Global climate change has resulted in a gradual sea level rise. Sea level rise can cause saline water to migrate upstream in estuaries and rivers, thereby threatening freshwater habitat and drinking-water supplies. In the present study, a three-dimensional hydrodynamic model was established to simulate salinity distributions and transport time scales in the Wu River estuary of central Taiwan. The model was calibrated and verified using tidal amplitudes and phases, time-series water surface elevation and salinity distributions in 2011. The results show that the model simulation and measured data are in good agreement. The validated model was then applied to calculate the salinity distribution, flushing time and residence time in response to a sea level rise of 38.27 cm. We found that the flushing time for high flow under the present condition was lower compared to the sea level rise scenario and that the flushing time for low flow under the present condition was higher compared to the sea level rise scenario. The residence time for the present condition and the sea level rise scenario was between 10.51 and 34.23 h and between 17.11 and 38.92 h, respectively. The simulated results reveal that the residence time of the Wu River estuary will increase when the sea level rises. The distance of salinity intrusion in the Wu River estuary will increase and move further upstream when the sea level rises, resulting in the limited availability of water of suitable quality for municipal and industrial uses.

A flood risk assessment of urban areas in Kaohsiung city along the Dianbao River was performed based on flood hazards and social vulnerability. In terms of hazard analysis, a rainfall-runoff model (HEC-HMS) was adopted to simulate discharges in the watershed, and the simulated discharges were utilized as inputs for the inundation model (FLO-2D). Comparisons between the observed and simulated discharges at the Wulilin Bridge flow station during Typhoon Kongrey (2013) and Typhoon Megi (2016) were used for the HEC-HMS model calibration and validation, respectively. The observed water levels at the Changrun Bridge station during Typhoon Kongrey and Typhoon Megi were utilized for the FLO-2D model calibration and validation, respectively. The results indicated that the simulated discharges and water levels reasonably reproduced the observations. The validated model was then applied to predict the inundation depths and extents under 50-, 100-, and 200-year rainfall return periods to form hazard maps. For social vulnerability, the fuzzy Delphi method and the analytic hierarchy process were employed to select the main factors affecting social vulnerability and to yield the weight of each social vulnerability factor. Subsequently, a social vulnerability map was built. A risk map was developed that compiled both flood hazards and social vulnerability levels. Based on the risk map, flood mitigation strategies with structural and nonstructural measures were proposed for consideration by decision-makers.

This study employs a deep learning-based object detection model, GoogleNet, to identify cracks in cultural heritage images. Subsequently, a semantic segmentation model, SegNet, is utilized to determine the location and extent of the cracks. To establish a scale ratio between image pixels and real-world dimensions, a parallel laser-based measurement approach is applied, enabling precise crack length calculations. The results indicate that the percentage error between crack lengths estimated using deep learning and those measured with a caliper is approximately 3%, demonstrating the feasibility and reliability of the proposed method. Additionally, the study examines the impact of iteration count, image quantity, and image category on the performance of GoogleNet and SegNet. While increasing the number of iterations significantly improves the models’ learning performance in the early stages, excessive iterations lead to overfitting. The optimal performance for GoogleNet was achieved at 75 iterations, whereas SegNet reached its best performance after 45,000 iterations. Similarly, while expanding the training dataset enhances model generalization, an excessive number of images may also contribute to overfitting. GoogleNet exhibited optimal performance with a training set of 66 images, while SegNet achieved the best segmentation accuracy when trained with 300 images. Furthermore, the study investigates the effect of different crack image categories by classifying datasets into four groups: general cracks, plain wall cracks, mottled wall cracks, and brick wall cracks. The findings reveal that training GoogleNet and SegNet with general crack images yielded the highest model performance, whereas training with a single crack category substantially reduced generalization capability.

