based simulation models is often available as a complementary data source. We develop a two-stage frequentist frame-work for fusing observations and simulations. In Stage 1, generalized extreme value (GEV) distributions are fitted independently at each site, with a nonstationary location parameter where appropriate to accommodate observed trends. In Stage 2, the parameter estimates from all sources are modeled jointly as a high-dimensional spatial process through a linear model of coregionalization (LMC). Cross-source correlations, estimated from spatially interspersed networks without co-located sites, provide the mechanism for information transfer; an analytic gradient for the resulting likelihood keeps estimation computationally practical. We apply the framework to U.S. coastal sea levels over 1979-2021, fusing 29 NOAA tide gauge records with 100 ADCIRC hydrodynamic simulation sites. Leave-one-out cross-validation shows a 35% reduction in 100-year return level RMSE relative to a gauge-only model. Geographic block cross-validation confirms that fusion benefits persist under spatial extrapolation. The approach is implemented in the R package evfuse.

The eastern North Carolina Coastal Area Management Act region is one of the most hurricane-prone areas of the United States. Hurricanes incur substantial damage and economic losses because structures located near the coast tend to be high value as well as particularly exposed. To bolster disaster mitigation and community resilience, it is crucial to understand how hurricane hazards drive social and economic impacts. We integrate detailed hazard simulations, property data, and labor compensation estimates to comprehensively analyze hurricanes’ economic impacts. This study investigates the spatial distribution of probabilistic hurricane hazards, and concomitant property losses and labor impacts, pinpointing particularly hard hit areas. Relationships between capital and labor losses, social vulnerability, and asset values reveal the latter as the primary determinant of overall economic consequences.

Coastal water level information is crucial for understanding flood occurrences and changing risks. Here, we validate the preliminary version (0.9) of NOAA’s Coastal Ocean Reanalysis (CORA), which is a 43-year reanalysis (1979–2021) of hourly coastal water levels for the Gulf of Mexico and Atlantic Ocean (i.e., the Gulf and East Coast region, or GEC). CORA-GEC v0.9 was conducted by the Renaissance Computing Institute using the coupled ADCIRC+SWAN coastal circulation and wave model. The model uses an unstructured mesh of nodes with varying spatial resolution that averages 400 m near the coast and is much coarser in the open ocean. Water level variations associated with tides and meteorological forcing are explicitly modeled, while lower-frequency water level variations are included by dynamically assimilating observations from NOAA’s National Water Level Observation Network. We compare CORA to water level observations that were either assimilated or not, and find that the reanalysis generally performs better than a state-of-the-art global ocean reanalysis (GLORYS12) in capturing the variability on monthly, seasonal, and interannual timescales as well as the long-term trend. The variability of hourly non-tidal residuals is also shown to be well resolved in CORA when compared to water level observations. Lastly, we present a case study of extreme water levels and coastal inundations around Miami, Florida to demonstrate an application of CORA for studying flood risks. Our assessment suggests that NOAA’s CORA-GEC v0.9 provides valuable information on water levels and flooding occurrence from 1979–2021 in areas that are experiencing changes across multiple time scales. CORA potentially can enhance flood risk assessment along parts of the U.S. Coast that do not have historical water level observations.

Two new and novel techniques to efficiently represent channels in large-scale finite element shallow water equation models are introduced. The first enables the discontinuous representation of bathymetric depth by allowing nodes along a discontinuity to have two land surface elevations, one representing the bottom of the discontinuity and the other the top of the discontinuity. Elemental integration proceeds using the nodal depth corresponding to whether the element is on the upper or lower side of the discontinuity and enables efficient treatment of steep sided features in a two-dimensional horizontal discretization. The second technique consolidates nodal equations at paired nodes across a channel. Doing this eliminates cross-channel variability and consequently eliminates the across-channel Courant-Friedrichs-Lewy stability constraint on the model time step. Together the two techniques allow channels of various sizes as well as other instances of steep topography to be embedded seamlessly, efficiently and in a fully coupled manner within an otherwise two-dimensional horizontal spatial discretization. This new capability is especially useful in the interface zone subject to both coastal and hydrological influences as it effectively captures bi-directional channelized flow, bi-directional flow between the channel and the floodplain, and coupled flow when the channel and floodplain are fully submerged. This paper describes the methodological development and presents a series of tests that verify the performance of this novel solution approach.

