Extended Data Figure 3: Interseismic GPS data corrections, slip deficit estimation and sensitivity tests for interseismic locking inversion.

a, Demonstration of the effect of sliver and shortening corrections on the interseismic GPS data. The left plot shows the data in the stable South American reference frame. Red vectors indicate the stations that the corrections were applied to. All stations underwent the sliver correction. Stations in the northeast and southeast underwent shortening corrections. b, Slip deficit estimation. The left panel shows our locking model; the central panel shows a compilation of events since 1877 according to ref. 1 plus the Antofagasta and Tocopilla earthquakes of 1995 and 2007 (refs 4, 7), as well as our solutions for the Pisagua mainshock and largest aftershock. The right panel shows moment density along the trench projected on latitude. The total accumulated moment corresponds to a M 8.97 event. This is about one-sixth of the moment released during the 1960 Valdivia M 9.5 event further south, but sixfold that released in our region of interest between 1877 and now according to the events listed in the central panel summing to a magnitude of 8.41. Even though the Pisagua sequence released a significant amount of the moment in the northernmost part, the remaining moment would still correspond to M 8.92. c, Model smoothness plotted against residual. The optimal smoothing factor of 0.05 in the corner of the L-curve resulted in a residual of 0.17 cm yr⁻¹. d, A selection of solutions with different smoothing factors. The central solution is the one we prefer. Black lines are 1-m isolines of the co-seismic slip distribution of the mainshock and the largest aftershock. e, Checkerboard tests of locking. Top: forward models consisting of three and two rows of locked patches. Lower panels: inverted locking patterns using the signal from the forward models at the GPS station positions applying the same uncertainties as in the actual observation data. For three locking rows, the trenchward row is clearly missed, whereas the areas closer to the station positions (magenta) are captured fairly well, the resolution being about 40 km.