Establishment of PWV Fusion Model Using Multi-source Data

  Objectives  Precipitable water vapor (PWV) information with high precision and high spatial-temporal resolution plays an important role in the study of extreme weather. The PWV obtained by traditional single water vapor detection technology has the defects of poor precision and low spatial-temporal resolution due to the limitations of its system design.

  Methods  To solve this problem, we propose a PWV hybrid model based on multi-source data, called the GSP(GPT2w+spherical harmonical function+polynomial fitting) model.In this model, the initial value of PWV is calculated by the GPT2w model, the residual sequence of PWV is fitted by a spherical harmonic function, and after that, deviation correction is performed for the residual PWV based on the polynomial fitting, and the Bartlett test is introduced to deter mine the optimal weights of multi-source data in the GSP model.

  Results  The data of 26 GNSS stations and 37 ERA-Interim grid points (1°×1°) in Yunnan Province, China has been selected to validate the GSP model, and the numerical results show that the accuracy improvement rate of the GSP model is 15%— 18% compared with the traditional polynomial fitting model. Compared with the ERA5 (0.25°×0.25°) data, the mean root mean square and Bias of GSP model are 1.64 mm and -0.25 mm, respectively.

  Conclusions  The above results show that the proposed GSP model has high accuracy and plays an important role in extreme weather warnings.