Analysis of GNSS time series noise characteristics and detection of station velocity outliers based on improved BIC and U-test

The presence of colored noise in GNSS coordinate time series can affect the accurate estimation of station velocity and velocity uncertainty.Accurate noise model identification is crucial for obtaining reliable and high-precision velocity estimates and their true confidence intervals.To address this, accurate noise model identification is critical for high-precision velocity estimates.The study selected 37 GNSS coordinate time series on the West Coast of the United States from 1999 to 2023, and we adopt a novel Bayesian Information Criterion with temporal prioritization (BIC_tp), extending the standard BIC by incorporating temporal factors to detect colored noise in GNSS coordinate time series, and compares it with BIC and the Akaike Information Criterion (AIC).The results show that BIC_tp effectively identifies distinct noise profiles, with the north (N) and east (E) components dominated by flicker noise plus random walk plus white noise (FNRWWN) and the up (U) component by power-law plus white noise (PLWN).Subsequently, the Mann-Whitney U-test and error propagation analysis assess the impact of these noise models on station velocity and annual amplitudes.Mean station velocity is significantly positive (0.07-0.20 mm/yr, 95 % confidence interval) with no outliers in velocity uncertainty.The U-test indicates that 70.3 % and 86.5 % of annual displacement a mplitudes in the N and E components, respectively, are below 1.0 mm, suggesting stable horizontal motion, while vertical (U) displacements show greater annual variations.Non-parametric testing enhances the detection of anomalous velocity and periodic terms, improving velocity and amplitude estimate accuracy by identifying data processing errors.