Subsurface Imaging via Joint Inversion
Objective
Inversion is the process of working backwards from observations to estimate the Earth's structure. Joint inversion utilizes multiple datasets (each sensitive to different properties) to build a single, consistent model. In this project, we extended the open-source BayHunter framework (an MCMC joint inversion tool for Surface Wave Dispersion and Receiver Functions) by adding a custom gravity anomaly module.
Methodology & Implementation
While SWD and RF are excellent for constraining velocity and interfaces, neither inherently accounts for density. To fix this, we integrated Gravity Anomaly to directly measure density variations, calculating it by summing the Bouguer Slab effect for each layer within the model.
During implementation, density was not isolated as a completely independent variable to be inverted on its own. Instead, we modified the standard linear relationship into a non-linear density equation dependent on both Vp and Vs. The model is then iteratively adjusted through the MCMC chain until the predicted and observed data match as closely as possible. This modified framework was initially validated using a Synthetic Test before being applied to real-world data from the HYB station.
Results & Limitations
Currently, the addition of gravity into the loop seems to be reducing the overall quality of the results rather than improving them. Specifically, the inclusion of gravity appears to artificially reduce the Vp/Vs ratio, which may be tied to the reference model being used or the reliance on the empirical Vp/Vs equation.
Future Work
This framework is still a work in progress. Future improvements include identifying the optimal parameters to stabilize the inversion, improving the gravity forward model itself, and running the inversion across multiple stations in a line to better isolate where the current model is failing. A crucial next step will be to isolate density as a completely separate variable to be inverted for on its own.