Case study on a novel approach to determining initial values for 3D water vapor tomography in the a priori water vapor field

Abstract

Insufficient prior information in GNSS water vapor tomographic models can lead to instability in the structure of the design matrix. The addition of high-precision initial values for water vapor density can effectively solve this problem, but common methods for obtaining water vapor density are relatively crude and not fully utilizing the original accuracy of the a priori water vapor field. To address this issue, a new method is proposed that takes into account the influence of the a priori information weight ratio in the tomographic model on tomographic analysis results, based on the negative exponential decrease of atmospheric water vapor density in the vertical direction with height. Moreover, the 3D water vapor tomographic technique based on the ray-tracing fast voxel traversal algorithm has been improved and implemented in the C++ programming language. To verify the feasibility, accuracy, and stability of the improved 3D water vapor tomographic technique and the proposed method, the experiment using actual data from the CORS station in Hong Kong in January and May and the corresponding grid data provided by ERA5 was conducted. As benchmarks, radio-sounding data from the experimental area was utilized. In May, the proposed method exhibited an average daily improvement in the root mean square error of 0.88% (UTC0) and 0.97% (UTC12), and an improvement in the interquartile range of the difference between the water vapor density values at each altitude layer (excluding the top and bottom layers) and the corresponding sounding data of 5.08% (UTC0) and 3.37% (UTC12). In January, the average daily improvement in the root mean square error was −13.78% (UTC0) and −4.84% (UTC12), and the improvement in the interquartile range of the difference between the water vapor density values at each altitude level and those calculated from the corresponding sounding data was −14.49% (UTC0) and −6.31% (UTC12). The results show that the proposed method and the commonly used method maintain high consistency in the accuracy of the water vapor density values solved by the system of tomographic equations and improve stability, especially in May when the water vapor density near the surface is high. The feasibility of the improved 3D water vapor tomographic technique based on the ray-tracing fast voxel traversal algorithm is also demonstrated.