Bayesian Inference of Local Paleosecular Variation From Sparse Paleomagnetic Data

Abstract We introduce a novel Bayesian method to construct local paleosecular variation (PSV) curves. By modeling the geomagnetic field as a Gaussian process, global models can be incorporated as informative prior distributions. Because we use Hamiltonian Monte‐Carlo methods, complex age distributions resulting for example from radiocarbon calibration can be incorporated in the model. This improves PSV reconstruction in poorly sampled intervals and allows age refinement of individual records. We validate the method with synthetic data and apply it to the Azores, demonstrating its effectiveness in data‐limited regions. The approach is implemented in an open‐source Python package.