Anthropogenic changes in tidal estuaries have significantly altered bathymetry and topography over the past half century. The geomorphic-driven changes in estuarine hydrodynamics and salt transport remain unclear. To explore this issue, a SELFE (Semi-implicit Eulaerian-Lagrangian Finite Element) -based model was developed and utilized in a case study in the Danshui River, Taiwan. The model was calibrated and validated using observed water level, current, and salinity data from 2015, 2016, and 2017. The performance of the SELFE model corresponded well to the measured data. Furthermore, the validated model was utilized to analyze the hydrodynamics, residual current, limit of salt intrusion, and residence time under the predevelopment (1981) and present (2015) conditions. The predicted results revealed that the time lag of water surface elevation at both high tide and low tide under the present condition was approximately 0.5–2 h shorter under the predevelopment condition. The residual circulation under the predevelopment condition was stronger than under the present condition for low flow, causing the limit of salt intrusion to extend further upstream under the predevelopment condition compared to the limit of salt intrusion under the present condition. The calculated residence time under the predevelopment condition was longer than the residence time under the present condition. The freshwater discharge input is a dominating factor affecting the salt intrusion and residence time in a tidal estuary. A regression correlation between the maximum distance of salt intrusion and freshwater discharge and a correlation between residence time and freshwater discharge were established to predict the limit of salt intrusion and residence time under the predevelopment and present conditions with different scenarios of freshwater discharge input.

The effects of environmental factors on suspended sediment plumes in the continental shelf out of the Danshuei River estuary were numerically investigated using an unstructured-grid three-dimensional hydrodynamic model (SCHISM) together with a suspended sediment (SS) module. The coupled model (SCHISM-SS) was calibrated and validated against the in situ measurement data in 2016. Consistent with the observation results, the model simulations satisfactorily reproduced the water levels, velocities, salinities, and suspended sediment concentrations. The model was then applied to explore the role of various environmental factors in the dynamics of suspended sediment plumes from the estuary to the adjacent coastal seas. These factors include tidal forcing, salinity, river discharge, and wind stress. Analysis and comparisons of different scenario results indicated that the suspended sediment plume was greatly affected by tides, e.g., a longer plume distance resulted from a larger flux under tidal motions. A higher sediment concentration in the plume in the offshore area was also found during the neap tide, relative to that observed during the spring tide. In addition, salinity affects the movement of density currents and the spread of the sediment plume, i.e., the plume distance is longer due to the residual circulation when a salinity difference is present. Further, an extreme river flow could occur during typhoon periods and would discharge a greater water volume into the coastal region, causing the suspended sediment plume to expand from the near shore. Finally, the directions of prevailing winds can slightly influence the sediment plumes.

Taiwan faces intense rainfall during typhoon seasons, leading to rapid increases in water level in rivers. Accurate flood forecasting in rivers is essential for protecting lives and property. The objective of this study is to develop a river flood forecasting model combining multiple additive regression trees (MART) and ensemble Kalman filtering (EnKF). MART, a machine learning technique, predicts water levels for internal boundary conditions, correcting a one-dimensional (1D) unsteady flow model. EnKF further refines these predictions, enabling precise real-time forecasts of water levels in the Danshui River system for up to three hours lead time. The model was calibrated and validated using observed data from four historical typhoons to evaluate its accuracy. For the present time at three water level stations in the Danshui River system, the root mean square error (RMSE) ranged from 0.088 to 0.343 m, while the coefficient of determination (R²) ranged from 0.954 to 0.999. The validated model (module 1) was divided into two additional modules: module 2, which combined the ensemble unsteady flow model with inner boundary correction and MART, and module 3, which featured an ensemble 1D unsteady flow model without inner boundary correction. These modules were employed to forecast water levels at three stations from the present time to 3 h lead time during Typhoon Muifa in 2022. The study revealed that the Tu-Ti-Kung-Pi station was less affected by inner boundaries due to significant tidal influences. Consequently, excluding the upstream and downstream boundaries, Tu-Ti-Kung-Pi station showed a superior RMSE trend from present time to 3 h lead time across all three modules. Conversely, the Taipei Bridge and Bailing Bridge stations began using inner boundary forecast values for correction from 1 h to 3 h lead times. This increased the uncertainty of the inner boundary, resulting in higher RMSE values for these locations in modules 1 and 2 compared to module 3.