Surface temperature and pressure are indispensable variables in Global Navigation Satellite System (GNSS) meteorology. The lack of meteorological observations located at or near the GNSS sites is a big challenge for the calculation of accurate zenith hydrostatic delay (ZHD). Therefore, various empirical models with different model forms were established to provide temperature and pressure values. In this study, the influence of different modelling factors, including model forms, temporal resolution of the data sources, and the spatial resolution of the data sources, is evaluated and the temperature and pressure model with the best performance is developed. On the basis of the meteorological parameters estimated by the above model, we analyzed the global performance of the three most commonly used ZHD models, that is, the Saastamoinen, Hopfield, and Black models. The numerical results show that the model with the idea of time-segmented modelling performs best, of which the global mean root mean square (RMS), mean absolute error (MAE), and standard deviation (SD) are 7.87/6.33/7.17 hPa and 2.95/2.31/2.79 K for pressure and temperature, respectively, using the data sources with temporal resolution of 2 h and spatial resolution of 2.5 • × 2 • in the reanalysis data comparison. In comparison with the radiosonde data, the mean RMS/MAE/SD are 7.02/5.24/6.46 hPa and 4.05/3.17/3.86 K for pressure and temperature, respectively. The Saastamoinen model with a global mean bias/RMS of 1.01/16.9 mm achieved the best ZHD estimated values compared with the other two ZHD models.ZHD is usually needed [4,5]. In addition, accurate ZHD is a prerequisite for obtaining water vapor information in GNSS meteorology, as ZWD is achieved by subtracting ZHD from ZTD [6][7][8][9][10][11].ZHD can be derived with millimeter-level accuracy from meteorological parameters and position at an observation station using a model, such as the Saastamoinen, Hopfield, and Black models [12][13][14]. However, most GNSS sites are not equipped with meteorological sensors, and there are often no collocated weather stations available for those GNSS sites. Thus, a standard model for providing precise and unbiased global pressure and temperature values is often used, for example, the global pressure and temperature (GPT) series models, TropGrid2 model, and the improved tropospheric grid (ITG) model [15][16][17][18][19].The GPT model is proposed by Boehm et al. [15], which is developed from three-year European Center for Medium-Range Weather Forecasts (ECMWF) reanalysis (ERA)-40 products, and provides temperature and pressure based on spherical harmonics up to degree and order 9, leading to coarse horizontal resolution of about 20º. A more precise ECMWF ERA-interim product was applied by Lagler et al. to establish the GPT2 model [16,20], in which the semi-annual harmonics was incorporated to better account for regions where very rainy periods or very dry periods dominate, making it an improved model to provide meteorological parameters resting upon a global 5º grid ...