Data-driven modeling for magnetic field variations using the GLO-MAP algorithm

This paper presents an application of the global-local orthogonal mapping (GLO-MAP) algorithm to derive data-driven models for magnetic field variations of the Earth. The GLO-MAP algorithm rigorously merges different independent local approximations that are based upon measured data to obtain a desired order, globally continuous approximation. We show the local magnetic field data acquired by ground-based survey and discuss details of the survey process. A potassium vapor magnetometer is used in the field experiments to obtain accurate observations that form the basis of the local modeling. Numerical results based on the experimental data show that the GLO-MAP algorithm can accurately and efficiently map the magnetic field variations, while a single global polynomial based modeling approach produces over 30% of the approximation error in the worst case.