Characterisation and Mitigation of Outliers in IGS repro3 GNSS ZTD Products

Abstract We evaluate the quality of zenith tropospheric delay (ZTD) estimates from four IGS Analysis Centres (ACs) – COD, GFZ, GRG, and TUG – participating in the third IGS reprocessing campaign (repro3). This study focuses on outliers, defined as estimates outside the 1–3 meter range. A small yet significant fraction of estimates fall outside this range for three ACs that did not constrain this parameter. Most outliers originate from a limited number of stations and frequently occur in clusters or during specific periods (e.g., 2000–2003). We find that 95% of outliers are unique to individual ACs, indicating dependence on their specific processing procedures. Outliers are typically linked to poor data quality, including large multipath effects, cycle slips, and/or strong data editing during pre-processing or processing. They often coincide with suboptimal ambiguity fixing statistics. Other parameters, such as station positions, may also be poorly estimated. outliers can be mitigated through robust screening of estimates. Screening based solely on formal errors may be insufficient, as some datasets processed with a zero-difference network strategy contain outliers despite small formal errors. As an alternative, screening using an independent reference ZTD data from atmospheric reanalysis is recommended.