An Efficient Tool to Assess Risk of Storm Surges Using Data Mining

The catastrophic damages caused by hurricane Katrina and Rita in 2005 focused new attention on accurately predicting storm surges and efficiently evaluating coastal risks. Numerous storm surge models (i.e., SLOSH, CH3D-SSMS, ADCIRC) have been developed and used in real-time for estimating local storm surges from an approaching hurricane. One of the state-of-the-art techniques is to run high resolution storm surge models such as ADCIRC utilizing hundreds of computers (or CPUs) in parallel (Fleming et al. 2007). Even with the parallel computing resources, it can take hours to predict high resolution local storm surge in real time once a hurricane advisory is issued from the National Hurricane Center (NHC). As an alternative, we have developed a robust and efficient method to predict local storm surge using data mining. This data mining method is programmed into a simplified Graphical User Interface (GUI) which operates in real time and it is desktop based, fast and very easy to use. The algorithm uses a weight based Storm Similarity Index (SSI) which is calculated by using current hurricane position, Central Pressure (C_P), Pressure Scale Radius (R_max) along with hurricane track, forecasted landfall location, storm forward speed, and forecasted storm track published by the National Hurricane Centre (NHC) and then correlated with the characteristics of hundreds of synthetic storm simulations archived in a central database. Based on the values of SSI (scales from 0 to 1), the GUI then identifies a group of synthetic storms that closely matches with the characteristics of the approaching hurricane and then display high resolution results (e.g., maximum surge elevation and hydrographs) in Google Earth environment. The method has been verified against two historical hurricanes, Katrina and Camille which made landfall in the Mississippi coast. With the help of this tool, the emergency personnel can quickly estimate high resolution local storm surge and can make quantitative decisions by evaluating "what-if-scenarios" starting two to three days ahead of the landfall.