• Arcement, G. J., and V. R. Schneider, 1989: Guide for selecting Manning’s roughness coefficients for natural channels and flood plains. U.S. Geological Survey Water Supply Paper 2339, U.S. Geological Survey, Denver, CO, 38 pp.

    • Search Google Scholar
    • Export Citation
  • Atkinson, J. H., J. J. Westerink, and J. M. Hervouet, 2004: Similarities between the wave equation and the quasi-bubble solutions to the shallow water equations. Int. J. Numer. Methods Fluids, 45 , 689714.

    • Search Google Scholar
    • Export Citation
  • Barnes, H. H., 1967: Roughness characteristics of natural channels. U.S. Geological Survey Water Supply Paper 1849, U.S. Geological Survey, Washington, DC, 213 pp.

    • Search Google Scholar
    • Export Citation
  • Blain, C. A., J. J. Westerink, and R. A. Luettich, 1994: The influence of domain size on the response characteristics of a hurricane storm surge model. J. Geophys. Res., 99 , (C9). 1846718479.

    • Search Google Scholar
    • Export Citation
  • Blain, C. A., J. J. Westerink, and R. A. Luettich, 1998: Grid convergence studies for the prediction of hurricane storm surge. Int. J. Numer. Methods Fluids, 26 , 369401.

    • Search Google Scholar
    • Export Citation
  • British Oceanographic Data Centre, 2003: General Bathymetric Chart of the Oceans, Centenary Edition. [Available online at http://www.bodc.ac.uk/products/bodc_products/gebco/].

    • Search Google Scholar
    • Export Citation
  • Cardone, V. J., and A. T. Cox, 2007: Tropical cyclone wind field forcing for surge models: Critical issues and sensitivities. Nat. Hazards, 51 , 2947. doi:10.1007/s11069-009-9369-0.

    • Search Google Scholar
    • Export Citation
  • Cardone, V. J., A. T. Cox, and G. Z. Forristall, 2007: Hindcasts of winds, waves and currents in the northern Gulf of Mexico in Hurricanes Katrina (2005) and Rita (2005). OTC 18652, Proc. 2007 Offshore Technology Conf., Houston, TX.

    • Search Google Scholar
    • Export Citation
  • Chow, V. T., 1959: Open-Channel Hydraulics. McGraw-Hill Book Company, 680 pp.

  • Cox, A. T., J. A. Greenwood, V. J. Cardone, and V. R. Swail, 1995: An interactive objective kinematic analysis system. Proc. Fourth Int. Workshop on Wave Hindcasting and Forecasting, Banff, Alberta, Canada, Atmospheric Environment Service, 109–118.

    • Search Google Scholar
    • Export Citation
  • Dawson, C., J. J. Westerink, J. C. Feyen, and D. Pothina, 2006: Continuous, discontinuous and coupled discontinuous-continuous Galerkin finite element methods for the shallow water equations. Int. J. Numer. Methods Fluids, 52 , 6388.

    • Search Google Scholar
    • Export Citation
  • Dietrich, J. C., and Coauthors, 2010: A high-resolution coupled riverine flow, tide, wind, wind wave, and storm surge model for southern Louisiana and Mississippi. Part II: Synoptic description and analysis of Hurricanes Katrina and Rita. Mon. Wea. Rev., 138 , 378404.

    • Search Google Scholar
    • Export Citation
  • Ebersole, B. A., J. J. Westerink, D. T. Resio, and R. G. Dean, 2007: Performance evaluation of the New Orleans and Southeast Louisiana Hurricane Protection System, Volume IV—The storm. Final Report of the Interagency Performance Evaluation Task Force, U.S. Army Corps of Engineers, Washington, DC, 263 pp.

    • Search Google Scholar
    • Export Citation
  • Garratt, J. R., 1977: Review of drag coefficients over oceans and continents. Mon. Wea. Rev., 105 , 915929.

  • Garster, J. K., B. Bergen, and D. Zilkoski, 2007: Performance evaluation of the New Orleans and Southeast Louisiana Hurricane Protection System, Volume II—Geodetic vertical and water level datums. Final Report of the Interagency Performance Evaluation Task Force, U.S. Army Corps of Engineers, Washington, DC, 157 pp.

    • Search Google Scholar
    • Export Citation
  • Gunther, H., 2005: WAM cycle 4.5 version 2.0. Institute for Coastal Research, GKSS Research Centre, Geesthacht, Germany, 38 pp.

  • Hagen, S. C., J. J. Westerink, R. L. Kolar, and O. Horstmann, 2001: Two dimensional unstructured mesh generation for tidal models. Int. J. Numer. Methods Fluids, 35 , 669686.

    • Search Google Scholar
    • Export Citation
  • Hartley, S., R. Pace III, J. B. Johnston, M. Swann, C. O’Neil, L. Handley, and L. Smith, 2000: A GAP analysis of Louisiana: Final report and data. U.S. Department of the Interior, U.S. Geological Survey, Lafayette, LA, 588 pp.

    • Search Google Scholar
    • Export Citation
  • Hendershott, M. C., 1981: Long waves and ocean tides. Evolution of Physical Oceanography, B. A. Warren and C. Wunsch, Eds., MIT Press, 292–341.

    • Search Google Scholar
    • Export Citation
  • Holland, G., 1980: An analytic model of the wind and pressure profiles in hurricanes. Mon. Wea. Rev., 108 , 12121218.

  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437471.

  • Komen, G., L. Cavaleri, M. Donelan, K. Hasselmann, S. Hasselmann, and P. A. E. M. Janssen, 1994: Dynamics and Modeling of Ocean Waves. Cambridge University Press, 560 pp.

    • Search Google Scholar
    • Export Citation
  • Kubatko, E. J., S. Bunya, C. Dawson, J. J. Westerink, and C. Mirabito, 2009: A performance comparison of continuous and discontinuous finite element shallow water models. J. Sci. Comput., 40 , 315339.

    • Search Google Scholar
    • Export Citation
  • Le Provost, C., F. Lyard, J. Molines, M. Genco, and F. Rabilloud, 1998: A hydrodynamic ocean tide model improved by assimilating a satellite altimeter-derived data set. J. Geophys. Res., 103 , 55135529.

    • Search Google Scholar
    • Export Citation
  • Louisiana State University, cited. 2004: Louisiana Lidar. [Available online at http://atlas.lsu.edu/lidar/].

  • Luettich, R. A., and J. J. Westerink, 2004: Formulation and numerical implementation of the 2D/3D ADCIRC finite element model version 44.XX. 74 pp. [Available online at http://adcirc.org/adcirc_theory_2004_12_08.pdf].

    • Search Google Scholar
    • Export Citation
  • Mississippi Automated Resource Information System, cited. 2006: Mississippi Island 10 meter by 10 meter DEM. [Available online at http://www.maris.state.ms.us/HTM/DownloadData/DEM.html].

    • Search Google Scholar
    • Export Citation
  • McGee, B. D., B. B. Goree, R. W. Tollett, B. K. Woodward, and W. H. Kress, 2006: Hurricane Rita surge data, southwestern Louisiana and southeastern Texas, September to November 2005. U.S. Geological Survey Data Series 220. [Available online at http://pubs.water.usgs.gov/ds220].

    • Search Google Scholar
    • Export Citation
  • Mukai, A., J. J. Westerink, R. Luettich Jr., and D. Mark, 2002: Eastcoast 2001: A tidal constituent database for the Western North Atlantic, Gulf of Mexico, and Caribbean Sea. Tech. Rep. ERDC/CHL TR-02-24, U.S. Army Corps of Engineers, 201 pp. [Available from ERDC Vicksburg (WES), U.S. Army Engineer Waterways Experiment Station (WES), ATTN: ERDC-ITL-K, 3909 Halls Ferry Rd., Vicksburg, MS 39180-6199].

    • Search Google Scholar
    • Export Citation
  • National Geophysical Data Center, 1988: Data announcement 88-MGG-02, digital relief of the surface of the Earth. National Oceanic and Atmospheric Administration, Boulder, CO. [Available online at http://www.ngdc.noaa.gov/mgg/global/etopo5.html].

    • Search Google Scholar
    • Export Citation
  • National Ocean Service, 1997: Hydrographic survey digital database. Vol. 1, 3rd ed. National Oceanic and Atmospheric Administration.

  • Powell, M., S. Houston, and T. Reinhold, 1996: Hurricane Andrew’s landfall in South Florida. Part I: Standardizing measurements for documentation of surface wind fields. Wea. Forecasting, 11 , 304328.

    • Search Google Scholar
    • Export Citation
  • Powell, M., S. Houston, L. Amat, and N. Morrisseau-Leroy, 1998: The HRD real-time hurricane wind analysis system. J. Wind Eng. Ind. Aerodyn., 77–78 , 5364.

    • Search Google Scholar
    • Export Citation
  • Powell, M. D., P. J. Vickery, and T. A. Reinhold, 2003: Reduced drag coefficient for high wind speeds in tropical cyclones. Nature, 422 , 279283.

    • Search Google Scholar
    • Export Citation
  • Powell, M. D., and Coauthors, 2008: Reconstruction of Hurricane Katrina’s wind fields for storm surge and wave hindcasting. Ocean Eng., in press, doi:10.1016/j.oceaneng.2009.08.014.

    • Search Google Scholar
    • Export Citation
  • Resio, D. T., 2007: White paper on estimating hurricane inundation probabilities. U.S. Army Engineering Research and Development Center, Vicksburg, MS, 125 pp.

    • Search Google Scholar
    • Export Citation
  • Sheremet, A., A. J. Mehta, B. Liu, and G. W. Stone, 2005: Wave-sediment interaction on a muddy inner shelf during Hurricane Claudette. Estuarine Coastal Shelf Sci., 63 , 225233.

    • Search Google Scholar
    • Export Citation
  • Smith, J. M., 2000: Benchmark tests of STWAVE. Proc. Sixth Int. Workshop on Wave Hindcasting and Forecasting, Monterey, CA, Environment Canada, 369–379.

    • Search Google Scholar
    • Export Citation
  • Smith, J. M., A. R. Sherlock, and D. T. Resio, 2001: STWAVE: Steady-state spectral wave model user’s manual for STWAVE, version 3.0. USACE, Engineer Research and Development Center, Tech. Rep. ERDC/CHL SR-01-1, Vicksburg, MS, 81 pp. [Available online at http://chl.erdc.usace.army.mil/Media/2/4/4/erdc-chl-sr-01-11.pdf].

    • Search Google Scholar
    • Export Citation
  • Smith, S. J., and J. M. Smith, 2001: Numerical modeling of waves at Ponce de Leon Inlet, Florida. J. Waterw. Port Coastal Ocean Div., 127 (3) 176184.

    • Search Google Scholar
    • Export Citation
  • Stone, G. W., A. Sheremet, X. Zhang, Q. He, B. Liu, and B. Strong, 2003: Landfall of two tropical systems seven days apart along south-central Louisiana, USA. Proc. Coastal Sediments ’03, Clearwater Beach, FL, University of South Florida/USGS/U.S. Army Corps of Engineers, 333–334.

    • Search Google Scholar
    • Export Citation
  • Thompson, E. F., and V. J. Cardone, 1996: Practical modeling of hurricane surface wind fields. J. Waterw. Port Coastal Ocean Div., 122 , 195205.

