A Novel Deep Learning Based Model for Tropical Intensity Estimation and Post-Disaster Management of Hurricanes

The prediction of severe weather events such as hurricanes is always a challenging task in the history climate research, and many deep learning models have been developed for predicting severity events. When disastrous hurricane strikes coastal region, it causes serious hazards to human life habitats also reflects prodigious amount economic losses. Therefore, necessary build improve accuracy avoid significant losses all aspects. However, impractical predict or monitor every storm formation real time. Though various techniques exist diagnosing tropical cyclone intensity convolutional neural networks (CNN), auto-encoders, recurrent network (RNN), etc., there are some challenges involved estimating intensity. This study emphasizes identify different categories perform post-disaster management. An improved (CNN) model used weakest strongest with values using infrared satellite imagery data wind speed from HURDAT2 database. achieves lower Root mean squared error (RMSE) value 7.6 knots Mean (MSE) 6.68 by adding batch normalization dropout layers CNN model. Further, crucial evaluate damage implementing advance measures planning resources. fine-tuning pre-trained visual geometry group (VGG 19) accomplished extent automatic annotation image Greater Houston. VGG 19 trained video datasets classifying types annotate event automatically. 98% achieved 97% results proved that proposed estimation its enhances ability, which can ultimately help scientists meteorologists comprehend Finally, mitigation steps reducing risks addressed.