Fig. 3

Results from frequency-domain robust principal component analysis of electric and magnetic fields from the array shown in Fig. 1, which includes the two sites from the southeastern USA used in Fig. 2. a Power spectrum of six dominant modes, given in signal-to-noise units expressed in decibels (dB). Signal power above 0 dB indicates that the mode is above the incoherent noise level and therefore significant. If the uniform source assumption that underlies the MT method held exactly, there would only be two significant modes (shown in red). In the FLR period band for this array (~ 20–40 s), the curve for the third mode becomes significant. Additional significant modes at longer periods are also associated with source effects (primarily in the vertical magnetic fields) due to increasing skin depths with respect to large-scale source structure (e.g., Jones and Spratt 2002). b–d Linear combinations of the first three spatial modes at a period of 26 s, displayed as complex vectors plotted on a map of array sites. The first two panels show linear combinations, defined following the approach in Egbert (1997), that best approximate uniform magnetic sources, linearly polarized geomagnetically N–S and E–W. Panel d represents the remaining signal in the first three modes. This mode exhibits a clear north–south spatial gradient and very large vertical magnetic components, consistent with expectations for a spatially localized FLR signature. In these plots, horizontal magnetic (electric) fields are shown with blue (red) arrows, with real parts solid and imaginary parts dashed. The vertical magnetic field is shown with a blue circle, scaled as for the horizontal components to indicate amplitude, with the position of the drawn radius denoting phase. To scale the lengths of electric and magnetic field vectors in a physically meaningful way, a reference Earth resistivity (given in each panel) is assumed