    • Search Google Scholar
    • Export Citation
  • Thompson, E. F., J. M. Smith, and H. C. Miller, 2004: Wave transformation modeling at Cape Fear River Entrance, North Carolina. J. Coastal Res., 20 (4) 11351154.

    • Search Google Scholar
    • Export Citation
  • URS, 2006a: Final coastal and riverine high-water marks collection for Hurricane Katrina in Louisiana. FEMA-1603-DR-LA, Task Orders 412 and 419, Federal Emergency Management Agency, Washington, DC, 76 pp.

    • Search Google Scholar
    • Export Citation
  • URS, 2006b: Final coastal and riverine high-water marks collection for Hurricane Rita in Louisiana. FEMA-1603-DR-LA, Task Orders 445 and 450, Federal Emergency Management Agency, Washington, DC, 79 pp.

    • Search Google Scholar
    • Export Citation
  • URS, 2008: Mississippi coastal analysis project compiled reports of HMTAP. Task Order 18, Federal Emergency Management Agency, Washington, DC, Vol. 1, 376 pp.

    • Search Google Scholar
    • Export Citation
  • U.S. Department of Defense, 1999: Digital nautical chart. National Imagery Mapping Agency, Washington, DC.

  • Villea, F. J., 2005: Mississippi GAP analysis project. GAP Analysis Bull. 13, U.S. Department of the Interior, U.S. Geological Survey. [Available online at http://www.gap.uidaho.edu/bulletins/13/Mississippi.htm].

    • Search Google Scholar
    • Export Citation
  • Vogelmann, J. E., S. M. Howard, L. Yang, C. R. Larson, B. K. Wylie, and N. Van Driel, 2001: Completion of the 1990s National Land Cover Data Set for the conterminous United States from Landsat thematic mapper data and ancillary data sources. Photogramm. Eng. Remote Sens., 67 , 650652.

    • Search Google Scholar
    • Export Citation
  • Westerink, J. J., R. A. Luettich, and J. C. Muccino, 1994: Modeling tides in the Western North Atlantic using unstructured graded grids. Tellus, 46A , 178199.

    • Search Google Scholar
    • Export Citation
  • Westerink, J. J., and Coauthors, 2008: A basin-to-channel-scale unstructured grid hurricane storm surge model applied to southern Louisiana. Mon. Wea. Rev., 136 , 833864.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    WAM model domain shown in red and nested STWAVE model domains shown in blue. In order from west to east, the five STWAVE domains are W, S, LP, SE, and MS-AL, as described in Table 1.

  • View in gallery

    ADCIRC SL15 model domain with bathymetry (m). Geographic locations of interest are indicated by the numbers identified in Table 2.

  • View in gallery

    Detail of the SL15 domain across southern Louisiana and Mississippi with bathymetry and topography [m relative to NAVD88 (2004.65)] with raised features such as levees, railroads, and highways shown in brown. Geographic locations of interest are indicated by numbers identified in Table 2.

  • View in gallery

    As in Fig. 3, but across southwestern Louisiana.

  • View in gallery

    As in Fig. 3, but across southeastern Louisiana and Mississippi.

  • View in gallery

    Detail of the SL15 grid across southern Louisiana and Mississippi with finite-element sizes shown in meters.

  • View in gallery

    Detail of the applied Manning n roughness coefficients for southern Louisiana.

  • View in gallery

    Sample of the applied directional wind reduction factor for uniform steady southerly winds for southern Louisiana. The coastline is outlined in white.

  • View in gallery

    Locations of the six USACE stations with stage–flow relationships that were compared to the computed water levels in Fig. 10. In numerical order, the six stations are Baton Rouge, Donaldsonville, New Orleans, Alliance, Empire, and Venice.

  • View in gallery

    Stage–flow relationships at six USACE stations along the Mississippi River. Measured data is shown as scatter points with associated best-fit curves. The predicted data is shown as connected blue dots.

  • View in gallery

    Comparison of amplitudes and phases as measured by NOAA and predicted by the SL15 model: (left) amplitudes and (right) phases. Each row of figures represents a region as indicated in Table 11 with difference estimates given in Table 13.

  • View in gallery

    Locations of the deep-water NDBC buoys used in the analysis of Hurricanes Katrina and Rita with offshore buoy identifier numbers.

  • View in gallery

    Wind speeds and directions during Hurricane Katrina at four offshore NDBC buoys with buoy identifiers. The measured data is shown with red dots, while the predicted results are shown with black lines.

  • View in gallery

    Wave heights and periods during Hurricane Katrina at six NDBC buoys with identifiers. The measured data is shown with red dots, while the predicted results are shown with black lines. The first four rows show comparisons to WAM results at selected offshore buoys, while the last two rows show comparison to STWAVE results at available coastal stations.

  • View in gallery

    Locations of the nine USACE, NOS, and NWS stations with hydrograph data for Hurricane Katrina. The nine stations are 1) Pass Manchac, 2) Bayou LaBranche, 3) Lake Pontchartrain Midlake Causeway, 4) 17th Street Canal, 5) Little Irish Bayou, 6) the IHNC Lock Staff Gauge, 7) Southwest Pass, 8) Mississippi River at Carrollton, and 9) Grand Isle. Colors indicate the differences between the modeled and measured peak surge. Green points indicate a match within 0.5 m. Red, orange, yellow, and light green circles indicate overpredictions by the model; green, blue, dark blue, and purple circles indicate underpredictions. The clear circle at station 5 indicates an incomplete hydrograph that does not allow for a peak point comparison.

  • View in gallery

    Hydrographs for the nine USACE, NOS, and NWS stations during Hurricane Katrina. The black lines are the computed water levels from the ADCIRC SL15 model, while the red lines are the measured data.

  • View in gallery

    Locations of USACE HWMs for Hurricane Katrina. Colors indicate the difference between the maximum computed water elevation from the ADCIRC SL15 hindcast and the measured high water mark. Green points indicate a match within 0.5 m. Red, orange, yellow, and light green circles indicate overpredictions by the model; green, blue, dark blue, and purple circles indicate underpredictions.

  • View in gallery

    As in Fig. 17, but for locations of URS HWMs for Hurricane Katrina.

  • View in gallery

    Comparisons between observed USACE high water marks and ADCIRC maximum surges during Hurricane Katrina at 206 locations shown in Fig. 17. Green points indicate a match within 0.5 m. Red, orange, yellow, and light green circles indicate overpredictions by the model; green, blue, dark blue, and purple circles indicate underpredictions. The slope of the best-fit line through all points is 0.99 and R2 value is 0.92.

  • View in gallery

    As in Fig. 19, but for comparisons between observed URS high water marks and ADCIRC maximum surges during Hurricane Katrina at 193 locations shown in Fig. 18. The slope of the best-fit line through all points is 1.02 and R2 value is 0.94.

  • View in gallery

    Wind speeds and directions during Hurricane Rita at four offshore NDBC buoys with buoy identifiers. The measured data is shown with red dots, while the predicted results are shown with black lines.

  • View in gallery

    Wave heights and periods during Hurricane Rita at five NDBC buoys with identifiers. The measured data is shown with red dots, while the predicted results are shown with black lines. The first four rows show WAM results at selected offshore buoys, while the last row shows STWAVE results at the available coastal station.

  • View in gallery

    Locations of the 23 USGS stations for Hurricane Rita. Colors indicate the difference between the maximum water elevation from the ADCIRC SL15 hindcast and the maximum water level from the USGS hydrograph data. Green points indicate a match within 0.5 m. Red, orange, yellow, and light green circles indicate overpredictions by the model; green, blue, dark blue, and purple circles indicate underpredictions. White points indicate stations where ADCIRC did not simulate storm surge.

  • View in gallery

    Hydrographs at the first 12 USGS stations for Hurricane Rita. The black lines are the computed water levels from the ADCIRC SL15 model, while the red dots are the measured data.

  • View in gallery

    As in Fig. 24, but for the last 11 USGS stations for Hurricane Rita.

  • View in gallery

    Locations of the 80 HWMs obtained from URS for Hurricane Rita. Colors at each location indicate the difference between the maximum elevation from the ADCIRC SL15 hindcast and the URS HWM. Green points indicate a match within 0.5 m. Red, orange, yellow, and light green circles indicate overpredictions by the model; green, blue, dark blue, and purple circles indicate underpredictions.

  • View in gallery

    Scatterplot of HWMs for Hurricane Rita. Green points indicate a match within 0.5 m. Red, orange, yellow, and light green circles indicate overpredictions by the model; green, blue, dark blue, and purple circles indicate underpredictions. The slope of the best-fit line through all points is 0.97 and the R2 value is 0.77.

  • View in gallery

    Scatterplot of HWMs for Hurricane Rita without data in Vermilion Bay. Green points indicate a match within 0.5 m. Red, orange, yellow, and light green circles indicate overpredictions by the model; green, blue, dark blue, and purple circles indicate underpredictions. The slope of the best-fit line through all points is 1.04 and the R2 value is 0.87.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 832 761 53
PDF Downloads 614 552 35

A High-Resolution Coupled Riverine Flow, Tide, Wind, Wind Wave, and Storm Surge Model for Southern Louisiana and Mississippi. Part I: Model Development and Validation

View More View Less
  • * Department of Civil Engineering and Geological Sciences, University of Notre Dame, Notre Dame, Indiana
  • | + Coastal Hydraulics Laboratory, U.S. Army Engineer Research and Development Center, Vicksburg, Mississippi
  • | # Arcadis, Inc., Denver, Colorado
  • | @ Institute of Marine Sciences, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina
  • | 5 Institute for Computational Engineering and Sciences, The University of Texas at Austin, Austin, Texas
  • | * *Oceanweather, Inc., Cos Cob, Connecticut
  • | ++ NOAA/Atlantic Oceanographic and Meteorological Laboratories/Hurricane Research Division, Miami, Florida
© Get Permissions
Full access

Abstract

A coupled system of wind, wind wave, and coastal circulation models has been implemented for southern Louisiana and Mississippi to simulate riverine flows, tides, wind waves, and hurricane storm surge in the region. The system combines the NOAA Hurricane Research Division Wind Analysis System (H*WIND) and the Interactive Objective Kinematic Analysis (IOKA) kinematic wind analyses, the Wave Model (WAM) offshore and Steady-State Irregular Wave (STWAVE) nearshore wind wave models, and the Advanced Circulation (ADCIRC) basin to channel-scale unstructured grid circulation model. The system emphasizes a high-resolution (down to 50 m) representation of the geometry, bathymetry, and topography; nonlinear coupling of all processes including wind wave radiation stress-induced set up; and objective specification of frictional parameters based on land-cover databases and commonly used parameters. Riverine flows and tides are validated for no storm conditions, while winds, wind waves, hydrographs, and high water marks are validated for Hurricanes Katrina and Rita.