This paper introduces an approach to evaluate the performance of a previously implemented or proposed hurricane evacuation plan that describes where and when official evacuation orders are issued. The approach involves use of the new integrated scenario-based evacuation (ISE) decision support tool to define a best track evacuation plan as a reference point and measure the performance of other plans in relation to that according to their ability to meet multiple stated objectives: minimizing risk to the population, travel time, and time people are away from their homes. Using North Carolina in Hurricane Florence (2018) as a case study, we demonstrate the process by evaluating performance of both the actual set of orders as executed and the orders that would have been recommended if the new ISE decision support tool had been used during the event. All three plans were evaluated for two cases-assuming the hurricane unfolds as it actually did, and if the hurricane had instead evolved like one of 21 other realistic scenarios. Results suggest the actual evacuation was quite good, and the ISE tool could have resulted in improved evacuation performance. (C) 2022 American Society of Civil Engineers.

Surrogate models are becoming increasingly popular for storm surge predictions. Using existing databases of storm simulations, developed typically during regional flood studies, these models provide fast-to-compute, data-driven approximations quantifying the expected storm surge for any new storm (not included in the training database). This paper considers the development of such a surrogate model for Delaware Bay, using a database of 156 simulations driven by synthetic tropical cyclones and offering predictions for a grid that includes close to 300,000 computational nodes within the geographical domain of interest. Kriging (Gaussian Process regression) is adopted as the surrogate modeling technique, and various relevant advancements are established. The appropriate parameterization of the synthetic storm database is examined. For this, instead of the storm features at landfall, the features when the storm is at closest distance to some representative point of the domain of interest are investigated as an alternative parametrization, and are found to produce a better surrogate. For nodes that remained dry for some of the database storms, imputation of the surge using a weighted k nearest neighbor (kNN) interpolation is considered to fill in the missing data. The use of a secondary, classification surrogate model, combining logistic principal component analysis and Kriging, is examined to address instances for which the imputed surge leads to misclassification of the node condition. Finally, concerns related to overfitting for the surrogate model are discussed, stemming from the small size of the available database. These concerns extend to both the calibration of the surrogate model hyper-parameters, as well as to the validation approaches adopted. During this process, the benefits from the use of principal component analysis as a dimensionality reduction technique, and the appropriate transformation and scaling of the surge output are examined in detail.

During coastal storms, forecasters and researchers use numerical models to predict the magnitude and extent of coastal flooding. These models must represent the large regions that may be affected by a storm, and thus, they can be computationally costly and may not use the highest geospatial resolution. However, predicted flood extents can be downscaled (by increasing resolution) as a post-processing step. Existing downscaling methods use either a static extrapolation of the flooding as a flat surface, or rely on subsequent simulations with nested, full-physics models at higher resolution. This research explores a middle way, in which the downscaling includes simplified physics to improve accuracy. Using results from a state-of-the-art model, we downscale its flood predictions with three methods: (1) static, in which the water surface elevations are extrapolated horizontally until they intersect the ground surface; (2) slopes, in which the gradient of the water surface is used; and (3) head loss, which accounts for energy losses due to land cover characteristics. The downscaling methods are then evaluated for forecasts and hindcasts of Hurricane Florence (2018), which caused widespread flooding in North Carolina. The static and slopes methods tend to over-estimate the flood extents. However, the head loss method generates a downscaled flooding extent that is a close match to the predictions from a higher-resolution, full-physics model. These results are encouraging for the use of these downscaling methods to support decision-making during coastal storms.

