Simultaneous estimation of hyperparameters and basement depth in gravity data inversion using the JAYA algorithm

Basement relief depth estimation is an important component in characterizing sedimentary basins, which serve as essential reservoir for geothermal energy, groundwater, and hydrocarbons. While residual gravity data inversion is a primary tool for this estimation, the process is inherently ill-posed and non-unique. Conventional global optimization methods (GOMs) often mitigate these issues through subjective filtering or smoothing constraints, yet they typically require exhaustive parameter tuning and high computational costs. To address these limitations, this study introduces a robust variant of the JAYA algorithm designed to optimize an objective function that incorporates the L2-norm of both the error and the model. A significant advantage of this approach is its ability to optimize Lagrange multiplier (known as hyperparameters) and basement depth parameter simultaneously, thereby improving objectivity and avoiding overly complex solutions. The performance of the JAYA algorithm was rigorously validated using synthetic gravity data (noise-free and noise-added) and real-world gravity anomalies from graben areas in Türkiye and Indonesia. Comparative analyses demonstrate that JAYA outperforms traditional GOMs, such as Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Differential Evolution (DE), in terms of convergence speed and structural robustness. Furthermore, the uncertainty appraisal analysis indicates that the proposed approach provides basement depth estimations within a reliable confidence interval. The resulting models show high correlation with published geophysical literature and lithological logs, confirming the algorithm’s efficacy as a precise and objective tool for subsurface resource exploration.