Authors | ; ; |
Data Type(s) | REU |
Natural Hazard Type(s) | Storm surge |
Date of Publication | 2021-08-16 |
Facilities | |
Awards | NSF, Natural Hazards Research Engineering Infrastructure, Network Coordination Office | 1612144 |
Keywords | UC Berkeley SimCenter, Machine Learning, GeoClaw, Neural Networks, Storm Surge |
DOI | 10.17603/ds2-jx64-ce09 |
License | Creative Commons Attribution |
This project studies the efficiency of a neural network-based model in comparison with a physics-based model. Hurricane Ike storm surge data was used within the GeoClaw package to produce a series of outputs, then a neural network was trained to replicate that data. The data in this project can be reused to further study the accuracy, loss, computational time, and storage compatibilities of neural networks in comparison to physics-based models. This project is unique as it is working to combine previously used forecasting methods to create a more efficient, hybrid method. The main audience is those researching storm surges and different forecasting models. A broader audience would include those interested in the uses of machine learning systems.