Uncertainty Quantification 2012: Modeling Hurrricane Storm SurgeSeptember 15, 2012
In a multipart minisymposium on model calibration and validation, Clint Dawson presented joint work of his group at UT Austin, Joannes Westerink of the University of Notre Dame, and Rick Luettich of the University of North Carolina at Chapel Hill on the development of numerical methods for modeling storm surge, and on their efforts to validate the models against historical data. Highlights of the talk and a brief discussion of ongoing work of Dawson's group with Ibrahim Hoteit of King Abdullah University of Science and Technology follow.
When a hurricane threatens the coastal United States, the first image to appear in the media is typically a swirling vortex of clouds observed from a satellite. The images that follow are often stock footage of winds blowing down trees and tearing through structures, and of waves crashing onto shore. There is usually much discussion of the "category" of the storm, which is a measure of the maximum sustained winds in the hurricane. What the general public does not realize is that the most destructive part of a hurricane is usually the storm surge--that is, the flooding caused by wind and waves pushing water inland. Furthermore, the hurricane category does not directly correlate to the magnitude of the surge. Storm surge can result in enormous loss of life and property, and can forever change coastal landscape and ecology. This has been evident in recent hurricane and tropical cyclone events in the United States and around the world. Therefore, the mathematical and computational modeling of hurricane storm surge is a topic of great interest.
High-fidelity simulation of storm surge requires several components: (1) an accurate description of the physical system, (2) inclusion of all physically relevant processes, (3) numerical resolution of the flow, and (4) accurate solution of the mathematical model. This leads to the development of highly resolved discretizations, which can capture local topography, bathymetry, bottom friction, and hydraulic conveyances that enhance flow and structures that impede flow. Furthermore, the numerical algorithms should be phase-accurate, should minimize numerical dissipation, and should be stable under highly advective flow regimes.
The applications of storm surge models generally fall into two categories: hindcasting and forecasting. Hindcast simulations are used to study the effects of historical hurricanes, to understand the physics of the events, and to guide decision makers in the development of new hurricane protection systems, such as levees and seawalls. Hindcast simulations are an integral part of determining which coastal communities qualify for federal flood insurance. They can also be used to calibrate and/or validate a storm surge model with data collected during the event. Hindcast simulations therefore rely on obtaining high-fidelity data for the hurricane wind field, bathymetry, coastal topography, and land-use characteristics for the flooded region. That is, hindcast simulations try to minimize uncertainty through extensive and accurate collection of data related to all aspects of the storm. These simulations are intensive both in the human capital required to collect and verify the inputs to the model, and in the computational resources required.
Forecast simulations are used to predict storm surge in real time as hurricanes approach land. They are used to guide emergency managers in ordering evacuations and deploying resources to the regions most likely to be affected. These simulations are fraught with uncertainties. The hurricane-force winds and storm track are uncertain, and in many cases the hurricane may threaten an area of the coast for which important data--on bathymetry, coastal topography, or bottom friction--either is not available or is unreliable. A forecast simulation must occur in real time; it must be completed and the results processed in time for emergency managers to make quick and ultimately life-saving decisions.
At the recent SIAM UQ meeting, Clint Dawson of the University of Texas at Austin described a collaborative effort of UT Austin, Joannes Westerink of the University of Notre Dame, and Rick Luettich of the University of North Carolina at Chapel Hill to develop efficient and accurate numerical methods for modeling storm surge, and to extensively validate these storm surge models against historical data. This effort has led to the development of the Advanced Circulation (ADCIRC) storm surge modeling system. The ADCIRC model has been used in the design of new levee protection systems in southern Louisiana, and in the development of new federal flood insurance maps for many states along the Gulf of Mexico and eastern coast of the U.S. A storm surge forecasting system, called the ADCIRC Surge Guidance System, or ASGS, is also being developed and implemented for real-time hurricane forecasting.
To overcome some of the uncertainties in the forecast models, Dawson's group at UT Austin is working with Ibrahim Hoteit of King Abdullah University of Science and Technology to develop novel data-assimilation methods that can incorporate data collected during the hurricane event into the ASGS forecast system. This research, which is very much in progress, will also rely on the development and deployment of new sensor and data collection apparati that can provide real-time surge, wind, and wave data.