Ann Based Tsunami Detection Algorithm

Abstract: Tsunamis cause damage by two mechanisms: the smashing force of a wall of water travelling at high speed, and the destructive power of a large volume of water draining off the land and carrying all with it, even if the wave did not look large. Earthquakes with an epicenter in the sea are not always tsunami genic. Direct detection in sea-level measurements is therefore essential to confirm the actual generation and propagation of a tsunami. Signals obtained from the bottom pressure recorders (BPRs) are commonly used for the automatic, real time detection of a tsunami within recorded signals. Only direct detection makes it possible to avoid false alarms and guarantees speed and accuracy in both the warning and the hazard assessment process. The tsunami detection algorithm works by first estimating the amplitudes of the pressure fluctuations within the tsunami frequency band and then testing these amplitudes against a threshold value. This project aims at developing a Tsunami warning algorithm using artificial neural network (ANN). . Proposed algorithm is compared to the one developed under the Deep-ocean Assessment and Reporting of Tsunamis (DART) program run by the U.S. National Oceanic and Atmospheric Administration (NOAA). The algorithm is designed to detect the occurrence of tsunami using the data obtained from the data buoy. The ANN used for this process is feed forward multilayer network with back propagation training. Simulation and experimental results show that an improvement in detection performance can be obtained by using the ANN algorithm.

Keywords: Tsunami forcasting;Neural networks;Back propogation; Feed forward method