This article introduces a new integrated scenario-based evacuation (ISE) framework to support hurricane evacuation decision making. It explicitly captures the dynamics, uncertainty, and human-natural system interactions that are fundamental to the challenge of hurricane evacuation, but have not been fully captured in previous formal evacuation models. The hazard is represented with an ensemble of probabilistic scenarios, population behavior with a dynamic decision model, and traffic with a dynamic user equilibrium model. The components are integrated in a multistage stochastic programming model that minimizes risk and travel times to provide a tree of evacuation order recommendations and an evaluation of the risk and travel time performance for that solution. The ISE framework recommendations offer an advance in the state of the art because they: (1) are based on an integrated hazard assessment (designed to ultimately include inland flooding), (2) explicitly balance the sometimes competing objectives of minimizing risk and minimizing travel time, (3) offer a well-hedged solution that is robust under the range of ways the hurricane might evolve, and (4) leverage the substantial value of increasing information (or decreasing degree of uncertainty) over the course of a hurricane event. A case study for Hurricane Isabel (2003) in eastern North Carolina is presented to demonstrate how the framework is applied, the type of results it can provide, and how it compares to available methods of a single scenario deterministic analysis and a two-stage stochastic program.

Modeling complex systems with governing physical equations is difficult in many areas due to computational challenges. Current research in Artificial Intelligence and Machine Learning (AI/ML) has produced evidence that data-driven learning algorithms can excel at tasks like classification and prediction where there is no known overarching mathematical model, e.g. in applications like computer vision and natural language processing. An emerging area of research suggests that scientific prediction can also benefit from data-driven AI/ML. However, this optimism is tempered by experimental challenges which indicate that exclusive reliance on historical data without inclusion of physical knowledge does not work well in practice. While at this early stage of research we do not yet have unifying frameworks for integration of physical knowledge with data-driven approaches, researchers working in different applications have discovered mathematical and computational tools that create effective solutions. It is hypothesized that the storm surge prediction problem may also benefit from such an integrated approach. We describe research on constraining an Artificial Neural Network (ANN) with a physics-based loss function that penalizes the ANN when it deviates from prediction patterns that would normally be expected from the input data of water level time series. This is an initial step in a new approach to the problem of storm surge prediction. With additional research to include geospatial constraints, knowledge of distribution shift over time in water levels, and wind velocity field, we may begin to see reduction in computational time without sacrificing accuracy when used alongside established numerical programs like ADCIRC. Increasing the speed of storm surge prediction can be highly beneficial to society and the mission of DHS since it can save lives and property in exposed areas. Department of Homeland Security Summer Research Team Final Faculty Report 2020 3 Storm Surge Prediction with Integration of Physical Knowledge in Machine Learning Models

The amount and extent of coastal flooding caused by hurricanes can be sensitive to the timing or speed of the storm. For storms moving parallel to the coast, the hazards can be stretched over a larger area. Hurricane Matthew was a powerful storm that impacted the southeastern U.S. during October 2016, moving mostly parallel to the coastline from Florida through North Carolina. In this study, three sources for atmospheric forcing are considered for a simulation of Matthew's water levels, which are validated against extensive observations, and then the storm's effects are explored on this long coastline. It is hypothesized that the spatial variability of Matthew's effects on total water levels is partly due to the surge interacting nonlinearly with tides. By changing the time of occurrence of the storm, differences in storm surge are observed in different regions due to the storm coinciding with other periods in the tidal cycles. These differences are found to be as large as 1 m and comparable to the tidal amplitude. A change in forward speed of the storm also should alter its associated flooding due to differences in the duration over which the storm impacts the coastal waters. With respect to the forward speed, the present study contributes to established results by considering the scenario of a shore-parallel hurricane. A faster storm caused an increase in peak water levels along the coast but a decrease in the overall volume of inundation. On the other hand, a slower storm pushed more water into the estuaries and bays and flooded a larger section of the coast. Implications for short-term forecasting and long-term design studies for storms moving parallel to long coastlines are discussed herein.

