Refining relative tropospheric delay modeling of GPT3 for large-height-difference RTK positioning

Real-Time Kinematic (RTK) positioning technique, with centimeter-level accuracy, is widely used in landslide deformation and building structural health monitoring. However, in large-height-difference scenarios where the height difference between the base and rover station is generally over 100 m, RTK positioning performance experiences a significant decline, mainly due to the deficiency in tropospheric delay modeling. In this study, we proposed a new method to refine the modeling of relative tropospheric delay corrections for the widely-used GPT3 by using ERA5 reanalysis data. The refined model, referred to as GPT3-e, shows an improvement of ∼33% for relative tropospheric modeling on a global scale compared to GPT3. Large-height-difference RTK positioning experiments further show that the vertical errors were reduced by 2.9 cm in January and 2.4 cm in July, with corresponding accuracy improvements of 64% and 57%, respectively, and slightly improves the ambiguity-fixing rate. As the refined model does not rely on external real-time products, it is compatible with the current RTK devices which makes the model easy to implement and will not significantly increase the cost.