<p>Still today vaguely defined, the South Atlantic Anomaly (SAA) is the vast<br>geographic region where the Earth&#8217;s magnetic field is weakest relative to an<br>ideal Earth-centered dipole field, and the inner radiation belt comes closest<br>to the planet. Nonetheless it represents a major concern to the space science<br>community, since the local reduced magnetic intensity often results in satellite<br>outages and radiation hazard to humans, especially in geomagnetically disturbed<br>periods.<br>Since 1958, relentless investigation of the various morphological and dynamic<br>features of the SAA has been taking place, robustly relying on field, plasma and<br>particle measurements from Low-Earth-Orbit (LEO) satellites since the late<br>1970s.<br>New readings provided by magnetometers operating at LEO altitudes show that,<br>within the past decade, an apparent second center of minimum field intensity<br>has begun to be clearly resolved southwest of Africa, suggesting a possible rapid<br>splitting of the SAA into two cells. In addition to magnetic determinations, the<br>tracking of fluxes of sub-MeV electrons that are lost to the atmosphere when<br>drifting into the SAA due to its increased bounce loss cone, offers a specular<br>view of the same phenomenon. This multi-messenger approach from different<br>platforms is best suited to catch fine details of the splitting.<br>Directly stemming from the data-adaptive Empirical Mode Decomposition (EMD)<br>developed at NASA in the 1990s for the analysis of non-stationary signals, the<br>Fast Iterative Filtering (FIF) class of signal mode decompositions is recently<br>taking center stage due to enhanced rigorous formalization in terms of con-<br>vergence and stability. Multidimensional and Multivariate FIF (MMFIF) is a<br>brand-new extension that handles multidimensional and multichannel datasets.<br>The application of MMFIF techniques to magnetic-field and particle data from<br>an ensemble of LEO satellites has allowed us to best characterize the dynamic<br>evolution of the SAA lobes in the 2010s, and compare it to analogous data in<br>the literature from the previous decades.</p>