Formal engineering hurricane evacuation studies have not typically considered inland flooding explicitly, though it has been shown repeatedly to be a major cause of damage and loss of life in hurricanes. In addition, coastal flooding and strong winds are often treated in a decoupled manner, so that the correlation between them is not captured. The recently introduced Integrated Scenario-based Evacuation (ISE) computational framework offers one approach to achieving evacuation decision support based on a representation of the hazard that considers coastal flooding, inland flooding, and wind in an integrated manner. Using a case study application of the ISE framework for Hurricane Matthew (2016) approaching the North Carolina coast, we evaluate the influence of including inland flooding on the resulting recommended evacuation plan (where and when official evacuation orders are to be issued) and the plan's performance in terms of risk reduction and travel time increase. Results provide insight into managing hurricane evacuation with consideration of inland flooding. They suggest that in some cases inland areas should be evacuated just as coastal areas are; the scenarios responsible for and the timing of inland flooding can differ from those for coastal areas; the response to the different hazards should be treated together as a system because they can interact in complex ways; and planning for inland flooding can help reduce risk substantially while not adding much to evacuee travel times because inland evacuees do not have to travel as far to safety.

Research performed to-date on data assimilation (DA) in storm surge modeling has found it to have limited value for predicting rapid surge responses (e.g., those accompanying tropical cyclones). In this paper, we submit that a well-resolved, barotropic hydrodynamic model is typically able to capture the surge event itself, leaving slower processes that determine the large scale, background water level as primary sources of water level error. These "unresolved drivers'' reflect physical processes not included in the model's governing equations or forcing terms, such as far field atmospheric forcing, baroclinic processes, major ocean currents, steric variations, or precipitation. We have developed a novel, efficient, optimal interpolation-based DA scheme, using observations from coastal water level gages, that dynamically corrects for the presence of unresolved drivers. The methodology is applied for Hurricane Matthew (2016) and results demonstrate it is highly effective at removing water level residuals, roughly halving overall surge errors for that storm. The method is computationally efficient, well-suited for either hindcast or forecast applications and extensible to more advanced techniques and datasets.

© 2019 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).CORRESPONDING AUTHOR: Robert Erhardt, erhardrj@wfu.eduA supplement to this article is available online (10.1175/BAMS-D-19-0073.2).

The evolution of a hurricane—how the track, intensity, forward speed, and resulting hazard effects on land (strong winds, flooding) develop over its lifetime—is often highly uncertain. Further, the uncertainty is dynamic because it is resolved as events unfold until ultimately the storm's evolution is known completely, and because the ensemble of forecasts changes over time. Emergency managers recognize these challenges and may engage in some combination of robust, adaptive, or repeated planning to address them. However, science- and engineering-based evacuation decision support models typically do not formally incorporate uncertainty. This article discusses the use of formal modeling to support robust, adaptive, and repeated decision-making during an impending hurricane. It also details a case study of Hurricane Isabel (2003) in North Carolina using the recently introduced Integrated Scenario-based Evacuation (ISE) computational framework to compare the effects of including each of the three features in the modeling. Findings suggest that making the evacuation planning robust, adaptive, and repeated should improve results by reducing both the numbers of people at risk and unnecessary evacuation orders and travel. The magnitude of those benefits, however, depends on uncertainty in, and evolution of, the attributes of the particular hurricane.

by Russell s. Vose, scott Applequist, MARk A. bouRAssA, sARA c. pRyoR, RebeccA J. bARthelMie, bRiAn blAnton, peteR D. bRoMiRski, hARolD e. bRooks, ARthuR t. DeGAetAno, RAnDAll M. Dole, DAViD R. eAsteRlinG, RobeRt e. Jensen, thoMAs R. kARl, RichARD W. kAtz, kAtheRine klink, MichAel c. kRuk, kenneth e. kunkel, MichAel c. MAccRAcken, thoMAs c. peteRson, kARsten shein, bRiDGet R. thoMAs, John e. WAlsh, XiAolAn l. WAnG, MichAel F. WehneR, DonAlD J. Wuebbles, AnD RobeRt s. younG

