ORCID ID

0000-0002-7962-8188

Date Awarded

2014

Document Type

Dissertation

Degree Name

Doctor of Philosophy (Ph.D.)

Department

Virginia Institute of Marine Science

Advisor

Harry V. Wang

Abstract

Coastal inundation initiated via storm surge by hurricanes and nor'easters along the U.S. East Coast is a substantial threat to residential properties, community infrastructure, and human life. During and after the storm, compounding with heavy precipitation and upland drainage, inundation can be caused by the combination of storm surge and river-induced inland flooding in various locations throughout the coastal plain. Thus, coastal inundation can be expanded from the open coast upstream into the tributaries of the New York Bay including the Hudson and East River systems. Given the cross-disciplinary nature of the dynamics (encompassing hydraulics, oceanography, and hydrology), and the complexity of the atmospheric forcing, a numerical model is the optimal approach for a comprehensive study of the hydrodynamics of coastal inundation.;This study will utilize the large-scale parallel SELFE model to simulate the storm surge and inundation caused by 2012 Hurricane Sandy utilizing different forecast wind and pressure fields. The large-scale numerical model made use of multiple inputs for atmospheric forcing and spatially covered a large domain area to account for large-scale oceanographic processes and output accurate model simulation of water levels. In a simultaneous effort, a street-level sub-grid inundation model coupled with Lidar-derived topography (UnTRIM 2) was employed to simulate localized flooding events in the New York Harbor.;Sub-grid modeling is a novel method by which water level elevations are efficiently calculated on a coarse computational grid, with discretized bathymetric depths and topographic heights stored on a sub-grid nested within each base grid cell, capable of addressing local friction parameters without resorting to solve the full set of equations. Sub-grid technology essentially allows velocity to be rationally and efficiently determined at the sub-grid level. This salient feature enables coastal flooding to be addressed in a single cross-scale model from the ocean to the upstream river channel without overly refining the grid resolution. to this end, high-resolution Digital Elevation Models (DEMs) were developed utilizing GIS from Lidar-derived topography for incorporation into a sub-grid model, for research into the plethora of practical research applications related to urban inundation in New York City.;SELFE large-scale storm tide simulations were successfully conducted for 2012 Hurricane Sandy using both the North American Regional Reanalysis (NARR), and the Regional Atmospheric Modeling System (RAMS) atmospheric hindcast model results as atmospheric inputs. Overall statistics using the 24km resolution NARR inputs observed an average R2 value of 0.8994, a relative error of 11.77%, and a root-mean-squared error of 32.69cm for 10 NOAA observation stations. The 4km RAMS inputs performed noticeably better at all 10 stations with aggregate statistics yielding an average R2 value of 0.9402, a relative error of 4.08%, and a rootmean-squared error of 19.22 cm. Since the RAMS atmospheric inputs possessed a higher spatial and temporal resolution than the NARR inputs for air pressure and wind speed, it was concluded that generally superior storm tide predictions could be expected from utilizing more reliable or better resolution atmospheric forecast products.;UnTRIM2 results were obtained via sub-grid simulation of 2012 Hurricane Sandy in the New York Harbor with high-resolution topography and building heights embedded in the model sub-grid for New York City. Model performance was assessed via comparison with various verified field measurements: (1) Temporal comparison of NOAA and USGS permanent water level gauges, (2) USGS rapid deployment water level gauges, along with a spatial inundation comparison using (3) USGS-collected high water marks, (4) FEMA-collected data regarding inundated schools, (5) calculated area and distance differentials using FEMA's maximum extent of inundation map, and (6) known locations of inundated subway entrances. Temporal results verified the effectiveness of the sub-grid model's wetting and drying scheme via seven over land rapid deployment gauges installed and collected by the USGS with a mean R2 of 0.9568, a relative error of 3.83%, and a root-mean-squared error of 18.15cm.;Spatial verification of the inundation depths predicted by the UnTRIM 2 model were addressed by comparison with 73 high water mark measurements collected by the USGS and by 80 FEMA-reported water level thicknesses at inundated schools throughout the sub-grid domain separated by state. Average statistics for the 73 USGS-recorded high water marks for New York and New Jersey were: 0.120+/-0.085m and 0.347+/-0.256m for root-mean-squared error +/- standard deviation, respectively. The larger differences and errors reported in the point to point comparisons for New Jersey relative to New York were largely due to the lack building representation in the sub-grid DEM for the New Jersey side of the Hudson River, and was a significant indication that the representation of buildings as a physical impediment to fluid flow is critical to urban inundation modeling.;A maximum difference threshold was imposed for distance and area comparisons with FEMA's Hurricane Sandy flood map using the average distance differential rounded to 40m. This was done to minimize the impact of missing or added infrastructure such as highway overpasses along with Lidar-derived data limitations of physical impediments to fluid flow not accounted for in the model's DEM. The difference in the absolute mean distance between the maximum extent predicted by the street-level sub-grid model and the FEMA maximum inundation observation was 21.207m or ≈4 sub-grid pixels at 5m resolution for the entire sub-grid domain. The final area comparison resulted in an 85.17% area (49,253,687m 2) spatial match, with 7.57% area (4,376,726m2) representing model over-prediction, and under-prediction area accounting for 7.27% (4,202,376m 2), with differences being attributed to lack of building representation in the FEMA maximum inundation map. Additionally, the implementation of the FEMA's spatial flood map data as a "bathtub" model derivative product of USGS interpolated high water marks and elevation data without regard for strong water current velocities or estuarine circulation can also account for regions with significant discrepancies.;Keywords: Sub-Grid, Stotul Surge, Inundation, New York Harbor, New York City, Jersey City, Conveyance Approach, Unstructured Grids, UnTRIM, SELFE, Lidar-Derived Topography.

DOI

https://dx.doi.org/doi:10.25773/v5-15yy-8z84

Rights

© The Author

Download

Share Feedback

Share

COinS