A digital twin system to classify tornado climatology data for predicting future disasters

Recognizing issues with existing tornado warning systems, we investigate alternative approaches. In particular, we present data science and advanced analytics solution—specifically, a digital twin system—to integrate data from various rich data sources, including tornado climatology data and realtime weather conditions preceding tornado warnings. By identifying the key data points that signal a warning, our solution enhances the prediction of tornado events. Evaluation on realworld US datasets highlight the advantages of incorporating a recurrent neural network (RNN) based classification forecasting model into our solution. Evaluation results further demonstrate the effectiveness and practicality of our solution in accurately forecasting tornado lead times, intensities and locations, aiding the building and developing smart and sustainable cities.