ICCE 2022

How to Cite

TEST OF LSTM NETWORKS IN LONG-TERM BEACH MORPHOLOGICAL CHANGES. (2023). Coastal Engineering Proceedings, 37, management.134. https://doi.org/10.9753/icce.v37.management.134


In the prediction of beach profile changes, for example, the long-term calculation of daily changes can only provide sufficiently reliable reproduction results for a few years. One of the reasons is that actual morphological change is caused by the superposition of complex processes that are unknown. The prediction of timeseries data by data-driven models in contrast to physical models has been applied to various fields in recent years with the spread of deep learning and is expected to be used as a tentative solution for problems that cannot be adequately predicted by physical modeling. Recurrent neural networks are being applied to predict shoreline change over several years (Montaño et al., 2020), but application to spatial morphological changes, such as beach profile changes, is rarely investigated due to the limitations of the large amount of observation data required for learning. Here, we used the LSTM (Long Short-Term Memory) network, one of the recurrent neural networks, and long-term beach profile data observed at the Hasaki coast, Japan for learning and prediction of beach profile changes.


Banno M, Nakamura S, Kosako T, Nakagawa Y, Yanagishima S, Kuriyama Y. (2020): Long-Term Observations of Beach Variability at Hasaki, Japan, Journal of Marine Science and Engineering, 8(11), 871.

Montaño, J., Coco, G., Antolínez, J.A.A. et al. (2020): Blind testing of shoreline evolution models. Sci Rep, 10, 2137.

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 International License.

Copyright (c) 2023 Masayuki Banno, Yoshiaki Kuriyama