Prediction of shoreline change using self-attention-based spatiotemporal network from optical satellite images

초록

Prediction of shoreline change is essential for coastal management due to concentrated coastal development and sea level rise caused by climate change. In this study, we extract shorelines from satellite images acquired over a long period of time and predict changes in the shoreline that respond to coastal hydrodynamics by modeling the relationship between the shoreline and external ocean forces. We obtain satellite images spanning over 40 years for two sand beaches on the east coast of Korea, where coastal erosion problems are emerging, and extract shorelines using a coarse mask-guided image segmentation network. In addition, ocean waves and tides are obtained and applied to modeling shoreline changes using a self-attention-based spatiotemporal network. The accuracy of shoreline prediction is evaluated using root mean square error, and correlation coefficient.

키워드