Accuracy Analysis of GPT2/GPT2w Models for SLR-Based Satellite Orbits Validation

The meteorological parameters such as the temperature and the pressure are crucial for tropospheric delay correction of Satellite Laser Ranging (SLR). Considering that the meteorological data of the SLR observations may be missing, this paper selects two common global meteorological models, GPT2 and GPT2w, to evaluate the accuracy of the meteorological model and apply them to SLR-based satellite orbit validation. In this study, we validated the accuracy of the empirical models with the observed meteorological data collected from the SLR stations. The results show that the RMSE of the pressure and temperature generated by GPT2 model is 5.61 hPa and 4.90 K, respectively. The RMSE of the pressure and temperature derived from the GPT2w is 5.58 hPa and 4.83 K, respectively. Then the SLR residuals are compared by applying the empirical meteorological models and the real observations. The results indicate that the RMS of the SLR residuals with the real meteorological data is 3.22 cm, and the RMS of the SLR residuals calculated with the GPT2 and GPT2w models are 3.45 cm and 3.42 cm, respectively. The GPT2w model performs slightly better than the GPT2 model in SLR data processing. It concludes that replacing the real meteorological data by the empirical models is feasible for SLR data processing.