Improvements in storm surge surrogate modeling for synthetic storm parameterization,node condition classification and implementation to small size databases

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.

     

The impact of extra-tropical transitioning on storm surge and waves in catastrophe risk modelling: application to the Japanese coastline

Catastrophe risk models are used to assess and manage the economic and societal impacts of natural perils such as tropical cyclones. Large ensembles of event simulations are required to generate useful model output. For example, to estimate the risk due to wind-driven storm surge and waves in tropical cyclone risk models, computationally efficient parametric representations of the wind forcing are required to enable the generation of large ensembles. This paper presents new results on the impact of including explicit representations of extra-tropical transitioning in parametric wind models used to force storm surge and wave simulations in a catastrophe risk modelling context. Extra-tropical transitioning is particularly important in modelling risk on the Japanese coastline, as roughly 40 % of typhoons hitting the Japanese mainland are transitioning before landfall. Using both a historical and idealized track set, we compare maximum storm surge and wave footprints along the Japanese coastline for models that include, and do not include, explicit representations of extra-tropical transitioning. We find that the inclusion of extra-tropical transitioning leads to lower storm surge (10–20 %) and waves (5–15 %) on the southern Japanese coast, with significantly higher storm surge and waves along the northern coast (25–50 %). The results of this paper demonstrate that useful risk assessment of coastal flood risk in Japan must consider the extra-tropical transitioning process.     

Dynamic issues in the SE South America storm surge modeling

The ability of the SMARA storm surge numerical prediction system to reproduce local effects in estuarine and coastal winds was recently improved by considering one-way coupling of the air–sea momentum exchange through the wave stress, and best forecasting practices for downscaling. The inclusion of long period atmospheric pressure forcing in tide and tide/surge calculations corrected a systematic error in the surge, produced by the South Atlantic Ocean quasi-stationary pressure patterns. The maximum forecast range for the storm surge at Buenos Aires provided by the real-time use of water level observations is approximately 12 h. The best available water level prediction is the 6-h forecast (nowcast) based on the closest water level observations. The 24-h forecast from the numerical models slightly improves this nowcast. Although the numerical forecast accuracy degrades after the first 48 h, the improvement to the full range observation-based prediction is maintained at the inner Río de la Plata area and extends to the first 3 days at the intermediate navigation channels.     

The impact of a storm surge on business establishments in the Houston MSA

This paper examines the possible storm surge damage from a major hurricane to hit the Houston Metropolitan Statistical Area (MSA.) Using storm surge analysis on a unique data set compiled from the Texas Workforce Commission (QCEW), the paper estimates the expected industry-level damage for each county in the Houston MSA. The advantages of using GIS to analyze the expected storm surge damage estimation is that it provides an accurate estimation of the number of affected employees and probable wages losses, by industry and county, based on QCEW data. The results indicate that the ‘Basic Chemical Manufacturing’ and ‘Oil and Gas Extraction’ industries incur the highest employee and payroll losses while the ‘Restaurants and Eateries’ has the largest establishment damage if a major hurricane were to hit the Houston MSA.     

On the use of NOAA's storm surge model,SLOSH, in managing coastal hazards — the experience in Puerto Rico

A numerical-dynamic, tropical storm surge model, SLOSH (Sea, Land, and Overland Surges from Hurricanes), was originally developed for real-time forecasting of hurricane storm surges on continental shelves, across inland water bodies and along coastlines and for inland routing of water -either from the sea or from inland water bodies. The model is two-dimensional, covering water bodies and inundated terrain. In the present version available at the University of Puerto Rico a curvilinear, polar coordinate grid scheme is used. The grid cells are approximately 3.2 × 3.2 km in size.The model has been used in a revision of all coastal Flood Insurance Rate Maps (FIRM) for Puerto Rico and the U.S. Virgin Islands, and in hurricane evacuation studies. The FIRM's, since they are based on the 100 year stillwater elevation, are also used by the state Planning Board for regulatory purposes. The hurricane evacuation studies are used by emergency planners and personnel to assign shelters, escape routes, and delimit coastal zones that need to be evacuated during a hurricane threat.Recently, the acquisition of data from hurricane Hugo has allowed the first comparison of model results and observations for Puerto Rico and the other islands.     

