STORM SURGE FORECASITNG USING NEURAL NETWORK
-
- KIM Sooyoul
- Tottori University
-
- MATSUMI Yoshiharu
- Tottori University
-
- IZUTA Yujiro
- Tottori University
-
- MASE Hajime
- Kyoto University
-
- MORI Nobuhito
- Kyoto University
-
- YASUDA Tomohiro
- Kyoto University
Bibliographic Information
- Other Title
-
- ニューラルネットワークによる高潮予測モデル
- ニューラルネットワーク ニ ヨル タカシオ ヨソク モデル
Search this article
Abstract
In the present study, we examine and develop storm surge forecasting models at Sakai Minato and Hamada using artificial neural network (ANN). The storm surge forecasting model aims to forecast a time series of surge levels with the 3 hrs to 30 hrs leadtime ahead every 3 hrs interval at one station. For training data, we gathered the five historical typhoons and their data, which are storm surge levels, sea level pressures, depression rate of sea level pressures and typhoon positions. In addition, those were collected from storm surge simulations using projected typhoons under the past and future climate change experiment. In order to investigate the accuracy of the forecasting models, we changed the number of hidden units in the ANN from 13 to 130 with every 13 interval. We showed that the 30 hrs forecasting model can predict a time series of storm surge levels with the correlation coefficient of 0.97 and the root mean square error of 0.029 m.
Journal
-
- Journal of Japan Society of Civil Engineers, Ser. B2 (Coastal Engineering)
-
Journal of Japan Society of Civil Engineers, Ser. B2 (Coastal Engineering) 71 (2), I_223-I_228, 2015
Japan Society of Civil Engineers