PREDICTION OF STORM SURGE LEVEL USING ARTIFICIAL NEURAL NETWORK: A CASE STUDY FOR TYPHOON HAIYAN

Abstract

Storm surge is considered as one of the greatest threats to life and property during a tropical cyclone, especially to a community living near the coastal area. The Philippines is particularly susceptible to the effects of coastal calamities like storm surges and tsunamis since it is an archipelago nation. One way to reduce the risk is to improve the ability of the community to monitor and forecast the hazard through technological research. As such, it is imperative to develop a numerical model that can predict and perform necessary calculations before a storm surge strikes in a coastal area. This paper utilized the Artificial Neural Network (ANN) to predict the storm surge level with 2013 Typhoon Haiyan (Yolanda) as a case study. The proposed model is tested, trained, and validated using the available 101 test data collected from the Guiuan station of PAGASA and NAMRIA. The collected data is composed of six (6) input variables and one (1) output variable. The input variables are the following: astronomical tide, central atmospheric pressure, rainfall intensity, wind radius, wind speed, and depth, while the output parameter is the storm surge level. The optimum mathematical model, as determined by the back-propagation technique in the artificial neural network (ANN) model, is Bayesian Regularization with twelve (12) hidden neurons, with a regression coefficient (R) of 0.99386 and a mean squared error (MSE) of 0.0051569, respectively. The results obtained are quite promising and demonstrate the potential application of the ANN model for disaster risk reduction during tropical storm activity.