含盐风暴潮溃堤洪水数值模拟及风险分析

为了研究三角洲河口风暴潮溃堤时的盐水运动规律,建立一、二维耦合的盐度数学模型对风暴潮溃堤时的盐水运动进行模拟。模型考虑洪泛区建筑物对盐水运动的影响以及溃口的渐变发展过程。用2008年多个测站的实测数据对河网模型的潮位和盐度计算结果进行了验证。将模型应用于珠江三角洲河网某近海溃口风暴潮溃堤的盐水运动模拟,并绘制了最大盐度等值面图。计算结果表明,该溃口大部分区域的溃堤积水盐度超过了4psu,因此,溃堤洪水的高盐度积水影响不容忽视。通过比较“溃堤”和“不溃堤”两种情况下的河网盐度计算结果,发现上游河道的溃堤分流增大了河道的纳潮量,促使涨潮量增大,增大了下游河网的咸潮上溯风险,减弱了上游来流对咸潮的压制效果。     

Hurricane storm surge simulations for Tampa Bay

Using a high resolution, three-dimensional, primitive equation, finite volume coastal ocean model with flooding and drying capabilities, supported by a merged bathymetric-topographic data set and driven by prototypical hurricane winds and atmospheric pressure fields, we investigated the storm surge responses for the Tampa Bay, Florida, vicinity and their sensitivities to point of landfall, direction and speed of approach, and intensity. All of these factors were found to be important. Flooding potential by wind stress and atmospheric pressure induced surge is significant for a category 2 hurricane and catastrophic for a category 4 hurricane. Tide, river, and wave effects are additive, making the potential for flood-induced damage even greater. Since storm surge sets up as a slope to the sea surface, the highest surge tends to occur over the upper reaches of the bay, Old Tampa Bay and Hillsborough Bay in particular. For point of landfall sensitivity, the worst case is when the hurricane center is positioned north of the bay mouth such that the maximum winds associated with the eye wall are at the bay mouth. Northerly (southerly) approaching storms yield larger (smaller) surges since the winds initially set up (set down) water level. As a hybrid between the landfall and direction sensitivity experiments, a storm transiting up the bay axis from southwest to northeast yields the smallest surge, debunking a misconception that this is the worst Tampa Bay flooding case. Hurricanes with slow (fast) translation speeds yield larger (smaller) surges within Tampa Bay due to the time required to redistribute mass.     

On the accuracy of atmospheric forcing for extra-tropical storm surge prediction

The ability of the SMARA storm surge numerical prediction system to reproduce local effects in estuarine and coastal winds was recently improved by considering one-way coupling of the air–sea momentum exchange through the wave stress, and best forecasting practices for downscaling. The inclusion of long period atmospheric pressure forcing in tide and tide/surge calculations corrected a systematic error in the surge, produced by the South Atlantic Ocean quasi-stationary pressure patterns. The maximum forecast range for the storm surge at Buenos Aires provided by the real-time use of water level observations is approximately 12 h. The best available water level prediction is the 6-h forecast (nowcast) based on the closest water level observations. The 24-h forecast from the numerical models slightly improves this nowcast. Although the numerical forecast accuracy degrades after the first 48 h, the improvement to the full range observation-based prediction is maintained at the inner Río de la Plata area and extends to the first 3 days at the intermediate navigation channels.     

Extremity analysis of storm surge for fixing safe design water level

Phenomenal storm surge levels associated with cyclones are common in East Coast of India. The coastal regions of Andhra Pradesh are in rapid stride of myriad marine infrastructural developments. The safe elevations of coastal structures need a long-term assessment of storm surge conditions. Hence, past 50 years (1949–1998), tropical cyclones hit the Bay are obtained from Fleet Naval Meteorological & Oceanographic Center, USA, and analyzed to assess the storm surge experienced around Kakinada and along south Andhra Pradesh coast. In this paper, authors implemented Rankin Hydromet Vortex model and Bretschneider’s wind stress formulation to hindcast the surge levels. It is seen from the hindcast data that the November, 1977 cyclone has generated highest surge of the order of 1.98 m. Extreme value analysis is carried out using Weibull distribution for long-term prediction. The results reveal that the surge for 1 in 100-year return period is 2.0 m. Further the highest surge in 50 years generated by the severe cyclone (1977) is numerically simulated using hydrodynamic model of Mike-21. The simulation results show that the Krishnapatnam, Nizampatnam and south of Kakinada have experienced a surge of 1.0, 1.5 and 0.75 m, respectively.     

