Correlated errors in GPS position time series: Implications for velocity estimates

摘要: [1] This study focuses on the effects of time correlation in weekly GPS position series velocity estimates. Time 2.5 to 13 years long from a homogeneously reprocessed solution 275 globally distributed stations are analyzed terms noise content and uncertainty assessment. Several models were tested, including power law Gauss-Markov processes. The best model describing our global data set was combination variable white with mean amplitudes ∼2 mm ∼6 mm, respectively, for sites considered. This provided vertical ∼0.3 mm/yr, 4–5 times larger than uncorrelated assumption. We demonstrated that correlated is dependent length and, especially, period. 2–3 oldest contain amplitude similar found 12 years. period should be taken into account when estimating content, comparing different estimations, or applying an external estimation assess uncertainty. showed dependency cannot explained by increasing tracking network ambiguity fixation rate but probably related amount quality recorded data.