A new tropospheric tomography model combining the pixel-based and function-based models

Abstract. As a new detection method of three-dimensional water vapor, the ground-based water vapor tomography technique using Global Navigation Satellite Systems (GNSS) observations can obtain the high spatial and temporal distribution information of tropospheric water vapor. Since the troposphere tomography was proposed, most previous studies belong to the pixel-based method, dividing the interest area into three-dimensional voxels of which the water vapor density of each voxel center is taken as the average water vapor density. However, the abovementioned method can only find the water vapor density value of the center of each voxel, which is unable to express the continuous change of water vapor in space and destroys the spatial continuity of water vapor variation. Moreover, when using the pixel-based method, too many voxels are needed to express the water vapor density, which leads to the problem of too many coefficients to be estimated. After analyzing the limitations of the traditional pixel-based troposphere tomography technique, this paper proposes a new GNSS tropospheric water vapor tomography model combining the pixel-based and function-based models for the first time. The tomographic experiences were validated using the data from 12 stations from the Hong Kong Satellite Positioning Reference Station Network (SatRef) collected between 25 March and 25 April 2014. The comparison between tomographic results and the European Centre for Medium-Range Weather Forecasts (ECMWF) data is mainly used to analyze the accuracy of the new model proposed in this paper under different conditions, for showing that this new model is superior to the traditional pixel-based model in terms of root-mean-square error (RMSE) and bias. The new model has more advantages than the traditional pixel-based model on the RMSE, especially when obtaining the water vapor in voxels without the penetration of GNSS rays, which is improved by 5.88 %. This model also solves the problem with more ease and convenience in expression.