CGTV-Tm: A High-Accuracy Gridded Atmospheric Weighted Mean Temperature Model Coupling Surface Temperature and Water Vapor Pressure over China.

The atmospheric weighted mean temperature (Tm) is critical for converting a zenith wet delay (ZWD) to precipitable water vapor (PWV). However, the existing Tm models still have certain limitations: Those driven by surface-measured parameters achieve high accuracy but depend heavily on in situ instruments, incurring high costs and lacking forecasting capability. Empirical models avoid measured data but fail to capture short-term Tm variations, leading to lower accuracy. Daily weather forecast data-which are low-cost, readily available, and reflective of short-term changes-offer a promising alternative. This study develops a gridded Tm model named CGTV-Tm, which couples temperature and water vapor pressure, using ERA5 reanalysis data over China (2019-2023). The model can be driven by daily weather forecast data. A dual vertical correction method is also proposed to improve performance. Validation against 2024 ERA5 and radiosonde data shows that CGTV-Tm achieves RMSEs of 2.38 K (vs. ERA5) and 2.64 K (vs. radiosonde), significantly outperforming the Bevis (3.61 K, 3.67 K), PTm (3.19 K, 2.94 K), and CGT-Tm (2.71 K, 3.08 K) models. When driven by daily weather forecast data, CGTV-Tm achieves an RMSE of 2.90 K, improving accuracy by 29.6% and 21.2% over the state-of-the-art empirical models GPT3 and HGPT2, respectively. These results demonstrate that CGTV-Tm not only surpasses traditional linear Tm models that rely solely on surface temperature but also, by using weather forecast data, it removes dependence on in situ instruments, offering a superior low-cost solution for real-time GNSS (Global Navigation Satellite System) PWV retrieval.