GABP-TroNet: Enhance Air-Ground Channel Modeling by Improving GZTDS Model for GNSS

Tropospheric delay is an error factor in high-precision positioning of Global Navigation Satellite Systems (GNSS).The simplified GNSS tropospheric delay model (GZTDS) predicts zenith tropospheric delay (ZTD) around the world and shows good global performance with a root mean square error (RMSE) of 3.46 cm. However, in some regions, the GZTDS model performs poorly. This paper intends to improve performance of the GZTDS model in the coastal regions of China by combining the GZTDS model with back propagation (BP) neural network based on genetic algorithm (GABP-TroNet). With the validation of data from the Vienna Mapping Function 1 (VMF1), the mean bias (MB) of the GABP-TroNet model is 0.08 cm and the RMSE is 3.11 cm, which improves the accuracy of predicting ZTD in the coastal regions of China compared to GZTDS.