An Assessment of GPT2w Model and Fusion of a Troposphere Model with in Situ Data

GPT2w is used to estimate slant tropospheric delay, and considered the best empirical model based on its nominal accuracy. Besides the model value of ZHD, this model delivers a blind value for meteorological parameters. We used the database from USNO to validate the marked accuracy of the model, and data from IGRA to access the accuracy of the blind meteorological elements. A system bias from T_m was detected. After the correction of this bias, the bias of model ZTD against the USNO ZTD rose from-1.38 mm to-0.3 mm. This paper presents a fusion of blind model and the in situ data. The input parameters of this new method are in situ P, t and hr, the corrected blind T_m and λ. This method performs better than the modified GPT2w model because of the in situ data, better than the Saast model as it profits from a improved ZWD model. Without the in situ data, the modified GPT2w is a good choice. If with available in situ data, the fusion method is recommended.