Graduation of Lucas Westrik
13 november 2020 14:00 - Locatie: Online - Door: Webredactie
Storm Surge Modelling due to Tropical Cyclone Activity
Development of an artificial neural network capable of predicting maximum storm surge heights for Hong Kong and Macau
Chair of graduation: Dr. ir. J. D. Bricker
Supervisors of graduation: Dr. ir. M. A. Diaz Loaiza (TU Delft), dr. ir. M. Zijlema (TU Delft), prof. ir. R. Ranasinghe (HE Delft), dr. R. Meynadier (AXA Paris)
This research focuses on the development of a neural network for water level predictions in case of an approaching tropical cyclone. First, a Delft3D model is developed capable of simulating storm surges for Hong Kong and Macau. This model is calibrated and validated extensively to ensure accurate model performance. The validated model set-up is then used in combination with a synthetic storm catalog to perform 1000 synthetic storm simulations. The results of these simulations provide the required data for the development of the neural network. The developed neural network can predict maximum water levels based of seven storm track parameters (latitude, longitude, eye pressure, max wind speed, radius of max winds, forward speed and direction). Multiple versions of the neural network are developed. The best performing network is capable of predicting maximum surge heights with a maximum error of 30 cm. Additionally, during this research, multiple Matlab tools have been developed for efficient and automatic model set-up and data processing. These tools can be adapted easy and used in other storm surge studies.