Impact of wind gusts on sea surface height in storm surge modelling, application to the North Sea

Storm surge models usually do not take into account the explicit effect of wind gusts on the sea surface height. However, as the wind speed enters quadratically into the shallow water equations, short-term fluctuations around the mean value do not average out. We investigate the impact of explicitly added gustiness on storm surge forecasts in the North Sea, using the WAQUA/DCSM model. The sensitivity of the model results to gustiness is tested with Monte Carlo simulations, and these are used to derive a parametrisation of the effect of gustiness on characteristics of storm surges. With the parametrisation and input from the ECMWF model archive, we run hindcasts for a few individual cases and also the 2007–2008 winter storm season. Although the explicit inclusion of gustiness increases the surge levels, it does not help to explain, and hence reduce, the errors in the model results. Moreover, the errors made by ignoring gustiness are small compared to other errors. We conclude that, at present, there is no need to include gustiness explicitly in storm surge calculations for the North Sea.     

渤海湾西部风暴潮漫滩数值模拟

基于ROMS海洋模式,结合近年的地质实测资料,建立了渤海湾西部地区风暴潮漫滩的数值模型。对模型进行验证后,对渤海湾西部区域重现期为50a、100a、200a及500a的风暴潮漫滩进行了数值模拟,分析了不同重现期风暴潮漫滩发展的动态过程及最大漫滩淹水范围。结果表明,数值模型基本能反映风暴潮的增水趋势,能够模拟风暴潮漫滩发生发展的动态过程。随着风暴潮强度的增加,渤海湾西部地区淹水范围具有从东海岸向西部内陆区域扩展的趋势。通过曲线拟合发现,风暴潮最大漫滩面积比值与高水位之间基本呈线性关系。     

A modeling study of coastal inundation induced by storm surge,sea-level rise,and subsidence in the Gulf of Mexico

The northern coasts of the Gulf of Mexico (GoM) are highly vulnerable to the direct threats of climate change, such as hurricane-induced storm surge, and such risks are exacerbated by land subsidence and global sea-level rise. This paper presents an application of a coastal storm surge model to study the coastal inundation process induced by tide and storm surge, and its response to the effects of land subsidence and sea-level rise in the northern Gulf coast. The unstructured-grid finite-volume coastal ocean model was used to simulate tides and hurricane-induced storm surges in the GoM. Simulated distributions of co-amplitude and co-phase lines for semi-diurnal and diurnal tides are in good agreement with previous modeling studies. The storm surges induced by four historical hurricanes (Rita, Katrina, Ivan, and Dolly) were simulated and compared to observed water levels at National Oceanic and Atmospheric Administration tide stations. Effects of coastal subsidence and future global sea-level rise on coastal inundation in the Louisiana coast were evaluated using a “change of inundation depth” parameter through sensitivity simulations that were based on a projected future subsidence scenario and 1-m global sea-level rise by the end of the century. Model results suggested that hurricane-induced storm surge height and coastal inundation could be exacerbated by future global sea-level rise and subsidence, and that responses of storm surge and coastal inundation to the effects of sea-level rise and subsidence are highly nonlinear and vary on temporal and spatial scales.     

Simulation of atmospheric states for a storm surge on the west coast of Korea: model comparison between MM5, WRF and COAMPS

High-quality informations on sea level pressure and sea surface wind stress are required to accurately predict storm surges over the Korean Peninsula. The storm surge on 31 March 2007 at Yeonggwang, on the western coast, was an abrupt response to mesocyclone development. In the present study, we attempted to obtain reliable surface winds and sea level pressures. Using an optimal physical parameterization for wind conditions, MM5, WRF and COAMPS were used to simulate the atmospheric states that accompanied the storm surge. The use of MM5, WRF and COAMPS simulations indicated the development of high winds in the strong pressure gradient due to an anticyclone and a mesocyclone in the southern part of the western coast. The response to this situation to the storm surge was sensitive. A low-level warm advection was examined as a possible causal mechanism for the development of a mesocyclone in the generating storm surge. The low-level warm temperature advection was simulated using the three models, but MM5 and WRF tended to underestimate the warm tongue and overestimate the wind speed. The WRF simulation was closer to the observed data than the other simulations in terms of wind speed and the intensity of the mesocyclone. It can be concluded that the magnitude of the storm surge at Yeonggwang was dependent, not only on the development of a mesocyclone but on ocean effects as well.     

