A Novel Regional Weighted Mean Temperature Model Through Multi-Source Data Fusion with Voronoi Diagrams: A Case Study in Northwest China

Objectives: The weighted mean temperature (T_m) is a key parameter for estimating precipitable water vapor (PWV) from Global Navigation Satellite System (GNSS) observations. The environment in northwest China is complicated, so it is necessary to establish a suitable atmospheric weighted mean temperature model. Methods: A novel Voronoi region regression model is presented by incorporating the Voronoi diagram concept into the traditional T_m regression model and a Voronoi region regression model is built using data from 24 radiosonde stations within the study region from 2017 to 2021. European Centre for Medium-Range Weather Forecasts reanalysis data and Global Geodetic Observing System Atmosphere T_m grid data are then integrated to generate a suitable for the Northwest regional T_m Augmentation Model (TAM). During the integration process, considering the fact that T_m decreases with elevation in the near-surface environment, a cubic polynomial model is proposed to correct the T_m values in the vertical direction. Results: The model is validated using data from radiosonde stations and GGOS T_m grid within the study region in 2022, suggesting that TAM model exhibits high accuracy and applicability in the Northwest region, with accuracy improvements of 19.7% and 50.7% compared to the Bevis model, 44.4% and 18.4% compared to GPT3-1 model, 53.8% and 41% compared to GPT3-5 respectively. Conclusions: The spatial distribution of T_m estimated by the TAM model exhibits optimal performance, especially in the Qinghai-Tibet Plateau and other high-altitude areas. It is believed that the TAM model holds great potential for high-precision water vapor monitoring with GNSS in mountainous areas.