Recent extreme rainfall events have revealed the transportation network’s vulnerabilities to road washouts. Currently, NC DOT reacts to these problems as are reported from the field. This inability to predict where washouts are likely to occur leads to long response times and inefficient positioning of resources. The availability of high quality statewide elevation data, historical rainfall records and advances in computer processing presents the opportunity to modify and develop programs to predict where washouts are likely to occur during extreme rainfall events. The purpose of this project is to develop models and test several approaches for predicting crossing washouts based on forecasted rainfall. A team of NCSU BAE engineers will first characterize and analyze historical washouts during extreme events. Then, detailed HEC-HMS models will be developed and calibrated and validated for one watershed in each physiographic region.
A user interface will be created to run the models using forecasted rainfall, relate the predicted discharge to potential washouts using water surface elevation-discharge relationships, and then output the results for display in a GIS map. The model output for a large number of historical events will then be used to test different machine learning algorithms for their ability to predict discharge and potential washout locations. The information on historical washouts and the model predictions will be used to develop a network of “safe” routes for each watershed. The results will help determine if existing hydrologic models can be leveraged to accurately predict potential washout locations and to evaluate if machine learning technology can be employed for accurate flood prediction. This project has the potential to substantially enhance NC DOT’s ability to respond to storm events and position resources appropriately. Results will be disseminated in NC DOT meetings, a training workshop for NC DOT personnel, and through extension factsheets and academic publications.