Uncertainty Quantification in Probabilistic Tsunami Hazard Assessment (PTHA): A Case Study in South China Sea

In this study, uncertainty quantification and parametric sensitivity analysis in Probabilistic Tsunami Hazard Assessment (PTHA) are performed for the South China Sea using a Monte Carlo approach. Uncertainties in parameters such as the magnitude–frequency distribution of the potential tsunami zone, geodetic information used to constrain the maximum magnitude, properties determining the slip distribution, scaling laws and dip of unit sources are considered, each of which is varied separately while the others are fixed. The Coefficient of Variation (COV) of tsunami amplitudes corresponding to a fixed annual probability at different coastal sites is used to represent the uncertainty of each parameter. In addition, the overall tsunami hazard and uncertainty are also presented by simultaneously varying all parameters in a Monte Carlo series. Our results suggest that a major contributor to the uncertainty is the magnitude frequency distribution parameters, especially the maximum magnitude. Geodetic information can be used to solve the problem that the maximum magnitude is underestimated owing to the scarcity of mega earthquakes in the historical earthquake catalog, while it will also lead to a large uncertainty if the parameters in it are not well determined. Among these geodetic parameters, the seismic coupling coefficient is the parameter that most influences the uncertainty as it is difficult to determine an accurate value. The effect of the slope parameter [Formula: see text] in the magnitude–frequency relationship is complicated because of its influence in determining the maximum magnitude and the number of earthquake events. In addition, the uncertainty associated with the down-dip width of the seismogenic zone is not remarkable but cannot be ignored, which is similar to that of the annual plate slip rate, while the effect of the rigidity is relatively small because its effect on the average slip of earthquake events will reduce the uncertainty. Compared to the uncertainty in determining the maximum magnitude, the effect of slip distribution properties such as Hurst exponent, correlation length and scaling exponent and the choice of scaling laws is relatively small. The uncertainty in the dip angle of the unit source is moderate and cannot be ignored, especially for sites along the strike of the subduction zone. Tsunami curves for four coastal sites indicate that the tsunami hazard in the South China Sea is subject to large variations when considering all uncertainties in the PTHA, with the COV in 2000 years at different sites being about 0.25, which means that the 95th percentile is about two times of the 5th percentile.