|Applying a Non-intrusive Measure Theoretic Inverse Analysis to Storm Surge|
Troy Butler, Colorado State University
Monday, September 17, 2012
4:00pm, room 223 Weber Building
Most deaths and monetary damage caused by hurricanes are primarily due to storm surge making the accurate forecasting of storm surge due to hurricanes and tropical cyclones a problem of critical importance. We present a state of the art modeling framework (ADCIRC) used in practice to predict maximum storm surge in the Gulf of Mexico and western north Atlantic. This model has undergone extensive validation with hindcast studies and the results are used by emergency managers to coordinate evacuations. The model itself is deterministic. However, there are several sources of uncertainty including knowledge of parameters defining friction and water depth that are critical to accurate forecasts of storm surge.
In this talk, we give background on the ADCIRC model and a method for modeling the infinite dimensional random fields that is useful for estimating critical parameters. We then discuss a non-intrusive alternative to a measure theoretic inverse sensitivity analysis that we use to identify highly probable input parameter configurations that can produce the noisy observable data of maximum water elevation. As a specific case study, we consider hurricane Katrina and model the bathymetry field around the Louisiana coastline.