## Current affiliation: Department of Systems Innovation, The University of Tokyo, Tokyo, Japan.

Corresponding author address: J. J. Westerink, Department of Civil Engineering and Geological Sciences, University of Notre Dame, 156 Fitzpatrick Hall, Notre Dame, IN 46556. Email: jjw@nd.edu

Abstract

A coupled system of wind, wind wave, and coastal circulation models has been implemented for southern Louisiana and Mississippi to simulate riverine flows, tides, wind waves, and hurricane storm surge in the region. The system combines the NOAA Hurricane Research Division Wind Analysis System (H*WIND) and the Interactive Objective Kinematic Analysis (IOKA) kinematic wind analyses, the Wave Model (WAM) offshore and Steady-State Irregular Wave (STWAVE) nearshore wind wave models, and the Advanced Circulation (ADCIRC) basin to channel-scale unstructured grid circulation model. The system emphasizes a high-resolution (down to 50 m) representation of the geometry, bathymetry, and topography; nonlinear coupling of all processes including wind wave radiation stress-induced set up; and objective specification of frictional parameters based on land-cover databases and commonly used parameters. Riverine flows and tides are validated for no storm conditions, while winds, wind waves, hydrographs, and high water marks are validated for Hurricanes Katrina and Rita.

## Current affiliation: Department of Systems Innovation, The University of Tokyo, Tokyo, Japan.

Corresponding author address: J. J. Westerink, Department of Civil Engineering and Geological Sciences, University of Notre Dame, 156 Fitzpatrick Hall, Notre Dame, IN 46556. Email: jjw@nd.edu

1. Introduction

Coastal Louisiana and Mississippi are especially prone to large hurricanes because of their geographic location in the north-central Gulf of Mexico. Between 1941 and 2008, the central Gulf was impacted by 16 major hurricanes including storms in 1941, 1957 (Audrey), 1964 (Hilda), 1965 (Betsy), 1969 (Camille), 1974 (Carmen), 1979 (Frederic), 1992 (Andrew), 1995 (Opal), 2002 (Lili), 2004 (Ivan), 2005 (Dennis, Katrina, and Rita), and most recently in 2008 (Gustav and Ike). It is estimated that this region is more than twice as likely to see a major Gulf hurricane compared to the adjacent coasts of Texas and Florida (Resio 2007). Wind-driven coastal surge from these large hurricanes was the most important contributor to devastating regional flooding, although maximum high water levels were also influenced by atmospheric pressure, tides, riverine currents, waves, and rainfall.

The central Gulf is not only statistically susceptible to more frequent hurricanes, but portions of this varied geographic system are vulnerable to developing especially large storm surge for a given set of storm characteristics because of the local geographic configuration. In particular, the east bank of the Mississippi River in southeastern Louisiana is characterized by a protruding delta on the Mississippi–Alabama shelf; the river itself; barrier islands; extensive levee, raised road, and railroad systems; low-lying topography; and large interconnected shallow lakes. Many of these features tend to amplify surge as water is blown from both the east and the south onto the shelf and then blocked by the delta, river banks, levees, and railroad beds. The regional surge in the lower Mississippi River is often propagated up the river, reaching New Orleans within hours. While the state of Mississippi is topographically more varied than Louisiana, with shallow estuaries and low-lying riverine basins interspersed with higher areas including a system of barrier islands lying to the south, Mississippi is also dramatically affected by the Mississippi River’s protrusion onto the shallow continental shelf. In fact, Pass Christian, Mississippi, experienced the largest storm surge ever recorded in the United States during Hurricane Katrina (Ebersole et al. 2007). Finally, western Louisiana is characterized by an east–west coastline, large inland lakes, and extensive low-lying wetlands. These features tend to diminish surge heights because only the southerly winds in the right center quadrant of the storm effectively push water against the coast, and the extensive low-lying wetlands may attenuate transient surges in this area.

To model coastal surge in this complex region, we must include all significant flow processes, accurately define the physical system, numerically resolve the system and the energetic flows, and apply accurate algorithms to solve the resulting mathematical model. The goal is to implement a modeling capability that represents the basic physics of the system as it is observed and does not require ad hoc model tuning of subgrid-scale coefficients, forcing functions, and/or boundary conditions.

The processes that affect storm surge inundation include winds, air–sea momentum transfer, atmospheric pressure, wind-driven waves, riverine flows, tides, and friction. Wind is the driving force of both wind waves and surge, and the characterization of the marine winds is paramount to obtaining accurate surge predictions. Wind wave generation and propagation, subsequent depth-limited breaking, and dissipation by vegetation in the nearshore or floodplain, and the associated transfer of the wind wave momentum through wave radiation stress gradient forcing, influence storm surge elevations and currents and modify the peak surge, the time of arrival of the peak surge, and drawdown. Water levels, currents, and wind waves affect the atmospheric boundary layer and the air–sea momentum transfer while water levels and currents affect the generation and transformation of waves. Riverine flows not only affect overall water levels, but can also affect the propagation of wind waves, tides, and surge up the rivers. Although tides are modest in the region and dominated by less energetic diurnal tides, they modify water levels and can do so nonlinearly. We consider the full nonlinear interaction of these processes to simulate wave and water level conditions throughout the domain.

Tides, waves, and surge are influenced by both basin-scale and local-scale geometric features and flow gradients. Astronomical tides in the Gulf of Mexico are affected by basin-wide generation and shelf dissipation processes, while inland propagation of these tides is affected by the details of the connecting channels and marshes. Storm surge in Lake Pontchartrain depends not only on local setup but also on the high-volume inflows from Lake Borgne through the Rigolets and Chef Menteur pass, and over the interlake marshes. In turn, the Lake Pontchartrain–Lake Borgne storm surge flow exchange depends on the water pushed onto the Mississippi–Alabama shelf, wind wave breaking-induced setup, the level of attenuation of surge into inland Mississippi, and local geometry and bathymetry.

The complexity of the entire system must be accurately defined and computationally resolved in the numerical models in order for the growth, propagation, and attenuation of waves, surge, tides, and riverine flows to be modeled correctly. High grid resolution is necessary when high spatial gradients exist in the geometric and topographic features as well as in the waves, surface elevations, and currents. The emergence of high-density observational data such as lidar and satellite photography has significantly improved the accurate characterization of topography, raised features, and surface roughness. In addition, dense soundings have improved the accurate characterization of the bathymetry.

In this paper, we describe the “SL15” storm surge model for Louisiana and Mississippi, which couples a sequence of well verified and validated wind, short-period wind wave, and coastal circulation models as an atmospheric–hydrodynamic modeling system. We independently validate each process with the available observational data, quantify differences between the component modeled and observational data and when possible estimate the uncertainty in the observational data itself. We stress that the validity of the coupled system relies on its ability to accurately represent the individual components and to then nonlinearly couple these components. We derive error estimates for the modeled river flows, tides, and Hurricane Katrina and Rita winds, waves, and surge levels. In a companion paper, we describe the detailed evolution and physics of winds, waves, surface elevation, and currents during Hurricanes Katrina and Rita (Dietrich et al. 2010).

2. Coupled wind, wind wave, tide, riverine flow model system

a. Kinematic winds

For hindcasting historical storms, we define wind fields using objectively analyzed measurements. Observational data comes from anemometers, airborne and land-based Doppler radar, airborne stepped-frequency microwave radiometer, buoys, ships, aircraft, coastal stations, and satellite measurements. For Katrina, the measured winds in the inner core are assimilated using the National Oceanic and Atmospheric Administration (NOAA) Hurricane Research Division Wind Analysis System (H*WIND) (Powell et al. 1996, 1998) and are then blended with Gulf-scale winds using an Interactive Objective Kinematic Analysis (IOKA) system (Cox et al. 1995; Cardone et al. 2007). H*WIND composites observations of wind velocity relative to the storm’s center and transforms them to a common reference condition of 10-m height, peak 1-min-averaged “sustained” wind speed, and marine exposure. A special set of H*WIND reanalyzed snapshots are available for Katrina (Powell et al. 2008). Peripheral winds are derived from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis project (Kalnay et al. 1996). Before inner-core and peripheral wind fields are blended, the inner core peak sustained winds are transformed to 30-min-average wind speeds using a gust model consistent with the H*WIND system. A final step is to inject local marine data, adjusted to a consistent 10-m elevation and neutral stability using the IOKA system. Lagrangian-based interpolation is used to produce the final wind fields on a regular 0.05° × 0.05° grid with snapshots every 15 min. Hurricane Rita inner-core wind fields are based mainly on TC96 mesoscale model (Thompson and Cardone 1996) solutions blended as described above into peripheral fields using IOKA (Cardone and Cox 2007; Cardone et al. 2007). Both hurricanes’ pressure fields used to drive the atmospheric pressure term in the circulation model are derived using a widely adopted parametric relationship (Holland 1980).

b. Deep-water wind wave model WAM

The Wave Model (WAM) is run to generate deep-water wave fields and directional spectra in a Gulf of Mexico domain. WAM is a third-generation discrete spectral wave model that solves the wave action balance equation and includes source-sink terms, atmospheric input, nonlinear wave–wave interactions, white-capping, bottom friction, and depth-limited wave breaking. The spatial and temporal variation of wave action in frequency and direction is solved over a fixed spatial grid (Komen et al. 1994). WAM has recently undergone major revisions to source term specification, multigrid nesting, and depth-limited breaking (Gunther 2005). The model computes directional wave spectra for 28 discrete frequency bands, and 24 directional bands centered every 15°.

The WAM model domain, shown in Fig. 1, extends over the entire Gulf of Mexico with a grid at 0.05° resolution. It is assumed that the wind waves are generated in the Gulf and that wave energy entering the Gulf and reaching the area of interest through the Florida and Yucatan Straits is minimal. Wave data within and outside of the Gulf indicates that the dominant wave energy is generated within the Gulf, along with the hurricane. The WAM model allows wave energy to propagate out of the Gulf through the Yucatan and Florida straits. The water depth is derived from the General Bathymetric Chart of the Oceans (British Oceanographic Data Centre 2003). The H*WIND–IOKA 30-min-averaged wind fields are linearly interpolated in time and space onto the WAM grid.

c. Nearshore wave model STWAVE

The nearshore wind wave model Steady-State Irregular Wave (STWAVE; Smith 2000; Smith et al. 2001; Smith and Smith 2001; Thompson et al. 2004) is used to generate and transform waves to the shore. STWAVE solves the steady-state conservation of spectral action balance along backward-traced wave rays. The source terms include wind input, nonlinear wave–wave interactions, dissipation within the wave field, and surf-zone breaking. The computed terms include wave propagation and source terms representing energy growth and decay in the spectrum. The assumptions made in STWAVE include a mild bottom slope; negligible wave reflection; steady waves, currents, and winds; linear refraction and shoaling, and a depth-uniform current. STWAVE can be implemented as either a half-plane model, where only waves propagating toward the coast are represented, or a full-plane model, allowing generation and propagation in all directions. Wave breaking in the surf zone limits the maximum wave height based on the local water depth and wave steepness.

Four or five STWAVE grids are used to simulate nearshore and coastal floodplain wind wave propagation and attenuation. These grids, also shown in Fig. 1 and summarized in Table 1, extend across coastal Louisiana, Mississippi, and Alabama. The spatial resolution of each STWAVE grid is 200 m. Bathymetry for all grids is interpolated from the Advanced Circulation (ADCIRC) model grid.

