A CNN-LSTM Method for the Morphology Evolution Prediction of Beach Mega-Nourishment

A CNN-LSTM Method for the Morphology Evolution Prediction of Beach Mega-Nourishment

Li, Yong (Hohai University)
van Oosterom, P.J.M. (TU Delft GIS Technologie)
Ge, Ying (Hohai University)
Zhang, Xiaoxiang (Hohai University)
Baart, F. (Deltares)

2020

Sand nourishment is widely adopted as an effective soft approach to provide long-term coastal safety, protect the ecology environment, and promote tourism and recreation. With the increase in frequency and expenses in beach nourishment worldwide, an adequate prediction of morphology evolution is greatly desired for coastline management. Based on detailed monitoring data of the mega-nourishment Sand Engine, this article integrates a convolutional neural network (CNN) and long short-term memory (LSTM) to predict the nourishment morphology evolution. The historical surveyed data are transformed into sequence grids, which are input into the CNN to obtain the spatial features of beach nourishment. The CNN is constructed by performing the convolutional and pooling operation on the historical data, which can extract actual spatial features and reduce network complexity. The output of the CNN is input to LSTM to learn the temporal relationship to predict future nourishment terrain using past time-series features. Finally, the LSTM output is decoded by the fully connected layer to obtain the prediction result. The complex spatiotemporal correlations among the input data are identified through effective training of the proposed model. The major contribution of this article is to propose a data-driven model that combines CNN and LSTM for the morphology evolution prediction of beach nourishment, and validate the effectiveness of the proposed model by comparing with the performances of other popular methods in predicting the nourishment changes.

Convolutional neural network
Long short-term memory
Mega-nourishment
Nourishment morphology evolution prediction

http://resolver.tudelft.nl/uuid:4e11d2c6-fdac-476e-89c9-b4ae20ca0019

https://doi.org/10.1109/ACCESS.2020.3030119

2169-3536

IEEE Access, 8, 184512-184523

Institutional Repository

journal article

© 2020 Yong Li, P.J.M. van Oosterom, Ying Ge, Xiaoxiang Zhang, F. Baart