Prediction of storm surge and coastal inundation using Artificial Neural Network – A case study for 1999 Odisha Super Cyclone

Tropical cyclone induced storm surge and associated onshore flooding poses significant danger and havoc to life, property and infrastructure during the time of landfall. Coastal belt along the East coast of India is thickly populated and also highly vulnerable to impact of tropical cyclones. Real-time forecasting system that provides reliable estimates on possible storm surge height, envelope and extent of onshore flooding has potential socio-economic benefits. Conventional methods use state-of-art numerical models or ensemble of models that are computationally expensive and highly time consuming during real-time operations. This study proposes an alternate approach using soft computing techniques such as Artificial Neural Network (ANN) for the prediction of storm surge and onshore flooding. The proposed network architecture is proven to be viable and highly cost-effective consistently maintaining high level of computational accuracy (>92%) thereby finding potential real-time application. As a case study, the efficacy of ANN model in simulating storm-tide and extent of onshore flooding associated with the 1999 Odisha Super cyclone have been examined. Pre-computed scenarios of storm-tide and inundation data were used to train ANN model for the entire Odisha coast with a success rate of 99%. After the training phase, computational time in prediction of storm surge and inundation is quite rapid (in order of seconds) as compared to any conventional model. Validation exercise performed to skill assess the robustness of ANN model using archived records of storm-tide and inundation obtained an accuracy of 92% and 94% respectively. Results obtained are quite encouraging demonstrating the efficacy of ANN model for real-time application and effectiveness for disaster risk reduction during tropical cyclone activity. Keywords: Artificial Neural Network, Odisha Super Cyclone, Storm tide, Coastal inundation

Abstract licence: CC BY-NC

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.