Manganese Transport and Sediment Remobilization in a Tidal Estuary: Parameter Sensitivity and Uncertainty Analysis

Tidal estuaries are characterized by complex hydrodynamic processes, while heavy metal transport is governed by multiple interacting physical mechanisms. These complexities introduce uncertainties in key model parameters and may compromise the reliability of model predictions. This study developed a three-dimensional hydrodynamic-heavy metal transport model (SCHISM-HM) for tidal estuaries and applied it to the Danshuei River estuarine system in Taiwan to systematically investigate manganese transport mechanisms, parameter sensitivity, and prediction uncertainty. The model was validated against field observations collected in 2020. The simulated hydrodynamic variables, including water level, velocity, and salinity, showed good agreement with observations, with skill scores (SS) exceeding 0.8. For total manganese concentrations, the mean absolute error (MAE) ranged from 0.001 to 0.006 mg/L, the root mean square error (RMSE) from 0.001 to 0.008 mg/L, and the SS from 0.909 to 0.999, demonstrating the capability of SCHISM-HM to reproduce the complex hydrodynamic conditions and heavy metal transport processes in the Danshuei River estuary. Sensitivity analysis revealed that both the partition coefficient (K_d) and suspended sediment settling velocity (w_s) were negatively correlated with total manganese concentrations in the water column. Increasing K_d from 0.1 to 0.4 L/mg resulted in reductions in peak total manganese concentrations of 32.03-93.75% across the monitoring stations. Conversely, decreasing w_s from 7.0 to 3.0 m/day increased total manganese concentrations at downstream stations by up to 382.08%. These results indicate that w_s exerts a stronger influence on total manganese concentrations than K_d, highlighting the critical role of suspended sediment transport in controlling heavy metal distributions. Model uncertainty was further quantified by integrating Monte Carlo simulation with Bayesian inference. The posterior distributions of both K_d and w_s were best represented by normal distributions, with correlation coefficients (R) exceeding 0.997. Uncertainty in simulated total manganese concentrations was predominantly associated with peak-concentration periods and exhibited pronounced spatial variability. In particular, the Taipei Bridge station, where river cross-sectional contraction enhances hydrodynamic forcing and sediment remobilization, exhibited the largest 90% confidence interval width (0.118 mg/L) and a relatively high probability of exceeding the water quality standard. Overall, the results demonstrate that bed sediment erosion and suspended sediment transport are key processes governing manganese dynamics in the Danshuei River estuary. The proposed modeling and uncertainty-analysis framework provides a quantitative basis for water pollution management and environmental risk assessment in tidal estuaries.