Inner core rotation and its variability from nonparametric modeling

We present a new approach for studying the inner core rotation and its variability by separating the underlying structure from its time evolution. This is achieved by fitting a large set of existing seismic BC‐DF traveltime data with a smoothing spline analysis, which is implemented in the statistics package R. This method allows us to separate the time‐independent mantle contribution from the time‐dependent core contribution without any a priori constraints and to also estimate the error of the fit. We add our newly acquired seismic data from the PASSCAL experiment ARCTIC in northern Alaska to PKP differential traveltime measurements previously obtained from South Sandwich Islands earthquakes and recorded at College (COL) and the Alaskan Seismic Network. This nearly doubles the number of measurements to a total of 1165 and increases the data coverage of the inner core structure and our time coverage (to about 55 years). The large number of measurements allows us to use the standard statistical method of bootstrapping to derive the rotation rate and thus to further separate the core structure from its time evolution. This approach has been successfully tested with synthetic data sets that feature both a nonlinear structure and variable rotation rates, while sharing the same earthquakes and stations geometry as our real data. Applying this method to the real data yields an average rotation rate of 0.39° ± 0.09 0.22 ° yr −1 (at 68% confidence level). Our results also suggest a nonzero acceleration of the inner core with an apparent change in the rotation rate from 0.24° to 0.56° yr −1 within the last 55 years. The minimum torque acting on the inner core is estimated 1.19 × 10 16 N m, which could result from the imbalance of the electromagnetic and gravitational torques.