Adaptive Algebraic Reconstruction Algorithms for GNSS Water Vapor Tomography

  Objectives   Global navigation satellite system(GNSS) tomography, characterized by reconstructing the three-dimensional distribution of the atmospheric water vapor, has proved its capacity for studying the extreme weather events. The ill-conditioned problem resulting from the GNSS acquisition geometry is the critical issue of the GNSS tomography system. Algebraic reconstruction techniques, with the advantages of simple iteration and fast convergence, has been widely applied to the tropospheric tomography. However, the error allocation principle based on the intersection length of GNSS rays with each ray-voxel affects the accuracy of tomographic results. An improved adaptive algebraic reconstruction techniques are proposed to address this problem.

  Methods   In view of the unreasonable error allocation in the traditional algorithm, we suggest a new error allocation principle based on the variation of water vapor density(WVD) in voxels for the adaptive algorithms, in which the product of the intersection length and WVD is considered as a new principle to redistribute the difference between the GNSS slant water vapor(SWV)and the reconstructed SWV. Besides, the weight matrix model of elevation angles is introduced to the improved algorithm to optimize the tomographic results.

  Results   The new algorithms are tested by measured data from Xuzhou continuously operating reference stations(CORS) network and radiosonde during July2016. Experimental results reveal that the WVD derived from the adaptive algorithms performs better than that of the general algorithms in root mean squared error(RMSE), standard deviation(STD) and mean absolute error(MAE), and the RMSE is decreased by 25.91%, 15.81%, and 24.64% for adaptive algebraic reconstruction technique(AART), adaptive multiplicative algebraic reconstruction technique(AMART)and adaptive simultaneous iterative reconstruction technique(ASIRT) respectively. Under the conditions of light rain, moderate rain and heavy rain, the adaptive algorithms retrieve better water vapor profile than the traditional ones.

  Conclusions   Overall, the water vapor profile derived from the adaptive algorithms agrees with the radiosonde distribution better than the traditional algorithms, which reveals the advantage of the proposed method on optimizing the tomography solutions.