Establishment and Accuracy Analysis of the Atmospheric Weighted Mean Temperature Model in Yunnan Province

The atmospheric weighted mean temperature (Tm) is a key parameter that determines the accuracy of GNSS water vapor inversion, and high-precision water vapor information is crucial for irrigation management, drought monitoring and disaster warning in modern precision agriculture. Considering that the current existing models are poorly applicable in various regions of Yunnan Province, this paper proposes a Tm modeling method that integrates linear models and deep learning: first, a linear regression model is constructed based on meteorological factors, and then the model residual is corrected by a long short-term memory network (lSTM). The study uses data from four sounding stations in Yunnan Province from 2019 to 2022 to establish a model, and uses sounding data from 2023 for accuracy evaluation. The results show that compared with the Bevis model, GPT3 model, and traditional linear regression model, the proposed method shows higher stability and applicability in different regions and seasons. The constructed high-precision Tm model can provide refined meteorological support for irrigation management and agricultural disaster warning.