Mapping joint hurricane wind and surge hazards for Charleston,South Carolina

Combined effects of hurricane wind and surge can pose significant threats to coastal cities. Although current design codes consider the joint occurrence of wind and surge, information on site-specific joint distributions of hurricane wind and surge along the US Coast is still sparse and limited. In this study, joint hazard maps for combined hurricane wind and surge for Charleston, South Carolina (SC), were developed. A stochastic Markov chain hurricane simulation program was utilized to generate 50,000 years of full-track hurricane events. The surface wind speeds and surge heights from individual hurricanes were computed using the Georgiou’s wind field model and the Sea, Lake and Overland Surges from Hurricanes (SLOSH) model, respectively. To validate the accuracy of the SLOSH model, the simulated surge levels were compared to the surge levels calculated by another state-of-the-art storm surge model, ADCIRC (Advanced Circulation), and the actual observed water elevations from historical hurricane events. Good agreements were found between the simulated and observed water elevations. The model surface wind speeds were also compared with the design wind speeds in ASCE 7-10 and were found to agree well with the design values. Using the peak wind speeds and maximum surge heights, the joint hazard surfaces and the joint hazard maps for Charleston, SC, were developed. As part of this study, an interactive computer program, which can be used to obtain the joint wind speed and surge height distributions for any location in terms of latitude and longitude in Charleston area, was created. These joint hazard surfaces and hazard maps can be used in a multi-hazard design or risk assessment framework to consider the combined effects of hurricane wind and surge.     

Implementation/optimization of moving least squares response surfaces for approximation of hurricane/storm surge and wave responses

One of the important recent advances in the field of hurricane/storm modelling has been the development of high-fidelity numerical simulation models for reliable and accurate prediction of wave and surge responses. The computational cost associated with these models has simultaneously created an incentive for researchers to investigate surrogate modelling (i.e. metamodeling) and interpolation/regression methodologies to efficiently approximate hurricane/storm responses exploiting existing databases of high-fidelity simulations. Moving least squares (MLS) response surfaces were recently proposed as such an approximation methodology, providing the ability to efficiently describe different responses of interest (such as surge and wave heights) in a large coastal region that may involve thousands of points for which the hurricane impact needs to be estimated. This paper discusses further implementation details and focuses on optimization characteristics of this surrogate modelling approach. The approximation of different response characteristics is considered, and special attention is given to predicting the storm surge for inland locations, for which the possibility of the location remaining dry needs to be additionally addressed. The optimal selection of the basis functions for the response surface and of the parameters of the MLS character of the approximation is discussed in detail, and the impact of the number of high-fidelity simulations informing the surrogate model is also investigated. Different normalizations of the response as well as choices for the objective function for the optimization problem are considered, and their impact on the accuracy of the resultant (under these choices) surrogate model is examined. Details for implementation of the methodology for efficient coastal risk assessment are reviewed, and the influence in the analysis of the model prediction error introduced through the surrogate modelling is discussed. A case study is provided, utilizing a recently developed database of high-fidelity simulations for the Hawaiian Islands.     

Simulation of the Hurricane Dennis storm surge and considerations for vertical resolution

A high-resolution storm surge model of Apalachee Bay in the northeastern Gulf of Mexico is developed using an unstructured grid finite-volume coastal ocean model (FVCOM). The model is applied to the case of Hurricane Dennis (July 2005). This storm caused underpredicted severe flooding of the Apalachee Bay coastal area and upriver inland communities. Accurate resolution of complicated geometry of the coastal region and waterways in the model reveals processes responsible for the unanticipated high storm tide in the area. Model results are validated with available observations of the storm tide. Model experiments suggest that during Dennis, excessive flooding in the coastal zone and the town of St. Marks, located up the St. Marks River, was caused by additive effects of coincident high tides (~10–15% of the total sea-level rise) and a propagating shelf wave (~30%) that added to the locally wind-generated surge. Wave setup, the biggest uncertainty, is estimated on the basis of empirical and analytical relations. The Dennis case is then used to test the sensitivity of the model solution to vertical discretization. A suite of model experiments is performed with varying numbers of vertical sigma (σ) levels, with different distribution of σ-levels within the water column and a varying bottom drag coefficient. The major finding is that the storm surge solution is more sensitive to resolution within the velocity shear zone at mid-depths compared to resolution of the upper and bottom layer or values of the bottom drag coefficient.     

