Development of an adaptive 4-D water vapour density model for the vertical constraints in GNSS tropospheric tomography.

Global Navigation Satellite Systems (GNSS) tropospheric tomography is a commonly used technique for the reconstruction of three-dimensional water vapour field, and a priori vertical constraint models are required for water vapour density (WVD) determination which plays a critical role in the quality of tomographic results. However, generalised exponential models were routinely used for vertical constraints and limited research was carried out in the GNSS tomography by taking epoch-by-epoch variations into consideration. In this study, an adaptive four-dimensional (4-D) WVD model for the vertical constraints in GNSS tropospheric tomography was developed based on both ERA5 and surface meteorological data in Hong Kong for each month during the period of 2015–2019, and the back-propagation neural network technique was used to develop the fitting model. Then, the WVD model was used to obtain the WVD of adjacent voxels in the vertical direction to alleviate the mis-representation of the generalised exponential model. The newly developed WVD model used in GNSS tropospheric tomography was validated using GNSS data from the Hong Kong region in the year 2020 and two tomographic epochs (00:00–00:30 UTC and 12:00–12:30 UTC) were evaluated each day. For each topographic epoch, the WVDs of the tomographic voxels including radiosonde profile are evaluated (10 voxels over 10 height layers) using radiosonde data as the reference and the WVDs of all tomographic voxels are evaluated (300 voxels over 10 height layers) using ERA5 data as the reference. Results showed that when radiosonde/ERA5 data were utilized as the references, corresponding monthly mean values of the root mean square errors (RMSEs) in the entire year reduced from 1.97/1.94 g/m³ of the traditional tomographic method to 1.56/1.36 g/m³ of the new method which showed approximately 21/30% improvements. These results suggest a better performance of the tomographic approach using the new WVD model for the vertical constraints proposed by this study by taking epoch-by-epoch information. [ABSTRACT FROM AUTHOR]