Strong motion baseline correction based on acceleration smoothness priors and co-located GNSS static displacement constraints

Strong-motion acceleration records are crucial for seismic research and vibration analysis. Still, baseline offsets often introduce drift in displacement integration estimates, compromising the accuracy of the coseismic displacement retrieval. By providing precise ground deformation signals, the high-rate Global Navigation Satellite System (GNSS) offers ideal baseline correction constraints. In this paper, we propose a baseline correction method based on acceleration smoothness priors and co-located high-rate GNSS static displacement constraints. First, accelerations are processed using the smoothness priors method (SPM). This method separates steady-state acceleration signals from potential non-periodic noise trends through regularization. Second, static displacements from co-located high-rate GNSS stations are applied as external constraints to refine a step-fitting function. The optimal baseline correction time parameters are iteratively determined through a grid search method. Finally, the displacement time series is then fitted with this constrained step-fitting function to achieve baseline deviation correction of acceleration records. A shake table experiment and two seismic events validated the proposed baseline correction method. In the shake table experiment, the corrected displacement time series returned to the zero line, preserving long-period and permanent displacement information, with a root mean square (RMS) of 0.585 cm and a correlation coefficient (CC) of 0.964. For the 2023 Turkey earthquake doublet, the corrected strong-motion displacements showed good agreement with GNSS data, achieving an average RMS of 2.720 cm and a CC of 0.799. For the 2021 Maduo Mw 7.4 event, the method yielded an RMS of 0.585 cm and a CC of 0.964, with an average RMS of 1.157 cm and a CC of 0.592 in all directions. This demonstrates its potential as a key technique for accurately retrieving source parameters and finite fault slip inversion from strong-motion accelerometer data.