Empirical modelling of site-specific errors in continuous GPS data

Continuous global positioning system (GPS) stations propagate biases and spurious signals into the derived parameter time series when the measurements are subject to site-specific effects, such as multipath. This is a particular problem in the investigation of geophysical and atmospheric phenomena where signals may be small in magnitude. A methodology to remove these erroneous signals from long-term time series will significantly increase the usefulness of the derived time series. This work provides the theoretical basis for use of an empirical site model (ESM) derived from post-fit phase residuals to mitigate unmodelled site-specific errors. Additionally, we also investigate the effectiveness of applying an ESM to a regional GPS network and a short baseline solution. Under most observing scenarios, we show that the ESM approach is predicted to improve the precision and accuracy of the site coordinates. However, it is important to note that we found some scenarios where the ESM can introduce a bias. For instance, when the antenna is mounted close to the ground. In this scenario, for a short baseline, we observed the introduction of a 4-mm bias in height. Use of an ESM for the same short baseline with an uncalibrated radome substantially improves the results by removing a large bias of over 10 mm in height. Similarly, application of an ESM derived from historic data yields similar improvements. This demonstrates that the ESM can be a powerful tool when applied to appropriate site-specific configurations and could potentially be implemented in routine GPS analysis for a broad range of applications.