Analysing uncertainty and sensitivity for riparian revegetation flood mitigation modelling in river networks

Uncertainties of hydraulic parameters in representing land-surface characteristics of revegetating riparian zones remains a critical limitation in hydrodynamic flood modelling, particularly when riparian revegetation is implemented as a nature-based solution (NbS) for downstream flood mitigation. This study presents an integrated and systematic framework for uncertainty and sensitivity analysis to evaluate the effectiveness of riparian revegetation in reducing downstream flood hazard, demonstrated in the Bundamba Creek sub-catchment, Australia. The framework combines efficient parameter space exploration using multiple sampling techniques, rain-on-grid hydrodynamic modelling, and an Adaptive Neuro-Fuzzy Inference System (ANFIS) surrogate model. Key hydraulic parameters, particularly Manning’s roughness associated with vegetation, are analysed within a realistic range of n = 0.07–0.26, representing sparse grass to dense riparian vegetation. An optimal sampling strategy is selected to ensure both global coverage and concentration within the most probable range of roughness ( n ≈ 0.10–0.20). The ANFIS model demonstrates strong predictive performance (NSE = 0.94; RMSE= 564 m 2 ), enabling rapid evaluation of a large number of scenarios. Results reveal that the most significant reduction in downstream high-hazard area occurs within critical roughness intervals ( n ≈ 0.13–0.16 and 0.22–0.24), where small increases in roughness lead to disproportionately large decreases in hazard extent. Sensitivity peaks at AEP = 1%, indicating that NbS effectiveness is maximized under moderate flood conditions, while the influence diminishes for extreme events. Uncertainty analysis demonstrates that maximum uncertainty occurs at intermediate-to-high rainfall thresholds, where flow transitions from channel-controlled to floodplain-dominated conditions, while uncertainty is significantly reduced during extreme events or very high rainfall. Ignoring uncertainty in Manning’s roughness may lead to misestimation of flood hazard, particularly near threshold rainfall conditions. The proposed framework provides a robust and computationally efficient approach for evaluating NbS performance while accounting for uncertainty at the catchment scale.