Open-water boundary conditions are obtained by extracting the wave energy spectra from the WAM solutions at the STWAVE boundary nodes. The wind fields are interpolated from the ADCIRC wind fields, which apply land effects to the H*WIND/IOKA marine wind fields. STWAVE is run at 30-min intervals for 2 days. The STWAVE computations include preliminary water levels interpolated from ADCIRC simulations forced only with wind, atmospheric pressure, riverine flows, and tides. Radiation stresses computed with STWAVE are added as input to a subsequent ADCIRC simulation.

d. ADCIRC model

The last component of the system is the ADCIRC unstructured coastal ocean circulation model, which is applied to compute surface water elevation and currents. The ADCIRC model solves the depth-integrated barotropic shallow-water equations in spherical coordinates using a finite-element solution (Luettich and Westerink 2004; Atkinson et al. 2004; Dawson et al. 2006; Westerink et al. 2008). The solution maintains both accuracy and robustness when applied to the wide range of scales of motion and wide range of hydrodynamic balances that exist when computing flows in the deep ocean transitioning to flows in inlets, floodplains, and rivers. The use of an unstructured grid allows for high localized grid resolution where solution gradients are large, and low grid resolution where solution gradients are small, minimizing both local and global error norms for a given computational cost.

e. SL15 domain and grid

The ADCIRC SL15 model domain, shown in Fig. 2, is an evolution of the earlier EC2001 U.S. East Coast and Gulf of Mexico astronomical tide model and the S08 southern Louisiana storm surge model (Blain et al. 1994; Mukai et al. 2002; Westerink et al. 2008). These models incorporate the western North Atlantic Ocean, the Gulf of Mexico, and the Caribbean Sea to allow for full dynamic coupling between oceans, continental shelves, and the coastal floodplain without necessitating that these complicated couplings be defined in the boundary conditions. The SL15 model extends the coverage of these earlier models to include all the floodplains of southern Louisiana and Mississippi. In addition, improved definitions of features, surface roughness, wave radiation stress, and grid resolution are incorporated. The highly resolved floodplain in the SL15 model extends from Beaumont, Texas, to Mobile Bay, Alabama. Areas in Texas and Alabama are included to allow storm surge that affects Louisiana and Mississippi to realistically attenuate and laterally spread into the adjacent states. In southern Louisiana and Mississippi, the northern land boundary extends along the 10–20-m elevation contours or major hydraulic controls. Details of the domain with bathymetry and topography as well as levees and raised roadways across southern Louisiana can be seen in Figs. 3 –5 with geographic places of interest listed in Table 2.

The computational grid resolves the tidal, wind, atmospheric pressure, and riverine flow forcing functions and flow processes from the ocean basins to the coastal floodplain. Effective resolution of tidal and hurricane response within the basins and on the shelf is determined by tidal wavelength, topographic length scale criteria, and hurricane size criteria (Westerink et al. 1994; Blain et al. 1998; Hagen et al. 2001). The grid applies localized refinement of the coastal floodplains and of the important hydraulic features, down to 50 m in critical channels and conveyances, as shown in Fig. 6. We accommodate the STWAVE forcing function by adding a swath of 50–200-m grid resolution along the coast, over barrier islands, and around Lake Pontchartrain where there are significant gradients in the wave radiation stresses and where forcing of surge through wave transformation is the largest. Barrier islands also need high grid refinement to resolve the very high currents that develop when these features are overtopped.

The SL15 computational grid contains 2 409 635 nodes and 4 721 496 elements. Grid resolution varies from 24 km in the Atlantic Ocean to about 50 m in Louisiana and Mississippi. Unstructured grids can resolve the critical features and the associated local flow processes with orders of magnitude fewer computational nodes than a structured grid.

f. SL15 bathymetry and topography

Geometry, topography, and bathymetry in the SL15 model are defined to replicate the prevailing conditions in August 2005 prior to Hurricane Katrina, with the exception of some of the barrier islands and the area between Lake Pontchartrain and Lake Borgne, which are included as post-Katrina September 2005 configurations. Open-ocean and shelf bathymetric depths are interpolated in order of preference from NOAA’s bathymetric sounding database, the Digital Nautical Charts database, and the 5-minute gridded elevations/bathymetry for the world (ETOPO5) database (National Ocean Service 1997; U.S. Department of Defense 1999; National Geophysical Data Center 1988; Mukai et al. 2002). Inland bathymetry is taken from regional bathymetric surveys from the U.S. Army Corps of Engineers New Orleans District (USACE-MVN).

Topography in the floodplain is obtained predominantly from the Atlas and the Mississippi Coastal Analysis Lidar Projects (Louisiana State University 2004; URS 2008). Where no data is available in the wetlands, the Louisiana Gap Analysis Project (LA-GAP) land-cover data (Hartley et al. 2000) is applied with estimated topographic heights of 0.80 m for marshland and −0.40 m for water. The U.S. Geological Survey (USGS) post-Katrina lidar data is applied to the Chandeleur Islands, and USACE post-Katrina lidar data is used for the Mississippi Sound Islands with the exception of Half Moon Island, Deer Island, and Singing River Island, where Mississippi Automated Resource Information System (MARIS) data is used (Mississippi Automated Resource Information System 2006). Levee and road systems that are barriers to flood propagation are features that fall below the defined grid scale, and represent a nonhydrostatic flow handled as a subgrid-scale weir (Westerink et al. 2008). All federal levees, many local and private levees, and road heights are defined using the USACE-MVN surveys. Road and railroad crown heights in Louisiana are generally taken from the Atlas lidar surveys. Note that the CSX railway between the Rigolets and Chef Menteur Pass in particular is important as a control in the flow between Lake Borgne and Lake Pontchartrain. According to the Atlas lidar surveys, the railway has a height of about 3.5 m. However, CSX railway personnel involved in the reconstruction indicated that the gravel bed was washed out during the storm and that the remaining compacted bed was at no more than 2 m, the elevation incorporated into the model. In addition, US 90 sustained some damage and estimates of the lowered values are made.

g. Vertical datum, LMSL, and steric water level adjustments

The North American Vertical Datum of 1988 (NAVD88) updated to the 2004.65 epoch, NAVD88 (2004.65) is used as the vertical reference. Topography is available relative to the original epoch, NAVD88, while federal levees and high water mark (HWM) data are available relative to NAVD88 (2004.65). Garster et al. (2007) computed the adjustment from local mean sea level (LMSL) to NAVD88 (2004.65) at 12 stations throughout southern Louisiana. The average adjustment at the 11 reliable stations is 0.134 m. Additionally, an examination of the datums at NOAA stations in the region reveals that LMSL regionally lies above mean lower low water (MLLW) by 0.152 m. Thus, bathymetric data, referenced to MLLW, has been regionally adjusted to NAVD88 (2004.65) by adding 0.018 m.

The computations themselves are referenced to NAVD88 (2004.65) by adding 0.134 m to the baseline LMSL reference of the model. Because the computations are barotropic, it is also necessary to account for the annual fluctuation in sea level due to thermal expansion of the upper layers of the Gulf of Mexico and by other effects. NOAA long-term stations at Dauphin Island, Mississippi, Grand Isle, Louisiana, and Sabine Pass, Texas, indicate that the increase in surface elevations is bimodal with station-averaged maximum mid-September water levels increasing to 0.158 m above the annual average (more information is available online at http://tidesandcurrents.noaa.gov/sltrends/sltrends.html). This expansion is also captured in harmonically decomposed tidal records by the long-term Sa and Ssa constituents, which show an average regional combined amplitude of 0.15 m with a standard deviation of 0.03 m. To make the seasonal sea surface adjustment for a specific storm, the regional long-term sea level station data is used at the date of landfall. Thus for Katrina, which occurred in late August, sea surface level increase above the annual average is regionally estimated as 0.10 m above LMSL, while for Rita, which made landfall on 24 September, the estimated increase is 0.15 m. Initial water levels in the model are therefore raised at the start of the computation with the combined average regional difference between LMSL and NAVD88 (2004.65) in addition to the steric increase. For Katrina, this adjustment equals 0.13 m + 0.10 m = 0.23 m. For Rita the adjustment equals 0.13 m + 0.15 m = 0.28 m.

h. Hydraulic friction

Bottom friction is computed by the standard quadratic parameterization of bottom stress using a Manning n formulation. Nodal Manning n coefficients are spatially assigned using land-cover definitions from the USGS LA-GAP in Louisiana, USGS Mississippi Gap Analysis Project (MS-GAP), and the USGS National Land Cover Data (NLCD) in Texas and Alabama (Hartley et al. 2000; Villea 2005; Vogelmann et al. 2001). The GAP data are preferred because the classification system, particularly in wetlands, is more detailed than the NLCD data. The Manning n associated with these land classifications, presented in Tables 3 –5, are selected or interpolated from standard hydraulic literature (Chow 1959; Barnes 1967; Arcement and Schneider 1989). For the open ocean, large inland lakes, sheltered estuaries, deep straight inlet channels, deep meandering rivers, and shallow meandering channels, n is assigned 0.02, 0.02, 0.025, 0.025, 0.025, and 0.045, respectively. Figure 7 shows the applied Manning n values in southern Louisiana. The bottom friction coefficient Cf has a defined lower limit equal to 0.003 in ocean and shelf waters in order to be consistent with Mukai et al. (2002). Lateral eddy viscosity is uniformly set to 5 m2 s−1 in water and 50 m2 s−1 on land.

i. Riverine forcing

At the Mississippi River at Baton Rouge, Louisiana, and at the Atchafalaya River at Simmesport, Louisiana, inflows are specified with a wave radiation boundary condition that ensures that neither surges nor tides artificially reflect back into the computational domain (Westerink et al. 2008). The river condition is spun up specifying a steady flow with no other forcings in the model, using a 0.5-day hyperbolic ramp, and running for 2.0–4.5 days to reach a dynamic steady state as summarized in Table 6. After this, the river radiation boundary condition is applied and other forcings are spun up. River flow rates for the simulations are specified in Table 7.

j. Tidal forcing

After the rivers have reached equilibrium, tides are spun up in the circulation model. Tides are forced on the Atlantic open-ocean boundary along the 60°W meridian with the seven dominant astronomical tidal constituents and include the diurnal O1, K1, and Q1 constituents and the semidiurnal M2, N2, S2, and K2 constituents, using data from Le Provost’s FES95.2 global model (Le Provost et al. 1998; Mukai et al. 2002). In addition, tidal potential functions are forced within the model domain for the same constituents. Periods, tidal potential constants, and the earth elasticity factors, which reduce the magnitude of the tidal potential forcing due to the earth tides, are listed in Table 8 (Hendershott 1981). Finally, the nodal factor and equilibrium argument for boundary and interior domain forcing tidal constituents are based on the starting time of the simulation. The resonant characteristics of the Gulf of Mexico and Caribbean Sea require a period of model simulation or spinup in order for the initial transients to physically dissipate and dynamically correct tides to be generated. Tidal spinup ramps and run times are detailed in Table 6.

k. Atmospheric and wave forcing

The IOKA–H*WIND wind field analyses provide marine wind exposure at 10-m height and 30-min-averaged winds. The wind surface stress is computed by a standard quadratic air–sea drag law. The air–sea drag coefficient is defined by Garratt’s drag formula, which is based largely on 10-min-averaged wind data (Garratt 1977). The IOKA–H*WIND winds are therefore adjusted to 10-min averages by noting that shorter sampling periods lead to higher-averaged winds and increasing them by a factor of 1.09 as recommended by Cardone. Cardone’s factor leads to almost identical 10-min winds as would be obtained by converting H*WIND peak 1-min winds to 10-min winds using Powell’s recommended conversion factor of 0.89 (Powell et al. 1996). The drag coefficient is limited to 0.0035 to represent sheeting processes. Powell et al. (2003) found upper-limit values based on GPS dropwindsondes as low as 0.0025 although there appears to be strong quadrantal variation, the limit may be higher in outer portions of the storm and values in shallow shelf waters are only now being obtained.

