GNSS water vapor tomography based on Kalman filter with optimized noise covariance.

GNSS water vapor tomography has emerged as a prominent technique for obtaining the three-dimensional distribution of atmospheric water vapor. It effectively compensates for the deficiency of GNSS precipitable water vapor that only reflects the two-dimensional distribution of water vapor and has become a hotspot in GNSS meteorology. The GNSS water vapor tomography based on the Kalman filter well considers the correlation between successive epochs and avoids the restrictions of too many additional constraints, but its accuracy and stability are often affected by the noise covariance. We propose an optimized noise covariance matrix method for GNSS water vapor tomography based on the Kalman filter. It constructs the state noise covariance using historical water vapor information derived from atmospheric reanalysis data and establishes the observation noise covariance considering the satellite elevation angle and the signal intercept crossing the tomographic voxels. The tomography experiments conducted in Hong Kong show that the root-mean-square error (RMSE) and mean absolute error (MAE) of the slant water vapor computed by the proposed method decreased by 38.8% and 34.9% compared with the traditional method. The proposed method improves average RMSE and MAE of 23.2% and 24.8% compared to the radiosonde data and 27.6% and 44.3% compared to ERA5 data. [ABSTRACT FROM AUTHOR]