For Army Corps of Engineers – 2005
Ensemble machine-learning-based was used to create flood probability indices for the states of Louisiana, Mississippi and Alabama (Gulf region). The study used the frequency ratio (FR) from the observational data and it was used in combination with support vector machine (SVM) using a radial basis function kernel in combination with weather data and hydro-geomorphometric parameters that are extracted using geomorphometric functions. Our predictions were of 97% accurate for flood probability in the region.
For Army Corps of Engineers (2007)
Statistical methods were used in combination with Support Vector Machine (SVM) to assess flood hazard over large areas specifically to extend the flood hazard zonation to the portion of the river networks. This method proved better over the locations where hydraulic models failed. A parametric model was developed to identify flood hazard gulf coast. The proposed method is based on a multivariate statistical analysis used in combination with artificial intelligence (AI) / machine learning algorithms using in input flood hazard maps delineated by the local authorities and terrain elevation data. The preliminary results obtained for several major catchments of the Louisiana and Mississippi states indicate good performance indicator. Results derived were very promising, giving the possibility to obtain reliable delineations of flood prone areas obtained in the rest of the southern US coastal areas.