The ADCIRC model corrects the IOKA–H*WIND marine winds to account for land roughness by making directional adjustments to the marine winds depending on upwind roughness, level of local inundation, and the presence of tree canopies (Westerink et al. 2008). The directional wind reduction is based on the USGS NLCD supplemented with GAP land-cover classification raster maps for areas identified as cypress forest, combined with land roughness lengths in Table 9. Wind boundary layer readjustments depend upon roughness conditions upwind of the location. Figure 8 shows sample directional roughness coefficients for steady uniform southerly winds. Furthermore, as inundation takes place, the land roughness elements are submerged and the drag is reduced. Finally, canopied areas are defined where there are deciduous forests, evergreen forests, mixed forests, woody wetlands, or cypress forests. Canopies are assumed to be so high that no water overtops them and thick enough for wind not to penetrate them.

The wind and pressure fields snapshots are applied every 15 min during the periods listed in Table 6 and are linearly interpolated in time between snaps. The STWAVE wave radiation stress fields are read every 30 min during the periods listed in Table 6 and are linearly interpolated in time and space.

l. System performance

The five STWAVE grids and the ADCIRC SL15 grid were run on a CRAY XT3 with 2.6 GHz Opteron processors (Sapphire; see online at http://www.erdc.hpc.mil). The five STWAVE grids were run with a relatively large time interval of 1800 s, and they required 2484 s day−1 of simulation on 96 computational cores. The ADCIRC SL15 grid was run with a relatively small time step of 1 s, and it required 4380 s day−1 of simulation on 256 computational cores. The ADCIRC model wall-clock times reduce linearly as the number of cores is increased (Kubatko et al. 2009).

3. River validation

The representation of the Mississippi River in the SL15 model was validated by comparing measured and predicted stages at stations from Baton Rouge to Venice, Louisiana, shown in Fig. 9. At each station, the USACE-MVN has measured stage–flow data, where water level is matched with the flow rate upriver at Tarbert Landing. Using data from multiple years, a best fit stage–flow curve can be derived at each station, as shown in Fig. 10.

SL15 model stage–flow curves, obtained by running a variety of steady flow rates on the Mississippi River, are also shown in Fig. 10. The model-predicted stages fall within the scatter of the measured data. It is only at the large flow rates that the SL15 model begins to over- or underpredict the stages. Table 10 summarizes the absolute average differences between the SL15 model stages and the measured data-derived best-fit curve. Table 10 also includes the uncertainty in the measured data by computing the absolute differences between the measured data and the measured data-derived best-fit curve. The differences between the SL15 stages and measured data-derived best-fit curve are on the same order as the estimated uncertainties in the measured data.

4. Tidal validation

The tides are weak in the Gulf of Mexico, with mixed semidiurnal and diurnal tides on the Florida shelf up to Apalachicola, Florida; diurnally dominated tides between Panama City Beach, Florida, and Port Fourchon, Louisiana; and mixed tides again being prevalent between Point au Fer Island, Louisiana, and Port Isabel, Texas. Along all three coastlines, the dominant constituents have amplitudes that are less than 0.2–0.4 m.

SL15 modeled tides are validated by comparing them to measurement-derived data at NOAA tidal harmonic constituent stations. These stations are listed in Table 11 and span the Florida Keys to Port Isabel. In Florida and Texas, where the SL15 domain does not include inland waters, stations are selected in open water. In the regions where the SL15 model does resolve inland water bodies, stations are selected in both open water and inland. Model time histories at the selected stations are analyzed harmonically over 60 days using the 23 constituents defined in Table 12.

A comparison is made between the NOAA-measured and the SL15-computed amplitudes and phases for the seven dominant constituents in Fig. 11. Difference bands are defined at 0.025 and 0.05 m for the amplitude plots and 10° and 20° for the phases. For the 10 stations in Florida, the constituents fall very near or inside the difference bands. For the stations in the other regions, the constituents group together and only the phases of the K2 constituent show significant differences.

Table 13 lists the correlation coefficients, R2, for the four groups of NOAA stations. The R2 coefficients are greater than 0.942, indicating an excellent match, with the exception of the non-Florida phases. When the K2 constituent is removed from the analysis, these values increase to greater than 0.937. Note that the K2 constituent is small and difficult to separate from the larger S2 constituent in a harmonic analysis of 60 days.

Table 14 shows the difference statistics between the NOAA-measured and the SL15-computed tidal data for the four groups of NOAA stations. Average difference, average absolute difference, and the standard deviation are shown for the amplitudes and phases. In addition, the dimensionless normalized root-mean-square (rms) difference is computed for the amplitudes and is defined as
i1520-0493-138-2-345-e1
where L is the number of elevation stations within a region, (x1, y1) is the station location, is the computed model elevation amplitude for constituent j, and is the NOAA-measured elevation amplitude for constituent j. In Table 14, the dimensional amplitude differences range from 0.002 to 0.010 m, and the dimensionless amplitude differences range from 0.023 to 0.057. The phase differences range from 1° to 26°. The phase behavior improves when the K2 constituent is excluded from the analysis.

We note that these quantities reflect the differences between the NOAA-measured and the SL15 model harmonic constituents and therefore include the uncertainties in the NOAA-measured data itself. To estimate the uncertainties in the NOAA-measured data, we compare the current (as of March 2007) NOAA-published harmonic data to previously measured and published NOAA harmonic constituent data. The normalized rms amplitude and absolute average phase differences in the NOAA data at stations with multiple measured values are listed in Table 15. Overall the normalized rms amplitude differences range between 0.013 and 0.041, the average phase differences range between 5.8° and 18.4°. The measurement data uncertainties estimated by the differences between the two NOAA datasets can be explained by the shifting geometry–bathymetry of coastal regions and the occurrences of nontidal events including wind-driven events, radiational heating cycles, and riverine discharges. The measurement uncertainties represent 35%–60% of the model-to-measurement amplitude differences for the majority of the constituents. For the model-to-measurement phase differences, the measurement uncertainties account for 50%–80%. The results in Table 15 indicate that a significant portion of the difference between the model and the measurement data can be attributed to uncertainties in the measurements themselves.

5. Hurricane Katrina validation

Hurricane Katrina is incomparable in U.S. recorded events in terms of surge levels and the quality and quantity of recorded data. Wind, wave, and water level data were collected during the event, and extensive postevent surveys of HWMs were made and referenced to NAVD88 (2004.65).

Wind speed and direction data collected during Hurricane Katrina at 12 NDBC buoys, shown in Fig. 12, are used to validate the IOKA–H*WIND wind fields. It should be noted that the NDBC buoy data are assimilated into the IOKA–H*WIND analysis, but that many other sources of data also influenced the analysis. Differences between the IOKA–H*WIND wind and that measured at the buoys is indicative of the analysis fidelity to all the input data. Comparisons at buoys close to the storm track are shown in Fig. 13. The IOKA–H*WIND winds match the oscillations in the wind speeds before the storm, the magnitude of the peak winds, and the rate at which the winds die down after the storm passes the buoys. A one-to-one comparison of available peak wind speeds at 11 buoys shows a best-fit slope of 0.99 and an R2 value of 0.93, indicating a good match between measured and predicted data.

At the same buoys, significant wave heights and peak wave periods are used to validate the WAM model as shown in Fig. 14. WAM matches the timing and magnitude of the peaks at the selected buoys, and a one-to-one comparison of peak significant wave heights at all 12 deep-water buoys shows a best-fit slope of 0.93 and an R2 value of 0.90. Station 42040 misses the quick peak at this buoy as do other wave models. It is unclear if the wind fields are regionally missing features, the models are unable to achieve the maximum wave heights or if the buoy data is biased for the two peak data points at this station. The results of the frequency spectra and the mean wave direction as a function of frequency comparisons have similar trends. Matching energy levels and mean wave directions across the entire frequency range for all NDBC sites show differences that are consistent with the peak data as well as with other third-generation wave models. We note that the peak significant wave height is the square root of the integrated energy spectrum.

STWAVE is validated by comparing computed significant wave heights and peak wave periods to limited measured data at two open-water Louisiana State University (LSU) Coastal Studies Institute stations: CSI05, located south of Isle Dernieres; and CSI06, located south of Timbalier Island (see online at http://wavcis.csi.lsu.edu/csi05.asp?table=WCIS05 and http://www.wavcis.lsu.edu/csi06.asp?table=WCIS06). Comparisons at these two coastal stations are also presented in Fig. 14. At CSI05, the computed wave heights and periods match the qualitative behavior of the storm, and their values lie within the scatter of the recordings. At CSI06, where the station failed during the peak of the storm, the computed wave heights and periods match the run-up to the storm.

During Katrina, the USACE-MVN, NOS, and NWS collected hydrograph data at nine stations shown in Fig. 15. This figure also shows the differences between ADCIRC computed and measured peak surge values at these stations. Figure 16 compares ADCIRC computed water levels against the measured time histories. Water levels at Pass Manchac on the west side of the lake compare to within 0.37 m of the measured values, showing excellent agreement in terms of timing and hydrograph features. The comparison at Bayou LaBranche shows good agreement in the timing of peaks and rising and drainage rates. The discrepancy, which is consistent in time, is attributed to a discrepancy in datum levels. Model results at Midlake in Lake Pontchartrain show two peaks occurring in the lake. The first peak is caused by winds from the north and northeast that pile water against the lake’s south shore, and the second peak is caused by the westerly winds pushing water toward the east side of the lake coupled with the massive intrusion of water from Lake Borgne during the storm’s second landfall. The comparison at the 17th Street Canal indicates that the model is underpredicting peak surge by about 0.6 m, but local Boussinesq models have indicated that there is more wave-driven setup, as much as 0.5 m, which cannot be captured with the current horizontal resolution. The model results at Little Irish Bayou on Lake Pontchartrain show rising water levels that match the recorded levels. Model and measured data at the Inner Harbor Navigation Canal (IHNC) lock staff gauge at the south end of the IHNC are well matched in terms of peak water levels and drawdown rates. The model does show a temporary drawdown prior to a second peak that is not fully matched in the data. This relates to localized drawdown on the west end of Lake Borgne that occurs as the storm passes, coupled with the model underprediction seen on the south side of Lake Pontchartrain. The comparison at Southwest Pass indicates that the modeled tides are well represented in the region and that the peak storm surge is overpredicted by about 0.4 m. The gauge at Carrollton adjacent to New Orleans indicates that the model captures the propagation of tides and surge up the Mississippi River. Finally, the comparison at Grand Isle shows good agreement. We note the excellent comparison of modeled and measured recession rates for stations in the Lake Pontchartrain–Lake Maurepas region, suggesting that the nonforced, but frictionally dominated recession process is well represented as water is withdrawn from these bodies through the Rigolets, Chef Menteur, and through Lake Borgne and off the shelf past the barrier islands.