Storm surge prediction models rely on an accurate representation of the wind conditions. In this paper, we examine the sensitivity of surge predictions to forecast uncertainties in the track and strength of a storm (storm strength is quantified by the power dissipation of the associated wind field). This analysis is performed using Hurricane Arthur (2014), a Category 2 hurricane, which made landfall along the North Carolina (NC) coast in early July 2014. Hindcast simulations of a coupled hydrodynamic-wave model are performed on a large un-structured mesh to analyze the surge impact of Arthur along the NC coastline. The effects of Arthur are best represented by a post-storm data assimilated wind product with parametric vortex winds providing a close approximation. Surge predictions driven by forecast advisories issued by the National Hurricane Center (NHC) during Arthur are analyzed. The storm track predictions from the NHC improve over time. However, successive advisories predict an unrealistic increase in the storm's strength. Due to these forecast errors, the global root mean square errors of the predicted wind speeds and water levels increase as the storm approaches landfall. The relative impacts of the track and strength errors on the surge predictions are assessed by replacing forecast storm parameters with the best known post-storm information about Arthur. In a "constant track" analysis, Arthur's post storm determined track is used in place of the track predictions of the different advisories but each advisory retains its size and intensity predictions. In a "constant storm strength" analysis, forecast wind and pressure parameters are replaced by corresponding parameters extracted from the post storm analysis while each advisory retains its forecast storm track. We observe a strong correlation between the forecast errors and the wind speed predictions. However, the correlation between these errors and the forecast water levels is weak signifying a non-linear response of the shallow coastal waters to meteorological forcing.

Hurricane track and intensity can change rapidly in unexpected ways, thus making predictions of hurricanes and related hazards uncertain. This inherent uncertainty often translates into suboptimal decision-making outcomes, such as unnecessary evacuation. Representing this uncertainty is thus critical in evacuation planning and related activities. We describe a physics-based hazard modeling approach that (1) dynamically accounts for the physical interactions among hazard components and (2) captures hurricane evolution uncertainty using an ensemble method. This loosely coupled model system provides a framework for probabilistic water inundation and wind speed levels for a new, risk-based approach to evacuation modeling, described in a companion article in this issue. It combines the Weather Research and Forecasting (WRF) meteorological model, the Coupled Routing and Excess STorage (CREST) hydrologic model, and the ADvanced CIRCulation (ADCIRC) storm surge, tide, and wind-wave model to compute inundation levels and wind speeds for an ensemble of hurricane predictions. Perturbations to WRF's initial and boundary conditions and different model physics/parameterizations generate an ensemble of storm solutions, which are then used to drive the coupled hydrologic + hydrodynamic models. Hurricane Isabel (2003) is used as a case study to illustrate the ensemble-based approach. The inundation, river runoff, and wind hazard results are strongly dependent on the accuracy of the mesoscale meteorological simulations, which improves with decreasing lead time to hurricane landfall. The ensemble envelope brackets the observed behavior while providing "best-case" and "worst-case" scenarios for the subsequent risk-based evacuation model.

Great uncertainty surrounds the future track, intensity, and resulting flooding and wind hazards associated with an offshore hurricane. As time progresses, that uncertainty is reduced, but so is the ability to evacuate or carry out other preparations. A quantitative assessment of how quickly the uncertainty is likely to resolve can thus provide useful input for emergency managers. If the impacts of a hurricane are likely to become more certain soon, an emergency manager may wait for more information. If not, the emergency manager may proceed with preparations. This paper introduces scenario-based hazard trees, a new tool that describes the resolution over time of uncertainty in a hurricane's impacts. The primary input is a scenario ensemble, commonly developed as part of a hurricane forecast. A linear integer program transforms the ensemble into a tree. The hazard trees provide a new dynamic way to characterize how uncertainty changes during the course of a hurricane, and can provide the required input for a multi-stage stochastic programming evacuation model. A case study is presented for eastern North Carolina during Hurricane Isabel in 2003.