An improved weighted mean temperature (T-m) model based on GPT2w with T-m lapse rate

摘要

Global pressure and temperature 2 wet (GPT2w) is an empirical model providing the mean values plus annual and semiannual amplitudes of weighted mean temperature (T-m), which makes it a widely used tool in converting zenith wet delay (ZWD) to precipitable water vapor (PWV) in GNSS meteorology. The model meets the needs of real-time T-m anywhere in the world without relying on any other meteorological observations compared with traditional T-m calculation methods. It outperforms the other empirical T-m models released in recent years. Due to the lack of the T-m vertical adjustment in the model, the accuracy of T-m estimated by the model is subject to certain constraints, especially at sites which have large altitude differences compared with the GPT2w grid points. We explored the T-m lapse rate for the vertical adjustment using 10 years of 37 monthly mean pressure level data from the European Center for Medium-Range Weather Forecasts (ECMWF) and extended the GPT2w model to a new one called the GPT2wh model. Three schemes with different height ranges were established to fit the T-m lapse rate, and the most appropriate scheme was selected by adopting the goodness of fit measures, including the coefficient of determination (R-squared) and the root mean square error (RMSE). In addition to the mean value, annual and semiannual amplitudes for T-m lapse rate on a regular 1 degrees grid were determined and stored in the GPT2wh model. The performance of the new model was assessed against the GPT2w model using different data sources in 2011, i.e., the ECMWF data and globally distributed radiosonde data. The numerical results show that the GPT2wh model outperforms the GPT2w model with an improved RMSE of 7.36/5.00/2.45/1.37/0.51/0.03 K at different height levels in the ECMWF comparison. In comparison with the radiosonde data, the mean RMSE of the GPT2wh model improves by 0.33 K from 4.16 to 3.83 K, i.e., an approximately 8% improvement against the GPT2w model. The impact of T-m on GNSS-PWV was analyzed, showing that the GPT2wh model can effectively improve the accuracy of the converted PWV.