Abstract
We present the Modified Barnacles Mating Optimizer (MBMO), a novel global optimization algorithm designed to enhance the accuracy of full-parametric, transdimensional, and joint inversion of earthquake-based Horizontal-to-Vertical Spectral Ratio (HVSR) and surface wave dispersion curves. MBMO incorporates advanced techniques to balance global exploration and local exploitation, facilitating efficient navigation of the model space and convergence toward the global optimum. Prior to inversion, we conducted modal analyses and sensitivity studies using a synthetic model to investigate the complexity and uncertainties of the inversion problem, guiding the development of MBMO. The algorithm's performance was rigorously evaluated using synthetic anomalies and field data from the Garner Valley Downhole Array (GVDA) site, with results objectively compared against the original BMO and the classical Particle Swarm Optimization (PSO). To enable transdimensional inversion without substantially increasing computational complexity, we introduced a Multiple Model Space Strategy (MMSS). The solutions from real data were interpreted alongside existing geophysical findings, with post-inversion uncertainty analyses conducted to assess the credibility of the results. Our findings demonstrate that MBMO outperforms both BMO and PSO in addressing complex inversion challenges. The application of MBMO to full-parametric, transdimensional, and joint inversion of earthquake-based HVSR and surface wave dispersion marks a significant advancement in geophysical parameter estimation, offering improved accuracy, reliability, and deeper insights into subsurface structures. Furthermore, the effectiveness of MBMO remains consistent across different objective function definitions-whether based on the average fitness of fitting multiple data sets or the multiplicative feature of different misfits.