The USACE collected 206 reliable HWMs and URS/Federal Emergency Management Agency (FEMA) collected 193 reliable HWMs during poststorm surveys with the locations and model to measurement differences shown in Figs. 17 –18, respectively (Ebersole et al. 2007; URS 2006a). The HWMs were collected as indicators of the “still-water levels” and thus did not include the active motion of wind waves but did include the effects of wave setup. The two sets of HWMs offer wide coverage of the impacted region. The overall match is good, with 70% of the USACE HWMs and 73% of the URS/FEMA HWMs matching the model results to within 0.5 m. Missing features, processes, and/or poor grid resolution are associated with the larger differences. For example along the west bank of the Mississippi River within Plaquemines Parish at Socola, Louisiana, as well as up and down river from this location, numerous HWMs within the levee system are substantially underpredicted because of the fact that we do not model levee breaching. Inadequate resolution in the circulation and wave models leads to the underprediction of wave induced setup on the south shore of Lake Pontchartrain as well as other locations with levees and raised roads. Farther inland, the model over- or underpredicts surge unless the area is connected to well-defined inland waterways, which allow surge to flow past or to the HWM locations. For far inland locations adjacent to steep topography, such as up the Pearl River basin, rainfall runoff may have significantly added to the surge levels.

Scatterplots of measured versus predicted HWMs are presented in Figs. 19 –20. For the USACE marks, the slope of the best-fit line is 0.99 and the correlation coefficient R2 is 0.92. For the URS marks, the slope of the best-fit line is 1.02 and R2 equals 0.94. Error statistics for Katrina are summarized in Table 16. For both datasets, the average absolute difference between modeled and measured HWMs is 0.36–0.4 m, and the standard deviation is 0.44–0.48 m. A portion of these differences can be attributed to uncertainties in the measured HWMs themselves. If two or more measured HWMs are hydraulically connected (defined as being within 500 m horizontally, having no barrier in between them, and having computed water levels within 0.1 m), then HWM uncertainties are estimated by examining the differences in these adjacent HWMs. Table 16 indicates that the estimated uncertainties in the measured HWMs are 20%–30% of the differences between the modeled and measured HWMs. When the HWM uncertainties are removed from the predicted to measured differences, then the estimated average absolute model error range is between 0.27 and 0.28 m, and the standard deviation is 0.42–0.44 m.

6. Hurricane Rita validation

Hurricane Rita was a large storm that made landfall at the western edge of Louisiana, with extensive inland penetration. Rita was also rich in both the quality and quantity of recorded data.

Wind data was collected at nine NDBC buoys shown in Fig. 12. Comparisons of wind speeds and directions at selected buoys are shown in Fig. 21. The IOKA winds match the oscillations in the wind speeds before the storm, the magnitude of the peak winds, and the rate at which the winds die down after the storm passes the buoy. The IOKA winds performed similarly at the other buoys, and a one-to-one comparison of peak wind speeds shows a best-fit slope of 0.97 and an R2 equal to 0.96.

At those same deep-water buoys, the significant wave heights and peak wave periods are used to validate WAM, and time series plots at selected stations are shown in Fig. 22. WAM matches the timing and magnitude of the peaks at the buoys, and a comparison of measured and predicted peak significant wave heights at the available nine stations shows a best-fit slope of 0.96 and an R2 value of 0.87.

STWAVE is validated by comparing its computed significant wave heights and peak periods to measured data at coastal station CSI05. As shown in Fig. 22, the model-predicted wave heights and periods lie within the scatter of the recordings. STWAVE computes a “double peak” in the wave heights and periods, because the winds shifted from southeasterly to southwesterly as Rita passed this station.

The USGS collected hydrograph data from 23 water-level sensors positioned as shown in Fig. 23 (McGee et al. 2006). This figure also shows the differences between measured and modeled peak still water levels at these sensors. The model’s ability to represent the drawdown, maximum water levels, and recession is shown in the hydrographs in Figs. 24 –25. Some stations, such as LA11, LA12, LC7, LC8a, LC11, and LC12, were located in regions that are normally dry, and thus only measured water levels above the vertical position at which they were placed. At most stations the features of the measured data are modeled well. At the stations where the recession curve was recorded, the modeled rate of dewatering, which is dominated by a balance between friction and water elevation gradients, is consistent with the observed rates. This indicates that bottom friction within the model provides an accurate representation of the actual role of bottom friction across these complex series of lakes and marshes. This is of critical importance to the accurate representation of inland surge decay in hurricanes such as Hurricane Rita.

At the few stations where the match is poor between the measured and predicted water levels, a lack of resolution is almost always the cause. The inlet into Sabine Pass, near station B15b, lacks the same level of horizontal resolution found elsewhere in the SL15 model. In addition there are vertical referencing uncertainties at this station. Stations LA2 and LA3 do not wet in the simulation and stations LA7 and LA8 flooded too early and by too much, because they are located along small tributaries that cannot be resolved at the 50-m resolution typically used in the model. Station LF3 also has narrow channel-scale connectivity–resolution problems. The model performs well around channels when sufficient resolution is included, such as for stations LC2a and LC2b along the wider Calcasieu Shipping Channel. These stations highlight the importance of resolution, topography, and vertical datum.

The maximum water levels can also be compared to FEMA/URS HWMs (URS 2006b). This analysis uses the 80 HWMs that were due only to storm surge with wave-induced setup and deemed by URS to be of good quality. The locations and model to measurement differences of these HWMs are shown in Fig. 26. The differences are within 0.5 m at 77% of the comparison locations across the state. A scatterplot of the HWMs is shown in Fig. 27. Overall, the slope of the best-fit line through all of the scatter points is 0.97, and the R2 is 0.77. The worst HWM comparisons are a cluster concentrated inside Vermilion Bay and are consistently underpredicted. Vermilion Bay may have problems related to the relatively low grid resolution in the region and/or its viscous muddy bottom (Sheremet et al. 2005; Stone et al. 2003), which may affect surge propagation, wind wave development and attenuation, and/or air–sea momentum transfer. A best-fit line for the 54 data points outside Vermilion Bay is presented in Fig. 28, showing a slope of 1.04, and a much improved R2 of 0.87.

Table 17 gives the average absolute difference between modeled and measured HWMs as 0.31 m, and the standard deviation as 0.40 m. However, both quantities improve when the HWMs near Vermilion Bay are excluded. Accounting for the uncertainty in the HWMs themselves, the estimated model average absolute errors range from 0.16 to 0.21 m with a standard deviation of 0.28–0.35 m.

7. Conclusions

Our coupled river, tide, wind, wind wave, and circulation model for southern Louisiana and Mississippi emphasizes an accurate representation of the physical features with grid resolution down to 50 m, the nonlinear coupling of the multiple processes that contribute to storm surge, an objective specification of frictional parameters that describe dissipation based on USGS GAP and NLCD land use data, wind adjustment based on upwind roughness, and robust and accurate boundary conditions achieved through nested model coupling in the case of the wave computations and through a basin-scale unstructured grid model for the circulation computations. Forcing functions, boundary conditions, geometric, topographic, bathymetric, and surface friction descriptors are defined within the system as they are observed and are not tuned to optimize the model to match observational data for waves or water levels.

The processes are validated separately for riverine flow and tides and concurrently for the hurricane events, validating winds, waves, hydrographs, and HWMs. Flow-stage relationships in the Mississippi River match measured best-fit relationships to within an average of 0.24 m. Tides along the Gulf Coast are also well represented by the model with the dominant diurnal tides being captured with an average absolute difference equal to 0.01 m. During the hurricane events, the kinematic wind analyses accurately represent the measured wind fields with an R2 of 0.93–0.96 while open-water significant wave heights correlate to measured values with R2 equal to 0.87–0.90. The HWMs during Katrina match measurements with an R2 equal to 0.92–0.94 and after accounting for measurement data uncertainties with an estimated average absolute error of 0.27–0.28 m and a standard deviation of 0.42–0.44 m. Rita HWMs match measurements with R2 equal to 0.77–0.87 and, after accounting for uncertainties in the measurement data, with an estimated average absolute error of 0.16–0.21 m and a standard deviation of 0.28–0.35 m. Finally, the hydrographs demonstrate that the model captures both the forced water level rises, and flood recession process even at far inland stations, indicating that friction is correctly represented.

The ability to model waves and water levels correctly is very dependent on providing a high level of grid resolution where gradients in topography, bathymetry, geometry, forcing functions, and elevation and current response functions are significant. Topography, inlets, channels, vertical structures, wave breaking zones, and high current gradient zones all require high levels of grid resolution. Most of the poor matches to data are attributable to poor grid resolution. This includes the upper regions of the Mississippi River, wave transformation zones on the south shore of Lake Pontchartrain, Vermilion Bay, and narrow channels that penetrate roads. In addition to resolution, physical processes are critical. Riverine flows, tides, and wave-driven setup are vital contributors to overall surge. However there are additional processes that should be added to further refine model skill. Upland locations in the vicinity of steep topography may be severely underpredicted due to the lack of rainfall–runoff processes. Interior portions of levee systems also require consideration of rainfall–runoff, wave overtopping flow rates, and breaching. Vermilion Bay and other similar fine sediment deltaic regions will require a detailed examination of how muddy sea beds affect waves and surge propagation and attenuation. In addition, better descriptors of air–sea momentum transfer tied to wave conditions will be beneficial. Finally vertical current structure can enhance or reduce water surface elevation.

The rapid advances in the observational systems such as lidar, satellite-based ocean vector winds and land-cover analysis, land-based and airborne Doppler radar, airborne microwave radiometers, computational algorithms, and computing platforms will continue to allow improvements in our ability to model coastal storm environments. We envision future models focusing on higher resolution, more physics within dynamically coupled systems, and improved parameterizations based on objective analyses of microscale data. Furthermore, these high-resolution hurricane models will be applied as forecasting tools using high-performance parallel computing environments.

Acknowledgments

Permission to publish this paper was granted by the Chief of Engineers, U.S. Army Corps of Engineers (USACE). This work was supported by the USACE Interagency Performance Evaluation Task Force; the Joint Coastal Surge Study in support of the USACE Louisiana Coastal Protection and Restoration Study, the USACE New Orleans District, and the USACE Hurricane Protection Office; the Federal Emergency Management Agency Region 6; and the USACE System-Wide Water Resources and MORPHOS Programs; the Department of Homeland Security Center of Excellence in Natural Hazards, Coastal Infrastructure, and Emergency Management; and the National Oceanic and Atmospheric Administration Integrated Ocean Observing System Program. Computational resources and support were provided by the U.S. Army Engineer Research and Development Center, Department of Defense Supercomputing Resource Center, and the University of Texas at Austin, Texas Advanced Computing Center. ADCIRC model development was supported by awards from the USACE, the National Science Foundation (DMS06-20696 and OCI-0746232), and the Office of Naval Research (N00014-06-1-0285).

REFERENCES

  • Arcement, G. J., and V. R. Schneider, 1989: Guide for selecting Manning’s roughness coefficients for natural channels and flood plains. U.S. Geological Survey Water Supply Paper 2339, U.S. Geological Survey, Denver, CO, 38 pp.

    • Search Google Scholar
    • Export Citation
  • Atkinson, J. H., J. J. Westerink, and J. M. Hervouet, 2004: Similarities between the wave equation and the quasi-bubble solutions to the shallow water equations. Int. J. Numer. Methods Fluids, 45 , 689714.

    • Search Google Scholar
    • Export Citation
  • Barnes, H. H., 1967: Roughness characteristics of natural channels. U.S. Geological Survey Water Supply Paper 1849, U.S. Geological Survey, Washington, DC, 213 pp.

