Improved GNSS processing and screening for ZTD monitoring in national network

Discontinuities in Zenith Total Delay (ZTD) series estimated from ground-based GNSS data are detrimental to climate monitoring applications. This study aims at improving the processing strategy to reprocess GNSS data from national networks. A network of 136 GNSS stations (including 32 stations from the EUREF Permanent Network and 104 stations from ASG-EUPOS network in Poland) was reprocessed for year 2014 using Bernese GNSS Software v.5.2. The average distance between the stations is about 70 km. In order to investigate the influence of network design strategy, three variants of processing were carried out and analyzed: 1) using network design strategy which works quite well for coordinates; 2) the standard "obs-max" strategy recommended for Bernese software; 3) a newly developed baselines design strategy optimized for ZTD estimation. The study shows that the network design has a strong impact on the quality and continuity of ZTD estimates. In case of sub-daily gaps in the measurements at reference stations, small clusters of stations can be disconnected from the main network in the first network strategy. This has little impact on coordinates, but offsets of a few centimeters in ZTD estimates and spikes in their formal errors can appear at all stations of the disconnected cluster. It is also responsible for significant discontinuities in the estimated ZTD series. This problem is observed also with "obs-max" strategy. The new network design strategy proposed here allows to avoid this problem. The reprocessed ZTD time series are much more continuous and homogeneous in comparison to the standard strategies. However, even when an optimal processing procedure is used, some outliers in the ZTD series still remain (e.g. due to short data gaps). Therefore, a post-processing screening procedure consisting of the removal of the first and the last ZTD estimates around observations gaps, range and outlier check of ZTD and formal errors was applied to remove these bad values. The screening removes less than 2% of the ZTD data. As a final screening and validation step, GNSS ZTD estimates are compared to ERA-Interim. We show that the agreement between GNSS and ERA-Interim results with the new baseline strategy and screening shows a significant improvement.