Influence of landscape restoration and degradation on storm surge and waves in southern Louisiana

The purpose of this investigation was to examine storm surge and wave reduction benefits of different environmental restoration features (marsh restoration and barrier island changes), as well as the impact of future wetland degradation on local surge and wave conditions. Storm surge simulations of two representative hurricanes were performed using the ADCIRC storm surge model with the inclusion of radiation stress gradients from the STWAVE nearshore wave model. Coupled model simulations were made for a number of landscape configurations that involved both restored and degraded wetland features. The impact of barrier island condition on hurricane surge and waves was also evaluated. Effects of landscape features were represented by changes in elevation and frictional resistance. Restoration and degradation of marsh resulted in decreases (for restoration cases) and increases (for degradation cases) in both surge and waves. The magnitude of change was correlated with the magnitude of the horizontal extent and elevation changes in the marsh. In general, the wave change patterns are consistent with the water level changes. Deflation of the Chandeleur Islands (barrier island chain) resulted in slightly increased surge. Results suggest that coastal marsh does have surge and wave reduction potential. Results also indicate that the impact of the landscape features is amplified in areas where there are levee “pockets.” Barrier islands and coastal ridges reduce wave heights, even if in a degraded condition and thus can reduce wave energy in wetland areas, protecting them from erosion.     

Application of a fast stochastic storm surge model on estimating the high water level frequency in the Lower Rhine Delta

In the Lower Rhine Delta of the Netherlands, the high water level is driven by a joint impact of the downstream storm surge and the upstream fluvial discharge, and affected by the operation of existing man-made structures. In scenario-based risk assessment, a large number of stochastic scenarios of storm surges are required for estimating the high water level frequency. In this article, a fast computing stochastic storm surge model is applied to the gauge station of Hook of Holland in the west of the Netherlands. A fixed number of tides are considered in this model based on the information of historical storm surge events. Based on this model, a large number of stochastic storm surge scenarios are derived and forced into a one-dimensional hydrodynamic model of the Netherlands, resulting in peak water levels in Rotterdam, the most vulnerable city in the delta. These peak water levels are statistically analyzed and converted to the high water level frequency curve in Rotterdam. The high water level frequency curve in Rotterdam tends to a much lower design water level compared to the official design water level that is used to design the dikes and structures for protection of the city. Moreover, there is a significant difference in the high water level frequency curves due to the fact that the stochastic storm surge model considers different numbers of tides. This highlights the critical impact of the storm surge duration on the high water level frequency in the Lower Rhine Delta.     

Advances in surrogate modeling for storm surge prediction: storm selection and addressing characteristics related to climate change

This paper establishes various advancements for the application of surrogate modeling techniques for storm surge prediction utilizing an existing database of high-fidelity, synthetic storms (tropical cyclones). Kriging, also known as Gaussian process regression, is specifically chosen as the surrogate model in this study. Emphasis is first placed on the storm selection for developing the database of synthetic storms. An adaptive, sequential selection is examined here that iteratively identifies the storm (or multiple storms) that is expected to provide the greatest enhancement of the prediction accuracy when that storm is added into the already available database. Appropriate error statistics are discussed for assessing convergence of this iterative selection, and its performance is compared to the joint probability method with optimal sampling, utilizing the required number of synthetic storms to achieve the same level of accuracy as comparison metric. The impact on risk estimation is also examined. The discussion then moves to adjustments of the surrogate modeling framework to support two implementation issues that might become more relevant due to climate change considerations: future storm intensification and sea level rise (SLR). For storm intensification, the use of the surrogate model for prediction extrapolation is examined. Tuning of the surrogate model characteristics using cross-validation techniques and modification of the tuning to prioritize storms with specific characteristics are proposed, whereas an augmentation of the database with new/additional storms is also considered. With respect to SLR, the recently developed database for the US Army Corps of Engineers’ North Atlantic Comprehensive Coastal Study is exploited to demonstrate how surrogate modeling can support predictions that include SLR considerations.