    • Search Google Scholar
    • Export Citation
  • Blain, C. A., J. J. Westerink, and R. A. Luettich, 1994: The influence of domain size on the response characteristics of a hurricane storm surge model. J. Geophys. Res., 99 , (C9). 1846718479.

    • Search Google Scholar
    • Export Citation
  • Blain, C. A., J. J. Westerink, and R. A. Luettich, 1998: Grid convergence studies for the prediction of hurricane storm surge. Int. J. Numer. Methods Fluids, 26 , 369401.

    • Search Google Scholar
    • Export Citation
  • British Oceanographic Data Centre, 2003: General Bathymetric Chart of the Oceans, Centenary Edition. [Available online at http://www.bodc.ac.uk/products/bodc_products/gebco/].

    • Search Google Scholar
    • Export Citation
  • Cardone, V. J., and A. T. Cox, 2007: Tropical cyclone wind field forcing for surge models: Critical issues and sensitivities. Nat. Hazards, 51 , 2947. doi:10.1007/s11069-009-9369-0.

    • Search Google Scholar
    • Export Citation
  • Cardone, V. J., A. T. Cox, and G. Z. Forristall, 2007: Hindcasts of winds, waves and currents in the northern Gulf of Mexico in Hurricanes Katrina (2005) and Rita (2005). OTC 18652, Proc. 2007 Offshore Technology Conf., Houston, TX.

    • Search Google Scholar
    • Export Citation
  • Chow, V. T., 1959: Open-Channel Hydraulics. McGraw-Hill Book Company, 680 pp.

  • Cox, A. T., J. A. Greenwood, V. J. Cardone, and V. R. Swail, 1995: An interactive objective kinematic analysis system. Proc. Fourth Int. Workshop on Wave Hindcasting and Forecasting, Banff, Alberta, Canada, Atmospheric Environment Service, 109–118.

    • Search Google Scholar
    • Export Citation
  • Dawson, C., J. J. Westerink, J. C. Feyen, and D. Pothina, 2006: Continuous, discontinuous and coupled discontinuous-continuous Galerkin finite element methods for the shallow water equations. Int. J. Numer. Methods Fluids, 52 , 6388.

    • Search Google Scholar
    • Export Citation
  • Dietrich, J. C., and Coauthors, 2010: A high-resolution coupled riverine flow, tide, wind, wind wave, and storm surge model for southern Louisiana and Mississippi. Part II: Synoptic description and analysis of Hurricanes Katrina and Rita. Mon. Wea. Rev., 138 , 378404.

    • Search Google Scholar
    • Export Citation
  • Ebersole, B. A., J. J. Westerink, D. T. Resio, and R. G. Dean, 2007: Performance evaluation of the New Orleans and Southeast Louisiana Hurricane Protection System, Volume IV—The storm. Final Report of the Interagency Performance Evaluation Task Force, U.S. Army Corps of Engineers, Washington, DC, 263 pp.

    • Search Google Scholar
    • Export Citation
  • Garratt, J. R., 1977: Review of drag coefficients over oceans and continents. Mon. Wea. Rev., 105 , 915929.

  • Garster, J. K., B. Bergen, and D. Zilkoski, 2007: Performance evaluation of the New Orleans and Southeast Louisiana Hurricane Protection System, Volume II—Geodetic vertical and water level datums. Final Report of the Interagency Performance Evaluation Task Force, U.S. Army Corps of Engineers, Washington, DC, 157 pp.

    • Search Google Scholar
    • Export Citation
  • Gunther, H., 2005: WAM cycle 4.5 version 2.0. Institute for Coastal Research, GKSS Research Centre, Geesthacht, Germany, 38 pp.

  • Hagen, S. C., J. J. Westerink, R. L. Kolar, and O. Horstmann, 2001: Two dimensional unstructured mesh generation for tidal models. Int. J. Numer. Methods Fluids, 35 , 669686.

    • Search Google Scholar
    • Export Citation
  • Hartley, S., R. Pace III, J. B. Johnston, M. Swann, C. O’Neil, L. Handley, and L. Smith, 2000: A GAP analysis of Louisiana: Final report and data. U.S. Department of the Interior, U.S. Geological Survey, Lafayette, LA, 588 pp.

    • Search Google Scholar
    • Export Citation
  • Hendershott, M. C., 1981: Long waves and ocean tides. Evolution of Physical Oceanography, B. A. Warren and C. Wunsch, Eds., MIT Press, 292–341.

    • Search Google Scholar
    • Export Citation
  • Holland, G., 1980: An analytic model of the wind and pressure profiles in hurricanes. Mon. Wea. Rev., 108 , 12121218.

  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437471.

  • Komen, G., L. Cavaleri, M. Donelan, K. Hasselmann, S. Hasselmann, and P. A. E. M. Janssen, 1994: Dynamics and Modeling of Ocean Waves. Cambridge University Press, 560 pp.

    • Search Google Scholar
    • Export Citation
  • Kubatko, E. J., S. Bunya, C. Dawson, J. J. Westerink, and C. Mirabito, 2009: A performance comparison of continuous and discontinuous finite element shallow water models. J. Sci. Comput., 40 , 315339.

    • Search Google Scholar
    • Export Citation
  • Le Provost, C., F. Lyard, J. Molines, M. Genco, and F. Rabilloud, 1998: A hydrodynamic ocean tide model improved by assimilating a satellite altimeter-derived data set. J. Geophys. Res., 103 , 55135529.

    • Search Google Scholar
    • Export Citation
  • Louisiana State University, cited. 2004: Louisiana Lidar. [Available online at http://atlas.lsu.edu/lidar/].

  • Luettich, R. A., and J. J. Westerink, 2004: Formulation and numerical implementation of the 2D/3D ADCIRC finite element model version 44.XX. 74 pp. [Available online at http://adcirc.org/adcirc_theory_2004_12_08.pdf].

    • Search Google Scholar
    • Export Citation
  • Mississippi Automated Resource Information System, cited. 2006: Mississippi Island 10 meter by 10 meter DEM. [Available online at http://www.maris.state.ms.us/HTM/DownloadData/DEM.html].

    • Search Google Scholar
    • Export Citation
  • McGee, B. D., B. B. Goree, R. W. Tollett, B. K. Woodward, and W. H. Kress, 2006: Hurricane Rita surge data, southwestern Louisiana and southeastern Texas, September to November 2005. U.S. Geological Survey Data Series 220. [Available online at http://pubs.water.usgs.gov/ds220].

    • Search Google Scholar
    • Export Citation
  • Mukai, A., J. J. Westerink, R. Luettich Jr., and D. Mark, 2002: Eastcoast 2001: A tidal constituent database for the Western North Atlantic, Gulf of Mexico, and Caribbean Sea. Tech. Rep. ERDC/CHL TR-02-24, U.S. Army Corps of Engineers, 201 pp. [Available from ERDC Vicksburg (WES), U.S. Army Engineer Waterways Experiment Station (WES), ATTN: ERDC-ITL-K, 3909 Halls Ferry Rd., Vicksburg, MS 39180-6199].

    • Search Google Scholar
    • Export Citation
  • National Geophysical Data Center, 1988: Data announcement 88-MGG-02, digital relief of the surface of the Earth. National Oceanic and Atmospheric Administration, Boulder, CO. [Available online at http://www.ngdc.noaa.gov/mgg/global/etopo5.html].

    • Search Google Scholar
    • Export Citation
  • National Ocean Service, 1997: Hydrographic survey digital database. Vol. 1, 3rd ed. National Oceanic and Atmospheric Administration.

  • Powell, M., S. Houston, and T. Reinhold, 1996: Hurricane Andrew’s landfall in South Florida. Part I: Standardizing measurements for documentation of surface wind fields. Wea. Forecasting, 11 , 304328.

    • Search Google Scholar
    • Export Citation
  • Powell, M., S. Houston, L. Amat, and N. Morrisseau-Leroy, 1998: The HRD real-time hurricane wind analysis system. J. Wind Eng. Ind. Aerodyn., 77–78 , 5364.

    • Search Google Scholar
    • Export Citation
  • Powell, M. D., P. J. Vickery, and T. A. Reinhold, 2003: Reduced drag coefficient for high wind speeds in tropical cyclones. Nature, 422 , 279283.

    • Search Google Scholar
    • Export Citation
  • Powell, M. D., and Coauthors, 2008: Reconstruction of Hurricane Katrina’s wind fields for storm surge and wave hindcasting. Ocean Eng., in press, doi:10.1016/j.oceaneng.2009.08.014.

    • Search Google Scholar
    • Export Citation
  • Resio, D. T., 2007: White paper on estimating hurricane inundation probabilities. U.S. Army Engineering Research and Development Center, Vicksburg, MS, 125 pp.

    • Search Google Scholar
    • Export Citation
  • Sheremet, A., A. J. Mehta, B. Liu, and G. W. Stone, 2005: Wave-sediment interaction on a muddy inner shelf during Hurricane Claudette. Estuarine Coastal Shelf Sci., 63 , 225233.

    • Search Google Scholar
    • Export Citation
  • Smith, J. M., 2000: Benchmark tests of STWAVE. Proc. Sixth Int. Workshop on Wave Hindcasting and Forecasting, Monterey, CA, Environment Canada, 369–379.

    • Search Google Scholar
    • Export Citation
  • Smith, J. M., A. R. Sherlock, and D. T. Resio, 2001: STWAVE: Steady-state spectral wave model user’s manual for STWAVE, version 3.0. USACE, Engineer Research and Development Center, Tech. Rep. ERDC/CHL SR-01-1, Vicksburg, MS, 81 pp. [Available online at http://chl.erdc.usace.army.mil/Media/2/4/4/erdc-chl-sr-01-11.pdf].

    • Search Google Scholar
    • Export Citation
  • Smith, S. J., and J. M. Smith, 2001: Numerical modeling of waves at Ponce de Leon Inlet, Florida. J. Waterw. Port Coastal Ocean Div., 127 (3) 176184.

    • Search Google Scholar
    • Export Citation
  • Stone, G. W., A. Sheremet, X. Zhang, Q. He, B. Liu, and B. Strong, 2003: Landfall of two tropical systems seven days apart along south-central Louisiana, USA. Proc. Coastal Sediments ’03, Clearwater Beach, FL, University of South Florida/USGS/U.S. Army Corps of Engineers, 333–334.

    • Search Google Scholar
    • Export Citation
  • Thompson, E. F., and V. J. Cardone, 1996: Practical modeling of hurricane surface wind fields. J. Waterw. Port Coastal Ocean Div., 122 , 195205.

    • Search Google Scholar
    • Export Citation
  • Thompson, E. F., J. M. Smith, and H. C. Miller, 2004: Wave transformation modeling at Cape Fear River Entrance, North Carolina. J. Coastal Res., 20 (4) 11351154.

    • Search Google Scholar
    • Export Citation
  • URS, 2006a: Final coastal and riverine high-water marks collection for Hurricane Katrina in Louisiana. FEMA-1603-DR-LA, Task Orders 412 and 419, Federal Emergency Management Agency, Washington, DC, 76 pp.

    • Search Google Scholar
    • Export Citation
  • URS, 2006b: Final coastal and riverine high-water marks collection for Hurricane Rita in Louisiana. FEMA-1603-DR-LA, Task Orders 445 and 450, Federal Emergency Management Agency, Washington, DC, 79 pp.

