Global stochastic model adaptation for BDS-only static precise point positioning using post-fit residual statistics

Abstract

Precise Point Positioning (PPP) performance is highly dependent on the consistency between the assumed measurement stochastic model and actual observation errors. This study implements and evaluates a lightweight global adaptive variance-scaling strategy for BDS-only static PPP. A single bounded scaling factor is used to uniformly adjust baseline elevation- and satellite-dependent variances, and is updated epoch by epoch via exponential smoothing of the root-mean-square of post-fit standardized residuals. Experiments were conducted using five globally distributed IGS MGEX stations, with three consecutive 24-hour arcs processed in four seasonal windows (DOY 001-003, 091-093, 182-184, and 274-276, 2025). B1I/B3I dual-frequency ionosphere-free observations and CODE precise orbit/clock products were adopted, while all processing options were kept identical except for enabling or disabling adaptive scaling. Positioning performance was evaluated by East-North-Up (ENU) RMS relative to SNX reference coordinates. Across all station-season groups, the adaptive solution shows non-negative coordinate-domain improvements in all components: the average East RMS decreases from 3.56cm to 3.23cm (about 9.5%), the average North RMS from 3.34cm to 3.25cm (about 2.5%), and the average Up RMS from 15.43cm to 15.00cm (about 3.1%). In addition, the outlier-screened epoch-wise normalized residual RMS is consistently higher under the adaptive solution, with adaptive-minus-original differences generally around 0.21-0.24 across stations and seasonal windows, indicating a systematic adjustment of residual-to-noise balance. These results demonstrate practical benefits of the proposed global adaptive strategy for BDS-only static PPP under the tested configuration.