Kevin R. Mackie

Hurricane storm surges pose a major threat to coastal and ocean engineering infrastructure, but are difficult to characterize using historical data alone; numerical models are essential for flood hazard assessment. The magnitude and extent of simulated storm surges are sensitive to both meteorological and geophysical factors, which are inherently uncertain and vary across space and time. In this study, we examine the temporal and spatial variability of storm surge sensitivities using first-order second-moment methods. We also use this method to understand how sensitivities vary with storm surge levels. Three historical hurricanes serve as case studies. Maximum surge levels are primarily influenced by hurricane intensity, size, and background sea level. However, sensitivities are dynamic and vary by location and evolve over the duration of a storm. Parameters most influential for peak surge are not always dominant at the time of maximum water level, and the importance of factors depends on surge magnitude, with some variables negligible at low surge levels becoming critical at high surge levels. This framework provides a systematic approach to quantifying storm surge sensitivity and can guide parameter selection in data-driven models, enhancing their ability to predict extreme surge events while accounting for temporal and spatial variability.

Compared to linear analysis approaches, nonlinear analysis of bridge models under large seismic demands has enabled more realistic prediction of global and local responses. However, the sensitivity to modeling assumptions, element and material formulations, software-dependent implementations, and parameter selection may lead to variable results between nonlinear analysis software packages. While previous work has led to a better understanding of how to model nonlinear static response of bridge components and systems, seismic loads and the corresponding material hysteresis introduce additional sources of variability in the nonlinear response. Two ordinary standard bridges in California were analyzed using simplified steel and concrete constitutive models in concentrated plasticity elements for the columns and nonlinear gap-link springs for the abutments. The models were implemented in SAP2000 and OpenSees and calibrated to achieve common material, section, and element-level nonlinear static responses. Subsequently, the models were extended to seismic excitation with common unloading-reloading rules for the material models as well as common definitions for damping and mass. Analysis results showed consistent drift, base shear, and moment-curvature response histories between the software packages. However, the nonlinear seismic response sensitivity for alternative column hinge formulations and abutment models should be investigated in future studies.

Reliable and robust numerical modeling of RC columns under lateral load is commonly performed using one-dimensional elements. Although the strain is assumed to vary linearly through the section, bond-slip and buckling of reinforcing bars at the joints/connections contribute substantially to the column response. This paper presents a three-dimensional (3-D) fiber-based model with explicit representation of the bond-slip that occurs at the column-footing interface for cast-in-place columns and at the interface between bars and mechanical couplers for precast columns with grouted sleeve (GS) connections. The model contains beam elements with fiber sections for concrete, beam elements for longitudinal steel bars and sleeves, and spring elements for interfacial bond-slip behavior. The novelty is that the model can be achieved using existing software implementations, retains the efficiency of fiber-based elements under cyclic loading, and parameters are calibrated based on component-level testing, unlike other resultant or hybrid models that require arbitrary backbone and cyclic property specification. An extensive experimental program was conducted to investigate the behavior of GS couplers and calibrate the bond-slip models. Four large-scale RC columns were then used to verify the proposed model by comparing the measured load-displacement, energy dissipation, curvature, and failure mode with the numerical responses.