    • Search Google Scholar
    • Export Citation
  • URS, 2008: Mississippi coastal analysis project compiled reports of HMTAP. Task Order 18, Federal Emergency Management Agency, Washington, DC, Vol. 1, 376 pp.

    • Search Google Scholar
    • Export Citation
  • U.S. Department of Defense, 1999: Digital nautical chart. National Imagery Mapping Agency, Washington, DC.

  • Villea, F. J., 2005: Mississippi GAP analysis project. GAP Analysis Bull. 13, U.S. Department of the Interior, U.S. Geological Survey. [Available online at http://www.gap.uidaho.edu/bulletins/13/Mississippi.htm].

    • Search Google Scholar
    • Export Citation
  • Vogelmann, J. E., S. M. Howard, L. Yang, C. R. Larson, B. K. Wylie, and N. Van Driel, 2001: Completion of the 1990s National Land Cover Data Set for the conterminous United States from Landsat thematic mapper data and ancillary data sources. Photogramm. Eng. Remote Sens., 67 , 650652.

    • Search Google Scholar
    • Export Citation
  • Westerink, J. J., R. A. Luettich, and J. C. Muccino, 1994: Modeling tides in the Western North Atlantic using unstructured graded grids. Tellus, 46A , 178199.

    • Search Google Scholar
    • Export Citation
  • Westerink, J. J., and Coauthors, 2008: A basin-to-channel-scale unstructured grid hurricane storm surge model applied to southern Louisiana. Mon. Wea. Rev., 136 , 833864.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

WAM model domain shown in red and nested STWAVE model domains shown in blue. In order from west to east, the five STWAVE domains are W, S, LP, SE, and MS-AL, as described in Table 1.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 2.
Fig. 2.

ADCIRC SL15 model domain with bathymetry (m). Geographic locations of interest are indicated by the numbers identified in Table 2.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 3.
Fig. 3.

Detail of the SL15 domain across southern Louisiana and Mississippi with bathymetry and topography [m relative to NAVD88 (2004.65)] with raised features such as levees, railroads, and highways shown in brown. Geographic locations of interest are indicated by numbers identified in Table 2.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 4.
Fig. 4.

As in Fig. 3, but across southwestern Louisiana.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 5.
Fig. 5.

As in Fig. 3, but across southeastern Louisiana and Mississippi.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 6.
Fig. 6.

Detail of the SL15 grid across southern Louisiana and Mississippi with finite-element sizes shown in meters.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 7.
Fig. 7.

Detail of the applied Manning n roughness coefficients for southern Louisiana.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 8.
Fig. 8.

Sample of the applied directional wind reduction factor for uniform steady southerly winds for southern Louisiana. The coastline is outlined in white.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 9.
Fig. 9.

Locations of the six USACE stations with stage–flow relationships that were compared to the computed water levels in Fig. 10. In numerical order, the six stations are Baton Rouge, Donaldsonville, New Orleans, Alliance, Empire, and Venice.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 10.
Fig. 10.

Stage–flow relationships at six USACE stations along the Mississippi River. Measured data is shown as scatter points with associated best-fit curves. The predicted data is shown as connected blue dots.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 11.
Fig. 11.

Comparison of amplitudes and phases as measured by NOAA and predicted by the SL15 model: (left) amplitudes and (right) phases. Each row of figures represents a region as indicated in Table 11 with difference estimates given in Table 13.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 12.
Fig. 12.

Locations of the deep-water NDBC buoys used in the analysis of Hurricanes Katrina and Rita with offshore buoy identifier numbers.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 13.
Fig. 13.

Wind speeds and directions during Hurricane Katrina at four offshore NDBC buoys with buoy identifiers. The measured data is shown with red dots, while the predicted results are shown with black lines.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 14.
Fig. 14.

Wave heights and periods during Hurricane Katrina at six NDBC buoys with identifiers. The measured data is shown with red dots, while the predicted results are shown with black lines. The first four rows show comparisons to WAM results at selected offshore buoys, while the last two rows show comparison to STWAVE results at available coastal stations.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 15.
Fig. 15.

Locations of the nine USACE, NOS, and NWS stations with hydrograph data for Hurricane Katrina. The nine stations are 1) Pass Manchac, 2) Bayou LaBranche, 3) Lake Pontchartrain Midlake Causeway, 4) 17th Street Canal, 5) Little Irish Bayou, 6) the IHNC Lock Staff Gauge, 7) Southwest Pass, 8) Mississippi River at Carrollton, and 9) Grand Isle. Colors indicate the differences between the modeled and measured peak surge. Green points indicate a match within 0.5 m. Red, orange, yellow, and light green circles indicate overpredictions by the model; green, blue, dark blue, and purple circles indicate underpredictions. The clear circle at station 5 indicates an incomplete hydrograph that does not allow for a peak point comparison.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 16.
Fig. 16.

Hydrographs for the nine USACE, NOS, and NWS stations during Hurricane Katrina. The black lines are the computed water levels from the ADCIRC SL15 model, while the red lines are the measured data.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 17.
Fig. 17.

Locations of USACE HWMs for Hurricane Katrina. Colors indicate the difference between the maximum computed water elevation from the ADCIRC SL15 hindcast and the measured high water mark. Green points indicate a match within 0.5 m. Red, orange, yellow, and light green circles indicate overpredictions by the model; green, blue, dark blue, and purple circles indicate underpredictions.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 18.
Fig. 18.

As in Fig. 17, but for locations of URS HWMs for Hurricane Katrina.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 19.
Fig. 19.

Comparisons between observed USACE high water marks and ADCIRC maximum surges during Hurricane Katrina at 206 locations shown in Fig. 17. Green points indicate a match within 0.5 m. Red, orange, yellow, and light green circles indicate overpredictions by the model; green, blue, dark blue, and purple circles indicate underpredictions. The slope of the best-fit line through all points is 0.99 and R2 value is 0.92.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 20.
Fig. 20.

As in Fig. 19, but for comparisons between observed URS high water marks and ADCIRC maximum surges during Hurricane Katrina at 193 locations shown in Fig. 18. The slope of the best-fit line through all points is 1.02 and R2 value is 0.94.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 21.
Fig. 21.

Wind speeds and directions during Hurricane Rita at four offshore NDBC buoys with buoy identifiers. The measured data is shown with red dots, while the predicted results are shown with black lines.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 22.
Fig. 22.

Wave heights and periods during Hurricane Rita at five NDBC buoys with identifiers. The measured data is shown with red dots, while the predicted results are shown with black lines. The first four rows show WAM results at selected offshore buoys, while the last row shows STWAVE results at the available coastal station.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 23.
Fig. 23.

Locations of the 23 USGS stations for Hurricane Rita. Colors indicate the difference between the maximum water elevation from the ADCIRC SL15 hindcast and the maximum water level from the USGS hydrograph data. Green points indicate a match within 0.5 m. Red, orange, yellow, and light green circles indicate overpredictions by the model; green, blue, dark blue, and purple circles indicate underpredictions. White points indicate stations where ADCIRC did not simulate storm surge.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 24.
Fig. 24.

Hydrographs at the first 12 USGS stations for Hurricane Rita. The black lines are the computed water levels from the ADCIRC SL15 model, while the red dots are the measured data.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 25.
Fig. 25.

As in Fig. 24, but for the last 11 USGS stations for Hurricane Rita.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 26.
Fig. 26.

Locations of the 80 HWMs obtained from URS for Hurricane Rita. Colors at each location indicate the difference between the maximum elevation from the ADCIRC SL15 hindcast and the URS HWM. Green points indicate a match within 0.5 m. Red, orange, yellow, and light green circles indicate overpredictions by the model; green, blue, dark blue, and purple circles indicate underpredictions.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 27.
Fig. 27.

Scatterplot of HWMs for Hurricane Rita. Green points indicate a match within 0.5 m. Red, orange, yellow, and light green circles indicate overpredictions by the model; green, blue, dark blue, and purple circles indicate underpredictions. The slope of the best-fit line through all points is 0.97 and the R2 value is 0.77.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Fig. 28.
Fig. 28.

Scatterplot of HWMs for Hurricane Rita without data in Vermilion Bay. Green points indicate a match within 0.5 m. Red, orange, yellow, and light green circles indicate overpredictions by the model; green, blue, dark blue, and purple circles indicate underpredictions. The slope of the best-fit line through all points is 1.04 and the R2 value is 0.87.

Citation: Monthly Weather Review 138, 2; 10.1175/2009MWR2906.1

Table 1.

STWAVE grid names, the origin of the southeast corner in degrees latitude and longitude (northwest corner for the Lake Pontchartrain grid), orientation in degrees measured counterclockwise from the parallel that runs through the origin, and an indicator if the full or half plane version of STWAVE was run.

Table 1.
Table 2.

Geographic location by type and number shown in Figs. 2 –5.

Table 2.
Table 3.

Manning-n values for LA-GAP classification.

Table 3.
Table 4.

Manning-n values for MS-GAP classification.

Table 4.
Table 5.

Manning-n values for 1992 NLCD classification.

Table 5.
Table 6.

ADCIRC model run segments for the river, tide, Hurricanes Katrina and Rita validation runs in days from the start of the simulation or in UTC time (hour/day/month). Duration of the application of the river ramp, time for the rivers to reach equilibrium, the duration of the tidal forcing ramp, the time for the tides to reach dynamic equilibrium, the duration of the wind and wave forcing are given.

Table 6.
Table 7.

River flow rates (m3 s−1) for the various simulations.

Table 7.
Table 8.

Principal tidal constituents with periods (hours), tidal potential constants (m), and associated effective earth elasticity factors.

Table 8.
Table 9.

The 1992 NLCD nominal land roughness values, z0-land. The *Class 95 is constructed from the GAP data for Louisiana. The NLCD did not have coverage for cypress wetland forest; therefore GAP datasets were merged into the NLCD and the cypress forest land type was imposed on top of the NLCD data.

Table 9.
Table 10.

Summary of average absolute differences (m) for the stage–flow relationships shown in Fig. 10.

Table 10.
Table 11.

NOAA stations used in the NOAA-measured to SL15-computed difference analysis for tidal constituents. The station IDs marked with asterisks (*) indicate stations whose longitude and latitude were shifted slightly in the ADCIRC SL15 model.

Table 11.
Table 12.

SL15 model harmonic constituents used to decompose model time histories into harmonic constituents.

Table 12.
Table 13.

Correlation coefficients R2 of SL15 computed harmonic constituents compared to NOAA-measured/analyzed values for the four groups of NOAA stations.

Table 13.
Table 14.

SL15 model and NOAA-measured/analyzed difference statistics for the four groups of NOAA stations. These difference values include measurement uncertainties. Average, average absolute, and standard deviation amplitude differences are in m and degrees, normalized root-mean-square difference is dimensionless.

Table 14.
Table 15.

Summary of SL15-computed and NOAA measurement–analysis differences for each harmonic constituent and NOAA-measured/analyzed data uncertainty estimates.

Table 15.
Table 16.

Summary of difference/error statistics for the Katrina HWM datasets. Average absolute differences/errors and standard deviations are given in meters.

Table 16.
Table 17.

Summary of difference/error statistics for the Rita HWM datasets. Average absolute differences/errors and standard deviations are given in meters.

Table 17.
Save