For this presentation, we retroactively forecast Hurricanes Sandy and Irene using an ensemble of 21 atmospheric forecasts from the Coupled Ocean-Atmosphere Mesoscale Prediction System Tropical Cyclone model (COAMPS-TC), for a range of lead-times out to 4 days prior to landfall. Model results for each successive forecast are analyzed using mathematical methods developed for analyzing high frequency financial data and using concepts based on sensor signal processing. The forecast time series and uncertainty are created by 1) using weights based on a correlation between observations and model results, 2) the extraction of features based on peak height and time of that peak, and finally 3) identifying the potentially best forecast from the 21 members based on an assessment of the model results versus observations using single value, Bhattacharyya distance, and Kullback-Leibler divergence methods. A Bayesian decision model is also used. Water level observations during these events at over 10 tide gauges are used to assess the strengths and weaknesses of each method. We close by discussing progress with our operational system, where we currently have ~70 meteorological model data sets for forcing, including those from the US National Weather Service, the European Center for Medium Range Forecasts, and several university-based meteorological forecasts.