In order to discuss the variation of saltwater intrusion and mixing types in terms of estuarine bed topography, planform shape, channel cross-section and multi-branch, a total of 25 numerical experiments were carried out with a conceptual estuary using a three-dimensional hydrodynamic simulator. The wavy bottom of the channel had less salinity intrusion length (SIL) compared with the flat bottom (reference case) under the constant tidal range and freshwater discharge. This is because of the decrease in velocity of the gravitational flow due to the bottom drag force as well as the trapping of saltwater in the bottom hollows. The SIL was increased in the case of funnel shaped estuary compared with the straight (constant width) channel. For the channel cross-section, the temporal variation of SIL was highest in the case of the triangular cross-section and the mixing condition was changed from partially mixed to stratified with the change in cross-sectional shape from triangular to parabolic and then to the rectangular cross-section. The results from the multi-branch indicated that the sub-channels with different length would affect the saltwater intrusion and mixing condition in the estuary. It was shown from the present numerical experiments that the saltwater distribution in the estuary was significantly affected by the planform shape, bed topography and channel cross-section.
Numerical wave prediction models require a large amount of computational power to timely complete the required calculations. Artificial Neural Networks (NN) have been introduced to perform predictions at a lesser computational cost and increased processing speed. Deep learning and specifically Convolutional Neural Networks (CNNs) have become accepted for various image recognition applications. Motivation for the examination of wave prediction by deep learning came from the success of CNNs in vision applications and the similarity of meteorological weather grid data to visual images. This study investigates a deep learning technique using the Japan Meteorological Agency’s Grid Point Value Mesoscale Model to predict wave height and period along Japanese coasts of the Sea of Japan. In particular, this study uses the Xception deep learning architecture with depthwise separable convolution to obtain improved wave height and period prediction over artificial neural networks, and gets overall success results.
In recent years, in Vietnamese Mekong Delta (VMD) riverbank erosion and collapse have been excessively occurring in many rivers, especially in small rivers, and threatening people living near the riverbank not only their properties but also their future, even their lives. Erosion and collapse are predicted to increase significantly under the influence of tidal range, sea level rise (SLR), and land subsidence. To confront with erosion and riverbank collapse, small rivers should be intensively studied together with large rivers as most recent studies. However, making a research on small rivers in VMD will be very difficult because of the lack of hydraulic data. The main objective of this study is to demonstrate how to conduct a practical flow modelling for a small tidal river in case of only time-series water level at the mouth available. The results of this study are concentrated in three important points. First, a new searching method for convenient interpolation methods was proposed to reproduce the river bathymetry with sparse depth samples. Second, it was found that Riemann boundary condition is very helpful in case of lack of upstream discharge data but need to be modified to be compatible with 2D flow model. Finally, it was demonstrated that flow model of a river can be easily simulated in long period by applying downstream tidal data and Riemann condition. This research will be helpful for other studies with similar field conditions in the future.