Understanding and reducing ZTD outliers in GNSS PPP-derived products

Zenith Total Delay (ZTD) estimates derived from GNSS observations are essential for atmospheric and geodetic applications. When processed using Precise Point Positioning (PPP), ZTD time series exhibit enhanced stability compared to network-based approaches. However, occasional outliers - ranging from a few centimetres to several meters - still occur, potentially degrading product quality and impacting downstream applications. The mechanisms driving these anomalies remain poorly understood, and their characterisation is critical for improving PPP-based ZTD products. This study examines the nature, origins, and possible mitigation strategies for such outliers in order to enhance the reliability of GNSS-derived tropospheric parameters.

We perform sensitivity tests using the CNES’s GINS software in PPP mode with integer ambiguity resolution, complemented by simplified PPP-like simulations, to identify the mechanisms underlying ZTD outliers. Particular attention is given to pre-processing procedures, which are critical for detecting and handling problematic observations and significantly impact ZTD accuracy. Building on this diagnostic phase, we explore parameter regularisation strategies aimed at mitigating the occurrence of ZTD outliers while preserving high processing quality. These analyses provide insights into both the origin of anomalies and practical approaches for improving the robustness of PPP-based tropospheric products.

In addition, we investigate complementary post-processing screening methods based either on purely statistical approaches or on the comparison with independent atmospheric reanalysis ZTD data. Combined with the strategies described above, these methods aim to reduce ZTD outliers while preserving geophysical variability. This integrated approach enhances GNSS positioning performance and improves the reliability of long-term GNSS-derived tropospheric time series, supporting climate monitoring and other atmospheric applications.